From 4637235183b80963536f2364e4d50fcb894886dd Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 14 Jan 2009 10:44:48 +1000 Subject: radeon/r200/r300: initial attempt to convert to common context code --- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 67 +++++++++++++++--------------- 1 file changed, 34 insertions(+), 33 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index ebea1fecdc..ef89d73bdd 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -52,7 +52,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_tcl.h" -static void flush_last_swtcl_prim( radeonContextPtr rmesa ); +static void flush_last_swtcl_prim(GLcontext *ctx); /* R100: xyzw, c0, c1/fog, stq[0..2] = 4+1+1+3*3 = 15 right? */ /* R200: xyzw, c0, c1/fog, strq[0..5] = 4+1+1+4*6 = 30 */ @@ -87,7 +87,7 @@ static GLuint radeon_cp_vc_frmts[3][2] = static void radeonSetVertexFormat( GLcontext *ctx ) { - radeonContextPtr rmesa = RADEON_CONTEXT( ctx ); + r100ContextPtr rmesa = R100_CONTEXT( ctx ); TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; DECLARE_RENDERINPUTS(index_bitset); @@ -204,7 +204,7 @@ static void radeonSetVertexFormat( GLcontext *ctx ) } } - if (!RENDERINPUTS_EQUAL( rmesa->tnl_index_bitset, index_bitset ) || + if (!RENDERINPUTS_EQUAL( rmesa->radeon.tnl_index_bitset, index_bitset ) || fmt_0 != rmesa->swtcl.vertex_format) { RADEON_NEWPRIM(rmesa); rmesa->swtcl.vertex_format = fmt_0; @@ -214,7 +214,7 @@ static void radeonSetVertexFormat( GLcontext *ctx ) rmesa->swtcl.vertex_attr_count, NULL, 0 ); rmesa->swtcl.vertex_size /= 4; - RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset ); + RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset ); if (RADEON_DEBUG & DEBUG_VERTS) fprintf( stderr, "%s: vertex_size= %d floats\n", __FUNCTION__, rmesa->swtcl.vertex_size); @@ -224,13 +224,13 @@ static void radeonSetVertexFormat( GLcontext *ctx ) static void radeonRenderStart( GLcontext *ctx ) { - radeonContextPtr rmesa = RADEON_CONTEXT( ctx ); + r100ContextPtr rmesa = R100_CONTEXT( ctx ); radeonSetVertexFormat( ctx ); if (rmesa->dma.flush != 0 && rmesa->dma.flush != flush_last_swtcl_prim) - rmesa->dma.flush( rmesa ); + rmesa->dma.flush( ctx ); } @@ -241,7 +241,7 @@ static void radeonRenderStart( GLcontext *ctx ) */ void radeonChooseVertexState( GLcontext *ctx ) { - radeonContextPtr rmesa = RADEON_CONTEXT( ctx ); + r100ContextPtr rmesa = R100_CONTEXT( ctx ); TNLcontext *tnl = TNL_CONTEXT(ctx); GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT]; @@ -254,7 +254,7 @@ void radeonChooseVertexState( GLcontext *ctx ) * rasterization fallback. As this function will be called again when we * leave a rasterization fallback, we can just skip it for now. */ - if (rmesa->Fallback != 0) + if (rmesa->radeon.Fallback != 0) return; /* HW perspective divide is a win, but tiny vertex formats are a @@ -284,8 +284,9 @@ void radeonChooseVertexState( GLcontext *ctx ) /* Flush vertices in the current dma region. */ -static void flush_last_swtcl_prim( radeonContextPtr rmesa ) +static void flush_last_swtcl_prim(GLcontext *ctx) { + r100ContextPtr rmesa = R100_CONTEXT(ctx); if (RADEON_DEBUG & DEBUG_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); @@ -293,7 +294,7 @@ static void flush_last_swtcl_prim( radeonContextPtr rmesa ) if (rmesa->dma.current.buf) { struct radeon_dma_region *current = &rmesa->dma.current; - GLuint current_offset = (rmesa->radeonScreen->gart_buffer_offset + + GLuint current_offset = (rmesa->radeon.radeonScreen->gart_buffer_offset + current->buf->buf->idx * RADEON_BUFFER_SIZE + current->start); @@ -326,7 +327,7 @@ static void flush_last_swtcl_prim( radeonContextPtr rmesa ) /* Alloc space in the current dma region. */ static INLINE void * -radeonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) +radeonAllocDmaLowVerts( r100ContextPtr rmesa, int nverts, int vsize ) { GLuint bytes = vsize * nverts; @@ -334,7 +335,7 @@ radeonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) radeonRefillCurrentDmaRegion( rmesa ); if (!rmesa->dma.flush) { - rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; + rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; rmesa->dma.flush = flush_last_swtcl_prim; } @@ -387,14 +388,14 @@ static const GLuint hw_prim[GL_POLYGON+1] = { }; static INLINE void -radeonDmaPrimitive( radeonContextPtr rmesa, GLenum prim ) +radeonDmaPrimitive( r100ContextPtr rmesa, GLenum prim ) { RADEON_NEWPRIM( rmesa ); rmesa->swtcl.hw_primitive = hw_prim[prim]; assert(rmesa->dma.current.ptr == rmesa->dma.current.start); } -#define LOCAL_VARS radeonContextPtr rmesa = RADEON_CONTEXT(ctx) +#define LOCAL_VARS r100ContextPtr rmesa = R100_CONTEXT(ctx) #define INIT( prim ) radeonDmaPrimitive( rmesa, prim ) #define FLUSH() RADEON_NEWPRIM( rmesa ) #define GET_CURRENT_VB_MAX_VERTS() \ @@ -418,7 +419,7 @@ radeonDmaPrimitive( radeonContextPtr rmesa, GLenum prim ) static GLboolean radeon_run_render( GLcontext *ctx, struct tnl_pipeline_stage *stage ) { - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; tnl_render_func *tab = TAG(render_tab_verts); @@ -496,12 +497,12 @@ static void radeonResetLineStipple( GLcontext *ctx ); #undef LOCAL_VARS #undef ALLOC_VERTS -#define CTX_ARG radeonContextPtr rmesa +#define CTX_ARG r100ContextPtr rmesa #define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size #define ALLOC_VERTS( n, size ) radeonAllocDmaLowVerts( rmesa, n, (size) * 4 ) #undef LOCAL_VARS #define LOCAL_VARS \ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \ + r100ContextPtr rmesa = R100_CONTEXT(ctx); \ const char *radeonverts = (char *)rmesa->swtcl.verts; #define VERT(x) (radeonVertex *)(radeonverts + ((x) * (vertsize) * sizeof(int))) #define VERTEX radeonVertex @@ -606,7 +607,7 @@ do { \ #undef INIT #define LOCAL_VARS(n) \ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \ + r100ContextPtr rmesa = R100_CONTEXT(ctx); \ GLuint color[n], spec[n]; \ GLuint coloroffset = rmesa->swtcl.coloroffset; \ GLuint specoffset = rmesa->swtcl.specoffset; \ @@ -673,7 +674,7 @@ static void init_rast_tab( void ) } while (0) #undef LOCAL_VARS #define LOCAL_VARS \ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \ + r100ContextPtr rmesa = R100_CONTEXT(ctx); \ const GLuint vertsize = rmesa->swtcl.vertex_size; \ const char *radeonverts = (char *)rmesa->swtcl.verts; \ const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \ @@ -700,11 +701,11 @@ static void init_rast_tab( void ) void radeonChooseRenderState( GLcontext *ctx ) { TNLcontext *tnl = TNL_CONTEXT(ctx); - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); GLuint index = 0; GLuint flags = ctx->_TriangleCaps; - if (!rmesa->TclFallback || rmesa->Fallback) + if (!rmesa->radeon.TclFallback || rmesa->radeon.Fallback) return; if (flags & DD_TRI_LIGHT_TWOSIDE) index |= RADEON_TWOSIDE_BIT; @@ -739,7 +740,7 @@ void radeonChooseRenderState( GLcontext *ctx ) static void radeonRasterPrimitive( GLcontext *ctx, GLuint hwprim ) { - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); if (rmesa->swtcl.hw_primitive != hwprim) { RADEON_NEWPRIM( rmesa ); @@ -749,7 +750,7 @@ static void radeonRasterPrimitive( GLcontext *ctx, GLuint hwprim ) static void radeonRenderPrimitive( GLcontext *ctx, GLenum prim ) { - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); rmesa->swtcl.render_primitive = prim; if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) radeonRasterPrimitive( ctx, reduced_hw_prim[prim] ); @@ -761,7 +762,7 @@ static void radeonRenderFinish( GLcontext *ctx ) static void radeonResetLineStipple( GLcontext *ctx ) { - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); RADEON_STATECHANGE( rmesa, lin ); } @@ -795,12 +796,12 @@ static const char *getFallbackString(GLuint bit) void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) { - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint oldfallback = rmesa->Fallback; + GLuint oldfallback = rmesa->radeon.Fallback; if (mode) { - rmesa->Fallback |= bit; + rmesa->radeon.Fallback |= bit; if (oldfallback == 0) { RADEON_FIREVERTICES( rmesa ); TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_TRUE ); @@ -813,7 +814,7 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) } } else { - rmesa->Fallback &= ~bit; + rmesa->radeon.Fallback &= ~bit; if (oldfallback == bit) { _swrast_flush( ctx ); tnl->Driver.Render.Start = radeonRenderStart; @@ -826,14 +827,14 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) tnl->Driver.Render.ResetLineStipple = radeonResetLineStipple; TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_FALSE ); - if (rmesa->TclFallback) { - /* These are already done if rmesa->TclFallback goes to + if (rmesa->radeon.TclFallback) { + /* These are already done if rmesa->radeon.TclFallback goes to * zero above. But not if it doesn't (RADEON_NO_TCL for * example?) */ _tnl_invalidate_vertex_state( ctx, ~0 ); _tnl_invalidate_vertices( ctx, ~0 ); - RENDERINPUTS_ZERO( rmesa->tnl_index_bitset ); + RENDERINPUTS_ZERO( rmesa->radeon.tnl_index_bitset ); radeonChooseVertexState( ctx ); radeonChooseRenderState( ctx ); } @@ -853,7 +854,7 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) void radeonInitSwtcl( GLcontext *ctx ) { TNLcontext *tnl = TNL_CONTEXT(ctx); - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); static int firsttime = 1; if (firsttime) { @@ -881,7 +882,7 @@ void radeonInitSwtcl( GLcontext *ctx ) void radeonDestroySwtcl( GLcontext *ctx ) { - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); if (rmesa->swtcl.indexed_verts.buf) radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, -- cgit v1.2.3 From b584b0728d3a001a142f76dde22f9e8ed7d2dd16 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 4 Feb 2009 15:58:08 +1000 Subject: radeon: make compile again. Not tested but ripped out lots of stuff unneeded anymore time to test later --- src/mesa/drivers/dri/radeon/radeon_context.c | 6 +- src/mesa/drivers/dri/radeon/radeon_context.h | 19 +-- src/mesa/drivers/dri/radeon/radeon_cs_legacy.c | 2 +- src/mesa/drivers/dri/radeon/radeon_ioctl.c | 109 +++++---------- src/mesa/drivers/dri/radeon/radeon_ioctl.h | 55 ++++---- src/mesa/drivers/dri/radeon/radeon_maos_verts.c | 10 +- src/mesa/drivers/dri/radeon/radeon_span.c | 136 ++++--------------- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 168 ++++++++---------------- src/mesa/drivers/dri/radeon/radeon_swtcl.h | 2 +- src/mesa/drivers/dri/radeon/radeon_tcl.c | 24 ++-- src/mesa/drivers/dri/radeon/radeon_tex.c | 3 +- src/mesa/drivers/dri/radeon/radeon_texstate.c | 21 +-- 12 files changed, 181 insertions(+), 374 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index 6b91bdce9c..aa1cd5e300 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -54,6 +54,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "drivers/common/driverfuncs.h" #include "radeon_context.h" +#include "common_cmdbuf.h" #include "radeon_ioctl.h" #include "radeon_state.h" #include "radeon_span.h" @@ -243,6 +244,7 @@ static void r100_init_vtbl(radeonContextPtr radeon) radeon->vtbl.update_draw_buffer = radeonUpdateDrawBuffer; radeon->vtbl.emit_cs_header = r100_vtbl_emit_cs_header; radeon->vtbl.emit_state = r100_vtbl_emit_state; + radeon->vtbl.swtcl_flush = r100_swtcl_flush; } /* Create the device specific context. @@ -332,7 +334,7 @@ radeonCreateContext( const __GLcontextModes *glVisual, rmesa->radeon.texture_depth = ( screen->cpp == 4 ) ? DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16; - rmesa->swtcl.RenderIndex = ~0; + rmesa->radeon.swtcl.RenderIndex = ~0; rmesa->hw.all_dirty = GL_TRUE; /* Set the maximum texture size small enough that we can guarentee that @@ -512,7 +514,7 @@ void radeonDestroyContext( __DRIcontextPrivate *driContextPriv ) radeonReleaseArrays( rmesa->radeon.glCtx, ~0 ); if (rmesa->radeon.dma.current) { radeonReleaseDmaRegion( &rmesa->radeon ); - radeonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ ); + rcommonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ ); } _mesa_vector4f_free( &rmesa->tcl.ObjClean ); diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h index ba5c57f121..6edbaf6910 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_context.h @@ -340,7 +340,7 @@ struct r100_state { #define RADEON_CMD_BUF_SZ (8*1024) - +#define R200_ELT_BUF_SZ (8*1024) /* radeon_tcl.c */ struct radeon_tcl_info { @@ -364,18 +364,17 @@ struct radeon_tcl_info { struct radeon_dma_region fog; struct radeon_dma_region tex[RADEON_MAX_TEXTURE_UNITS]; struct radeon_dma_region norm; + + struct radeon_bo *elt_dma_bo; + int elt_dma_offset; /** Offset into this buffer object, in bytes */ + int elt_used; }; /* radeon_swtcl.c */ -struct radeon_swtcl_info { - GLuint RenderIndex; - GLuint vertex_size; +struct r100_swtcl_info { GLuint vertex_format; - struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; - GLuint vertex_attr_count; - GLubyte *verts; /* Fallback rasterization functions @@ -384,10 +383,6 @@ struct radeon_swtcl_info { radeon_line_func draw_line; radeon_tri_func draw_tri; - GLuint hw_primitive; - GLenum render_primitive; - GLuint numverts; - /** * Offset of the 4UB color data within a hardware (swtcl) vertex. */ @@ -448,7 +443,7 @@ struct r100_context { /* radeon_swtcl.c */ - struct radeon_swtcl_info swtcl; + struct r100_swtcl_info swtcl; GLboolean using_hyperz; GLboolean texmicrotile; diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c index fe4f90e0b1..1c885c1612 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c @@ -313,7 +313,7 @@ static int cs_emit(struct radeon_cs *cs) cmd.boxes = (drm_clip_rect_t *) csm->ctx->pClipRects; } - // dump_cmdbuf(cs); + //dump_cmdbuf(cs); r = drmCommandWrite(cs->csm->fd, DRM_RADEON_CMDBUF, &cmd, sizeof(cmd)); if (r) { diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index b101668f77..a823c79d6c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -43,6 +43,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "swrast/swrast.h" #include "radeon_context.h" +#include "common_cmdbuf.h" +#include "radeon_cs.h" #include "radeon_state.h" #include "radeon_ioctl.h" #include "radeon_tcl.h" @@ -186,7 +188,7 @@ void radeonEmitState( r100ContextPtr rmesa ) * for enough space for the case of emitting all state, and inline the * radeonAllocCmdBuf code here without all the checks. */ - radeonEnsureCmdBufSpace(rmesa, rmesa->hw.max_state_size); + rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->hw.max_state_size, __FUNCTION__); dest = rmesa->store.cmd_buf + rmesa->store.cmd_used; /* We always always emit zbs, this is due to a bug found by keithw in @@ -243,8 +245,7 @@ extern void radeonEmitVbufPrim( r100ContextPtr rmesa, GLuint primitive, GLuint vertex_nr ) { - drm_radeon_cmd_header_t *cmd; - + BATCH_LOCALS(&rmesa->radeon); assert(!(primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND)); @@ -254,43 +255,16 @@ extern void radeonEmitVbufPrim( r100ContextPtr rmesa, fprintf(stderr, "%s cmd_used/4: %d\n", __FUNCTION__, rmesa->store.cmd_used/4); - cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, VBUF_BUFSZ, - __FUNCTION__ ); -#if RADEON_OLD_PACKETS - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP; - cmd[1].i = RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM | (3 << 16); - cmd[2].i = rmesa->ioctl.vertex_offset; - cmd[3].i = vertex_nr; - cmd[4].i = vertex_format; - cmd[5].i = (primitive | - RADEON_CP_VC_CNTL_PRIM_WALK_LIST | - RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA | - RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE | - (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT)); - - if (RADEON_DEBUG & DEBUG_PRIMS) - fprintf(stderr, "%s: header 0x%x offt 0x%x vfmt 0x%x vfcntl %x \n", - __FUNCTION__, - cmd[1].i, cmd[2].i, cmd[4].i, cmd[5].i); -#else - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP; - cmd[1].i = RADEON_CP_PACKET3_3D_DRAW_VBUF | (1 << 16); - cmd[2].i = vertex_format; - cmd[3].i = (primitive | - RADEON_CP_VC_CNTL_PRIM_WALK_LIST | - RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA | - RADEON_CP_VC_CNTL_MAOS_ENABLE | - RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE | - (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT)); - - - if (RADEON_DEBUG & DEBUG_PRIMS) - fprintf(stderr, "%s: header 0x%x vfmt 0x%x vfcntl %x \n", - __FUNCTION__, - cmd[1].i, cmd[2].i, cmd[3].i); -#endif + BEGIN_BATCH(3); + OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_DRAW_VBUF, 0); + OUT_BATCH(vertex_format); + OUT_BATCH(primitive | + RADEON_CP_VC_CNTL_PRIM_WALK_LIST | + RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA | + RADEON_CP_VC_CNTL_MAOS_ENABLE | + RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE | + (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT)); + END_BATCH(); } @@ -346,35 +320,14 @@ GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa, radeonEmitState( rmesa ); - cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, - ELTS_BUFSZ(min_nr), - __FUNCTION__ ); -#if RADEON_OLD_PACKETS - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP; - cmd[1].i = RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM; - cmd[2].i = rmesa->ioctl.vertex_offset; - cmd[3].i = 0xffff; - cmd[4].i = vertex_format; - cmd[5].i = (primitive | - RADEON_CP_VC_CNTL_PRIM_WALK_IND | - RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA | - RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE); - - retval = (GLushort *)(cmd+6); -#else - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP; - cmd[1].i = RADEON_CP_PACKET3_3D_DRAW_INDX; - cmd[2].i = vertex_format; - cmd[3].i = (primitive | - RADEON_CP_VC_CNTL_PRIM_WALK_IND | - RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA | - RADEON_CP_VC_CNTL_MAOS_ENABLE | - RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE); - - retval = (GLushort *)(cmd+4); -#endif + rmesa->tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom, + 0, R200_ELT_BUF_SZ, 4, + RADEON_GEM_DOMAIN_GTT, 0); + rmesa->tcl.elt_dma_offset = 0; + rmesa->tcl.elt_used = min_nr * 2; + + radeon_bo_map(rmesa->tcl.elt_dma_bo, 1); + retval = rmesa->tcl.elt_dma_bo->ptr + rmesa->tcl.elt_dma_offset; if (RADEON_DEBUG & DEBUG_PRIMS) fprintf(stderr, "%s: header 0x%x vfmt 0x%x prim %x \n", @@ -400,21 +353,19 @@ void radeonEmitVertexAOS( r100ContextPtr rmesa, rmesa->ioctl.vertex_size = vertex_size; rmesa->ioctl.vertex_offset = offset; #else - drm_radeon_cmd_header_t *cmd; + BATCH_LOCALS(&rmesa->radeon); if (RADEON_DEBUG & (DEBUG_PRIMS|DEBUG_IOCTL)) fprintf(stderr, "%s: vertex_size 0x%x offset 0x%x \n", __FUNCTION__, vertex_size, offset); - cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, VERT_AOS_BUFSZ, - __FUNCTION__ ); - - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3; - cmd[1].i = RADEON_CP_PACKET3_3D_LOAD_VBPNTR | (2 << 16); - cmd[2].i = 1; - cmd[3].i = vertex_size | (vertex_size << 8); - cmd[4].i = offset; + BEGIN_BATCH(5); + OUT_BATCH_PACKET3(RADEON_CP_PACKET3_3D_LOAD_VBPNTR, 2); + OUT_BATCH(1); + OUT_BATCH(vertex_size | (vertex_size << 8)); + OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); +} #endif } diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h index b4bc9b1144..87d64fb728 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h @@ -153,36 +153,37 @@ do { \ #define VBUF_BUFSZ (4 * sizeof(int)) #endif -/* Ensure that a minimum amount of space is available in the command buffer. - * This is used to ensure atomicity of state updates with the rendering requests - * that rely on them. - * - * An alternative would be to implement a "soft lock" such that when the buffer - * wraps at an inopportune time, we grab the lock, flush the current buffer, - * and hang on to the lock until the critical section is finished and we flush - * the buffer again and unlock. - */ -static INLINE void radeonEnsureCmdBufSpace( r100ContextPtr rmesa, - int bytes ) -{ - if (rmesa->store.cmd_used + bytes > RADEON_CMD_BUF_SZ) - radeonFlushCmdBuf( rmesa, __FUNCTION__ ); - assert( bytes <= RADEON_CMD_BUF_SZ ); -} -/* Alloc space in the command buffer - */ -static INLINE char *radeonAllocCmdBuf( r100ContextPtr rmesa, - int bytes, const char *where ) +static inline uint32_t cmdpacket3(int cmd_type) { - if (rmesa->store.cmd_used + bytes > RADEON_CMD_BUF_SZ) - radeonFlushCmdBuf( rmesa, __FUNCTION__ ); + drm_radeon_cmd_header_t cmd; + + cmd.i = 0; + cmd.header.cmd_type = cmd_type; + + return (uint32_t)cmd.i; - { - char *head = rmesa->store.cmd_buf + rmesa->store.cmd_used; - rmesa->store.cmd_used += bytes; - return head; - } } +#define OUT_BATCH_PACKET3(packet, num_extra) do { \ + if (!b_l_rmesa->radeonScreen->kernel_mm) { \ + OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3)); \ + OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ + } else { \ + OUT_BATCH(CP_PACKET2); \ + OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ + } \ + } while(0) + +#define OUT_BATCH_PACKET3_CLIP(packet, num_extra) do { \ + if (!b_l_rmesa->radeonScreen->kernel_mm) { \ + OUT_BATCH(cmdpacket3(RADEON_CMD_PACKET3_CLIP)); \ + OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ + } else { \ + OUT_BATCH(CP_PACKET2); \ + OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ + } \ + } while(0) + + #endif /* __RADEON_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 2abf644ef2..5a0bbaacc9 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -380,8 +380,8 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (rmesa->tcl.indexed_verts.buf) radeonReleaseArrays( ctx, ~0 ); - radeonAllocDmaRegion( rmesa, - &rmesa->tcl.indexed_verts, + radeonAllocDmaRegion( &rmesa->radeon, + 0, &rmesa->tcl.indexed_verts, VB->Count * setup_tab[i].vertex_size * 4, 4); @@ -425,7 +425,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) rmesa->tcl.indexed_verts.start ); rmesa->tcl.vertex_format = setup_tab[i].vertex_format; - rmesa->tcl.indexed_verts.aos_start = GET_START( &rmesa->tcl.indexed_verts ); + // rmesa->tcl.indexed_verts.aos_start = GET_START( &rmesa->tcl.indexed_verts ); rmesa->tcl.indexed_verts.aos_size = setup_tab[i].vertex_size; rmesa->tcl.indexed_verts.aos_stride = setup_tab[i].vertex_size; @@ -444,6 +444,6 @@ void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) _tnl_print_vert_flags( __FUNCTION__, newinputs ); #endif - if (newinputs) - radeonReleaseDmaRegion( rmesa, &rmesa->tcl.indexed_verts, __FUNCTION__ ); + /// if (newinputs) + /// radeonReleaseDmaRegion( rmesa, &rmesa->tcl.indexed_verts, __FUNCTION__ ); } diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index a850baea29..332a0ca69b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -59,21 +59,18 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * information. */ #define LOCAL_VARS \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - const __DRIdrawablePrivate *dPriv = drb->dPriv; \ + struct radeon_renderbuffer *rrb = (void *) rb; \ + const __DRIdrawablePrivate *dPriv = rrb->dPriv; \ const GLuint bottom = dPriv->h - 1; \ - GLubyte *buf = (GLubyte *) drb->flippedData \ - + (dPriv->y * drb->flippedPitch + dPriv->x) * drb->cpp; \ GLuint p; \ - (void) p; + (void) p; #define LOCAL_DEPTH_VARS \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - const __DRIdrawablePrivate *dPriv = drb->dPriv; \ + struct radeon_renderbuffer *rrb = (void *) rb; \ + const __DRIdrawablePrivate *dPriv = rrb->dPriv; \ const GLuint bottom = dPriv->h - 1; \ GLuint xo = dPriv->x; \ - GLuint yo = dPriv->y; \ - GLubyte *buf = (GLubyte *) drb->Base.Data; + GLuint yo = dPriv->y; #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS @@ -94,7 +91,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define TAG(x) radeon##x##_RGB565 #define TAG2(x,y) radeon##x##_RGB565##y -#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 2) +#define GET_PTR(X,Y) radeon_ptr16(rrb, (X), (Y)) #include "spantmp2.h" /* 32 bit, ARGB8888 color spanline and pixel functions @@ -104,82 +101,18 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define TAG(x) radeon##x##_ARGB8888 #define TAG2(x,y) radeon##x##_ARGB8888##y -#define GET_PTR(X,Y) (buf + ((Y) * drb->flippedPitch + (X)) * 4) +#define GET_PTR(X,Y) radeon_ptr32(rrb, (X), (Y)) #include "spantmp2.h" -/* ================================================================ - * Depth buffer - */ - -/* The Radeon family has depth tiling on all the time, so we have to convert - * the x,y coordinates into the memory bus address (mba) in the same - * manner as the engine. In each case, the linear block address (ba) - * is calculated, and then wired with x and y to produce the final - * memory address. - * The chip will do address translation on its own if the surface registers - * are set up correctly. It is not quite enough to get it working with hyperz - * too... - */ - -static GLuint radeon_mba_z32(const driRenderbuffer * drb, GLint x, GLint y) -{ - GLuint pitch = drb->pitch; - if (drb->depthHasSurface) { - return 4 * (x + y * pitch); - } else { - GLuint ba, address = 0; /* a[0..1] = 0 */ - -#ifdef COMPILE_R300 - ba = (y / 8) * (pitch / 8) + (x / 8); -#else - ba = (y / 16) * (pitch / 16) + (x / 16); -#endif - - address |= (x & 0x7) << 2; /* a[2..4] = x[0..2] */ - address |= (y & 0x3) << 5; /* a[5..6] = y[0..1] */ - address |= (((x & 0x10) >> 2) ^ (y & 0x4)) << 5; /* a[7] = x[4] ^ y[2] */ - address |= (ba & 0x3) << 8; /* a[8..9] = ba[0..1] */ - - address |= (y & 0x8) << 7; /* a[10] = y[3] */ - address |= (((x & 0x8) << 1) ^ (y & 0x10)) << 7; /* a[11] = x[3] ^ y[4] */ - address |= (ba & ~0x3) << 10; /* a[12..] = ba[2..] */ - - return address; - } -} - -static INLINE GLuint -radeon_mba_z16(const driRenderbuffer * drb, GLint x, GLint y) -{ - GLuint pitch = drb->pitch; - if (drb->depthHasSurface) { - return 2 * (x + y * pitch); - } else { - GLuint ba, address = 0; /* a[0] = 0 */ - - ba = (y / 16) * (pitch / 32) + (x / 32); - - address |= (x & 0x7) << 1; /* a[1..3] = x[0..2] */ - address |= (y & 0x7) << 4; /* a[4..6] = y[0..2] */ - address |= (x & 0x8) << 4; /* a[7] = x[3] */ - address |= (ba & 0x3) << 8; /* a[8..9] = ba[0..1] */ - address |= (y & 0x8) << 7; /* a[10] = y[3] */ - address |= ((x & 0x10) ^ (y & 0x10)) << 7; /* a[11] = x[4] ^ y[4] */ - address |= (ba & ~0x3) << 10; /* a[12..] = ba[2..] */ - - return address; - } -} - /* 16-bit depth buffer functions */ #define VALUE_TYPE GLushort #define WRITE_DEPTH( _x, _y, d ) \ - *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo )) = d; + *(GLushort *)radeon_ptr(rrb, _x + xo, _y + yo) = d #define READ_DEPTH( d, _x, _y ) \ - d = *(GLushort *)(buf + radeon_mba_z16( drb, _x + xo, _y + yo )); + d = *(GLushort *)radeon_ptr(rrb, _x + xo, _y + yo) #define TAG(x) radeon##x##_z16 #include "depthtmp.h" @@ -203,24 +136,25 @@ do { \ #else #define WRITE_DEPTH( _x, _y, d ) \ do { \ - GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \ - GLuint tmp = *(GLuint *)(buf + offset); \ + GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x + xo, _y + yo); \ + GLuint tmp = *_ptr; \ tmp &= 0xff000000; \ tmp |= ((d) & 0x00ffffff); \ - *(GLuint *)(buf + offset) = tmp; \ + *_ptr = tmp; \ } while (0) #endif #ifdef COMPILE_R300 #define READ_DEPTH( d, _x, _y ) \ - do { \ + do { \ d = (*(GLuint *)(buf + radeon_mba_z32( drb, _x + xo, \ _y + yo )) & 0xffffff00) >> 8; \ }while(0) #else #define READ_DEPTH( d, _x, _y ) \ - d = *(GLuint *)(buf + radeon_mba_z32( drb, _x + xo, \ - _y + yo )) & 0x00ffffff; + do { \ + d = (*(GLuint*)(radeon_ptr32(rrb, _x + xo, _y + yo)) & 0x00ffffff); \ + } while (0) #endif #define TAG(x) radeon##x##_z24_s8 @@ -244,11 +178,11 @@ do { \ #else #define WRITE_STENCIL( _x, _y, d ) \ do { \ - GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \ - GLuint tmp = *(GLuint *)(buf + offset); \ + GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x + xo, _y + yo); \ + GLuint tmp = *_ptr; \ tmp &= 0x00ffffff; \ tmp |= (((d) & 0xff) << 24); \ - *(GLuint *)(buf + offset) = tmp; \ + *_ptr = tmp; \ } while (0) #endif @@ -262,8 +196,8 @@ do { \ #else #define READ_STENCIL( d, _x, _y ) \ do { \ - GLuint offset = radeon_mba_z32( drb, _x + xo, _y + yo ); \ - GLuint tmp = *(GLuint *)(buf + offset); \ + GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x + xo, _y + yo); \ + GLuint tmp = *_ptr; \ d = (tmp & 0xff000000) >> 24; \ } while (0) #endif @@ -271,32 +205,6 @@ do { \ #define TAG(x) radeon##x##_z24_s8 #include "stenciltmp.h" -/* Move locking out to get reasonable span performance (10x better - * than doing this in HW_LOCK above). WaitForIdle() is the main - * culprit. - */ - -static void radeonSpanRenderStart(GLcontext * ctx) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); -#ifdef COMPILE_R300 - r300ContextPtr r300 = (r300ContextPtr) rmesa; - R300_FIREVERTICES(r300); -#else - r100ContextPtr r100 = (r100ContextPtr) rmesa; - RADEON_FIREVERTICES(r100); -#endif - LOCK_HARDWARE(rmesa); - radeonWaitForIdleLocked(rmesa); -} - -static void radeonSpanRenderFinish(GLcontext * ctx) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - _swrast_flush(ctx); - UNLOCK_HARDWARE(rmesa); -} - void radeonInitSpanFuncs(GLcontext * ctx) { struct swrast_device_driver *swdd = diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index ef89d73bdd..4b49c141a0 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -52,8 +52,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_tcl.h" -static void flush_last_swtcl_prim(GLcontext *ctx); - /* R100: xyzw, c0, c1/fog, stq[0..2] = 4+1+1+3*3 = 15 right? */ /* R200: xyzw, c0, c1/fog, strq[0..5] = 4+1+1+4*6 = 30 */ #define RADEON_MAX_TNL_VERTEX_SIZE (15 * sizeof(GLfloat)) /* for mesa _tnl stage */ @@ -64,18 +62,18 @@ static void flush_last_swtcl_prim(GLcontext *ctx); #define EMIT_ATTR( ATTR, STYLE, F0 ) \ do { \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = (ATTR); \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = (STYLE); \ - rmesa->swtcl.vertex_attr_count++; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR); \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE); \ + rmesa->radeon.swtcl.vertex_attr_count++; \ fmt_0 |= F0; \ } while (0) #define EMIT_PAD( N ) \ do { \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].attrib = 0; \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].format = EMIT_PAD; \ - rmesa->swtcl.vertex_attrs[rmesa->swtcl.vertex_attr_count].offset = (N); \ - rmesa->swtcl.vertex_attr_count++; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = 0; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = EMIT_PAD; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].offset = (N); \ + rmesa->radeon.swtcl.vertex_attr_count++; \ } while (0) static GLuint radeon_cp_vc_frmts[3][2] = @@ -106,7 +104,7 @@ static void radeonSetVertexFormat( GLcontext *ctx ) } assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL ); - rmesa->swtcl.vertex_attr_count = 0; + rmesa->radeon.swtcl.vertex_attr_count = 0; /* EMIT_ATTR's must be in order as they tell t_vertex.c how to * build up a hardware vertex. @@ -208,16 +206,16 @@ static void radeonSetVertexFormat( GLcontext *ctx ) fmt_0 != rmesa->swtcl.vertex_format) { RADEON_NEWPRIM(rmesa); rmesa->swtcl.vertex_format = fmt_0; - rmesa->swtcl.vertex_size = + rmesa->radeon.swtcl.vertex_size = _tnl_install_attrs( ctx, - rmesa->swtcl.vertex_attrs, - rmesa->swtcl.vertex_attr_count, + rmesa->radeon.swtcl.vertex_attrs, + rmesa->radeon.swtcl.vertex_attr_count, NULL, 0 ); - rmesa->swtcl.vertex_size /= 4; + rmesa->radeon.swtcl.vertex_size /= 4; RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset ); if (RADEON_DEBUG & DEBUG_VERTS) fprintf( stderr, "%s: vertex_size= %d floats\n", - __FUNCTION__, rmesa->swtcl.vertex_size); + __FUNCTION__, rmesa->radeon.swtcl.vertex_size); } } @@ -228,9 +226,9 @@ static void radeonRenderStart( GLcontext *ctx ) radeonSetVertexFormat( ctx ); - if (rmesa->dma.flush != 0 && - rmesa->dma.flush != flush_last_swtcl_prim) - rmesa->dma.flush( ctx ); + if (rmesa->radeon.dma.flush != 0 && + rmesa->radeon.dma.flush != rcommon_flush_last_swtcl_prim) + rmesa->radeon.dma.flush( ctx ); } @@ -281,81 +279,29 @@ void radeonChooseVertexState( GLcontext *ctx ) } } - -/* Flush vertices in the current dma region. - */ -static void flush_last_swtcl_prim(GLcontext *ctx) +void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) { r100ContextPtr rmesa = R100_CONTEXT(ctx); - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - - rmesa->dma.flush = NULL; - - if (rmesa->dma.current.buf) { - struct radeon_dma_region *current = &rmesa->dma.current; - GLuint current_offset = (rmesa->radeon.radeonScreen->gart_buffer_offset + - current->buf->buf->idx * RADEON_BUFFER_SIZE + - current->start); - - assert (!(rmesa->swtcl.hw_primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND)); - - assert (current->start + - rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == - current->ptr); - - if (rmesa->dma.current.start != rmesa->dma.current.ptr) { - radeonEnsureCmdBufSpace( rmesa, VERT_AOS_BUFSZ + - rmesa->hw.max_state_size + VBUF_BUFSZ ); - - radeonEmitVertexAOS( rmesa, - rmesa->swtcl.vertex_size, - current_offset); - - radeonEmitVbufPrim( rmesa, - rmesa->swtcl.vertex_format, - rmesa->swtcl.hw_primitive, - rmesa->swtcl.numverts); - } - - rmesa->swtcl.numverts = 0; - current->start = current->ptr; - } -} - - -/* Alloc space in the current dma region. - */ -static INLINE void * -radeonAllocDmaLowVerts( r100ContextPtr rmesa, int nverts, int vsize ) -{ - GLuint bytes = vsize * nverts; - if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) - radeonRefillCurrentDmaRegion( rmesa ); + rcommonEnsureCmdBufSpace(&rmesa->radeon, + rmesa->hw.max_state_size + (12*sizeof(int)), + __FUNCTION__); - if (!rmesa->dma.flush) { - rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; - rmesa->dma.flush = flush_last_swtcl_prim; - } - - assert( vsize == rmesa->swtcl.vertex_size * 4 ); - assert( rmesa->dma.flush == flush_last_swtcl_prim ); - assert (rmesa->dma.current.start + - rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == - rmesa->dma.current.ptr); + radeonEmitState(rmesa); + radeonEmitVertexAOS( rmesa, + rmesa->radeon.swtcl.vertex_size, + // rmesa->radeon.dma.current, + current_offset); - { - GLubyte *head = (GLubyte *)(rmesa->dma.current.address + rmesa->dma.current.ptr); - rmesa->dma.current.ptr += bytes; - rmesa->swtcl.numverts += nverts; - return head; - } + + radeonEmitVbufPrim( rmesa, + rmesa->swtcl.vertex_format, + rmesa->radeon.swtcl.hw_primitive, + rmesa->radeon.swtcl.numverts); } - /* * Render unclipped vertex buffers by emitting vertices directly to * dma buffers. Use strip/fan hardware primitives where possible. @@ -391,19 +337,19 @@ static INLINE void radeonDmaPrimitive( r100ContextPtr rmesa, GLenum prim ) { RADEON_NEWPRIM( rmesa ); - rmesa->swtcl.hw_primitive = hw_prim[prim]; - assert(rmesa->dma.current.ptr == rmesa->dma.current.start); + rmesa->radeon.swtcl.hw_primitive = hw_prim[prim]; + // assert(rmesa->radeon.dma.current.ptr == rmesa->radeon.dma.current.start); } #define LOCAL_VARS r100ContextPtr rmesa = R100_CONTEXT(ctx) #define INIT( prim ) radeonDmaPrimitive( rmesa, prim ) #define FLUSH() RADEON_NEWPRIM( rmesa ) -#define GET_CURRENT_VB_MAX_VERTS() \ - (((int)rmesa->dma.current.end - (int)rmesa->dma.current.ptr) / (rmesa->swtcl.vertex_size*4)) +#define GET_CURRENT_VB_MAX_VERTS() 10\ +// (((int)rmesa->radeon.dma.current.end - (int)rmesa->radeon.dma.current.ptr) / (rmesa->radeon.swtcl.vertex_size*4)) #define GET_SUBSEQUENT_VB_MAX_VERTS() \ - ((RADEON_BUFFER_SIZE) / (rmesa->swtcl.vertex_size*4)) + ((RADEON_BUFFER_SIZE) / (rmesa->radeon.swtcl.vertex_size*4)) #define ALLOC_VERTS( nr ) \ - radeonAllocDmaLowVerts( rmesa, nr, rmesa->swtcl.vertex_size * 4 ) + rcommonAllocDmaLowVerts( &rmesa->radeon, nr, rmesa->radeon.swtcl.vertex_size * 4 ) #define EMIT_VERTS( ctx, j, nr, buf ) \ _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf) @@ -428,7 +374,7 @@ static GLboolean radeon_run_render( GLcontext *ctx, if (rmesa->swtcl.indexed_verts.buf) RELEASE_ELT_VERTS(); - if (rmesa->swtcl.RenderIndex != 0 || + if (rmesa->radeon.swtcl.RenderIndex != 0 || !radeon_dma_validate_render( ctx, VB )) return GL_TRUE; @@ -498,12 +444,12 @@ static void radeonResetLineStipple( GLcontext *ctx ); #undef LOCAL_VARS #undef ALLOC_VERTS #define CTX_ARG r100ContextPtr rmesa -#define GET_VERTEX_DWORDS() rmesa->swtcl.vertex_size -#define ALLOC_VERTS( n, size ) radeonAllocDmaLowVerts( rmesa, n, (size) * 4 ) +#define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size +#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, (size) * 4 ) #undef LOCAL_VARS #define LOCAL_VARS \ r100ContextPtr rmesa = R100_CONTEXT(ctx); \ - const char *radeonverts = (char *)rmesa->swtcl.verts; + const char *radeonverts = (char *)rmesa->radeon.swtcl.verts; #define VERT(x) (radeonVertex *)(radeonverts + ((x) * (vertsize) * sizeof(int))) #define VERTEX radeonVertex #undef TAG @@ -561,7 +507,7 @@ static struct { #define VERT_Y(_v) _v->v.y #define VERT_Z(_v) _v->v.z #define AREA_IS_CCW( a ) (a < 0) -#define GET_VERTEX(e) (rmesa->swtcl.verts + ((e) * rmesa->swtcl.vertex_size * sizeof(int))) +#define GET_VERTEX(e) (rmesa->radeon.swtcl.verts + ((e) * rmesa->radeon.swtcl.vertex_size * sizeof(int))) #define VERT_SET_RGBA( v, c ) \ do { \ @@ -618,7 +564,7 @@ do { \ ***********************************************************************/ #define RASTERIZE(x) radeonRasterPrimitive( ctx, reduced_hw_prim[x] ) -#define RENDER_PRIMITIVE rmesa->swtcl.render_primitive +#define RENDER_PRIMITIVE rmesa->radeon.swtcl.render_primitive #undef TAG #define TAG(x) x #include "tnl_dd/t_dd_unfilled.h" @@ -675,8 +621,8 @@ static void init_rast_tab( void ) #undef LOCAL_VARS #define LOCAL_VARS \ r100ContextPtr rmesa = R100_CONTEXT(ctx); \ - const GLuint vertsize = rmesa->swtcl.vertex_size; \ - const char *radeonverts = (char *)rmesa->swtcl.verts; \ + const GLuint vertsize = rmesa->radeon.swtcl.vertex_size; \ + const char *radeonverts = (char *)rmesa->radeon.swtcl.verts; \ const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \ const GLboolean stipple = ctx->Line.StippleFlag; \ (void) elt; (void) stipple; @@ -711,7 +657,7 @@ void radeonChooseRenderState( GLcontext *ctx ) if (flags & DD_TRI_LIGHT_TWOSIDE) index |= RADEON_TWOSIDE_BIT; if (flags & DD_TRI_UNFILLED) index |= RADEON_UNFILLED_BIT; - if (index != rmesa->swtcl.RenderIndex) { + if (index != rmesa->radeon.swtcl.RenderIndex) { tnl->Driver.Render.Points = rast_tab[index].points; tnl->Driver.Render.Line = rast_tab[index].line; tnl->Driver.Render.ClippedLine = rast_tab[index].line; @@ -728,7 +674,7 @@ void radeonChooseRenderState( GLcontext *ctx ) tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon; } - rmesa->swtcl.RenderIndex = index; + rmesa->radeon.swtcl.RenderIndex = index; } } @@ -742,16 +688,16 @@ static void radeonRasterPrimitive( GLcontext *ctx, GLuint hwprim ) { r100ContextPtr rmesa = R100_CONTEXT(ctx); - if (rmesa->swtcl.hw_primitive != hwprim) { + if (rmesa->radeon.swtcl.hw_primitive != hwprim) { RADEON_NEWPRIM( rmesa ); - rmesa->swtcl.hw_primitive = hwprim; + rmesa->radeon.swtcl.hw_primitive = hwprim; } } static void radeonRenderPrimitive( GLcontext *ctx, GLenum prim ) { r100ContextPtr rmesa = R100_CONTEXT(ctx); - rmesa->swtcl.render_primitive = prim; + rmesa->radeon.swtcl.render_primitive = prim; if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) radeonRasterPrimitive( ctx, reduced_hw_prim[prim] ); } @@ -806,7 +752,7 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) RADEON_FIREVERTICES( rmesa ); TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_TRUE ); _swsetup_Wakeup( ctx ); - rmesa->swtcl.RenderIndex = ~0; + rmesa->radeon.swtcl.RenderIndex = ~0; if (RADEON_DEBUG & DEBUG_FALLBACKS) { fprintf(stderr, "Radeon begin rasterization fallback: 0x%x %s\n", bit, getFallbackString(bit)); @@ -873,10 +819,10 @@ void radeonInitSwtcl( GLcontext *ctx ) _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, RADEON_MAX_TNL_VERTEX_SIZE); - rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf; - rmesa->swtcl.RenderIndex = ~0; - rmesa->swtcl.render_primitive = GL_TRIANGLES; - rmesa->swtcl.hw_primitive = 0; + rmesa->radeon.swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf; + rmesa->radeon.swtcl.RenderIndex = ~0; + rmesa->radeon.swtcl.render_primitive = GL_TRIANGLES; + rmesa->radeon.swtcl.hw_primitive = 0; } @@ -884,7 +830,7 @@ void radeonDestroySwtcl( GLcontext *ctx ) { r100ContextPtr rmesa = R100_CONTEXT(ctx); - if (rmesa->swtcl.indexed_verts.buf) - radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, - __FUNCTION__ ); + // if (rmesa->swtcl.indexed_verts.buf) + // radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, + // __FUNCTION__ ); } diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.h b/src/mesa/drivers/dri/radeon/radeon_swtcl.h index e485052ad7..3ada9890b3 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.h +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.h @@ -63,5 +63,5 @@ extern void radeon_translate_vertex( GLcontext *ctx, extern void radeon_print_vertex( GLcontext *ctx, const radeonVertex *v ); - +extern void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset); #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index 81c0c84257..a54e471065 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -145,15 +145,15 @@ static GLboolean discrete_prim[0x10] = { static GLushort *radeonAllocElts( r100ContextPtr rmesa, GLuint nr ) { - if (rmesa->dma.flush) - rmesa->dma.flush( rmesa->radeon.glCtx ); + if (rmesa->radeon.dma.flush) + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); - radeonEnsureCmdBufSpace(rmesa, AOS_BUFSZ(rmesa->tcl.nr_aos_components) + + rcommonEnsureCmdBufSpace(&rmesa->radeon, AOS_BUFSZ(rmesa->tcl.nr_aos_components) + rmesa->hw.max_state_size + ELTS_BUFSZ(nr)); radeonEmitAOS( rmesa, - rmesa->tcl.aos_components, - rmesa->tcl.nr_aos_components, 0 ); + rmesa->tcl.aos_components, + rmesa->tcl.nr_aos_components, 0 ); return radeonAllocEltsOpenEnded( rmesa, rmesa->tcl.vertex_format, @@ -182,12 +182,14 @@ static void radeonEmitPrim( GLcontext *ctx, rmesa->hw.max_state_size + VBUF_BUFSZ ); radeonEmitAOS( rmesa, + rmesa->tcl.aos_components, rmesa->tcl.nr_aos_components, start ); /* Why couldn't this packet have taken an offset param? */ radeonEmitVbufPrim( rmesa, + 0, rmesa->tcl.hw_primitive, count - start ); } @@ -508,15 +510,15 @@ static void transition_to_hwtnl( GLcontext *ctx ) tnl->Driver.NotifyMaterialChange = radeonUpdateMaterial; - if ( rmesa->dma.flush ) - rmesa->dma.flush( rmesa->radeon.glCtx ); + if ( rmesa->radeon.dma.flush ) + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); - rmesa->dma.flush = NULL; + rmesa->radeon.dma.flush = NULL; rmesa->swtcl.vertex_format = 0; - if (rmesa->swtcl.indexed_verts.buf) - radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, - __FUNCTION__ ); + // if (rmesa->swtcl.indexed_verts.buf) + // radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, + // __FUNCTION__ ); if (RADEON_DEBUG & DEBUG_FALLBACKS) fprintf(stderr, "Radeon end tcl fallback\n"); diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c index c6a911abef..9ef08fbe68 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex.c @@ -44,6 +44,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/texobj.h" #include "radeon_context.h" +#include "radeon_mipmap_tree.h" #include "radeon_state.h" #include "radeon_ioctl.h" #include "radeon_swtcl.h" @@ -453,7 +454,7 @@ void radeonInitTextureFuncs( struct dd_function_table *functions ) functions->CompressedTexImage2D = radeonCompressedTexImage2D; functions->CompressedTexSubImage2D = radeonCompressedTexSubImage2D; - functions->GenerateMipmap = radeon_generate_mipmap; + functions->GenerateMipmap = radeonGenerateMipmap; functions->NewTextureImage = radeonNewTextureImage; functions->FreeTexImageData = radeonFreeTexImageData; diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index 9ede92b831..3c28d70e37 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -43,6 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/enums.h" #include "radeon_context.h" +#include "radeon_mipmap_tree.h" #include "radeon_state.h" #include "radeon_ioctl.h" #include "radeon_swtcl.h" @@ -75,10 +76,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define VALID_FORMAT(f) ( ((f) <= MESA_FORMAT_RGBA_DXT5) \ && (tx_table[f].format != 0xffffffff) ) -static const struct { +struct tx_table { GLuint format, filter; -} -tx_table[] = +}; + +static const struct tx_table tx_table[] = { _ALPHA(RGBA8888), _ALPHA_REV(RGBA8888), @@ -901,13 +903,13 @@ void radeonSetTexOffset(__DRIcontext * pDRICtx, GLint texname, RADEON_TXFORMAT_CUBIC_MAP_ENABLE | \ RADEON_TXFORMAT_NON_POWER2) - +#if 0 static void import_tex_obj_state( r100ContextPtr rmesa, int unit, radeonTexObjPtr texobj ) { /* do not use RADEON_DB_STATE to avoid stale texture caches */ - int *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0]; + uint32_t *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0]; GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT]; RADEON_STATECHANGE( rmesa, tex[unit] ); @@ -955,7 +957,7 @@ static void import_tex_obj_state( r100ContextPtr rmesa, texobj->dirty_state &= ~(1<image_override) { if (VALID_FORMAT(firstImage->TexFormat->MesaFormat)) { - const struct tx_table *table = _mesa_little_endian() ? tx_table_le : - tx_table_be; + const struct tx_table *table = tx_table; t->pp_txformat &= ~(RADEON_TXFORMAT_FORMAT_MASK | RADEON_TXFORMAT_ALPHA_IN_MAP); @@ -1399,8 +1400,8 @@ static void setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t) (log2Height << RADEON_FACE_HEIGHT_4_SHIFT)); } - t->pp_txsize = (((firstImage->Width - 1) << RADEON_PP_TX_WIDTHMASK_SHIFT) - | ((firstImage->Height - 1) << RADEON_PP_TX_HEIGHTMASK_SHIFT)); + t->pp_txsize = (((firstImage->Width - 1) << RADEON_TEX_USIZE_SHIFT) + | ((firstImage->Height - 1) << RADEON_TEX_VSIZE_SHIFT)); if ( !t->image_override ) { if (firstImage->IsCompressed) -- cgit v1.2.3 From 3fafaf8959681cc41c988607bb6e387bab4fe1b5 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Mon, 9 Feb 2009 03:50:38 +1000 Subject: radeon: make more r100 work --- src/mesa/drivers/dri/radeon/common_context.h | 1 + src/mesa/drivers/dri/radeon/radeon_context.h | 15 +- src/mesa/drivers/dri/radeon/radeon_cs_legacy.c | 2 +- src/mesa/drivers/dri/radeon/radeon_ioctl.c | 278 +++++++++++----- src/mesa/drivers/dri/radeon/radeon_ioctl.h | 3 +- src/mesa/drivers/dri/radeon/radeon_maos_arrays.c | 84 ++--- src/mesa/drivers/dri/radeon/radeon_maos_verts.c | 29 +- src/mesa/drivers/dri/radeon/radeon_state_init.c | 383 +++++++++++++++++++++-- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 3 +- src/mesa/drivers/dri/radeon/radeon_tcl.c | 38 ++- src/mesa/drivers/dri/radeon/radeon_tex.c | 2 +- src/mesa/drivers/dri/radeon/radeon_texstate.c | 46 ++- 12 files changed, 670 insertions(+), 214 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/radeon/common_context.h b/src/mesa/drivers/dri/radeon/common_context.h index 90abca0bc1..618e74d458 100644 --- a/src/mesa/drivers/dri/radeon/common_context.h +++ b/src/mesa/drivers/dri/radeon/common_context.h @@ -269,6 +269,7 @@ struct radeon_swtcl_info { struct radeon_ioctl { GLuint vertex_offset; + struct radeon_bo *bo; GLuint vertex_size; }; diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h index 6edbaf6910..358095a2b9 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_context.h @@ -334,11 +334,6 @@ struct r100_state { struct radeon_texture_state texture; }; -#define GET_START(rvb) (rmesa->radeon.radeonScreen->gart_buffer_offset + \ - (rvb)->address - rmesa->dma.buf0_address + \ - (rvb)->start) - - #define RADEON_CMD_BUF_SZ (8*1024) #define R200_ELT_BUF_SZ (8*1024) /* radeon_tcl.c @@ -352,12 +347,14 @@ struct radeon_tcl_info { */ GLvector4f ObjClean; - struct radeon_dma_region *aos_components[8]; + struct radeon_aos aos[8]; GLuint nr_aos_components; GLuint *Elts; - struct radeon_dma_region indexed_verts; + struct radeon_bo *indexed_bo; + +// struct radeon_dma_region indexed_verts; struct radeon_dma_region obj; struct radeon_dma_region rgba; struct radeon_dma_region spec; @@ -365,8 +362,8 @@ struct radeon_tcl_info { struct radeon_dma_region tex[RADEON_MAX_TEXTURE_UNITS]; struct radeon_dma_region norm; - struct radeon_bo *elt_dma_bo; - int elt_dma_offset; /** Offset into this buffer object, in bytes */ + int elt_cmd_offset; /** Offset into the cmdbuf */ + int elt_cmd_start; int elt_used; }; diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c index 1c885c1612..fbb93a8ba1 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c @@ -313,7 +313,7 @@ static int cs_emit(struct radeon_cs *cs) cmd.boxes = (drm_clip_rect_t *) csm->ctx->pClipRects; } - //dump_cmdbuf(cs); + //dump_cmdbuf(cs); r = drmCommandWrite(cs->csm->fd, DRM_RADEON_CMDBUF, &cmd, sizeof(cmd)); if (r) { diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index 4ab297c5ce..1a33595884 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -59,7 +59,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define RADEON_TIMEOUT 512 #define RADEON_IDLE_RETRY 16 -#define DEBUG_CMDBUF 0 +#define DEBUG_CMDBUF 1 static void radeonSaveHwState( r100ContextPtr rmesa ) { @@ -190,10 +190,6 @@ static INLINE void radeonEmitAtoms(r100ContextPtr r100, GLboolean dirty) void radeonEmitState( r100ContextPtr rmesa ) { - struct radeon_state_atom *atom; - char *dest; - uint32_t dwords; - if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) fprintf(stderr, "%s\n", __FUNCTION__); @@ -212,7 +208,6 @@ void radeonEmitState( r100ContextPtr rmesa ) * radeonAllocCmdBuf code here without all the checks. */ rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->hw.max_state_size, __FUNCTION__); - dest = rmesa->store.cmd_buf + rmesa->store.cmd_used; /* We always always emit zbs, this is due to a bug found by keithw in the hardware and rediscovered after Erics changes by me. @@ -250,12 +245,26 @@ extern void radeonEmitVbufPrim( r100ContextPtr rmesa, radeonEmitState( rmesa ); - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s cmd_used/4: %d\n", __FUNCTION__, - rmesa->store.cmd_used/4); + // if (RADEON_DEBUG & DEBUG_IOCTL) + // fprintf(stderr, "%s cmd_used/4: %d\n", __FUNCTION__, + // rmesa->store.cmd_used/4); + + +#if RADEON_OLD_PACKETS + BEGIN_BATCH(6); + OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM, 3); + OUT_BATCH_RELOC(rmesa->ioctl.vertex_offset, rmesa->ioctl.bo, rmesa->ioctl.vertex_offset, RADEON_GEM_DOMAIN_GTT, 0, 0); + OUT_BATCH(vertex_nr); + OUT_BATCH(vertex_format); + OUT_BATCH(primitive | RADEON_CP_VC_CNTL_PRIM_WALK_LIST | + RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA | + RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE | + (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT)); + END_BATCH(); - BEGIN_BATCH(3); - OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_DRAW_VBUF, 0); +#else + BEGIN_BATCH(4); + OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_DRAW_VBUF, 1); OUT_BATCH(vertex_format); OUT_BATCH(primitive | RADEON_CP_VC_CNTL_PRIM_WALK_LIST | @@ -264,45 +273,89 @@ extern void radeonEmitVbufPrim( r100ContextPtr rmesa, RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE | (vertex_nr << RADEON_CP_VC_CNTL_NUM_SHIFT)); END_BATCH(); +#endif } +static void radeonFireEB(r100ContextPtr rmesa, int vertex_count, int vertex_format, int type) +{ + BATCH_LOCALS(&rmesa->radeon); + + if (vertex_count > 0) { + BEGIN_BATCH(8); + OUT_BATCH_PACKET3(RADEON_CP_PACKET3_3D_DRAW_INDX, 0); + OUT_BATCH(vertex_format); + OUT_BATCH(RADEON_CP_VC_CNTL_PRIM_WALK_IND | + ((vertex_count + 0) << 16) | + type); + + assert(0); // RADEON HAS NO INDX_BUFFERs +#if 0 + + + if (!rmesa->radeon.radeonScreen->kernel_mm) { + OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2); + OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810); + OUT_BATCH_RELOC(rmesa->tcl.elt_dma_offset, + rmesa->tcl.elt_dma_bo, + rmesa->tcl.elt_dma_offset, + RADEON_GEM_DOMAIN_GTT, 0, 0); + OUT_BATCH(vertex_count/2); + } else { + OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2); + OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810); + OUT_BATCH(rmesa->tcl.elt_dma_offset); + OUT_BATCH(vertex_count/2); + radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, + rmesa->tcl.elt_dma_bo, + RADEON_GEM_DOMAIN_GTT, 0, 0); + } +#endif + END_BATCH(); + } +} void radeonFlushElts( GLcontext *ctx ) { r100ContextPtr rmesa = R100_CONTEXT(ctx); - int *cmd = (int *)(rmesa->store.cmd_buf + rmesa->store.elts_start); + BATCH_LOCALS(&rmesa->radeon); int dwords; -#if RADEON_OLD_PACKETS - int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 24)) / 2; -#else - int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 16)) / 2; -#endif - + uint32_t *cmd = (uint32_t *)(rmesa->radeon.cmdbuf.cs->packets + rmesa->tcl.elt_cmd_start); + int nr = (rmesa->radeon.cmdbuf.cs->section_ndw - rmesa->radeon.cmdbuf.cs->section_cdw) * 2; + if (RADEON_DEBUG & DEBUG_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); assert( rmesa->radeon.dma.flush == radeonFlushElts ); rmesa->radeon.dma.flush = NULL; + /* Cope with odd number of elts: */ - rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2; - dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4; + // rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2; + // dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4; + dwords = nr / 2; + + rmesa->radeon.cmdbuf.cs->cdw += dwords; #if RADEON_OLD_PACKETS - cmd[1] |= (dwords - 3) << 16; + cmd[1] |= (dwords + 3) << 16; cmd[5] |= nr << RADEON_CP_VC_CNTL_NUM_SHIFT; #else - cmd[1] |= (dwords - 3) << 16; + cmd[1] |= (dwords) << 16; cmd[3] |= nr << RADEON_CP_VC_CNTL_NUM_SHIFT; #endif + fprintf(stderr,"nr is %d cmd1 is %08x\n", nr, cmd[1]); + + rmesa->radeon.cmdbuf.cs->section_cdw += dwords; + END_BATCH(); + if (RADEON_DEBUG & DEBUG_SYNC) { fprintf(stderr, "%s: Syncing\n", __FUNCTION__); radeonFinish( rmesa->radeon.glCtx ); } -} +} GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa, GLuint vertex_format, @@ -310,46 +363,65 @@ GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa, GLuint min_nr ) { GLushort *retval; - + BATCH_LOCALS(&rmesa->radeon); if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s %d\n", __FUNCTION__, min_nr); + fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive); assert((primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND)); radeonEmitState( rmesa ); - rmesa->tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom, - 0, R200_ELT_BUF_SZ, 4, - RADEON_GEM_DOMAIN_GTT, 0); - rmesa->tcl.elt_dma_offset = 0; + rmesa->tcl.elt_cmd_start = rmesa->radeon.cmdbuf.cs->cdw; + +#if RADEON_OLD_PACKETS + BEGIN_BATCH_NO_AUTOSTATE(2+ELTS_BUFSZ(min_nr)/4); + OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM, 0); + OUT_BATCH_RELOC(rmesa->ioctl.vertex_offset, rmesa->ioctl.bo, rmesa->ioctl.vertex_offset, RADEON_GEM_DOMAIN_GTT, 0, 0); + OUT_BATCH(0xffff); + OUT_BATCH(vertex_format); + OUT_BATCH(primitive | + RADEON_CP_VC_CNTL_PRIM_WALK_IND | + RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA | + RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE); + +#else + BEGIN_BATCH_NO_AUTOSTATE(ELTS_BUFSZ(min_nr)/4); + OUT_BATCH_PACKET3_CLIP(RADEON_CP_PACKET3_DRAW_INDX, 0); + OUT_BATCH(vertex_format); + OUT_BATCH(primitive | + RADEON_CP_VC_CNTL_PRIM_WALK_IND | + RADEON_CP_VC_CNTL_COLOR_ORDER_RGBA | + RADEON_CP_VC_CNTL_MAOS_ENABLE | + RADEON_CP_VC_CNTL_VTX_FMT_RADEON_MODE); +#endif + + + rmesa->tcl.elt_cmd_offset = rmesa->radeon.cmdbuf.cs->cdw; rmesa->tcl.elt_used = min_nr * 2; - radeon_bo_map(rmesa->tcl.elt_dma_bo, 1); - retval = rmesa->tcl.elt_dma_bo->ptr + rmesa->tcl.elt_dma_offset; + retval = (GLushort *)(rmesa->radeon.cmdbuf.cs->packets + rmesa->tcl.elt_cmd_offset); + + fprintf(stderr," %d elt start %d offset %d\n", min_nr, rmesa->tcl.elt_cmd_start, rmesa->tcl.elt_cmd_offset); if (RADEON_DEBUG & DEBUG_PRIMS) - fprintf(stderr, "%s: header vfmt 0x%x prim %x \n", - __FUNCTION__, - vertex_format, primitive); + fprintf(stderr, "%s: header prim %x \n", + __FUNCTION__, primitive); assert(!rmesa->radeon.dma.flush); rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; rmesa->radeon.dma.flush = radeonFlushElts; - // rmesa->store.elts_start = ((char *)cmd) - rmesa->store.cmd_buf; - return retval; } - - void radeonEmitVertexAOS( r100ContextPtr rmesa, GLuint vertex_size, + struct radeon_bo *bo, GLuint offset ) { #if RADEON_OLD_PACKETS - rmesa->ioctl.vertex_size = vertex_size; rmesa->ioctl.vertex_offset = offset; + rmesa->ioctl.bo = bo; #else BATCH_LOCALS(&rmesa->radeon); @@ -357,68 +429,124 @@ void radeonEmitVertexAOS( r100ContextPtr rmesa, fprintf(stderr, "%s: vertex_size 0x%x offset 0x%x \n", __FUNCTION__, vertex_size, offset); - BEGIN_BATCH(5); + BEGIN_BATCH(7); OUT_BATCH_PACKET3(RADEON_CP_PACKET3_3D_LOAD_VBPNTR, 2); OUT_BATCH(1); OUT_BATCH(vertex_size | (vertex_size << 8)); OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0); END_BATCH(); -} + #endif } void radeonEmitAOS( r100ContextPtr rmesa, - struct radeon_dma_region **component, GLuint nr, GLuint offset ) { #if RADEON_OLD_PACKETS assert( nr == 1 ); - assert( component[0]->aos_size == component[0]->aos_stride ); - rmesa->ioctl.vertex_size = component[0]->aos_size; + // assert( rmesa->radeon.aos[0]->aos_size == component[0]->aos_stride ); + // rmesa->ioctl.vertex_offset = + // (component[0]->aos_start + offset * component[0]->aos_stride * 4); + rmesa->ioctl.bo = rmesa->tcl.aos[0].bo; rmesa->ioctl.vertex_offset = - (component[0]->aos_start + offset * component[0]->aos_stride * 4); + (rmesa->tcl.aos[0].offset + offset * rmesa->tcl.aos[0].stride * 4); #else - drm_radeon_cmd_header_t *cmd; - int sz = AOS_BUFSZ(nr); + BATCH_LOCALS(&rmesa->radeon); + uint32_t voffset; + // int sz = AOS_BUFSZ(nr); + int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2; int i; - int *tmp; if (RADEON_DEBUG & DEBUG_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); - - cmd = (drm_radeon_cmd_header_t *)radeonAllocCmdBuf( rmesa, sz, - __FUNCTION__ ); - cmd[0].i = 0; - cmd[0].header.cmd_type = RADEON_CMD_PACKET3; - cmd[1].i = RADEON_CP_PACKET3_3D_LOAD_VBPNTR | (((sz / sizeof(int))-3) << 16); - cmd[2].i = nr; - tmp = &cmd[0].i; - cmd += 3; - - for (i = 0 ; i < nr ; i++) { - if (i & 1) { - cmd[0].i |= ((component[i]->aos_stride << 24) | - (component[i]->aos_size << 16)); - cmd[2].i = (component[i]->aos_start + - offset * component[i]->aos_stride * 4); - cmd += 3; + BEGIN_BATCH(sz+2+(nr * 2)); + OUT_BATCH_PACKET3(RADEON_CP_PACKET3_3D_LOAD_VBPNTR, sz - 1); + OUT_BATCH(nr); + + if (!rmesa->radeon.radeonScreen->kernel_mm) { + for (i = 0; i + 1 < nr; i += 2) { + OUT_BATCH((rmesa->tcl.aos[i].components << 0) | + (rmesa->tcl.aos[i].stride << 8) | + (rmesa->tcl.aos[i + 1].components << 16) | + (rmesa->tcl.aos[i + 1].stride << 24)); + + voffset = rmesa->tcl.aos[i + 0].offset + + offset * 4 * rmesa->tcl.aos[i + 0].stride; + OUT_BATCH_RELOC(voffset, + rmesa->tcl.aos[i].bo, + voffset, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + voffset = rmesa->tcl.aos[i + 1].offset + + offset * 4 * rmesa->tcl.aos[i + 1].stride; + OUT_BATCH_RELOC(voffset, + rmesa->tcl.aos[i+1].bo, + voffset, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + } + + if (nr & 1) { + OUT_BATCH((rmesa->tcl.aos[nr - 1].components << 0) | + (rmesa->tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->tcl.aos[nr - 1].offset + + offset * 4 * rmesa->tcl.aos[nr - 1].stride; + OUT_BATCH_RELOC(voffset, + rmesa->tcl.aos[nr - 1].bo, + voffset, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + } + } else { + for (i = 0; i + 1 < nr; i += 2) { + OUT_BATCH((rmesa->tcl.aos[i].components << 0) | + (rmesa->tcl.aos[i].stride << 8) | + (rmesa->tcl.aos[i + 1].components << 16) | + (rmesa->tcl.aos[i + 1].stride << 24)); + + voffset = rmesa->tcl.aos[i + 0].offset + + offset * 4 * rmesa->tcl.aos[i + 0].stride; + OUT_BATCH(voffset); + voffset = rmesa->tcl.aos[i + 1].offset + + offset * 4 * rmesa->tcl.aos[i + 1].stride; + OUT_BATCH(voffset); + } + + if (nr & 1) { + OUT_BATCH((rmesa->tcl.aos[nr - 1].components << 0) | + (rmesa->tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->tcl.aos[nr - 1].offset + + offset * 4 * rmesa->tcl.aos[nr - 1].stride; + OUT_BATCH(voffset); } - else { - cmd[0].i = ((component[i]->aos_stride << 8) | - (component[i]->aos_size << 0)); - cmd[1].i = (component[i]->aos_start + - offset * component[i]->aos_stride * 4); + for (i = 0; i + 1 < nr; i += 2) { + voffset = rmesa->tcl.aos[i + 0].offset + + offset * 4 * rmesa->tcl.aos[i + 0].stride; + radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, + rmesa->tcl.aos[i+0].bo, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + voffset = rmesa->tcl.aos[i + 1].offset + + offset * 4 * rmesa->tcl.aos[i + 1].stride; + radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, + rmesa->tcl.aos[i+1].bo, + RADEON_GEM_DOMAIN_GTT, + 0, 0); + } + if (nr & 1) { + voffset = rmesa->tcl.aos[nr - 1].offset + + offset * 4 * rmesa->tcl.aos[nr - 1].stride; + radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, + rmesa->tcl.aos[nr-1].bo, + RADEON_GEM_DOMAIN_GTT, + 0, 0); } } + END_BATCH(); - if (RADEON_DEBUG & DEBUG_VERTS) { - fprintf(stderr, "%s:\n", __FUNCTION__); - for (i = 0 ; i < sz ; i++) - fprintf(stderr, " %d: %x\n", i, tmp[i]); - } #endif } diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h index 87d64fb728..b7f0729494 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h @@ -43,6 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. extern void radeonEmitState( r100ContextPtr rmesa ); extern void radeonEmitVertexAOS( r100ContextPtr rmesa, GLuint vertex_size, + struct radeon_bo *bo, GLuint offset ); extern void radeonEmitVbufPrim( r100ContextPtr rmesa, @@ -58,8 +59,8 @@ extern GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa, GLuint primitive, GLuint min_nr ); + extern void radeonEmitAOS( r100ContextPtr rmesa, - struct radeon_dma_region **regions, GLuint n, GLuint offset ); diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index fa3d0335fc..7f5da16b03 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -40,7 +40,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "swrast_setup/swrast_setup.h" #include "math/m_translate.h" #include "tnl/tnl.h" -#include "tnl/tcontext.h" #include "radeon_context.h" #include "radeon_ioctl.h" @@ -49,23 +48,18 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_maos.h" #include "radeon_tcl.h" -static void emit_vecfog( GLcontext *ctx, - struct radeon_dma_region *rvb, - char *data, - int stride, - int count ) +static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, + GLvoid *data, int stride, int count) { int i; - GLfloat *out; - + uint32_t *out; + int size = 1; radeonContextPtr rmesa = RADEON_CONTEXT(ctx); if (RADEON_DEBUG & DEBUG_VERTS) fprintf(stderr, "%s count %d stride %d\n", __FUNCTION__, count, stride); - assert (!rvb->buf); - if (stride == 0) { radeonAllocDmaRegion( rmesa, &aos->bo, &aos->offset, size * 4, 32 ); count = 1; @@ -125,15 +119,12 @@ static void emit_stq_vec(uint32_t *out, GLvoid *data, int stride, int count) -static void emit_tex_vector( GLcontext *ctx, - struct radeon_aos *aos, - GLvoid *data, - int size, - int stride, - int count ) +static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, + GLvoid *data, int size, int stride, int count) { radeonContextPtr rmesa = RADEON_CONTEXT(ctx); int emitsize; + uint32_t *out; if (RADEON_DEBUG & DEBUG_VERTS) fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size); @@ -160,6 +151,7 @@ static void emit_tex_vector( GLcontext *ctx, /* Emit the data */ + out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: emit_s0_vec( out, data, stride, count ); @@ -188,9 +180,8 @@ static void emit_tex_vector( GLcontext *ctx, */ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) { - radeonContextPtr rmesa = RADEON_CONTEXT( ctx ); + r100ContextPtr rmesa = R100_CONTEXT( ctx ); struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb; - struct radeon_dma_region **component = rmesa->tcl.aos_components; GLuint nr = 0; GLuint vfmt = 0; GLuint count = VB->Count; @@ -204,7 +195,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (1) { if (!rmesa->tcl.obj.buf) rcommon_emit_vector( ctx, - &rmesa->tcl.obj, + &(rmesa->tcl.aos[nr]), (char *)VB->ObjPtr->data, VB->ObjPtr->size, VB->ObjPtr->stride, @@ -217,21 +208,21 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) default: break; } - component[nr++] = &rmesa->tcl.obj; + nr++; } if (inputs & VERT_BIT_NORMAL) { if (!rmesa->tcl.norm.buf) rcommon_emit_vector( ctx, - &(rmesa->tcl.norm), + &(rmesa->tcl.aos[nr]), (char *)VB->NormalPtr->data, 3, VB->NormalPtr->stride, count); vfmt |= RADEON_CP_VC_FRMT_N0; - component[nr++] = &rmesa->tcl.norm; + nr++; } if (inputs & VERT_BIT_COLOR0) { @@ -250,14 +241,13 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.rgba.buf) rcommon_emit_vector( ctx, - &(rmesa->tcl.rgba), + &(rmesa->tcl.aos[nr]), (char *)VB->ColorPtr[0]->data, emitsize, VB->ColorPtr[0]->stride, count); - - component[nr++] = &rmesa->tcl.rgba; + nr++; } @@ -265,7 +255,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.spec.buf) { rcommon_emit_vector( ctx, - &rmesa->tcl.spec, + &(rmesa->tcl.aos[nr]), (char *)VB->SecondaryColorPtr[0]->data, 3, VB->SecondaryColorPtr[0]->stride, @@ -273,7 +263,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) } vfmt |= RADEON_CP_VC_FRMT_FPSPEC; - component[nr++] = &rmesa->tcl.spec; + nr++; } /* FIXME: not sure if this is correct. May need to stitch this together with @@ -282,13 +272,13 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (inputs & VERT_BIT_FOG) { if (!rmesa->tcl.fog.buf) emit_vecfog( ctx, - &(rmesa->tcl.fog), + &(rmesa->tcl.aos[nr]), (char *)VB->FogCoordPtr->data, VB->FogCoordPtr->stride, count); vfmt |= RADEON_CP_VC_FRMT_FPFOG; - component[nr++] = &rmesa->tcl.fog; + nr++; } @@ -299,11 +289,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (inputs & VERT_BIT_TEX(unit)) { if (!rmesa->tcl.tex[unit].buf) emit_tex_vector( ctx, - &(rmesa->tcl.tex[unit]), + &(rmesa->tcl.aos[nr]), (char *)VB->TexCoordPtr[unit]->data, VB->TexCoordPtr[unit]->size, VB->TexCoordPtr[unit]->stride, count ); + nr++; vfmt |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least @@ -321,7 +312,6 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; } - component[nr++] = &rmesa->tcl.tex[unit]; } } @@ -337,31 +327,13 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) { - radeonContextPtr rmesa = RADEON_CONTEXT( ctx ); - GLuint unit; - -#if 0 - if (RADEON_DEBUG & DEBUG_VERTS) - _tnl_print_vert_flags( __FUNCTION__, newinputs ); -#endif - - if (newinputs & VERT_BIT_POS) - radeonReleaseDmaRegion( rmesa, &rmesa->tcl.obj, __FUNCTION__ ); - - if (newinputs & VERT_BIT_NORMAL) - radeonReleaseDmaRegion( rmesa, &rmesa->tcl.norm, __FUNCTION__ ); - - if (newinputs & VERT_BIT_COLOR0) - radeonReleaseDmaRegion( rmesa, &rmesa->tcl.rgba, __FUNCTION__ ); - - if (newinputs & VERT_BIT_COLOR1) - radeonReleaseDmaRegion( rmesa, &rmesa->tcl.spec, __FUNCTION__ ); - - if (newinputs & VERT_BIT_FOG) - radeonReleaseDmaRegion( rmesa, &rmesa->tcl.fog, __FUNCTION__ ); + r100ContextPtr rmesa = R100_CONTEXT( ctx ); + int i; - for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) { - if (newinputs & VERT_BIT_TEX(unit)) - radeonReleaseDmaRegion( rmesa, &rmesa->tcl.tex[unit], __FUNCTION__ ); + for (i = 0; i < rmesa->tcl.nr_aos_components; i++) { + if (rmesa->tcl.aos[i].bo) { + radeon_bo_unref(rmesa->tcl.aos[i].bo); + rmesa->tcl.aos[i].bo = NULL; + } } } diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 5a0bbaacc9..bb90a7b111 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -374,14 +374,15 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) break; if (rmesa->tcl.vertex_format == setup_tab[i].vertex_format && - rmesa->tcl.indexed_verts.buf) + rmesa->tcl.aos[0].bo) return; - if (rmesa->tcl.indexed_verts.buf) + if (rmesa->tcl.aos[0].bo) radeonReleaseArrays( ctx, ~0 ); radeonAllocDmaRegion( &rmesa->radeon, - 0, &rmesa->tcl.indexed_verts, + &rmesa->tcl.aos[0].bo, + &rmesa->tcl.aos[0].offset, VB->Count * setup_tab[i].vertex_size * 4, 4); @@ -421,15 +422,11 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) setup_tab[i].emit( ctx, 0, VB->Count, - rmesa->tcl.indexed_verts.address + - rmesa->tcl.indexed_verts.start ); + rmesa->tcl.aos[0].bo->ptr + rmesa->tcl.aos[0].offset); + // rmesa->tcl.aos[0].size = setup_tab[i].vertex_size; + rmesa->tcl.aos[0].stride = setup_tab[i].vertex_size; rmesa->tcl.vertex_format = setup_tab[i].vertex_format; - // rmesa->tcl.indexed_verts.aos_start = GET_START( &rmesa->tcl.indexed_verts ); - rmesa->tcl.indexed_verts.aos_size = setup_tab[i].vertex_size; - rmesa->tcl.indexed_verts.aos_stride = setup_tab[i].vertex_size; - - rmesa->tcl.aos_components[0] = &rmesa->tcl.indexed_verts; rmesa->tcl.nr_aos_components = 1; } @@ -438,12 +435,10 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) { r100ContextPtr rmesa = R100_CONTEXT( ctx ); + int i; -#if 0 - if (RADEON_DEBUG & DEBUG_VERTS) - _tnl_print_vert_flags( __FUNCTION__, newinputs ); -#endif - - /// if (newinputs) - /// radeonReleaseDmaRegion( rmesa, &rmesa->tcl.indexed_verts, __FUNCTION__ ); + for (i = 0; i < rmesa->tcl.nr_aos_components; i++) { + radeon_bo_unref(rmesa->tcl.aos[i].bo); + rmesa->tcl.aos[i].bo = NULL; + } } diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index 02ea616a59..0a9e814191 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -38,14 +38,128 @@ #include "swrast_setup/swrast_setup.h" #include "radeon_context.h" +#include "common_cmdbuf.h" +#include "radeon_cs.h" +#include "radeon_mipmap_tree.h" #include "radeon_ioctl.h" #include "radeon_state.h" #include "radeon_tcl.h" #include "radeon_tex.h" #include "radeon_swtcl.h" +#include "../r200/r200_reg.h" + #include "xmlpool.h" +/* New (1.3) state mechanism. 3 commands (packet, scalar, vector) in + * 1.3 cmdbuffers allow all previous state to be updated as well as + * the tcl scalar and vector areas. + */ +static struct { + int start; + int len; + const char *name; +} packet[RADEON_MAX_STATE_PACKETS] = { + {RADEON_PP_MISC, 7, "RADEON_PP_MISC"}, + {RADEON_PP_CNTL, 3, "RADEON_PP_CNTL"}, + {RADEON_RB3D_COLORPITCH, 1, "RADEON_RB3D_COLORPITCH"}, + {RADEON_RE_LINE_PATTERN, 2, "RADEON_RE_LINE_PATTERN"}, + {RADEON_SE_LINE_WIDTH, 1, "RADEON_SE_LINE_WIDTH"}, + {RADEON_PP_LUM_MATRIX, 1, "RADEON_PP_LUM_MATRIX"}, + {RADEON_PP_ROT_MATRIX_0, 2, "RADEON_PP_ROT_MATRIX_0"}, + {RADEON_RB3D_STENCILREFMASK, 3, "RADEON_RB3D_STENCILREFMASK"}, + {RADEON_SE_VPORT_XSCALE, 6, "RADEON_SE_VPORT_XSCALE"}, + {RADEON_SE_CNTL, 2, "RADEON_SE_CNTL"}, + {RADEON_SE_CNTL_STATUS, 1, "RADEON_SE_CNTL_STATUS"}, + {RADEON_RE_MISC, 1, "RADEON_RE_MISC"}, + {RADEON_PP_TXFILTER_0, 6, "RADEON_PP_TXFILTER_0"}, + {RADEON_PP_BORDER_COLOR_0, 1, "RADEON_PP_BORDER_COLOR_0"}, + {RADEON_PP_TXFILTER_1, 6, "RADEON_PP_TXFILTER_1"}, + {RADEON_PP_BORDER_COLOR_1, 1, "RADEON_PP_BORDER_COLOR_1"}, + {RADEON_PP_TXFILTER_2, 6, "RADEON_PP_TXFILTER_2"}, + {RADEON_PP_BORDER_COLOR_2, 1, "RADEON_PP_BORDER_COLOR_2"}, + {RADEON_SE_ZBIAS_FACTOR, 2, "RADEON_SE_ZBIAS_FACTOR"}, + {RADEON_SE_TCL_OUTPUT_VTX_FMT, 11, "RADEON_SE_TCL_OUTPUT_VTX_FMT"}, + {RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED, 17, + "RADEON_SE_TCL_MATERIAL_EMMISSIVE_RED"}, + {R200_PP_TXCBLEND_0, 4, "R200_PP_TXCBLEND_0"}, + {R200_PP_TXCBLEND_1, 4, "R200_PP_TXCBLEND_1"}, + {R200_PP_TXCBLEND_2, 4, "R200_PP_TXCBLEND_2"}, + {R200_PP_TXCBLEND_3, 4, "R200_PP_TXCBLEND_3"}, + {R200_PP_TXCBLEND_4, 4, "R200_PP_TXCBLEND_4"}, + {R200_PP_TXCBLEND_5, 4, "R200_PP_TXCBLEND_5"}, + {R200_PP_TXCBLEND_6, 4, "R200_PP_TXCBLEND_6"}, + {R200_PP_TXCBLEND_7, 4, "R200_PP_TXCBLEND_7"}, + {R200_SE_TCL_LIGHT_MODEL_CTL_0, 6, "R200_SE_TCL_LIGHT_MODEL_CTL_0"}, + {R200_PP_TFACTOR_0, 6, "R200_PP_TFACTOR_0"}, + {R200_SE_VTX_FMT_0, 4, "R200_SE_VTX_FMT_0"}, + {R200_SE_VAP_CNTL, 1, "R200_SE_VAP_CNTL"}, + {R200_SE_TCL_MATRIX_SEL_0, 5, "R200_SE_TCL_MATRIX_SEL_0"}, + {R200_SE_TCL_TEX_PROC_CTL_2, 5, "R200_SE_TCL_TEX_PROC_CTL_2"}, + {R200_SE_TCL_UCP_VERT_BLEND_CTL, 1, "R200_SE_TCL_UCP_VERT_BLEND_CTL"}, + {R200_PP_TXFILTER_0, 6, "R200_PP_TXFILTER_0"}, + {R200_PP_TXFILTER_1, 6, "R200_PP_TXFILTER_1"}, + {R200_PP_TXFILTER_2, 6, "R200_PP_TXFILTER_2"}, + {R200_PP_TXFILTER_3, 6, "R200_PP_TXFILTER_3"}, + {R200_PP_TXFILTER_4, 6, "R200_PP_TXFILTER_4"}, + {R200_PP_TXFILTER_5, 6, "R200_PP_TXFILTER_5"}, + {R200_PP_TXOFFSET_0, 1, "R200_PP_TXOFFSET_0"}, + {R200_PP_TXOFFSET_1, 1, "R200_PP_TXOFFSET_1"}, + {R200_PP_TXOFFSET_2, 1, "R200_PP_TXOFFSET_2"}, + {R200_PP_TXOFFSET_3, 1, "R200_PP_TXOFFSET_3"}, + {R200_PP_TXOFFSET_4, 1, "R200_PP_TXOFFSET_4"}, + {R200_PP_TXOFFSET_5, 1, "R200_PP_TXOFFSET_5"}, + {R200_SE_VTE_CNTL, 1, "R200_SE_VTE_CNTL"}, + {R200_SE_TCL_OUTPUT_VTX_COMP_SEL, 1, + "R200_SE_TCL_OUTPUT_VTX_COMP_SEL"}, + {R200_PP_TAM_DEBUG3, 1, "R200_PP_TAM_DEBUG3"}, + {R200_PP_CNTL_X, 1, "R200_PP_CNTL_X"}, + {R200_RB3D_DEPTHXY_OFFSET, 1, "R200_RB3D_DEPTHXY_OFFSET"}, + {R200_RE_AUX_SCISSOR_CNTL, 1, "R200_RE_AUX_SCISSOR_CNTL"}, + {R200_RE_SCISSOR_TL_0, 2, "R200_RE_SCISSOR_TL_0"}, + {R200_RE_SCISSOR_TL_1, 2, "R200_RE_SCISSOR_TL_1"}, + {R200_RE_SCISSOR_TL_2, 2, "R200_RE_SCISSOR_TL_2"}, + {R200_SE_VAP_CNTL_STATUS, 1, "R200_SE_VAP_CNTL_STATUS"}, + {R200_SE_VTX_STATE_CNTL, 1, "R200_SE_VTX_STATE_CNTL"}, + {R200_RE_POINTSIZE, 1, "R200_RE_POINTSIZE"}, + {R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0, 4, + "R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0"}, + {R200_PP_CUBIC_FACES_0, 1, "R200_PP_CUBIC_FACES_0"}, /* 61 */ + {R200_PP_CUBIC_OFFSET_F1_0, 5, "R200_PP_CUBIC_OFFSET_F1_0"}, /* 62 */ + {R200_PP_CUBIC_FACES_1, 1, "R200_PP_CUBIC_FACES_1"}, + {R200_PP_CUBIC_OFFSET_F1_1, 5, "R200_PP_CUBIC_OFFSET_F1_1"}, + {R200_PP_CUBIC_FACES_2, 1, "R200_PP_CUBIC_FACES_2"}, + {R200_PP_CUBIC_OFFSET_F1_2, 5, "R200_PP_CUBIC_OFFSET_F1_2"}, + {R200_PP_CUBIC_FACES_3, 1, "R200_PP_CUBIC_FACES_3"}, + {R200_PP_CUBIC_OFFSET_F1_3, 5, "R200_PP_CUBIC_OFFSET_F1_3"}, + {R200_PP_CUBIC_FACES_4, 1, "R200_PP_CUBIC_FACES_4"}, + {R200_PP_CUBIC_OFFSET_F1_4, 5, "R200_PP_CUBIC_OFFSET_F1_4"}, + {R200_PP_CUBIC_FACES_5, 1, "R200_PP_CUBIC_FACES_5"}, + {R200_PP_CUBIC_OFFSET_F1_5, 5, "R200_PP_CUBIC_OFFSET_F1_5"}, + {RADEON_PP_TEX_SIZE_0, 2, "RADEON_PP_TEX_SIZE_0"}, + {RADEON_PP_TEX_SIZE_1, 2, "RADEON_PP_TEX_SIZE_1"}, + {RADEON_PP_TEX_SIZE_2, 2, "RADEON_PP_TEX_SIZE_2"}, + {R200_RB3D_BLENDCOLOR, 3, "R200_RB3D_BLENDCOLOR"}, + {R200_SE_TCL_POINT_SPRITE_CNTL, 1, "R200_SE_TCL_POINT_SPRITE_CNTL"}, + {RADEON_PP_CUBIC_FACES_0, 1, "RADEON_PP_CUBIC_FACES_0"}, + {RADEON_PP_CUBIC_OFFSET_T0_0, 5, "RADEON_PP_CUBIC_OFFSET_T0_0"}, + {RADEON_PP_CUBIC_FACES_1, 1, "RADEON_PP_CUBIC_FACES_1"}, + {RADEON_PP_CUBIC_OFFSET_T1_0, 5, "RADEON_PP_CUBIC_OFFSET_T1_0"}, + {RADEON_PP_CUBIC_FACES_2, 1, "RADEON_PP_CUBIC_FACES_2"}, + {RADEON_PP_CUBIC_OFFSET_T2_0, 5, "RADEON_PP_CUBIC_OFFSET_T2_0"}, + {R200_PP_TRI_PERF, 2, "R200_PP_TRI_PERF"}, + {R200_PP_TXCBLEND_8, 32, "R200_PP_AFS_0"}, /* 85 */ + {R200_PP_TXCBLEND_0, 32, "R200_PP_AFS_1"}, + {R200_PP_TFACTOR_0, 8, "R200_ATF_TFACTOR"}, + {R200_PP_TXFILTER_0, 8, "R200_PP_TXCTLALL_0"}, + {R200_PP_TXFILTER_1, 8, "R200_PP_TXCTLALL_1"}, + {R200_PP_TXFILTER_2, 8, "R200_PP_TXCTLALL_2"}, + {R200_PP_TXFILTER_3, 8, "R200_PP_TXCTLALL_3"}, + {R200_PP_TXFILTER_4, 8, "R200_PP_TXCTLALL_4"}, + {R200_PP_TXFILTER_5, 8, "R200_PP_TXCTLALL_5"}, + {R200_VAP_PVS_CNTL_1, 2, "R200_VAP_PVS_CNTL"}, +}; + /* ============================================================= * State initialization */ @@ -65,12 +179,17 @@ void radeonPrintDirty( r100ContextPtr rmesa, const char *msg ) fprintf(stderr, "\n"); } -static int cmdpkt( int id ) +static int cmdpkt( r100ContextPtr rmesa, int id ) { drm_radeon_cmd_header_t h; - h.i = 0; - h.packet.cmd_type = RADEON_CMD_PACKET; - h.packet.packet_id = id; + + if (rmesa->radeon.radeonScreen->kernel_mm) { + return CP_PACKET0(packet[id].start, packet[id].len - 1); + } else { + h.i = 0; + h.packet.cmd_type = RADEON_CMD_PACKET; + h.packet.packet_id = id; + } return h.i; } @@ -146,7 +265,200 @@ CHECK( txr0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_RECT_BIT)) CHECK( txr1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_RECT_BIT)) CHECK( txr2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_RECT_BIT)) +#define OUT_VEC(hdr, data) do { \ + drm_radeon_cmd_header_t h; \ + h.i = hdr; \ + OUT_BATCH(CP_PACKET0(RADEON_SE_TCL_STATE_FLUSH, 0)); \ + OUT_BATCH(0); \ + OUT_BATCH(CP_PACKET0(R200_SE_TCL_VECTOR_INDX_REG, 0)); \ + OUT_BATCH(h.vectors.offset | (h.vectors.stride << RADEON_VEC_INDX_OCTWORD_STRIDE_SHIFT)); \ + OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_VECTOR_DATA_REG, h.vectors.count - 1)); \ + OUT_BATCH_TABLE((data), h.vectors.count); \ + } while(0) + +#define OUT_SCL(hdr, data) do { \ + drm_radeon_cmd_header_t h; \ + h.i = hdr; \ + OUT_BATCH(CP_PACKET0(R200_SE_TCL_SCALAR_INDX_REG, 0)); \ + OUT_BATCH((h.scalars.offset) | (h.scalars.stride << RADEON_SCAL_INDX_DWORD_STRIDE_SHIFT)); \ + OUT_BATCH(CP_PACKET0_ONE(R200_SE_TCL_SCALAR_DATA_REG, h.scalars.count - 1)); \ + OUT_BATCH_TABLE((data), h.scalars.count); \ + } while(0) + +static void scl_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r100ContextPtr r100 = R100_CONTEXT(ctx); + BATCH_LOCALS(&r100->radeon); + uint32_t dwords = atom->cmd_size; + + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_SCL(atom->cmd[0], atom->cmd+1); + END_BATCH(); +} + + +static void vec_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r100ContextPtr r100 = R100_CONTEXT(ctx); + BATCH_LOCALS(&r100->radeon); + uint32_t dwords = atom->cmd_size; + + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_VEC(atom->cmd[0], atom->cmd+1); + END_BATCH(); +} + +static void ctx_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r100ContextPtr r100 = R100_CONTEXT(ctx); + BATCH_LOCALS(&r100->radeon); + struct radeon_renderbuffer *rrb; + uint32_t cbpitch; + uint32_t zbpitch; + uint32_t dwords = atom->cmd_size; + GLframebuffer *fb = r100->radeon.dri.drawable->driverPrivate; + + /* output the first 7 bytes of context */ + BEGIN_BATCH_NO_AUTOSTATE(dwords + 4); + OUT_BATCH_TABLE(atom->cmd, 5); + + rrb = r100->radeon.state.depth.rrb; + if (!rrb) { + OUT_BATCH(0); + OUT_BATCH(0); + } else { + zbpitch = (rrb->pitch / rrb->cpp); + OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_BATCH(zbpitch); + } + + OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]); + OUT_BATCH(atom->cmd[CTX_CMD_1]); + OUT_BATCH(atom->cmd[CTX_PP_CNTL]); + OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]); + + rrb = r100->radeon.state.color.rrb; + if (r100->radeon.radeonScreen->driScreen->dri2.enabled) { + rrb = (struct radeon_renderbuffer *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + } + if (!rrb || !rrb->bo) { + OUT_BATCH(atom->cmd[CTX_RB3D_COLOROFFSET]); + } else { + OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + } + + OUT_BATCH(atom->cmd[CTX_CMD_2]); + + if (!rrb || !rrb->bo) { + OUT_BATCH(atom->cmd[CTX_RB3D_COLORPITCH]); + } else { + cbpitch = (rrb->pitch / rrb->cpp); + if (rrb->cpp == 4) + ; + else + ; + if (r100->radeon.sarea->tiling_enabled) + cbpitch |= R200_COLOR_TILE_ENABLE; + OUT_BATCH(cbpitch); + } + + END_BATCH(); +} +static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r100ContextPtr r100 = R100_CONTEXT(ctx); + BATCH_LOCALS(&r100->radeon); + struct radeon_renderbuffer *rrb, *drb; + uint32_t cbpitch = 0; + uint32_t zbpitch = 0; + uint32_t dwords = atom->cmd_size; + GLframebuffer *fb = r100->radeon.dri.drawable->driverPrivate; + + rrb = r100->radeon.state.color.rrb; + if (r100->radeon.radeonScreen->driScreen->dri2.enabled) { + rrb = (struct radeon_renderbuffer *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + } + if (rrb) { + assert(rrb->bo != NULL); + cbpitch = (rrb->pitch / rrb->cpp); + if (r100->radeon.sarea->tiling_enabled) + cbpitch |= R200_COLOR_TILE_ENABLE; + } + + drb = r100->radeon.state.depth.rrb; + if (drb) + zbpitch = (drb->pitch / drb->cpp); + + /* output the first 7 bytes of context */ + BEGIN_BATCH_NO_AUTOSTATE(dwords); + + /* In the CS case we need to split this up */ + OUT_BATCH(CP_PACKET0(packet[0].start, 3)); + OUT_BATCH_TABLE((atom->cmd + 1), 4); + if (drb) { + OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHOFFSET, 0)); + OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + + OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHPITCH, 0)); + OUT_BATCH(zbpitch); + } + + OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZSTENCILCNTL, 0)); + OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]); + OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 1)); + OUT_BATCH(atom->cmd[CTX_PP_CNTL]); + OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]); + + + if (rrb) { + OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLOROFFSET, 0)); + OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + } + + if (rrb) { + if (rrb->cpp == 4) + ; + else + ; + OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLORPITCH, 0)); + OUT_BATCH(cbpitch); + } + + // if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM) { + // OUT_BATCH_TABLE((atom->cmd + 14), 4); + // } + + END_BATCH(); +} + + + +static void tex_emit(GLcontext *ctx, struct radeon_state_atom *atom) +{ + r100ContextPtr r100 = R100_CONTEXT(ctx); + BATCH_LOCALS(&r100->radeon); + uint32_t dwords = atom->cmd_size; + int i = atom->idx; + radeonTexObj *t = r100->state.texture.unit[i].texobj; + + if (!t) + return; + + BEGIN_BATCH_NO_AUTOSTATE(dwords + 2); + OUT_BATCH_TABLE(atom->cmd, 3); + if (t && !t->image_override) { + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + RADEON_GEM_DOMAIN_VRAM, 0, 0); + } else if (!t) { + + + OUT_BATCH(atom->cmd[10]); + } + + OUT_BATCH_TABLE((atom->cmd+4), 5); + END_BATCH(); +} /* Initialize the context's hardware state. */ @@ -221,6 +533,10 @@ void radeonInitState( r100ContextPtr rmesa ) /* Allocate state buffers: */ ALLOC_STATE( ctx, always, CTX_STATE_SIZE, "CTX/context", 0 ); + if (rmesa->radeon.radeonScreen->kernel_mm) + rmesa->hw.ctx.emit = ctx_emit_cs; + else + rmesa->hw.ctx.emit = ctx_emit; ALLOC_STATE( lin, always, LIN_STATE_SIZE, "LIN/line", 0 ); ALLOC_STATE( msk, always, MSK_STATE_SIZE, "MSK/mask", 0 ); ALLOC_STATE( vpt, always, VPT_STATE_SIZE, "VPT/viewport", 0 ); @@ -236,6 +552,9 @@ void radeonInitState( r100ContextPtr rmesa ) ALLOC_STATE( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0 ); ALLOC_STATE( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0 ); ALLOC_STATE( tex[2], tex2, TEX_STATE_SIZE, "TEX/tex-2", 0 ); + + for (i = 0; i < 3; i++) + rmesa->hw.tex[i].emit = tex_emit; if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR100) { ALLOC_STATE( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0 ); @@ -276,35 +595,35 @@ void radeonInitState( r100ContextPtr rmesa ) /* Fill in the packet headers: */ - rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(RADEON_EMIT_PP_MISC); - rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(RADEON_EMIT_PP_CNTL); - rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(RADEON_EMIT_RB3D_COLORPITCH); - rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(RADEON_EMIT_RE_LINE_PATTERN); - rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(RADEON_EMIT_SE_LINE_WIDTH); - rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(RADEON_EMIT_RB3D_STENCILREFMASK); - rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(RADEON_EMIT_SE_VPORT_XSCALE); - rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(RADEON_EMIT_SE_CNTL); - rmesa->hw.set.cmd[SET_CMD_1] = cmdpkt(RADEON_EMIT_SE_CNTL_STATUS); - rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(RADEON_EMIT_RE_MISC); - rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_0); - rmesa->hw.tex[0].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_0); - rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_1); - rmesa->hw.tex[1].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_1); - rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(RADEON_EMIT_PP_TXFILTER_2); - rmesa->hw.tex[2].cmd[TEX_CMD_1] = cmdpkt(RADEON_EMIT_PP_BORDER_COLOR_2); - rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_0); - rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T0); - rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_1); - rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T1); - rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(RADEON_EMIT_PP_CUBIC_FACES_2); - rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(RADEON_EMIT_PP_CUBIC_OFFSETS_T2); - rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(RADEON_EMIT_SE_ZBIAS_FACTOR); - rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT); + rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_MISC); + rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CNTL); + rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(rmesa, RADEON_EMIT_RB3D_COLORPITCH); + rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_LINE_PATTERN); + rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_LINE_WIDTH); + rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RB3D_STENCILREFMASK); + rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_VPORT_XSCALE); + rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL); + rmesa->hw.set.cmd[SET_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_SE_CNTL_STATUS); + rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_RE_MISC); + rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_0); + rmesa->hw.tex[0].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_0); + rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_1); + rmesa->hw.tex[1].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_1); + rmesa->hw.tex[2].cmd[TEX_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TXFILTER_2); + rmesa->hw.tex[2].cmd[TEX_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_BORDER_COLOR_2); + rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_0); + rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T0); + rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_1); + rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T1); + rmesa->hw.cube[2].cmd[CUBE_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_FACES_2); + rmesa->hw.cube[2].cmd[CUBE_CMD_1] = cmdpkt(rmesa, RADEON_EMIT_PP_CUBIC_OFFSETS_T2); + rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_ZBIAS_FACTOR); + rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_SE_TCL_OUTPUT_VTX_FMT); rmesa->hw.mtl.cmd[MTL_CMD_0] = - cmdpkt(RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED); - rmesa->hw.txr[0].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_0); - rmesa->hw.txr[1].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_1); - rmesa->hw.txr[2].cmd[TXR_CMD_0] = cmdpkt(RADEON_EMIT_PP_TEX_SIZE_2); + cmdpkt(rmesa, RADEON_EMIT_SE_TCL_MATERIAL_EMMISSIVE_RED); + rmesa->hw.txr[0].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_0); + rmesa->hw.txr[1].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_1); + rmesa->hw.txr[2].cmd[TXR_CMD_0] = cmdpkt(rmesa, RADEON_EMIT_PP_TEX_SIZE_2); rmesa->hw.grd.cmd[GRD_CMD_0] = cmdscl( RADEON_SS_VERT_GUARD_CLIP_ADJ_ADDR, 1, 4 ); rmesa->hw.fog.cmd[FOG_CMD_0] = diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 4b49c141a0..615621dd98 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -46,6 +46,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_pipeline.h" #include "radeon_context.h" +#include "common_cmdbuf.h" #include "radeon_ioctl.h" #include "radeon_state.h" #include "radeon_swtcl.h" @@ -291,7 +292,7 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(rmesa); radeonEmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - // rmesa->radeon.dma.current, + rmesa->radeon.dma.current, current_offset); diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index a54e471065..4f3acc7daf 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -43,6 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_pipeline.h" #include "radeon_context.h" +#include "common_cmdbuf.h" #include "radeon_state.h" #include "radeon_ioctl.h" #include "radeon_tex.h" @@ -145,19 +146,31 @@ static GLboolean discrete_prim[0x10] = { static GLushort *radeonAllocElts( r100ContextPtr rmesa, GLuint nr ) { - if (rmesa->radeon.dma.flush) - rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); +#if 0 + if (rmesa->radeon.dma.flush == radeonFlushElts && + rmesa->tcl.elt_used + nr*2 < R200_ELT_BUF_SZ) { - rcommonEnsureCmdBufSpace(&rmesa->radeon, AOS_BUFSZ(rmesa->tcl.nr_aos_components) + - rmesa->hw.max_state_size + ELTS_BUFSZ(nr)); + GLushort *dest = (GLushort *)(rmesa->tcl.elt_dma_bo->ptr + + rmesa->tcl.elt_used); - radeonEmitAOS( rmesa, - rmesa->tcl.aos_components, - rmesa->tcl.nr_aos_components, 0 ); + rmesa->tcl.elt_used += nr*2; + + return dest; + } + else { +#endif + if (rmesa->radeon.dma.flush) + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); + + rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->hw.max_state_size + ELTS_BUFSZ(nr) + + AOS_BUFSZ(rmesa->tcl.nr_aos_components), __FUNCTION__); + + radeonEmitAOS( rmesa, + rmesa->tcl.nr_aos_components, 0 ); - return radeonAllocEltsOpenEnded( rmesa, - rmesa->tcl.vertex_format, - rmesa->tcl.hw_primitive, nr ); + return radeonAllocEltsOpenEnded( rmesa, rmesa->tcl.vertex_format, + rmesa->tcl.hw_primitive, nr ); + // } } #define CLOSE_ELTS() RADEON_NEWPRIM( rmesa ) @@ -179,17 +192,16 @@ static void radeonEmitPrim( GLcontext *ctx, rcommonEnsureCmdBufSpace( &rmesa->radeon, AOS_BUFSZ(rmesa->tcl.nr_aos_components) + - rmesa->hw.max_state_size + VBUF_BUFSZ ); + rmesa->hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); radeonEmitAOS( rmesa, - rmesa->tcl.aos_components, rmesa->tcl.nr_aos_components, start ); /* Why couldn't this packet have taken an offset param? */ radeonEmitVbufPrim( rmesa, - 0, + rmesa->tcl.vertex_format, rmesa->tcl.hw_primitive, count - start ); } diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c index 9ef08fbe68..8784dcc5ed 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex.c @@ -315,7 +315,7 @@ static void radeonTexParameter( GLcontext *ctx, GLenum target, struct gl_texture_object *texObj, GLenum pname, const GLfloat *params ) { - radeonTexObjPtr t = (radeonTexObjPtr) texObj->DriverData; + radeonTexObj* t = radeon_tex_obj(texObj); if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) { fprintf( stderr, "%s( %s )\n", __FUNCTION__, diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index 3c28d70e37..e76b52437f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -903,7 +903,7 @@ void radeonSetTexOffset(__DRIcontext * pDRICtx, GLint texname, RADEON_TXFORMAT_CUBIC_MAP_ENABLE | \ RADEON_TXFORMAT_NON_POWER2) -#if 0 + static void import_tex_obj_state( r100ContextPtr rmesa, int unit, radeonTexObjPtr texobj ) @@ -921,7 +921,7 @@ static void import_tex_obj_state( r100ContextPtr rmesa, cmd[TEX_PP_TXOFFSET] = texobj->pp_txoffset; cmd[TEX_PP_BORDER_COLOR] = texobj->pp_border_color; - if (texobj->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) { + if (texobj->base.Target == GL_TEXTURE_RECTANGLE_NV) { GLuint *txr_cmd = RADEON_DB_STATE( txr[unit] ); txr_cmd[TXR_PP_TEX_SIZE] = texobj->pp_txsize; /* NPOT only! */ txr_cmd[TXR_PP_TEX_PITCH] = texobj->pp_txpitch; /* NPOT only! */ @@ -931,10 +931,11 @@ static void import_tex_obj_state( r100ContextPtr rmesa, else { se_coord_fmt &= ~(RADEON_VTX_ST0_NONPARAMETRIC << unit); - if (texobj->base.tObj->Target == GL_TEXTURE_CUBE_MAP) { - int *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0]; - GLuint bytesPerFace = texobj->base.totalSize / 6; - ASSERT(texobj->base.totalSize % 6 == 0); + if (texobj->base.Target == GL_TEXTURE_CUBE_MAP) { + uint32_t *cube_cmd = &rmesa->hw.cube[unit].cmd[CUBE_CMD_0]; + // GLuint bytesPerFace = texobj->base.totalSize / 6; + // ASSERT(texobj->base.totalSize % 6 == 0); + GLuint bytesPerFace = 1; // TODO RADEON_STATECHANGE( rmesa, cube[unit] ); cube_cmd[CUBE_PP_CUBIC_FACES] = texobj->pp_cubic_faces; @@ -957,8 +958,6 @@ static void import_tex_obj_state( r100ContextPtr rmesa, texobj->dirty_state &= ~(1<dirty_state & (1<validated = GL_TRUE; + return GL_TRUE; +} + static GLboolean radeonUpdateTextureUnit( GLcontext *ctx, int unit ) { + r100ContextPtr rmesa = R100_CONTEXT(ctx); struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; + if (!ctx->Texture.Unit[unit]._ReallyEnabled) + return GL_TRUE; + + if (!radeon_validate_texture(ctx, ctx->Texture.Unit[unit]._Current, unit)) { + _mesa_warning(ctx, + "failed to validate texture for unit %d.\n", + unit); + rmesa->state.texture.unit[unit].texobj = NULL; + return GL_FALSE; + } + rmesa->state.texture.unit[unit].texobj = radeon_tex_obj(ctx->Texture.Unit[unit]._Current); + return GL_TRUE; } void radeonUpdateTextureState( GLcontext *ctx ) -- cgit v1.2.3 From 1090d206de011a67d236d8c4ae32d2d42b2f6337 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Thu, 12 Feb 2009 21:16:39 +1000 Subject: radeon/r200/r300: another big merge upheavel. This merges lots of the hw state atom emission and firevertices code. it also removes a lot of the extra radeon crap from r300 and merge scissor --- src/mesa/drivers/dri/r200/r200_cmdbuf.c | 195 +++++------------------ src/mesa/drivers/dri/r200/r200_context.c | 32 +--- src/mesa/drivers/dri/r200/r200_context.h | 14 +- src/mesa/drivers/dri/r200/r200_ioctl.c | 61 +------ src/mesa/drivers/dri/r200/r200_ioctl.h | 14 +- src/mesa/drivers/dri/r200/r200_state.c | 29 +--- src/mesa/drivers/dri/r200/r200_state_init.c | 15 +- src/mesa/drivers/dri/r200/r200_swtcl.c | 10 +- src/mesa/drivers/dri/r200/r200_tcl.c | 6 +- src/mesa/drivers/dri/r200/r200_tex.c | 2 +- src/mesa/drivers/dri/r300/Makefile | 4 - src/mesa/drivers/dri/r300/r300_cmdbuf.c | 119 ++------------ src/mesa/drivers/dri/r300/r300_context.c | 32 ++-- src/mesa/drivers/dri/r300/r300_context.h | 9 +- src/mesa/drivers/dri/r300/r300_emit.c | 1 - src/mesa/drivers/dri/r300/r300_emit.h | 32 ++-- src/mesa/drivers/dri/r300/r300_ioctl.c | 7 +- src/mesa/drivers/dri/r300/r300_render.c | 2 - src/mesa/drivers/dri/r300/r300_state.c | 16 +- src/mesa/drivers/dri/r300/r300_state.h | 15 +- src/mesa/drivers/dri/r300/r300_swtcl.c | 2 +- src/mesa/drivers/dri/r300/r300_tex.c | 2 +- src/mesa/drivers/dri/r300/r300_texmem.c | 58 ------- src/mesa/drivers/dri/r300/r300_texstate.c | 1 - src/mesa/drivers/dri/r300/radeon_context.c | 75 --------- src/mesa/drivers/dri/r300/radeon_context.h | 43 ----- src/mesa/drivers/dri/r300/radeon_ioctl.c | 69 -------- src/mesa/drivers/dri/r300/radeon_ioctl.h | 57 ------- src/mesa/drivers/dri/r300/radeon_state.c | 109 ------------- src/mesa/drivers/dri/r300/radeon_state.h | 39 ----- src/mesa/drivers/dri/radeon/common_cmdbuf.h | 14 +- src/mesa/drivers/dri/radeon/common_context.h | 27 ++-- src/mesa/drivers/dri/radeon/common_misc.c | 155 ++++++++++++++++-- src/mesa/drivers/dri/radeon/common_misc.h | 5 + src/mesa/drivers/dri/radeon/radeon_context.c | 46 ++---- src/mesa/drivers/dri/radeon/radeon_context.h | 24 +-- src/mesa/drivers/dri/radeon/radeon_ioctl.c | 202 +++--------------------- src/mesa/drivers/dri/radeon/radeon_ioctl.h | 14 +- src/mesa/drivers/dri/radeon/radeon_screen.c | 2 +- src/mesa/drivers/dri/radeon/radeon_state.c | 26 +-- src/mesa/drivers/dri/radeon/radeon_state_init.c | 12 +- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 12 +- src/mesa/drivers/dri/radeon/radeon_tcl.c | 8 +- src/mesa/drivers/dri/radeon/radeon_tex.c | 2 +- 44 files changed, 353 insertions(+), 1266 deletions(-) delete mode 100644 src/mesa/drivers/dri/r300/r300_texmem.c delete mode 100644 src/mesa/drivers/dri/r300/radeon_context.c delete mode 100644 src/mesa/drivers/dri/r300/radeon_ioctl.c delete mode 100644 src/mesa/drivers/dri/r300/radeon_ioctl.h delete mode 100644 src/mesa/drivers/dri/r300/radeon_state.c delete mode 100644 src/mesa/drivers/dri/r300/radeon_state.h (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c index eba5d6e019..aa64232a93 100644 --- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c +++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c @@ -47,8 +47,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_sanity.h" #include "radeon_reg.h" -#define DEBUG_CMDBUF 0 - /* The state atoms will be emitted in the order they appear in the atom list, * so this step is important. */ @@ -58,161 +56,54 @@ void r200SetUpAtomList( r200ContextPtr rmesa ) mtu = rmesa->radeon.glCtx->Const.MaxTextureUnits; - make_empty_list(&rmesa->hw.atomlist); - rmesa->hw.atomlist.name = "atom-list"; - - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ctx ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.set ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.lin ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msk ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpt ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vtx ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vap ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vte ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msc ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.cst ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.zbs ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tcl ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.msl ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tcg ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.grd ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.fog ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tam ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tf ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.atf ); + make_empty_list(&rmesa->radeon.hw.atomlist); + rmesa->radeon.hw.atomlist.name = "atom-list"; + + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ctx ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.set ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.lin ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msk ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpt ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vtx ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vap ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vte ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msc ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.cst ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.zbs ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tcl ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.msl ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tcg ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.grd ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.fog ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tam ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tf ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.atf ); for (i = 0; i < mtu; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.tex[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.tex[i] ); for (i = 0; i < mtu; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.cube[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.cube[i] ); for (i = 0; i < 6; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.pix[i] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.afs[0] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.afs[1] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.pix[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.afs[0] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.afs[1] ); for (i = 0; i < 8; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.lit[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.lit[i] ); for (i = 0; i < 3 + mtu; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.mat[i] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.eye ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.glt ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.mat[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.eye ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.glt ); for (i = 0; i < 2; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.mtl[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.mtl[i] ); for (i = 0; i < 6; ++i) - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ucp[i] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.spr ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.ptp ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.prf ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.pvs ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpp[0] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpp[1] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpi[0] ); - insert_at_tail( &rmesa->hw.atomlist, &rmesa->hw.vpi[1] ); -} - -static void r200SaveHwState( r200ContextPtr rmesa ) -{ - struct radeon_state_atom *atom; - char * dest = rmesa->backup_store.cmd_buf; - uint32_t dwords; - if (R200_DEBUG & DEBUG_STATE) - fprintf(stderr, "%s\n", __FUNCTION__); - - rmesa->backup_store.cmd_used = 0; - - foreach( atom, &rmesa->hw.atomlist ) { - dwords = atom->check( rmesa->radeon.glCtx, atom ); - if ( dwords ) { - int size = atom->cmd_size * 4; - - if (atom->emit) { - (*atom->emit)(rmesa->radeon.glCtx, atom); - } else { - memcpy( dest, atom->cmd, size); - dest += size; - rmesa->backup_store.cmd_used += size; - } - if (R200_DEBUG & DEBUG_STATE) - radeon_print_state_atom( atom ); - } - } - - assert( rmesa->backup_store.cmd_used <= R200_CMD_BUF_SZ ); - if (R200_DEBUG & DEBUG_STATE) - fprintf(stderr, "Returning to r200EmitState\n"); -} - -static INLINE void r200EmitAtoms(r200ContextPtr r200, GLboolean dirty) -{ - BATCH_LOCALS(&r200->radeon); - struct radeon_state_atom *atom; - int dwords; - - /* Emit actual atoms */ - foreach(atom, &r200->hw.atomlist) { - if ((atom->dirty || r200->hw.all_dirty) == dirty) { - dwords = (*atom->check) (r200->radeon.glCtx, atom); - if (dwords) { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - radeon_print_state_atom(atom); - } - if (atom->emit) { - (*atom->emit)(r200->radeon.glCtx, atom); - } else { - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(atom->cmd, dwords); - END_BATCH(); - } - atom->dirty = GL_FALSE; - } else { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - fprintf(stderr, " skip state %s\n", - atom->name); - } - } - } - } - - COMMIT_BATCH(); -} - -void r200EmitState( r200ContextPtr rmesa ) -{ - char *dest; - int mtu; - struct radeon_state_atom *atom; - uint32_t dwords; - - if (R200_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (rmesa->save_on_next_emit) { - r200SaveHwState(rmesa); - rmesa->save_on_next_emit = GL_FALSE; - } - - if (rmesa->radeon.cmdbuf.cs->cdw && !rmesa->hw.is_dirty && !rmesa->hw.all_dirty) - return; - - mtu = rmesa->radeon.glCtx->Const.MaxTextureUnits; - - /* To avoid going across the entire set of states multiple times, just check - * for enough space for the case of emitting all state, and inline the - * r200AllocCmdBuf code here without all the checks. - */ - rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->hw.max_state_size, __FUNCTION__); - - if (!rmesa->radeon.cmdbuf.cs->cdw) { - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin reemit state\n"); - - r200EmitAtoms(rmesa, GL_FALSE); - } - - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin dirty state\n"); - - r200EmitAtoms(rmesa, GL_TRUE); - rmesa->hw.is_dirty = GL_FALSE; - rmesa->hw.all_dirty = GL_FALSE; + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ucp[i] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.spr ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.ptp ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.prf ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.pvs ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpp[0] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpp[1] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[0] ); + insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[1] ); } /* Fire a section of the retained (indexed_verts) buffer as a regular @@ -227,7 +118,7 @@ void r200EmitVbufPrim( r200ContextPtr rmesa, assert(!(primitive & R200_VF_PRIM_WALK_IND)); - r200EmitState( rmesa ); + radeonEmitState(&rmesa->radeon); if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS)) fprintf(stderr, "%s cmd_used/4: %d prim %x nr %d\n", __FUNCTION__, @@ -313,7 +204,7 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa, assert((primitive & R200_VF_PRIM_WALK_IND)); - r200EmitState( rmesa ); + radeonEmitState(&rmesa->radeon); rmesa->tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom, 0, R200_ELT_BUF_SZ, 4, diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c index 00274fbd43..1f5a9f2427 100644 --- a/src/mesa/drivers/dri/r200/r200_context.c +++ b/src/mesa/drivers/dri/r200/r200_context.c @@ -61,6 +61,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_maos.h" #include "r200_vertprog.h" +#include "radeon_span.h" + #define need_GL_ARB_multisample #define need_GL_ARB_texture_compression #define need_GL_ARB_vertex_buffer_object @@ -259,43 +261,17 @@ static void r200_get_lock(radeonContextPtr radeon) } } - -static void r200_vtbl_flush(GLcontext *ctx) -{ - R200_FIREVERTICES(R200_CONTEXT(ctx)); -} - -static void r200_vtbl_flush_vertices(radeonContextPtr rmesa) -{ - R200_FIREVERTICES(((r200ContextPtr)rmesa)); -} - -static void r200_vtbl_set_all_dirty(GLcontext *ctx) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - rmesa->hw.all_dirty = GL_TRUE; -} - static void r200_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa) { } -static void r200_vtbl_emit_state(radeonContextPtr rmesa) -{ - r200EmitState((r200ContextPtr)rmesa); -} - static void r200_init_vtbl(radeonContextPtr radeon) { radeon->vtbl.get_lock = r200_get_lock; radeon->vtbl.update_viewport_offset = r200UpdateViewportOffset; - radeon->vtbl.flush = r200_vtbl_flush; - radeon->vtbl.flush_vertices = r200_vtbl_flush_vertices; - radeon->vtbl.set_all_dirty = r200_vtbl_set_all_dirty; radeon->vtbl.update_draw_buffer = r200UpdateDrawBuffer; radeon->vtbl.emit_cs_header = r200_vtbl_emit_cs_header; - radeon->vtbl.emit_state = r200_vtbl_emit_state; radeon->vtbl.swtcl_flush = r200_swtcl_flush; } @@ -390,7 +366,7 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16; rmesa->radeon.swtcl.RenderIndex = ~0; - rmesa->hw.all_dirty = 1; + rmesa->radeon.hw.all_dirty = 1; /* Set the maximum texture size small enough that we can guarentee that * all texture units can bind a maximal texture and have all of them in @@ -565,7 +541,7 @@ void r200DestroyContext( __DRIcontextPrivate *driContextPriv ) /* check if we're deleting the currently bound context */ if (rmesa == current) { - R200_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); _mesa_make_current(NULL, NULL, NULL); } diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h index 862751bd40..96f47ea92b 100644 --- a/src/mesa/drivers/dri/r200/r200_context.h +++ b/src/mesa/drivers/dri/r200/r200_context.h @@ -469,9 +469,6 @@ struct r200_texture_state { struct r200_hw_state { - /* Head of the linked list of state atoms. */ - struct radeon_state_atom atomlist; - /* Hardware state, stored as cmdbuf commands: * -- Need to doublebuffer for * - reviving state after loss of context @@ -512,9 +509,6 @@ struct r200_hw_state { struct radeon_state_atom atf; struct radeon_state_atom spr; struct radeon_state_atom ptp; - - int max_state_size; /* Number of bytes necessary for a full state emit. */ - GLboolean is_dirty, all_dirty; }; struct r200_state { @@ -600,16 +594,10 @@ struct r200_context { */ struct radeon_ioctl ioctl; struct radeon_store store; - /* A full state emit as of the first state emit in the main store, in case - * the context is lost. - */ - struct radeon_store backup_store; /* Clientdata textures; */ - GLuint prefer_gart_client_texturing; - - GLboolean save_on_next_emit; + GLuint prefer_gart_client_texturing; /* TCL stuff */ diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c index 0b7f664457..8f64c8c6d9 100644 --- a/src/mesa/drivers/dri/r200/r200_ioctl.c +++ b/src/mesa/drivers/dri/r200/r200_ioctl.c @@ -57,34 +57,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R200_TIMEOUT 512 #define R200_IDLE_RETRY 16 - -/* At this point we were in FlushCmdBufLocked but we had lost our context, so - * we need to unwire our current cmdbuf, hook the one with the saved state in - * it, flush it, and then put the current one back. This is so commands at the - * start of a cmdbuf can rely on the state being kept from the previous one. - */ -static void r200BackUpAndEmitLostStateLocked( r200ContextPtr rmesa ) -{ - GLuint nr_released_bufs; - struct radeon_store saved_store; - - if (rmesa->backup_store.cmd_used == 0) - return; - - if (R200_DEBUG & DEBUG_STATE) - fprintf(stderr, "Emitting backup state on lost context\n"); - - rmesa->radeon.lost_context = GL_FALSE; - - nr_released_bufs = rmesa->radeon.dma.nr_released_bufs; - saved_store = rmesa->store; - rmesa->radeon.dma.nr_released_bufs = 0; - rmesa->store = rmesa->backup_store; - rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ ); - rmesa->radeon.dma.nr_released_bufs = nr_released_bufs; - rmesa->store = saved_store; -} - static void r200UserClear(GLcontext *ctx, GLuint flags) { if (flags & (RADEON_FRONT | RADEON_BACK)) { @@ -291,36 +263,9 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask ) else r200KernelClear(ctx, flags); - rmesa->hw.all_dirty = GL_TRUE; -} - - -void r200Flush( GLcontext *ctx ) -{ - r200ContextPtr rmesa = R200_CONTEXT( ctx ); - - if (R200_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (rmesa->radeon.dma.flush) - rmesa->radeon.dma.flush( ctx ); - - r200EmitState( rmesa ); - - if (rmesa->radeon.cmdbuf.cs->cdw) - rcommonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ ); + rmesa->radeon.hw.all_dirty = GL_TRUE; } -/* Make sure all commands have been sent to the hardware and have - * completed processing. - */ -void r200Finish( GLcontext *ctx ) -{ - r200Flush( ctx ); - radeon_common_finish(ctx); -} - - /* This version of AllocateMemoryMESA allocates only GART memory, and * only does so after the point at which the driver has been * initialized. @@ -458,7 +403,7 @@ GLuint r200GartOffsetFromVirtual( r200ContextPtr rmesa, const GLvoid *pointer ) void r200InitIoctlFuncs( struct dd_function_table *functions ) { functions->Clear = r200Clear; - functions->Finish = r200Finish; - functions->Flush = r200Flush; + functions->Finish = radeonFinish; + functions->Flush = radeonFlush; } diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.h b/src/mesa/drivers/dri/r200/r200_ioctl.h index 316acef459..6b114971c3 100644 --- a/src/mesa/drivers/dri/r200/r200_ioctl.h +++ b/src/mesa/drivers/dri/r200/r200_ioctl.h @@ -47,7 +47,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "common_cmdbuf.h" -extern void r200EmitState( r200ContextPtr rmesa ); extern void r200EmitVertexAOS( r200ContextPtr rmesa, GLuint vertex_size, struct radeon_bo *bo, @@ -101,7 +100,7 @@ do { \ do { \ R200_NEWPRIM( rmesa ); \ rmesa->hw.ATOM.dirty = GL_TRUE; \ - rmesa->hw.is_dirty = GL_TRUE; \ + rmesa->radeon.hw.is_dirty = GL_TRUE; \ } while (0) #define R200_DB_STATE( ATOM ) \ @@ -116,7 +115,7 @@ static INLINE int R200_DB_STATECHANGE( GLuint *tmp; R200_NEWPRIM( rmesa ); atom->dirty = GL_TRUE; - rmesa->hw.is_dirty = GL_TRUE; + rmesa->radeon.hw.is_dirty = GL_TRUE; tmp = atom->cmd; atom->cmd = atom->lastcmd; atom->lastcmd = tmp; @@ -127,15 +126,6 @@ static INLINE int R200_DB_STATECHANGE( } -/* Fire the buffered vertices no matter what. - */ -#define R200_FIREVERTICES( rmesa ) \ -do { \ - if ( rmesa->radeon.cmdbuf.cs->cdw || rmesa->radeon.dma.flush ) { \ - r200Flush( rmesa->radeon.glCtx ); \ - } \ -} while (0) - /* Command lengths. Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ * are available, you will also be adding an rmesa->state.max_state_size because * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts. diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c index f3d809d62c..5c41018cdb 100644 --- a/src/mesa/drivers/dri/r200/r200_state.c +++ b/src/mesa/drivers/dri/r200/r200_state.c @@ -524,25 +524,6 @@ static void r200Fogfv( GLcontext *ctx, GLenum pname, const GLfloat *param ) } } - -/* ============================================================= - * Scissoring - */ - - -static void r200Scissor( GLcontext *ctx, - GLint x, GLint y, GLsizei w, GLsizei h ) -{ - r200ContextPtr rmesa = R200_CONTEXT(ctx); - - if ( ctx->Scissor.Enabled ) { - R200_FIREVERTICES( rmesa ); /* don't pipeline cliprect changes */ - radeonUpdateScissor( ctx ); - } - -} - - /* ============================================================= * Culling */ @@ -787,7 +768,7 @@ static void r200PolygonStipple( GLcontext *ctx, const GLubyte *mask ) /* TODO: push this into cmd mechanism */ - R200_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); LOCK_HARDWARE( &rmesa->radeon ); /* FIXME: Use window x,y offsets into stipple RAM. @@ -1638,7 +1619,7 @@ void r200UpdateWindow( GLcontext *ctx ) float_ui32_type sz = { v[MAT_SZ] * rmesa->radeon.state.depth.scale }; float_ui32_type tz = { v[MAT_TZ] * rmesa->radeon.state.depth.scale }; - R200_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); R200_STATECHANGE( rmesa, vpt ); rmesa->hw.vpt.cmd[VPT_SE_VPORT_XSCALE] = sx.ui32; @@ -1783,7 +1764,7 @@ static void r200DrawBuffer( GLcontext *ctx, GLenum mode ) fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr( mode )); - R200_FIREVERTICES(rmesa); /* don't pipeline cliprect changes */ + radeon_firevertices(&rmesa->radeon); /* don't pipeline cliprect changes */ if (ctx->DrawBuffer->_NumColorDrawBuffers != 1) { /* 0 (GL_NONE) buffers or multiple color drawing buffers */ @@ -2057,7 +2038,7 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state ) } case GL_SCISSOR_TEST: - R200_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); rmesa->radeon.state.scissor.enabled = state; radeonUpdateScissor( ctx ); break; @@ -2575,7 +2556,7 @@ void r200InitStateFuncs( struct dd_function_table *functions ) functions->PointParameterfv = r200PointParameter; functions->PointSize = r200PointSize; functions->RenderMode = r200RenderMode; - functions->Scissor = r200Scissor; + functions->Scissor = radeonScissor; functions->ShadeModel = r200ShadeModel; functions->StencilFuncSeparate = r200StencilFuncSeparate; functions->StencilMaskSeparate = r200StencilMaskSeparate; diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c index 1e904ea432..d11baacae9 100644 --- a/src/mesa/drivers/dri/r200/r200_state_init.c +++ b/src/mesa/drivers/dri/r200/r200_state_init.c @@ -46,8 +46,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_buffer.h" #include "radeon_mipmap_tree.h" #include "radeon_cs.h" -#include "common_context.h" -#include "common_cmdbuf.h" +#include "common_misc.h" #include "r200_context.h" #include "r200_ioctl.h" #include "r200_state.h" @@ -177,8 +176,8 @@ void r200PrintDirty( r200ContextPtr rmesa, const char *msg ) fprintf(stderr, msg); fprintf(stderr, ": "); - foreach(l, &rmesa->hw.atomlist) { - if (l->dirty || rmesa->hw.all_dirty) + foreach(l, &rmesa->radeon.hw.atomlist) { + if (l->dirty || rmesa->radeon.hw.all_dirty) fprintf(stderr, "%s, ", l->name); } @@ -649,7 +648,7 @@ void r200InitState( r200ContextPtr rmesa ) rmesa->state.pixel.readPitch = rmesa->radeon.state.color.drawPitch; #endif - rmesa->hw.max_state_size = 0; + rmesa->radeon.hw.max_state_size = 0; #define ALLOC_STATE( ATOM, CHK, SZ, NM, IDX ) \ do { \ @@ -660,7 +659,7 @@ void r200InitState( r200ContextPtr rmesa ) rmesa->hw.ATOM.idx = IDX; \ rmesa->hw.ATOM.check = check_##CHK; \ rmesa->hw.ATOM.dirty = GL_FALSE; \ - rmesa->hw.max_state_size += SZ * sizeof(int); \ + rmesa->radeon.hw.max_state_size += SZ * sizeof(int); \ } while (0) @@ -1405,7 +1404,7 @@ void r200InitState( r200ContextPtr rmesa ) r200LightingSpaceChange( ctx ); - rmesa->hw.all_dirty = GL_TRUE; + rmesa->radeon.hw.all_dirty = GL_TRUE; - rcommonInitCmdBuf(&rmesa->radeon, rmesa->hw.max_state_size); + rcommonInitCmdBuf(&rmesa->radeon); } diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index a39072d1d2..df786561e1 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -204,8 +204,6 @@ static void r200SetVertexFormat( GLcontext *ctx ) static void r200RenderStart( GLcontext *ctx ) { - r200ContextPtr rmesa = R200_CONTEXT( ctx ); - r200SetVertexFormat( ctx ); } @@ -271,11 +269,11 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) { r200ContextPtr rmesa = R200_CONTEXT(ctx); rcommonEnsureCmdBufSpace(&rmesa->radeon, - rmesa->hw.max_state_size + (12*sizeof(int)), + rmesa->radeon.hw.max_state_size + (12*sizeof(int)), __FUNCTION__); - r200EmitState(rmesa); + radeonEmitState(&rmesa->radeon); r200EmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, rmesa->radeon.dma.current, @@ -649,7 +647,7 @@ void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode ) if (mode) { rmesa->radeon.Fallback |= bit; if (oldfallback == 0) { - R200_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); TCL_FALLBACK( ctx, R200_TCL_FALLBACK_RASTER, GL_TRUE ); _swsetup_Wakeup( ctx ); rmesa->radeon.swtcl.RenderIndex = ~0; @@ -914,6 +912,4 @@ void r200InitSwtcl( GLcontext *ctx ) void r200DestroySwtcl( GLcontext *ctx ) { - r200ContextPtr rmesa = R200_CONTEXT(ctx); - } diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c index 4c4df14ef8..8e0fb14e70 100644 --- a/src/mesa/drivers/dri/r200/r200_tcl.c +++ b/src/mesa/drivers/dri/r200/r200_tcl.c @@ -123,7 +123,7 @@ static GLboolean discrete_prim[0x10] = { #define RESET_STIPPLE() do { \ R200_STATECHANGE( rmesa, lin ); \ - r200EmitState( rmesa ); \ + radeonEmitState(&rmesa->radeon); \ } while (0) #define AUTO_STIPPLE( mode ) do { \ @@ -134,7 +134,7 @@ static GLboolean discrete_prim[0x10] = { else \ rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &= \ ~R200_LINE_PATTERN_AUTO_RESET; \ - r200EmitState( rmesa ); \ + radeonEmitState(&rmesa->radeon); \ } while (0) @@ -189,7 +189,7 @@ static void r200EmitPrim( GLcontext *ctx, // fprintf(stderr,"Emit prim %d\n", rmesa->tcl.nr_aos_components); rcommonEnsureCmdBufSpace( &rmesa->radeon, AOS_BUFSZ(rmesa->tcl.nr_aos_components) + - rmesa->hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); + rmesa->radeon.hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); r200EmitAOS( rmesa, rmesa->tcl.nr_aos_components, diff --git a/src/mesa/drivers/dri/r200/r200_tex.c b/src/mesa/drivers/dri/r200/r200_tex.c index cf0a32bf89..c0d1723c77 100644 --- a/src/mesa/drivers/dri/r200/r200_tex.c +++ b/src/mesa/drivers/dri/r200/r200_tex.c @@ -420,7 +420,7 @@ static void r200DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) if (rmesa) { int i; - R200_FIREVERTICES(rmesa); + radeon_firevertices(&rmesa->radeon); for ( i = 0 ; i < rmesa->radeon.glCtx->Const.MaxTextureUnits ; i++ ) { if ( t == rmesa->state.texture.unit[i].texobj ) { rmesa->state.texture.unit[i].texobj = NULL; diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile index 4042005d03..abdd1da490 100644 --- a/src/mesa/drivers/dri/r300/Makefile +++ b/src/mesa/drivers/dri/r300/Makefile @@ -26,11 +26,7 @@ DRIVER_SOURCES = \ radeon_screen.c \ common_lock.c \ common_misc.c \ - radeon_context.c \ - radeon_ioctl.c \ radeon_span.c \ - radeon_state.c \ - r300_context.c \ r300_ioctl.c \ r300_cmdbuf.c \ r300_state.c \ diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c index 303efe3ada..d6acb39d9f 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -45,7 +45,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_drm.h" #include "radeon_buffer.h" -#include "radeon_ioctl.h" #include "r300_context.h" #include "r300_ioctl.h" #include "radeon_reg.h" @@ -61,111 +60,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R300_VAP_PVS_UPLOAD_ADDRESS 0x2200 # define RADEON_ONE_REG_WR (1 << 15) -// Set this to 1 for extremely verbose debugging of command buffers -#define DEBUG_CMDBUF 0 - /** # of dwords reserved for additional instructions that may need to be written * during flushing. */ #define SPACE_FOR_FLUSHING 4 -static void r300PrintStateAtom(r300ContextPtr r300, - struct radeon_state_atom *state) -{ - int i; - int dwords = (*state->check) (r300->radeon.glCtx, state); - - fprintf(stderr, " emit %s %d/%d\n", state->name, dwords, state->cmd_size); - - if (RADEON_DEBUG & DEBUG_VERBOSE) { - for (i = 0; i < dwords; i++) { - fprintf(stderr, " %s[%d]: %08x\n", - state->name, i, state->cmd[i]); - } - } -} - -/** - * Emit all atoms with a dirty field equal to dirty. - * - * The caller must have ensured that there is enough space in the command - * buffer. - */ -static INLINE void r300EmitAtoms(r300ContextPtr r300, GLboolean dirty) -{ - BATCH_LOCALS(&r300->radeon); - struct radeon_state_atom *atom; - int dwords; - - cp_wait(r300, R300_WAIT_3D | R300_WAIT_3D_CLEAN); - BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH(cmdpacket0(r300->radeon.radeonScreen, R300_TX_INVALTAGS, 1)); - OUT_BATCH(R300_TX_FLUSH); - END_BATCH(); - end_3d(r300); - - /* Emit actual atoms */ - foreach(atom, &r300->hw.atomlist) { - if ((atom->dirty || r300->hw.all_dirty) == dirty) { - dwords = (*atom->check) (r300->radeon.glCtx, atom); - if (dwords) { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - r300PrintStateAtom(r300, atom); - } - if (atom->emit) { - (*atom->emit)(r300->radeon.glCtx, atom); - } else { - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(atom->cmd, dwords); - END_BATCH(); - } - atom->dirty = GL_FALSE; - } else { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - fprintf(stderr, " skip state %s\n", - atom->name); - } - } - } - } - - COMMIT_BATCH(); -} - -/** - * Copy dirty hardware state atoms into the command buffer. - * - * We also copy out clean state if we're at the start of a buffer. That makes - * it easy to recover from lost contexts. - */ -void r300EmitState(r300ContextPtr r300) -{ - if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_PRIMS)) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (r300->radeon.cmdbuf.cs->cdw && !r300->hw.is_dirty && !r300->hw.all_dirty) - return; - - /* To avoid going across the entire set of states multiple times, just check - * for enough space for the case of emitting all state. - */ - rcommonEnsureCmdBufSpace(&r300->radeon, r300->hw.max_state_size, __FUNCTION__); - - if (!r300->radeon.cmdbuf.cs->cdw) { - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin reemit state\n"); - - r300EmitAtoms(r300, GL_FALSE); - } - - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin dirty state\n"); - - r300EmitAtoms(r300, GL_TRUE); - r300->hw.is_dirty = GL_FALSE; - r300->hw.all_dirty = GL_FALSE; -} - static unsigned packet0_count(r300ContextPtr r300, uint32_t *pkt) { if (r300->radeon.radeonScreen->kernel_mm) { @@ -405,8 +304,8 @@ int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom) r300->hw.ATOM.idx = (IDX); \ r300->hw.ATOM.check = check_##CHK; \ r300->hw.ATOM.dirty = GL_FALSE; \ - r300->hw.max_state_size += (SZ); \ - insert_at_tail(&r300->hw.atomlist, &r300->hw.ATOM); \ + r300->radeon.hw.max_state_size += (SZ); \ + insert_at_tail(&r300->radeon.hw.atomlist, &r300->hw.ATOM); \ } while (0) /** * Allocate memory for the command buffer and initialize the state atom @@ -425,7 +324,7 @@ void r300InitCmdBuf(r300ContextPtr r300) if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) is_r500 = 1; - r300->hw.max_state_size = 2 + 2; /* reserve extra space for WAIT_IDLE and tex cache flush */ + r300->radeon.hw.max_state_size = 2 + 2; /* reserve extra space for WAIT_IDLE and tex cache flush */ mtu = r300->radeon.glCtx->Const.MaxTextureUnits; if (RADEON_DEBUG & DEBUG_TEXTURE) { @@ -433,8 +332,8 @@ void r300InitCmdBuf(r300ContextPtr r300) } /* Setup the atom linked list */ - make_empty_list(&r300->hw.atomlist); - r300->hw.atomlist.name = "atom-list"; + make_empty_list(&r300->radeon.hw.atomlist); + r300->radeon.hw.atomlist.name = "atom-list"; /* Initialize state atoms */ ALLOC_STATE(vpt, always, R300_VPT_CMDSIZE, 0); @@ -708,10 +607,10 @@ void r300InitCmdBuf(r300ContextPtr r300) r300->hw.tex.border_color.cmd[R300_TEX_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_TX_BORDER_COLOR_0, 0); - r300->hw.is_dirty = GL_TRUE; - r300->hw.all_dirty = GL_TRUE; + r300->radeon.hw.is_dirty = GL_TRUE; + r300->radeon.hw.all_dirty = GL_TRUE; - rcommonInitCmdBuf(&r300->radeon, r300->hw.max_state_size); + rcommonInitCmdBuf(&r300->radeon); } /** @@ -721,7 +620,7 @@ void r300DestroyCmdBuf(r300ContextPtr r300) { struct radeon_state_atom *atom; - foreach(atom, &r300->hw.atomlist) { + foreach(atom, &r300->radeon.hw.atomlist) { FREE(atom->cmd); } diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c index 3e47888e45..e4dc37debc 100644 --- a/src/mesa/drivers/dri/r300/r300_context.c +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -55,7 +55,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "drivers/common/driverfuncs.h" -#include "radeon_ioctl.h" #include "radeon_span.h" #include "r300_context.h" #include "r300_cmdbuf.h" @@ -198,22 +197,6 @@ static void r300_get_lock(radeonContextPtr rmesa) } } -static void r300_vtbl_flush(GLcontext *ctx) -{ - r300Flush(ctx); -} - -static void r300_vtbl_set_all_dirty(GLcontext *ctx) -{ - r300ContextPtr rmesa = R300_CONTEXT(ctx); - rmesa->hw.all_dirty = GL_TRUE; -} - -static void r300_vtbl_emit_state(radeonContextPtr rmesa) -{ - r300EmitState((r300ContextPtr)rmesa); -} - static void r300_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa) { /* please flush pipe do all pending work */ @@ -243,22 +226,25 @@ static void r300_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmes R300_WAIT_3D | R300_WAIT_3D_CLEAN)); } -static void r300_vtbl_flush_vertices(radeonContextPtr rmesa) +static void r300_vtbl_pre_emit_atoms(radeonContextPtr radeon) { - R300_FIREVERTICES(((r300ContextPtr)rmesa)); + BATCH_LOCALS(radeon); + cp_wait(radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); + BEGIN_BATCH_NO_AUTOSTATE(2); + OUT_BATCH(cmdpacket0(radeon->radeonScreen, R300_TX_INVALTAGS, 1)); + OUT_BATCH(R300_TX_FLUSH); + END_BATCH(); + end_3d(radeon); } static void r300_init_vtbl(radeonContextPtr radeon) { radeon->vtbl.get_lock = r300_get_lock; radeon->vtbl.update_viewport_offset = r300UpdateViewportOffset; - radeon->vtbl.flush = r300_vtbl_flush; - radeon->vtbl.set_all_dirty = r300_vtbl_set_all_dirty; radeon->vtbl.update_draw_buffer = r300UpdateDrawBuffer; radeon->vtbl.emit_cs_header = r300_vtbl_emit_cs_header; - radeon->vtbl.emit_state = r300_vtbl_emit_state; - radeon->vtbl.flush_vertices = r300_vtbl_flush_vertices; radeon->vtbl.swtcl_flush = r300_swtcl_flush; + radeon->vtbl.pre_emit_atoms = r300_vtbl_pre_emit_atoms; } diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 9e97a136e3..75393bcd14 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -42,6 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_drm.h" #include "dri_util.h" #include "texmem.h" +#include "common_context.h" #include "radeon_context.h" #include "radeon_bo.h" @@ -53,7 +54,7 @@ struct r300_context; typedef struct r300_context r300ContextRec; typedef struct r300_context *r300ContextPtr; -#include "radeon_lock.h" + #include "main/mm.h" /* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html . @@ -304,12 +305,6 @@ struct r300_texture_state { * Cache for hardware register state. */ struct r300_hw_state { - struct radeon_state_atom atomlist; - - GLboolean is_dirty; - GLboolean all_dirty; - int max_state_size; /* in dwords */ - struct radeon_state_atom vpt; /* viewport (1D98) */ struct radeon_state_atom vap_cntl; struct radeon_state_atom vap_index_offset; /* 0x208c r5xx only */ diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c index 4a18397aa8..20afae6a09 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.c +++ b/src/mesa/drivers/dri/r300/r300_emit.c @@ -46,7 +46,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_context.h" #include "r300_context.h" -#include "radeon_ioctl.h" #include "r300_state.h" #include "r300_emit.h" #include "r300_ioctl.h" diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h index 290bd614c5..6c02646605 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.h +++ b/src/mesa/drivers/dri/r300/r300_emit.h @@ -146,36 +146,36 @@ static INLINE uint32_t cmdpacify(struct radeon_screen *rscrn) /** * Must be sent to switch to 2d commands */ -void static INLINE end_3d(r300ContextPtr rmesa) +void static INLINE end_3d(radeonContextPtr radeon) { - BATCH_LOCALS(&rmesa->radeon); + BATCH_LOCALS(radeon); - if (!rmesa->radeon.radeonScreen->kernel_mm) { - BEGIN_BATCH_NO_AUTOSTATE(1); - OUT_BATCH(cmdpacify(rmesa->radeon.radeonScreen)); - END_BATCH(); - } + if (!radeon->radeonScreen->kernel_mm) { + BEGIN_BATCH_NO_AUTOSTATE(1); + OUT_BATCH(cmdpacify(radeon->radeonScreen)); + END_BATCH(); + } } void static INLINE cp_delay(r300ContextPtr rmesa, unsigned short count) { BATCH_LOCALS(&rmesa->radeon); - if (!rmesa->radeon.radeonScreen->kernel_mm) { - BEGIN_BATCH_NO_AUTOSTATE(1); - OUT_BATCH(cmdcpdelay(rmesa->radeon.radeonScreen, count)); - END_BATCH(); - } + if (!rmesa->radeon.radeonScreen->kernel_mm) { + BEGIN_BATCH_NO_AUTOSTATE(1); + OUT_BATCH(cmdcpdelay(rmesa->radeon.radeonScreen, count)); + END_BATCH(); + } } -void static INLINE cp_wait(r300ContextPtr rmesa, unsigned char flags) +void static INLINE cp_wait(radeonContextPtr radeon, unsigned char flags) { - BATCH_LOCALS(&rmesa->radeon); + BATCH_LOCALS(radeon); uint32_t wait_until; - if (!rmesa->radeon.radeonScreen->kernel_mm) { + if (!radeon->radeonScreen->kernel_mm) { BEGIN_BATCH_NO_AUTOSTATE(1); - OUT_BATCH(cmdwait(rmesa->radeon.radeonScreen, flags)); + OUT_BATCH(cmdwait(radeon->radeonScreen, flags)); END_BATCH(); } else { switch(flags) { diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c index f07252fbe2..ee33415095 100644 --- a/src/mesa/drivers/dri/r300/r300_ioctl.c +++ b/src/mesa/drivers/dri/r300/r300_ioctl.c @@ -48,7 +48,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_buffer.h" #include "r300_context.h" -#include "radeon_ioctl.h" #include "r300_ioctl.h" #include "r300_cmdbuf.h" #include "r300_state.h" @@ -94,8 +93,8 @@ static void r300ClearBuffer(r300ContextPtr r300, int flags, } /* TODO in bufmgr */ - cp_wait(r300, R300_WAIT_3D | R300_WAIT_3D_CLEAN); - end_3d(rmesa); + cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); + end_3d(&rmesa->radeon); if (flags & CLEARBUFFER_COLOR) { assert(rrb != 0); @@ -195,7 +194,7 @@ static void r300ClearBuffer(r300ContextPtr r300, int flags, } r300EmitCacheFlush(rmesa); - cp_wait(rmesa, R300_WAIT_3D | R300_WAIT_3D_CLEAN); + cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); R300_STATECHANGE(r300, cb); R300_STATECHANGE(r300, cmk); diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index 23d068c339..393bf754ff 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -66,8 +66,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_vp_build.h" #include "radeon_reg.h" #include "radeon_macros.h" -#include "radeon_ioctl.h" -#include "radeon_state.h" #include "r300_context.h" #include "r300_ioctl.h" #include "r300_state.h" diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index bdb67a1afd..1e7d9af11d 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -53,8 +53,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "vbo/vbo.h" #include "tnl/tnl.h" -#include "radeon_ioctl.h" -#include "radeon_state.h" #include "radeon_buffer.h" #include "r300_context.h" #include "r300_ioctl.h" @@ -1087,7 +1085,7 @@ static void r300UpdateWindow(GLcontext * ctx) GLfloat sz = v[MAT_SZ] * rmesa->radeon.state.depth.scale; GLfloat tz = v[MAT_TZ] * rmesa->radeon.state.depth.scale; - R300_FIREVERTICES(rmesa); + radeon_firevertices(&rmesa->radeon); R300_STATECHANGE(rmesa, vpt); rmesa->hw.vpt.cmd[R300_VPT_XSCALE] = r300PackFloat32(sx); @@ -2112,6 +2110,7 @@ static void r300SetupVertexProgram(r300ContextPtr rmesa) */ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state) { + r300ContextPtr rmesa = R300_CONTEXT(ctx); if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "%s( %s = %s )\n", __FUNCTION__, _mesa_lookup_enum_by_nr(cap), @@ -2157,8 +2156,12 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state) case GL_POLYGON_OFFSET_FILL: r300SetPolygonOffsetState(ctx, state); break; + case GL_SCISSOR_TEST: + radeon_firevertices(&rmesa->radeon); + rmesa->radeon.state.scissor.enabled = state; + radeonUpdateScissor( ctx ); + break; default: - radeonEnable(ctx, cap, state); break; } } @@ -2405,7 +2408,7 @@ static void r300ResetHwState(r300ContextPtr r300) r300->hw.vps.cmd[R300_VPS_ZERO_3] = 0; } - r300->hw.all_dirty = GL_TRUE; + r300->radeon.hw.all_dirty = GL_TRUE; } void r300UpdateShaders(r300ContextPtr rmesa) @@ -2747,7 +2750,7 @@ static void r300DrawBuffer( GLcontext *ctx, GLenum mode ) fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr( mode )); - R300_FIREVERTICES(rmesa); /* don't pipeline cliprect changes */ + radeon_firevertices(&rmesa->radeon); /* don't pipeline cliprect changes */ radeonSetCliprects( &rmesa->radeon ); radeonUpdatePageFlipping(&rmesa->radeon); @@ -2803,6 +2806,7 @@ void r300InitStateFuncs(struct dd_function_table *functions) functions->RenderMode = r300RenderMode; functions->ClipPlane = r300ClipPlane; + functions->Scissor = radeonScissor; functions->DrawBuffer = r300DrawBuffer; functions->ReadBuffer = r300ReadBuffer; diff --git a/src/mesa/drivers/dri/r300/r300_state.h b/src/mesa/drivers/dri/r300/r300_state.h index 4d0a25fb0e..247a20ee51 100644 --- a/src/mesa/drivers/dri/r300/r300_state.h +++ b/src/mesa/drivers/dri/r300/r300_state.h @@ -39,26 +39,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R300_NEWPRIM( rmesa ) \ do { \ + if ( rmesa->radeon.dma.flush ) \ + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); \ } while (0) #define R300_STATECHANGE(r300, atom) \ do { \ R300_NEWPRIM(r300); \ r300->hw.atom.dirty = GL_TRUE; \ - r300->hw.is_dirty = GL_TRUE; \ + r300->radeon.hw.is_dirty = GL_TRUE; \ } while(0) -#define R300_PRINT_STATE(r300, atom) \ - r300PrintStateAtom(r300, &r300->hw.atom) - -/* Fire the buffered vertices no matter what. - TODO: This has not been implemented yet - */ -#define R300_FIREVERTICES( r300 ) \ - do { \ - r300Flush( (r300)->radeon.glCtx ); \ - } while (0) - // r300_state.c extern int future_hw_tcl_on; void _tnl_UpdateFixedFunctionProgram (GLcontext * ctx); diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index 98208cd77e..34e752178a 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -626,7 +626,7 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r300ContextPtr rmesa = R300_CONTEXT(ctx); rcommonEnsureCmdBufSpace(&rmesa->radeon, - rmesa->hw.max_state_size + (12*sizeof(int)), + rmesa->radeon.hw.max_state_size + (12*sizeof(int)), __FUNCTION__); r300EmitState(rmesa); r300EmitVertexAOS(rmesa, diff --git a/src/mesa/drivers/dri/r300/r300_tex.c b/src/mesa/drivers/dri/r300/r300_tex.c index 989aa462ae..e316e2d822 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.c +++ b/src/mesa/drivers/dri/r300/r300_tex.c @@ -263,7 +263,7 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) if (rmesa) { int i; - R300_FIREVERTICES(rmesa); + radeon_firevertices(&rmesa->radeon); for(i = 0; i < R300_MAX_TEXTURE_UNITS; ++i) if (rmesa->hw.textures[i] == t) diff --git a/src/mesa/drivers/dri/r300/r300_texmem.c b/src/mesa/drivers/dri/r300/r300_texmem.c deleted file mode 100644 index 53eeca1386..0000000000 --- a/src/mesa/drivers/dri/r300/r300_texmem.c +++ /dev/null @@ -1,58 +0,0 @@ -/************************************************************************** - -Copyright (C) Tungsten Graphics 2002. All Rights Reserved. -The Weather Channel, Inc. funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 -license. This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation on the rights to use, copy, modify, merge, publish, -distribute, sub license, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND -NON-INFRINGEMENT. IN NO EVENT SHALL ATI, VA LINUX SYSTEMS AND/OR THEIR -SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER -IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR -IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE -SOFTWARE. - -**************************************************************************/ - -/** - * \file - * - * \author Gareth Hughes - * - * \author Kevin E. Martin - */ - -#include - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/context.h" -#include "main/colormac.h" -#include "main/macros.h" -#include "main/simple_list.h" -#include "radeon_reg.h" /* gets definition for usleep */ -#include "r300_context.h" -#include "r300_state.h" -#include "r300_cmdbuf.h" -#include "r300_emit.h" -#include "r300_mipmap_tree.h" -#include "radeon_ioctl.h" -#include "r300_tex.h" -#include "r300_ioctl.h" -#include /* for usleep() */ - - diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c index 6e2b1309c5..50ecc570bd 100644 --- a/src/mesa/drivers/dri/r300/r300_texstate.c +++ b/src/mesa/drivers/dri/r300/r300_texstate.c @@ -47,7 +47,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" #include "r300_ioctl.h" -#include "radeon_ioctl.h" #include "radeon_mipmap_tree.h" #include "radeon_cs.h" #include "r300_tex.h" diff --git a/src/mesa/drivers/dri/r300/radeon_context.c b/src/mesa/drivers/dri/r300/radeon_context.c deleted file mode 100644 index 13a7682416..0000000000 --- a/src/mesa/drivers/dri/r300/radeon_context.c +++ /dev/null @@ -1,75 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/** - * \file radeon_context.c - * Common context initialization. - * - * \author Keith Whitwell - */ - -#include - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/context.h" -#include "main/state.h" -#include "main/matrix.h" -#include "main/framebuffer.h" - -#include "drivers/common/driverfuncs.h" -#include "swrast/swrast.h" - -#include "radeon_buffer.h" -#include "radeon_screen.h" -#include "radeon_ioctl.h" -#include "radeon_macros.h" -#include "radeon_reg.h" - -#include "radeon_state.h" -#include "r300_state.h" - -#include "utils.h" -#include "vblank.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ -#include "drirenderbuffer.h" - -/* Force the context `c' to be unbound from its buffer. - */ -GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv) -{ - radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; - - if (RADEON_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx %p\n", __FUNCTION__, - radeon->glCtx); - - return GL_TRUE; -} - diff --git a/src/mesa/drivers/dri/r300/radeon_context.h b/src/mesa/drivers/dri/r300/radeon_context.h index 736bf7c649..005edf8448 100644 --- a/src/mesa/drivers/dri/r300/radeon_context.h +++ b/src/mesa/drivers/dri/r300/radeon_context.h @@ -67,20 +67,6 @@ extern void radeonFallback(GLcontext * ctx, GLuint bit, GLboolean mode); /* TCL fallbacks */ extern void radeonTclFallback(GLcontext * ctx, GLuint bit, GLboolean mode); -#define RADEON_TCL_FALLBACK_RASTER 0x0001 /* rasterization */ -#define RADEON_TCL_FALLBACK_UNFILLED 0x0002 /* unfilled tris */ -#define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE 0x0004 /* twoside tris */ -#define RADEON_TCL_FALLBACK_MATERIAL 0x0008 /* material in vb */ -#define RADEON_TCL_FALLBACK_TEXGEN_0 0x0010 /* texgen, unit 0 */ -#define RADEON_TCL_FALLBACK_TEXGEN_1 0x0020 /* texgen, unit 1 */ -#define RADEON_TCL_FALLBACK_TEXGEN_2 0x0040 /* texgen, unit 2 */ -#define RADEON_TCL_FALLBACK_TEXGEN_3 0x0080 /* texgen, unit 3 */ -#define RADEON_TCL_FALLBACK_TEXGEN_4 0x0100 /* texgen, unit 4 */ -#define RADEON_TCL_FALLBACK_TEXGEN_5 0x0200 /* texgen, unit 5 */ -#define RADEON_TCL_FALLBACK_TCL_DISABLE 0x0400 /* user disable */ -#define RADEON_TCL_FALLBACK_BITMAP 0x0800 /* draw bitmap with points */ -#define RADEON_TCL_FALLBACK_VERTEX_PROGRAM 0x1000 /* vertex program active */ - #if R200_MERGED #define TCL_FALLBACK( ctx, bit, mode ) radeonTclFallback( ctx, bit, mode ) #else @@ -88,33 +74,4 @@ extern void radeonTclFallback(GLcontext * ctx, GLuint bit, GLboolean mode); #endif -extern void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv, - int x, int y, int w, int h); -extern GLboolean radeonInitContext(radeonContextPtr radeon, - struct dd_function_table *functions, - const __GLcontextModes * glVisual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); -extern void radeonCleanupContext(radeonContextPtr radeon); -extern GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); -extern GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv); -void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable); - -/* ================================================================ - * Debugging: - */ -#define DO_DEBUG 1 - -#if DO_DEBUG -extern int RADEON_DEBUG; -#else -#define RADEON_DEBUG 0 -#endif - -#define DEBUG_SYNC 0x1000 -#define DEBUG_PIXEL 0x2000 -#define DEBUG_MEMORY 0x4000 - #endif /* __RADEON_CONTEXT_H__ */ diff --git a/src/mesa/drivers/dri/r300/radeon_ioctl.c b/src/mesa/drivers/dri/r300/radeon_ioctl.c deleted file mode 100644 index 120d6c03ab..0000000000 --- a/src/mesa/drivers/dri/r300/radeon_ioctl.c +++ /dev/null @@ -1,69 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell - */ - -#include -#include - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/macros.h" -#include "main/context.h" -#include "swrast/swrast.h" -#include "r300_context.h" -#include "radeon_ioctl.h" -#include "radeon_buffer.h" -#include "r300_ioctl.h" -#include "r300_state.h" -#include "radeon_reg.h" - -#include "drirenderbuffer.h" -#include "vblank.h" - -void radeonFlush(GLcontext * ctx) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - - if (IS_R300_CLASS(radeon->radeonScreen)) - r300Flush(ctx); -} - - -/* Make sure all commands have been sent to the hardware and have - * completed processing. - */ -void radeonFinish(GLcontext * ctx) -{ - radeonFlush(ctx); - radeon_common_finish(ctx); -} diff --git a/src/mesa/drivers/dri/r300/radeon_ioctl.h b/src/mesa/drivers/dri/r300/radeon_ioctl.h deleted file mode 100644 index 3add775b82..0000000000 --- a/src/mesa/drivers/dri/r300/radeon_ioctl.h +++ /dev/null @@ -1,57 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell - */ - -#ifndef __RADEON_IOCTL_H__ -#define __RADEON_IOCTL_H__ - -#include "main/simple_list.h" -#include "radeon_dri.h" -#include "radeon_lock.h" - -#include "xf86drm.h" -#include "drm.h" -#if 0 -#include "r200context.h" -#endif -#include "radeon_drm.h" - -extern void radeonCopyBuffer(__DRIdrawablePrivate * drawable, - const drm_clip_rect_t * rect); -extern void radeonPageFlip(__DRIdrawablePrivate * drawable); -extern void radeonFlush(GLcontext * ctx); -extern void radeonFinish(GLcontext * ctx); -extern void radeonWaitForIdleLocked(radeonContextPtr radeon); -extern uint32_t radeonGetAge(radeonContextPtr radeon); - -#endif /* __RADEON_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/r300/radeon_state.c b/src/mesa/drivers/dri/r300/radeon_state.c deleted file mode 100644 index 0ee5313043..0000000000 --- a/src/mesa/drivers/dri/r300/radeon_state.c +++ /dev/null @@ -1,109 +0,0 @@ -/************************************************************************** - -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell - */ - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/api_arrayelt.h" -#include "main/enums.h" -#include "main/framebuffer.h" -#include "main/colormac.h" -#include "main/light.h" - -#include "swrast/swrast.h" -#include "vbo/vbo.h" -#include "tnl/tnl.h" -#include "tnl/t_pipeline.h" -#include "swrast_setup/swrast_setup.h" - -#include "radeon_ioctl.h" -#include "radeon_state.h" -#include "r300_ioctl.h" - - -/* ============================================================= - * Scissoring - */ - -static void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h) -{ - if (ctx->Scissor.Enabled) { - /* We don't pipeline cliprect changes */ - r300Flush(ctx); - radeonUpdateScissor(ctx); - } -} - -/** - * Handle common enable bits. - * Called as a fallback by r200Enable/r300Enable. - */ -void radeonEnable(GLcontext* ctx, GLenum cap, GLboolean state) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - - switch(cap) { - case GL_SCISSOR_TEST: - /* We don't pipeline cliprect & scissor changes */ - r300Flush(ctx); - - radeon->state.scissor.enabled = state; - radeonUpdateScissor(ctx); - break; - - default: - return; - } -} - - -/** - * Initialize default state. - * This function is called once at context init time from - * r200InitState/r300InitState - */ -void radeonInitState(radeonContextPtr radeon) -{ - radeon->Fallback = 0; -} - - -/** - * Initialize common state functions. - * Called by r200InitStateFuncs/r300InitStateFuncs - */ -void radeonInitStateFuncs(struct dd_function_table *functions) -{ - functions->Scissor = radeonScissor; -} diff --git a/src/mesa/drivers/dri/r300/radeon_state.h b/src/mesa/drivers/dri/r300/radeon_state.h deleted file mode 100644 index 279b108412..0000000000 --- a/src/mesa/drivers/dri/r300/radeon_state.h +++ /dev/null @@ -1,39 +0,0 @@ -/* -Copyright (C) 2004 Nicolai Haehnle. All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Nicolai Haehnle - */ - -#ifndef __RADEON_STATE_H__ -#define __RADEON_STATE_H__ - -extern void radeonEnable(GLcontext* ctx, GLenum cap, GLboolean state); - -extern void radeonInitState(radeonContextPtr radeon); -extern void radeonInitStateFuncs(struct dd_function_table* functions); - -#endif diff --git a/src/mesa/drivers/dri/radeon/common_cmdbuf.h b/src/mesa/drivers/dri/radeon/common_cmdbuf.h index 071e29ee02..5526934209 100644 --- a/src/mesa/drivers/dri/radeon/common_cmdbuf.h +++ b/src/mesa/drivers/dri/radeon/common_cmdbuf.h @@ -1,10 +1,12 @@ #ifndef COMMON_CMDBUF_H #define COMMON_CMDBUF_H +#include "radeon_cs.h" + void rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller); int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller); int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller); -void rcommonInitCmdBuf(radeonContextPtr rmesa, int max_state_size); +void rcommonInitCmdBuf(radeonContextPtr rmesa); void rcommonDestroyCmdBuf(radeonContextPtr rmesa); void rcommonBeginBatch(radeonContextPtr rmesa, @@ -128,4 +130,14 @@ void rcommonBeginBatch(radeonContextPtr rmesa, /** Write a 32 bit float to the ring; requires 1 dword. */ #define OUT_BATCH_FLOAT32(f) \ OUT_BATCH(radeonPackFloat32((f))); + + +/* Fire the buffered vertices no matter what. + */ +static INLINE void radeon_firevertices(radeonContextPtr radeon) +{ + if (radeon->cmdbuf.cs->cdw || radeon->dma.flush ) + radeonFlush(radeon->glCtx); +} + #endif diff --git a/src/mesa/drivers/dri/radeon/common_context.h b/src/mesa/drivers/dri/radeon/common_context.h index 618e74d458..c3a445e0d2 100644 --- a/src/mesa/drivers/dri/radeon/common_context.h +++ b/src/mesa/drivers/dri/radeon/common_context.h @@ -111,6 +111,13 @@ struct radeon_state_atom { void (*emit) (GLcontext *, struct radeon_state_atom *atom); }; +struct radeon_hw_state { + /* Head of the linked list of state atoms. */ + struct radeon_state_atom atomlist; + int max_state_size; /* Number of bytes necessary for a full state emit. */ + GLboolean is_dirty, all_dirty; +}; + /* Texture related */ typedef struct _radeon_texture_image radeon_texture_image; @@ -206,17 +213,6 @@ struct radeon_dma_buffer { drmBufPtr buf; }; -/* A retained region, eg vertices for indexed vertices. - */ -struct radeon_dma_region { - struct radeon_dma_buffer *buf; - char *address; /* == buf->address */ - int start, end, ptr; /* offsets from start of buf */ - int aos_start; - int aos_stride; - int aos_size; -}; - struct radeon_aos { struct radeon_bo *bo; /** Buffer object where vertex data is stored */ int offset; /** Offset into buffer object, in bytes */ @@ -388,6 +384,7 @@ struct radeon_context { float initialMaxAnisotropy; struct radeon_dma dma; + struct radeon_hw_state hw; /* Rasterization and vertex state: */ GLuint TclFallback; @@ -434,13 +431,11 @@ struct radeon_context { struct { void (*get_lock)(radeonContextPtr radeon); void (*update_viewport_offset)(GLcontext *ctx); - void (*flush)(GLcontext *ctx); - void (*set_all_dirty)(GLcontext *ctx); void (*update_draw_buffer)(GLcontext *ctx); void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa); - void (*emit_state)(radeonContextPtr rmesa); - void (*flush_vertices)(radeonContextPtr rmesa); void (*swtcl_flush)(GLcontext *ctx, uint32_t offset); + void (*pre_emit_atoms)(radeonContextPtr rmesa); + void (*pre_emit_state)(radeonContextPtr rmesa); } vtbl; }; @@ -502,4 +497,6 @@ extern int RADEON_DEBUG; #define RADEON_DEBUG 0 #endif +#include "common_misc.h" +#include "common_cmdbuf.h" #endif diff --git a/src/mesa/drivers/dri/radeon/common_misc.c b/src/mesa/drivers/dri/radeon/common_misc.c index 9456f2a438..12536b1672 100644 --- a/src/mesa/drivers/dri/radeon/common_misc.c +++ b/src/mesa/drivers/dri/radeon/common_misc.c @@ -42,6 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/colormac.h" #include "main/light.h" #include "main/framebuffer.h" +#include "main/simple_list.h" #include "swrast/swrast.h" #include "vbo/vbo.h" @@ -81,6 +82,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. int RADEON_DEBUG = (0); #endif +#define DEBUG_CMDBUF 0 + /* ============================================================= * Scissoring */ @@ -215,6 +218,21 @@ void radeonUpdateScissor( GLcontext *ctx ) } } +/* ============================================================= + * Scissoring + */ + +void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + if (ctx->Scissor.Enabled) { + /* We don't pipeline cliprect changes */ + radeon_firevertices(radeon); + radeonUpdateScissor(ctx); + } +} + + /* ================================================================ * SwapBuffers with client-side throttling */ @@ -367,7 +385,7 @@ void radeonCopyBuffer( __DRIdrawablePrivate *dPriv, fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *) rmesa->glCtx ); } - rmesa->vtbl.flush(rmesa->glCtx); + radeon_firevertices(rmesa); LOCK_HARDWARE( rmesa ); /* Throttle the frame rate -- only allow one pending swap buffers @@ -437,7 +455,7 @@ void radeonCopyBuffer( __DRIdrawablePrivate *dPriv, } rmesa->swap_ust = ust; - rmesa->vtbl.set_all_dirty(rmesa->glCtx); + rmesa->hw.all_dirty = GL_TRUE; } } @@ -465,7 +483,7 @@ void radeonPageFlip( __DRIdrawablePrivate *dPriv ) rmesa->sarea->pfCurrentPage); } - rmesa->vtbl.flush(rmesa->glCtx); + radeon_firevertices(rmesa); LOCK_HARDWARE( rmesa ); @@ -516,16 +534,111 @@ void radeonPageFlip( __DRIdrawablePrivate *dPriv ) rmesa->vtbl.update_draw_buffer(rmesa->glCtx); } +void radeonFlush(GLcontext *ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (radeon->dma.flush) + radeon->dma.flush( ctx ); + + radeonEmitState(radeon); + + if (radeon->cmdbuf.cs->cdw) + rcommonFlushCmdBuf(radeon, __FUNCTION__); +} + +static INLINE void radeonEmitAtoms(radeonContextPtr radeon, GLboolean dirty) +{ + BATCH_LOCALS(radeon); + struct radeon_state_atom *atom; + int dwords; + + if (radeon->vtbl.pre_emit_atoms) + radeon->vtbl.pre_emit_atoms(radeon); + + /* Emit actual atoms */ + foreach(atom, &radeon->hw.atomlist) { + if ((atom->dirty || radeon->hw.all_dirty) == dirty) { + dwords = (*atom->check) (radeon->glCtx, atom); + if (dwords) { + if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { + radeon_print_state_atom(atom); + } + if (atom->emit) { + (*atom->emit)(radeon->glCtx, atom); + } else { + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_BATCH_TABLE(atom->cmd, dwords); + END_BATCH(); + } + atom->dirty = GL_FALSE; + } else { + if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { + fprintf(stderr, " skip state %s\n", + atom->name); + } + } + } + } + + COMMIT_BATCH(); +} + +void radeonEmitState(radeonContextPtr radeon) +{ + if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (radeon->vtbl.pre_emit_state) + radeon->vtbl.pre_emit_state(radeon); + + /* this code used to return here but now it emits zbs */ + if (radeon->cmdbuf.cs->cdw && !radeon->hw.is_dirty && !radeon->hw.all_dirty) + return; + + /* To avoid going across the entire set of states multiple times, just check + * for enough space for the case of emitting all state, and inline the + * radeonAllocCmdBuf code here without all the checks. + */ + rcommonEnsureCmdBufSpace(radeon, radeon->hw.max_state_size, __FUNCTION__); + + /* We always always emit zbs, this is due to a bug found by keithw in + the hardware and rediscovered after Erics changes by me. + if you ever touch this code make sure you emit zbs otherwise + you get tcl lockups on at least M7/7500 class of chips - airlied */ + /* special r100 case */ + // rmesa->hw.zbs.dirty=1; + + if (!radeon->cmdbuf.cs->cdw) { + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin reemit state\n"); + + radeonEmitAtoms(radeon, GL_FALSE); + } + + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin dirty state\n"); + + radeonEmitAtoms(radeon, GL_TRUE); + radeon->hw.is_dirty = GL_FALSE; + radeon->hw.all_dirty = GL_FALSE; + +} + /* Make sure all commands have been sent to the hardware and have * completed processing. */ -void radeon_common_finish(GLcontext * ctx) +void radeonFinish(GLcontext * ctx) { radeonContextPtr radeon = RADEON_CONTEXT(ctx); struct gl_framebuffer *fb = ctx->DrawBuffer; int i; + radeonFlush(ctx); + if (radeon->radeonScreen->kernel_mm) { for (i = 0; i < fb->_NumColorDrawBuffers; i++) { struct radeon_renderbuffer *rrb; @@ -611,7 +724,7 @@ int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller) rmesa->cmdbuf.flushing = 1; if (rmesa->cmdbuf.cs->cdw) { ret = radeon_cs_emit(rmesa->cmdbuf.cs); - rmesa->vtbl.set_all_dirty(rmesa->glCtx); + rmesa->hw.all_dirty = GL_TRUE; } radeon_cs_erase(rmesa->cmdbuf.cs); rmesa->cmdbuf.flushing = 0; @@ -650,20 +763,18 @@ void rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *ca } } -void rcommonInitCmdBuf(radeonContextPtr rmesa, int max_state_size) +void rcommonInitCmdBuf(radeonContextPtr rmesa) { GLuint size; /* Initialize command buffer */ size = 256 * driQueryOptioni(&rmesa->optionCache, "command_buffer_size"); - if (size < 2 * max_state_size) { - size = 2 * max_state_size + 65535; + if (size < 2 * rmesa->hw.max_state_size) { + size = 2 * rmesa->hw.max_state_size + 65535; } if (size > 64 * 256) size = 64 * 256; - size = 64 * 1024 / 4; - if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) { fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%zd\n", sizeof(drm_r300_cmd_header_t)); @@ -671,7 +782,7 @@ void rcommonInitCmdBuf(radeonContextPtr rmesa, int max_state_size) sizeof(drm_radeon_cmd_buffer_t)); fprintf(stderr, "Allocating %d bytes command buffer (max state is %d bytes)\n", - size * 4, max_state_size * 4); + size * 4, rmesa->hw.max_state_size * 4); } if (rmesa->radeonScreen->kernel_mm) { @@ -725,7 +836,7 @@ void rcommonBeginBatch(radeonContextPtr rmesa, int n, if (!rmesa->cmdbuf.cs->cdw && dostate) { if (RADEON_DEBUG & DEBUG_IOCTL) fprintf(stderr, "Reemit state after flush (from %s)\n", function); - rmesa->vtbl.emit_state(rmesa); + radeonEmitState(rmesa); } radeon_cs_begin(rmesa->cmdbuf.cs, n, file, function, line); } @@ -927,6 +1038,20 @@ void radeonCleanupContext(radeonContextPtr radeon) } } +/* Force the context `c' to be unbound from its buffer. + */ +GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv) +{ + radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; + + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "%s ctx %p\n", __FUNCTION__, + radeon->glCtx); + + return GL_TRUE; +} + + static void radeon_make_kernel_renderbuffer_current(radeonContextPtr radeon, GLframebuffer *draw) @@ -1810,7 +1935,7 @@ static void radeon_teximage( radeonTexObj* t = radeon_tex_obj(texObj); radeon_texture_image* image = get_radeon_texture_image(texImage); - rmesa->vtbl.flush_vertices(rmesa); + radeon_firevertices(rmesa); t->validated = GL_FALSE; @@ -1968,7 +2093,7 @@ static void radeon_texsubimage(GLcontext* ctx, int dims, int level, radeonTexObj* t = radeon_tex_obj(texObj); radeon_texture_image* image = get_radeon_texture_image(texImage); - rmesa->vtbl.flush_vertices(rmesa); + radeon_firevertices(rmesa); t->validated = GL_FALSE; pixels = _mesa_validate_pbo_teximage(ctx, dims, @@ -2373,7 +2498,7 @@ void radeonSpanRenderStart(GLcontext * ctx) radeonContextPtr rmesa = RADEON_CONTEXT(ctx); int i; - rmesa->vtbl.flush_vertices(rmesa); + radeon_firevertices(rmesa); for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { if (ctx->Texture.Unit[i]._ReallyEnabled) diff --git a/src/mesa/drivers/dri/radeon/common_misc.h b/src/mesa/drivers/dri/radeon/common_misc.h index 80b34aa6bd..ebf287a031 100644 --- a/src/mesa/drivers/dri/radeon/common_misc.h +++ b/src/mesa/drivers/dri/radeon/common_misc.h @@ -6,6 +6,7 @@ void radeonRecalcScissorRects(radeonContextPtr radeon); void radeonSetCliprects(radeonContextPtr radeon); void radeonUpdateScissor( GLcontext *ctx ); +void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h); void radeonWaitForIdleLocked(radeonContextPtr radeon); extern uint32_t radeonGetAge(radeonContextPtr radeon); @@ -26,6 +27,7 @@ GLboolean radeonInitContext(radeonContextPtr radeon, void *sharedContextPrivate); void radeonCleanupContext(radeonContextPtr radeon); +GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv); void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable); GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, @@ -132,6 +134,9 @@ void rcommon_flush_last_swtcl_prim(GLcontext *ctx); void *rcommonAllocDmaLowVerts(radeonContextPtr rmesa, int nverts, int vsize); +void radeonFlush(GLcontext *ctx); +void radeonFinish(GLcontext * ctx); +void radeonEmitState(radeonContextPtr radeon); static inline struct radeon_renderbuffer *radeon_get_depthbuffer(radeonContextPtr rmesa) { diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index 6fa01f4ded..b25f036c44 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -215,29 +215,17 @@ static void r100_get_lock(radeonContextPtr radeon) } } -static void r100_vtbl_flush(GLcontext *ctx) -{ - RADEON_FIREVERTICES(R100_CONTEXT(ctx)); -} - -static void r100_vtbl_set_all_dirty(GLcontext *ctx) -{ - r100ContextPtr rmesa = R100_CONTEXT(ctx); - rmesa->hw.all_dirty = GL_TRUE; -} - static void r100_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa) { } -static void r100_vtbl_emit_state(radeonContextPtr rmesa) +static void r100_vtbl_pre_emit_state(radeonContextPtr radeon) { - radeonEmitState((r100ContextPtr)rmesa); -} - -static void r100_vtbl_flush_vertices(radeonContextPtr rmesa) -{ - RADEON_FIREVERTICES(((r100ContextPtr)rmesa)); + r100ContextPtr rmesa = (r100ContextPtr)radeon; + + /* r100 always needs to emit ZBS to avoid TCL lockups */ + rmesa->hw.zbs.dirty = 1; + radeon->hw.is_dirty = 1; } @@ -245,13 +233,10 @@ static void r100_init_vtbl(radeonContextPtr radeon) { radeon->vtbl.get_lock = r100_get_lock; radeon->vtbl.update_viewport_offset = radeonUpdateViewportOffset; - radeon->vtbl.flush = r100_vtbl_flush; - radeon->vtbl.set_all_dirty = r100_vtbl_set_all_dirty; radeon->vtbl.update_draw_buffer = radeonUpdateDrawBuffer; radeon->vtbl.emit_cs_header = r100_vtbl_emit_cs_header; - radeon->vtbl.emit_state = r100_vtbl_emit_state; radeon->vtbl.swtcl_flush = r100_swtcl_flush; - radeon->vtbl.flush_vertices = r100_vtbl_flush_vertices; + radeon->vtbl.pre_emit_state = r100_vtbl_pre_emit_state; } /* Create the device specific context. @@ -344,7 +329,7 @@ radeonCreateContext( const __GLcontextModes *glVisual, DRI_CONF_TEXTURE_DEPTH_32 : DRI_CONF_TEXTURE_DEPTH_16; rmesa->radeon.swtcl.RenderIndex = ~0; - rmesa->hw.all_dirty = GL_TRUE; + rmesa->radeon.hw.all_dirty = GL_TRUE; /* Set the maximum texture size small enough that we can guarentee that * all texture units can bind a maximal texture and have all of them in @@ -503,7 +488,7 @@ void radeonDestroyContext( __DRIcontextPrivate *driContextPriv ) /* check if we're deleting the currently bound context */ if (rmesa == current) { - RADEON_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); _mesa_make_current(NULL, NULL, NULL); } @@ -553,16 +538,3 @@ void radeonDestroyContext( __DRIcontextPrivate *driContextPriv ) } } - -/* Force the context `c' to be unbound from its buffer. - */ -GLboolean -radeonUnbindContext( __DRIcontextPrivate *driContextPriv ) -{ - r100ContextPtr rmesa = (r100ContextPtr) driContextPriv->driverPrivate; - - if (RADEON_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *) rmesa->radeon.glCtx); - - return GL_TRUE; -} diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h index 358095a2b9..d05e47f16e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_context.h @@ -296,10 +296,7 @@ struct radeon_texture_state { #define SHN_SHININESS 1 #define SHN_STATE_SIZE 2 -struct radeon_hw_state { - /* Head of the linked list of state atoms. */ - struct radeon_state_atom atomlist; - +struct r100_hw_state { /* Hardware state, stored as cmdbuf commands: * -- Need to doublebuffer for * - eliding noop statechange loops? (except line stipple count) @@ -324,8 +321,6 @@ struct radeon_hw_state { struct radeon_state_atom glt; struct radeon_state_atom txr[3]; /* for NPOT */ - int max_state_size; /* Number of bytes necessary for a full state emit. */ - GLboolean is_dirty, all_dirty; }; @@ -354,14 +349,6 @@ struct radeon_tcl_info { struct radeon_bo *indexed_bo; -// struct radeon_dma_region indexed_verts; - struct radeon_dma_region obj; - struct radeon_dma_region rgba; - struct radeon_dma_region spec; - struct radeon_dma_region fog; - struct radeon_dma_region tex[RADEON_MAX_TEXTURE_UNITS]; - struct radeon_dma_region norm; - int elt_cmd_offset; /** Offset into the cmdbuf */ int elt_cmd_start; int elt_used; @@ -391,8 +378,6 @@ struct r100_swtcl_info { GLuint specoffset; GLboolean needproj; - - struct radeon_dma_region indexed_verts; }; @@ -411,17 +396,13 @@ struct r100_context { /* Driver and hardware state management */ - struct radeon_hw_state hw; + struct r100_hw_state hw; struct r100_state state; /* Vertex buffers */ struct radeon_ioctl ioctl; struct radeon_store store; - /* A full state emit as of the first state emit in the main store, in case - * the context is lost. - */ - struct radeon_store backup_store; /* TCL stuff */ @@ -455,7 +436,6 @@ struct r100_context { GLuint c_textureBytes; GLuint c_vertexBuffers; - GLboolean save_on_next_emit; }; #define R100_CONTEXT(ctx) ((r100ContextPtr)(ctx->DriverCtx)) diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index 45d750dbf4..d0cd9ce7a6 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -59,60 +59,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define RADEON_TIMEOUT 512 #define RADEON_IDLE_RETRY 16 -#define DEBUG_CMDBUF 1 - -static void radeonSaveHwState( r100ContextPtr rmesa ) -{ - struct radeon_state_atom *atom; - char * dest = rmesa->backup_store.cmd_buf; - - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "%s\n", __FUNCTION__); - - rmesa->backup_store.cmd_used = 0; - - foreach( atom, &rmesa->hw.atomlist ) { - if ( atom->check( rmesa->radeon.glCtx, 0 ) ) { - int size = atom->cmd_size * 4; - memcpy( dest, atom->cmd, size); - dest += size; - rmesa->backup_store.cmd_used += size; - if (RADEON_DEBUG & DEBUG_STATE) - radeon_print_state_atom( atom ); - } - } - - assert( rmesa->backup_store.cmd_used <= RADEON_CMD_BUF_SZ ); - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Returning to radeonEmitState\n"); -} - -/* At this point we were in FlushCmdBufLocked but we had lost our context, so - * we need to unwire our current cmdbuf, hook the one with the saved state in - * it, flush it, and then put the current one back. This is so commands at the - * start of a cmdbuf can rely on the state being kept from the previous one. - */ -static void radeonBackUpAndEmitLostStateLocked( r100ContextPtr rmesa ) -{ - GLuint nr_released_bufs; - struct radeon_store saved_store; - - if (rmesa->backup_store.cmd_used == 0) - return; - - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Emitting backup state on lost context\n"); - - rmesa->radeon.lost_context = GL_FALSE; - - nr_released_bufs = rmesa->radeon.dma.nr_released_bufs; - saved_store = rmesa->store; - rmesa->radeon.dma.nr_released_bufs = 0; - rmesa->store = rmesa->backup_store; - rcommonFlushCmdBufLocked( &rmesa->radeon, __FUNCTION__ ); - rmesa->radeon.dma.nr_released_bufs = nr_released_bufs; - rmesa->store = saved_store; -} /* ============================================================= * Kernel command buffer handling @@ -125,107 +71,33 @@ void radeonSetUpAtomList( r100ContextPtr rmesa ) { int i, mtu = rmesa->radeon.glCtx->Const.MaxTextureUnits; - make_empty_list(&rmesa->hw.atomlist); - rmesa->hw.atomlist.name = "atom-list"; + make_empty_list(&rmesa->radeon.hw.atomlist); + rmesa->radeon.hw.atomlist.name = "atom-list"; - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.ctx); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.set); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.lin); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.msk); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.vpt); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.tcl); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.msc); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.ctx); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.set); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.lin); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.msk); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.vpt); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.tcl); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.msc); for (i = 0; i < mtu; ++i) { - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.tex[i]); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.txr[i]); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.cube[i]); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.tex[i]); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.txr[i]); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.cube[i]); } - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.zbs); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.mtl); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.zbs); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.mtl); for (i = 0; i < 3 + mtu; ++i) - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.mat[i]); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.mat[i]); for (i = 0; i < 8; ++i) - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.lit[i]); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.lit[i]); for (i = 0; i < 6; ++i) - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.ucp[i]); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.eye); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.grd); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.fog); - insert_at_tail(&rmesa->hw.atomlist, &rmesa->hw.glt); -} - -static INLINE void radeonEmitAtoms(r100ContextPtr r100, GLboolean dirty) -{ - BATCH_LOCALS(&r100->radeon); - struct radeon_state_atom *atom; - int dwords; - - /* Emit actual atoms */ - foreach(atom, &r100->hw.atomlist) { - if ((atom->dirty || r100->hw.all_dirty) == dirty) { - dwords = (*atom->check) (r100->radeon.glCtx, atom); - if (dwords) { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - radeon_print_state_atom(atom); - } - if (atom->emit) { - (*atom->emit)(r100->radeon.glCtx, atom); - } else { - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(atom->cmd, dwords); - END_BATCH(); - } - atom->dirty = GL_FALSE; - } else { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - fprintf(stderr, " skip state %s\n", - atom->name); - } - } - } - } - - COMMIT_BATCH(); -} - -void radeonEmitState( r100ContextPtr rmesa ) -{ - if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (rmesa->save_on_next_emit) { - radeonSaveHwState(rmesa); - rmesa->save_on_next_emit = GL_FALSE; - } - - /* this code used to return here but now it emits zbs */ - - /* To avoid going across the entire set of states multiple times, just check - * for enough space for the case of emitting all state, and inline the - * radeonAllocCmdBuf code here without all the checks. - */ - rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->hw.max_state_size, __FUNCTION__); - - /* We always always emit zbs, this is due to a bug found by keithw in - the hardware and rediscovered after Erics changes by me. - if you ever touch this code make sure you emit zbs otherwise - you get tcl lockups on at least M7/7500 class of chips - airlied */ - rmesa->hw.zbs.dirty=1; - - if (!rmesa->radeon.cmdbuf.cs->cdw) { - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin reemit state\n"); - - radeonEmitAtoms(rmesa, GL_FALSE); - } - - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin dirty state\n"); - - radeonEmitAtoms(rmesa, GL_TRUE); - rmesa->hw.is_dirty = GL_FALSE; - rmesa->hw.all_dirty = GL_FALSE; - + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.ucp[i]); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.eye); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.grd); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.fog); + insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.glt); } /* Fire a section of the retained (indexed_verts) buffer as a regular @@ -240,7 +112,7 @@ extern void radeonEmitVbufPrim( r100ContextPtr rmesa, assert(!(primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND)); - radeonEmitState( rmesa ); + radeonEmitState(&rmesa->radeon); #if RADEON_OLD_PACKETS BEGIN_BATCH(8); @@ -318,7 +190,7 @@ GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa, assert((primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND)); - radeonEmitState( rmesa ); + radeonEmitState(&rmesa->radeon); rmesa->tcl.elt_cmd_start = rmesa->radeon.cmdbuf.cs->cdw; @@ -679,35 +551,9 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask ) } UNLOCK_HARDWARE( &rmesa->radeon ); - rmesa->hw.all_dirty = GL_TRUE; -} - -void radeonFlush( GLcontext *ctx ) -{ - r100ContextPtr rmesa = R100_CONTEXT( ctx ); - - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (rmesa->radeon.dma.flush) - rmesa->radeon.dma.flush( ctx ); - - radeonEmitState( rmesa ); - - if (rmesa->radeon.cmdbuf.cs->cdw) - rcommonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ ); -} - -/* Make sure all commands have been sent to the hardware and have - * completed processing. - */ -void radeonFinish( GLcontext *ctx ) -{ - radeonFlush( ctx ); - radeon_common_finish(ctx); + rmesa->radeon.hw.all_dirty = GL_TRUE; } - void radeonInitIoctlFuncs( GLcontext *ctx ) { ctx->Driver.Clear = radeonClear; diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h index 4e93804646..6d616bf804 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h @@ -40,7 +40,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_lock.h" #include "radeon_cs_legacy.h" -extern void radeonEmitState( r100ContextPtr rmesa ); extern void radeonEmitVertexAOS( r100ContextPtr rmesa, GLuint vertex_size, struct radeon_bo *bo, @@ -104,7 +103,7 @@ do { \ do { \ RADEON_NEWPRIM( rmesa ); \ rmesa->hw.ATOM.dirty = GL_TRUE; \ - rmesa->hw.is_dirty = GL_TRUE; \ + rmesa->radeon.hw.is_dirty = GL_TRUE; \ } while (0) #define RADEON_DB_STATE( ATOM ) \ @@ -118,7 +117,7 @@ static INLINE int RADEON_DB_STATECHANGE(r100ContextPtr rmesa, GLuint *tmp; RADEON_NEWPRIM( rmesa ); atom->dirty = GL_TRUE; - rmesa->hw.is_dirty = GL_TRUE; + rmesa->radeon.hw.is_dirty = GL_TRUE; tmp = atom->cmd; atom->cmd = atom->lastcmd; atom->lastcmd = tmp; @@ -128,15 +127,6 @@ static INLINE int RADEON_DB_STATECHANGE(r100ContextPtr rmesa, return 0; } -/* Fire the buffered vertices no matter what. - */ -#define RADEON_FIREVERTICES( rmesa ) \ -do { \ - if (rmesa->radeon.cmdbuf.cs->cdw || rmesa->radeon.dma.flush ) { \ - radeonFlush( rmesa->radeon.glCtx ); \ - } \ -} while (0) - /* Command lengths. Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ * are available, you will also be adding an rmesa->state.max_state_size because * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts. diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index 497582ee23..7397218fb6 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -1523,7 +1523,7 @@ const struct __DriverAPIRec driDriverAPI = { .DestroyBuffer = radeonDestroyBuffer, .SwapBuffers = radeonSwapBuffers, .MakeCurrent = radeonMakeCurrent, - .UnbindContext = r200UnbindContext, + .UnbindContext = radeonUnbindContext, .GetSwapInfo = getSwapInfo, .GetDrawableMSC = driDrawableGetMSC32, .WaitForMSC = driWaitForMSC32, diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c index 1d683e5b3a..a9ec1d51c7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.c +++ b/src/mesa/drivers/dri/radeon/radeon_state.c @@ -47,6 +47,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "swrast_setup/swrast_setup.h" #include "radeon_context.h" +#include "common_cmdbuf.h" #include "radeon_ioctl.h" #include "radeon_state.h" #include "radeon_tcl.h" @@ -406,23 +407,6 @@ static void radeonFogfv( GLcontext *ctx, GLenum pname, const GLfloat *param ) } } - -/* ============================================================= - * Scissoring - */ -static void radeonScissor( GLcontext *ctx, - GLint x, GLint y, GLsizei w, GLsizei h ) -{ - r100ContextPtr rmesa = R100_CONTEXT(ctx); - - if ( ctx->Scissor.Enabled ) { - RADEON_FIREVERTICES( rmesa ); /* don't pipeline cliprect changes */ - radeonUpdateScissor( ctx ); - } - -} - - /* ============================================================= * Culling */ @@ -567,7 +551,7 @@ static void radeonPolygonStipple( GLcontext *ctx, const GLubyte *mask ) /* TODO: push this into cmd mechanism */ - RADEON_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); LOCK_HARDWARE( &rmesa->radeon ); /* FIXME: Use window x,y offsets into stipple RAM. @@ -1415,7 +1399,7 @@ void radeonUpdateWindow( GLcontext *ctx ) float_ui32_type sz = { v[MAT_SZ] * rmesa->radeon.state.depth.scale }; float_ui32_type tz = { v[MAT_TZ] * rmesa->radeon.state.depth.scale }; - RADEON_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); RADEON_STATECHANGE( rmesa, vpt ); rmesa->hw.vpt.cmd[VPT_SE_VPORT_XSCALE] = sx.ui32; @@ -1561,7 +1545,7 @@ static void radeonDrawBuffer( GLcontext *ctx, GLenum mode ) fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr( mode )); - RADEON_FIREVERTICES(rmesa); /* don't pipeline cliprect changes */ + radeon_firevertices(&rmesa->radeon); /* don't pipeline cliprect changes */ if (ctx->DrawBuffer->_NumColorDrawBuffers != 1) { /* 0 (GL_NONE) buffers or multiple color drawing buffers */ @@ -1843,7 +1827,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state ) } case GL_SCISSOR_TEST: - RADEON_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); rmesa->radeon.state.scissor.enabled = state; radeonUpdateScissor( ctx ); break; diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index f12d9bdb50..1cfb539e71 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -171,8 +171,8 @@ void radeonPrintDirty( r100ContextPtr rmesa, const char *msg ) fprintf(stderr, msg); fprintf(stderr, ": "); - foreach(l, &rmesa->hw.atomlist) { - if (l->dirty || rmesa->hw.all_dirty) + foreach(l, &rmesa->radeon.hw.atomlist) { + if (l->dirty || rmesa->radeon.hw.all_dirty) fprintf(stderr, "%s, ", l->name); } @@ -512,7 +512,7 @@ void radeonInitState( r100ContextPtr rmesa ) drawPitch = rmesa->radeon.radeonScreen->frontPitch; } - rmesa->hw.max_state_size = 0; + rmesa->radeon.hw.max_state_size = 0; #define ALLOC_STATE_IDX( ATOM, CHK, SZ, NM, FLAG, IDX ) \ do { \ @@ -524,7 +524,7 @@ void radeonInitState( r100ContextPtr rmesa ) rmesa->hw.ATOM.check = check_##CHK; \ rmesa->hw.ATOM.dirty = GL_TRUE; \ rmesa->hw.ATOM.idx = IDX; \ - rmesa->hw.max_state_size += SZ * sizeof(int); \ + rmesa->radeon.hw.max_state_size += SZ * sizeof(int); \ } while (0) #define ALLOC_STATE( ATOM, CHK, SZ, NM, FLAG ) \ @@ -932,7 +932,7 @@ void radeonInitState( r100ContextPtr rmesa ) rmesa->hw.eye.cmd[EYE_Z] = IEEE_ONE; rmesa->hw.eye.cmd[EYE_RESCALE_FACTOR] = IEEE_ONE; - rmesa->hw.all_dirty = GL_TRUE; + rmesa->radeon.hw.all_dirty = GL_TRUE; - rcommonInitCmdBuf(&rmesa->radeon, rmesa->hw.max_state_size); + rcommonInitCmdBuf(&rmesa->radeon); } diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 615621dd98..995ab2099e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -285,11 +285,11 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r100ContextPtr rmesa = R100_CONTEXT(ctx); rcommonEnsureCmdBufSpace(&rmesa->radeon, - rmesa->hw.max_state_size + (12*sizeof(int)), + rmesa->radeon.hw.max_state_size + (12*sizeof(int)), __FUNCTION__); - radeonEmitState(rmesa); + radeonEmitState(&rmesa->radeon); radeonEmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, rmesa->radeon.dma.current, @@ -372,9 +372,6 @@ static GLboolean radeon_run_render( GLcontext *ctx, tnl_render_func *tab = TAG(render_tab_verts); GLuint i; - if (rmesa->swtcl.indexed_verts.buf) - RELEASE_ELT_VERTS(); - if (rmesa->radeon.swtcl.RenderIndex != 0 || !radeon_dma_validate_render( ctx, VB )) return GL_TRUE; @@ -750,7 +747,7 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) if (mode) { rmesa->radeon.Fallback |= bit; if (oldfallback == 0) { - RADEON_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_TRUE ); _swsetup_Wakeup( ctx ); rmesa->radeon.swtcl.RenderIndex = ~0; @@ -831,7 +828,4 @@ void radeonDestroySwtcl( GLcontext *ctx ) { r100ContextPtr rmesa = R100_CONTEXT(ctx); - // if (rmesa->swtcl.indexed_verts.buf) - // radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, - // __FUNCTION__ ); } diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index daa88f43c7..7281800ce2 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -126,7 +126,7 @@ static GLboolean discrete_prim[0x10] = { #define RESET_STIPPLE() do { \ RADEON_STATECHANGE( rmesa, lin ); \ - radeonEmitState( rmesa ); \ + radeonEmitState(&rmesa->radeon); \ } while (0) #define AUTO_STIPPLE( mode ) do { \ @@ -137,7 +137,7 @@ static GLboolean discrete_prim[0x10] = { else \ rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &= \ ~RADEON_LINE_PATTERN_AUTO_RESET; \ - radeonEmitState( rmesa ); \ + radeonEmitState(&rmesa->radeon); \ } while (0) @@ -149,7 +149,7 @@ static GLushort *radeonAllocElts( r100ContextPtr rmesa, GLuint nr ) if (rmesa->radeon.dma.flush) rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); - rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->hw.max_state_size + ELTS_BUFSZ(nr) + + rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->radeon.hw.max_state_size + ELTS_BUFSZ(nr) + AOS_BUFSZ(rmesa->tcl.nr_aos_components), __FUNCTION__); radeonEmitAOS( rmesa, @@ -178,7 +178,7 @@ static void radeonEmitPrim( GLcontext *ctx, rcommonEnsureCmdBufSpace( &rmesa->radeon, AOS_BUFSZ(rmesa->tcl.nr_aos_components) + - rmesa->hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); + rmesa->radeon.hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); radeonEmitAOS( rmesa, rmesa->tcl.nr_aos_components, diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c index a916b63a27..e1b988bf4d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex.c @@ -381,7 +381,7 @@ static void radeonDeleteTexture( GLcontext *ctx, } if ( rmesa ) { - RADEON_FIREVERTICES( rmesa ); + radeon_firevertices(&rmesa->radeon); for ( i = 0 ; i < rmesa->radeon.glCtx->Const.MaxTextureUnits ; i++ ) { if ( t == rmesa->state.texture.unit[i].texobj ) { rmesa->state.texture.unit[i].texobj = NULL; -- cgit v1.2.3 From 23d3559bd4ece1fcab5513ebdaa38600d6654374 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Thu, 12 Feb 2009 22:38:10 +1000 Subject: radeon: renaming and headers cleanup --- src/mesa/drivers/dri/radeon/Makefile | 17 +- src/mesa/drivers/dri/radeon/common_cmdbuf.h | 143 -- src/mesa/drivers/dri/radeon/common_context.h | 502 ---- src/mesa/drivers/dri/radeon/common_lock.c | 126 - src/mesa/drivers/dri/radeon/common_lock.h | 120 - src/mesa/drivers/dri/radeon/common_misc.c | 2715 -------------------- src/mesa/drivers/dri/radeon/common_misc.h | 166 -- src/mesa/drivers/dri/radeon/radeon_bo_legacy.c | 2 +- src/mesa/drivers/dri/radeon/radeon_cmdbuf.h | 143 ++ src/mesa/drivers/dri/radeon/radeon_common.c | 843 ++++++ src/mesa/drivers/dri/radeon/radeon_common.h | 56 + .../drivers/dri/radeon/radeon_common_context.c | 584 +++++ .../drivers/dri/radeon/radeon_common_context.h | 513 ++++ src/mesa/drivers/dri/radeon/radeon_context.c | 42 +- src/mesa/drivers/dri/radeon/radeon_context.h | 3 +- src/mesa/drivers/dri/radeon/radeon_cs_legacy.c | 4 +- src/mesa/drivers/dri/radeon/radeon_cs_legacy.h | 2 +- src/mesa/drivers/dri/radeon/radeon_dma.c | 323 +++ src/mesa/drivers/dri/radeon/radeon_dma.h | 51 + src/mesa/drivers/dri/radeon/radeon_ioctl.c | 2 +- src/mesa/drivers/dri/radeon/radeon_lock.c | 124 + src/mesa/drivers/dri/radeon/radeon_lock.h | 81 +- src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h | 2 +- src/mesa/drivers/dri/radeon/radeon_screen.c | 4 +- src/mesa/drivers/dri/radeon/radeon_span.c | 226 +- src/mesa/drivers/dri/radeon/radeon_state.c | 1 - src/mesa/drivers/dri/radeon/radeon_state_init.c | 1 - src/mesa/drivers/dri/radeon/radeon_swtcl.c | 1 - src/mesa/drivers/dri/radeon/radeon_tcl.c | 2 +- src/mesa/drivers/dri/radeon/radeon_texture.c | 849 ++++++ src/mesa/drivers/dri/radeon/radeon_texture.h | 108 + 31 files changed, 3916 insertions(+), 3840 deletions(-) delete mode 100644 src/mesa/drivers/dri/radeon/common_cmdbuf.h delete mode 100644 src/mesa/drivers/dri/radeon/common_context.h delete mode 100644 src/mesa/drivers/dri/radeon/common_lock.c delete mode 100644 src/mesa/drivers/dri/radeon/common_lock.h delete mode 100644 src/mesa/drivers/dri/radeon/common_misc.c delete mode 100644 src/mesa/drivers/dri/radeon/common_misc.h create mode 100644 src/mesa/drivers/dri/radeon/radeon_cmdbuf.h create mode 100644 src/mesa/drivers/dri/radeon/radeon_common.c create mode 100644 src/mesa/drivers/dri/radeon/radeon_common.h create mode 100644 src/mesa/drivers/dri/radeon/radeon_common_context.c create mode 100644 src/mesa/drivers/dri/radeon/radeon_common_context.h create mode 100644 src/mesa/drivers/dri/radeon/radeon_dma.c create mode 100644 src/mesa/drivers/dri/radeon/radeon_dma.h create mode 100644 src/mesa/drivers/dri/radeon/radeon_lock.c create mode 100644 src/mesa/drivers/dri/radeon/radeon_texture.c create mode 100644 src/mesa/drivers/dri/radeon/radeon_texture.h (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile index 2d00f521a6..16bc063baf 100644 --- a/src/mesa/drivers/dri/radeon/Makefile +++ b/src/mesa/drivers/dri/radeon/Makefile @@ -8,6 +8,16 @@ LIBNAME = radeon_dri.so MINIGLX_SOURCES = server/radeon_dri.c +RADEON_COMMON_SOURCES = \ + radeon_texture.c \ + radeon_common_context.c \ + radeon_dma.c \ + radeon_lock.c \ + radeon_bo_legacy.c \ + radeon_cs_legacy.c \ + radeon_mipmap_tree.c \ + radeon_span.c + DRIVER_SOURCES = \ radeon_context.c \ radeon_ioctl.c \ @@ -18,14 +28,9 @@ DRIVER_SOURCES = \ radeon_texstate.c \ radeon_tcl.c \ radeon_swtcl.c \ - radeon_span.c \ radeon_maos.c \ radeon_sanity.c \ - common_lock.c \ - common_misc.c \ - radeon_bo_legacy.c \ - radeon_cs_legacy.c \ - radeon_mipmap_tree.c + $(RADEON_COMMON_SOURCES) C_SOURCES = \ $(COMMON_SOURCES) \ diff --git a/src/mesa/drivers/dri/radeon/common_cmdbuf.h b/src/mesa/drivers/dri/radeon/common_cmdbuf.h deleted file mode 100644 index 5526934209..0000000000 --- a/src/mesa/drivers/dri/radeon/common_cmdbuf.h +++ /dev/null @@ -1,143 +0,0 @@ -#ifndef COMMON_CMDBUF_H -#define COMMON_CMDBUF_H - -#include "radeon_cs.h" - -void rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller); -int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller); -int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller); -void rcommonInitCmdBuf(radeonContextPtr rmesa); -void rcommonDestroyCmdBuf(radeonContextPtr rmesa); - -void rcommonBeginBatch(radeonContextPtr rmesa, - int n, - int dostate, - const char *file, - const char *function, - int line); - -#define RADEON_CP_PACKET3_NOP 0xC0001000 -#define RADEON_CP_PACKET3_NEXT_CHAR 0xC0001900 -#define RADEON_CP_PACKET3_PLY_NEXTSCAN 0xC0001D00 -#define RADEON_CP_PACKET3_SET_SCISSORS 0xC0001E00 -#define RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM 0xC0002300 -#define RADEON_CP_PACKET3_LOAD_MICROCODE 0xC0002400 -#define RADEON_CP_PACKET3_WAIT_FOR_IDLE 0xC0002600 -#define RADEON_CP_PACKET3_3D_DRAW_VBUF 0xC0002800 -#define RADEON_CP_PACKET3_3D_DRAW_IMMD 0xC0002900 -#define RADEON_CP_PACKET3_3D_DRAW_INDX 0xC0002A00 -#define RADEON_CP_PACKET3_LOAD_PALETTE 0xC0002C00 -#define RADEON_CP_PACKET3_3D_LOAD_VBPNTR 0xC0002F00 -#define RADEON_CP_PACKET3_CNTL_PAINT 0xC0009100 -#define RADEON_CP_PACKET3_CNTL_BITBLT 0xC0009200 -#define RADEON_CP_PACKET3_CNTL_SMALLTEXT 0xC0009300 -#define RADEON_CP_PACKET3_CNTL_HOSTDATA_BLT 0xC0009400 -#define RADEON_CP_PACKET3_CNTL_POLYLINE 0xC0009500 -#define RADEON_CP_PACKET3_CNTL_POLYSCANLINES 0xC0009800 -#define RADEON_CP_PACKET3_CNTL_PAINT_MULTI 0xC0009A00 -#define RADEON_CP_PACKET3_CNTL_BITBLT_MULTI 0xC0009B00 -#define RADEON_CP_PACKET3_CNTL_TRANS_BITBLT 0xC0009C00 - -#define CP_PACKET2 (2 << 30) -#define CP_PACKET0(reg, n) (RADEON_CP_PACKET0 | ((n)<<16) | ((reg)>>2)) -#define CP_PACKET0_ONE(reg, n) (RADEON_CP_PACKET0 | RADEON_CP_PACKET0_ONE_REG_WR | ((n)<<16) | ((reg)>>2)) -#define CP_PACKET3( pkt, n ) \ - (RADEON_CP_PACKET3 | (pkt) | ((n) << 16)) - -/** - * Every function writing to the command buffer needs to declare this - * to get the necessary local variables. - */ -#define BATCH_LOCALS(rmesa) \ - const radeonContextPtr b_l_rmesa = rmesa - -/** - * Prepare writing n dwords to the command buffer, - * including producing any necessary state emits on buffer wraparound. - */ -#define BEGIN_BATCH(n) rcommonBeginBatch(b_l_rmesa, n, 1, __FILE__, __FUNCTION__, __LINE__) - -/** - * Same as BEGIN_BATCH, but do not cause automatic state emits. - */ -#define BEGIN_BATCH_NO_AUTOSTATE(n) rcommonBeginBatch(b_l_rmesa, n, 0, __FILE__, __FUNCTION__, __LINE__) - -/** - * Write one dword to the command buffer. - */ -#define OUT_BATCH(data) \ - do { \ - radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, data);\ - } while(0) - -/** - * Write a relocated dword to the command buffer. - */ -#define OUT_BATCH_RELOC(data, bo, offset, rd, wd, flags) \ - do { \ - if (0 && offset) { \ - fprintf(stderr, "(%s:%s:%d) offset : %d\n", \ - __FILE__, __FUNCTION__, __LINE__, offset); \ - } \ - radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, offset); \ - radeon_cs_write_reloc(b_l_rmesa->cmdbuf.cs, \ - bo, rd, wd, flags); \ - if (!b_l_rmesa->radeonScreen->kernel_mm) \ - b_l_rmesa->cmdbuf.cs->section_cdw += 2; \ - } while(0) - - -/** - * Write n dwords from ptr to the command buffer. - */ -#define OUT_BATCH_TABLE(ptr,n) \ - do { \ - int _i; \ - for (_i=0; _i < n; _i++) {\ - radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, ptr[_i]);\ - }\ - } while(0) - -/** - * Finish writing dwords to the command buffer. - * The number of (direct or indirect) OUT_BATCH calls between the previous - * BEGIN_BATCH and END_BATCH must match the number specified at BEGIN_BATCH time. - */ -#define END_BATCH() \ - do { \ - radeon_cs_end(b_l_rmesa->cmdbuf.cs, __FILE__, __FUNCTION__, __LINE__);\ - } while(0) - -/** - * After the last END_BATCH() of rendering, this indicates that flushing - * the command buffer now is okay. - */ -#define COMMIT_BATCH() \ - do { \ - } while(0) - - -/** Single register write to command buffer; requires 2 dwords. */ -#define OUT_BATCH_REGVAL(reg, val) \ - OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), 1)); \ - OUT_BATCH((val)) - -/** Continuous register range write to command buffer; requires 1 dword, - * expects count dwords afterwards for register contents. */ -#define OUT_BATCH_REGSEQ(reg, count) \ - OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), (count))); - -/** Write a 32 bit float to the ring; requires 1 dword. */ -#define OUT_BATCH_FLOAT32(f) \ - OUT_BATCH(radeonPackFloat32((f))); - - -/* Fire the buffered vertices no matter what. - */ -static INLINE void radeon_firevertices(radeonContextPtr radeon) -{ - if (radeon->cmdbuf.cs->cdw || radeon->dma.flush ) - radeonFlush(radeon->glCtx); -} - -#endif diff --git a/src/mesa/drivers/dri/radeon/common_context.h b/src/mesa/drivers/dri/radeon/common_context.h deleted file mode 100644 index c3a445e0d2..0000000000 --- a/src/mesa/drivers/dri/radeon/common_context.h +++ /dev/null @@ -1,502 +0,0 @@ - -#ifndef COMMON_CONTEXT_H -#define COMMON_CONTEXT_H - -#include "main/mm.h" -#include "math/m_vector.h" -#include "texmem.h" -#include "tnl/t_context.h" -#include "main/colormac.h" - -#include "radeon_screen.h" -#include "radeon_drm.h" -#include "dri_util.h" -#include "tnl/t_vertex.h" - -/* This union is used to avoid warnings/miscompilation - with float to uint32_t casts due to strict-aliasing */ -typedef union { GLfloat f; uint32_t ui32; } float_ui32_type; - -struct radeon_context; -typedef struct radeon_context radeonContextRec; -typedef struct radeon_context *radeonContextPtr; - - -#define TEX_0 0x1 -#define TEX_1 0x2 -#define TEX_2 0x4 -#define TEX_3 0x8 -#define TEX_4 0x10 -#define TEX_5 0x20 - -/* Rasterizing fallbacks */ -/* See correponding strings in r200_swtcl.c */ -#define RADEON_FALLBACK_TEXTURE 0x0001 -#define RADEON_FALLBACK_DRAW_BUFFER 0x0002 -#define RADEON_FALLBACK_STENCIL 0x0004 -#define RADEON_FALLBACK_RENDER_MODE 0x0008 -#define RADEON_FALLBACK_BLEND_EQ 0x0010 -#define RADEON_FALLBACK_BLEND_FUNC 0x0020 -#define RADEON_FALLBACK_DISABLE 0x0040 -#define RADEON_FALLBACK_BORDER_MODE 0x0080 - -#define R200_FALLBACK_TEXTURE 0x01 -#define R200_FALLBACK_DRAW_BUFFER 0x02 -#define R200_FALLBACK_STENCIL 0x04 -#define R200_FALLBACK_RENDER_MODE 0x08 -#define R200_FALLBACK_DISABLE 0x10 -#define R200_FALLBACK_BORDER_MODE 0x20 - -#define RADEON_TCL_FALLBACK_RASTER 0x1 /* rasterization */ -#define RADEON_TCL_FALLBACK_UNFILLED 0x2 /* unfilled tris */ -#define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE 0x4 /* twoside tris */ -#define RADEON_TCL_FALLBACK_MATERIAL 0x8 /* material in vb */ -#define RADEON_TCL_FALLBACK_TEXGEN_0 0x10 /* texgen, unit 0 */ -#define RADEON_TCL_FALLBACK_TEXGEN_1 0x20 /* texgen, unit 1 */ -#define RADEON_TCL_FALLBACK_TEXGEN_2 0x40 /* texgen, unit 2 */ -#define RADEON_TCL_FALLBACK_TCL_DISABLE 0x80 /* user disable */ -#define RADEON_TCL_FALLBACK_FOGCOORDSPEC 0x100 /* fogcoord, sep. spec light */ - -/* The blit width for texture uploads - */ -#define BLIT_WIDTH_BYTES 1024 - -/* Use the templated vertex format: - */ -#define COLOR_IS_RGBA -#define TAG(x) radeon##x -#include "tnl_dd/t_dd_vertex.h" -#undef TAG - -struct radeon_colorbuffer_state { - GLuint clear; - int roundEnable; - struct radeon_renderbuffer *rrb; -}; - -struct radeon_depthbuffer_state { - GLuint clear; - GLfloat scale; - struct radeon_renderbuffer *rrb; -}; - -struct radeon_scissor_state { - drm_clip_rect_t rect; - GLboolean enabled; - - GLuint numClipRects; /* Cliprects active */ - GLuint numAllocedClipRects; /* Cliprects available */ - drm_clip_rect_t *pClipRects; -}; - -struct radeon_stencilbuffer_state { - GLboolean hwBuffer; - GLuint clear; /* rb3d_stencilrefmask value */ -}; - -struct radeon_stipple_state { - GLuint mask[32]; -}; - -struct radeon_state_atom { - struct radeon_state_atom *next, *prev; - const char *name; /* for debug */ - int cmd_size; /* size in bytes */ - GLuint idx; - GLuint is_tcl; - GLuint *cmd; /* one or more cmd's */ - GLuint *lastcmd; /* one or more cmd's */ - GLboolean dirty; /* dirty-mark in emit_state_list */ - int (*check) (GLcontext *, struct radeon_state_atom *atom); /* is this state active? */ - void (*emit) (GLcontext *, struct radeon_state_atom *atom); -}; - -struct radeon_hw_state { - /* Head of the linked list of state atoms. */ - struct radeon_state_atom atomlist; - int max_state_size; /* Number of bytes necessary for a full state emit. */ - GLboolean is_dirty, all_dirty; -}; - - -/* Texture related */ -typedef struct _radeon_texture_image radeon_texture_image; - -struct _radeon_texture_image { - struct gl_texture_image base; - - /** - * If mt != 0, the image is stored in hardware format in the - * given mipmap tree. In this case, base.Data may point into the - * mapping of the buffer object that contains the mipmap tree. - * - * If mt == 0, the image is stored in normal memory pointed to - * by base.Data. - */ - struct _radeon_mipmap_tree *mt; - struct radeon_bo *bo; - - int mtlevel; /** if mt != 0, this is the image's level in the mipmap tree */ - int mtface; /** if mt != 0, this is the image's face in the mipmap tree */ -}; - - -static INLINE radeon_texture_image *get_radeon_texture_image(struct gl_texture_image *image) -{ - return (radeon_texture_image*)image; -} - - -typedef struct radeon_tex_obj radeonTexObj, *radeonTexObjPtr; - -#define RADEON_TXO_MICRO_TILE (1 << 3) - -/* Texture object in locally shared texture space. - */ -struct radeon_tex_obj { - // driTextureObject base; - struct gl_texture_object base; - struct _radeon_mipmap_tree *mt; - - /** - * This is true if we've verified that the mipmap tree above is complete - * and so on. - */ - GLboolean validated; - - GLuint override_offset; - GLboolean image_override; /* Image overridden by GLX_EXT_tfp */ - GLuint tile_bits; /* hw texture tile bits used on this texture */ - struct radeon_bo *bo; - - GLuint bufAddr; /* Offset to start of locally - shared texture block */ - - GLuint dirty_state; /* Flags (1 per texunit) for - whether or not this texobj - has dirty hardware state - (pp_*) that needs to be - brought into the - texunit. */ - - drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS]; - /* Six, for the cube faces */ - - - - GLuint pp_txfilter; /* hardware register values */ - GLuint pp_txformat; - GLuint pp_txformat_x; - GLuint pp_txoffset; /* Image location in texmem. - All cube faces follow. */ - GLuint pp_txsize; /* npot only */ - GLuint pp_txpitch; /* npot only */ - GLuint pp_border_color; - GLuint pp_cubic_faces; /* cube face 1,2,3,4 log2 sizes */ - - GLuint pp_txfilter_1; /* r300 */ - - GLboolean border_fallback; - - -}; - -static INLINE radeonTexObj* radeon_tex_obj(struct gl_texture_object *texObj) -{ - return (radeonTexObj*)texObj; -} - -/* Need refcounting on dma buffers: - */ -struct radeon_dma_buffer { - int refcount; /* the number of retained regions in buf */ - drmBufPtr buf; -}; - -struct radeon_aos { - struct radeon_bo *bo; /** Buffer object where vertex data is stored */ - int offset; /** Offset into buffer object, in bytes */ - int components; /** Number of components per vertex */ - int stride; /** Stride in dwords (may be 0 for repeating) */ - int count; /** Number of vertices */ -}; - -struct radeon_dma { - /* Active dma region. Allocations for vertices and retained - * regions come from here. Also used for emitting random vertices, - * these may be flushed by calling flush_current(); - */ - struct radeon_bo *current; /** Buffer that DMA memory is allocated from */ - int current_used; /** Number of bytes allocated and forgotten about */ - int current_vertexptr; /** End of active vertex region */ - - /** - * If current_vertexptr != current_used then flush must be non-zero. - * flush must be called before non-active vertex allocations can be - * performed. - */ - void (*flush) (GLcontext *); - - /* Number of "in-flight" DMA buffers, i.e. the number of buffers - * for which a DISCARD command is currently queued in the command buffer -. - */ - GLuint nr_released_bufs; -}; - -/* radeon_swtcl.c - */ -struct radeon_swtcl_info { - - GLuint RenderIndex; - GLuint vertex_size; - GLubyte *verts; - - /* Fallback rasterization functions - */ - GLuint hw_primitive; - GLenum render_primitive; - GLuint numverts; - - struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; - GLuint vertex_attr_count; - -}; - -struct radeon_ioctl { - GLuint vertex_offset; - struct radeon_bo *bo; - GLuint vertex_size; -}; - -#define RADEON_MAX_PRIMS 64 - -struct radeon_prim { - GLuint start; - GLuint end; - GLuint prim; -}; - -static INLINE GLuint radeonPackColor(GLuint cpp, - GLubyte r, GLubyte g, - GLubyte b, GLubyte a) -{ - switch (cpp) { - case 2: - return PACK_COLOR_565(r, g, b); - case 4: - return PACK_COLOR_8888(a, r, g, b); - default: - return 0; - } -} - -#define MAX_CMD_BUF_SZ (16*1024) - -#define MAX_DMA_BUF_SZ (64*1024) - -struct radeon_store { - GLuint statenr; - GLuint primnr; - char cmd_buf[MAX_CMD_BUF_SZ]; - int cmd_used; - int elts_start; -}; - -struct radeon_dri_mirror { - __DRIcontextPrivate *context; /* DRI context */ - __DRIscreenPrivate *screen; /* DRI screen */ - - /** - * DRI drawable bound to this context for drawing. - */ - __DRIdrawablePrivate *drawable; - - /** - * DRI drawable bound to this context for reading. - */ - __DRIdrawablePrivate *readable; - - drm_context_t hwContext; - drm_hw_lock_t *hwLock; - int fd; - int drmMinor; -}; - -#define DEBUG_TEXTURE 0x001 -#define DEBUG_STATE 0x002 -#define DEBUG_IOCTL 0x004 -#define DEBUG_PRIMS 0x008 -#define DEBUG_VERTS 0x010 -#define DEBUG_FALLBACKS 0x020 -#define DEBUG_VFMT 0x040 -#define DEBUG_CODEGEN 0x080 -#define DEBUG_VERBOSE 0x100 -#define DEBUG_DRI 0x200 -#define DEBUG_DMA 0x400 -#define DEBUG_SANITY 0x800 -#define DEBUG_SYNC 0x1000 -#define DEBUG_PIXEL 0x2000 -#define DEBUG_MEMORY 0x4000 - - - -typedef void (*radeon_tri_func) (radeonContextPtr, - radeonVertex *, - radeonVertex *, radeonVertex *); - -typedef void (*radeon_line_func) (radeonContextPtr, - radeonVertex *, radeonVertex *); - -typedef void (*radeon_point_func) (radeonContextPtr, radeonVertex *); - -struct radeon_state { - struct radeon_colorbuffer_state color; - struct radeon_depthbuffer_state depth; - struct radeon_scissor_state scissor; - struct radeon_stencilbuffer_state stencil; -}; - -/** - * This structure holds the command buffer while it is being constructed. - * - * The first batch of commands in the buffer is always the state that needs - * to be re-emitted when the context is lost. This batch can be skipped - * otherwise. - */ -struct radeon_cmdbuf { - struct radeon_cs_manager *csm; - struct radeon_cs *cs; - int size; /** # of dwords total */ - unsigned int flushing:1; /** whether we're currently in FlushCmdBufLocked */ -}; - -struct radeon_context { - GLcontext *glCtx; - radeonScreenPtr radeonScreen; /* Screen private DRI data */ - - /* Texture object bookkeeping - */ - unsigned nr_heaps; - driTexHeap * texture_heaps[ RADEON_NR_TEX_HEAPS ]; - driTextureObject swapped; - int texture_depth; - float initialMaxAnisotropy; - - struct radeon_dma dma; - struct radeon_hw_state hw; - /* Rasterization and vertex state: - */ - GLuint TclFallback; - GLuint Fallback; - GLuint NewGLState; - DECLARE_RENDERINPUTS(tnl_index_bitset); /* index of bits for last tnl_install_attrs */ - - /* Page flipping */ - GLuint doPageFlip; - - /* Drawable, cliprect and scissor information */ - GLuint numClipRects; /* Cliprects for the draw buffer */ - drm_clip_rect_t *pClipRects; - unsigned int lastStamp; - GLboolean lost_context; - drm_radeon_sarea_t *sarea; /* Private SAREA data */ - - /* Mirrors of some DRI state */ - struct radeon_dri_mirror dri; - - /* Busy waiting */ - GLuint do_usleeps; - GLuint do_irqs; - GLuint irqsEmitted; - drm_radeon_irq_wait_t iw; - - /* buffer swap */ - int64_t swap_ust; - int64_t swap_missed_ust; - - GLuint swap_count; - GLuint swap_missed_count; - - /* Derived state - for r300 only */ - struct radeon_state state; - - struct radeon_swtcl_info swtcl; - /* Configuration cache - */ - driOptionCache optionCache; - - struct radeon_cmdbuf cmdbuf; - - struct { - void (*get_lock)(radeonContextPtr radeon); - void (*update_viewport_offset)(GLcontext *ctx); - void (*update_draw_buffer)(GLcontext *ctx); - void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa); - void (*swtcl_flush)(GLcontext *ctx, uint32_t offset); - void (*pre_emit_atoms)(radeonContextPtr rmesa); - void (*pre_emit_state)(radeonContextPtr rmesa); - } vtbl; -}; - -#define RADEON_CONTEXT(glctx) ((radeonContextPtr)(ctx->DriverCtx)) - -/** - * This function takes a float and packs it into a uint32_t - */ -static INLINE uint32_t radeonPackFloat32(float fl) -{ - union { - float fl; - uint32_t u; - } u; - - u.fl = fl; - return u.u; -} - -/* This is probably wrong for some values, I need to test this - * some more. Range checking would be a good idea also.. - * - * But it works for most things. I'll fix it later if someone - * else with a better clue doesn't - */ -static INLINE uint32_t radeonPackFloat24(float f) -{ - float mantissa; - int exponent; - uint32_t float24 = 0; - - if (f == 0.0) - return 0; - - mantissa = frexpf(f, &exponent); - - /* Handle -ve */ - if (mantissa < 0) { - float24 |= (1 << 23); - mantissa = mantissa * -1.0; - } - /* Handle exponent, bias of 63 */ - exponent += 62; - float24 |= (exponent << 16); - /* Kill 7 LSB of mantissa */ - float24 |= (radeonPackFloat32(mantissa) & 0x7FFFFF) >> 7; - - return float24; -} - -/* ================================================================ - * Debugging: - */ -#define DO_DEBUG 1 - -#if DO_DEBUG -extern int RADEON_DEBUG; -#else -#define RADEON_DEBUG 0 -#endif - -#include "common_misc.h" -#include "common_cmdbuf.h" -#endif diff --git a/src/mesa/drivers/dri/radeon/common_lock.c b/src/mesa/drivers/dri/radeon/common_lock.c deleted file mode 100644 index 91600ac4cd..0000000000 --- a/src/mesa/drivers/dri/radeon/common_lock.c +++ /dev/null @@ -1,126 +0,0 @@ -/************************************************************************** - -Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and - VA Linux Systems Inc., Fremont, California. -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Gareth Hughes - * Keith Whitwell - * Kevin E. Martin - */ - -#include "main/glheader.h" -#include "main/mtypes.h" -#include "main/colormac.h" -#include "dri_util.h" -#include "radeon_screen.h" -#include "common_context.h" -#include "common_lock.h" -#include "common_misc.h" - -#include "drirenderbuffer.h" - -#if DEBUG_LOCKING -char *prevLockFile = NULL; -int prevLockLine = 0; -#endif - -/* Turn on/off page flipping according to the flags in the sarea: - */ -void radeonUpdatePageFlipping(radeonContextPtr rmesa) -{ - int use_back; - __DRIdrawablePrivate *const drawable = rmesa->dri.drawable; - GLframebuffer *fb = drawable->driverPrivate; - - rmesa->doPageFlip = rmesa->sarea->pfState; - if (rmesa->glCtx->WinSysDrawBuffer) { - rmesa->vtbl.update_draw_buffer(rmesa->glCtx); - } - - use_back = rmesa->glCtx->DrawBuffer ? - (rmesa->glCtx->DrawBuffer->_ColorDrawBufferIndexes[0] == - BUFFER_BACK_LEFT) : 1; - use_back ^= (rmesa->sarea->pfCurrentPage == 1); - - if (use_back) - rmesa->state.color.rrb = (void *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; - else - rmesa->state.color.rrb = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; - - rmesa->state.depth.rrb = (void *)fb->Attachment[BUFFER_DEPTH].Renderbuffer; -} - -/* Update the hardware state. This is called if another context has - * grabbed the hardware lock, which includes the X server. This - * function also updates the driver's window state after the X server - * moves, resizes or restacks a window -- the change will be reflected - * in the drawable position and clip rects. Since the X server grabs - * the hardware lock when it changes the window state, this routine will - * automatically be called after such a change. - */ -void radeonGetLock(radeonContextPtr rmesa, GLuint flags) -{ - __DRIdrawablePrivate *const drawable = rmesa->dri.drawable; - __DRIdrawablePrivate *const readable = rmesa->dri.readable; - __DRIscreenPrivate *sPriv = rmesa->dri.screen; - drm_radeon_sarea_t *sarea = rmesa->sarea; - - assert(drawable != NULL); - - drmGetLock(rmesa->dri.fd, rmesa->dri.hwContext, flags); - - /* The window might have moved, so we might need to get new clip - * rects. - * - * NOTE: This releases and regrabs the hw lock to allow the X server - * to respond to the DRI protocol request for new drawable info. - * Since the hardware state depends on having the latest drawable - * clip rects, all state checking must be done _after_ this call. - */ - DRI_VALIDATE_DRAWABLE_INFO(sPriv, drawable); - if (drawable != readable) { - DRI_VALIDATE_DRAWABLE_INFO(sPriv, readable); - } - - if (rmesa->lastStamp != drawable->lastStamp) { - radeonUpdatePageFlipping(rmesa); - radeonSetCliprects(rmesa); - rmesa->vtbl.update_viewport_offset(rmesa->glCtx); - driUpdateFramebufferSize(rmesa->glCtx, drawable); - } - - rmesa->vtbl.get_lock(rmesa); - - rmesa->lost_context = GL_TRUE; -} diff --git a/src/mesa/drivers/dri/radeon/common_lock.h b/src/mesa/drivers/dri/radeon/common_lock.h deleted file mode 100644 index 431b076f02..0000000000 --- a/src/mesa/drivers/dri/radeon/common_lock.h +++ /dev/null @@ -1,120 +0,0 @@ -/************************************************************************** - -Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and - VA Linux Systems Inc., Fremont, California. -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Gareth Hughes - * Keith Whitwell - * Kevin E. Martin - */ - -#ifndef COMMON_LOCK_H -#define COMMON_LOCK_H - -#include "main/colormac.h" -#include "radeon_screen.h" -#include "common_context.h" - -extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags); - -/* Turn DEBUG_LOCKING on to find locking conflicts. - */ -#define DEBUG_LOCKING 0 - -#if DEBUG_LOCKING -extern char *prevLockFile; -extern int prevLockLine; - -#define DEBUG_LOCK() \ - do { \ - prevLockFile = (__FILE__); \ - prevLockLine = (__LINE__); \ - } while (0) - -#define DEBUG_RESET() \ - do { \ - prevLockFile = 0; \ - prevLockLine = 0; \ - } while (0) - -#define DEBUG_CHECK_LOCK() \ - do { \ - if ( prevLockFile ) { \ - fprintf( stderr, \ - "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n", \ - prevLockFile, prevLockLine, __FILE__, __LINE__ ); \ - exit( 1 ); \ - } \ - } while (0) - -#else - -#define DEBUG_LOCK() -#define DEBUG_RESET() -#define DEBUG_CHECK_LOCK() - -#endif - -/* - * !!! We may want to separate locks from locks with validation. This - * could be used to improve performance for those things commands that - * do not do any drawing !!! - */ - -/* Lock the hardware and validate our state. - */ -#define LOCK_HARDWARE( rmesa ) \ - do { \ - char __ret = 0; \ - DEBUG_CHECK_LOCK(); \ - if (!(rmesa)->radeonScreen->driScreen->dri2.enabled) { \ - DRM_CAS( (rmesa)->dri.hwLock, (rmesa)->dri.hwContext, \ - (DRM_LOCK_HELD | (rmesa)->dri.hwContext), __ret ); \ - if ( __ret ) \ - radeonGetLock( (rmesa), 0 ); \ - } \ - DEBUG_LOCK(); \ - } while (0) - -#define UNLOCK_HARDWARE( rmesa ) \ - do { \ - if (!(rmesa)->radeonScreen->driScreen->dri2.enabled) { \ - DRM_UNLOCK( (rmesa)->dri.fd, \ - (rmesa)->dri.hwLock, \ - (rmesa)->dri.hwContext ); \ - DEBUG_RESET(); \ - } \ - } while (0) - -#endif diff --git a/src/mesa/drivers/dri/radeon/common_misc.c b/src/mesa/drivers/dri/radeon/common_misc.c deleted file mode 100644 index f0035e54cf..0000000000 --- a/src/mesa/drivers/dri/radeon/common_misc.c +++ /dev/null @@ -1,2715 +0,0 @@ -/************************************************************************** - -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE -LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION -OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Keith Whitwell - */ - -#include -#include "main/glheader.h" -#include "main/imports.h" -#include "main/context.h" -#include "main/api_arrayelt.h" -#include "main/enums.h" -#include "main/colormac.h" -#include "main/light.h" -#include "main/framebuffer.h" -#include "main/simple_list.h" - -#include "swrast/swrast.h" -#include "vbo/vbo.h" -#include "tnl/tnl.h" -#include "tnl/t_pipeline.h" -#include "swrast_setup/swrast_setup.h" -#include "main/mipmap.h" -#include "main/texformat.h" -#include "main/texstore.h" -#include "main/teximage.h" -#include "main/texobj.h" - -#include "drirenderbuffer.h" -#include "vblank.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ - -#include "radeon_bo.h" -#include "radeon_cs.h" -#include "radeon_bo_legacy.h" -#include "radeon_cs_legacy.h" -#include "radeon_bo_gem.h" -#include "radeon_cs_gem.h" -#include "dri_util.h" -#include "radeon_drm.h" -#include "radeon_buffer.h" -#include "radeon_screen.h" -#include "common_context.h" -#include "common_misc.h" -#include "common_lock.h" -#include "common_cmdbuf.h" -#include "radeon_mipmap_tree.h" -#include "radeon_span.h" - -#define DRIVER_DATE "20090101" - -#ifndef RADEON_DEBUG -int RADEON_DEBUG = (0); -#endif - -#define DEBUG_CMDBUF 0 - -/* ============================================================= - * Scissoring - */ - -static GLboolean intersect_rect(drm_clip_rect_t * out, - drm_clip_rect_t * a, drm_clip_rect_t * b) -{ - *out = *a; - if (b->x1 > out->x1) - out->x1 = b->x1; - if (b->y1 > out->y1) - out->y1 = b->y1; - if (b->x2 < out->x2) - out->x2 = b->x2; - if (b->y2 < out->y2) - out->y2 = b->y2; - if (out->x1 >= out->x2) - return GL_FALSE; - if (out->y1 >= out->y2) - return GL_FALSE; - return GL_TRUE; -} - -void radeonRecalcScissorRects(radeonContextPtr radeon) -{ - drm_clip_rect_t *out; - int i; - - /* Grow cliprect store? - */ - if (radeon->state.scissor.numAllocedClipRects < radeon->numClipRects) { - while (radeon->state.scissor.numAllocedClipRects < - radeon->numClipRects) { - radeon->state.scissor.numAllocedClipRects += 1; /* zero case */ - radeon->state.scissor.numAllocedClipRects *= 2; - } - - if (radeon->state.scissor.pClipRects) - FREE(radeon->state.scissor.pClipRects); - - radeon->state.scissor.pClipRects = - MALLOC(radeon->state.scissor.numAllocedClipRects * - sizeof(drm_clip_rect_t)); - - if (radeon->state.scissor.pClipRects == NULL) { - radeon->state.scissor.numAllocedClipRects = 0; - return; - } - } - - out = radeon->state.scissor.pClipRects; - radeon->state.scissor.numClipRects = 0; - - for (i = 0; i < radeon->numClipRects; i++) { - if (intersect_rect(out, - &radeon->pClipRects[i], - &radeon->state.scissor.rect)) { - radeon->state.scissor.numClipRects++; - out++; - } - } -} - -/** - * Update cliprects and scissors. - */ -void radeonSetCliprects(radeonContextPtr radeon) -{ - __DRIdrawablePrivate *const drawable = radeon->dri.drawable; - __DRIdrawablePrivate *const readable = radeon->dri.readable; - GLframebuffer *const draw_fb = (GLframebuffer*)drawable->driverPrivate; - GLframebuffer *const read_fb = (GLframebuffer*)readable->driverPrivate; - - if (!radeon->radeonScreen->driScreen->dri2.enabled) { - if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) { - /* Can't ignore 2d windows if we are page flipping. */ - if (drawable->numBackClipRects == 0 || radeon->doPageFlip || - radeon->sarea->pfCurrentPage == 1) { - radeon->numClipRects = drawable->numClipRects; - radeon->pClipRects = drawable->pClipRects; - } else { - radeon->numClipRects = drawable->numBackClipRects; - radeon->pClipRects = drawable->pBackClipRects; - } - } else { - /* front buffer (or none, or multiple buffers */ - radeon->numClipRects = drawable->numClipRects; - radeon->pClipRects = drawable->pClipRects; - } - } - - if ((draw_fb->Width != drawable->w) || - (draw_fb->Height != drawable->h)) { - _mesa_resize_framebuffer(radeon->glCtx, draw_fb, - drawable->w, drawable->h); - draw_fb->Initialized = GL_TRUE; - } - - if (drawable != readable) { - if ((read_fb->Width != readable->w) || - (read_fb->Height != readable->h)) { - _mesa_resize_framebuffer(radeon->glCtx, read_fb, - readable->w, readable->h); - read_fb->Initialized = GL_TRUE; - } - } - - if (radeon->state.scissor.enabled) - radeonRecalcScissorRects(radeon); - - radeon->lastStamp = drawable->lastStamp; -} - -void radeonUpdateScissor( GLcontext *ctx ) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - - if ( rmesa->dri.drawable ) { - __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; - - int x = ctx->Scissor.X; - int y = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height; - int w = ctx->Scissor.X + ctx->Scissor.Width - 1; - int h = dPriv->h - ctx->Scissor.Y - 1; - - rmesa->state.scissor.rect.x1 = x + dPriv->x; - rmesa->state.scissor.rect.y1 = y + dPriv->y; - rmesa->state.scissor.rect.x2 = w + dPriv->x + 1; - rmesa->state.scissor.rect.y2 = h + dPriv->y + 1; - - radeonRecalcScissorRects( rmesa ); - } -} - -/* ============================================================= - * Scissoring - */ - -void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - if (ctx->Scissor.Enabled) { - /* We don't pipeline cliprect changes */ - radeon_firevertices(radeon); - radeonUpdateScissor(ctx); - } -} - - -/* ================================================================ - * SwapBuffers with client-side throttling - */ - -static uint32_t radeonGetLastFrame(radeonContextPtr radeon) -{ - drm_radeon_getparam_t gp; - int ret; - uint32_t frame; - - gp.param = RADEON_PARAM_LAST_FRAME; - gp.value = (int *)&frame; - ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM, - &gp, sizeof(gp)); - if (ret) { - fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, - ret); - exit(1); - } - - return frame; -} - -uint32_t radeonGetAge(radeonContextPtr radeon) -{ - drm_radeon_getparam_t gp; - int ret; - uint32_t age; - - gp.param = RADEON_PARAM_LAST_CLEAR; - gp.value = (int *)&age; - ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM, - &gp, sizeof(gp)); - if (ret) { - fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, - ret); - exit(1); - } - - return age; -} - -static void radeonEmitIrqLocked(radeonContextPtr radeon) -{ - drm_radeon_irq_emit_t ie; - int ret; - - ie.irq_seq = &radeon->iw.irq_seq; - ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_IRQ_EMIT, - &ie, sizeof(ie)); - if (ret) { - fprintf(stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__, - ret); - exit(1); - } -} - -static void radeonWaitIrq(radeonContextPtr radeon) -{ - int ret; - - do { - ret = drmCommandWrite(radeon->dri.fd, DRM_RADEON_IRQ_WAIT, - &radeon->iw, sizeof(radeon->iw)); - } while (ret && (errno == EINTR || errno == EBUSY)); - - if (ret) { - fprintf(stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, - ret); - exit(1); - } -} - -static void radeonWaitForFrameCompletion(radeonContextPtr radeon) -{ - drm_radeon_sarea_t *sarea = radeon->sarea; - - if (radeon->do_irqs) { - if (radeonGetLastFrame(radeon) < sarea->last_frame) { - if (!radeon->irqsEmitted) { - while (radeonGetLastFrame(radeon) < - sarea->last_frame) ; - } else { - UNLOCK_HARDWARE(radeon); - radeonWaitIrq(radeon); - LOCK_HARDWARE(radeon); - } - radeon->irqsEmitted = 10; - } - - if (radeon->irqsEmitted) { - radeonEmitIrqLocked(radeon); - radeon->irqsEmitted--; - } - } else { - while (radeonGetLastFrame(radeon) < sarea->last_frame) { - UNLOCK_HARDWARE(radeon); - if (radeon->do_usleeps) - DO_USLEEP(1); - LOCK_HARDWARE(radeon); - } - } -} - -/* wait for idle */ -void radeonWaitForIdleLocked(radeonContextPtr radeon) -{ - int ret; - int i = 0; - - do { - ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_CP_IDLE); - if (ret) - DO_USLEEP(1); - } while (ret && ++i < 100); - - if (ret < 0) { - UNLOCK_HARDWARE(radeon); - fprintf(stderr, "Error: R300 timed out... exiting\n"); - exit(-1); - } -} - -static void radeonWaitForIdle(radeonContextPtr radeon) -{ - LOCK_HARDWARE(radeon); - radeonWaitForIdleLocked(radeon); - UNLOCK_HARDWARE(radeon); -} - - -/* Copy the back color buffer to the front color buffer. - */ -void radeonCopyBuffer( __DRIdrawablePrivate *dPriv, - const drm_clip_rect_t *rect) -{ - radeonContextPtr rmesa; - GLint nbox, i, ret; - GLboolean missed_target; - int64_t ust; - __DRIscreenPrivate *psp; - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; - - if ( RADEON_DEBUG & DEBUG_IOCTL ) { - fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *) rmesa->glCtx ); - } - - radeon_firevertices(rmesa); - LOCK_HARDWARE( rmesa ); - - /* Throttle the frame rate -- only allow one pending swap buffers - * request at a time. - */ - radeonWaitForFrameCompletion( rmesa ); - if (!rect) - { - UNLOCK_HARDWARE( rmesa ); - driWaitForVBlank( dPriv, & missed_target ); - LOCK_HARDWARE( rmesa ); - } - - nbox = dPriv->numClipRects; /* must be in locked region */ - - for ( i = 0 ; i < nbox ; ) { - GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox ); - drm_clip_rect_t *box = dPriv->pClipRects; - drm_clip_rect_t *b = rmesa->sarea->boxes; - GLint n = 0; - - for ( ; i < nr ; i++ ) { - - *b = box[i]; - - if (rect) - { - if (rect->x1 > b->x1) - b->x1 = rect->x1; - if (rect->y1 > b->y1) - b->y1 = rect->y1; - if (rect->x2 < b->x2) - b->x2 = rect->x2; - if (rect->y2 < b->y2) - b->y2 = rect->y2; - - if (b->x1 >= b->x2 || b->y1 >= b->y2) - continue; - } - - b++; - n++; - } - rmesa->sarea->nbox = n; - - if (!n) - continue; - - ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP ); - - if ( ret ) { - fprintf( stderr, "DRM_RADEON_SWAP_BUFFERS: return = %d\n", ret ); - UNLOCK_HARDWARE( rmesa ); - exit( 1 ); - } - } - - UNLOCK_HARDWARE( rmesa ); - if (!rect) - { - psp = dPriv->driScreenPriv; - rmesa->swap_count++; - (*psp->systemTime->getUST)( & ust ); - if ( missed_target ) { - rmesa->swap_missed_count++; - rmesa->swap_missed_ust = ust - rmesa->swap_ust; - } - - rmesa->swap_ust = ust; - rmesa->hw.all_dirty = GL_TRUE; - - } -} - -void radeonPageFlip( __DRIdrawablePrivate *dPriv ) -{ - radeonContextPtr rmesa; - GLint ret; - GLboolean missed_target; - __DRIscreenPrivate *psp; - struct radeon_renderbuffer *rrb; - GLframebuffer *fb = dPriv->driverPrivate; - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; - rrb = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; - - psp = dPriv->driScreenPriv; - - if ( RADEON_DEBUG & DEBUG_IOCTL ) { - fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__, - rmesa->sarea->pfCurrentPage); - } - - radeon_firevertices(rmesa); - - LOCK_HARDWARE( rmesa ); - - if (!dPriv->numClipRects) { - UNLOCK_HARDWARE(rmesa); - usleep(10000); /* throttle invisible client 10ms */ - return; - } - - drm_clip_rect_t *box = dPriv->pClipRects; - drm_clip_rect_t *b = rmesa->sarea->boxes; - b[0] = box[0]; - rmesa->sarea->nbox = 1; - - /* Throttle the frame rate -- only allow a few pending swap buffers - * request at a time. - */ - radeonWaitForFrameCompletion( rmesa ); - UNLOCK_HARDWARE( rmesa ); - driWaitForVBlank( dPriv, & missed_target ); - if ( missed_target ) { - rmesa->swap_missed_count++; - (void) (*psp->systemTime->getUST)( & rmesa->swap_missed_ust ); - } - LOCK_HARDWARE( rmesa ); - - ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_FLIP ); - - UNLOCK_HARDWARE( rmesa ); - - if ( ret ) { - fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret ); - exit( 1 ); - } - - rmesa->swap_count++; - (void) (*psp->systemTime->getUST)( & rmesa->swap_ust ); - - /* Get ready for drawing next frame. Update the renderbuffers' - * flippedOffset/Pitch fields so we draw into the right place. - */ - driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer, - rmesa->sarea->pfCurrentPage); - - rmesa->state.color.rrb = rrb; - - if (rmesa->vtbl.update_draw_buffer) - rmesa->vtbl.update_draw_buffer(rmesa->glCtx); -} - -void radeonFlush(GLcontext *ctx) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (radeon->dma.flush) - radeon->dma.flush( ctx ); - - radeonEmitState(radeon); - - if (radeon->cmdbuf.cs->cdw) - rcommonFlushCmdBuf(radeon, __FUNCTION__); -} - -static INLINE void radeonEmitAtoms(radeonContextPtr radeon, GLboolean dirty) -{ - BATCH_LOCALS(radeon); - struct radeon_state_atom *atom; - int dwords; - - if (radeon->vtbl.pre_emit_atoms) - radeon->vtbl.pre_emit_atoms(radeon); - - /* Emit actual atoms */ - foreach(atom, &radeon->hw.atomlist) { - if ((atom->dirty || radeon->hw.all_dirty) == dirty) { - dwords = (*atom->check) (radeon->glCtx, atom); - if (dwords) { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - radeon_print_state_atom(atom); - } - if (atom->emit) { - (*atom->emit)(radeon->glCtx, atom); - } else { - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(atom->cmd, dwords); - END_BATCH(); - } - atom->dirty = GL_FALSE; - } else { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - fprintf(stderr, " skip state %s\n", - atom->name); - } - } - } - } - - COMMIT_BATCH(); -} - -void radeonEmitState(radeonContextPtr radeon) -{ - if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (radeon->vtbl.pre_emit_state) - radeon->vtbl.pre_emit_state(radeon); - - /* this code used to return here but now it emits zbs */ - if (radeon->cmdbuf.cs->cdw && !radeon->hw.is_dirty && !radeon->hw.all_dirty) - return; - - /* To avoid going across the entire set of states multiple times, just check - * for enough space for the case of emitting all state, and inline the - * radeonAllocCmdBuf code here without all the checks. - */ - rcommonEnsureCmdBufSpace(radeon, radeon->hw.max_state_size, __FUNCTION__); - - /* We always always emit zbs, this is due to a bug found by keithw in - the hardware and rediscovered after Erics changes by me. - if you ever touch this code make sure you emit zbs otherwise - you get tcl lockups on at least M7/7500 class of chips - airlied */ - /* special r100 case */ - // rmesa->hw.zbs.dirty=1; - - if (!radeon->cmdbuf.cs->cdw) { - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin reemit state\n"); - - radeonEmitAtoms(radeon, GL_FALSE); - } - - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin dirty state\n"); - - radeonEmitAtoms(radeon, GL_TRUE); - radeon->hw.is_dirty = GL_FALSE; - radeon->hw.all_dirty = GL_FALSE; - -} - - -/* Make sure all commands have been sent to the hardware and have - * completed processing. - */ -void radeonFinish(GLcontext * ctx) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - struct gl_framebuffer *fb = ctx->DrawBuffer; - int i; - - radeonFlush(ctx); - - if (radeon->radeonScreen->kernel_mm) { - for (i = 0; i < fb->_NumColorDrawBuffers; i++) { - struct radeon_renderbuffer *rrb; - rrb = (struct radeon_renderbuffer *)fb->_ColorDrawBuffers[i]; - if (rrb->bo) - radeon_bo_wait(rrb->bo); - } - } else if (radeon->do_irqs) { - LOCK_HARDWARE(radeon); - radeonEmitIrqLocked(radeon); - UNLOCK_HARDWARE(radeon); - radeonWaitIrq(radeon); - } else { - radeonWaitForIdle(radeon); - } -} - -/** - * Swap front and back buffer. - */ -void radeonSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - radeonContextPtr radeon; - GLcontext *ctx; - - radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; - ctx = radeon->glCtx; - - if (ctx->Visual.doubleBufferMode) { - _mesa_notifySwapBuffers(ctx);/* flush pending rendering comands */ - if (radeon->doPageFlip) { - radeonPageFlip(dPriv); - } else { - radeonCopyBuffer(dPriv, NULL); - } - } - } else { - /* XXX this shouldn't be an error but we can't handle it for now */ - _mesa_problem(NULL, "%s: drawable has no context!", - __FUNCTION__); - } -} - -void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv, - int x, int y, int w, int h ) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - radeonContextPtr radeon; - GLcontext *ctx; - - radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; - ctx = radeon->glCtx; - - if (ctx->Visual.doubleBufferMode) { - drm_clip_rect_t rect; - rect.x1 = x + dPriv->x; - rect.y1 = (dPriv->h - y - h) + dPriv->y; - rect.x2 = rect.x1 + w; - rect.y2 = rect.y1 + h; - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - radeonCopyBuffer(dPriv, &rect); - } - } else { - /* XXX this shouldn't be an error but we can't handle it for now */ - _mesa_problem(NULL, "%s: drawable has no context!", - __FUNCTION__); - } -} - -/* cmdbuffer */ -/** - * Send the current command buffer via ioctl to the hardware. - */ -int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller) -{ - int ret = 0; - - if (rmesa->cmdbuf.flushing) { - fprintf(stderr, "Recursive call into r300FlushCmdBufLocked!\n"); - exit(-1); - } - rmesa->cmdbuf.flushing = 1; - if (rmesa->cmdbuf.cs->cdw) { - ret = radeon_cs_emit(rmesa->cmdbuf.cs); - rmesa->hw.all_dirty = GL_TRUE; - } - radeon_cs_erase(rmesa->cmdbuf.cs); - rmesa->cmdbuf.flushing = 0; - return ret; -} - -int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller) -{ - int ret; - - radeonReleaseDmaRegion(rmesa); - - LOCK_HARDWARE(rmesa); - ret = rcommonFlushCmdBufLocked(rmesa, caller); - UNLOCK_HARDWARE(rmesa); - - if (ret) { - fprintf(stderr, "drmRadeonCmdBuffer: %d\n", ret); - _mesa_exit(ret); - } - - return ret; -} - -/** - * Make sure that enough space is available in the command buffer - * by flushing if necessary. - * - * \param dwords The number of dwords we need to be free on the command buffer - */ -void rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller) -{ - if ((rmesa->cmdbuf.cs->cdw + dwords + 128) > rmesa->cmdbuf.size || - radeon_cs_need_flush(rmesa->cmdbuf.cs)) { - rcommonFlushCmdBuf(rmesa, caller); - } -} - -void rcommonInitCmdBuf(radeonContextPtr rmesa) -{ - GLuint size; - /* Initialize command buffer */ - size = 256 * driQueryOptioni(&rmesa->optionCache, - "command_buffer_size"); - if (size < 2 * rmesa->hw.max_state_size) { - size = 2 * rmesa->hw.max_state_size + 65535; - } - if (size > 64 * 256) - size = 64 * 256; - - if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) { - fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%zd\n", - sizeof(drm_r300_cmd_header_t)); - fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%zd\n", - sizeof(drm_radeon_cmd_buffer_t)); - fprintf(stderr, - "Allocating %d bytes command buffer (max state is %d bytes)\n", - size * 4, rmesa->hw.max_state_size * 4); - } - - if (rmesa->radeonScreen->kernel_mm) { - int fd = rmesa->radeonScreen->driScreen->fd; - rmesa->cmdbuf.csm = radeon_cs_manager_gem_ctor(fd); - } else { - rmesa->cmdbuf.csm = radeon_cs_manager_legacy_ctor(rmesa); - } - if (rmesa->cmdbuf.csm == NULL) { - /* FIXME: fatal error */ - return; - } - rmesa->cmdbuf.cs = radeon_cs_create(rmesa->cmdbuf.csm, size); - assert(rmesa->cmdbuf.cs != NULL); - rmesa->cmdbuf.size = size; - - if (!rmesa->radeonScreen->kernel_mm) { - radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, rmesa->radeonScreen->texSize[0]); - radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, rmesa->radeonScreen->gartTextures.size); - } else { - struct drm_radeon_gem_info mminfo; - - if (!drmCommandWriteRead(rmesa->dri.fd, DRM_RADEON_GEM_INFO, &mminfo, sizeof(mminfo))) - { - radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, mminfo.vram_size); - radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, mminfo.gart_size); - } - } - -} -/** - * Destroy the command buffer - */ -void rcommonDestroyCmdBuf(radeonContextPtr rmesa) -{ - radeon_cs_destroy(rmesa->cmdbuf.cs); - if (rmesa->radeonScreen->driScreen->dri2.enabled || rmesa->radeonScreen->kernel_mm) { - radeon_cs_manager_gem_dtor(rmesa->cmdbuf.csm); - } else { - radeon_cs_manager_legacy_dtor(rmesa->cmdbuf.csm); - } -} - -void rcommonBeginBatch(radeonContextPtr rmesa, int n, - int dostate, - const char *file, - const char *function, - int line) -{ - rcommonEnsureCmdBufSpace(rmesa, n, function); - if (!rmesa->cmdbuf.cs->cdw && dostate) { - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "Reemit state after flush (from %s)\n", function); - radeonEmitState(rmesa); - } - radeon_cs_begin(rmesa->cmdbuf.cs, n, file, function, line); -} - - - -/* Return various strings for glGetString(). - */ -static const GLubyte *radeonGetString(GLcontext * ctx, GLenum name) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - static char buffer[128]; - - switch (name) { - case GL_VENDOR: - if (IS_R300_CLASS(radeon->radeonScreen)) - return (GLubyte *) "DRI R300 Project"; - else - return (GLubyte *) "Tungsten Graphics, Inc."; - - case GL_RENDERER: - { - unsigned offset; - GLuint agp_mode = (radeon->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 : - radeon->radeonScreen->AGPMode; - const char* chipname; - - if (IS_R300_CLASS(radeon->radeonScreen)) - chipname = "R300"; - else if (IS_R200_CLASS(radeon->radeonScreen)) - chipname = "R200"; - else - chipname = "R100"; - - offset = driGetRendererString(buffer, chipname, DRIVER_DATE, - agp_mode); - - if (IS_R300_CLASS(radeon->radeonScreen)) { - sprintf(&buffer[offset], " %sTCL", - (radeon->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) - ? "" : "NO-"); - } else { - sprintf(&buffer[offset], " %sTCL", - !(radeon->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE) - ? "" : "NO-"); - } - - if (radeon->radeonScreen->driScreen->dri2.enabled) - strcat(buffer, " DRI2"); - - return (GLubyte *) buffer; - } - - default: - return NULL; - } -} - -/* Initialize the driver's misc functions. - */ -static void radeonInitDriverFuncs(struct dd_function_table *functions) -{ - functions->GetString = radeonGetString; -} - -/** - * Create and initialize all common fields of the context, - * including the Mesa context itself. - */ -GLboolean radeonInitContext(radeonContextPtr radeon, - struct dd_function_table* functions, - const __GLcontextModes * glVisual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); - GLcontext* ctx; - GLcontext* shareCtx; - int fthrottle_mode; - - /* Fill in additional standard functions. */ - radeonInitDriverFuncs(functions); - - radeon->radeonScreen = screen; - /* Allocate and initialize the Mesa context */ - if (sharedContextPrivate) - shareCtx = ((radeonContextPtr)sharedContextPrivate)->glCtx; - else - shareCtx = NULL; - radeon->glCtx = _mesa_create_context(glVisual, shareCtx, - functions, (void *)radeon); - if (!radeon->glCtx) - return GL_FALSE; - - ctx = radeon->glCtx; - driContextPriv->driverPrivate = radeon; - - /* DRI fields */ - radeon->dri.context = driContextPriv; - radeon->dri.screen = sPriv; - radeon->dri.drawable = NULL; - radeon->dri.readable = NULL; - radeon->dri.hwContext = driContextPriv->hHWContext; - radeon->dri.hwLock = &sPriv->pSAREA->lock; - radeon->dri.fd = sPriv->fd; - radeon->dri.drmMinor = sPriv->drm_version.minor; - - radeon->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA + - screen->sarea_priv_offset); - - /* Setup IRQs */ - fthrottle_mode = driQueryOptioni(&radeon->optionCache, "fthrottle_mode"); - radeon->iw.irq_seq = -1; - radeon->irqsEmitted = 0; - radeon->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS && - radeon->radeonScreen->irq); - - radeon->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); - - if (!radeon->do_irqs) - fprintf(stderr, - "IRQ's not enabled, falling back to %s: %d %d\n", - radeon->do_usleeps ? "usleeps" : "busy waits", - fthrottle_mode, radeon->radeonScreen->irq); - - (*sPriv->systemTime->getUST) (&radeon->swap_ust); - - return GL_TRUE; -} - -/** - * Cleanup common context fields. - * Called by r200DestroyContext/r300DestroyContext - */ -void radeonCleanupContext(radeonContextPtr radeon) -{ - FILE *track; - struct radeon_renderbuffer *rb; - GLframebuffer *fb; - - /* free the Mesa context */ - _mesa_destroy_context(radeon->glCtx); - - fb = (void*)radeon->dri.drawable->driverPrivate; - rb = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void *)fb->Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - fb = (void*)radeon->dri.readable->driverPrivate; - rb = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void *)fb->Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - - /* _mesa_destroy_context() might result in calls to functions that - * depend on the DriverCtx, so don't set it to NULL before. - * - * radeon->glCtx->DriverCtx = NULL; - */ - - - - /* free the option cache */ - driDestroyOptionCache(&radeon->optionCache); - - rcommonDestroyCmdBuf(radeon); - - if (radeon->state.scissor.pClipRects) { - FREE(radeon->state.scissor.pClipRects); - radeon->state.scissor.pClipRects = 0; - } - track = fopen("/tmp/tracklog", "w"); - if (track) { - radeon_tracker_print(&radeon->radeonScreen->bom->tracker, track); - fclose(track); - } -} - -/* Force the context `c' to be unbound from its buffer. - */ -GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv) -{ - radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; - - if (RADEON_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx %p\n", __FUNCTION__, - radeon->glCtx); - - return GL_TRUE; -} - - -static void -radeon_make_kernel_renderbuffer_current(radeonContextPtr radeon, - GLframebuffer *draw) -{ - /* if radeon->fake */ - struct radeon_renderbuffer *rb; - - if ((rb = (void *)draw->Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) { - if (!rb->bo) { - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - radeon->radeonScreen->frontOffset, - 0, - 0, - RADEON_GEM_DOMAIN_VRAM, - 0); - } - rb->cpp = radeon->radeonScreen->cpp; - rb->pitch = radeon->radeonScreen->frontPitch * rb->cpp; - } - if ((rb = (void *)draw->Attachment[BUFFER_BACK_LEFT].Renderbuffer)) { - if (!rb->bo) { - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - radeon->radeonScreen->backOffset, - 0, - 0, - RADEON_GEM_DOMAIN_VRAM, - 0); - } - rb->cpp = radeon->radeonScreen->cpp; - rb->pitch = radeon->radeonScreen->backPitch * rb->cpp; - } - if ((rb = (void *)draw->Attachment[BUFFER_DEPTH].Renderbuffer)) { - if (!rb->bo) { - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - radeon->radeonScreen->depthOffset, - 0, - 0, - RADEON_GEM_DOMAIN_VRAM, - 0); - } - rb->cpp = radeon->radeonScreen->cpp; - rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp; - } - if ((rb = (void *)draw->Attachment[BUFFER_STENCIL].Renderbuffer)) { - if (!rb->bo) { - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - radeon->radeonScreen->depthOffset, - 0, - 0, - RADEON_GEM_DOMAIN_VRAM, - 0); - } - rb->cpp = radeon->radeonScreen->cpp; - rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp; - } -} - -static void -radeon_make_renderbuffer_current(radeonContextPtr radeon, - GLframebuffer *draw) -{ - int size = 4096*4096*4; - /* if radeon->fake */ - struct radeon_renderbuffer *rb; - - if (radeon->radeonScreen->kernel_mm) { - radeon_make_kernel_renderbuffer_current(radeon, draw); - return; - } - - - if ((rb = (void *)draw->Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) { - if (!rb->bo) { - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - radeon->radeonScreen->frontOffset + - radeon->radeonScreen->fbLocation, - size, - 4096, - RADEON_GEM_DOMAIN_VRAM, - 0); - } - rb->cpp = radeon->radeonScreen->cpp; - rb->pitch = radeon->radeonScreen->frontPitch * rb->cpp; - } - if ((rb = (void *)draw->Attachment[BUFFER_BACK_LEFT].Renderbuffer)) { - if (!rb->bo) { - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - radeon->radeonScreen->backOffset + - radeon->radeonScreen->fbLocation, - size, - 4096, - RADEON_GEM_DOMAIN_VRAM, - 0); - } - rb->cpp = radeon->radeonScreen->cpp; - rb->pitch = radeon->radeonScreen->backPitch * rb->cpp; - } - if ((rb = (void *)draw->Attachment[BUFFER_DEPTH].Renderbuffer)) { - if (!rb->bo) { - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - radeon->radeonScreen->depthOffset + - radeon->radeonScreen->fbLocation, - size, - 4096, - RADEON_GEM_DOMAIN_VRAM, - 0); - } - rb->cpp = radeon->radeonScreen->cpp; - rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp; - } - if ((rb = (void *)draw->Attachment[BUFFER_STENCIL].Renderbuffer)) { - if (!rb->bo) { - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - radeon->radeonScreen->depthOffset + - radeon->radeonScreen->fbLocation, - size, - 4096, - RADEON_GEM_DOMAIN_VRAM, - 0); - } - rb->cpp = radeon->radeonScreen->cpp; - rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp; - } -} - - -void -radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) -{ - unsigned int attachments[10]; - __DRIbuffer *buffers; - __DRIscreen *screen; - struct radeon_renderbuffer *rb; - int i, count; - GLframebuffer *draw; - radeonContextPtr radeon; - - if (RADEON_DEBUG & DEBUG_DRI) - fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable); - - draw = drawable->driverPrivate; - screen = context->driScreenPriv; - radeon = (radeonContextPtr) context->driverPrivate; - i = 0; - if ((rb = (void *)draw->Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) { - attachments[i++] = __DRI_BUFFER_FRONT_LEFT; - } - if ((rb = (void *)draw->Attachment[BUFFER_BACK_LEFT].Renderbuffer)) { - attachments[i++] = __DRI_BUFFER_BACK_LEFT; - } - if ((rb = (void *)draw->Attachment[BUFFER_DEPTH].Renderbuffer)) { - attachments[i++] = __DRI_BUFFER_DEPTH; - } - - buffers = (*screen->dri2.loader->getBuffers)(drawable, - &drawable->w, - &drawable->h, - attachments, i, - &count, - drawable->loaderPrivate); - if (buffers == NULL) - return; - - /* set one cliprect to cover the whole drawable */ - drawable->x = 0; - drawable->y = 0; - drawable->backX = 0; - drawable->backY = 0; - drawable->numClipRects = 1; - drawable->pClipRects[0].x1 = 0; - drawable->pClipRects[0].y1 = 0; - drawable->pClipRects[0].x2 = drawable->w; - drawable->pClipRects[0].y2 = drawable->h; - drawable->numBackClipRects = 1; - drawable->pBackClipRects[0].x1 = 0; - drawable->pBackClipRects[0].y1 = 0; - drawable->pBackClipRects[0].x2 = drawable->w; - drawable->pBackClipRects[0].y2 = drawable->h; - for (i = 0; i < count; i++) { - switch (buffers[i].attachment) { - case __DRI_BUFFER_FRONT_LEFT: - rb = (void *)draw->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; - if (rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb->cpp = buffers[i].cpp; - rb->pitch = buffers[i].pitch; - rb->width = drawable->w; - rb->height = drawable->h; - rb->has_surface = 0; - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - buffers[i].name, - 0, - 0, - RADEON_GEM_DOMAIN_VRAM, - buffers[i].flags); - if (rb->bo == NULL) { - fprintf(stderr, "failled to attach front %d\n", - buffers[i].name); - } - break; - case __DRI_BUFFER_BACK_LEFT: - rb = (void *)draw->Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb->cpp = buffers[i].cpp; - rb->pitch = buffers[i].pitch; - rb->width = drawable->w; - rb->height = drawable->h; - rb->has_surface = 0; - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - buffers[i].name, - 0, - 0, - RADEON_GEM_DOMAIN_VRAM, - buffers[i].flags); - break; - case __DRI_BUFFER_DEPTH: - rb = (void *)draw->Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb->cpp = buffers[i].cpp; - rb->pitch = buffers[i].pitch; - rb->width = drawable->w; - rb->height = drawable->h; - rb->has_surface = 0; - rb->bo = radeon_bo_open(radeon->radeonScreen->bom, - buffers[i].name, - 0, - 0, - RADEON_GEM_DOMAIN_VRAM, - buffers[i].flags); - break; - case __DRI_BUFFER_STENCIL: - break; - case __DRI_BUFFER_ACCUM: - default: - fprintf(stderr, - "unhandled buffer attach event, attacment type %d\n", - buffers[i].attachment); - return; - } - } - radeon = (radeonContextPtr) context->driverPrivate; - driUpdateFramebufferSize(radeon->glCtx, drawable); -} - -/* Force the context `c' to be the current context and associate with it - * buffer `b'. - */ -GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv) -{ - radeonContextPtr radeon; - GLframebuffer *dfb, *rfb; - - if (!driContextPriv) { - if (RADEON_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx is null\n", __FUNCTION__); - _mesa_make_current(NULL, NULL, NULL); - return GL_TRUE; - } - radeon = (radeonContextPtr) driContextPriv->driverPrivate; - dfb = driDrawPriv->driverPrivate; - rfb = driReadPriv->driverPrivate; - - if (driContextPriv->driScreenPriv->dri2.enabled) { - radeon_update_renderbuffers(driContextPriv, driDrawPriv); - if (driDrawPriv != driReadPriv) - radeon_update_renderbuffers(driContextPriv, driReadPriv); - radeon->state.color.rrb = - (void *)dfb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; - radeon->state.depth.rrb = - (void *)dfb->Attachment[BUFFER_DEPTH].Renderbuffer; - } else { - radeon_make_renderbuffer_current(radeon, dfb); - } - - - if (RADEON_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx %p dfb %p rfb %p\n", __FUNCTION__, radeon->glCtx, dfb, rfb); - - driUpdateFramebufferSize(radeon->glCtx, driDrawPriv); - if (driReadPriv != driDrawPriv) - driUpdateFramebufferSize(radeon->glCtx, driReadPriv); - - - - _mesa_make_current(radeon->glCtx, dfb, rfb); - - if (radeon->dri.drawable != driDrawPriv) { - if (driDrawPriv->swap_interval == (unsigned)-1) { - driDrawPriv->vblFlags = - (radeon->radeonScreen->irq != 0) - ? driGetDefaultVBlankFlags(&radeon-> - optionCache) - : VBLANK_FLAG_NO_IRQ; - - driDrawableInitVBlank(driDrawPriv); - } - } - - radeon->dri.readable = driReadPriv; - - if (radeon->dri.drawable != driDrawPriv || - radeon->lastStamp != driDrawPriv->lastStamp) { - radeon->dri.drawable = driDrawPriv; - - radeonSetCliprects(radeon); - radeon->vtbl.update_viewport_offset(radeon->glCtx); - } - - _mesa_update_state(radeon->glCtx); - - if (!driContextPriv->driScreenPriv->dri2.enabled) { - radeonUpdatePageFlipping(radeon); - } - - if (RADEON_DEBUG & DEBUG_DRI) - fprintf(stderr, "End %s\n", __FUNCTION__); - return GL_TRUE; -} - - -#if defined(USE_X86_ASM) -#define COPY_DWORDS( dst, src, nr ) \ -do { \ - int __tmp; \ - __asm__ __volatile__( "rep ; movsl" \ - : "=%c" (__tmp), "=D" (dst), "=S" (__tmp) \ - : "0" (nr), \ - "D" ((long)dst), \ - "S" ((long)src) ); \ -} while (0) -#else -#define COPY_DWORDS( dst, src, nr ) \ -do { \ - int j; \ - for ( j = 0 ; j < nr ; j++ ) \ - dst[j] = ((int *)src)[j]; \ - dst += nr; \ -} while (0) -#endif - -static void radeonEmitVec4(uint32_t *out, GLvoid * data, int stride, int count) -{ - int i; - - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 4) - COPY_DWORDS(out, data, count); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out++; - data += stride; - } -} - -void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count) -{ - int i; - - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 8) - COPY_DWORDS(out, data, count * 2); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data + 4); - out += 2; - data += stride; - } -} - -void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count) -{ - int i; - - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 12) { - COPY_DWORDS(out, data, count * 3); - } - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data + 4); - out[2] = *(int *)(data + 8); - out += 3; - data += stride; - } -} - -static void radeonEmitVec16(uint32_t *out, GLvoid * data, int stride, int count) -{ - int i; - - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 16) - COPY_DWORDS(out, data, count * 4); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data + 4); - out[2] = *(int *)(data + 8); - out[3] = *(int *)(data + 12); - out += 4; - data += stride; - } -} - -void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, - GLvoid * data, int size, int stride, int count) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - uint32_t *out; - - if (stride == 0) { - radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32); - count = 1; - aos->stride = 0; - } else { - radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * count * 4, 32); - aos->stride = size; - } - - aos->components = size; - aos->count = count; - - out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); - switch (size) { - case 1: radeonEmitVec4(out, data, stride, count); break; - case 2: radeonEmitVec8(out, data, stride, count); break; - case 3: radeonEmitVec12(out, data, stride, count); break; - case 4: radeonEmitVec16(out, data, stride, count); break; - default: - assert(0); - break; - } -} - - -void radeon_print_state_atom( struct radeon_state_atom *state ) -{ - int i; - - fprintf(stderr, "emit %s/%d\n", state->name, state->cmd_size); - - if (RADEON_DEBUG & DEBUG_VERBOSE) - for (i = 0 ; i < state->cmd_size ; i++) - fprintf(stderr, "\t%s[%d]: %x\n", state->name, i, state->cmd[i]); - -} - -/* textures */ -/** - * Allocate an empty texture image object. - */ -struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx) -{ - return CALLOC(sizeof(radeon_texture_image)); -} - -/** - * Free memory associated with this texture image. - */ -void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage) -{ - radeon_texture_image* image = get_radeon_texture_image(timage); - - if (image->mt) { - radeon_miptree_unreference(image->mt); - image->mt = 0; - assert(!image->base.Data); - } else { - _mesa_free_texture_image_data(ctx, timage); - } - if (image->bo) { - radeon_bo_unref(image->bo); - image->bo = NULL; - } - if (timage->Data) { - _mesa_free_texmemory(timage->Data); - timage->Data = NULL; - } -} - -/* Set Data pointer and additional data for mapped texture image */ -static void teximage_set_map_data(radeon_texture_image *image) -{ - radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; - - image->base.Data = image->mt->bo->ptr + lvl->faces[image->mtface].offset; - image->base.RowStride = lvl->rowstride / image->mt->bpp; -} - - -/** - * Map a single texture image for glTexImage and friends. - */ -void radeon_teximage_map(radeon_texture_image *image, GLboolean write_enable) -{ - if (image->mt) { - assert(!image->base.Data); - - radeon_bo_map(image->mt->bo, write_enable); - teximage_set_map_data(image); - } -} - - -void radeon_teximage_unmap(radeon_texture_image *image) -{ - if (image->mt) { - assert(image->base.Data); - - image->base.Data = 0; - radeon_bo_unmap(image->mt->bo); - } -} - -/** - * Map a validated texture for reading during software rendering. - */ -void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj) -{ - radeonTexObj* t = radeon_tex_obj(texObj); - int face, level; - - /* for r100 3D sw fallbacks don't have mt */ - if (!t->mt) - return; - - radeon_bo_map(t->mt->bo, GL_FALSE); - for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) - teximage_set_map_data(get_radeon_texture_image(texObj->Image[face][level])); - } -} - -void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj) -{ - radeonTexObj* t = radeon_tex_obj(texObj); - int face, level; - - /* for r100 3D sw fallbacks don't have mt */ - if (!t->mt) - return; - - for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) - texObj->Image[face][level]->Data = 0; - } - radeon_bo_unmap(t->mt->bo); -} - -GLuint radeon_face_for_target(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - return (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; - default: - return 0; - } -} - -/** - * Wraps Mesa's implementation to ensure that the base level image is mapped. - * - * This relies on internal details of _mesa_generate_mipmap, in particular - * the fact that the memory for recreated texture images is always freed. - */ -static void radeon_generate_mipmap(GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj) -{ - radeonTexObj* t = radeon_tex_obj(texObj); - GLuint nr_faces = (t->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - int i, face; - - - _mesa_generate_mipmap(ctx, target, texObj); - - for (face = 0; face < nr_faces; face++) { - for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) { - radeon_texture_image *image; - - image = get_radeon_texture_image(texObj->Image[face][i]); - - if (image == NULL) - break; - - image->mtlevel = i; - image->mtface = face; - - radeon_miptree_unreference(image->mt); - image->mt = NULL; - } - } - -} - -void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj) -{ - GLuint face = radeon_face_for_target(target); - radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[face][texObj->BaseLevel]); - - radeon_teximage_map(baseimage, GL_FALSE); - radeon_generate_mipmap(ctx, target, texObj); - radeon_teximage_unmap(baseimage); -} - - -/* try to find a format which will only need a memcopy */ -static const struct gl_texture_format *radeonChoose8888TexFormat(radeonContextPtr rmesa, - GLenum srcFormat, - GLenum srcType) -{ - const GLuint ui = 1; - const GLubyte littleEndian = *((const GLubyte *)&ui); - - /* r100 can only do this */ - if (IS_R100_CLASS(rmesa->radeonScreen)) - return _dri_texformat_argb8888; - - if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) || - (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) || - (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) || - (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) { - return &_mesa_texformat_rgba8888; - } else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) || - (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) || - (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) || - (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) { - return &_mesa_texformat_rgba8888_rev; - } else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && !littleEndian) || - srcType == GL_UNSIGNED_INT_8_8_8_8)) { - return &_mesa_texformat_argb8888_rev; - } else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && littleEndian) || - srcType == GL_UNSIGNED_INT_8_8_8_8_REV)) { - return &_mesa_texformat_argb8888; - } else - return _dri_texformat_argb8888; -} - -const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - const GLboolean do32bpt = - (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32); - const GLboolean force16bpt = - (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16); - (void)format; - -#if 0 - fprintf(stderr, "InternalFormat=%s(%d) type=%s format=%s\n", - _mesa_lookup_enum_by_nr(internalFormat), internalFormat, - _mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format)); - fprintf(stderr, "do32bpt=%d force16bpt=%d\n", do32bpt, force16bpt); -#endif - - switch (internalFormat) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - switch (type) { - case GL_UNSIGNED_INT_10_10_10_2: - case GL_UNSIGNED_INT_2_10_10_10_REV: - return do32bpt ? _dri_texformat_argb8888 : - _dri_texformat_argb1555; - case GL_UNSIGNED_SHORT_4_4_4_4: - case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return _dri_texformat_argb4444; - case GL_UNSIGNED_SHORT_5_5_5_1: - case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return _dri_texformat_argb1555; - default: - return do32bpt ? radeonChoose8888TexFormat(rmesa, format, type) : - _dri_texformat_argb4444; - } - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - switch (type) { - case GL_UNSIGNED_SHORT_4_4_4_4: - case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return _dri_texformat_argb4444; - case GL_UNSIGNED_SHORT_5_5_5_1: - case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return _dri_texformat_argb1555; - case GL_UNSIGNED_SHORT_5_6_5: - case GL_UNSIGNED_SHORT_5_6_5_REV: - return _dri_texformat_rgb565; - default: - return do32bpt ? _dri_texformat_argb8888 : - _dri_texformat_rgb565; - } - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return !force16bpt ? - radeonChoose8888TexFormat(rmesa, format,type) : - _dri_texformat_argb4444; - - case GL_RGBA4: - case GL_RGBA2: - return _dri_texformat_argb4444; - - case GL_RGB5_A1: - return _dri_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return !force16bpt ? _dri_texformat_argb8888 : - _dri_texformat_rgb565; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - return _dri_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - return _dri_texformat_a8; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - return _dri_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return _dri_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - return _dri_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_APPLE || - type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; - - case GL_ALPHA16F_ARB: - return &_mesa_texformat_alpha_float16; - case GL_ALPHA32F_ARB: - return &_mesa_texformat_alpha_float32; - case GL_LUMINANCE16F_ARB: - return &_mesa_texformat_luminance_float16; - case GL_LUMINANCE32F_ARB: - return &_mesa_texformat_luminance_float32; - case GL_LUMINANCE_ALPHA16F_ARB: - return &_mesa_texformat_luminance_alpha_float16; - case GL_LUMINANCE_ALPHA32F_ARB: - return &_mesa_texformat_luminance_alpha_float32; - case GL_INTENSITY16F_ARB: - return &_mesa_texformat_intensity_float16; - case GL_INTENSITY32F_ARB: - return &_mesa_texformat_intensity_float32; - case GL_RGB16F_ARB: - return &_mesa_texformat_rgba_float16; - case GL_RGB32F_ARB: - return &_mesa_texformat_rgba_float32; - case GL_RGBA16F_ARB: - return &_mesa_texformat_rgba_float16; - case GL_RGBA32F_ARB: - return &_mesa_texformat_rgba_float32; - - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: -#if 0 - switch (type) { - case GL_UNSIGNED_BYTE: - case GL_UNSIGNED_SHORT: - return &_mesa_texformat_z16; - case GL_UNSIGNED_INT: - return &_mesa_texformat_z32; - case GL_UNSIGNED_INT_24_8_EXT: - default: - return &_mesa_texformat_z24_s8; - } -#else - return &_mesa_texformat_z16; -#endif - - default: - _mesa_problem(ctx, - "unexpected internalFormat 0x%x in r300ChooseTextureFormat", - (int)internalFormat); - return NULL; - } - - return NULL; /* never get here */ -} - -/** - * All glTexImage calls go through this function. - */ -static void radeon_teximage( - GLcontext *ctx, int dims, - GLint face, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLsizei imageSize, - GLenum format, GLenum type, const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, - int compressed) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - radeonTexObj* t = radeon_tex_obj(texObj); - radeon_texture_image* image = get_radeon_texture_image(texImage); - - radeon_firevertices(rmesa); - - t->validated = GL_FALSE; - - /* Choose and fill in the texture format for this image */ - texImage->TexFormat = radeonChooseTextureFormat(ctx, internalFormat, format, type); - _mesa_set_fetch_functions(texImage, dims); - - if (texImage->TexFormat->TexelBytes == 0) { - texImage->IsCompressed = GL_TRUE; - texImage->CompressedSize = - ctx->Driver.CompressedTextureSize(ctx, texImage->Width, - texImage->Height, texImage->Depth, - texImage->TexFormat->MesaFormat); - } else { - texImage->IsCompressed = GL_FALSE; - texImage->CompressedSize = 0; - } - - /* Allocate memory for image */ - radeonFreeTexImageData(ctx, texImage); /* Mesa core only clears texImage->Data but not image->mt */ - - if (!t->mt) - radeon_try_alloc_miptree(rmesa, t, texImage, face, level); - if (t->mt && radeon_miptree_matches_image(t->mt, texImage, face, level)) { - image->mt = t->mt; - image->mtlevel = level - t->mt->firstLevel; - image->mtface = face; - radeon_miptree_reference(t->mt); - } else { - int size; - if (texImage->IsCompressed) { - size = texImage->CompressedSize; - } else { - size = texImage->Width * texImage->Height * texImage->Depth * texImage->TexFormat->TexelBytes; - } - texImage->Data = _mesa_alloc_texmemory(size); - } - - /* Upload texture image; note that the spec allows pixels to be NULL */ - if (compressed) { - pixels = _mesa_validate_pbo_compressed_teximage( - ctx, imageSize, pixels, packing, "glCompressedTexImage"); - } else { - pixels = _mesa_validate_pbo_teximage( - ctx, dims, width, height, depth, - format, type, pixels, packing, "glTexImage"); - } - - if (pixels) { - radeon_teximage_map(image, GL_TRUE); - - if (compressed) { - memcpy(texImage->Data, pixels, imageSize); - } else { - GLuint dstRowStride; - if (image->mt) { - radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; - dstRowStride = lvl->rowstride; - } else { - dstRowStride = texImage->Width * texImage->TexFormat->TexelBytes; - } - if (!texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); - } - - } - - /* SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - radeon_generate_mipmap(ctx, texObj->Target, texObj); - } - - if (pixels) - radeon_teximage_unmap(image); - - _mesa_unmap_teximage_pbo(ctx, packing); - - -} - -void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - radeon_teximage(ctx, 1, 0, level, internalFormat, width, 1, 1, - 0, format, type, pixels, packing, texObj, texImage, 0); -} - -void radeonTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) - -{ - GLuint face = radeon_face_for_target(target); - - radeon_teximage(ctx, 2, face, level, internalFormat, width, height, 1, - 0, format, type, pixels, packing, texObj, texImage, 0); -} - -void radeonCompressedTexImage2D(GLcontext * ctx, GLenum target, - GLint level, GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid * data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - GLuint face = radeon_face_for_target(target); - - radeon_teximage(ctx, 2, face, level, internalFormat, width, height, 1, - imageSize, 0, 0, data, 0, texObj, texImage, 1); -} - -void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - radeon_teximage(ctx, 3, 0, level, internalFormat, width, height, depth, - 0, format, type, pixels, packing, texObj, texImage, 0); -} - -/** - * Update a subregion of the given texture image. - */ -static void radeon_texsubimage(GLcontext* ctx, int dims, int level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, - int compressed) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - radeonTexObj* t = radeon_tex_obj(texObj); - radeon_texture_image* image = get_radeon_texture_image(texImage); - - radeon_firevertices(rmesa); - - t->validated = GL_FALSE; - pixels = _mesa_validate_pbo_teximage(ctx, dims, - width, height, depth, format, type, pixels, packing, "glTexSubImage1D"); - - if (pixels) { - GLint dstRowStride; - radeon_teximage_map(image, GL_TRUE); - - if (image->mt) { - radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; - dstRowStride = lvl->rowstride; - } else { - dstRowStride = texImage->RowStride * texImage->TexFormat->TexelBytes; - } - - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); - - - } - - /* GL_SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - radeon_generate_mipmap(ctx, texObj->Target, texObj); - } - radeon_teximage_unmap(image); - - _mesa_unmap_teximage_pbo(ctx, packing); - - -} - -void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - radeon_texsubimage(ctx, 1, level, xoffset, 0, 0, width, 1, 1, - format, type, pixels, packing, texObj, texImage, 0); -} - -void radeonTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - radeon_texsubimage(ctx, 2, level, xoffset, yoffset, 0, width, height, - 1, format, type, pixels, packing, texObj, texImage, - 0); -} - -void radeonCompressedTexSubImage2D(GLcontext * ctx, GLenum target, - GLint level, GLint xoffset, - GLint yoffset, GLsizei width, - GLsizei height, GLenum format, - GLsizei imageSize, const GLvoid * data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - radeon_texsubimage(ctx, 2, level, xoffset, yoffset, 0, width, height, 1, - format, 0, data, 0, texObj, texImage, 1); -} - - -void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - radeon_texsubimage(ctx, 3, level, xoffset, yoffset, zoffset, width, height, depth, - format, type, pixels, packing, texObj, texImage, 0); -} - -static void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride, - GLuint numrows, GLuint rowsize) -{ - assert(rowsize <= dststride); - assert(rowsize <= srcstride); - - if (rowsize == srcstride && rowsize == dststride) { - memcpy(dst, src, numrows*rowsize); - } else { - GLuint i; - for(i = 0; i < numrows; ++i) { - memcpy(dst, src, rowsize); - dst += dststride; - src += srcstride; - } - } -} - - -/** - * Ensure that the given image is stored in the given miptree from now on. - */ -static void migrate_image_to_miptree(radeon_mipmap_tree *mt, radeon_texture_image *image, int face, int level) -{ - radeon_mipmap_level *dstlvl = &mt->levels[level - mt->firstLevel]; - unsigned char *dest; - - assert(image->mt != mt); - assert(dstlvl->width == image->base.Width); - assert(dstlvl->height == image->base.Height); - assert(dstlvl->depth == image->base.Depth); - - - radeon_bo_map(mt->bo, GL_TRUE); - dest = mt->bo->ptr + dstlvl->faces[face].offset; - - if (image->mt) { - /* Format etc. should match, so we really just need a memcpy(). - * In fact, that memcpy() could be done by the hardware in many - * cases, provided that we have a proper memory manager. - */ - radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel]; - - assert(srclvl->size == dstlvl->size); - assert(srclvl->rowstride == dstlvl->rowstride); - - radeon_bo_map(image->mt->bo, GL_FALSE); - - memcpy(dest, - image->mt->bo->ptr + srclvl->faces[face].offset, - dstlvl->size); - radeon_bo_unmap(image->mt->bo); - - radeon_miptree_unreference(image->mt); - } else { - uint srcrowstride = image->base.Width * image->base.TexFormat->TexelBytes; - -// if (mt->tilebits) -// WARN_ONCE("%s: tiling not supported yet", __FUNCTION__); - - copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride, - image->base.Height * image->base.Depth, srcrowstride); - - _mesa_free_texmemory(image->base.Data); - image->base.Data = 0; - } - - radeon_bo_unmap(mt->bo); - - image->mt = mt; - image->mtface = face; - image->mtlevel = level; - radeon_miptree_reference(image->mt); -} - -int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - radeonTexObj *t = radeon_tex_obj(texObj); - radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[0][texObj->BaseLevel]); - int face, level; - - if (t->validated || t->image_override) - return GL_TRUE; - - if (RADEON_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s: Validating texture %p now\n", __FUNCTION__, texObj); - - if (baseimage->base.Border > 0) - return GL_FALSE; - - /* Ensure a matching miptree exists. - * - * Differing mipmap trees can result when the app uses TexImage to - * change texture dimensions. - * - * Prefer to use base image's miptree if it - * exists, since that most likely contains more valid data (remember - * that the base level is usually significantly larger than the rest - * of the miptree, so cubemaps are the only possible exception). - */ - if (baseimage->mt && - baseimage->mt != t->mt && - radeon_miptree_matches_texture(baseimage->mt, &t->base)) { - radeon_miptree_unreference(t->mt); - t->mt = baseimage->mt; - radeon_miptree_reference(t->mt); - } else if (t->mt && !radeon_miptree_matches_texture(t->mt, &t->base)) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } - - if (!t->mt) { - if (RADEON_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, " Allocate new miptree\n"); - radeon_try_alloc_miptree(rmesa, t, &baseimage->base, 0, texObj->BaseLevel); - if (!t->mt) { - _mesa_problem(ctx, "r300_validate_texture failed to alloc miptree"); - return GL_FALSE; - } - } - - /* Ensure all images are stored in the single main miptree */ - for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) { - radeon_texture_image *image = get_radeon_texture_image(texObj->Image[face][level]); - if (RADEON_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, " face %i, level %i... %p vs %p ", face, level, t->mt, image->mt); - if (t->mt == image->mt) { - if (RADEON_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "OK\n"); - continue; - } - - if (RADEON_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "migrating\n"); - migrate_image_to_miptree(t->mt, image, face, level); - } - } - - return GL_TRUE; -} - - -GLubyte *radeon_ptr32(const struct radeon_renderbuffer * rrb, - GLint x, GLint y) -{ - GLubyte *ptr = rrb->bo->ptr; - const __DRIdrawablePrivate *dPriv = rrb->dPriv; - uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE; - GLint offset; - GLint nmacroblkpl; - GLint nmicroblkpl; - - x += dPriv->x; - y += dPriv->y; - - if (rrb->has_surface || !(rrb->bo->flags & mask)) { - offset = x * rrb->cpp + y * rrb->pitch; - } else { - offset = 0; - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) { - if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) { - nmacroblkpl = rrb->pitch >> 5; - offset += ((y >> 4) * nmacroblkpl) << 11; - offset += ((y & 15) >> 1) << 8; - offset += (y & 1) << 4; - offset += (x >> 5) << 11; - offset += ((x & 31) >> 2) << 5; - offset += (x & 3) << 2; - } else { - nmacroblkpl = rrb->pitch >> 6; - offset += ((y >> 3) * nmacroblkpl) << 11; - offset += (y & 7) << 8; - offset += (x >> 6) << 11; - offset += ((x & 63) >> 3) << 5; - offset += (x & 7) << 2; - } - } else { - nmicroblkpl = ((rrb->pitch + 31) & ~31) >> 5; - offset += (y * nmicroblkpl) << 5; - offset += (x >> 3) << 5; - offset += (x & 7) << 2; - } - } - return &ptr[offset]; -} - -GLubyte *radeon_ptr16(const struct radeon_renderbuffer * rrb, - GLint x, GLint y) -{ - GLubyte *ptr = rrb->bo->ptr; - const __DRIdrawablePrivate *dPriv = rrb->dPriv; - uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE; - GLint offset; - GLint nmacroblkpl; - GLint nmicroblkpl; - - x += dPriv->x; - y += dPriv->y; - - if (rrb->has_surface || !(rrb->bo->flags & mask)) { - offset = x * rrb->cpp + y * rrb->pitch; - } else { - offset = 0; - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) { - if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) { - nmacroblkpl = rrb->pitch >> 6; - offset += ((y >> 4) * nmacroblkpl) << 11; - offset += ((y & 15) >> 1) << 8; - offset += (y & 1) << 4; - offset += (x >> 6) << 11; - offset += ((x & 63) >> 3) << 5; - offset += (x & 7) << 1; - } else { - nmacroblkpl = rrb->pitch >> 7; - offset += ((y >> 3) * nmacroblkpl) << 11; - offset += (y & 7) << 8; - offset += (x >> 7) << 11; - offset += ((x & 127) >> 4) << 5; - offset += (x & 15) << 2; - } - } else { - nmicroblkpl = ((rrb->pitch + 31) & ~31) >> 5; - offset += (y * nmicroblkpl) << 5; - offset += (x >> 4) << 5; - offset += (x & 15) << 2; - } - } - return &ptr[offset]; -} - -GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb, - GLint x, GLint y) -{ - GLubyte *ptr = rrb->bo->ptr; - const __DRIdrawablePrivate *dPriv = rrb->dPriv; - uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE; - GLint offset; - GLint microblkxs; - GLint macroblkxs; - GLint nmacroblkpl; - GLint nmicroblkpl; - - x += dPriv->x; - y += dPriv->y; - - if (rrb->has_surface || !(rrb->bo->flags & mask)) { - offset = x * rrb->cpp + y * rrb->pitch; - } else { - offset = 0; - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) { - if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) { - microblkxs = 16 / rrb->cpp; - macroblkxs = 128 / rrb->cpp; - nmacroblkpl = rrb->pitch / macroblkxs; - offset += ((y >> 4) * nmacroblkpl) << 11; - offset += ((y & 15) >> 1) << 8; - offset += (y & 1) << 4; - offset += (x / macroblkxs) << 11; - offset += ((x & (macroblkxs - 1)) / microblkxs) << 5; - offset += (x & (microblkxs - 1)) * rrb->cpp; - } else { - microblkxs = 32 / rrb->cpp; - macroblkxs = 256 / rrb->cpp; - nmacroblkpl = rrb->pitch / macroblkxs; - offset += ((y >> 3) * nmacroblkpl) << 11; - offset += (y & 7) << 8; - offset += (x / macroblkxs) << 11; - offset += ((x & (macroblkxs - 1)) / microblkxs) << 5; - offset += (x & (microblkxs - 1)) * rrb->cpp; - } - } else { - microblkxs = 32 / rrb->cpp; - nmicroblkpl = ((rrb->pitch + 31) & ~31) >> 5; - offset += (y * nmicroblkpl) << 5; - offset += (x / microblkxs) << 5; - offset += (x & (microblkxs - 1)) * rrb->cpp; - } - } - return &ptr[offset]; -} - - -static void map_buffer(struct gl_renderbuffer *rb, GLboolean write) -{ - struct radeon_renderbuffer *rrb = (void*)rb; - int r; - - if (rrb->bo) { - r = radeon_bo_map(rrb->bo, write); - if (r) { - fprintf(stderr, "(%s) error(%d) mapping buffer.\n", - __FUNCTION__, r); - } - } -} - -static void unmap_buffer(struct gl_renderbuffer *rb) -{ - struct radeon_renderbuffer *rrb = (void*)rb; - - if (rrb->bo) { - radeon_bo_unmap(rrb->bo); - } -} - -void radeonSpanRenderStart(GLcontext * ctx) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - int i; - - radeon_firevertices(rmesa); - - for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) - ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[i]._Current); - } - - /* color draw buffers */ - for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) { - map_buffer(ctx->DrawBuffer->_ColorDrawBuffers[i], GL_TRUE); - } - - map_buffer(ctx->ReadBuffer->_ColorReadBuffer, GL_FALSE); - - if (ctx->DrawBuffer->_DepthBuffer) { - map_buffer(ctx->DrawBuffer->_DepthBuffer->Wrapped, GL_TRUE); - } - if (ctx->DrawBuffer->_StencilBuffer) - map_buffer(ctx->DrawBuffer->_StencilBuffer->Wrapped, GL_TRUE); - - /* The locking and wait for idle should really only be needed in classic mode. - * In a future memory manager based implementation, this should become - * unnecessary due to the fact that mapping our buffers, textures, etc. - * should implicitly wait for any previous rendering commands that must - * be waited on. */ - LOCK_HARDWARE(rmesa); - radeonWaitForIdleLocked(rmesa); -} - -void radeonSpanRenderFinish(GLcontext * ctx) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - int i; - _swrast_flush(ctx); - UNLOCK_HARDWARE(rmesa); - - for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) - ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[i]._Current); - } - - /* color draw buffers */ - for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) - unmap_buffer(ctx->DrawBuffer->_ColorDrawBuffers[i]); - - unmap_buffer(ctx->ReadBuffer->_ColorReadBuffer); - - if (ctx->DrawBuffer->_DepthBuffer) - unmap_buffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); - if (ctx->DrawBuffer->_StencilBuffer) - unmap_buffer(ctx->DrawBuffer->_StencilBuffer->Wrapped); -} - -void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) -{ - struct radeon_cs_space_check bos[1]; - int flushed = 0, ret; - - size = MAX2(size, MAX_DMA_BUF_SZ * 16); - - if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (rmesa->dma.flush) { - rmesa->dma.flush(rmesa->glCtx); - } - - if (rmesa->dma.nr_released_bufs > 4) { - rcommonFlushCmdBuf(rmesa, __FUNCTION__); - rmesa->dma.nr_released_bufs = 0; - } - - if (rmesa->dma.current) { - radeon_bo_unmap(rmesa->dma.current); - radeon_bo_unref(rmesa->dma.current); - rmesa->dma.current = 0; - } - -again_alloc: - rmesa->dma.current = radeon_bo_open(rmesa->radeonScreen->bom, - 0, size, 4, RADEON_GEM_DOMAIN_GTT, - 0); - - if (!rmesa->dma.current) { - rcommonFlushCmdBuf(rmesa, __FUNCTION__); - rmesa->dma.nr_released_bufs = 0; - goto again_alloc; - } - - rmesa->dma.current_used = 0; - rmesa->dma.current_vertexptr = 0; - - bos[0].bo = rmesa->dma.current; - bos[0].read_domains = RADEON_GEM_DOMAIN_GTT; - bos[0].write_domain =0 ; - bos[0].new_accounted = 0; - - ret = radeon_cs_space_check(rmesa->cmdbuf.cs, bos, 1); - if (ret == RADEON_CS_SPACE_OP_TO_BIG) { - fprintf(stderr,"Got OPEARTION TO BIG ILLEGAL - this cannot happen"); - assert(0); - } else if (ret == RADEON_CS_SPACE_FLUSH) { - rcommonFlushCmdBuf(rmesa, __FUNCTION__); - if (flushed) { - fprintf(stderr,"flushed but still no space\n"); - assert(0); - } - flushed = 1; - goto again_alloc; - } - radeon_bo_map(rmesa->dma.current, 1); -} - -/* Allocates a region from rmesa->dma.current. If there isn't enough - * space in current, grab a new buffer (and discard what was left of current) - */ -void radeonAllocDmaRegion(radeonContextPtr rmesa, - struct radeon_bo **pbo, int *poffset, - int bytes, int alignment) -{ - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s %d\n", __FUNCTION__, bytes); - - if (rmesa->dma.flush) - rmesa->dma.flush(rmesa->glCtx); - - assert(rmesa->dma.current_used == rmesa->dma.current_vertexptr); - - alignment--; - rmesa->dma.current_used = (rmesa->dma.current_used + alignment) & ~alignment; - - if (!rmesa->dma.current || rmesa->dma.current_used + bytes > rmesa->dma.current->size) - radeonRefillCurrentDmaRegion(rmesa, (bytes + 15) & ~15); - - *poffset = rmesa->dma.current_used; - *pbo = rmesa->dma.current; - radeon_bo_ref(*pbo); - - /* Always align to at least 16 bytes */ - rmesa->dma.current_used = (rmesa->dma.current_used + bytes + 15) & ~15; - rmesa->dma.current_vertexptr = rmesa->dma.current_used; - - assert(rmesa->dma.current_used <= rmesa->dma.current->size); -} - -void radeonReleaseDmaRegion(radeonContextPtr rmesa) -{ - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s %p\n", __FUNCTION__, rmesa->dma.current); - if (rmesa->dma.current) { - rmesa->dma.nr_released_bufs++; - radeon_bo_unmap(rmesa->dma.current); - radeon_bo_unref(rmesa->dma.current); - } - rmesa->dma.current = NULL; -} - - -/* Flush vertices in the current dma region. - */ -void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - struct radeon_dma *dma = &rmesa->dma; - - - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - dma->flush = NULL; - - if (dma->current) { - GLuint current_offset = dma->current_used; - - assert (dma->current_used + - rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == - dma->current_vertexptr); - - if (dma->current_used != dma->current_vertexptr) { - dma->current_used = dma->current_vertexptr; - - rmesa->vtbl.swtcl_flush(ctx, current_offset); - } - rmesa->swtcl.numverts = 0; - } -} -/* Alloc space in the current dma region. - */ -void * -rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) -{ - GLuint bytes = vsize * nverts; - void *head; - - if (!rmesa->dma.current || rmesa->dma.current_vertexptr + bytes > rmesa->dma.current->size) { - radeonRefillCurrentDmaRegion(rmesa, bytes); - } - - if (!rmesa->dma.flush) { - rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; - rmesa->dma.flush = rcommon_flush_last_swtcl_prim; - } - - ASSERT( vsize == rmesa->swtcl.vertex_size * 4 ); - ASSERT( rmesa->dma.flush == rcommon_flush_last_swtcl_prim ); - ASSERT( rmesa->dma.current_used + - rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == - rmesa->dma.current_vertexptr ); - -// fprintf(stderr,"current %p %x\n", rmesa->radeon.dma.current->ptr, -// rmesa->radeon.dma.current_vertexptr); - head = (rmesa->dma.current->ptr + rmesa->dma.current_vertexptr); - rmesa->dma.current_vertexptr += bytes; - rmesa->swtcl.numverts += nverts; - return head; -} diff --git a/src/mesa/drivers/dri/radeon/common_misc.h b/src/mesa/drivers/dri/radeon/common_misc.h deleted file mode 100644 index ebf287a031..0000000000 --- a/src/mesa/drivers/dri/radeon/common_misc.h +++ /dev/null @@ -1,166 +0,0 @@ -#ifndef COMMON_MISC_H -#define COMMON_MISC_H - -#include "common_context.h" -#include "radeon_buffer.h" -void radeonRecalcScissorRects(radeonContextPtr radeon); -void radeonSetCliprects(radeonContextPtr radeon); -void radeonUpdateScissor( GLcontext *ctx ); -void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h); - -void radeonWaitForIdleLocked(radeonContextPtr radeon); -extern uint32_t radeonGetAge(radeonContextPtr radeon); -void radeonCopyBuffer( __DRIdrawablePrivate *dPriv, - const drm_clip_rect_t *rect); -void radeonPageFlip( __DRIdrawablePrivate *dPriv ); -void radeon_common_finish(GLcontext * ctx); -void radeonSwapBuffers(__DRIdrawablePrivate * dPriv); -void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv, - int x, int y, int w, int h ); - -void radeonUpdatePageFlipping(radeonContextPtr rmesa); - -GLboolean radeonInitContext(radeonContextPtr radeon, - struct dd_function_table* functions, - const __GLcontextModes * glVisual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - -void radeonCleanupContext(radeonContextPtr radeon); -GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv); -void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable); -GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); - -void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count); -void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count); - -void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, - GLvoid * data, int size, int stride, int count); -void radeon_print_state_atom( struct radeon_state_atom *state ); - -struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx); -void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage); - -void radeon_teximage_map(radeon_texture_image *image, GLboolean write_enable); -void radeon_teximage_unmap(radeon_texture_image *image); -void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj); -void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj); -void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj); -int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj); -GLuint radeon_face_for_target(GLenum target); -const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type); - -void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); -void radeonTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); -void radeonCompressedTexImage2D(GLcontext * ctx, GLenum target, - GLint level, GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid * data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); -void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); -void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); -void radeonTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); -void radeonCompressedTexSubImage2D(GLcontext * ctx, GLenum target, - GLint level, GLint xoffset, - GLint yoffset, GLsizei width, - GLsizei height, GLenum format, - GLsizei imageSize, const GLvoid * data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void radeonSpanRenderStart(GLcontext * ctx); -void radeonSpanRenderFinish(GLcontext * ctx); -GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb, - GLint x, GLint y); -GLubyte *radeon_ptr16(const struct radeon_renderbuffer * rrb, - GLint x, GLint y); -GLubyte *radeon_ptr32(const struct radeon_renderbuffer * rrb, - GLint x, GLint y); -void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size); -void radeonAllocDmaRegion(radeonContextPtr rmesa, - struct radeon_bo **pbo, int *poffset, - int bytes, int alignment); -void radeonReleaseDmaRegion(radeonContextPtr rmesa); - -void rcommon_flush_last_swtcl_prim(GLcontext *ctx); - -void *rcommonAllocDmaLowVerts(radeonContextPtr rmesa, int nverts, int vsize); - -void radeonFlush(GLcontext *ctx); -void radeonFinish(GLcontext * ctx); -void radeonEmitState(radeonContextPtr radeon); - -static inline struct radeon_renderbuffer *radeon_get_depthbuffer(radeonContextPtr rmesa) -{ - struct radeon_renderbuffer *rrb; - rrb = rmesa->state.depth.rrb; - if (!rrb) - return NULL; - - return rrb; -} - -static inline struct radeon_renderbuffer *radeon_get_colorbuffer(radeonContextPtr rmesa) -{ - struct radeon_renderbuffer *rrb; - GLframebuffer *fb = rmesa->dri.drawable->driverPrivate; - - rrb = rmesa->state.color.rrb; - if (rmesa->radeonScreen->driScreen->dri2.enabled) { - rrb = (struct radeon_renderbuffer *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; - } - if (!rrb) - return NULL; - return rrb; -} - - -#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c index a716f1059b..b7658353af 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c @@ -47,7 +47,7 @@ #include "radeon_drm.h" #include "radeon_bo.h" #include "radeon_bo_legacy.h" -#include "common_context.h" +#include "radeon_common.h" struct bo_legacy { struct radeon_bo base; diff --git a/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h b/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h new file mode 100644 index 0000000000..5526934209 --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h @@ -0,0 +1,143 @@ +#ifndef COMMON_CMDBUF_H +#define COMMON_CMDBUF_H + +#include "radeon_cs.h" + +void rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller); +int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller); +int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller); +void rcommonInitCmdBuf(radeonContextPtr rmesa); +void rcommonDestroyCmdBuf(radeonContextPtr rmesa); + +void rcommonBeginBatch(radeonContextPtr rmesa, + int n, + int dostate, + const char *file, + const char *function, + int line); + +#define RADEON_CP_PACKET3_NOP 0xC0001000 +#define RADEON_CP_PACKET3_NEXT_CHAR 0xC0001900 +#define RADEON_CP_PACKET3_PLY_NEXTSCAN 0xC0001D00 +#define RADEON_CP_PACKET3_SET_SCISSORS 0xC0001E00 +#define RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM 0xC0002300 +#define RADEON_CP_PACKET3_LOAD_MICROCODE 0xC0002400 +#define RADEON_CP_PACKET3_WAIT_FOR_IDLE 0xC0002600 +#define RADEON_CP_PACKET3_3D_DRAW_VBUF 0xC0002800 +#define RADEON_CP_PACKET3_3D_DRAW_IMMD 0xC0002900 +#define RADEON_CP_PACKET3_3D_DRAW_INDX 0xC0002A00 +#define RADEON_CP_PACKET3_LOAD_PALETTE 0xC0002C00 +#define RADEON_CP_PACKET3_3D_LOAD_VBPNTR 0xC0002F00 +#define RADEON_CP_PACKET3_CNTL_PAINT 0xC0009100 +#define RADEON_CP_PACKET3_CNTL_BITBLT 0xC0009200 +#define RADEON_CP_PACKET3_CNTL_SMALLTEXT 0xC0009300 +#define RADEON_CP_PACKET3_CNTL_HOSTDATA_BLT 0xC0009400 +#define RADEON_CP_PACKET3_CNTL_POLYLINE 0xC0009500 +#define RADEON_CP_PACKET3_CNTL_POLYSCANLINES 0xC0009800 +#define RADEON_CP_PACKET3_CNTL_PAINT_MULTI 0xC0009A00 +#define RADEON_CP_PACKET3_CNTL_BITBLT_MULTI 0xC0009B00 +#define RADEON_CP_PACKET3_CNTL_TRANS_BITBLT 0xC0009C00 + +#define CP_PACKET2 (2 << 30) +#define CP_PACKET0(reg, n) (RADEON_CP_PACKET0 | ((n)<<16) | ((reg)>>2)) +#define CP_PACKET0_ONE(reg, n) (RADEON_CP_PACKET0 | RADEON_CP_PACKET0_ONE_REG_WR | ((n)<<16) | ((reg)>>2)) +#define CP_PACKET3( pkt, n ) \ + (RADEON_CP_PACKET3 | (pkt) | ((n) << 16)) + +/** + * Every function writing to the command buffer needs to declare this + * to get the necessary local variables. + */ +#define BATCH_LOCALS(rmesa) \ + const radeonContextPtr b_l_rmesa = rmesa + +/** + * Prepare writing n dwords to the command buffer, + * including producing any necessary state emits on buffer wraparound. + */ +#define BEGIN_BATCH(n) rcommonBeginBatch(b_l_rmesa, n, 1, __FILE__, __FUNCTION__, __LINE__) + +/** + * Same as BEGIN_BATCH, but do not cause automatic state emits. + */ +#define BEGIN_BATCH_NO_AUTOSTATE(n) rcommonBeginBatch(b_l_rmesa, n, 0, __FILE__, __FUNCTION__, __LINE__) + +/** + * Write one dword to the command buffer. + */ +#define OUT_BATCH(data) \ + do { \ + radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, data);\ + } while(0) + +/** + * Write a relocated dword to the command buffer. + */ +#define OUT_BATCH_RELOC(data, bo, offset, rd, wd, flags) \ + do { \ + if (0 && offset) { \ + fprintf(stderr, "(%s:%s:%d) offset : %d\n", \ + __FILE__, __FUNCTION__, __LINE__, offset); \ + } \ + radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, offset); \ + radeon_cs_write_reloc(b_l_rmesa->cmdbuf.cs, \ + bo, rd, wd, flags); \ + if (!b_l_rmesa->radeonScreen->kernel_mm) \ + b_l_rmesa->cmdbuf.cs->section_cdw += 2; \ + } while(0) + + +/** + * Write n dwords from ptr to the command buffer. + */ +#define OUT_BATCH_TABLE(ptr,n) \ + do { \ + int _i; \ + for (_i=0; _i < n; _i++) {\ + radeon_cs_write_dword(b_l_rmesa->cmdbuf.cs, ptr[_i]);\ + }\ + } while(0) + +/** + * Finish writing dwords to the command buffer. + * The number of (direct or indirect) OUT_BATCH calls between the previous + * BEGIN_BATCH and END_BATCH must match the number specified at BEGIN_BATCH time. + */ +#define END_BATCH() \ + do { \ + radeon_cs_end(b_l_rmesa->cmdbuf.cs, __FILE__, __FUNCTION__, __LINE__);\ + } while(0) + +/** + * After the last END_BATCH() of rendering, this indicates that flushing + * the command buffer now is okay. + */ +#define COMMIT_BATCH() \ + do { \ + } while(0) + + +/** Single register write to command buffer; requires 2 dwords. */ +#define OUT_BATCH_REGVAL(reg, val) \ + OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), 1)); \ + OUT_BATCH((val)) + +/** Continuous register range write to command buffer; requires 1 dword, + * expects count dwords afterwards for register contents. */ +#define OUT_BATCH_REGSEQ(reg, count) \ + OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), (count))); + +/** Write a 32 bit float to the ring; requires 1 dword. */ +#define OUT_BATCH_FLOAT32(f) \ + OUT_BATCH(radeonPackFloat32((f))); + + +/* Fire the buffered vertices no matter what. + */ +static INLINE void radeon_firevertices(radeonContextPtr radeon) +{ + if (radeon->cmdbuf.cs->cdw || radeon->dma.flush ) + radeonFlush(radeon->glCtx); +} + +#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c new file mode 100644 index 0000000000..87b6dac40b --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -0,0 +1,843 @@ +/************************************************************************** + +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +/* + - Scissor implementation + - buffer swap/copy ioctls + - finish/flush + - state emission + - cmdbuffer management +*/ + +#include +#include "main/glheader.h" +#include "main/imports.h" +#include "main/context.h" +#include "main/api_arrayelt.h" +#include "main/enums.h" +#include "main/colormac.h" +#include "main/light.h" +#include "main/framebuffer.h" +#include "main/simple_list.h" + +#include "swrast/swrast.h" +#include "vbo/vbo.h" +#include "tnl/tnl.h" +#include "tnl/t_pipeline.h" +#include "swrast_setup/swrast_setup.h" + +#include "dri_util.h" +#include "drirenderbuffer.h" +#include "vblank.h" + +#include "radeon_common.h" +#include "radeon_common.h" +#include "radeon_bo.h" +#include "radeon_cs.h" +#include "radeon_bo_legacy.h" +#include "radeon_cs_legacy.h" +#include "radeon_bo_gem.h" +#include "radeon_cs_gem.h" +#include "radeon_drm.h" +#include "radeon_mipmap_tree.h" + +#define DEBUG_CMDBUF 0 + +/* ============================================================= + * Scissoring + */ + +static GLboolean intersect_rect(drm_clip_rect_t * out, + drm_clip_rect_t * a, drm_clip_rect_t * b) +{ + *out = *a; + if (b->x1 > out->x1) + out->x1 = b->x1; + if (b->y1 > out->y1) + out->y1 = b->y1; + if (b->x2 < out->x2) + out->x2 = b->x2; + if (b->y2 < out->y2) + out->y2 = b->y2; + if (out->x1 >= out->x2) + return GL_FALSE; + if (out->y1 >= out->y2) + return GL_FALSE; + return GL_TRUE; +} + +void radeonRecalcScissorRects(radeonContextPtr radeon) +{ + drm_clip_rect_t *out; + int i; + + /* Grow cliprect store? + */ + if (radeon->state.scissor.numAllocedClipRects < radeon->numClipRects) { + while (radeon->state.scissor.numAllocedClipRects < + radeon->numClipRects) { + radeon->state.scissor.numAllocedClipRects += 1; /* zero case */ + radeon->state.scissor.numAllocedClipRects *= 2; + } + + if (radeon->state.scissor.pClipRects) + FREE(radeon->state.scissor.pClipRects); + + radeon->state.scissor.pClipRects = + MALLOC(radeon->state.scissor.numAllocedClipRects * + sizeof(drm_clip_rect_t)); + + if (radeon->state.scissor.pClipRects == NULL) { + radeon->state.scissor.numAllocedClipRects = 0; + return; + } + } + + out = radeon->state.scissor.pClipRects; + radeon->state.scissor.numClipRects = 0; + + for (i = 0; i < radeon->numClipRects; i++) { + if (intersect_rect(out, + &radeon->pClipRects[i], + &radeon->state.scissor.rect)) { + radeon->state.scissor.numClipRects++; + out++; + } + } +} + +/** + * Update cliprects and scissors. + */ +void radeonSetCliprects(radeonContextPtr radeon) +{ + __DRIdrawablePrivate *const drawable = radeon->dri.drawable; + __DRIdrawablePrivate *const readable = radeon->dri.readable; + GLframebuffer *const draw_fb = (GLframebuffer*)drawable->driverPrivate; + GLframebuffer *const read_fb = (GLframebuffer*)readable->driverPrivate; + + if (!radeon->radeonScreen->driScreen->dri2.enabled) { + if (draw_fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) { + /* Can't ignore 2d windows if we are page flipping. */ + if (drawable->numBackClipRects == 0 || radeon->doPageFlip || + radeon->sarea->pfCurrentPage == 1) { + radeon->numClipRects = drawable->numClipRects; + radeon->pClipRects = drawable->pClipRects; + } else { + radeon->numClipRects = drawable->numBackClipRects; + radeon->pClipRects = drawable->pBackClipRects; + } + } else { + /* front buffer (or none, or multiple buffers */ + radeon->numClipRects = drawable->numClipRects; + radeon->pClipRects = drawable->pClipRects; + } + } + + if ((draw_fb->Width != drawable->w) || + (draw_fb->Height != drawable->h)) { + _mesa_resize_framebuffer(radeon->glCtx, draw_fb, + drawable->w, drawable->h); + draw_fb->Initialized = GL_TRUE; + } + + if (drawable != readable) { + if ((read_fb->Width != readable->w) || + (read_fb->Height != readable->h)) { + _mesa_resize_framebuffer(radeon->glCtx, read_fb, + readable->w, readable->h); + read_fb->Initialized = GL_TRUE; + } + } + + if (radeon->state.scissor.enabled) + radeonRecalcScissorRects(radeon); + + radeon->lastStamp = drawable->lastStamp; +} + +void radeonUpdateScissor( GLcontext *ctx ) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + + if ( rmesa->dri.drawable ) { + __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; + + int x = ctx->Scissor.X; + int y = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height; + int w = ctx->Scissor.X + ctx->Scissor.Width - 1; + int h = dPriv->h - ctx->Scissor.Y - 1; + + rmesa->state.scissor.rect.x1 = x + dPriv->x; + rmesa->state.scissor.rect.y1 = y + dPriv->y; + rmesa->state.scissor.rect.x2 = w + dPriv->x + 1; + rmesa->state.scissor.rect.y2 = h + dPriv->y + 1; + + radeonRecalcScissorRects( rmesa ); + } +} + +/* ============================================================= + * Scissoring + */ + +void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + if (ctx->Scissor.Enabled) { + /* We don't pipeline cliprect changes */ + radeon_firevertices(radeon); + radeonUpdateScissor(ctx); + } +} + + +/* ================================================================ + * SwapBuffers with client-side throttling + */ + +static uint32_t radeonGetLastFrame(radeonContextPtr radeon) +{ + drm_radeon_getparam_t gp; + int ret; + uint32_t frame; + + gp.param = RADEON_PARAM_LAST_FRAME; + gp.value = (int *)&frame; + ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM, + &gp, sizeof(gp)); + if (ret) { + fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, + ret); + exit(1); + } + + return frame; +} + +uint32_t radeonGetAge(radeonContextPtr radeon) +{ + drm_radeon_getparam_t gp; + int ret; + uint32_t age; + + gp.param = RADEON_PARAM_LAST_CLEAR; + gp.value = (int *)&age; + ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM, + &gp, sizeof(gp)); + if (ret) { + fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, + ret); + exit(1); + } + + return age; +} + +static void radeonEmitIrqLocked(radeonContextPtr radeon) +{ + drm_radeon_irq_emit_t ie; + int ret; + + ie.irq_seq = &radeon->iw.irq_seq; + ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_IRQ_EMIT, + &ie, sizeof(ie)); + if (ret) { + fprintf(stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__, + ret); + exit(1); + } +} + +static void radeonWaitIrq(radeonContextPtr radeon) +{ + int ret; + + do { + ret = drmCommandWrite(radeon->dri.fd, DRM_RADEON_IRQ_WAIT, + &radeon->iw, sizeof(radeon->iw)); + } while (ret && (errno == EINTR || errno == EBUSY)); + + if (ret) { + fprintf(stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, + ret); + exit(1); + } +} + +static void radeonWaitForFrameCompletion(radeonContextPtr radeon) +{ + drm_radeon_sarea_t *sarea = radeon->sarea; + + if (radeon->do_irqs) { + if (radeonGetLastFrame(radeon) < sarea->last_frame) { + if (!radeon->irqsEmitted) { + while (radeonGetLastFrame(radeon) < + sarea->last_frame) ; + } else { + UNLOCK_HARDWARE(radeon); + radeonWaitIrq(radeon); + LOCK_HARDWARE(radeon); + } + radeon->irqsEmitted = 10; + } + + if (radeon->irqsEmitted) { + radeonEmitIrqLocked(radeon); + radeon->irqsEmitted--; + } + } else { + while (radeonGetLastFrame(radeon) < sarea->last_frame) { + UNLOCK_HARDWARE(radeon); + if (radeon->do_usleeps) + DO_USLEEP(1); + LOCK_HARDWARE(radeon); + } + } +} + +/* wait for idle */ +void radeonWaitForIdleLocked(radeonContextPtr radeon) +{ + int ret; + int i = 0; + + do { + ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_CP_IDLE); + if (ret) + DO_USLEEP(1); + } while (ret && ++i < 100); + + if (ret < 0) { + UNLOCK_HARDWARE(radeon); + fprintf(stderr, "Error: R300 timed out... exiting\n"); + exit(-1); + } +} + +static void radeonWaitForIdle(radeonContextPtr radeon) +{ + LOCK_HARDWARE(radeon); + radeonWaitForIdleLocked(radeon); + UNLOCK_HARDWARE(radeon); +} + + +/* Copy the back color buffer to the front color buffer. + */ +void radeonCopyBuffer( __DRIdrawablePrivate *dPriv, + const drm_clip_rect_t *rect) +{ + radeonContextPtr rmesa; + GLint nbox, i, ret; + GLboolean missed_target; + int64_t ust; + __DRIscreenPrivate *psp; + + assert(dPriv); + assert(dPriv->driContextPriv); + assert(dPriv->driContextPriv->driverPrivate); + + rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; + + if ( RADEON_DEBUG & DEBUG_IOCTL ) { + fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *) rmesa->glCtx ); + } + + radeon_firevertices(rmesa); + LOCK_HARDWARE( rmesa ); + + /* Throttle the frame rate -- only allow one pending swap buffers + * request at a time. + */ + radeonWaitForFrameCompletion( rmesa ); + if (!rect) + { + UNLOCK_HARDWARE( rmesa ); + driWaitForVBlank( dPriv, & missed_target ); + LOCK_HARDWARE( rmesa ); + } + + nbox = dPriv->numClipRects; /* must be in locked region */ + + for ( i = 0 ; i < nbox ; ) { + GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox ); + drm_clip_rect_t *box = dPriv->pClipRects; + drm_clip_rect_t *b = rmesa->sarea->boxes; + GLint n = 0; + + for ( ; i < nr ; i++ ) { + + *b = box[i]; + + if (rect) + { + if (rect->x1 > b->x1) + b->x1 = rect->x1; + if (rect->y1 > b->y1) + b->y1 = rect->y1; + if (rect->x2 < b->x2) + b->x2 = rect->x2; + if (rect->y2 < b->y2) + b->y2 = rect->y2; + + if (b->x1 >= b->x2 || b->y1 >= b->y2) + continue; + } + + b++; + n++; + } + rmesa->sarea->nbox = n; + + if (!n) + continue; + + ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP ); + + if ( ret ) { + fprintf( stderr, "DRM_RADEON_SWAP_BUFFERS: return = %d\n", ret ); + UNLOCK_HARDWARE( rmesa ); + exit( 1 ); + } + } + + UNLOCK_HARDWARE( rmesa ); + if (!rect) + { + psp = dPriv->driScreenPriv; + rmesa->swap_count++; + (*psp->systemTime->getUST)( & ust ); + if ( missed_target ) { + rmesa->swap_missed_count++; + rmesa->swap_missed_ust = ust - rmesa->swap_ust; + } + + rmesa->swap_ust = ust; + rmesa->hw.all_dirty = GL_TRUE; + + } +} + +void radeonPageFlip( __DRIdrawablePrivate *dPriv ) +{ + radeonContextPtr rmesa; + GLint ret; + GLboolean missed_target; + __DRIscreenPrivate *psp; + struct radeon_renderbuffer *rrb; + GLframebuffer *fb = dPriv->driverPrivate; + + assert(dPriv); + assert(dPriv->driContextPriv); + assert(dPriv->driContextPriv->driverPrivate); + + rmesa = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; + rrb = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; + + psp = dPriv->driScreenPriv; + + if ( RADEON_DEBUG & DEBUG_IOCTL ) { + fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__, + rmesa->sarea->pfCurrentPage); + } + + radeon_firevertices(rmesa); + + LOCK_HARDWARE( rmesa ); + + if (!dPriv->numClipRects) { + UNLOCK_HARDWARE(rmesa); + usleep(10000); /* throttle invisible client 10ms */ + return; + } + + drm_clip_rect_t *box = dPriv->pClipRects; + drm_clip_rect_t *b = rmesa->sarea->boxes; + b[0] = box[0]; + rmesa->sarea->nbox = 1; + + /* Throttle the frame rate -- only allow a few pending swap buffers + * request at a time. + */ + radeonWaitForFrameCompletion( rmesa ); + UNLOCK_HARDWARE( rmesa ); + driWaitForVBlank( dPriv, & missed_target ); + if ( missed_target ) { + rmesa->swap_missed_count++; + (void) (*psp->systemTime->getUST)( & rmesa->swap_missed_ust ); + } + LOCK_HARDWARE( rmesa ); + + ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_FLIP ); + + UNLOCK_HARDWARE( rmesa ); + + if ( ret ) { + fprintf( stderr, "DRM_RADEON_FLIP: return = %d\n", ret ); + exit( 1 ); + } + + rmesa->swap_count++; + (void) (*psp->systemTime->getUST)( & rmesa->swap_ust ); + + /* Get ready for drawing next frame. Update the renderbuffers' + * flippedOffset/Pitch fields so we draw into the right place. + */ + driFlipRenderbuffers(rmesa->glCtx->WinSysDrawBuffer, + rmesa->sarea->pfCurrentPage); + + rmesa->state.color.rrb = rrb; + + if (rmesa->vtbl.update_draw_buffer) + rmesa->vtbl.update_draw_buffer(rmesa->glCtx); +} + + +/** + * Swap front and back buffer. + */ +void radeonSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + radeonContextPtr radeon; + GLcontext *ctx; + + radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; + ctx = radeon->glCtx; + + if (ctx->Visual.doubleBufferMode) { + _mesa_notifySwapBuffers(ctx);/* flush pending rendering comands */ + if (radeon->doPageFlip) { + radeonPageFlip(dPriv); + } else { + radeonCopyBuffer(dPriv, NULL); + } + } + } else { + /* XXX this shouldn't be an error but we can't handle it for now */ + _mesa_problem(NULL, "%s: drawable has no context!", + __FUNCTION__); + } +} + +void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv, + int x, int y, int w, int h ) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + radeonContextPtr radeon; + GLcontext *ctx; + + radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; + ctx = radeon->glCtx; + + if (ctx->Visual.doubleBufferMode) { + drm_clip_rect_t rect; + rect.x1 = x + dPriv->x; + rect.y1 = (dPriv->h - y - h) + dPriv->y; + rect.x2 = rect.x1 + w; + rect.y2 = rect.y1 + h; + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + radeonCopyBuffer(dPriv, &rect); + } + } else { + /* XXX this shouldn't be an error but we can't handle it for now */ + _mesa_problem(NULL, "%s: drawable has no context!", + __FUNCTION__); + } +} + + +static void radeon_print_state_atom( struct radeon_state_atom *state ) +{ + int i; + + fprintf(stderr, "emit %s/%d\n", state->name, state->cmd_size); + + if (RADEON_DEBUG & DEBUG_VERBOSE) + for (i = 0 ; i < state->cmd_size ; i++) + fprintf(stderr, "\t%s[%d]: %x\n", state->name, i, state->cmd[i]); + +} + +static INLINE void radeonEmitAtoms(radeonContextPtr radeon, GLboolean dirty) +{ + BATCH_LOCALS(radeon); + struct radeon_state_atom *atom; + int dwords; + + if (radeon->vtbl.pre_emit_atoms) + radeon->vtbl.pre_emit_atoms(radeon); + + /* Emit actual atoms */ + foreach(atom, &radeon->hw.atomlist) { + if ((atom->dirty || radeon->hw.all_dirty) == dirty) { + dwords = (*atom->check) (radeon->glCtx, atom); + if (dwords) { + if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { + radeon_print_state_atom(atom); + } + if (atom->emit) { + (*atom->emit)(radeon->glCtx, atom); + } else { + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_BATCH_TABLE(atom->cmd, dwords); + END_BATCH(); + } + atom->dirty = GL_FALSE; + } else { + if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { + fprintf(stderr, " skip state %s\n", + atom->name); + } + } + } + } + + COMMIT_BATCH(); +} + +void radeonEmitState(radeonContextPtr radeon) +{ + if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (radeon->vtbl.pre_emit_state) + radeon->vtbl.pre_emit_state(radeon); + + /* this code used to return here but now it emits zbs */ + if (radeon->cmdbuf.cs->cdw && !radeon->hw.is_dirty && !radeon->hw.all_dirty) + return; + + /* To avoid going across the entire set of states multiple times, just check + * for enough space for the case of emitting all state, and inline the + * radeonAllocCmdBuf code here without all the checks. + */ + rcommonEnsureCmdBufSpace(radeon, radeon->hw.max_state_size, __FUNCTION__); + + if (!radeon->cmdbuf.cs->cdw) { + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin reemit state\n"); + + radeonEmitAtoms(radeon, GL_FALSE); + } + + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin dirty state\n"); + + radeonEmitAtoms(radeon, GL_TRUE); + radeon->hw.is_dirty = GL_FALSE; + radeon->hw.all_dirty = GL_FALSE; + +} + + +void radeonFlush(GLcontext *ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (radeon->dma.flush) + radeon->dma.flush( ctx ); + + radeonEmitState(radeon); + + if (radeon->cmdbuf.cs->cdw) + rcommonFlushCmdBuf(radeon, __FUNCTION__); +} + +/* Make sure all commands have been sent to the hardware and have + * completed processing. + */ +void radeonFinish(GLcontext * ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + struct gl_framebuffer *fb = ctx->DrawBuffer; + int i; + + radeonFlush(ctx); + + if (radeon->radeonScreen->kernel_mm) { + for (i = 0; i < fb->_NumColorDrawBuffers; i++) { + struct radeon_renderbuffer *rrb; + rrb = (struct radeon_renderbuffer *)fb->_ColorDrawBuffers[i]; + if (rrb->bo) + radeon_bo_wait(rrb->bo); + } + } else if (radeon->do_irqs) { + LOCK_HARDWARE(radeon); + radeonEmitIrqLocked(radeon); + UNLOCK_HARDWARE(radeon); + radeonWaitIrq(radeon); + } else { + radeonWaitForIdle(radeon); + } +} + +/* cmdbuffer */ +/** + * Send the current command buffer via ioctl to the hardware. + */ +int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller) +{ + int ret = 0; + + if (rmesa->cmdbuf.flushing) { + fprintf(stderr, "Recursive call into r300FlushCmdBufLocked!\n"); + exit(-1); + } + rmesa->cmdbuf.flushing = 1; + if (rmesa->cmdbuf.cs->cdw) { + ret = radeon_cs_emit(rmesa->cmdbuf.cs); + rmesa->hw.all_dirty = GL_TRUE; + } + radeon_cs_erase(rmesa->cmdbuf.cs); + rmesa->cmdbuf.flushing = 0; + return ret; +} + +int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller) +{ + int ret; + + radeonReleaseDmaRegion(rmesa); + + LOCK_HARDWARE(rmesa); + ret = rcommonFlushCmdBufLocked(rmesa, caller); + UNLOCK_HARDWARE(rmesa); + + if (ret) { + fprintf(stderr, "drmRadeonCmdBuffer: %d\n", ret); + _mesa_exit(ret); + } + + return ret; +} + +/** + * Make sure that enough space is available in the command buffer + * by flushing if necessary. + * + * \param dwords The number of dwords we need to be free on the command buffer + */ +void rcommonEnsureCmdBufSpace(radeonContextPtr rmesa, int dwords, const char *caller) +{ + if ((rmesa->cmdbuf.cs->cdw + dwords + 128) > rmesa->cmdbuf.size || + radeon_cs_need_flush(rmesa->cmdbuf.cs)) { + rcommonFlushCmdBuf(rmesa, caller); + } +} + +void rcommonInitCmdBuf(radeonContextPtr rmesa) +{ + GLuint size; + /* Initialize command buffer */ + size = 256 * driQueryOptioni(&rmesa->optionCache, + "command_buffer_size"); + if (size < 2 * rmesa->hw.max_state_size) { + size = 2 * rmesa->hw.max_state_size + 65535; + } + if (size > 64 * 256) + size = 64 * 256; + + if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) { + fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%zd\n", + sizeof(drm_r300_cmd_header_t)); + fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%zd\n", + sizeof(drm_radeon_cmd_buffer_t)); + fprintf(stderr, + "Allocating %d bytes command buffer (max state is %d bytes)\n", + size * 4, rmesa->hw.max_state_size * 4); + } + + if (rmesa->radeonScreen->kernel_mm) { + int fd = rmesa->radeonScreen->driScreen->fd; + rmesa->cmdbuf.csm = radeon_cs_manager_gem_ctor(fd); + } else { + rmesa->cmdbuf.csm = radeon_cs_manager_legacy_ctor(rmesa); + } + if (rmesa->cmdbuf.csm == NULL) { + /* FIXME: fatal error */ + return; + } + rmesa->cmdbuf.cs = radeon_cs_create(rmesa->cmdbuf.csm, size); + assert(rmesa->cmdbuf.cs != NULL); + rmesa->cmdbuf.size = size; + + if (!rmesa->radeonScreen->kernel_mm) { + radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, rmesa->radeonScreen->texSize[0]); + radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, rmesa->radeonScreen->gartTextures.size); + } else { + struct drm_radeon_gem_info mminfo; + + if (!drmCommandWriteRead(rmesa->dri.fd, DRM_RADEON_GEM_INFO, &mminfo, sizeof(mminfo))) + { + radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, mminfo.vram_size); + radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, mminfo.gart_size); + } + } + +} +/** + * Destroy the command buffer + */ +void rcommonDestroyCmdBuf(radeonContextPtr rmesa) +{ + radeon_cs_destroy(rmesa->cmdbuf.cs); + if (rmesa->radeonScreen->driScreen->dri2.enabled || rmesa->radeonScreen->kernel_mm) { + radeon_cs_manager_gem_dtor(rmesa->cmdbuf.csm); + } else { + radeon_cs_manager_legacy_dtor(rmesa->cmdbuf.csm); + } +} + +void rcommonBeginBatch(radeonContextPtr rmesa, int n, + int dostate, + const char *file, + const char *function, + int line) +{ + rcommonEnsureCmdBufSpace(rmesa, n, function); + if (!rmesa->cmdbuf.cs->cdw && dostate) { + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "Reemit state after flush (from %s)\n", function); + radeonEmitState(rmesa); + } + radeon_cs_begin(rmesa->cmdbuf.cs, n, file, function, line); +} + + + diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h new file mode 100644 index 0000000000..cc9d579ea1 --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_common.h @@ -0,0 +1,56 @@ +#ifndef COMMON_MISC_H +#define COMMON_MISC_H + +#include "radeon_common_context.h" +#include "radeon_buffer.h" +#include "radeon_dma.h" +#include "radeon_texture.h" + +void radeonRecalcScissorRects(radeonContextPtr radeon); +void radeonSetCliprects(radeonContextPtr radeon); +void radeonUpdateScissor( GLcontext *ctx ); +void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h); + +void radeonWaitForIdleLocked(radeonContextPtr radeon); +extern uint32_t radeonGetAge(radeonContextPtr radeon); +void radeonCopyBuffer( __DRIdrawablePrivate *dPriv, + const drm_clip_rect_t *rect); +void radeonPageFlip( __DRIdrawablePrivate *dPriv ); +void radeonSwapBuffers(__DRIdrawablePrivate * dPriv); +void radeonCopySubBuffer(__DRIdrawablePrivate * dPriv, + int x, int y, int w, int h ); + +void radeonUpdatePageFlipping(radeonContextPtr rmesa); + +void radeonFlush(GLcontext *ctx); +void radeonFinish(GLcontext * ctx); +void radeonEmitState(radeonContextPtr radeon); + +static inline struct radeon_renderbuffer *radeon_get_depthbuffer(radeonContextPtr rmesa) +{ + struct radeon_renderbuffer *rrb; + rrb = rmesa->state.depth.rrb; + if (!rrb) + return NULL; + + return rrb; +} + +static inline struct radeon_renderbuffer *radeon_get_colorbuffer(radeonContextPtr rmesa) +{ + struct radeon_renderbuffer *rrb; + GLframebuffer *fb = rmesa->dri.drawable->driverPrivate; + + rrb = rmesa->state.color.rrb; + if (rmesa->radeonScreen->driScreen->dri2.enabled) { + rrb = (struct radeon_renderbuffer *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + } + if (!rrb) + return NULL; + return rrb; +} + +#include "radeon_cmdbuf.h" + + +#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c new file mode 100644 index 0000000000..8acde2b90c --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -0,0 +1,584 @@ +/************************************************************************** + +Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and + VA Linux Systems Inc., Fremont, California. +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +#include "radeon_common.h" +#include "xmlpool.h" /* for symbolic values of enum-type options */ +#include "utils.h" +#include "drirenderbuffer.h" +#include "vblank.h" + +#define DRIVER_DATE "20090101" + +#ifndef RADEON_DEBUG +int RADEON_DEBUG = (0); +#endif + +/* Return various strings for glGetString(). + */ +static const GLubyte *radeonGetString(GLcontext * ctx, GLenum name) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + static char buffer[128]; + + switch (name) { + case GL_VENDOR: + if (IS_R300_CLASS(radeon->radeonScreen)) + return (GLubyte *) "DRI R300 Project"; + else + return (GLubyte *) "Tungsten Graphics, Inc."; + + case GL_RENDERER: + { + unsigned offset; + GLuint agp_mode = (radeon->radeonScreen->card_type==RADEON_CARD_PCI) ? 0 : + radeon->radeonScreen->AGPMode; + const char* chipname; + + if (IS_R300_CLASS(radeon->radeonScreen)) + chipname = "R300"; + else if (IS_R200_CLASS(radeon->radeonScreen)) + chipname = "R200"; + else + chipname = "R100"; + + offset = driGetRendererString(buffer, chipname, DRIVER_DATE, + agp_mode); + + if (IS_R300_CLASS(radeon->radeonScreen)) { + sprintf(&buffer[offset], " %sTCL", + (radeon->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) + ? "" : "NO-"); + } else { + sprintf(&buffer[offset], " %sTCL", + !(radeon->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE) + ? "" : "NO-"); + } + + if (radeon->radeonScreen->driScreen->dri2.enabled) + strcat(buffer, " DRI2"); + + return (GLubyte *) buffer; + } + + default: + return NULL; + } +} + +/* Initialize the driver's misc functions. + */ +static void radeonInitDriverFuncs(struct dd_function_table *functions) +{ + functions->GetString = radeonGetString; +} + +/** + * Create and initialize all common fields of the context, + * including the Mesa context itself. + */ +GLboolean radeonInitContext(radeonContextPtr radeon, + struct dd_function_table* functions, + const __GLcontextModes * glVisual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) +{ + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); + GLcontext* ctx; + GLcontext* shareCtx; + int fthrottle_mode; + + /* Fill in additional standard functions. */ + radeonInitDriverFuncs(functions); + + radeon->radeonScreen = screen; + /* Allocate and initialize the Mesa context */ + if (sharedContextPrivate) + shareCtx = ((radeonContextPtr)sharedContextPrivate)->glCtx; + else + shareCtx = NULL; + radeon->glCtx = _mesa_create_context(glVisual, shareCtx, + functions, (void *)radeon); + if (!radeon->glCtx) + return GL_FALSE; + + ctx = radeon->glCtx; + driContextPriv->driverPrivate = radeon; + + /* DRI fields */ + radeon->dri.context = driContextPriv; + radeon->dri.screen = sPriv; + radeon->dri.drawable = NULL; + radeon->dri.readable = NULL; + radeon->dri.hwContext = driContextPriv->hHWContext; + radeon->dri.hwLock = &sPriv->pSAREA->lock; + radeon->dri.fd = sPriv->fd; + radeon->dri.drmMinor = sPriv->drm_version.minor; + + radeon->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA + + screen->sarea_priv_offset); + + /* Setup IRQs */ + fthrottle_mode = driQueryOptioni(&radeon->optionCache, "fthrottle_mode"); + radeon->iw.irq_seq = -1; + radeon->irqsEmitted = 0; + radeon->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS && + radeon->radeonScreen->irq); + + radeon->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); + + if (!radeon->do_irqs) + fprintf(stderr, + "IRQ's not enabled, falling back to %s: %d %d\n", + radeon->do_usleeps ? "usleeps" : "busy waits", + fthrottle_mode, radeon->radeonScreen->irq); + + (*sPriv->systemTime->getUST) (&radeon->swap_ust); + + return GL_TRUE; +} + +/** + * Cleanup common context fields. + * Called by r200DestroyContext/r300DestroyContext + */ +void radeonCleanupContext(radeonContextPtr radeon) +{ + FILE *track; + struct radeon_renderbuffer *rb; + GLframebuffer *fb; + + /* free the Mesa context */ + _mesa_destroy_context(radeon->glCtx); + + fb = (void*)radeon->dri.drawable->driverPrivate; + rb = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; + if (rb && rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb = (void *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + if (rb && rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb = (void *)fb->Attachment[BUFFER_DEPTH].Renderbuffer; + if (rb && rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + fb = (void*)radeon->dri.readable->driverPrivate; + rb = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; + if (rb && rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb = (void *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + if (rb && rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb = (void *)fb->Attachment[BUFFER_DEPTH].Renderbuffer; + if (rb && rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + + /* _mesa_destroy_context() might result in calls to functions that + * depend on the DriverCtx, so don't set it to NULL before. + * + * radeon->glCtx->DriverCtx = NULL; + */ + + + + /* free the option cache */ + driDestroyOptionCache(&radeon->optionCache); + + rcommonDestroyCmdBuf(radeon); + + if (radeon->state.scissor.pClipRects) { + FREE(radeon->state.scissor.pClipRects); + radeon->state.scissor.pClipRects = 0; + } + track = fopen("/tmp/tracklog", "w"); + if (track) { + radeon_tracker_print(&radeon->radeonScreen->bom->tracker, track); + fclose(track); + } +} + +/* Force the context `c' to be unbound from its buffer. + */ +GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv) +{ + radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; + + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "%s ctx %p\n", __FUNCTION__, + radeon->glCtx); + + return GL_TRUE; +} + + +static void +radeon_make_kernel_renderbuffer_current(radeonContextPtr radeon, + GLframebuffer *draw) +{ + /* if radeon->fake */ + struct radeon_renderbuffer *rb; + + if ((rb = (void *)draw->Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) { + if (!rb->bo) { + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + radeon->radeonScreen->frontOffset, + 0, + 0, + RADEON_GEM_DOMAIN_VRAM, + 0); + } + rb->cpp = radeon->radeonScreen->cpp; + rb->pitch = radeon->radeonScreen->frontPitch * rb->cpp; + } + if ((rb = (void *)draw->Attachment[BUFFER_BACK_LEFT].Renderbuffer)) { + if (!rb->bo) { + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + radeon->radeonScreen->backOffset, + 0, + 0, + RADEON_GEM_DOMAIN_VRAM, + 0); + } + rb->cpp = radeon->radeonScreen->cpp; + rb->pitch = radeon->radeonScreen->backPitch * rb->cpp; + } + if ((rb = (void *)draw->Attachment[BUFFER_DEPTH].Renderbuffer)) { + if (!rb->bo) { + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + radeon->radeonScreen->depthOffset, + 0, + 0, + RADEON_GEM_DOMAIN_VRAM, + 0); + } + rb->cpp = radeon->radeonScreen->cpp; + rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp; + } + if ((rb = (void *)draw->Attachment[BUFFER_STENCIL].Renderbuffer)) { + if (!rb->bo) { + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + radeon->radeonScreen->depthOffset, + 0, + 0, + RADEON_GEM_DOMAIN_VRAM, + 0); + } + rb->cpp = radeon->radeonScreen->cpp; + rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp; + } +} + +static void +radeon_make_renderbuffer_current(radeonContextPtr radeon, + GLframebuffer *draw) +{ + int size = 4096*4096*4; + /* if radeon->fake */ + struct radeon_renderbuffer *rb; + + if (radeon->radeonScreen->kernel_mm) { + radeon_make_kernel_renderbuffer_current(radeon, draw); + return; + } + + + if ((rb = (void *)draw->Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) { + if (!rb->bo) { + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + radeon->radeonScreen->frontOffset + + radeon->radeonScreen->fbLocation, + size, + 4096, + RADEON_GEM_DOMAIN_VRAM, + 0); + } + rb->cpp = radeon->radeonScreen->cpp; + rb->pitch = radeon->radeonScreen->frontPitch * rb->cpp; + } + if ((rb = (void *)draw->Attachment[BUFFER_BACK_LEFT].Renderbuffer)) { + if (!rb->bo) { + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + radeon->radeonScreen->backOffset + + radeon->radeonScreen->fbLocation, + size, + 4096, + RADEON_GEM_DOMAIN_VRAM, + 0); + } + rb->cpp = radeon->radeonScreen->cpp; + rb->pitch = radeon->radeonScreen->backPitch * rb->cpp; + } + if ((rb = (void *)draw->Attachment[BUFFER_DEPTH].Renderbuffer)) { + if (!rb->bo) { + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + radeon->radeonScreen->depthOffset + + radeon->radeonScreen->fbLocation, + size, + 4096, + RADEON_GEM_DOMAIN_VRAM, + 0); + } + rb->cpp = radeon->radeonScreen->cpp; + rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp; + } + if ((rb = (void *)draw->Attachment[BUFFER_STENCIL].Renderbuffer)) { + if (!rb->bo) { + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + radeon->radeonScreen->depthOffset + + radeon->radeonScreen->fbLocation, + size, + 4096, + RADEON_GEM_DOMAIN_VRAM, + 0); + } + rb->cpp = radeon->radeonScreen->cpp; + rb->pitch = radeon->radeonScreen->depthPitch * rb->cpp; + } +} + + +void +radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) +{ + unsigned int attachments[10]; + __DRIbuffer *buffers; + __DRIscreen *screen; + struct radeon_renderbuffer *rb; + int i, count; + GLframebuffer *draw; + radeonContextPtr radeon; + + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable); + + draw = drawable->driverPrivate; + screen = context->driScreenPriv; + radeon = (radeonContextPtr) context->driverPrivate; + i = 0; + if ((rb = (void *)draw->Attachment[BUFFER_FRONT_LEFT].Renderbuffer)) { + attachments[i++] = __DRI_BUFFER_FRONT_LEFT; + } + if ((rb = (void *)draw->Attachment[BUFFER_BACK_LEFT].Renderbuffer)) { + attachments[i++] = __DRI_BUFFER_BACK_LEFT; + } + if ((rb = (void *)draw->Attachment[BUFFER_DEPTH].Renderbuffer)) { + attachments[i++] = __DRI_BUFFER_DEPTH; + } + + buffers = (*screen->dri2.loader->getBuffers)(drawable, + &drawable->w, + &drawable->h, + attachments, i, + &count, + drawable->loaderPrivate); + if (buffers == NULL) + return; + + /* set one cliprect to cover the whole drawable */ + drawable->x = 0; + drawable->y = 0; + drawable->backX = 0; + drawable->backY = 0; + drawable->numClipRects = 1; + drawable->pClipRects[0].x1 = 0; + drawable->pClipRects[0].y1 = 0; + drawable->pClipRects[0].x2 = drawable->w; + drawable->pClipRects[0].y2 = drawable->h; + drawable->numBackClipRects = 1; + drawable->pBackClipRects[0].x1 = 0; + drawable->pBackClipRects[0].y1 = 0; + drawable->pBackClipRects[0].x2 = drawable->w; + drawable->pBackClipRects[0].y2 = drawable->h; + for (i = 0; i < count; i++) { + switch (buffers[i].attachment) { + case __DRI_BUFFER_FRONT_LEFT: + rb = (void *)draw->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; + if (rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb->cpp = buffers[i].cpp; + rb->pitch = buffers[i].pitch; + rb->width = drawable->w; + rb->height = drawable->h; + rb->has_surface = 0; + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + buffers[i].name, + 0, + 0, + RADEON_GEM_DOMAIN_VRAM, + buffers[i].flags); + if (rb->bo == NULL) { + fprintf(stderr, "failled to attach front %d\n", + buffers[i].name); + } + break; + case __DRI_BUFFER_BACK_LEFT: + rb = (void *)draw->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + if (rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb->cpp = buffers[i].cpp; + rb->pitch = buffers[i].pitch; + rb->width = drawable->w; + rb->height = drawable->h; + rb->has_surface = 0; + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + buffers[i].name, + 0, + 0, + RADEON_GEM_DOMAIN_VRAM, + buffers[i].flags); + break; + case __DRI_BUFFER_DEPTH: + rb = (void *)draw->Attachment[BUFFER_DEPTH].Renderbuffer; + if (rb->bo) { + radeon_bo_unref(rb->bo); + rb->bo = NULL; + } + rb->cpp = buffers[i].cpp; + rb->pitch = buffers[i].pitch; + rb->width = drawable->w; + rb->height = drawable->h; + rb->has_surface = 0; + rb->bo = radeon_bo_open(radeon->radeonScreen->bom, + buffers[i].name, + 0, + 0, + RADEON_GEM_DOMAIN_VRAM, + buffers[i].flags); + break; + case __DRI_BUFFER_STENCIL: + break; + case __DRI_BUFFER_ACCUM: + default: + fprintf(stderr, + "unhandled buffer attach event, attacment type %d\n", + buffers[i].attachment); + return; + } + } + radeon = (radeonContextPtr) context->driverPrivate; + driUpdateFramebufferSize(radeon->glCtx, drawable); +} + +/* Force the context `c' to be the current context and associate with it + * buffer `b'. + */ +GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) +{ + radeonContextPtr radeon; + GLframebuffer *dfb, *rfb; + + if (!driContextPriv) { + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "%s ctx is null\n", __FUNCTION__); + _mesa_make_current(NULL, NULL, NULL); + return GL_TRUE; + } + radeon = (radeonContextPtr) driContextPriv->driverPrivate; + dfb = driDrawPriv->driverPrivate; + rfb = driReadPriv->driverPrivate; + + if (driContextPriv->driScreenPriv->dri2.enabled) { + radeon_update_renderbuffers(driContextPriv, driDrawPriv); + if (driDrawPriv != driReadPriv) + radeon_update_renderbuffers(driContextPriv, driReadPriv); + radeon->state.color.rrb = + (void *)dfb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + radeon->state.depth.rrb = + (void *)dfb->Attachment[BUFFER_DEPTH].Renderbuffer; + } else { + radeon_make_renderbuffer_current(radeon, dfb); + } + + + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "%s ctx %p dfb %p rfb %p\n", __FUNCTION__, radeon->glCtx, dfb, rfb); + + driUpdateFramebufferSize(radeon->glCtx, driDrawPriv); + if (driReadPriv != driDrawPriv) + driUpdateFramebufferSize(radeon->glCtx, driReadPriv); + + + + _mesa_make_current(radeon->glCtx, dfb, rfb); + + if (radeon->dri.drawable != driDrawPriv) { + if (driDrawPriv->swap_interval == (unsigned)-1) { + driDrawPriv->vblFlags = + (radeon->radeonScreen->irq != 0) + ? driGetDefaultVBlankFlags(&radeon-> + optionCache) + : VBLANK_FLAG_NO_IRQ; + + driDrawableInitVBlank(driDrawPriv); + } + } + + radeon->dri.readable = driReadPriv; + + if (radeon->dri.drawable != driDrawPriv || + radeon->lastStamp != driDrawPriv->lastStamp) { + radeon->dri.drawable = driDrawPriv; + + radeonSetCliprects(radeon); + radeon->vtbl.update_viewport_offset(radeon->glCtx); + } + + _mesa_update_state(radeon->glCtx); + + if (!driContextPriv->driScreenPriv->dri2.enabled) { + radeonUpdatePageFlipping(radeon); + } + + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "End %s\n", __FUNCTION__); + return GL_TRUE; +} + diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h new file mode 100644 index 0000000000..b88aeab051 --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -0,0 +1,513 @@ + +#ifndef COMMON_CONTEXT_H +#define COMMON_CONTEXT_H + +#include "main/mm.h" +#include "math/m_vector.h" +#include "texmem.h" +#include "tnl/t_context.h" +#include "main/colormac.h" + +#include "radeon_screen.h" +#include "radeon_drm.h" +#include "dri_util.h" +#include "tnl/t_vertex.h" + +/* This union is used to avoid warnings/miscompilation + with float to uint32_t casts due to strict-aliasing */ +typedef union { GLfloat f; uint32_t ui32; } float_ui32_type; + +struct radeon_context; +typedef struct radeon_context radeonContextRec; +typedef struct radeon_context *radeonContextPtr; + + +#define TEX_0 0x1 +#define TEX_1 0x2 +#define TEX_2 0x4 +#define TEX_3 0x8 +#define TEX_4 0x10 +#define TEX_5 0x20 + +/* Rasterizing fallbacks */ +/* See correponding strings in r200_swtcl.c */ +#define RADEON_FALLBACK_TEXTURE 0x0001 +#define RADEON_FALLBACK_DRAW_BUFFER 0x0002 +#define RADEON_FALLBACK_STENCIL 0x0004 +#define RADEON_FALLBACK_RENDER_MODE 0x0008 +#define RADEON_FALLBACK_BLEND_EQ 0x0010 +#define RADEON_FALLBACK_BLEND_FUNC 0x0020 +#define RADEON_FALLBACK_DISABLE 0x0040 +#define RADEON_FALLBACK_BORDER_MODE 0x0080 + +#define R200_FALLBACK_TEXTURE 0x01 +#define R200_FALLBACK_DRAW_BUFFER 0x02 +#define R200_FALLBACK_STENCIL 0x04 +#define R200_FALLBACK_RENDER_MODE 0x08 +#define R200_FALLBACK_DISABLE 0x10 +#define R200_FALLBACK_BORDER_MODE 0x20 + +#define RADEON_TCL_FALLBACK_RASTER 0x1 /* rasterization */ +#define RADEON_TCL_FALLBACK_UNFILLED 0x2 /* unfilled tris */ +#define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE 0x4 /* twoside tris */ +#define RADEON_TCL_FALLBACK_MATERIAL 0x8 /* material in vb */ +#define RADEON_TCL_FALLBACK_TEXGEN_0 0x10 /* texgen, unit 0 */ +#define RADEON_TCL_FALLBACK_TEXGEN_1 0x20 /* texgen, unit 1 */ +#define RADEON_TCL_FALLBACK_TEXGEN_2 0x40 /* texgen, unit 2 */ +#define RADEON_TCL_FALLBACK_TCL_DISABLE 0x80 /* user disable */ +#define RADEON_TCL_FALLBACK_FOGCOORDSPEC 0x100 /* fogcoord, sep. spec light */ + +/* The blit width for texture uploads + */ +#define BLIT_WIDTH_BYTES 1024 + +/* Use the templated vertex format: + */ +#define COLOR_IS_RGBA +#define TAG(x) radeon##x +#include "tnl_dd/t_dd_vertex.h" +#undef TAG + +struct radeon_colorbuffer_state { + GLuint clear; + int roundEnable; + struct radeon_renderbuffer *rrb; +}; + +struct radeon_depthbuffer_state { + GLuint clear; + GLfloat scale; + struct radeon_renderbuffer *rrb; +}; + +struct radeon_scissor_state { + drm_clip_rect_t rect; + GLboolean enabled; + + GLuint numClipRects; /* Cliprects active */ + GLuint numAllocedClipRects; /* Cliprects available */ + drm_clip_rect_t *pClipRects; +}; + +struct radeon_stencilbuffer_state { + GLboolean hwBuffer; + GLuint clear; /* rb3d_stencilrefmask value */ +}; + +struct radeon_stipple_state { + GLuint mask[32]; +}; + +struct radeon_state_atom { + struct radeon_state_atom *next, *prev; + const char *name; /* for debug */ + int cmd_size; /* size in bytes */ + GLuint idx; + GLuint is_tcl; + GLuint *cmd; /* one or more cmd's */ + GLuint *lastcmd; /* one or more cmd's */ + GLboolean dirty; /* dirty-mark in emit_state_list */ + int (*check) (GLcontext *, struct radeon_state_atom *atom); /* is this state active? */ + void (*emit) (GLcontext *, struct radeon_state_atom *atom); +}; + +struct radeon_hw_state { + /* Head of the linked list of state atoms. */ + struct radeon_state_atom atomlist; + int max_state_size; /* Number of bytes necessary for a full state emit. */ + GLboolean is_dirty, all_dirty; +}; + + +/* Texture related */ +typedef struct _radeon_texture_image radeon_texture_image; + +struct _radeon_texture_image { + struct gl_texture_image base; + + /** + * If mt != 0, the image is stored in hardware format in the + * given mipmap tree. In this case, base.Data may point into the + * mapping of the buffer object that contains the mipmap tree. + * + * If mt == 0, the image is stored in normal memory pointed to + * by base.Data. + */ + struct _radeon_mipmap_tree *mt; + struct radeon_bo *bo; + + int mtlevel; /** if mt != 0, this is the image's level in the mipmap tree */ + int mtface; /** if mt != 0, this is the image's face in the mipmap tree */ +}; + + +static INLINE radeon_texture_image *get_radeon_texture_image(struct gl_texture_image *image) +{ + return (radeon_texture_image*)image; +} + + +typedef struct radeon_tex_obj radeonTexObj, *radeonTexObjPtr; + +#define RADEON_TXO_MICRO_TILE (1 << 3) + +/* Texture object in locally shared texture space. + */ +struct radeon_tex_obj { + // driTextureObject base; + struct gl_texture_object base; + struct _radeon_mipmap_tree *mt; + + /** + * This is true if we've verified that the mipmap tree above is complete + * and so on. + */ + GLboolean validated; + + GLuint override_offset; + GLboolean image_override; /* Image overridden by GLX_EXT_tfp */ + GLuint tile_bits; /* hw texture tile bits used on this texture */ + struct radeon_bo *bo; + + GLuint bufAddr; /* Offset to start of locally + shared texture block */ + + GLuint dirty_state; /* Flags (1 per texunit) for + whether or not this texobj + has dirty hardware state + (pp_*) that needs to be + brought into the + texunit. */ + + drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS]; + /* Six, for the cube faces */ + + + + GLuint pp_txfilter; /* hardware register values */ + GLuint pp_txformat; + GLuint pp_txformat_x; + GLuint pp_txoffset; /* Image location in texmem. + All cube faces follow. */ + GLuint pp_txsize; /* npot only */ + GLuint pp_txpitch; /* npot only */ + GLuint pp_border_color; + GLuint pp_cubic_faces; /* cube face 1,2,3,4 log2 sizes */ + + GLuint pp_txfilter_1; /* r300 */ + + GLboolean border_fallback; + + +}; + +static INLINE radeonTexObj* radeon_tex_obj(struct gl_texture_object *texObj) +{ + return (radeonTexObj*)texObj; +} + +/* Need refcounting on dma buffers: + */ +struct radeon_dma_buffer { + int refcount; /* the number of retained regions in buf */ + drmBufPtr buf; +}; + +struct radeon_aos { + struct radeon_bo *bo; /** Buffer object where vertex data is stored */ + int offset; /** Offset into buffer object, in bytes */ + int components; /** Number of components per vertex */ + int stride; /** Stride in dwords (may be 0 for repeating) */ + int count; /** Number of vertices */ +}; + +struct radeon_dma { + /* Active dma region. Allocations for vertices and retained + * regions come from here. Also used for emitting random vertices, + * these may be flushed by calling flush_current(); + */ + struct radeon_bo *current; /** Buffer that DMA memory is allocated from */ + int current_used; /** Number of bytes allocated and forgotten about */ + int current_vertexptr; /** End of active vertex region */ + + /** + * If current_vertexptr != current_used then flush must be non-zero. + * flush must be called before non-active vertex allocations can be + * performed. + */ + void (*flush) (GLcontext *); + + /* Number of "in-flight" DMA buffers, i.e. the number of buffers + * for which a DISCARD command is currently queued in the command buffer +. + */ + GLuint nr_released_bufs; +}; + +/* radeon_swtcl.c + */ +struct radeon_swtcl_info { + + GLuint RenderIndex; + GLuint vertex_size; + GLubyte *verts; + + /* Fallback rasterization functions + */ + GLuint hw_primitive; + GLenum render_primitive; + GLuint numverts; + + struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; + GLuint vertex_attr_count; + +}; + +struct radeon_ioctl { + GLuint vertex_offset; + struct radeon_bo *bo; + GLuint vertex_size; +}; + +#define RADEON_MAX_PRIMS 64 + +struct radeon_prim { + GLuint start; + GLuint end; + GLuint prim; +}; + +static INLINE GLuint radeonPackColor(GLuint cpp, + GLubyte r, GLubyte g, + GLubyte b, GLubyte a) +{ + switch (cpp) { + case 2: + return PACK_COLOR_565(r, g, b); + case 4: + return PACK_COLOR_8888(a, r, g, b); + default: + return 0; + } +} + +#define MAX_CMD_BUF_SZ (16*1024) + +#define MAX_DMA_BUF_SZ (64*1024) + +struct radeon_store { + GLuint statenr; + GLuint primnr; + char cmd_buf[MAX_CMD_BUF_SZ]; + int cmd_used; + int elts_start; +}; + +struct radeon_dri_mirror { + __DRIcontextPrivate *context; /* DRI context */ + __DRIscreenPrivate *screen; /* DRI screen */ + + /** + * DRI drawable bound to this context for drawing. + */ + __DRIdrawablePrivate *drawable; + + /** + * DRI drawable bound to this context for reading. + */ + __DRIdrawablePrivate *readable; + + drm_context_t hwContext; + drm_hw_lock_t *hwLock; + int fd; + int drmMinor; +}; + +#define DEBUG_TEXTURE 0x001 +#define DEBUG_STATE 0x002 +#define DEBUG_IOCTL 0x004 +#define DEBUG_PRIMS 0x008 +#define DEBUG_VERTS 0x010 +#define DEBUG_FALLBACKS 0x020 +#define DEBUG_VFMT 0x040 +#define DEBUG_CODEGEN 0x080 +#define DEBUG_VERBOSE 0x100 +#define DEBUG_DRI 0x200 +#define DEBUG_DMA 0x400 +#define DEBUG_SANITY 0x800 +#define DEBUG_SYNC 0x1000 +#define DEBUG_PIXEL 0x2000 +#define DEBUG_MEMORY 0x4000 + + + +typedef void (*radeon_tri_func) (radeonContextPtr, + radeonVertex *, + radeonVertex *, radeonVertex *); + +typedef void (*radeon_line_func) (radeonContextPtr, + radeonVertex *, radeonVertex *); + +typedef void (*radeon_point_func) (radeonContextPtr, radeonVertex *); + +struct radeon_state { + struct radeon_colorbuffer_state color; + struct radeon_depthbuffer_state depth; + struct radeon_scissor_state scissor; + struct radeon_stencilbuffer_state stencil; +}; + +/** + * This structure holds the command buffer while it is being constructed. + * + * The first batch of commands in the buffer is always the state that needs + * to be re-emitted when the context is lost. This batch can be skipped + * otherwise. + */ +struct radeon_cmdbuf { + struct radeon_cs_manager *csm; + struct radeon_cs *cs; + int size; /** # of dwords total */ + unsigned int flushing:1; /** whether we're currently in FlushCmdBufLocked */ +}; + +struct radeon_context { + GLcontext *glCtx; + radeonScreenPtr radeonScreen; /* Screen private DRI data */ + + /* Texture object bookkeeping + */ + unsigned nr_heaps; + driTexHeap * texture_heaps[ RADEON_NR_TEX_HEAPS ]; + driTextureObject swapped; + int texture_depth; + float initialMaxAnisotropy; + + struct radeon_dma dma; + struct radeon_hw_state hw; + /* Rasterization and vertex state: + */ + GLuint TclFallback; + GLuint Fallback; + GLuint NewGLState; + DECLARE_RENDERINPUTS(tnl_index_bitset); /* index of bits for last tnl_install_attrs */ + + /* Page flipping */ + GLuint doPageFlip; + + /* Drawable, cliprect and scissor information */ + GLuint numClipRects; /* Cliprects for the draw buffer */ + drm_clip_rect_t *pClipRects; + unsigned int lastStamp; + GLboolean lost_context; + drm_radeon_sarea_t *sarea; /* Private SAREA data */ + + /* Mirrors of some DRI state */ + struct radeon_dri_mirror dri; + + /* Busy waiting */ + GLuint do_usleeps; + GLuint do_irqs; + GLuint irqsEmitted; + drm_radeon_irq_wait_t iw; + + /* buffer swap */ + int64_t swap_ust; + int64_t swap_missed_ust; + + GLuint swap_count; + GLuint swap_missed_count; + + /* Derived state - for r300 only */ + struct radeon_state state; + + struct radeon_swtcl_info swtcl; + /* Configuration cache + */ + driOptionCache optionCache; + + struct radeon_cmdbuf cmdbuf; + + struct { + void (*get_lock)(radeonContextPtr radeon); + void (*update_viewport_offset)(GLcontext *ctx); + void (*update_draw_buffer)(GLcontext *ctx); + void (*emit_cs_header)(struct radeon_cs *cs, radeonContextPtr rmesa); + void (*swtcl_flush)(GLcontext *ctx, uint32_t offset); + void (*pre_emit_atoms)(radeonContextPtr rmesa); + void (*pre_emit_state)(radeonContextPtr rmesa); + } vtbl; +}; + +#define RADEON_CONTEXT(glctx) ((radeonContextPtr)(ctx->DriverCtx)) + +/** + * This function takes a float and packs it into a uint32_t + */ +static INLINE uint32_t radeonPackFloat32(float fl) +{ + union { + float fl; + uint32_t u; + } u; + + u.fl = fl; + return u.u; +} + +/* This is probably wrong for some values, I need to test this + * some more. Range checking would be a good idea also.. + * + * But it works for most things. I'll fix it later if someone + * else with a better clue doesn't + */ +static INLINE uint32_t radeonPackFloat24(float f) +{ + float mantissa; + int exponent; + uint32_t float24 = 0; + + if (f == 0.0) + return 0; + + mantissa = frexpf(f, &exponent); + + /* Handle -ve */ + if (mantissa < 0) { + float24 |= (1 << 23); + mantissa = mantissa * -1.0; + } + /* Handle exponent, bias of 63 */ + exponent += 62; + float24 |= (exponent << 16); + /* Kill 7 LSB of mantissa */ + float24 |= (radeonPackFloat32(mantissa) & 0x7FFFFF) >> 7; + + return float24; +} + +GLboolean radeonInitContext(radeonContextPtr radeon, + struct dd_function_table* functions, + const __GLcontextModes * glVisual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); + +void radeonCleanupContext(radeonContextPtr radeon); +GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv); +void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable); +GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); + +/* ================================================================ + * Debugging: + */ +#define DO_DEBUG 1 + +#if DO_DEBUG +extern int RADEON_DEBUG; +#else +#define RADEON_DEBUG 0 +#endif + +#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index b25f036c44..0296cffc20 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -53,8 +53,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "drivers/common/driverfuncs.h" +#include "radeon_common.h" #include "radeon_context.h" -#include "common_cmdbuf.h" #include "radeon_ioctl.h" #include "radeon_state.h" #include "radeon_span.h" @@ -77,36 +77,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ -/* Return various strings for glGetString(). - */ -static const GLubyte *radeonGetString( GLcontext *ctx, GLenum name ) -{ - r100ContextPtr rmesa = R100_CONTEXT(ctx); - static char buffer[128]; - unsigned offset; - GLuint agp_mode = (rmesa->radeon.radeonScreen->card_type==RADEON_CARD_PCI) ? 0 : - rmesa->radeon.radeonScreen->AGPMode; - - switch ( name ) { - case GL_VENDOR: - return (GLubyte *)"Tungsten Graphics, Inc."; - - case GL_RENDERER: - offset = driGetRendererString( buffer, "Radeon", DRIVER_DATE, - agp_mode ); - - sprintf( & buffer[ offset ], " %sTCL", - !(rmesa->radeon.TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE) - ? "" : "NO-" ); - - return (GLubyte *)buffer; - - default: - return NULL; - } -} - - /* Extension strings exported by the R100 driver. */ const struct dri_extension card_extensions[] = @@ -163,15 +133,6 @@ static const struct tnl_pipeline_stage *radeon_pipeline[] = { NULL, }; - - -/* Initialize the driver's misc functions. - */ -static void radeonInitDriverFuncs( struct dd_function_table *functions ) -{ - functions->GetString = radeonGetString; -} - static const struct dri_debug_control debug_control[] = { { "fall", DEBUG_FALLBACKS }, @@ -292,7 +253,6 @@ radeonCreateContext( const __GLcontextModes *glVisual, * (the texture functions are especially important) */ _mesa_init_driver_functions( &functions ); - radeonInitDriverFuncs( &functions ); radeonInitTextureFuncs( &functions ); if (!radeonInitContext(&rmesa->radeon, &functions, diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h index d05e47f16e..2efabd1f55 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_context.h @@ -53,8 +53,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/colormac.h" #include "radeon_screen.h" -#include "common_context.h" -#include "common_misc.h" +#include "radeon_common.h" struct r100_context; diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c index fbb93a8ba1..2db3f7a68d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c @@ -31,8 +31,6 @@ */ #include - -#include "common_context.h" #include "radeon_cs.h" #include "radeon_cs_legacy.h" #include "radeon_bo_legacy.h" @@ -234,7 +232,7 @@ static int cs_process_relocs(struct radeon_cs *cs) } cs->packets[relocs[i].indices[j]] += soffset; if (cs->packets[relocs[i].indices[j]] >= eoffset) { - radeon_bo_debug(relocs[i].base.bo, 12); + /* radeon_bo_debug(relocs[i].base.bo, 12); */ fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n", relocs[i].base.bo, soffset, eoffset); fprintf(stderr, "above end: %p 0x%08X 0x%08X\n", diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h index 4da2479eee..19d904174f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h +++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.h @@ -33,7 +33,7 @@ #define RADEON_CS_LEGACY_H #include "radeon_cs.h" -#include "common_context.h" +#include "radeon_common.h" struct radeon_cs_manager *radeon_cs_manager_legacy_ctor(struct radeon_context *ctx); void radeon_cs_manager_legacy_dtor(struct radeon_cs_manager *csm); diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c new file mode 100644 index 0000000000..393b12182d --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -0,0 +1,323 @@ +/************************************************************************** + +Copyright (C) 2004 Nicolai Haehnle. +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the "Software"), +to deal in the Software without restriction, including without limitation +on the rights to use, copy, modify, merge, publish, distribute, sub +license, and/or sell copies of the Software, and to permit persons to whom +the Software is furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice (including the next +paragraph) shall be included in all copies or substantial portions of the +Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL +ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, +DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR +OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE +USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +#include "radeon_common.h" + +#if defined(USE_X86_ASM) +#define COPY_DWORDS( dst, src, nr ) \ +do { \ + int __tmp; \ + __asm__ __volatile__( "rep ; movsl" \ + : "=%c" (__tmp), "=D" (dst), "=S" (__tmp) \ + : "0" (nr), \ + "D" ((long)dst), \ + "S" ((long)src) ); \ +} while (0) +#else +#define COPY_DWORDS( dst, src, nr ) \ +do { \ + int j; \ + for ( j = 0 ; j < nr ; j++ ) \ + dst[j] = ((int *)src)[j]; \ + dst += nr; \ +} while (0) +#endif + +static void radeonEmitVec4(uint32_t *out, GLvoid * data, int stride, int count) +{ + int i; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s count %d stride %d out %p data %p\n", + __FUNCTION__, count, stride, (void *)out, (void *)data); + + if (stride == 4) + COPY_DWORDS(out, data, count); + else + for (i = 0; i < count; i++) { + out[0] = *(int *)data; + out++; + data += stride; + } +} + +void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count) +{ + int i; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s count %d stride %d out %p data %p\n", + __FUNCTION__, count, stride, (void *)out, (void *)data); + + if (stride == 8) + COPY_DWORDS(out, data, count * 2); + else + for (i = 0; i < count; i++) { + out[0] = *(int *)data; + out[1] = *(int *)(data + 4); + out += 2; + data += stride; + } +} + +void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count) +{ + int i; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s count %d stride %d out %p data %p\n", + __FUNCTION__, count, stride, (void *)out, (void *)data); + + if (stride == 12) { + COPY_DWORDS(out, data, count * 3); + } + else + for (i = 0; i < count; i++) { + out[0] = *(int *)data; + out[1] = *(int *)(data + 4); + out[2] = *(int *)(data + 8); + out += 3; + data += stride; + } +} + +static void radeonEmitVec16(uint32_t *out, GLvoid * data, int stride, int count) +{ + int i; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s count %d stride %d out %p data %p\n", + __FUNCTION__, count, stride, (void *)out, (void *)data); + + if (stride == 16) + COPY_DWORDS(out, data, count * 4); + else + for (i = 0; i < count; i++) { + out[0] = *(int *)data; + out[1] = *(int *)(data + 4); + out[2] = *(int *)(data + 8); + out[3] = *(int *)(data + 12); + out += 4; + data += stride; + } +} + +void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, + GLvoid * data, int size, int stride, int count) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + uint32_t *out; + + if (stride == 0) { + radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * 4, 32); + count = 1; + aos->stride = 0; + } else { + radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * count * 4, 32); + aos->stride = size; + } + + aos->components = size; + aos->count = count; + + out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); + switch (size) { + case 1: radeonEmitVec4(out, data, stride, count); break; + case 2: radeonEmitVec8(out, data, stride, count); break; + case 3: radeonEmitVec12(out, data, stride, count); break; + case 4: radeonEmitVec16(out, data, stride, count); break; + default: + assert(0); + break; + } +} + +void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) +{ + struct radeon_cs_space_check bos[1]; + int flushed = 0, ret; + + size = MAX2(size, MAX_DMA_BUF_SZ * 16); + + if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (rmesa->dma.flush) { + rmesa->dma.flush(rmesa->glCtx); + } + + if (rmesa->dma.nr_released_bufs > 4) { + rcommonFlushCmdBuf(rmesa, __FUNCTION__); + rmesa->dma.nr_released_bufs = 0; + } + + if (rmesa->dma.current) { + radeon_bo_unmap(rmesa->dma.current); + radeon_bo_unref(rmesa->dma.current); + rmesa->dma.current = 0; + } + +again_alloc: + rmesa->dma.current = radeon_bo_open(rmesa->radeonScreen->bom, + 0, size, 4, RADEON_GEM_DOMAIN_GTT, + 0); + + if (!rmesa->dma.current) { + rcommonFlushCmdBuf(rmesa, __FUNCTION__); + rmesa->dma.nr_released_bufs = 0; + goto again_alloc; + } + + rmesa->dma.current_used = 0; + rmesa->dma.current_vertexptr = 0; + + bos[0].bo = rmesa->dma.current; + bos[0].read_domains = RADEON_GEM_DOMAIN_GTT; + bos[0].write_domain =0 ; + bos[0].new_accounted = 0; + + ret = radeon_cs_space_check(rmesa->cmdbuf.cs, bos, 1); + if (ret == RADEON_CS_SPACE_OP_TO_BIG) { + fprintf(stderr,"Got OPEARTION TO BIG ILLEGAL - this cannot happen"); + assert(0); + } else if (ret == RADEON_CS_SPACE_FLUSH) { + rcommonFlushCmdBuf(rmesa, __FUNCTION__); + if (flushed) { + fprintf(stderr,"flushed but still no space\n"); + assert(0); + } + flushed = 1; + goto again_alloc; + } + radeon_bo_map(rmesa->dma.current, 1); +} + +/* Allocates a region from rmesa->dma.current. If there isn't enough + * space in current, grab a new buffer (and discard what was left of current) + */ +void radeonAllocDmaRegion(radeonContextPtr rmesa, + struct radeon_bo **pbo, int *poffset, + int bytes, int alignment) +{ + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s %d\n", __FUNCTION__, bytes); + + if (rmesa->dma.flush) + rmesa->dma.flush(rmesa->glCtx); + + assert(rmesa->dma.current_used == rmesa->dma.current_vertexptr); + + alignment--; + rmesa->dma.current_used = (rmesa->dma.current_used + alignment) & ~alignment; + + if (!rmesa->dma.current || rmesa->dma.current_used + bytes > rmesa->dma.current->size) + radeonRefillCurrentDmaRegion(rmesa, (bytes + 15) & ~15); + + *poffset = rmesa->dma.current_used; + *pbo = rmesa->dma.current; + radeon_bo_ref(*pbo); + + /* Always align to at least 16 bytes */ + rmesa->dma.current_used = (rmesa->dma.current_used + bytes + 15) & ~15; + rmesa->dma.current_vertexptr = rmesa->dma.current_used; + + assert(rmesa->dma.current_used <= rmesa->dma.current->size); +} + +void radeonReleaseDmaRegion(radeonContextPtr rmesa) +{ + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s %p\n", __FUNCTION__, rmesa->dma.current); + if (rmesa->dma.current) { + rmesa->dma.nr_released_bufs++; + radeon_bo_unmap(rmesa->dma.current); + radeon_bo_unref(rmesa->dma.current); + } + rmesa->dma.current = NULL; +} + + +/* Flush vertices in the current dma region. + */ +void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + struct radeon_dma *dma = &rmesa->dma; + + + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + dma->flush = NULL; + + if (dma->current) { + GLuint current_offset = dma->current_used; + + assert (dma->current_used + + rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == + dma->current_vertexptr); + + if (dma->current_used != dma->current_vertexptr) { + dma->current_used = dma->current_vertexptr; + + rmesa->vtbl.swtcl_flush(ctx, current_offset); + } + rmesa->swtcl.numverts = 0; + } +} +/* Alloc space in the current dma region. + */ +void * +rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) +{ + GLuint bytes = vsize * nverts; + void *head; + + if (!rmesa->dma.current || rmesa->dma.current_vertexptr + bytes > rmesa->dma.current->size) { + radeonRefillCurrentDmaRegion(rmesa, bytes); + } + + if (!rmesa->dma.flush) { + rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; + rmesa->dma.flush = rcommon_flush_last_swtcl_prim; + } + + ASSERT( vsize == rmesa->swtcl.vertex_size * 4 ); + ASSERT( rmesa->dma.flush == rcommon_flush_last_swtcl_prim ); + ASSERT( rmesa->dma.current_used + + rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == + rmesa->dma.current_vertexptr ); + + head = (rmesa->dma.current->ptr + rmesa->dma.current_vertexptr); + rmesa->dma.current_vertexptr += bytes; + rmesa->swtcl.numverts += nverts; + return head; +} diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.h b/src/mesa/drivers/dri/radeon/radeon_dma.h new file mode 100644 index 0000000000..cee3744fed --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_dma.h @@ -0,0 +1,51 @@ +/************************************************************************** + +Copyright (C) 2004 Nicolai Haehnle. +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the "Software"), +to deal in the Software without restriction, including without limitation +on the rights to use, copy, modify, merge, publish, distribute, sub +license, and/or sell copies of the Software, and to permit persons to whom +the Software is furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice (including the next +paragraph) shall be included in all copies or substantial portions of the +Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL +ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, +DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR +OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE +USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +#ifndef RADEON_DMA_H +#define RADEON_DMA_H + +void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count); +void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count); + +void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, + GLvoid * data, int size, int stride, int count); + +void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size); +void radeonAllocDmaRegion(radeonContextPtr rmesa, + struct radeon_bo **pbo, int *poffset, + int bytes, int alignment); +void radeonReleaseDmaRegion(radeonContextPtr rmesa); + +void rcommon_flush_last_swtcl_prim(GLcontext *ctx); + +void *rcommonAllocDmaLowVerts(radeonContextPtr rmesa, int nverts, int vsize); +#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index d0cd9ce7a6..99b1ef332f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -43,7 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "swrast/swrast.h" #include "radeon_context.h" -#include "common_cmdbuf.h" +#include "radeon_common.h" #include "radeon_cs.h" #include "radeon_state.h" #include "radeon_ioctl.h" diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.c b/src/mesa/drivers/dri/radeon/radeon_lock.c new file mode 100644 index 0000000000..9a7e76b245 --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_lock.c @@ -0,0 +1,124 @@ +/************************************************************************** + +Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and + VA Linux Systems Inc., Fremont, California. +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Gareth Hughes + * Keith Whitwell + * Kevin E. Martin + */ + +#include "main/glheader.h" +#include "main/mtypes.h" +#include "main/colormac.h" +#include "dri_util.h" +#include "radeon_screen.h" +#include "radeon_common.h" +#include "radeon_lock.h" +#include "drirenderbuffer.h" + +#if DEBUG_LOCKING +char *prevLockFile = NULL; +int prevLockLine = 0; +#endif + +/* Turn on/off page flipping according to the flags in the sarea: + */ +void radeonUpdatePageFlipping(radeonContextPtr rmesa) +{ + int use_back; + __DRIdrawablePrivate *const drawable = rmesa->dri.drawable; + GLframebuffer *fb = drawable->driverPrivate; + + rmesa->doPageFlip = rmesa->sarea->pfState; + if (rmesa->glCtx->WinSysDrawBuffer) { + rmesa->vtbl.update_draw_buffer(rmesa->glCtx); + } + + use_back = rmesa->glCtx->DrawBuffer ? + (rmesa->glCtx->DrawBuffer->_ColorDrawBufferIndexes[0] == + BUFFER_BACK_LEFT) : 1; + use_back ^= (rmesa->sarea->pfCurrentPage == 1); + + if (use_back) + rmesa->state.color.rrb = (void *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + else + rmesa->state.color.rrb = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; + + rmesa->state.depth.rrb = (void *)fb->Attachment[BUFFER_DEPTH].Renderbuffer; +} + +/* Update the hardware state. This is called if another context has + * grabbed the hardware lock, which includes the X server. This + * function also updates the driver's window state after the X server + * moves, resizes or restacks a window -- the change will be reflected + * in the drawable position and clip rects. Since the X server grabs + * the hardware lock when it changes the window state, this routine will + * automatically be called after such a change. + */ +void radeonGetLock(radeonContextPtr rmesa, GLuint flags) +{ + __DRIdrawablePrivate *const drawable = rmesa->dri.drawable; + __DRIdrawablePrivate *const readable = rmesa->dri.readable; + __DRIscreenPrivate *sPriv = rmesa->dri.screen; + drm_radeon_sarea_t *sarea = rmesa->sarea; + + assert(drawable != NULL); + + drmGetLock(rmesa->dri.fd, rmesa->dri.hwContext, flags); + + /* The window might have moved, so we might need to get new clip + * rects. + * + * NOTE: This releases and regrabs the hw lock to allow the X server + * to respond to the DRI protocol request for new drawable info. + * Since the hardware state depends on having the latest drawable + * clip rects, all state checking must be done _after_ this call. + */ + DRI_VALIDATE_DRAWABLE_INFO(sPriv, drawable); + if (drawable != readable) { + DRI_VALIDATE_DRAWABLE_INFO(sPriv, readable); + } + + if (rmesa->lastStamp != drawable->lastStamp) { + radeonUpdatePageFlipping(rmesa); + radeonSetCliprects(rmesa); + rmesa->vtbl.update_viewport_offset(rmesa->glCtx); + driUpdateFramebufferSize(rmesa->glCtx, drawable); + } + + rmesa->vtbl.get_lock(rmesa); + + rmesa->lost_context = GL_TRUE; +} diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.h b/src/mesa/drivers/dri/radeon/radeon_lock.h index f2d17a54d4..f5ebb8dd54 100644 --- a/src/mesa/drivers/dri/radeon/radeon_lock.h +++ b/src/mesa/drivers/dri/radeon/radeon_lock.h @@ -39,9 +39,82 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * Kevin E. Martin */ -#ifndef __RADEON_LOCK_H__ -#define __RADEON_LOCK_H__ +#ifndef COMMON_LOCK_H +#define COMMON_LOCK_H -#include "common_lock.h" +#include "main/colormac.h" +#include "radeon_screen.h" +#include "radeon_common.h" -#endif /* __RADEON_LOCK_H__ */ +extern void radeonGetLock(radeonContextPtr rmesa, GLuint flags); + +/* Turn DEBUG_LOCKING on to find locking conflicts. + */ +#define DEBUG_LOCKING 0 + +#if DEBUG_LOCKING +extern char *prevLockFile; +extern int prevLockLine; + +#define DEBUG_LOCK() \ + do { \ + prevLockFile = (__FILE__); \ + prevLockLine = (__LINE__); \ + } while (0) + +#define DEBUG_RESET() \ + do { \ + prevLockFile = 0; \ + prevLockLine = 0; \ + } while (0) + +#define DEBUG_CHECK_LOCK() \ + do { \ + if ( prevLockFile ) { \ + fprintf( stderr, \ + "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n", \ + prevLockFile, prevLockLine, __FILE__, __LINE__ ); \ + exit( 1 ); \ + } \ + } while (0) + +#else + +#define DEBUG_LOCK() +#define DEBUG_RESET() +#define DEBUG_CHECK_LOCK() + +#endif + +/* + * !!! We may want to separate locks from locks with validation. This + * could be used to improve performance for those things commands that + * do not do any drawing !!! + */ + +/* Lock the hardware and validate our state. + */ +#define LOCK_HARDWARE( rmesa ) \ + do { \ + char __ret = 0; \ + DEBUG_CHECK_LOCK(); \ + if (!(rmesa)->radeonScreen->driScreen->dri2.enabled) { \ + DRM_CAS( (rmesa)->dri.hwLock, (rmesa)->dri.hwContext, \ + (DRM_LOCK_HELD | (rmesa)->dri.hwContext), __ret ); \ + if ( __ret ) \ + radeonGetLock( (rmesa), 0 ); \ + } \ + DEBUG_LOCK(); \ + } while (0) + +#define UNLOCK_HARDWARE( rmesa ) \ + do { \ + if (!(rmesa)->radeonScreen->driScreen->dri2.enabled) { \ + DRM_UNLOCK( (rmesa)->dri.fd, \ + (rmesa)->dri.hwLock, \ + (rmesa)->dri.hwContext ); \ + DEBUG_RESET(); \ + } \ + } while (0) + +#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h index 875e5e5aee..43dfa48aa7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h @@ -28,7 +28,7 @@ #ifndef __RADEON_MIPMAP_TREE_H_ #define __RADEON_MIPMAP_TREE_H_ -#include "common_context.h" +#include "radeon_common.h" typedef struct _radeon_mipmap_tree radeon_mipmap_tree; typedef struct _radeon_mipmap_level radeon_mipmap_level; diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index 7397218fb6..7f0d526ed2 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -47,7 +47,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_macros.h" #include "radeon_screen.h" #include "radeon_buffer.h" -#include "common_misc.h" +#include "radeon_common.h" #include "radeon_span.h" #if !RADEON_COMMON #include "radeon_context.h" @@ -362,7 +362,7 @@ static const __DRItexBufferExtension r300TexBufferExtension = { }; #endif -int radeon_set_screen_flags(radeonScreenPtr screen, int device_id) +static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id) { screen->chip_flags = 0; switch ( device_id ) { diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index bd613f4779..ebd18a1634 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -43,14 +43,154 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/glheader.h" #include "swrast/swrast.h" -#include "common_context.h" -#include "common_misc.h" +#include "radeon_common.h" +#include "radeon_lock.h" #include "radeon_span.h" #include "radeon_buffer.h" #define DBG 0 +static GLubyte *radeon_ptr32(const struct radeon_renderbuffer * rrb, + GLint x, GLint y) +{ + GLubyte *ptr = rrb->bo->ptr; + const __DRIdrawablePrivate *dPriv = rrb->dPriv; + uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE; + GLint offset; + GLint nmacroblkpl; + GLint nmicroblkpl; + + x += dPriv->x; + y += dPriv->y; + + if (rrb->has_surface || !(rrb->bo->flags & mask)) { + offset = x * rrb->cpp + y * rrb->pitch; + } else { + offset = 0; + if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) { + if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) { + nmacroblkpl = rrb->pitch >> 5; + offset += ((y >> 4) * nmacroblkpl) << 11; + offset += ((y & 15) >> 1) << 8; + offset += (y & 1) << 4; + offset += (x >> 5) << 11; + offset += ((x & 31) >> 2) << 5; + offset += (x & 3) << 2; + } else { + nmacroblkpl = rrb->pitch >> 6; + offset += ((y >> 3) * nmacroblkpl) << 11; + offset += (y & 7) << 8; + offset += (x >> 6) << 11; + offset += ((x & 63) >> 3) << 5; + offset += (x & 7) << 2; + } + } else { + nmicroblkpl = ((rrb->pitch + 31) & ~31) >> 5; + offset += (y * nmicroblkpl) << 5; + offset += (x >> 3) << 5; + offset += (x & 7) << 2; + } + } + return &ptr[offset]; +} + +static GLubyte *radeon_ptr16(const struct radeon_renderbuffer * rrb, + GLint x, GLint y) +{ + GLubyte *ptr = rrb->bo->ptr; + const __DRIdrawablePrivate *dPriv = rrb->dPriv; + uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE; + GLint offset; + GLint nmacroblkpl; + GLint nmicroblkpl; + + x += dPriv->x; + y += dPriv->y; + + if (rrb->has_surface || !(rrb->bo->flags & mask)) { + offset = x * rrb->cpp + y * rrb->pitch; + } else { + offset = 0; + if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) { + if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) { + nmacroblkpl = rrb->pitch >> 6; + offset += ((y >> 4) * nmacroblkpl) << 11; + offset += ((y & 15) >> 1) << 8; + offset += (y & 1) << 4; + offset += (x >> 6) << 11; + offset += ((x & 63) >> 3) << 5; + offset += (x & 7) << 1; + } else { + nmacroblkpl = rrb->pitch >> 7; + offset += ((y >> 3) * nmacroblkpl) << 11; + offset += (y & 7) << 8; + offset += (x >> 7) << 11; + offset += ((x & 127) >> 4) << 5; + offset += (x & 15) << 2; + } + } else { + nmicroblkpl = ((rrb->pitch + 31) & ~31) >> 5; + offset += (y * nmicroblkpl) << 5; + offset += (x >> 4) << 5; + offset += (x & 15) << 2; + } + } + return &ptr[offset]; +} + +static GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb, + GLint x, GLint y) +{ + GLubyte *ptr = rrb->bo->ptr; + const __DRIdrawablePrivate *dPriv = rrb->dPriv; + uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE; + GLint offset; + GLint microblkxs; + GLint macroblkxs; + GLint nmacroblkpl; + GLint nmicroblkpl; + + x += dPriv->x; + y += dPriv->y; + + if (rrb->has_surface || !(rrb->bo->flags & mask)) { + offset = x * rrb->cpp + y * rrb->pitch; + } else { + offset = 0; + if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) { + if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE) { + microblkxs = 16 / rrb->cpp; + macroblkxs = 128 / rrb->cpp; + nmacroblkpl = rrb->pitch / macroblkxs; + offset += ((y >> 4) * nmacroblkpl) << 11; + offset += ((y & 15) >> 1) << 8; + offset += (y & 1) << 4; + offset += (x / macroblkxs) << 11; + offset += ((x & (macroblkxs - 1)) / microblkxs) << 5; + offset += (x & (microblkxs - 1)) * rrb->cpp; + } else { + microblkxs = 32 / rrb->cpp; + macroblkxs = 256 / rrb->cpp; + nmacroblkpl = rrb->pitch / macroblkxs; + offset += ((y >> 3) * nmacroblkpl) << 11; + offset += (y & 7) << 8; + offset += (x / macroblkxs) << 11; + offset += ((x & (macroblkxs - 1)) / microblkxs) << 5; + offset += (x & (microblkxs - 1)) * rrb->cpp; + } + } else { + microblkxs = 32 / rrb->cpp; + nmicroblkpl = ((rrb->pitch + 31) & ~31) >> 5; + offset += (y * nmicroblkpl) << 5; + offset += (x / microblkxs) << 5; + offset += (x & (microblkxs - 1)) * rrb->cpp; + } + } + return &ptr[offset]; +} + + /* * Note that all information needed to access pixels in a renderbuffer * should be obtained through the gl_renderbuffer parameter, not per-context @@ -215,6 +355,88 @@ do { \ #define TAG(x) radeon##x##_z24_s8 #include "stenciltmp.h" + +static void map_buffer(struct gl_renderbuffer *rb, GLboolean write) +{ + struct radeon_renderbuffer *rrb = (void*)rb; + int r; + + if (rrb->bo) { + r = radeon_bo_map(rrb->bo, write); + if (r) { + fprintf(stderr, "(%s) error(%d) mapping buffer.\n", + __FUNCTION__, r); + } + } +} + +static void unmap_buffer(struct gl_renderbuffer *rb) +{ + struct radeon_renderbuffer *rrb = (void*)rb; + + if (rrb->bo) { + radeon_bo_unmap(rrb->bo); + } +} + +static void radeonSpanRenderStart(GLcontext * ctx) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + int i; + + radeon_firevertices(rmesa); + + for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled) + ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[i]._Current); + } + + /* color draw buffers */ + for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) { + map_buffer(ctx->DrawBuffer->_ColorDrawBuffers[i], GL_TRUE); + } + + map_buffer(ctx->ReadBuffer->_ColorReadBuffer, GL_FALSE); + + if (ctx->DrawBuffer->_DepthBuffer) { + map_buffer(ctx->DrawBuffer->_DepthBuffer->Wrapped, GL_TRUE); + } + if (ctx->DrawBuffer->_StencilBuffer) + map_buffer(ctx->DrawBuffer->_StencilBuffer->Wrapped, GL_TRUE); + + /* The locking and wait for idle should really only be needed in classic mode. + * In a future memory manager based implementation, this should become + * unnecessary due to the fact that mapping our buffers, textures, etc. + * should implicitly wait for any previous rendering commands that must + * be waited on. */ + LOCK_HARDWARE(rmesa); + radeonWaitForIdleLocked(rmesa); +} + +static void radeonSpanRenderFinish(GLcontext * ctx) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + int i; + _swrast_flush(ctx); + UNLOCK_HARDWARE(rmesa); + + for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled) + ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[i]._Current); + } + + /* color draw buffers */ + for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) + unmap_buffer(ctx->DrawBuffer->_ColorDrawBuffers[i]); + + unmap_buffer(ctx->ReadBuffer->_ColorReadBuffer); + + if (ctx->DrawBuffer->_DepthBuffer) + unmap_buffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + if (ctx->DrawBuffer->_StencilBuffer) + unmap_buffer(ctx->DrawBuffer->_StencilBuffer->Wrapped); +} + void radeonInitSpanFuncs(GLcontext * ctx) { struct swrast_device_driver *swdd = diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c index 7f90bc8090..86d872030b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.c +++ b/src/mesa/drivers/dri/radeon/radeon_state.c @@ -47,7 +47,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "swrast_setup/swrast_setup.h" #include "radeon_context.h" -#include "common_cmdbuf.h" #include "radeon_ioctl.h" #include "radeon_state.h" #include "radeon_tcl.h" diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index 1cfb539e71..a397c0824f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -38,7 +38,6 @@ #include "swrast_setup/swrast_setup.h" #include "radeon_context.h" -#include "common_cmdbuf.h" #include "radeon_cs.h" #include "radeon_mipmap_tree.h" #include "radeon_ioctl.h" diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 995ab2099e..af933a35f2 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -46,7 +46,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_pipeline.h" #include "radeon_context.h" -#include "common_cmdbuf.h" #include "radeon_ioctl.h" #include "radeon_state.h" #include "radeon_swtcl.h" diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index 7281800ce2..5887ab355d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -42,8 +42,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/tnl.h" #include "tnl/t_pipeline.h" +#include "radeon_common.h" #include "radeon_context.h" -#include "common_cmdbuf.h" #include "radeon_state.h" #include "radeon_ioctl.h" #include "radeon_tex.h" diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c new file mode 100644 index 0000000000..beb21e7b43 --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_texture.c @@ -0,0 +1,849 @@ +/* + * Copyright (C) 2008 Nicolai Haehnle. + * Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + * + * The Weather Channel (TM) funded Tungsten Graphics to develop the + * initial release of the Radeon 8500 driver under the XFree86 license. + * This notice must be preserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "main/glheader.h" +#include "main/imports.h" +#include "main/context.h" +#include "main/mipmap.h" +#include "main/texformat.h" +#include "main/texstore.h" +#include "main/teximage.h" +#include "main/texobj.h" + +#include "xmlpool.h" /* for symbolic values of enum-type options */ + +#include "radeon_common.h" + +#include "radeon_mipmap_tree.h" + +/* textures */ +/** + * Allocate an empty texture image object. + */ +struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx) +{ + return CALLOC(sizeof(radeon_texture_image)); +} + +/** + * Free memory associated with this texture image. + */ +void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage) +{ + radeon_texture_image* image = get_radeon_texture_image(timage); + + if (image->mt) { + radeon_miptree_unreference(image->mt); + image->mt = 0; + assert(!image->base.Data); + } else { + _mesa_free_texture_image_data(ctx, timage); + } + if (image->bo) { + radeon_bo_unref(image->bo); + image->bo = NULL; + } + if (timage->Data) { + _mesa_free_texmemory(timage->Data); + timage->Data = NULL; + } +} + +/* Set Data pointer and additional data for mapped texture image */ +static void teximage_set_map_data(radeon_texture_image *image) +{ + radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; + + image->base.Data = image->mt->bo->ptr + lvl->faces[image->mtface].offset; + image->base.RowStride = lvl->rowstride / image->mt->bpp; +} + + +/** + * Map a single texture image for glTexImage and friends. + */ +void radeon_teximage_map(radeon_texture_image *image, GLboolean write_enable) +{ + if (image->mt) { + assert(!image->base.Data); + + radeon_bo_map(image->mt->bo, write_enable); + teximage_set_map_data(image); + } +} + + +void radeon_teximage_unmap(radeon_texture_image *image) +{ + if (image->mt) { + assert(image->base.Data); + + image->base.Data = 0; + radeon_bo_unmap(image->mt->bo); + } +} + +/** + * Map a validated texture for reading during software rendering. + */ +void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj) +{ + radeonTexObj* t = radeon_tex_obj(texObj); + int face, level; + + /* for r100 3D sw fallbacks don't have mt */ + if (!t->mt) + return; + + radeon_bo_map(t->mt->bo, GL_FALSE); + for(face = 0; face < t->mt->faces; ++face) { + for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) + teximage_set_map_data(get_radeon_texture_image(texObj->Image[face][level])); + } +} + +void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj) +{ + radeonTexObj* t = radeon_tex_obj(texObj); + int face, level; + + /* for r100 3D sw fallbacks don't have mt */ + if (!t->mt) + return; + + for(face = 0; face < t->mt->faces; ++face) { + for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) + texObj->Image[face][level]->Data = 0; + } + radeon_bo_unmap(t->mt->bo); +} + +GLuint radeon_face_for_target(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: + return (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; + default: + return 0; + } +} + +/** + * Wraps Mesa's implementation to ensure that the base level image is mapped. + * + * This relies on internal details of _mesa_generate_mipmap, in particular + * the fact that the memory for recreated texture images is always freed. + */ +static void radeon_generate_mipmap(GLcontext *ctx, GLenum target, + struct gl_texture_object *texObj) +{ + radeonTexObj* t = radeon_tex_obj(texObj); + GLuint nr_faces = (t->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + int i, face; + + + _mesa_generate_mipmap(ctx, target, texObj); + + for (face = 0; face < nr_faces; face++) { + for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) { + radeon_texture_image *image; + + image = get_radeon_texture_image(texObj->Image[face][i]); + + if (image == NULL) + break; + + image->mtlevel = i; + image->mtface = face; + + radeon_miptree_unreference(image->mt); + image->mt = NULL; + } + } + +} + +void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj) +{ + GLuint face = radeon_face_for_target(target); + radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[face][texObj->BaseLevel]); + + radeon_teximage_map(baseimage, GL_FALSE); + radeon_generate_mipmap(ctx, target, texObj); + radeon_teximage_unmap(baseimage); +} + + +/* try to find a format which will only need a memcopy */ +static const struct gl_texture_format *radeonChoose8888TexFormat(radeonContextPtr rmesa, + GLenum srcFormat, + GLenum srcType) +{ + const GLuint ui = 1; + const GLubyte littleEndian = *((const GLubyte *)&ui); + + /* r100 can only do this */ + if (IS_R100_CLASS(rmesa->radeonScreen)) + return _dri_texformat_argb8888; + + if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8) || + (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) || + (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) || + (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) { + return &_mesa_texformat_rgba8888; + } else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) || + (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) || + (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) || + (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) { + return &_mesa_texformat_rgba8888_rev; + } else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && !littleEndian) || + srcType == GL_UNSIGNED_INT_8_8_8_8)) { + return &_mesa_texformat_argb8888_rev; + } else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && littleEndian) || + srcType == GL_UNSIGNED_INT_8_8_8_8_REV)) { + return &_mesa_texformat_argb8888; + } else + return _dri_texformat_argb8888; +} + +const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + const GLboolean do32bpt = + (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_32); + const GLboolean force16bpt = + (rmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FORCE_16); + (void)format; + +#if 0 + fprintf(stderr, "InternalFormat=%s(%d) type=%s format=%s\n", + _mesa_lookup_enum_by_nr(internalFormat), internalFormat, + _mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format)); + fprintf(stderr, "do32bpt=%d force16bpt=%d\n", do32bpt, force16bpt); +#endif + + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + switch (type) { + case GL_UNSIGNED_INT_10_10_10_2: + case GL_UNSIGNED_INT_2_10_10_10_REV: + return do32bpt ? _dri_texformat_argb8888 : + _dri_texformat_argb1555; + case GL_UNSIGNED_SHORT_4_4_4_4: + case GL_UNSIGNED_SHORT_4_4_4_4_REV: + return _dri_texformat_argb4444; + case GL_UNSIGNED_SHORT_5_5_5_1: + case GL_UNSIGNED_SHORT_1_5_5_5_REV: + return _dri_texformat_argb1555; + default: + return do32bpt ? radeonChoose8888TexFormat(rmesa, format, type) : + _dri_texformat_argb4444; + } + + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + switch (type) { + case GL_UNSIGNED_SHORT_4_4_4_4: + case GL_UNSIGNED_SHORT_4_4_4_4_REV: + return _dri_texformat_argb4444; + case GL_UNSIGNED_SHORT_5_5_5_1: + case GL_UNSIGNED_SHORT_1_5_5_5_REV: + return _dri_texformat_argb1555; + case GL_UNSIGNED_SHORT_5_6_5: + case GL_UNSIGNED_SHORT_5_6_5_REV: + return _dri_texformat_rgb565; + default: + return do32bpt ? _dri_texformat_argb8888 : + _dri_texformat_rgb565; + } + + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return !force16bpt ? + radeonChoose8888TexFormat(rmesa, format,type) : + _dri_texformat_argb4444; + + case GL_RGBA4: + case GL_RGBA2: + return _dri_texformat_argb4444; + + case GL_RGB5_A1: + return _dri_texformat_argb1555; + + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return !force16bpt ? _dri_texformat_argb8888 : + _dri_texformat_rgb565; + + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + return _dri_texformat_rgb565; + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + return _dri_texformat_a8; + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + return _dri_texformat_l8; + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + return _dri_texformat_al88; + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + return _dri_texformat_i8; + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_APPLE || + type == GL_UNSIGNED_BYTE) + return &_mesa_texformat_ycbcr; + else + return &_mesa_texformat_ycbcr_rev; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; + + case GL_ALPHA16F_ARB: + return &_mesa_texformat_alpha_float16; + case GL_ALPHA32F_ARB: + return &_mesa_texformat_alpha_float32; + case GL_LUMINANCE16F_ARB: + return &_mesa_texformat_luminance_float16; + case GL_LUMINANCE32F_ARB: + return &_mesa_texformat_luminance_float32; + case GL_LUMINANCE_ALPHA16F_ARB: + return &_mesa_texformat_luminance_alpha_float16; + case GL_LUMINANCE_ALPHA32F_ARB: + return &_mesa_texformat_luminance_alpha_float32; + case GL_INTENSITY16F_ARB: + return &_mesa_texformat_intensity_float16; + case GL_INTENSITY32F_ARB: + return &_mesa_texformat_intensity_float32; + case GL_RGB16F_ARB: + return &_mesa_texformat_rgba_float16; + case GL_RGB32F_ARB: + return &_mesa_texformat_rgba_float32; + case GL_RGBA16F_ARB: + return &_mesa_texformat_rgba_float16; + case GL_RGBA32F_ARB: + return &_mesa_texformat_rgba_float32; + + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: +#if 0 + switch (type) { + case GL_UNSIGNED_BYTE: + case GL_UNSIGNED_SHORT: + return &_mesa_texformat_z16; + case GL_UNSIGNED_INT: + return &_mesa_texformat_z32; + case GL_UNSIGNED_INT_24_8_EXT: + default: + return &_mesa_texformat_z24_s8; + } +#else + return &_mesa_texformat_z16; +#endif + + default: + _mesa_problem(ctx, + "unexpected internalFormat 0x%x in r300ChooseTextureFormat", + (int)internalFormat); + return NULL; + } + + return NULL; /* never get here */ +} + +/** + * All glTexImage calls go through this function. + */ +static void radeon_teximage( + GLcontext *ctx, int dims, + GLint face, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLsizei imageSize, + GLenum format, GLenum type, const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, + int compressed) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + radeonTexObj* t = radeon_tex_obj(texObj); + radeon_texture_image* image = get_radeon_texture_image(texImage); + + radeon_firevertices(rmesa); + + t->validated = GL_FALSE; + + /* Choose and fill in the texture format for this image */ + texImage->TexFormat = radeonChooseTextureFormat(ctx, internalFormat, format, type); + _mesa_set_fetch_functions(texImage, dims); + + if (texImage->TexFormat->TexelBytes == 0) { + texImage->IsCompressed = GL_TRUE; + texImage->CompressedSize = + ctx->Driver.CompressedTextureSize(ctx, texImage->Width, + texImage->Height, texImage->Depth, + texImage->TexFormat->MesaFormat); + } else { + texImage->IsCompressed = GL_FALSE; + texImage->CompressedSize = 0; + } + + /* Allocate memory for image */ + radeonFreeTexImageData(ctx, texImage); /* Mesa core only clears texImage->Data but not image->mt */ + + if (!t->mt) + radeon_try_alloc_miptree(rmesa, t, texImage, face, level); + if (t->mt && radeon_miptree_matches_image(t->mt, texImage, face, level)) { + image->mt = t->mt; + image->mtlevel = level - t->mt->firstLevel; + image->mtface = face; + radeon_miptree_reference(t->mt); + } else { + int size; + if (texImage->IsCompressed) { + size = texImage->CompressedSize; + } else { + size = texImage->Width * texImage->Height * texImage->Depth * texImage->TexFormat->TexelBytes; + } + texImage->Data = _mesa_alloc_texmemory(size); + } + + /* Upload texture image; note that the spec allows pixels to be NULL */ + if (compressed) { + pixels = _mesa_validate_pbo_compressed_teximage( + ctx, imageSize, pixels, packing, "glCompressedTexImage"); + } else { + pixels = _mesa_validate_pbo_teximage( + ctx, dims, width, height, depth, + format, type, pixels, packing, "glTexImage"); + } + + if (pixels) { + radeon_teximage_map(image, GL_TRUE); + + if (compressed) { + memcpy(texImage->Data, pixels, imageSize); + } else { + GLuint dstRowStride; + if (image->mt) { + radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; + dstRowStride = lvl->rowstride; + } else { + dstRowStride = texImage->Width * texImage->TexFormat->TexelBytes; + } + if (!texImage->TexFormat->StoreImage(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + } + + } + + /* SGIS_generate_mipmap */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + radeon_generate_mipmap(ctx, texObj->Target, texObj); + } + + if (pixels) + radeon_teximage_unmap(image); + + _mesa_unmap_teximage_pbo(ctx, packing); + + +} + +void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + radeon_teximage(ctx, 1, 0, level, internalFormat, width, 1, 1, + 0, format, type, pixels, packing, texObj, texImage, 0); +} + +void radeonTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) + +{ + GLuint face = radeon_face_for_target(target); + + radeon_teximage(ctx, 2, face, level, internalFormat, width, height, 1, + 0, format, type, pixels, packing, texObj, texImage, 0); +} + +void radeonCompressedTexImage2D(GLcontext * ctx, GLenum target, + GLint level, GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid * data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + GLuint face = radeon_face_for_target(target); + + radeon_teximage(ctx, 2, face, level, internalFormat, width, height, 1, + imageSize, 0, 0, data, 0, texObj, texImage, 1); +} + +void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + radeon_teximage(ctx, 3, 0, level, internalFormat, width, height, depth, + 0, format, type, pixels, packing, texObj, texImage, 0); +} + +/** + * Update a subregion of the given texture image. + */ +static void radeon_texsubimage(GLcontext* ctx, int dims, int level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, + int compressed) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + radeonTexObj* t = radeon_tex_obj(texObj); + radeon_texture_image* image = get_radeon_texture_image(texImage); + + radeon_firevertices(rmesa); + + t->validated = GL_FALSE; + pixels = _mesa_validate_pbo_teximage(ctx, dims, + width, height, depth, format, type, pixels, packing, "glTexSubImage1D"); + + if (pixels) { + GLint dstRowStride; + radeon_teximage_map(image, GL_TRUE); + + if (image->mt) { + radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; + dstRowStride = lvl->rowstride; + } else { + dstRowStride = texImage->RowStride * texImage->TexFormat->TexelBytes; + } + + if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); + + + } + + /* GL_SGIS_generate_mipmap */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + radeon_generate_mipmap(ctx, texObj->Target, texObj); + } + radeon_teximage_unmap(image); + + _mesa_unmap_teximage_pbo(ctx, packing); + + +} + +void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + radeon_texsubimage(ctx, 1, level, xoffset, 0, 0, width, 1, 1, + format, type, pixels, packing, texObj, texImage, 0); +} + +void radeonTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + radeon_texsubimage(ctx, 2, level, xoffset, yoffset, 0, width, height, + 1, format, type, pixels, packing, texObj, texImage, + 0); +} + +void radeonCompressedTexSubImage2D(GLcontext * ctx, GLenum target, + GLint level, GLint xoffset, + GLint yoffset, GLsizei width, + GLsizei height, GLenum format, + GLsizei imageSize, const GLvoid * data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + radeon_texsubimage(ctx, 2, level, xoffset, yoffset, 0, width, height, 1, + format, 0, data, 0, texObj, texImage, 1); +} + + +void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + radeon_texsubimage(ctx, 3, level, xoffset, yoffset, zoffset, width, height, depth, + format, type, pixels, packing, texObj, texImage, 0); +} + +static void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride, + GLuint numrows, GLuint rowsize) +{ + assert(rowsize <= dststride); + assert(rowsize <= srcstride); + + if (rowsize == srcstride && rowsize == dststride) { + memcpy(dst, src, numrows*rowsize); + } else { + GLuint i; + for(i = 0; i < numrows; ++i) { + memcpy(dst, src, rowsize); + dst += dststride; + src += srcstride; + } + } +} + + +/** + * Ensure that the given image is stored in the given miptree from now on. + */ +static void migrate_image_to_miptree(radeon_mipmap_tree *mt, radeon_texture_image *image, int face, int level) +{ + radeon_mipmap_level *dstlvl = &mt->levels[level - mt->firstLevel]; + unsigned char *dest; + + assert(image->mt != mt); + assert(dstlvl->width == image->base.Width); + assert(dstlvl->height == image->base.Height); + assert(dstlvl->depth == image->base.Depth); + + + radeon_bo_map(mt->bo, GL_TRUE); + dest = mt->bo->ptr + dstlvl->faces[face].offset; + + if (image->mt) { + /* Format etc. should match, so we really just need a memcpy(). + * In fact, that memcpy() could be done by the hardware in many + * cases, provided that we have a proper memory manager. + */ + radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel]; + + assert(srclvl->size == dstlvl->size); + assert(srclvl->rowstride == dstlvl->rowstride); + + radeon_bo_map(image->mt->bo, GL_FALSE); + + memcpy(dest, + image->mt->bo->ptr + srclvl->faces[face].offset, + dstlvl->size); + radeon_bo_unmap(image->mt->bo); + + radeon_miptree_unreference(image->mt); + } else { + uint srcrowstride = image->base.Width * image->base.TexFormat->TexelBytes; + +// if (mt->tilebits) +// WARN_ONCE("%s: tiling not supported yet", __FUNCTION__); + + copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride, + image->base.Height * image->base.Depth, srcrowstride); + + _mesa_free_texmemory(image->base.Data); + image->base.Data = 0; + } + + radeon_bo_unmap(mt->bo); + + image->mt = mt; + image->mtface = face; + image->mtlevel = level; + radeon_miptree_reference(image->mt); +} + +int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + radeonTexObj *t = radeon_tex_obj(texObj); + radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[0][texObj->BaseLevel]); + int face, level; + + if (t->validated || t->image_override) + return GL_TRUE; + + if (RADEON_DEBUG & DEBUG_TEXTURE) + fprintf(stderr, "%s: Validating texture %p now\n", __FUNCTION__, texObj); + + if (baseimage->base.Border > 0) + return GL_FALSE; + + /* Ensure a matching miptree exists. + * + * Differing mipmap trees can result when the app uses TexImage to + * change texture dimensions. + * + * Prefer to use base image's miptree if it + * exists, since that most likely contains more valid data (remember + * that the base level is usually significantly larger than the rest + * of the miptree, so cubemaps are the only possible exception). + */ + if (baseimage->mt && + baseimage->mt != t->mt && + radeon_miptree_matches_texture(baseimage->mt, &t->base)) { + radeon_miptree_unreference(t->mt); + t->mt = baseimage->mt; + radeon_miptree_reference(t->mt); + } else if (t->mt && !radeon_miptree_matches_texture(t->mt, &t->base)) { + radeon_miptree_unreference(t->mt); + t->mt = 0; + } + + if (!t->mt) { + if (RADEON_DEBUG & DEBUG_TEXTURE) + fprintf(stderr, " Allocate new miptree\n"); + radeon_try_alloc_miptree(rmesa, t, &baseimage->base, 0, texObj->BaseLevel); + if (!t->mt) { + _mesa_problem(ctx, "r300_validate_texture failed to alloc miptree"); + return GL_FALSE; + } + } + + /* Ensure all images are stored in the single main miptree */ + for(face = 0; face < t->mt->faces; ++face) { + for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) { + radeon_texture_image *image = get_radeon_texture_image(texObj->Image[face][level]); + if (RADEON_DEBUG & DEBUG_TEXTURE) + fprintf(stderr, " face %i, level %i... %p vs %p ", face, level, t->mt, image->mt); + if (t->mt == image->mt) { + if (RADEON_DEBUG & DEBUG_TEXTURE) + fprintf(stderr, "OK\n"); + continue; + } + + if (RADEON_DEBUG & DEBUG_TEXTURE) + fprintf(stderr, "migrating\n"); + migrate_image_to_miptree(t->mt, image, face, level); + } + } + + return GL_TRUE; +} diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.h b/src/mesa/drivers/dri/radeon/radeon_texture.h new file mode 100644 index 0000000000..ff0e1d061e --- /dev/null +++ b/src/mesa/drivers/dri/radeon/radeon_texture.h @@ -0,0 +1,108 @@ +/* + * Copyright (C) 2008 Nicolai Haehnle. + * Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + * + * The Weather Channel (TM) funded Tungsten Graphics to develop the + * initial release of the Radeon 8500 driver under the XFree86 license. + * This notice must be preserved. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef RADEON_TEXTURE_H +#define RADEON_TEXTURE_H +struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx); +void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage); + +void radeon_teximage_map(radeon_texture_image *image, GLboolean write_enable); +void radeon_teximage_unmap(radeon_texture_image *image); +void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj); +void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj); +void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj); +int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj); +GLuint radeon_face_for_target(GLenum target); +const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type); + +void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); +void radeonTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); +void radeonCompressedTexImage2D(GLcontext * ctx, GLenum target, + GLint level, GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid * data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); +void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); +void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); +void radeonTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); +void radeonCompressedTexSubImage2D(GLcontext * ctx, GLenum target, + GLint level, GLint xoffset, + GLint yoffset, GLsizei width, + GLsizei height, GLenum format, + GLsizei imageSize, const GLvoid * data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); +#endif -- cgit v1.2.3 From f577c8e462fc924ea436d129ad64c8a1226b5f9c Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Tue, 24 Mar 2009 18:32:46 +1000 Subject: radeon/r200/r300: fix warnings --- src/mesa/drivers/dri/r200/r200_cmdbuf.c | 15 +-------------- src/mesa/drivers/dri/r200/r200_context.c | 1 - src/mesa/drivers/dri/r200/r200_ioctl.c | 1 - src/mesa/drivers/dri/r300/r300_state.c | 6 +----- src/mesa/drivers/dri/radeon/radeon_common_context.c | 3 ++- src/mesa/drivers/dri/radeon/radeon_fbo.c | 5 +++-- src/mesa/drivers/dri/radeon/radeon_lock.c | 2 -- src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c | 16 ++++++++-------- src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h | 3 +-- src/mesa/drivers/dri/radeon/radeon_span.c | 3 +-- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 2 -- src/mesa/drivers/dri/radeon/radeon_texstate.c | 7 ------- 12 files changed, 17 insertions(+), 47 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c index ae31bcb831..83375c88a7 100644 --- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c +++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c @@ -112,7 +112,6 @@ void r200EmitVbufPrim( r200ContextPtr rmesa, GLuint primitive, GLuint vertex_nr ) { - drm_radeon_cmd_header_t *cmd; BATCH_LOCALS(&rmesa->radeon); assert(!(primitive & R200_VF_PRIM_WALK_IND)); @@ -164,8 +163,7 @@ static void r200FireEB(r200ContextPtr rmesa, int vertex_count, int type) void r200FlushElts(GLcontext *ctx) { - r200ContextPtr rmesa = R200_CONTEXT(ctx); - int dwords; + r200ContextPtr rmesa = R200_CONTEXT(ctx); int nr, elt_used = rmesa->tcl.elt_used; if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS)) @@ -345,14 +343,3 @@ void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset) } END_BATCH(); } - -void r200FireAOS(r200ContextPtr rmesa, int vertex_count, int type) -{ - BATCH_LOCALS(&rmesa->radeon); - - BEGIN_BATCH(3); - OUT_BATCH_PACKET3(R200_CP_CMD_3D_DRAW_VBUF_2, 0); - OUT_BATCH(R200_VF_PRIM_WALK_LIST | (vertex_count << 16) | type); - END_BATCH(); -} - diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c index d3898d0900..564e168290 100644 --- a/src/mesa/drivers/dri/r200/r200_context.c +++ b/src/mesa/drivers/dri/r200/r200_context.c @@ -243,7 +243,6 @@ static void r200_get_lock(radeonContextPtr radeon) { r200ContextPtr rmesa = (r200ContextPtr)radeon; drm_radeon_sarea_t *sarea = radeon->sarea; - int i; R200_STATECHANGE( rmesa, ctx ); if (rmesa->radeon.sarea->tiling_enabled) { diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c index ccb56202f6..b8e40da475 100644 --- a/src/mesa/drivers/dri/r200/r200_ioctl.c +++ b/src/mesa/drivers/dri/r200/r200_ioctl.c @@ -193,7 +193,6 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask ) __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable; GLuint flags = 0; GLuint color_mask = 0; - GLint ret; GLuint orig_mask = mask; if ( R200_DEBUG & DEBUG_IOCTL ) { diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index f49b43c207..6796d36d4c 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -42,6 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/macros.h" #include "main/context.h" #include "main/dd.h" +#include "main/framebuffer.h" #include "main/simple_list.h" #include "main/api_arrayelt.h" #include "main/texformat.h" @@ -2029,7 +2030,6 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state) static void r300ResetHwState(r300ContextPtr r300) { GLcontext *ctx = r300->radeon.glCtx; - struct radeon_renderbuffer *rrb; int has_tcl = 1; if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) @@ -2039,7 +2039,6 @@ static void r300ResetHwState(r300ContextPtr r300) fprintf(stderr, "%s\n", __FUNCTION__); radeon_firevertices(&r300->radeon); - //r300UpdateWindow(ctx); r300ColorMask(ctx, ctx->Color.ColorMask[RCOMP], @@ -2524,9 +2523,6 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state) */ void r300InitState(r300ContextPtr r300) { - GLcontext *ctx = r300->radeon.glCtx; - GLuint depth_fmt; - memset(&(r300->state.texture), 0, sizeof(r300->state.texture)); r300ResetHwState(r300); diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index 5b23473ab3..9964de7c7c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -36,6 +36,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "xmlpool.h" /* for symbolic values of enum-type options */ #include "utils.h" #include "vblank.h" +#include "drirenderbuffer.h" #include "main/state.h" #define DRIVER_DATE "20090101" @@ -398,7 +399,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) struct radeon_framebuffer *draw; radeonContextPtr radeon; char *regname; - struct radeon_bo *depth_bo, *bo; + struct radeon_bo *depth_bo = NULL, *bo; if (RADEON_DEBUG & DEBUG_DRI) fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable); diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c index 40945855d9..96ba5d3c60 100644 --- a/src/mesa/drivers/dri/radeon/radeon_fbo.c +++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c @@ -510,8 +510,9 @@ radeon_render_texture(GLcontext * ctx, att->TextureLevel); if (att->Texture->Target == GL_TEXTURE_3D) { - const GLuint *offsets = radeon_miptree_depth_offsets(radeon_image->mt, - att->TextureLevel); + GLuint offsets[6]; + radeon_miptree_depth_offsets(radeon_image->mt, att->TextureLevel, + offsets); imageOffset += offsets[att->Zoffset]; } diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.c b/src/mesa/drivers/dri/radeon/radeon_lock.c index e1bb2cd6e2..fe19218d7a 100644 --- a/src/mesa/drivers/dri/radeon/radeon_lock.c +++ b/src/mesa/drivers/dri/radeon/radeon_lock.c @@ -61,7 +61,6 @@ void radeonGetLock(radeonContextPtr rmesa, GLuint flags) __DRIdrawablePrivate *const drawable = rmesa->dri.drawable; __DRIdrawablePrivate *const readable = rmesa->dri.readable; __DRIscreenPrivate *sPriv = rmesa->dri.screen; - drm_radeon_sarea_t *sarea = rmesa->sarea; assert(drawable != NULL); @@ -92,7 +91,6 @@ void radeonGetLock(radeonContextPtr rmesa, GLuint flags) void radeon_lock_hardware(radeonContextPtr radeon) { - __DRIdrawable *dPriv = radeon->dri.drawable; char ret = 0; struct radeon_framebuffer *rfb = NULL; struct radeon_renderbuffer *rrb = NULL; diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c index 228629e3c4..34d6261706 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c @@ -94,8 +94,6 @@ static void compute_tex_image_offset(radeon_mipmap_tree *mt, /* Find image size in bytes */ if (mt->compressed) { /* TODO: Is this correct? Need test cases for compressed textures! */ - GLuint align; - lvl->rowstride = (lvl->width * mt->bpp + 63) & ~63; lvl->size = radeon_compressed_texture_size(mt->radeon->glCtx, lvl->width, lvl->height, lvl->depth, mt->compressed); @@ -365,14 +363,16 @@ void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t, * * These functions present that view to mesa: */ -const GLuint * -radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level) +void +radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets) { - static const GLuint zero = 0; if (mt->target != GL_TEXTURE_3D || mt->faces == 1) - return &zero; - else - return mt->levels[level].faces[0].offset; + offsets[0] = 0; + else { + int i; + for (i = 0; i < 6; i++) + offsets[i] = mt->levels[level].faces[i].offset; + } } GLuint diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h index d9ad5ad39a..697010bc02 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h @@ -94,6 +94,5 @@ void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t, struct gl_texture_image *texImage, GLuint face, GLuint level); GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt, GLuint face, GLuint level); -const GLuint * -radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level); +void radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets); #endif /* __RADEON_MIPMAP_TREE_H_ */ diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index 3d2c5da4c0..0b4eaae02b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -377,7 +377,7 @@ do { \ #include "stenciltmp.h" -void map_unmap_rb(struct gl_renderbuffer *rb, int flag) +static void map_unmap_rb(struct gl_renderbuffer *rb, int flag) { struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb); int r; @@ -403,7 +403,6 @@ void map_unmap_rb(struct gl_renderbuffer *rb, int flag) static void radeon_map_unmap_buffers(GLcontext *ctx, GLboolean map) { - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); GLuint i, j; /* color draw buffers */ diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index af933a35f2..2484006f1c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -825,6 +825,4 @@ void radeonInitSwtcl( GLcontext *ctx ) void radeonDestroySwtcl( GLcontext *ctx ) { - r100ContextPtr rmesa = R100_CONTEXT(ctx); - } diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index 78e2a08559..9931f01b63 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -775,11 +775,6 @@ void radeonSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPri static void disable_tex_obj_state( r100ContextPtr rmesa, int unit ) { - /* do not use RADEON_DB_STATE to avoid stale texture caches */ - uint32_t *cmd = &rmesa->hw.tex[unit].cmd[TEX_CMD_0]; - GLuint se_coord_fmt = rmesa->hw.set.cmd[SET_SE_COORDFMT]; - GLuint *txr_cmd = RADEON_DB_STATE( txr[unit] ); - RADEON_STATECHANGE( rmesa, tex[unit] ); RADEON_STATECHANGE( rmesa, tcl ); @@ -1142,8 +1137,6 @@ static GLboolean radeon_validate_texture(GLcontext *ctx, struct gl_texture_objec static GLboolean radeonUpdateTextureUnit( GLcontext *ctx, int unit ) { r100ContextPtr rmesa = R100_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - if (ctx->Texture.Unit[unit]._ReallyEnabled & TEXTURE_3D_BIT) { return GL_FALSE; -- cgit v1.2.3 From e00ef43d796f0ae0247b1072bf0aa8cdd8e3034d Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Thu, 2 Apr 2009 14:25:07 +1000 Subject: radeon/r200/r300: collapse context destruction down to a common path. Context destruction was nearly the same over all the drivers, so collapse it down. --- src/mesa/drivers/dri/r200/r200_cmdbuf.c | 96 ++++++++++----------- src/mesa/drivers/dri/r200/r200_context.c | 58 ------------- src/mesa/drivers/dri/r200/r200_context.h | 6 -- src/mesa/drivers/dri/r200/r200_maos.h | 1 - src/mesa/drivers/dri/r200/r200_maos_arrays.c | 24 ++---- src/mesa/drivers/dri/r200/r200_state.h | 4 - src/mesa/drivers/dri/r200/r200_state_init.c | 16 ---- src/mesa/drivers/dri/r200/r200_swtcl.c | 4 - src/mesa/drivers/dri/r200/r200_swtcl.h | 1 - src/mesa/drivers/dri/r200/r200_tcl.c | 16 ++-- src/mesa/drivers/dri/r300/r300_cmdbuf.c | 13 --- src/mesa/drivers/dri/r300/r300_cmdbuf.h | 2 - src/mesa/drivers/dri/r300/r300_context.c | 48 ----------- src/mesa/drivers/dri/r300/r300_context.h | 4 - src/mesa/drivers/dri/r300/r300_emit.c | 21 +---- src/mesa/drivers/dri/r300/r300_emit.h | 1 - src/mesa/drivers/dri/r300/r300_render.c | 98 +++++++++++----------- src/mesa/drivers/dri/radeon/Makefile | 3 +- .../drivers/dri/radeon/radeon_common_context.c | 90 +++++++++++++++----- .../drivers/dri/radeon/radeon_common_context.h | 12 ++- src/mesa/drivers/dri/radeon/radeon_context.c | 54 ++---------- src/mesa/drivers/dri/radeon/radeon_context.h | 24 ++---- src/mesa/drivers/dri/radeon/radeon_dma.c | 17 ++++ src/mesa/drivers/dri/radeon/radeon_dma.h | 1 + src/mesa/drivers/dri/radeon/radeon_ioctl.c | 76 ++++++++--------- src/mesa/drivers/dri/radeon/radeon_maos.h | 1 - src/mesa/drivers/dri/radeon/radeon_maos_arrays.c | 13 --- src/mesa/drivers/dri/radeon/radeon_maos_verts.c | 29 ++----- src/mesa/drivers/dri/radeon/radeon_screen.c | 46 +++------- src/mesa/drivers/dri/radeon/radeon_state.h | 3 - src/mesa/drivers/dri/radeon/radeon_state_init.c | 16 ---- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 4 - src/mesa/drivers/dri/radeon/radeon_swtcl.h | 1 - src/mesa/drivers/dri/radeon/radeon_tcl.c | 8 +- 34 files changed, 286 insertions(+), 525 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c index 83375c88a7..3a11a448ec 100644 --- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c +++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c @@ -143,18 +143,18 @@ static void r200FireEB(r200ContextPtr rmesa, int vertex_count, int type) if (!rmesa->radeon.radeonScreen->kernel_mm) { OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2); OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810); - OUT_BATCH_RELOC(rmesa->tcl.elt_dma_offset, - rmesa->tcl.elt_dma_bo, - rmesa->tcl.elt_dma_offset, + OUT_BATCH_RELOC(rmesa->radeon.tcl.elt_dma_offset, + rmesa->radeon.tcl.elt_dma_bo, + rmesa->radeon.tcl.elt_dma_offset, RADEON_GEM_DOMAIN_GTT, 0, 0); OUT_BATCH(vertex_count/2); } else { OUT_BATCH_PACKET3(R200_CP_CMD_INDX_BUFFER, 2); OUT_BATCH((0x80 << 24) | (0 << 16) | 0x810); - OUT_BATCH(rmesa->tcl.elt_dma_offset); + OUT_BATCH(rmesa->radeon.tcl.elt_dma_offset); OUT_BATCH(vertex_count/2); radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->tcl.elt_dma_bo, + rmesa->radeon.tcl.elt_dma_bo, RADEON_GEM_DOMAIN_GTT, 0, 0); } END_BATCH(); @@ -176,12 +176,12 @@ void r200FlushElts(GLcontext *ctx) nr = elt_used / 2; - radeon_bo_unmap(rmesa->tcl.elt_dma_bo); + radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo); r200FireEB(rmesa, nr, rmesa->tcl.hw_primitive); - radeon_bo_unref(rmesa->tcl.elt_dma_bo); - rmesa->tcl.elt_dma_bo = NULL; + radeon_bo_unref(rmesa->radeon.tcl.elt_dma_bo); + rmesa->radeon.tcl.elt_dma_bo = NULL; if (R200_DEBUG & DEBUG_SYNC) { fprintf(stderr, "%s: Syncing\n", __FUNCTION__); @@ -203,14 +203,14 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa, radeonEmitState(&rmesa->radeon); - rmesa->tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom, + rmesa->radeon.tcl.elt_dma_bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom, 0, R200_ELT_BUF_SZ, 4, RADEON_GEM_DOMAIN_GTT, 0); - rmesa->tcl.elt_dma_offset = 0; + rmesa->radeon.tcl.elt_dma_offset = 0; rmesa->tcl.elt_used = min_nr * 2; - radeon_bo_map(rmesa->tcl.elt_dma_bo, 1); - retval = rmesa->tcl.elt_dma_bo->ptr + rmesa->tcl.elt_dma_offset; + radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1); + retval = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset; if (R200_DEBUG & DEBUG_PRIMS) @@ -264,79 +264,79 @@ void r200EmitAOS(r200ContextPtr rmesa, GLuint nr, GLuint offset) if (!rmesa->radeon.radeonScreen->kernel_mm) { for (i = 0; i + 1 < nr; i += 2) { - OUT_BATCH((rmesa->tcl.aos[i].components << 0) | - (rmesa->tcl.aos[i].stride << 8) | - (rmesa->tcl.aos[i + 1].components << 16) | - (rmesa->tcl.aos[i + 1].stride << 24)); + OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) | + (rmesa->radeon.tcl.aos[i].stride << 8) | + (rmesa->radeon.tcl.aos[i + 1].components << 16) | + (rmesa->radeon.tcl.aos[i + 1].stride << 24)); - voffset = rmesa->tcl.aos[i + 0].offset + - offset * 4 * rmesa->tcl.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; OUT_BATCH_RELOC(voffset, - rmesa->tcl.aos[i].bo, + rmesa->radeon.tcl.aos[i].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); - voffset = rmesa->tcl.aos[i + 1].offset + - offset * 4 * rmesa->tcl.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; OUT_BATCH_RELOC(voffset, - rmesa->tcl.aos[i+1].bo, + rmesa->radeon.tcl.aos[i+1].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); } if (nr & 1) { - OUT_BATCH((rmesa->tcl.aos[nr - 1].components << 0) | - (rmesa->tcl.aos[nr - 1].stride << 8)); - voffset = rmesa->tcl.aos[nr - 1].offset + - offset * 4 * rmesa->tcl.aos[nr - 1].stride; + OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) | + (rmesa->radeon.tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; OUT_BATCH_RELOC(voffset, - rmesa->tcl.aos[nr - 1].bo, + rmesa->radeon.tcl.aos[nr - 1].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); } } else { for (i = 0; i + 1 < nr; i += 2) { - OUT_BATCH((rmesa->tcl.aos[i].components << 0) | - (rmesa->tcl.aos[i].stride << 8) | - (rmesa->tcl.aos[i + 1].components << 16) | - (rmesa->tcl.aos[i + 1].stride << 24)); + OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) | + (rmesa->radeon.tcl.aos[i].stride << 8) | + (rmesa->radeon.tcl.aos[i + 1].components << 16) | + (rmesa->radeon.tcl.aos[i + 1].stride << 24)); - voffset = rmesa->tcl.aos[i + 0].offset + - offset * 4 * rmesa->tcl.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; OUT_BATCH(voffset); - voffset = rmesa->tcl.aos[i + 1].offset + - offset * 4 * rmesa->tcl.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; OUT_BATCH(voffset); } if (nr & 1) { - OUT_BATCH((rmesa->tcl.aos[nr - 1].components << 0) | - (rmesa->tcl.aos[nr - 1].stride << 8)); - voffset = rmesa->tcl.aos[nr - 1].offset + - offset * 4 * rmesa->tcl.aos[nr - 1].stride; + OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) | + (rmesa->radeon.tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; OUT_BATCH(voffset); } for (i = 0; i + 1 < nr; i += 2) { - voffset = rmesa->tcl.aos[i + 0].offset + - offset * 4 * rmesa->tcl.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->tcl.aos[i+0].bo, + rmesa->radeon.tcl.aos[i+0].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); - voffset = rmesa->tcl.aos[i + 1].offset + - offset * 4 * rmesa->tcl.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->tcl.aos[i+1].bo, + rmesa->radeon.tcl.aos[i+1].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); } if (nr & 1) { - voffset = rmesa->tcl.aos[nr - 1].offset + - offset * 4 * rmesa->tcl.aos[nr - 1].stride; + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->tcl.aos[nr-1].bo, + rmesa->radeon.tcl.aos[nr-1].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); } diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c index 564e168290..f80f0d8ac7 100644 --- a/src/mesa/drivers/dri/r200/r200_context.c +++ b/src/mesa/drivers/dri/r200/r200_context.c @@ -491,61 +491,3 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, } -/* Destroy the device specific context. - */ -/* Destroy the Mesa and driver specific context data. - */ -void r200DestroyContext( __DRIcontextPrivate *driContextPriv ) -{ - GET_CURRENT_CONTEXT(ctx); - r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate; - r200ContextPtr current = ctx ? R200_CONTEXT(ctx) : NULL; - - /* check if we're deleting the currently bound context */ - if (rmesa == current) { - radeon_firevertices(&rmesa->radeon); - _mesa_make_current(NULL, NULL, NULL); - } - - /* Free r200 context resources */ - assert(rmesa); /* should never be null */ - if ( rmesa ) { - - _swsetup_DestroyContext( rmesa->radeon.glCtx ); - _tnl_DestroyContext( rmesa->radeon.glCtx ); - _vbo_DestroyContext( rmesa->radeon.glCtx ); - _swrast_DestroyContext( rmesa->radeon.glCtx ); - - r200DestroySwtcl( rmesa->radeon.glCtx ); - r200ReleaseArrays( rmesa->radeon.glCtx, ~0 ); - - if (rmesa->radeon.dma.current) { - radeonReleaseDmaRegion( &rmesa->radeon ); - rcommonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ ); - } - - if (rmesa->radeon.state.scissor.pClipRects) { - FREE(rmesa->radeon.state.scissor.pClipRects); - rmesa->radeon.state.scissor.pClipRects = NULL; - } - - radeonCleanupContext(&rmesa->radeon); - - FREE( rmesa ); - } -} - - - -/* Force the context `c' to be unbound from its buffer. - */ -GLboolean -r200UnbindContext( __DRIcontextPrivate *driContextPriv ) -{ - r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate; - - if (R200_DEBUG & DEBUG_DRI) - fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)rmesa->radeon.glCtx); - - return GL_TRUE; -} diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h index fcbe725d6f..6267293817 100644 --- a/src/mesa/drivers/dri/r200/r200_context.h +++ b/src/mesa/drivers/dri/r200/r200_context.h @@ -526,14 +526,8 @@ struct r200_state { struct r200_tcl_info { GLuint hw_primitive; -/* hw can handle 12 components max */ - struct radeon_aos aos[12]; - GLuint nr_aos_components; - GLuint *Elts; - struct radeon_bo *elt_dma_bo; - int elt_dma_offset; /** Offset into this buffer object, in bytes */ int elt_used; }; diff --git a/src/mesa/drivers/dri/r200/r200_maos.h b/src/mesa/drivers/dri/r200/r200_maos.h index d3ed06d402..16a70475e1 100644 --- a/src/mesa/drivers/dri/r200/r200_maos.h +++ b/src/mesa/drivers/dri/r200/r200_maos.h @@ -38,6 +38,5 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_context.h" extern void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev ); -extern void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs ); #endif diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c index 5dbc202330..383a0c4b0d 100644 --- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c +++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c @@ -142,17 +142,17 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev ) case 3: /* special handling to fix up fog. Will get us into trouble with vbos...*/ assert(attrib == VERT_ATTRIB_FOG); - if (!rmesa->tcl.aos[i].bo) { + if (!rmesa->radeon.tcl.aos[i].bo) { if (ctx->VertexProgram._Enabled) rcommon_emit_vector( ctx, - &(rmesa->tcl.aos[nr]), + &(rmesa->radeon.tcl.aos[nr]), (char *)VB->AttribPtr[attrib]->data, 1, VB->AttribPtr[attrib]->stride, count); else r200_emit_vecfog( ctx, - &(rmesa->tcl.aos[nr]), + &(rmesa->radeon.tcl.aos[nr]), (char *)VB->AttribPtr[attrib]->data, VB->AttribPtr[attrib]->stride, count); @@ -199,9 +199,9 @@ void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev ) default: assert(0); } - if (!rmesa->tcl.aos[nr].bo) { + if (!rmesa->radeon.tcl.aos[nr].bo) { rcommon_emit_vector( ctx, - &(rmesa->tcl.aos[nr]), + &(rmesa->radeon.tcl.aos[nr]), (char *)VB->AttribPtr[attrib]->data, emitsize, VB->AttribPtr[attrib]->stride, @@ -220,18 +220,6 @@ after_emit: rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = vfmt1; } - rmesa->tcl.nr_aos_components = nr; + rmesa->radeon.tcl.aos_count = nr; } - -void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs ) -{ - r200ContextPtr rmesa = R200_CONTEXT( ctx ); - int i; - for (i = 0; i < rmesa->tcl.nr_aos_components; i++) { - if (rmesa->tcl.aos[i].bo) { - radeon_bo_unref(rmesa->tcl.aos[i].bo); - rmesa->tcl.aos[i].bo = NULL; - } - } -} diff --git a/src/mesa/drivers/dri/r200/r200_state.h b/src/mesa/drivers/dri/r200/r200_state.h index 1dddbfdbfe..23cf8aea66 100644 --- a/src/mesa/drivers/dri/r200/r200_state.h +++ b/src/mesa/drivers/dri/r200/r200_state.h @@ -49,10 +49,6 @@ extern void r200UpdateDrawBuffer(GLcontext *ctx); extern GLboolean r200ValidateState( GLcontext *ctx ); -extern void r200PrintDirty( r200ContextPtr rmesa, - const char *msg ); - - extern void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode ); #define FALLBACK( rmesa, bit, mode ) do { \ if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n", \ diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c index 2400abe838..75262e46bd 100644 --- a/src/mesa/drivers/dri/r200/r200_state_init.c +++ b/src/mesa/drivers/dri/r200/r200_state_init.c @@ -166,22 +166,6 @@ static struct { /* ============================================================= * State initialization */ - -void r200PrintDirty( r200ContextPtr rmesa, const char *msg ) -{ - struct radeon_state_atom *l; - - fprintf(stderr, msg); - fprintf(stderr, ": "); - - foreach(l, &rmesa->radeon.hw.atomlist) { - if (l->dirty || rmesa->radeon.hw.all_dirty) - fprintf(stderr, "%s, ", l->name); - } - - fprintf(stderr, "\n"); -} - static int cmdpkt( r200ContextPtr rmesa, int id ) { drm_radeon_cmd_header_t h; diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index b006409987..712da98077 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -908,7 +908,3 @@ void r200InitSwtcl( GLcontext *ctx ) rmesa->radeon.swtcl.hw_primitive = 0; } - -void r200DestroySwtcl( GLcontext *ctx ) -{ -} diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.h b/src/mesa/drivers/dri/r200/r200_swtcl.h index a4051a4b76..b0905879d7 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.h +++ b/src/mesa/drivers/dri/r200/r200_swtcl.h @@ -39,7 +39,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_context.h" extern void r200InitSwtcl( GLcontext *ctx ); -extern void r200DestroySwtcl( GLcontext *ctx ); extern void r200ChooseRenderState( GLcontext *ctx ); extern void r200ChooseVertexState( GLcontext *ctx ); diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c index 8e0fb14e70..580370933e 100644 --- a/src/mesa/drivers/dri/r200/r200_tcl.c +++ b/src/mesa/drivers/dri/r200/r200_tcl.c @@ -145,7 +145,7 @@ static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr ) if (rmesa->radeon.dma.flush == r200FlushElts && rmesa->tcl.elt_used + nr*2 < R200_ELT_BUF_SZ) { - GLushort *dest = (GLushort *)(rmesa->tcl.elt_dma_bo->ptr + + GLushort *dest = (GLushort *)(rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->tcl.elt_used); rmesa->tcl.elt_used += nr*2; @@ -156,10 +156,10 @@ static GLushort *r200AllocElts( r200ContextPtr rmesa, GLuint nr ) if (rmesa->radeon.dma.flush) rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); - rcommonEnsureCmdBufSpace(&rmesa->radeon, AOS_BUFSZ(rmesa->tcl.nr_aos_components), __FUNCTION__); + rcommonEnsureCmdBufSpace(&rmesa->radeon, AOS_BUFSZ(rmesa->radeon.tcl.aos_count), __FUNCTION__); r200EmitAOS( rmesa, - rmesa->tcl.nr_aos_components, 0 ); + rmesa->radeon.tcl.aos_count, 0 ); return r200AllocEltsOpenEnded( rmesa, rmesa->tcl.hw_primitive, nr ); } @@ -186,13 +186,13 @@ static void r200EmitPrim( GLcontext *ctx, r200ContextPtr rmesa = R200_CONTEXT( ctx ); r200TclPrimitive( ctx, prim, hwprim ); - // fprintf(stderr,"Emit prim %d\n", rmesa->tcl.nr_aos_components); + // fprintf(stderr,"Emit prim %d\n", rmesa->radeon.tcl.aos_count); rcommonEnsureCmdBufSpace( &rmesa->radeon, - AOS_BUFSZ(rmesa->tcl.nr_aos_components) + + AOS_BUFSZ(rmesa->radeon.tcl.aos_count) + rmesa->radeon.hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); r200EmitAOS( rmesa, - rmesa->tcl.nr_aos_components, + rmesa->radeon.tcl.aos_count, start ); /* Why couldn't this packet have taken an offset param? @@ -481,7 +481,7 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx, /* Do the actual work: */ - r200ReleaseArrays( ctx, ~0 /* stage->changed_inputs */ ); + radeonReleaseArrays( ctx, ~0 /* stage->changed_inputs */ ); r200EmitArrays( ctx, vimap_rev ); rmesa->tcl.Elts = VB->Elts; @@ -545,7 +545,7 @@ static void transition_to_swtnl( GLcontext *ctx ) tnl->Driver.NotifyMaterialChange = _mesa_validate_all_lighting_tables; - r200ReleaseArrays( ctx, ~0 ); + radeonReleaseArrays( ctx, ~0 ); /* Still using the D3D based hardware-rasterizer from the radeon; * need to put the card into D3D mode to make it work: diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c index fc8a2e7431..1ecbeea489 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -649,16 +649,3 @@ void r300InitCmdBuf(r300ContextPtr r300) rcommonInitCmdBuf(&r300->radeon); } - -/** - * Destroy the command buffer and state atoms. - */ -void r300DestroyCmdBuf(r300ContextPtr r300) -{ - struct radeon_state_atom *atom; - - foreach(atom, &r300->radeon.hw.atomlist) { - FREE(atom->cmd); - } - -} diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h index b7798eb97b..3786813de3 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.h +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h @@ -39,8 +39,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" extern void r300InitCmdBuf(r300ContextPtr r300); -extern void r300DestroyCmdBuf(r300ContextPtr r300); - void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom); int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom); diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c index 5e271c6010..06db7ab8ff 100644 --- a/src/mesa/drivers/dri/r300/r300_context.c +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -470,51 +470,3 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual, return GL_TRUE; } -/* Destroy the device specific context. - */ -void r300DestroyContext(__DRIcontextPrivate * driContextPriv) -{ - GET_CURRENT_CONTEXT(ctx); - r300ContextPtr r300 = (r300ContextPtr) driContextPriv->driverPrivate; - radeonContextPtr radeon = (radeonContextPtr) r300; - radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL; - - if (RADEON_DEBUG & DEBUG_DRI) { - fprintf(stderr, "Destroying context !\n"); - } - - /* check if we're deleting the currently bound context */ - if (&r300->radeon == current) { - radeonFlush(r300->radeon.glCtx); - _mesa_make_current(NULL, NULL, NULL); - } - - /* Free r300 context resources */ - assert(r300); /* should never be null */ - - if (r300) { - _swsetup_DestroyContext(r300->radeon.glCtx); - _tnl_DestroyContext(r300->radeon.glCtx); - _vbo_DestroyContext(r300->radeon.glCtx); - _swrast_DestroyContext(r300->radeon.glCtx); - - radeon_firevertices(&r300->radeon); - - if (radeon->state.scissor.pClipRects) { - FREE(radeon->state.scissor.pClipRects); - radeon->state.scissor.pClipRects = NULL; - } - - r300DestroyCmdBuf(r300); - - radeonCleanupContext(&r300->radeon); - - - /* the memory manager might be accessed when Mesa frees the shared - * state, so don't destroy it earlier - */ - - - FREE(r300); - } -} diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 37718f5415..5ef59d258e 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -633,11 +633,7 @@ struct r300_state { struct r300_texture_state texture; int sw_tcl_inputs[VERT_ATTRIB_MAX]; struct r300_vertex_shader_state vertex_shader; - struct radeon_aos aos[R300_MAX_AOS_ARRAYS]; - int aos_count; - struct radeon_bo *elt_dma_bo; /** Buffer object that contains element indices */ - int elt_dma_offset; /** Offset into this buffer object, in bytes */ DECLARE_RENDERINPUTS(render_inputs_bitset); /* actual render inputs that R300 was configured for. They are the same as tnl->render_inputs for fixed pipeline */ diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c index 4bf0e7a189..3ce0ba68c9 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.c +++ b/src/mesa/drivers/dri/r300/r300_emit.c @@ -302,7 +302,7 @@ int r300EmitArrays(GLcontext * ctx) for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) { swizzle[i][ci] = ci; } - rcommon_emit_vector(ctx, &rmesa->state.aos[i], + rcommon_emit_vector(ctx, &rmesa->radeon.tcl.aos[i], vb->AttribPtr[tab[i]]->data, vb->AttribPtr[tab[i]]->size, vb->AttribPtr[tab[i]]->stride, count); @@ -343,28 +343,11 @@ int r300EmitArrays(GLcontext * ctx) rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten); - rmesa->state.aos_count = nr; + rmesa->radeon.tcl.aos_count = nr; return R300_FALLBACK_NONE; } -void r300ReleaseArrays(GLcontext * ctx) -{ - r300ContextPtr rmesa = R300_CONTEXT(ctx); - int i; - - if (rmesa->state.elt_dma_bo) { - radeon_bo_unref(rmesa->state.elt_dma_bo); - rmesa->state.elt_dma_bo = NULL; - } - for (i = 0; i < rmesa->state.aos_count; i++) { - if (rmesa->state.aos[i].bo) { - radeon_bo_unref(rmesa->state.aos[i].bo); - rmesa->state.aos[i].bo = NULL; - } - } -} - void r300EmitCacheFlush(r300ContextPtr rmesa) { BATCH_LOCALS(&rmesa->radeon); diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h index 6bc8f8e944..80c22d5e9a 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.h +++ b/src/mesa/drivers/dri/r300/r300_emit.h @@ -218,7 +218,6 @@ void static INLINE cp_wait(radeonContextPtr radeon, unsigned char flags) extern int r300EmitArrays(GLcontext * ctx); -extern void r300ReleaseArrays(GLcontext * ctx); extern int r300PrimitiveType(r300ContextPtr rmesa, int prim); extern int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim); diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index 829d088033..924305dd12 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -175,12 +175,12 @@ static void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts) r300ContextPtr rmesa = R300_CONTEXT(ctx); void *out; - radeonAllocDmaRegion(&rmesa->radeon, &rmesa->state.elt_dma_bo, - &rmesa->state.elt_dma_offset, n_elts * 4, 4); - radeon_bo_map(rmesa->state.elt_dma_bo, 1); - out = rmesa->state.elt_dma_bo->ptr + rmesa->state.elt_dma_offset; + radeonAllocDmaRegion(&rmesa->radeon, &rmesa->radeon.tcl.elt_dma_bo, + &rmesa->radeon.tcl.elt_dma_offset, n_elts * 4, 4); + radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1); + out = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset; memcpy(out, elts, n_elts * 4); - radeon_bo_unmap(rmesa->state.elt_dma_bo); + radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo); } static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type) @@ -199,19 +199,19 @@ static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type) OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2); OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) | (R300_VAP_PORT_IDX0 >> 2)); - OUT_BATCH_RELOC(rmesa->state.elt_dma_offset, - rmesa->state.elt_dma_bo, - rmesa->state.elt_dma_offset, + OUT_BATCH_RELOC(rmesa->radeon.tcl.elt_dma_offset, + rmesa->radeon.tcl.elt_dma_bo, + rmesa->radeon.tcl.elt_dma_offset, RADEON_GEM_DOMAIN_GTT, 0, 0); OUT_BATCH(vertex_count); } else { OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2); OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) | (R300_VAP_PORT_IDX0 >> 2)); - OUT_BATCH(rmesa->state.elt_dma_offset); + OUT_BATCH(rmesa->radeon.tcl.elt_dma_offset); OUT_BATCH(vertex_count); radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->state.elt_dma_bo, + rmesa->radeon.tcl.elt_dma_bo, RADEON_GEM_DOMAIN_GTT, 0, 0); } END_BATCH(); @@ -236,34 +236,34 @@ static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset) OUT_BATCH(nr); for (i = 0; i + 1 < nr; i += 2) { - OUT_BATCH((rmesa->state.aos[i].components << 0) | - (rmesa->state.aos[i].stride << 8) | - (rmesa->state.aos[i + 1].components << 16) | - (rmesa->state.aos[i + 1].stride << 24)); + OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) | + (rmesa->radeon.tcl.aos[i].stride << 8) | + (rmesa->radeon.tcl.aos[i + 1].components << 16) | + (rmesa->radeon.tcl.aos[i + 1].stride << 24)); - voffset = rmesa->state.aos[i + 0].offset + - offset * 4 * rmesa->state.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; OUT_BATCH_RELOC(voffset, - rmesa->state.aos[i].bo, + rmesa->radeon.tcl.aos[i].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); - voffset = rmesa->state.aos[i + 1].offset + - offset * 4 * rmesa->state.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; OUT_BATCH_RELOC(voffset, - rmesa->state.aos[i+1].bo, + rmesa->radeon.tcl.aos[i+1].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); } if (nr & 1) { - OUT_BATCH((rmesa->state.aos[nr - 1].components << 0) | - (rmesa->state.aos[nr - 1].stride << 8)); - voffset = rmesa->state.aos[nr - 1].offset + - offset * 4 * rmesa->state.aos[nr - 1].stride; + OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) | + (rmesa->radeon.tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; OUT_BATCH_RELOC(voffset, - rmesa->state.aos[nr - 1].bo, + rmesa->radeon.tcl.aos[nr - 1].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); @@ -276,45 +276,45 @@ static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset) OUT_BATCH(nr); for (i = 0; i + 1 < nr; i += 2) { - OUT_BATCH((rmesa->state.aos[i].components << 0) | - (rmesa->state.aos[i].stride << 8) | - (rmesa->state.aos[i + 1].components << 16) | - (rmesa->state.aos[i + 1].stride << 24)); + OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) | + (rmesa->radeon.tcl.aos[i].stride << 8) | + (rmesa->radeon.tcl.aos[i + 1].components << 16) | + (rmesa->radeon.tcl.aos[i + 1].stride << 24)); - voffset = rmesa->state.aos[i + 0].offset + - offset * 4 * rmesa->state.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; OUT_BATCH(voffset); - voffset = rmesa->state.aos[i + 1].offset + - offset * 4 * rmesa->state.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; OUT_BATCH(voffset); } if (nr & 1) { - OUT_BATCH((rmesa->state.aos[nr - 1].components << 0) | - (rmesa->state.aos[nr - 1].stride << 8)); - voffset = rmesa->state.aos[nr - 1].offset + - offset * 4 * rmesa->state.aos[nr - 1].stride; + OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) | + (rmesa->radeon.tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; OUT_BATCH(voffset); } for (i = 0; i + 1 < nr; i += 2) { - voffset = rmesa->state.aos[i + 0].offset + - offset * 4 * rmesa->state.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->state.aos[i+0].bo, + rmesa->radeon.tcl.aos[i+0].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); - voffset = rmesa->state.aos[i + 1].offset + - offset * 4 * rmesa->state.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->state.aos[i+1].bo, + rmesa->radeon.tcl.aos[i+1].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); } if (nr & 1) { - voffset = rmesa->state.aos[nr - 1].offset + - offset * 4 * rmesa->state.aos[nr - 1].stride; + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->state.aos[nr-1].bo, + rmesa->radeon.tcl.aos[nr-1].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); } @@ -370,10 +370,10 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx, * arrays. *sigh* */ r300EmitElts(ctx, vb->Elts, num_verts); - r300EmitAOS(rmesa, rmesa->state.aos_count, start); + r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, start); r300FireEB(rmesa, num_verts, type); } else { - r300EmitAOS(rmesa, rmesa->state.aos_count, start); + r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, start); r300FireAOS(rmesa, num_verts, type); } COMMIT_BATCH(); @@ -408,7 +408,7 @@ static GLboolean r300RunRender(GLcontext * ctx, r300EmitCacheFlush(rmesa); - r300ReleaseArrays(ctx); + radeonReleaseArrays(ctx, ~0); return GL_FALSE; } diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile index f469c6f43a..ba409ba813 100644 --- a/src/mesa/drivers/dri/radeon/Makefile +++ b/src/mesa/drivers/dri/radeon/Makefile @@ -19,7 +19,8 @@ RADEON_COMMON_SOURCES = \ radeon_bo_legacy.c \ radeon_cs_legacy.c \ radeon_mipmap_tree.c \ - radeon_span.c + radeon_span.c \ + radeon_fbo.c DRIVER_SOURCES = \ radeon_context.c \ diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index 9103c8c0f6..4d61b00345 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -37,8 +37,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "utils.h" #include "vblank.h" #include "drirenderbuffer.h" +#include "main/context.h" #include "main/framebuffer.h" #include "main/state.h" +#include "main/simple_list.h" +#include "swrast/swrast.h" +#include "swrast_setup/swrast_setup.h" +#include "tnl/tnl.h" #define DRIVER_DATE "20090101" @@ -175,39 +180,81 @@ GLboolean radeonInitContext(radeonContextPtr radeon, return GL_TRUE; } + + +/** + * Destroy the command buffer and state atoms. + */ +static void radeon_destroy_atom_list(radeonContextPtr radeon) +{ + struct radeon_state_atom *atom; + + foreach(atom, &radeon->hw.atomlist) { + FREE(atom->cmd); + if (atom->lastcmd) + FREE(atom->lastcmd); + } + +} + /** * Cleanup common context fields. * Called by r200DestroyContext/r300DestroyContext */ -void radeonCleanupContext(radeonContextPtr radeon) +void radeonDestroyContext(__DRIcontextPrivate *driContextPriv ) { #ifdef RADEON_BO_TRACK FILE *track; #endif - struct radeon_framebuffer *rfb; - - radeonDestroyBuffer(radeon->dri.drawable); - radeonDestroyBuffer(radeon->dri.readable); - - /* free the Mesa context */ - _mesa_destroy_context(radeon->glCtx); - - /* _mesa_destroy_context() might result in calls to functions that - * depend on the DriverCtx, so don't set it to NULL before. - * - * radeon->glCtx->DriverCtx = NULL; - */ - + GET_CURRENT_CONTEXT(ctx); + radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; + radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL; + if (radeon == current) { + radeon_firevertices(radeon); + _mesa_make_current(NULL, NULL, NULL); + } + + assert(radeon); + if (radeon) { - /* free the option cache */ - driDestroyOptionCache(&radeon->optionCache); + if (radeon->dma.current) { + radeonReleaseDmaRegion( radeon ); + rcommonFlushCmdBuf( radeon, __FUNCTION__ ); + } - rcommonDestroyCmdBuf(radeon); + radeonReleaseArrays(ctx, ~0); - if (radeon->state.scissor.pClipRects) { - FREE(radeon->state.scissor.pClipRects); - radeon->state.scissor.pClipRects = 0; + if (radeon->vtbl.free_context) + radeon->vtbl.free_context(radeon->glCtx); + _swsetup_DestroyContext( radeon->glCtx ); + _tnl_DestroyContext( radeon->glCtx ); + _vbo_DestroyContext( radeon->glCtx ); + _swrast_DestroyContext( radeon->glCtx ); + + radeonDestroyBuffer(radeon->dri.drawable); + radeonDestroyBuffer(radeon->dri.readable); + + /* free atom list */ + /* free the Mesa context */ + _mesa_destroy_context(radeon->glCtx); + + /* _mesa_destroy_context() might result in calls to functions that + * depend on the DriverCtx, so don't set it to NULL before. + * + * radeon->glCtx->DriverCtx = NULL; + */ + /* free the option cache */ + driDestroyOptionCache(&radeon->optionCache); + + rcommonDestroyCmdBuf(radeon); + + radeon_destroy_atom_list(radeon); + + if (radeon->state.scissor.pClipRects) { + FREE(radeon->state.scissor.pClipRects); + radeon->state.scissor.pClipRects = 0; + } } #ifdef RADEON_BO_TRACK track = fopen("/tmp/tracklog", "w"); @@ -216,6 +263,7 @@ void radeonCleanupContext(radeonContextPtr radeon) fclose(track); } #endif + FREE(radeon); } /* Force the context `c' to be unbound from its buffer. diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index 0ce72c9198..c6e6be7484 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -290,6 +290,14 @@ struct radeon_swtcl_info { }; +#define RADEON_MAX_AOS_ARRAYS 16 +struct radeon_tcl_info { + struct radeon_aos aos[RADEON_MAX_AOS_ARRAYS]; + GLuint aos_count; + struct radeon_bo *elt_dma_bo; /** Buffer object that contains element indices */ + int elt_dma_offset; /** Offset into this buffer object, in bytes */ +}; + struct radeon_ioctl { GLuint vertex_offset; struct radeon_bo *bo; @@ -367,7 +375,6 @@ struct radeon_dri_mirror { #define DEBUG_MEMORY 0x4000 - typedef void (*radeon_tri_func) (radeonContextPtr, radeonVertex *, radeonVertex *, radeonVertex *); @@ -436,6 +443,7 @@ struct radeon_context { struct radeon_state state; struct radeon_swtcl_info swtcl; + struct radeon_tcl_info tcl; /* Configuration cache */ driOptionCache optionCache; @@ -468,6 +476,7 @@ struct radeon_context { void (*pre_emit_atoms)(radeonContextPtr rmesa); void (*pre_emit_state)(radeonContextPtr rmesa); void (*fallback)(GLcontext *ctx, GLuint bit, GLboolean mode); + void (*free_context)(GLcontext *ctx); } vtbl; }; @@ -530,6 +539,7 @@ void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv); +extern void radeonDestroyContext(__DRIcontextPrivate * driContextPriv); /* ================================================================ * Debugging: diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index cad2c3e0c1..2600c78df3 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -188,6 +188,11 @@ static void r100_vtbl_pre_emit_state(radeonContextPtr radeon) radeon->hw.is_dirty = 1; } +static void r100_vtbl_free_context(GLcontext *ctx) +{ + r100ContextPtr rmesa = R100_CONTEXT(ctx); + _mesa_vector4f_free( &rmesa->tcl.ObjClean ); +} static void r100_init_vtbl(radeonContextPtr radeon) { @@ -202,7 +207,7 @@ static void r100_init_vtbl(radeonContextPtr radeon) /* Create the device specific context. */ GLboolean -radeonCreateContext( const __GLcontextModes *glVisual, +r100CreateContext( const __GLcontextModes *glVisual, __DRIcontextPrivate *driContextPriv, void *sharedContextPrivate) { @@ -397,50 +402,3 @@ radeonCreateContext( const __GLcontextModes *glVisual, } return GL_TRUE; } - - -/* Destroy the device specific context. - */ -/* Destroy the Mesa and driver specific context data. - */ -void radeonDestroyContext( __DRIcontextPrivate *driContextPriv ) -{ - GET_CURRENT_CONTEXT(ctx); - r100ContextPtr rmesa = (r100ContextPtr) driContextPriv->driverPrivate; - r100ContextPtr current = ctx ? R100_CONTEXT(ctx) : NULL; - - /* check if we're deleting the currently bound context */ - if (rmesa == current) { - radeon_firevertices(&rmesa->radeon); - _mesa_make_current(NULL, NULL, NULL); - } - - /* Free radeon context resources */ - assert(rmesa); /* should never be null */ - if ( rmesa ) { - - _swsetup_DestroyContext( rmesa->radeon.glCtx ); - _tnl_DestroyContext( rmesa->radeon.glCtx ); - _vbo_DestroyContext( rmesa->radeon.glCtx ); - _swrast_DestroyContext( rmesa->radeon.glCtx ); - - radeonDestroySwtcl( rmesa->radeon.glCtx ); - radeonReleaseArrays( rmesa->radeon.glCtx, ~0 ); - if (rmesa->radeon.dma.current) { - radeonReleaseDmaRegion( &rmesa->radeon ); - rcommonFlushCmdBuf( &rmesa->radeon, __FUNCTION__ ); - } - - _mesa_vector4f_free( &rmesa->tcl.ObjClean ); - - if (rmesa->radeon.state.scissor.pClipRects) { - FREE(rmesa->radeon.state.scissor.pClipRects); - rmesa->radeon.state.scissor.pClipRects = NULL; - } - - radeonCleanupContext(&rmesa->radeon); - - FREE( rmesa ); - } -} - diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h index 5235a6e374..1795d8bdb6 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_context.h @@ -332,7 +332,7 @@ struct r100_state { #define R200_ELT_BUF_SZ (8*1024) /* radeon_tcl.c */ -struct radeon_tcl_info { +struct r100_tcl_info { GLuint vertex_format; GLuint hw_primitive; @@ -341,14 +341,9 @@ struct radeon_tcl_info { */ GLvector4f ObjClean; - struct radeon_aos aos[8]; - GLuint nr_aos_components; - GLuint *Elts; - struct radeon_bo *indexed_bo; - - int elt_cmd_offset; /** Offset into the cmdbuf */ + int elt_cmd_offset; int elt_cmd_start; int elt_used; }; @@ -416,7 +411,7 @@ struct r100_context { /* radeon_tcl.c */ - struct radeon_tcl_info tcl; + struct r100_tcl_info tcl; /* radeon_swtcl.c */ @@ -443,15 +438,10 @@ struct r100_context { #define RADEON_OLD_PACKETS 1 -extern void radeonDestroyContext(__DRIcontextPrivate * driContextPriv); -extern GLboolean radeonCreateContext(const __GLcontextModes * glVisual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); -extern GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); -extern GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv); - +extern GLboolean r100CreateContext( const __GLcontextModes *glVisual, + __DRIcontextPrivate *driContextPriv, + void *sharedContextPrivate); + #endif /* __RADEON_CONTEXT_H__ */ diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 01fc20bd79..47f789e9cd 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -328,3 +328,20 @@ restart: rmesa->swtcl.numverts += nverts; return head; } + +void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) +{ + radeonContextPtr radeon = RADEON_CONTEXT( ctx ); + int i; + + if (radeon->tcl.elt_dma_bo) { + radeon_bo_unref(radeon->tcl.elt_dma_bo); + radeon->tcl.elt_dma_bo = NULL; + } + for (i = 0; i < radeon->tcl.aos_count; i++) { + if (radeon->tcl.aos[i].bo) { + radeon_bo_unref(radeon->tcl.aos[i].bo); + radeon->tcl.aos[i].bo = NULL; + } + } +} diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.h b/src/mesa/drivers/dri/radeon/radeon_dma.h index cee3744fed..06e388fc1d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.h +++ b/src/mesa/drivers/dri/radeon/radeon_dma.h @@ -48,4 +48,5 @@ void radeonReleaseDmaRegion(radeonContextPtr rmesa); void rcommon_flush_last_swtcl_prim(GLcontext *ctx); void *rcommonAllocDmaLowVerts(radeonContextPtr rmesa, int nverts, int vsize); +void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ); #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index f18aa1a4da..b5fde6d3de 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -313,9 +313,9 @@ void radeonEmitAOS( r100ContextPtr rmesa, { #if RADEON_OLD_PACKETS assert( nr == 1 ); - rmesa->ioctl.bo = rmesa->tcl.aos[0].bo; + rmesa->ioctl.bo = rmesa->radeon.tcl.aos[0].bo; rmesa->ioctl.vertex_offset = - (rmesa->tcl.aos[0].offset + offset * rmesa->tcl.aos[0].stride * 4); + (rmesa->radeon.tcl.aos[0].offset + offset * rmesa->radeon.tcl.aos[0].stride * 4); #else BATCH_LOCALS(&rmesa->radeon); uint32_t voffset; @@ -332,79 +332,79 @@ void radeonEmitAOS( r100ContextPtr rmesa, if (!rmesa->radeon.radeonScreen->kernel_mm) { for (i = 0; i + 1 < nr; i += 2) { - OUT_BATCH((rmesa->tcl.aos[i].components << 0) | - (rmesa->tcl.aos[i].stride << 8) | - (rmesa->tcl.aos[i + 1].components << 16) | - (rmesa->tcl.aos[i + 1].stride << 24)); + OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) | + (rmesa->radeon.tcl.aos[i].stride << 8) | + (rmesa->radeon.tcl.aos[i + 1].components << 16) | + (rmesa->radeon.tcl.aos[i + 1].stride << 24)); - voffset = rmesa->tcl.aos[i + 0].offset + - offset * 4 * rmesa->tcl.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; OUT_BATCH_RELOC(voffset, - rmesa->tcl.aos[i].bo, + rmesa->radeon.tcl.aos[i].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); - voffset = rmesa->tcl.aos[i + 1].offset + - offset * 4 * rmesa->tcl.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; OUT_BATCH_RELOC(voffset, - rmesa->tcl.aos[i+1].bo, + rmesa->radeon.tcl.aos[i+1].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); } if (nr & 1) { - OUT_BATCH((rmesa->tcl.aos[nr - 1].components << 0) | - (rmesa->tcl.aos[nr - 1].stride << 8)); - voffset = rmesa->tcl.aos[nr - 1].offset + - offset * 4 * rmesa->tcl.aos[nr - 1].stride; + OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) | + (rmesa->radeon.tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; OUT_BATCH_RELOC(voffset, - rmesa->tcl.aos[nr - 1].bo, + rmesa->radeon.tcl.aos[nr - 1].bo, voffset, RADEON_GEM_DOMAIN_GTT, 0, 0); } } else { for (i = 0; i + 1 < nr; i += 2) { - OUT_BATCH((rmesa->tcl.aos[i].components << 0) | - (rmesa->tcl.aos[i].stride << 8) | - (rmesa->tcl.aos[i + 1].components << 16) | - (rmesa->tcl.aos[i + 1].stride << 24)); + OUT_BATCH((rmesa->radeon.tcl.aos[i].components << 0) | + (rmesa->radeon.tcl.aos[i].stride << 8) | + (rmesa->radeon.tcl.aos[i + 1].components << 16) | + (rmesa->radeon.tcl.aos[i + 1].stride << 24)); - voffset = rmesa->tcl.aos[i + 0].offset + - offset * 4 * rmesa->tcl.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; OUT_BATCH(voffset); - voffset = rmesa->tcl.aos[i + 1].offset + - offset * 4 * rmesa->tcl.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; OUT_BATCH(voffset); } if (nr & 1) { - OUT_BATCH((rmesa->tcl.aos[nr - 1].components << 0) | - (rmesa->tcl.aos[nr - 1].stride << 8)); - voffset = rmesa->tcl.aos[nr - 1].offset + - offset * 4 * rmesa->tcl.aos[nr - 1].stride; + OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) | + (rmesa->radeon.tcl.aos[nr - 1].stride << 8)); + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; OUT_BATCH(voffset); } for (i = 0; i + 1 < nr; i += 2) { - voffset = rmesa->tcl.aos[i + 0].offset + - offset * 4 * rmesa->tcl.aos[i + 0].stride; + voffset = rmesa->radeon.tcl.aos[i + 0].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->tcl.aos[i+0].bo, + rmesa->radeon.tcl.aos[i+0].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); - voffset = rmesa->tcl.aos[i + 1].offset + - offset * 4 * rmesa->tcl.aos[i + 1].stride; + voffset = rmesa->radeon.tcl.aos[i + 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->tcl.aos[i+1].bo, + rmesa->radeon.tcl.aos[i+1].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); } if (nr & 1) { - voffset = rmesa->tcl.aos[nr - 1].offset + - offset * 4 * rmesa->tcl.aos[nr - 1].stride; + voffset = rmesa->radeon.tcl.aos[nr - 1].offset + + offset * 4 * rmesa->radeon.tcl.aos[nr - 1].stride; radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs, - rmesa->tcl.aos[nr-1].bo, + rmesa->radeon.tcl.aos[nr-1].bo, RADEON_GEM_DOMAIN_GTT, 0, 0); } diff --git a/src/mesa/drivers/dri/radeon/radeon_maos.h b/src/mesa/drivers/dri/radeon/radeon_maos.h index b8935e84a0..b88eb198d5 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos.h @@ -38,6 +38,5 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_context.h" extern void radeonEmitArrays( GLcontext *ctx, GLuint inputs ); -extern void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ); #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 7f5da16b03..7c6ea0530e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -324,16 +324,3 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) rmesa->tcl.vertex_format = vfmt; } - -void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) -{ - r100ContextPtr rmesa = R100_CONTEXT( ctx ); - int i; - - for (i = 0; i < rmesa->tcl.nr_aos_components; i++) { - if (rmesa->tcl.aos[i].bo) { - radeon_bo_unref(rmesa->tcl.aos[i].bo); - rmesa->tcl.aos[i].bo = NULL; - } - } -} diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index d468a97200..78ec119302 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -374,15 +374,15 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) break; if (rmesa->tcl.vertex_format == setup_tab[i].vertex_format && - rmesa->tcl.aos[0].bo) + rmesa->radeon.tcl.aos[0].bo) return; - if (rmesa->tcl.aos[0].bo) + if (rmesa->radeon.tcl.aos[0].bo) radeonReleaseArrays( ctx, ~0 ); radeonAllocDmaRegion( &rmesa->radeon, - &rmesa->tcl.aos[0].bo, - &rmesa->tcl.aos[0].offset, + &rmesa->radeon.tcl.aos[0].bo, + &rmesa->radeon.tcl.aos[0].offset, VB->Count * setup_tab[i].vertex_size * 4, 4); @@ -422,25 +422,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) setup_tab[i].emit( ctx, 0, VB->Count, - rmesa->tcl.aos[0].bo->ptr + rmesa->tcl.aos[0].offset); + rmesa->radeon.tcl.aos[0].bo->ptr + rmesa->radeon.tcl.aos[0].offset); - // rmesa->tcl.aos[0].size = setup_tab[i].vertex_size; - rmesa->tcl.aos[0].stride = setup_tab[i].vertex_size; + // rmesa->radeon.tcl.aos[0].size = setup_tab[i].vertex_size; + rmesa->radeon.tcl.aos[0].stride = setup_tab[i].vertex_size; rmesa->tcl.vertex_format = setup_tab[i].vertex_format; - rmesa->tcl.nr_aos_components = 1; + rmesa->radeon.tcl.aos_count = 1; } - -void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) -{ - r100ContextPtr rmesa = R100_CONTEXT( ctx ); - int i; - - for (i = 0; i < rmesa->tcl.nr_aos_components; i++) { - if (rmesa->tcl.aos[i].bo) { - radeon_bo_unref(rmesa->tcl.aos[i].bo); - rmesa->tcl.aos[i].bo = NULL; - } - } -} diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index ea400a672d..ecfdce9d01 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -1332,7 +1332,6 @@ radeonDestroyBuffer(__DRIdrawablePrivate *driDrawPriv) _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL); } -#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) /** * Choose the appropriate CreateContext function based on the chipset. * Eventually, all drivers will go through this process. @@ -1343,25 +1342,21 @@ static GLboolean radeonCreateContext(const __GLcontextModes * glVisual, { __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); - +#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) if (IS_R300_CLASS(screen)) return r300CreateContext(glVisual, driContextPriv, sharedContextPriv); - return GL_FALSE; -} - -/** - * Choose the appropriate DestroyContext function based on the chipset. - */ -static void radeonDestroyContext(__DRIcontextPrivate * driContextPriv) -{ - radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; - - if (IS_R300_CLASS(radeon->radeonScreen)) - return r300DestroyContext(driContextPriv); -} +#endif +#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R200) + if (IS_R200_CLASS(screen)) + return r200CreateContext(glVisual, driContextPriv, sharedContextPriv); +#endif +#if !RADEON_COMMON + return r100CreateContext(glVisual, driContextPriv, sharedContextPriv); #endif + return GL_FALSE; +} /** @@ -1547,7 +1542,6 @@ getSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo ) return 0; } -#if !RADEON_COMMON || (RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)) const struct __DriverAPIRec driDriverAPI = { .InitScreen = radeonInitScreen, .DestroyScreen = radeonDestroyScreen, @@ -1567,24 +1561,4 @@ const struct __DriverAPIRec driDriverAPI = { /* DRI2 */ .InitScreen2 = radeonInitScreen2, }; -#else -const struct __DriverAPIRec driDriverAPI = { - .InitScreen = radeonInitScreen, - .DestroyScreen = radeonDestroyScreen, - .CreateContext = r200CreateContext, - .DestroyContext = r200DestroyContext, - .CreateBuffer = radeonCreateBuffer, - .DestroyBuffer = radeonDestroyBuffer, - .SwapBuffers = radeonSwapBuffers, - .MakeCurrent = radeonMakeCurrent, - .UnbindContext = radeonUnbindContext, - .GetSwapInfo = getSwapInfo, - .GetDrawableMSC = driDrawableGetMSC32, - .WaitForMSC = driWaitForMSC32, - .WaitForSBC = NULL, - .SwapBuffersMSC = NULL, - .CopySubBuffer = radeonCopySubBuffer, - .InitScreen2 = radeonInitScreen2, -}; -#endif diff --git a/src/mesa/drivers/dri/radeon/radeon_state.h b/src/mesa/drivers/dri/radeon/radeon_state.h index 17c2b11222..f05fa827d7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.h +++ b/src/mesa/drivers/dri/radeon/radeon_state.h @@ -52,9 +52,6 @@ extern void radeonUploadTexMatrix( r100ContextPtr rmesa, extern void radeonValidateState( GLcontext *ctx ); -extern void radeonPrintDirty( r100ContextPtr rmesa, - const char *msg ); - extern void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ); #define FALLBACK( rmesa, bit, mode ) do { \ diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index 3d0cd8d3f8..174a7e1862 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -161,22 +161,6 @@ static struct { /* ============================================================= * State initialization */ - -void radeonPrintDirty( r100ContextPtr rmesa, const char *msg ) -{ - struct radeon_state_atom *l; - - fprintf(stderr, msg); - fprintf(stderr, ": "); - - foreach(l, &rmesa->radeon.hw.atomlist) { - if (l->dirty || rmesa->radeon.hw.all_dirty) - fprintf(stderr, "%s, ", l->name); - } - - fprintf(stderr, "\n"); -} - static int cmdpkt( r100ContextPtr rmesa, int id ) { drm_radeon_cmd_header_t h; diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 2484006f1c..e31f045991 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -822,7 +822,3 @@ void radeonInitSwtcl( GLcontext *ctx ) rmesa->radeon.swtcl.hw_primitive = 0; } - -void radeonDestroySwtcl( GLcontext *ctx ) -{ -} diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.h b/src/mesa/drivers/dri/radeon/radeon_swtcl.h index 3ada9890b3..da89158eeb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.h +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.h @@ -40,7 +40,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_context.h" extern void radeonInitSwtcl( GLcontext *ctx ); -extern void radeonDestroySwtcl( GLcontext *ctx ); extern void radeonChooseRenderState( GLcontext *ctx ); extern void radeonChooseVertexState( GLcontext *ctx ); diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index 5887ab355d..df6708f05e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -150,10 +150,10 @@ static GLushort *radeonAllocElts( r100ContextPtr rmesa, GLuint nr ) rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->radeon.hw.max_state_size + ELTS_BUFSZ(nr) + - AOS_BUFSZ(rmesa->tcl.nr_aos_components), __FUNCTION__); + AOS_BUFSZ(rmesa->radeon.tcl.aos_count), __FUNCTION__); radeonEmitAOS( rmesa, - rmesa->tcl.nr_aos_components, 0 ); + rmesa->radeon.tcl.aos_count, 0 ); return radeonAllocEltsOpenEnded( rmesa, rmesa->tcl.vertex_format, rmesa->tcl.hw_primitive, nr ); @@ -177,11 +177,11 @@ static void radeonEmitPrim( GLcontext *ctx, radeonTclPrimitive( ctx, prim, hwprim ); rcommonEnsureCmdBufSpace( &rmesa->radeon, - AOS_BUFSZ(rmesa->tcl.nr_aos_components) + + AOS_BUFSZ(rmesa->radeon.tcl.aos_count) + rmesa->radeon.hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); radeonEmitAOS( rmesa, - rmesa->tcl.nr_aos_components, + rmesa->radeon.tcl.aos_count, start ); /* Why couldn't this packet have taken an offset param? -- cgit v1.2.3 From bbf2b5c4ffcb6755d34a5b698445aecf604e45fb Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Fri, 14 Aug 2009 22:10:24 +0300 Subject: radeon: Optimize memory handling for dma operations. We keep dma buffer objects in list untill they have been unused for many draw operations. Current limit of having 100 flushes is just guess for good performance/memory trade off. Moving WARN_ONCE macro to common context because it is used in multiple drivers. Signed-off-by: Pauli Nieminen --- src/mesa/drivers/dri/r200/r200_state.c | 12 +- src/mesa/drivers/dri/r200/r200_swtcl.c | 3 +- src/mesa/drivers/dri/r300/r300_context.h | 16 --- src/mesa/drivers/dri/r300/r300_draw.c | 3 +- src/mesa/drivers/dri/r300/r300_swtcl.c | 3 +- src/mesa/drivers/dri/r300/r300_texstate.c | 3 +- src/mesa/drivers/dri/r600/r600_context.h | 16 --- src/mesa/drivers/dri/r600/r600_texstate.c | 3 +- src/mesa/drivers/dri/r600/r700_ioctl.c | 3 +- src/mesa/drivers/dri/radeon/radeon_common.c | 4 +- .../drivers/dri/radeon/radeon_common_context.c | 4 +- .../drivers/dri/radeon/radeon_common_context.h | 34 ++++- src/mesa/drivers/dri/radeon/radeon_dma.c | 142 +++++++++++++++------ src/mesa/drivers/dri/radeon/radeon_dma.h | 4 +- src/mesa/drivers/dri/radeon/radeon_state.c | 3 +- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 3 +- 16 files changed, 163 insertions(+), 93 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c index 5a6fd20d8c..ffc1a95745 100644 --- a/src/mesa/drivers/dri/r200/r200_state.c +++ b/src/mesa/drivers/dri/r200/r200_state.c @@ -2289,8 +2289,11 @@ static GLboolean r200ValidateBuffers(GLcontext *ctx) { r200ContextPtr rmesa = R200_CONTEXT(ctx); struct radeon_renderbuffer *rrb; + struct radeon_dma_bo *dma_bo; int i, ret; + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); radeon_cs_space_reset_bos(rmesa->radeon.cmdbuf.cs); rrb = radeon_get_colorbuffer(&rmesa->radeon); @@ -2323,9 +2326,12 @@ static GLboolean r200ValidateBuffers(GLcontext *ctx) RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0); } - ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0); - if (ret) - return GL_FALSE; + dma_bo = first_elem(&rmesa->radeon.dma.reserved); + { + ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, dma_bo->bo, RADEON_GEM_DOMAIN_GTT, 0); + if (ret) + return GL_FALSE; + } return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 83e70b586d..1b23891140 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -39,6 +39,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/image.h" #include "main/imports.h" #include "main/macros.h" +#include "main/simple_list.h" #include "swrast/s_context.h" #include "swrast/s_fog.h" @@ -275,7 +276,7 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); r200EmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - rmesa->radeon.dma.current, + first_elem(&rmesa->radeon.dma.reserved)->bo, current_offset); diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 65e0d46611..339b304558 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -51,22 +51,6 @@ typedef struct r300_context r300ContextRec; typedef struct r300_context *r300ContextPtr; -/* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html . - I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble - with other compilers ... GLUE! -*/ -#define WARN_ONCE(a, ...) { \ - static int warn##__LINE__=1; \ - if(warn##__LINE__){ \ - fprintf(stderr, "*********************************WARN_ONCE*********************************\n"); \ - fprintf(stderr, "File %s function %s line %d\n", \ - __FILE__, __FUNCTION__, __LINE__); \ - fprintf(stderr, a, ## __VA_ARGS__);\ - fprintf(stderr, "***************************************************************************\n"); \ - warn##__LINE__=0;\ - } \ - } - #include "r300_vertprog.h" diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c index d34f33b328..dbf5384d55 100644 --- a/src/mesa/drivers/dri/r300/r300_draw.c +++ b/src/mesa/drivers/dri/r300/r300_draw.c @@ -31,6 +31,7 @@ #include "main/state.h" #include "main/api_validate.h" #include "main/enums.h" +#include "main/simple_list.h" #include "r300_reg.h" #include "r300_context.h" @@ -510,7 +511,7 @@ static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *ar } r300->radeon.tcl.aos_count = vbuf->num_attribs; - ret = radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs, r300->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0); + ret = radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs, first_elem(&r300->radeon.dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0); if (ret) r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, GL_TRUE); } diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index a634cb5192..9d6f756879 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -39,6 +39,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_emit.h" #include "r300_tex.h" #include "r300_render.h" +#include "main/simple_list.h" #define EMIT_ATTR( ATTR, STYLE ) \ do { \ @@ -617,7 +618,7 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r300_emit_scissor(ctx); r300EmitVertexAOS(rmesa, rmesa->radeon.swtcl.vertex_size, - rmesa->radeon.dma.current, + first_elem(&rmesa->radeon.dma.reserved)->bo, current_offset); r300EmitVbufPrim(rmesa, diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c index 6f489ace7b..f030451b28 100644 --- a/src/mesa/drivers/dri/r300/r300_texstate.c +++ b/src/mesa/drivers/dri/r300/r300_texstate.c @@ -43,6 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/teximage.h" #include "main/texobj.h" #include "main/enums.h" +#include "main/simple_list.h" #include "r300_context.h" #include "r300_state.h" @@ -323,7 +324,7 @@ GLboolean r300ValidateBuffers(GLcontext * ctx) RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0); } - ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0); + ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, first_elem(&rmesa->radeon.dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0); if (ret) return GL_FALSE; return GL_TRUE; diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index 30ddce682c..4373254dd6 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -61,22 +61,6 @@ GLboolean r700SendSQConfig(context_t *context); #include "main/mm.h" -/* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html . - I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble - with other compilers ... GLUE! -*/ -#define WARN_ONCE(a, ...) { \ - static int warn##__LINE__=1; \ - if(warn##__LINE__){ \ - fprintf(stderr, "*********************************WARN_ONCE*********************************\n"); \ - fprintf(stderr, "File %s function %s line %d\n", \ - __FILE__, __FUNCTION__, __LINE__); \ - fprintf(stderr, a, ## __VA_ARGS__);\ - fprintf(stderr, "***************************************************************************\n"); \ - warn##__LINE__=0;\ - } \ - } - /************ DMA BUFFERS **************/ /* The blit width for texture uploads diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c index ee9b64ee43..1057d7d8bb 100644 --- a/src/mesa/drivers/dri/r600/r600_texstate.c +++ b/src/mesa/drivers/dri/r600/r600_texstate.c @@ -43,6 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/teximage.h" #include "main/texobj.h" #include "main/enums.h" +#include "main/simple_list.h" #include "r600_context.h" #include "r700_state.h" @@ -685,7 +686,7 @@ GLboolean r600ValidateBuffers(GLcontext * ctx) RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0); } - ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0); + ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, first_elem(&rmesa->radeon.dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0); if (ret) return GL_FALSE; return GL_TRUE; diff --git a/src/mesa/drivers/dri/r600/r700_ioctl.c b/src/mesa/drivers/dri/r600/r700_ioctl.c index 23cc128d6d..e0e506d665 100644 --- a/src/mesa/drivers/dri/r600/r700_ioctl.c +++ b/src/mesa/drivers/dri/r600/r700_ioctl.c @@ -31,6 +31,7 @@ #include "main/imports.h" #include "main/macros.h" #include "main/context.h" +#include "main/simple_list.h" #include "swrast/swrast.h" #include "radeon_common.h" @@ -52,7 +53,7 @@ static void r700Flush(GLcontext *ctx) we have no DMA buffer allocated. then no point flushing anything at all. */ - if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && !radeon->dma.current) + if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && is_empty_list(&radeon->dma.reserved)) return; if (radeon->dma.flush) diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index 0e9a1ae5d8..b5b4fed8fa 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -1043,7 +1043,7 @@ void radeonFlush(GLcontext *ctx) we have no DMA buffer allocated. then no point flushing anything at all. */ - if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && !radeon->dma.current) + if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && is_empty_list(&radeon->dma.reserved)) return; if (radeon->dma.flush) @@ -1152,7 +1152,7 @@ int rcommonFlushCmdBuf(radeonContextPtr rmesa, const char *caller) { int ret; - radeonReleaseDmaRegion(rmesa); + radeonReleaseDmaRegions(rmesa); LOCK_HARDWARE(rmesa); ret = rcommonFlushCmdBufLocked(rmesa, caller); diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index 35622099bf..13711963e2 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -264,6 +264,7 @@ GLboolean radeonInitContext(radeonContextPtr radeon, } make_empty_list(&radeon->query.not_flushed_head); + radeon_init_dma(radeon); return GL_TRUE; } @@ -309,10 +310,11 @@ void radeonDestroyContext(__DRIcontextPrivate *driContextPriv ) assert(radeon); if (radeon) { - if (radeon->dma.current) { + if (!is_empty_list(&radeon->dma.reserved)) { rcommonFlushCmdBuf( radeon, __FUNCTION__ ); } + radeonFreeDmaRegions(radeon); radeonReleaseArrays(radeon->glCtx, ~0); meta_destroy_metaops(&radeon->meta); if (radeon->vtbl.free_context) diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index a9480cd2e4..39fab66ca7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -18,6 +18,22 @@ struct radeon_context; #include "radeon_bocs_wrapper.h" +/* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html . + I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble + with other compilers ... GLUE! +*/ +#define WARN_ONCE(a, ...) { \ + static int warn##__LINE__=1; \ + if(warn##__LINE__){ \ + fprintf(stderr, "*********************************WARN_ONCE*********************************\n"); \ + fprintf(stderr, "File %s function %s line %d\n", \ + __FILE__, __FUNCTION__, __LINE__); \ + fprintf(stderr, a, ## __VA_ARGS__);\ + fprintf(stderr, "***************************************************************************\n"); \ + warn##__LINE__=0;\ + } \ + } + /* This union is used to avoid warnings/miscompilation with float to uint32_t casts due to strict-aliasing */ typedef union { GLfloat f; uint32_t ui32; } float_ui32_type; @@ -281,12 +297,22 @@ struct radeon_aos { int count; /** Number of vertices */ }; +#define DMA_BO_FREE_TIME 100 + +struct radeon_dma_bo { + struct radeon_dma_bo *next, *prev; + struct radeon_bo *bo; + int expire_counter; +}; + struct radeon_dma { /* Active dma region. Allocations for vertices and retained * regions come from here. Also used for emitting random vertices, * these may be flushed by calling flush_current(); */ - struct radeon_bo *current; /** Buffer that DMA memory is allocated from */ + struct radeon_dma_bo free; + struct radeon_dma_bo wait; + struct radeon_dma_bo reserved; int current_used; /** Number of bytes allocated and forgotten about */ int current_vertexptr; /** End of active vertex region */ @@ -296,12 +322,6 @@ struct radeon_dma { * performed. */ void (*flush) (GLcontext *); - - /* Number of "in-flight" DMA buffers, i.e. the number of buffers - * for which a DISCARD command is currently queued in the command buffer -. - */ - GLuint nr_released_bufs; }; /* radeon_swtcl.c diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 5e755c51ed..93bcae2454 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -31,6 +31,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ #include "radeon_common.h" +#include "main/simple_list.h" #if defined(USE_X86_ASM) #define COPY_DWORDS( dst, src, nr ) \ @@ -161,9 +162,15 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, } } -void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) +void radeon_init_dma(radeonContextPtr rmesa) { + make_empty_list(&rmesa->dma.free); + make_empty_list(&rmesa->dma.wait); + make_empty_list(&rmesa->dma.reserved); +} +void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) +{ size = MAX2(size, MAX_DMA_BUF_SZ); if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) @@ -173,43 +180,41 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) rmesa->dma.flush(rmesa->glCtx); } - if (rmesa->dma.nr_released_bufs > 4) { - rcommonFlushCmdBuf(rmesa, __FUNCTION__); - rmesa->dma.nr_released_bufs = 0; - } - - if (rmesa->dma.current) { - radeon_bo_unmap(rmesa->dma.current); - radeon_bo_unref(rmesa->dma.current); - rmesa->dma.current = 0; - } + if (is_empty_list(&rmesa->dma.free)) { + struct radeon_dma_bo *dma_bo = CALLOC(sizeof(struct radeon_dma_bo)); + assert(dma_bo); again_alloc: - rmesa->dma.current = radeon_bo_open(rmesa->radeonScreen->bom, + dma_bo->bo = radeon_bo_open(rmesa->radeonScreen->bom, 0, size, 4, RADEON_GEM_DOMAIN_GTT, 0); - if (!rmesa->dma.current) { - rcommonFlushCmdBuf(rmesa, __FUNCTION__); - rmesa->dma.nr_released_bufs = 0; - goto again_alloc; + if (!dma_bo->bo) { + rcommonFlushCmdBuf(rmesa, __FUNCTION__); + goto again_alloc; + } + insert_at_head(&rmesa->dma.reserved, dma_bo); + } else { + struct radeon_dma_bo *dma_bo = last_elem(&rmesa->dma.free); + assert(dma_bo->bo->cref == 1); + remove_from_list(dma_bo); + insert_at_head(&rmesa->dma.reserved, dma_bo); } rmesa->dma.current_used = 0; rmesa->dma.current_vertexptr = 0; if (radeon_cs_space_check_with_bo(rmesa->cmdbuf.cs, - rmesa->dma.current, + first_elem(&rmesa->dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0)) fprintf(stderr,"failure to revalidate BOs - badness\n"); - if (!rmesa->dma.current) { + if (is_empty_list(&rmesa->dma.reserved)) { /* Cmd buff have been flushed in radeon_revalidate_bos */ - rmesa->dma.nr_released_bufs = 0; goto again_alloc; } - radeon_bo_map(rmesa->dma.current, 1); + radeon_bo_map(first_elem(&rmesa->dma.reserved)->bo, 1); } /* Allocates a region from rmesa->dma.current. If there isn't enough @@ -230,30 +235,87 @@ void radeonAllocDmaRegion(radeonContextPtr rmesa, alignment--; rmesa->dma.current_used = (rmesa->dma.current_used + alignment) & ~alignment; - if (!rmesa->dma.current || rmesa->dma.current_used + bytes > rmesa->dma.current->size) + if (is_empty_list(&rmesa->dma.reserved) + || rmesa->dma.current_used + bytes > first_elem(&rmesa->dma.reserved)->bo->size) radeonRefillCurrentDmaRegion(rmesa, (bytes + 15) & ~15); *poffset = rmesa->dma.current_used; - *pbo = rmesa->dma.current; + *pbo = first_elem(&rmesa->dma.reserved)->bo; radeon_bo_ref(*pbo); /* Always align to at least 16 bytes */ rmesa->dma.current_used = (rmesa->dma.current_used + bytes + 15) & ~15; rmesa->dma.current_vertexptr = rmesa->dma.current_used; - assert(rmesa->dma.current_used <= rmesa->dma.current->size); + assert(rmesa->dma.current_used <= first_elem(&rmesa->dma.reserved)->bo->size); +} + +void radeonFreeDmaRegions(radeonContextPtr rmesa) +{ + struct radeon_dma_bo *dma_bo; + struct radeon_dma_bo *temp; + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + + foreach_s(dma_bo, temp, &rmesa->dma.free) { + remove_from_list(dma_bo); + radeon_bo_unref(dma_bo->bo); + FREE(dma_bo); + } + + foreach_s(dma_bo, temp, &rmesa->dma.reserved) { + remove_from_list(dma_bo); + radeon_bo_unmap(dma_bo->bo); + radeon_bo_unref(dma_bo->bo); + FREE(dma_bo); + } } -void radeonReleaseDmaRegion(radeonContextPtr rmesa) +void radeonReleaseDmaRegions(radeonContextPtr rmesa) { + struct radeon_dma_bo *dma_bo; + struct radeon_dma_bo *temp; + const int expire_at = ++rmesa->dma.free.expire_counter + DMA_BO_FREE_TIME; + const int time = rmesa->dma.free.expire_counter; if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s %p\n", __FUNCTION__, rmesa->dma.current); - if (rmesa->dma.current) { - rmesa->dma.nr_released_bufs++; - radeon_bo_unmap(rmesa->dma.current); - radeon_bo_unref(rmesa->dma.current); + fprintf(stderr, "%s\n", __FUNCTION__); + + /* move waiting bos to free list. + wait list provides gpu time to handle data before reuse */ + foreach_s(dma_bo, temp, &rmesa->dma.wait) { + if (dma_bo->expire_counter == time) { + WARN_ONCE("Leaking dma buffer object!\n"); + radeon_bo_unref(dma_bo->bo); + remove_from_list(dma_bo); + FREE(dma_bo); + continue; + } + if (dma_bo->bo->cref > 1) + continue; + remove_from_list(dma_bo); + dma_bo->expire_counter = expire_at; + insert_at_tail(&rmesa->dma.free, dma_bo); + } + + /* unmap the last dma region */ + if (!is_empty_list(&rmesa->dma.reserved)) + radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); + /* move reserved to wait list */ + foreach_s(dma_bo, temp, &rmesa->dma.reserved) { + remove_from_list(dma_bo); + dma_bo->expire_counter = expire_at; + insert_at_tail(&rmesa->dma.wait, dma_bo); + } + + /* free bos that have been unused for some time */ + foreach_s(dma_bo, temp, &rmesa->dma.free) { + if (dma_bo->expire_counter != time) + break; + remove_from_list(dma_bo); + radeon_bo_unref(dma_bo->bo); + FREE(dma_bo); } - rmesa->dma.current = NULL; + } @@ -266,10 +328,10 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s %p\n", __FUNCTION__, dma->current); + fprintf(stderr, "%s\n", __FUNCTION__); dma->flush = NULL; - if (dma->current) { + if (!is_empty_list(&dma->reserved)) { GLuint current_offset = dma->current_used; assert (dma->current_used + @@ -292,7 +354,10 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) GLuint bytes = vsize * nverts; void *head; restart: - if (!rmesa->dma.current || rmesa->dma.current_vertexptr + bytes > rmesa->dma.current->size) { + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + if (is_empty_list(&rmesa->dma.reserved) + || rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) { radeonRefillCurrentDmaRegion(rmesa, bytes); } @@ -302,7 +367,7 @@ restart: rmesa->hw.max_state_size + (20*sizeof(int)), __FUNCTION__); /* if cmdbuf flushed DMA restart */ - if (!rmesa->dma.current) + if (is_empty_list(&rmesa->dma.reserved)) goto restart; rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; rmesa->dma.flush = rcommon_flush_last_swtcl_prim; @@ -314,7 +379,7 @@ restart: rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == rmesa->dma.current_vertexptr ); - head = (rmesa->dma.current->ptr + rmesa->dma.current_vertexptr); + head = (first_elem(&rmesa->dma.reserved)->bo->ptr + rmesa->dma.current_vertexptr); rmesa->dma.current_vertexptr += bytes; rmesa->swtcl.numverts += nverts; return head; @@ -324,18 +389,17 @@ void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) { radeonContextPtr radeon = RADEON_CONTEXT( ctx ); int i; + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); if (radeon->dma.flush) { radeon->dma.flush(radeon->glCtx); } - if (radeon->tcl.elt_dma_bo) { - radeon_bo_unref(radeon->tcl.elt_dma_bo); - radeon->tcl.elt_dma_bo = NULL; - } for (i = 0; i < radeon->tcl.aos_count; i++) { if (radeon->tcl.aos[i].bo) { radeon_bo_unref(radeon->tcl.aos[i].bo); radeon->tcl.aos[i].bo = NULL; + } } } diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.h b/src/mesa/drivers/dri/radeon/radeon_dma.h index 55509ed00c..21f956e321 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.h +++ b/src/mesa/drivers/dri/radeon/radeon_dma.h @@ -42,13 +42,15 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, const GLvoid * data, int size, int stride, int count); void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size); +void radeon_init_dma(radeonContextPtr rmesa); void radeonAllocDmaRegion(radeonContextPtr rmesa, struct radeon_bo **pbo, int *poffset, int bytes, int alignment); -void radeonReleaseDmaRegion(radeonContextPtr rmesa); +void radeonReleaseDmaRegions(radeonContextPtr rmesa); void rcommon_flush_last_swtcl_prim(GLcontext *ctx); void *rcommonAllocDmaLowVerts(radeonContextPtr rmesa, int nverts, int vsize); +void radeonFreeDmaRegions(radeonContextPtr rmesa); void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ); #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c index 0d1728b747..56f82bdb0b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.c +++ b/src/mesa/drivers/dri/radeon/radeon_state.c @@ -40,6 +40,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/state.h" #include "main/context.h" #include "main/framebuffer.h" +#include "main/simple_list.h" #include "vbo/vbo.h" #include "tnl/tnl.h" @@ -2099,7 +2100,7 @@ static GLboolean r100ValidateBuffers(GLcontext *ctx) RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0); } - ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, rmesa->radeon.dma.current, RADEON_GEM_DOMAIN_GTT, 0); + ret = radeon_cs_space_check_with_bo(rmesa->radeon.cmdbuf.cs, first_elem(&rmesa->radeon.dma.reserved)->bo, RADEON_GEM_DOMAIN_GTT, 0); if (ret) return GL_FALSE; return GL_TRUE; diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index e31f045991..58b3be9391 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -38,6 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/enums.h" #include "main/imports.h" #include "main/macros.h" +#include "main/simple_list.h" #include "swrast_setup/swrast_setup.h" #include "math/m_translate.h" @@ -291,7 +292,7 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); radeonEmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - rmesa->radeon.dma.current, + first_elem(&rmesa->radeon.dma.reserved)->bo, current_offset); -- cgit v1.2.3 From d1a0ece9077b3de49c293a04c220b995424cef28 Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Wed, 12 Aug 2009 20:28:54 +0300 Subject: radeon: Port flush preventation code from r200 to r100. - emit prediction - fixes to emit_sizes for data - clean up of excesive use of radeonEnsureCmdBufferSpace Signed-off-by: Pauli Nieminen --- src/mesa/drivers/dri/radeon/radeon_dma.c | 2 +- src/mesa/drivers/dri/radeon/radeon_ioctl.h | 12 +- src/mesa/drivers/dri/radeon/radeon_state_init.c | 266 +++++++++++++++--------- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 2 +- src/mesa/drivers/dri/radeon/radeon_tcl.c | 72 ++++++- 5 files changed, 246 insertions(+), 108 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 7e6b74add8..79b2b77d30 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -431,7 +431,7 @@ restart: if (!rmesa->dma.flush) { /* make sure we have enough space to use this in cmdbuf */ rcommonEnsureCmdBufSpace(rmesa, - rmesa->hw.max_state_size + (20*sizeof(int)), + radeonCountEmitSize( rmesa ) + (20*sizeof(int)), __FUNCTION__); /* if cmdbuf flushed DMA restart */ if (is_empty_list(&rmesa->dma.reserved)) diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h index 18805d4c57..deb53ae313 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h @@ -132,16 +132,18 @@ static INLINE int RADEON_DB_STATECHANGE(r100ContextPtr rmesa, * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts. */ #if RADEON_OLD_PACKETS -#define AOS_BUFSZ(nr) ((3 + ((nr / 2) * 3) + ((nr & 1) * 2)) * sizeof(int)) +#define AOS_BUFSZ(nr) ((3 + ((nr / 2) * 3) + ((nr & 1) * 2))+nr*2) #define VERT_AOS_BUFSZ (0) #define ELTS_BUFSZ(nr) (24 + nr * 2) -#define VBUF_BUFSZ (6 * sizeof(int)) +#define VBUF_BUFSZ (8) #else -#define AOS_BUFSZ(nr) ((3 + ((nr / 2) * 3) + ((nr & 1) * 2)) * sizeof(int)) -#define VERT_AOS_BUFSZ (5 * sizeof(int)) +#define AOS_BUFSZ(nr) ((3 + ((nr / 2) * 3) + ((nr & 1) * 2) + nr*2)) +#define VERT_AOS_BUFSZ (5) #define ELTS_BUFSZ(nr) (16 + nr * 2) -#define VBUF_BUFSZ (4 * sizeof(int)) +#define VBUF_BUFSZ (4) #endif +#define SCISSOR_BUFSZ (8) +#define INDEX_BUFSZ (7) static inline uint32_t cmdpacket3(int cmd_type) diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index 57aa7f1ca4..e70dbbe06f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -197,55 +197,84 @@ static int cmdscl( int offset, int stride, int count ) return h.i; } -#define CHECK( NM, FLAG ) \ +#define CHECK( NM, FLAG, ADD ) \ static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \ { \ - return FLAG ? atom->cmd_size : 0; \ + return FLAG ? atom->cmd_size + (ADD) : 0; \ } -#define TCL_CHECK( NM, FLAG ) \ +#define TCL_CHECK( NM, FLAG, ADD ) \ static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \ { \ r100ContextPtr rmesa = R100_CONTEXT(ctx); \ - return (!rmesa->radeon.TclFallback && (FLAG)) ? atom->cmd_size : 0; \ + return (!rmesa->radeon.TclFallback && (FLAG)) ? atom->cmd_size + (ADD) : 0; \ } -CHECK( always, GL_TRUE ) -CHECK( never, GL_FALSE ) -CHECK( tex0, ctx->Texture.Unit[0]._ReallyEnabled ) -CHECK( tex1, ctx->Texture.Unit[1]._ReallyEnabled ) +CHECK( always, GL_TRUE, 0 ) +CHECK( always_add2, GL_TRUE, 2 ) +CHECK( never, GL_FALSE, 0 ) +CHECK( tex0_mm, ctx->Texture.Unit[0]._ReallyEnabled, 3 ) +CHECK( tex1_mm, ctx->Texture.Unit[1]._ReallyEnabled, 3 ) /* need this for the cubic_map on disabled unit 2 bug, maybe r100 only? */ -CHECK( tex2, ctx->Texture._EnabledUnits ) -CHECK( cube0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_CUBE_BIT)) -CHECK( cube1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_CUBE_BIT)) -CHECK( cube2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_CUBE_BIT)) -CHECK( fog, ctx->Fog.Enabled ) -TCL_CHECK( tcl, GL_TRUE ) -TCL_CHECK( tcl_tex0, ctx->Texture.Unit[0]._ReallyEnabled ) -TCL_CHECK( tcl_tex1, ctx->Texture.Unit[1]._ReallyEnabled ) -TCL_CHECK( tcl_tex2, ctx->Texture.Unit[2]._ReallyEnabled ) -TCL_CHECK( tcl_lighting, ctx->Light.Enabled ) -TCL_CHECK( tcl_eyespace_or_lighting, ctx->_NeedEyeCoords || ctx->Light.Enabled ) -TCL_CHECK( tcl_lit0, ctx->Light.Enabled && ctx->Light.Light[0].Enabled ) -TCL_CHECK( tcl_lit1, ctx->Light.Enabled && ctx->Light.Light[1].Enabled ) -TCL_CHECK( tcl_lit2, ctx->Light.Enabled && ctx->Light.Light[2].Enabled ) -TCL_CHECK( tcl_lit3, ctx->Light.Enabled && ctx->Light.Light[3].Enabled ) -TCL_CHECK( tcl_lit4, ctx->Light.Enabled && ctx->Light.Light[4].Enabled ) -TCL_CHECK( tcl_lit5, ctx->Light.Enabled && ctx->Light.Light[5].Enabled ) -TCL_CHECK( tcl_lit6, ctx->Light.Enabled && ctx->Light.Light[6].Enabled ) -TCL_CHECK( tcl_lit7, ctx->Light.Enabled && ctx->Light.Light[7].Enabled ) -TCL_CHECK( tcl_ucp0, (ctx->Transform.ClipPlanesEnabled & 0x1) ) -TCL_CHECK( tcl_ucp1, (ctx->Transform.ClipPlanesEnabled & 0x2) ) -TCL_CHECK( tcl_ucp2, (ctx->Transform.ClipPlanesEnabled & 0x4) ) -TCL_CHECK( tcl_ucp3, (ctx->Transform.ClipPlanesEnabled & 0x8) ) -TCL_CHECK( tcl_ucp4, (ctx->Transform.ClipPlanesEnabled & 0x10) ) -TCL_CHECK( tcl_ucp5, (ctx->Transform.ClipPlanesEnabled & 0x20) ) -TCL_CHECK( tcl_eyespace_or_fog, ctx->_NeedEyeCoords || ctx->Fog.Enabled ) - -CHECK( txr0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_RECT_BIT)) -CHECK( txr1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_RECT_BIT)) -CHECK( txr2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_RECT_BIT)) +CHECK( tex2_mm, ctx->Texture._EnabledUnits, 3 ) +CHECK( tex0, ctx->Texture.Unit[0]._ReallyEnabled, 2 ) +CHECK( tex1, ctx->Texture.Unit[1]._ReallyEnabled, 2 ) +CHECK( tex2, ctx->Texture._EnabledUnits, 2 ) +CHECK( cube0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_CUBE_BIT), 3 + 3*5 - CUBE_STATE_SIZE ) +CHECK( cube1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_CUBE_BIT), 3 + 3*5 - CUBE_STATE_SIZE ) +CHECK( cube2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_CUBE_BIT), 3 + 3*5 - CUBE_STATE_SIZE ) +CHECK( cube0_mm, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_CUBE_BIT), 2 + 4*5 - CUBE_STATE_SIZE ) +CHECK( cube1_mm, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_CUBE_BIT), 2 + 4*5 - CUBE_STATE_SIZE ) +CHECK( cube2_mm, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_CUBE_BIT), 2 + 4*5 - CUBE_STATE_SIZE ) +CHECK( fog, ctx->Fog.Enabled, 0 ) +CHECK( fog_add4, ctx->Fog.Enabled, 4 ) +TCL_CHECK( tcl, GL_TRUE, 0 ) +TCL_CHECK( tcl_add4, GL_TRUE, 4 ) +TCL_CHECK( tcl_tex0, ctx->Texture.Unit[0]._ReallyEnabled, 0 ) +TCL_CHECK( tcl_tex1, ctx->Texture.Unit[1]._ReallyEnabled, 0 ) +TCL_CHECK( tcl_tex2, ctx->Texture.Unit[2]._ReallyEnabled, 0 ) +TCL_CHECK( tcl_tex0_add4, ctx->Texture.Unit[0]._ReallyEnabled, 4 ) +TCL_CHECK( tcl_tex1_add4, ctx->Texture.Unit[1]._ReallyEnabled, 4 ) +TCL_CHECK( tcl_tex2_add4, ctx->Texture.Unit[2]._ReallyEnabled, 4 ) +TCL_CHECK( tcl_lighting, ctx->Light.Enabled, 0 ) +TCL_CHECK( tcl_lighting_add4, ctx->Light.Enabled, 4 ) +TCL_CHECK( tcl_eyespace_or_lighting, ctx->_NeedEyeCoords || ctx->Light.Enabled, 0 ) +TCL_CHECK( tcl_eyespace_or_lighting_add4, ctx->_NeedEyeCoords || ctx->Light.Enabled, 4 ) +TCL_CHECK( tcl_lit0, ctx->Light.Enabled && ctx->Light.Light[0].Enabled, 0 ) +TCL_CHECK( tcl_lit1, ctx->Light.Enabled && ctx->Light.Light[1].Enabled, 0 ) +TCL_CHECK( tcl_lit2, ctx->Light.Enabled && ctx->Light.Light[2].Enabled, 0 ) +TCL_CHECK( tcl_lit3, ctx->Light.Enabled && ctx->Light.Light[3].Enabled, 0 ) +TCL_CHECK( tcl_lit4, ctx->Light.Enabled && ctx->Light.Light[4].Enabled, 0 ) +TCL_CHECK( tcl_lit5, ctx->Light.Enabled && ctx->Light.Light[5].Enabled, 0 ) +TCL_CHECK( tcl_lit6, ctx->Light.Enabled && ctx->Light.Light[6].Enabled, 0 ) +TCL_CHECK( tcl_lit7, ctx->Light.Enabled && ctx->Light.Light[7].Enabled, 0 ) +TCL_CHECK( tcl_lit0_add6, ctx->Light.Enabled && ctx->Light.Light[0].Enabled, 6 ) +TCL_CHECK( tcl_lit1_add6, ctx->Light.Enabled && ctx->Light.Light[1].Enabled, 6 ) +TCL_CHECK( tcl_lit2_add6, ctx->Light.Enabled && ctx->Light.Light[2].Enabled, 6 ) +TCL_CHECK( tcl_lit3_add6, ctx->Light.Enabled && ctx->Light.Light[3].Enabled, 6 ) +TCL_CHECK( tcl_lit4_add6, ctx->Light.Enabled && ctx->Light.Light[4].Enabled, 6 ) +TCL_CHECK( tcl_lit5_add6, ctx->Light.Enabled && ctx->Light.Light[5].Enabled, 6 ) +TCL_CHECK( tcl_lit6_add6, ctx->Light.Enabled && ctx->Light.Light[6].Enabled, 6 ) +TCL_CHECK( tcl_lit7_add6, ctx->Light.Enabled && ctx->Light.Light[7].Enabled, 6 ) +TCL_CHECK( tcl_ucp0, (ctx->Transform.ClipPlanesEnabled & 0x1), 0 ) +TCL_CHECK( tcl_ucp1, (ctx->Transform.ClipPlanesEnabled & 0x2), 0 ) +TCL_CHECK( tcl_ucp2, (ctx->Transform.ClipPlanesEnabled & 0x4), 0 ) +TCL_CHECK( tcl_ucp3, (ctx->Transform.ClipPlanesEnabled & 0x8), 0 ) +TCL_CHECK( tcl_ucp4, (ctx->Transform.ClipPlanesEnabled & 0x10), 0 ) +TCL_CHECK( tcl_ucp5, (ctx->Transform.ClipPlanesEnabled & 0x20), 0 ) +TCL_CHECK( tcl_ucp0_add4, (ctx->Transform.ClipPlanesEnabled & 0x1), 4 ) +TCL_CHECK( tcl_ucp1_add4, (ctx->Transform.ClipPlanesEnabled & 0x2), 4 ) +TCL_CHECK( tcl_ucp2_add4, (ctx->Transform.ClipPlanesEnabled & 0x4), 4 ) +TCL_CHECK( tcl_ucp3_add4, (ctx->Transform.ClipPlanesEnabled & 0x8), 4 ) +TCL_CHECK( tcl_ucp4_add4, (ctx->Transform.ClipPlanesEnabled & 0x10), 4 ) +TCL_CHECK( tcl_ucp5_add4, (ctx->Transform.ClipPlanesEnabled & 0x20), 4 ) +TCL_CHECK( tcl_eyespace_or_fog, ctx->_NeedEyeCoords || ctx->Fog.Enabled, 0 ) +TCL_CHECK( tcl_eyespace_or_fog_add4, ctx->_NeedEyeCoords || ctx->Fog.Enabled, 4 ) + +CHECK( txr0, (ctx->Texture.Unit[0]._ReallyEnabled & TEXTURE_RECT_BIT), 0 ) +CHECK( txr1, (ctx->Texture.Unit[1]._ReallyEnabled & TEXTURE_RECT_BIT), 0 ) +CHECK( txr2, (ctx->Texture.Unit[2]._ReallyEnabled & TEXTURE_RECT_BIT), 0 ) #define OUT_VEC(hdr, data) do { \ drm_radeon_cmd_header_t h; \ @@ -271,9 +300,8 @@ static void scl_emit(GLcontext *ctx, struct radeon_state_atom *atom) { r100ContextPtr r100 = R100_CONTEXT(ctx); BATCH_LOCALS(&r100->radeon); - uint32_t dwords = atom->cmd_size; + uint32_t dwords = atom->check(ctx, atom); - dwords += 2; BEGIN_BATCH_NO_AUTOSTATE(dwords); OUT_SCL(atom->cmd[0], atom->cmd+1); END_BATCH(); @@ -284,9 +312,8 @@ static void vec_emit(GLcontext *ctx, struct radeon_state_atom *atom) { r100ContextPtr r100 = R100_CONTEXT(ctx); BATCH_LOCALS(&r100->radeon); - uint32_t dwords = atom->cmd_size; + uint32_t dwords = atom->check(ctx, atom); - dwords += 4; BEGIN_BATCH_NO_AUTOSTATE(dwords); OUT_VEC(atom->cmd[0], atom->cmd+1); END_BATCH(); @@ -297,9 +324,8 @@ static void lit_emit(GLcontext *ctx, struct radeon_state_atom *atom) { r100ContextPtr r100 = R100_CONTEXT(ctx); BATCH_LOCALS(&r100->radeon); - uint32_t dwords = atom->cmd_size; + uint32_t dwords = atom->check(ctx, atom); - dwords += 6; BEGIN_BATCH_NO_AUTOSTATE(dwords); OUT_VEC(atom->cmd[LIT_CMD_0], atom->cmd+1); OUT_SCL(atom->cmd[LIT_CMD_1], atom->cmd+LIT_CMD_1+1); @@ -313,10 +339,10 @@ static void ctx_emit(GLcontext *ctx, struct radeon_state_atom *atom) struct radeon_renderbuffer *rrb; uint32_t cbpitch; uint32_t zbpitch, depth_fmt; - uint32_t dwords = atom->cmd_size; + uint32_t dwords = atom->check(ctx, atom); /* output the first 7 bytes of context */ - BEGIN_BATCH_NO_AUTOSTATE(dwords + 4); + BEGIN_BATCH_NO_AUTOSTATE(dwords); OUT_BATCH_TABLE(atom->cmd, 5); rrb = radeon_get_depthbuffer(&r100->radeon); @@ -371,6 +397,28 @@ static void ctx_emit(GLcontext *ctx, struct radeon_state_atom *atom) END_BATCH(); } +static int check_always_ctx( GLcontext *ctx, struct radeon_state_atom *atom) +{ + r100ContextPtr r100 = R100_CONTEXT(ctx); + struct radeon_renderbuffer *rrb, *drb; + uint32_t dwords; + + rrb = radeon_get_colorbuffer(&r100->radeon); + if (!rrb || !rrb->bo) { + return 0; + } + + drb = radeon_get_depthbuffer(&r100->radeon); + + dwords = 10; + if (drb) + dwords += 6; + if (rrb) + dwords += 8; + + return dwords; +} + static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) { r100ContextPtr r100 = R100_CONTEXT(ctx); @@ -378,7 +426,7 @@ static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) struct radeon_renderbuffer *rrb, *drb; uint32_t cbpitch = 0; uint32_t zbpitch = 0; - uint32_t dwords = atom->cmd_size; + uint32_t dwords = atom->check(ctx, atom); uint32_t depth_fmt; rrb = radeon_get_colorbuffer(&r100->radeon); @@ -418,12 +466,6 @@ static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) } - /* output the first 7 bytes of context */ - dwords = 10; - if (drb) - dwords += 6; - if (rrb) - dwords += 8; BEGIN_BATCH_NO_AUTOSTATE(dwords); /* In the CS case we need to split this up */ @@ -474,7 +516,7 @@ static void cube_emit(GLcontext *ctx, struct radeon_state_atom *atom) { r100ContextPtr r100 = R100_CONTEXT(ctx); BATCH_LOCALS(&r100->radeon); - uint32_t dwords = 3; + uint32_t dwords = atom->check(ctx, atom); int i = atom->idx, j; radeonTexObj *t = r100->state.texture.unit[i].texobj; radeon_mipmap_level *lvl; @@ -488,7 +530,7 @@ static void cube_emit(GLcontext *ctx, struct radeon_state_atom *atom) if (!t->mt) return; - BEGIN_BATCH_NO_AUTOSTATE(dwords + (5 * 3)); + BEGIN_BATCH_NO_AUTOSTATE(dwords); OUT_BATCH_TABLE(atom->cmd, 3); lvl = &t->mt->levels[0]; for (j = 0; j < 5; j++) { @@ -502,7 +544,7 @@ static void cube_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) { r100ContextPtr r100 = R100_CONTEXT(ctx); BATCH_LOCALS(&r100->radeon); - uint32_t dwords = 2; + uint32_t dwords = atom->check(ctx, atom); int i = atom->idx, j; radeonTexObj *t = r100->state.texture.unit[i].texobj; radeon_mipmap_level *lvl; @@ -523,7 +565,7 @@ static void cube_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) default: case 0: base_reg = RADEON_PP_CUBIC_OFFSET_T0_0; break; }; - BEGIN_BATCH_NO_AUTOSTATE(dwords + (5 * 4)); + BEGIN_BATCH_NO_AUTOSTATE(dwords); OUT_BATCH_TABLE(atom->cmd, 2); lvl = &t->mt->levels[0]; for (j = 0; j < 5; j++) { @@ -666,9 +708,10 @@ void radeonInitState( r100ContextPtr rmesa ) /* Allocate state buffers: */ ALLOC_STATE( ctx, always, CTX_STATE_SIZE, "CTX/context", 0 ); - if (rmesa->radeon.radeonScreen->kernel_mm) + if (rmesa->radeon.radeonScreen->kernel_mm) { rmesa->hw.ctx.emit = ctx_emit_cs; - else + rmesa->hw.ctx.check = check_always_ctx; + } else rmesa->hw.ctx.emit = ctx_emit; ALLOC_STATE( lin, always, LIN_STATE_SIZE, "LIN/line", 0 ); ALLOC_STATE( msk, always, MSK_STATE_SIZE, "MSK/mask", 0 ); @@ -678,13 +721,63 @@ void radeonInitState( r100ContextPtr rmesa ) ALLOC_STATE( zbs, always, ZBS_STATE_SIZE, "ZBS/zbias", 0 ); ALLOC_STATE( tcl, always, TCL_STATE_SIZE, "TCL/tcl", 1 ); ALLOC_STATE( mtl, tcl_lighting, MTL_STATE_SIZE, "MTL/material", 1 ); - ALLOC_STATE( grd, always, GRD_STATE_SIZE, "GRD/guard-band", 1 ); - ALLOC_STATE( fog, fog, FOG_STATE_SIZE, "FOG/fog", 1 ); - ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 1 ); - ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 1 ); - ALLOC_STATE_IDX( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0, 0); - ALLOC_STATE_IDX( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0, 1); - ALLOC_STATE_IDX( tex[2], tex2, TEX_STATE_SIZE, "TEX/tex-2", 0, 2); + if (rmesa->radeon.radeonScreen->kernel_mm) { + ALLOC_STATE( grd, always_add2, GRD_STATE_SIZE, "GRD/guard-band", 1 ); + ALLOC_STATE( fog, fog_add4, FOG_STATE_SIZE, "FOG/fog", 1 ); + ALLOC_STATE( glt, tcl_lighting_add4, GLT_STATE_SIZE, "GLT/light-global", 1 ); + ALLOC_STATE( eye, tcl_lighting_add4, EYE_STATE_SIZE, "EYE/eye-vector", 1 ); + ALLOC_STATE_IDX( tex[0], tex0_mm, TEX_STATE_SIZE, "TEX/tex-0", 0, 0); + ALLOC_STATE_IDX( tex[1], tex1_mm, TEX_STATE_SIZE, "TEX/tex-1", 0, 1); + ALLOC_STATE_IDX( tex[2], tex2_mm, TEX_STATE_SIZE, "TEX/tex-2", 0, 2); + ALLOC_STATE( mat[0], tcl_add4, MAT_STATE_SIZE, "MAT/modelproject", 1 ); + ALLOC_STATE( mat[1], tcl_eyespace_or_fog_add4, MAT_STATE_SIZE, "MAT/modelview", 1 ); + ALLOC_STATE( mat[2], tcl_eyespace_or_lighting_add4, MAT_STATE_SIZE, "MAT/it-modelview", 1 ); + ALLOC_STATE( mat[3], tcl_tex0_add4, MAT_STATE_SIZE, "MAT/texmat0", 1 ); + ALLOC_STATE( mat[4], tcl_tex1_add4, MAT_STATE_SIZE, "MAT/texmat1", 1 ); + ALLOC_STATE( mat[5], tcl_tex2_add4, MAT_STATE_SIZE, "MAT/texmat2", 1 ); + ALLOC_STATE( lit[0], tcl_lit0_add6, LIT_STATE_SIZE, "LIT/light-0", 1 ); + ALLOC_STATE( lit[1], tcl_lit1_add6, LIT_STATE_SIZE, "LIT/light-1", 1 ); + ALLOC_STATE( lit[2], tcl_lit2_add6, LIT_STATE_SIZE, "LIT/light-2", 1 ); + ALLOC_STATE( lit[3], tcl_lit3_add6, LIT_STATE_SIZE, "LIT/light-3", 1 ); + ALLOC_STATE( lit[4], tcl_lit4_add6, LIT_STATE_SIZE, "LIT/light-4", 1 ); + ALLOC_STATE( lit[5], tcl_lit5_add6, LIT_STATE_SIZE, "LIT/light-5", 1 ); + ALLOC_STATE( lit[6], tcl_lit6_add6, LIT_STATE_SIZE, "LIT/light-6", 1 ); + ALLOC_STATE( lit[7], tcl_lit7_add6, LIT_STATE_SIZE, "LIT/light-7", 1 ); + ALLOC_STATE( ucp[0], tcl_ucp0_add4, UCP_STATE_SIZE, "UCP/userclip-0", 1 ); + ALLOC_STATE( ucp[1], tcl_ucp1_add4, UCP_STATE_SIZE, "UCP/userclip-1", 1 ); + ALLOC_STATE( ucp[2], tcl_ucp2_add4, UCP_STATE_SIZE, "UCP/userclip-2", 1 ); + ALLOC_STATE( ucp[3], tcl_ucp3_add4, UCP_STATE_SIZE, "UCP/userclip-3", 1 ); + ALLOC_STATE( ucp[4], tcl_ucp4_add4, UCP_STATE_SIZE, "UCP/userclip-4", 1 ); + ALLOC_STATE( ucp[5], tcl_ucp5_add4, UCP_STATE_SIZE, "UCP/userclip-5", 1 ); + } else { + ALLOC_STATE( grd, always, GRD_STATE_SIZE, "GRD/guard-band", 1 ); + ALLOC_STATE( fog, fog, FOG_STATE_SIZE, "FOG/fog", 1 ); + ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 1 ); + ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 1 ); + ALLOC_STATE_IDX( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0, 0); + ALLOC_STATE_IDX( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0, 1); + ALLOC_STATE_IDX( tex[2], tex2, TEX_STATE_SIZE, "TEX/tex-2", 0, 2); + ALLOC_STATE( mat[0], tcl, MAT_STATE_SIZE, "MAT/modelproject", 1 ); + ALLOC_STATE( mat[1], tcl_eyespace_or_fog, MAT_STATE_SIZE, "MAT/modelview", 1 ); + ALLOC_STATE( mat[2], tcl_eyespace_or_lighting, MAT_STATE_SIZE, "MAT/it-modelview", 1 ); + ALLOC_STATE( mat[3], tcl_tex0, MAT_STATE_SIZE, "MAT/texmat0", 1 ); + ALLOC_STATE( mat[4], tcl_tex1, MAT_STATE_SIZE, "MAT/texmat1", 1 ); + ALLOC_STATE( mat[5], tcl_tex2, MAT_STATE_SIZE, "MAT/texmat2", 1 ); + ALLOC_STATE( lit[0], tcl_lit0, LIT_STATE_SIZE, "LIT/light-0", 1 ); + ALLOC_STATE( lit[1], tcl_lit1, LIT_STATE_SIZE, "LIT/light-1", 1 ); + ALLOC_STATE( lit[2], tcl_lit2, LIT_STATE_SIZE, "LIT/light-2", 1 ); + ALLOC_STATE( lit[3], tcl_lit3, LIT_STATE_SIZE, "LIT/light-3", 1 ); + ALLOC_STATE( lit[4], tcl_lit4, LIT_STATE_SIZE, "LIT/light-4", 1 ); + ALLOC_STATE( lit[5], tcl_lit5, LIT_STATE_SIZE, "LIT/light-5", 1 ); + ALLOC_STATE( lit[6], tcl_lit6, LIT_STATE_SIZE, "LIT/light-6", 1 ); + ALLOC_STATE( lit[7], tcl_lit7, LIT_STATE_SIZE, "LIT/light-7", 1 ); + ALLOC_STATE( ucp[0], tcl_ucp0, UCP_STATE_SIZE, "UCP/userclip-0", 1 ); + ALLOC_STATE( ucp[1], tcl_ucp1, UCP_STATE_SIZE, "UCP/userclip-1", 1 ); + ALLOC_STATE( ucp[2], tcl_ucp2, UCP_STATE_SIZE, "UCP/userclip-2", 1 ); + ALLOC_STATE( ucp[3], tcl_ucp3, UCP_STATE_SIZE, "UCP/userclip-3", 1 ); + ALLOC_STATE( ucp[4], tcl_ucp4, UCP_STATE_SIZE, "UCP/userclip-4", 1 ); + ALLOC_STATE( ucp[5], tcl_ucp5, UCP_STATE_SIZE, "UCP/userclip-5", 1 ); + } for (i = 0; i < 3; i++) { if (rmesa->radeon.radeonScreen->kernel_mm) @@ -694,14 +787,19 @@ void radeonInitState( r100ContextPtr rmesa ) } if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR100) { - ALLOC_STATE_IDX( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 ); - ALLOC_STATE_IDX( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 ); - ALLOC_STATE_IDX( cube[2], cube2, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 ); - for (i = 0; i < 3; i++) - if (rmesa->radeon.radeonScreen->kernel_mm) - rmesa->hw.cube[i].emit = cube_emit_cs; - else - rmesa->hw.cube[i].emit = cube_emit; + if (rmesa->radeon.radeonScreen->kernel_mm) { + ALLOC_STATE_IDX( cube[0], cube0_mm, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 ); + ALLOC_STATE_IDX( cube[1], cube1_mm, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 ); + ALLOC_STATE_IDX( cube[2], cube2_mm, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 ); + for (i = 0; i < 3; i++) + rmesa->hw.cube[i].emit = cube_emit_cs; + } else { + ALLOC_STATE_IDX( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 ); + ALLOC_STATE_IDX( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 ); + ALLOC_STATE_IDX( cube[2], cube2, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 ); + for (i = 0; i < 3; i++) + rmesa->hw.cube[i].emit = cube_emit; + } } else { @@ -709,26 +807,6 @@ void radeonInitState( r100ContextPtr rmesa ) ALLOC_STATE_IDX( cube[1], never, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 ); ALLOC_STATE_IDX( cube[2], never, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 ); } - ALLOC_STATE( mat[0], tcl, MAT_STATE_SIZE, "MAT/modelproject", 1 ); - ALLOC_STATE( mat[1], tcl_eyespace_or_fog, MAT_STATE_SIZE, "MAT/modelview", 1 ); - ALLOC_STATE( mat[2], tcl_eyespace_or_lighting, MAT_STATE_SIZE, "MAT/it-modelview", 1 ); - ALLOC_STATE( mat[3], tcl_tex0, MAT_STATE_SIZE, "MAT/texmat0", 1 ); - ALLOC_STATE( mat[4], tcl_tex1, MAT_STATE_SIZE, "MAT/texmat1", 1 ); - ALLOC_STATE( mat[5], tcl_tex2, MAT_STATE_SIZE, "MAT/texmat2", 1 ); - ALLOC_STATE( ucp[0], tcl_ucp0, UCP_STATE_SIZE, "UCP/userclip-0", 1 ); - ALLOC_STATE( ucp[1], tcl_ucp1, UCP_STATE_SIZE, "UCP/userclip-1", 1 ); - ALLOC_STATE( ucp[2], tcl_ucp2, UCP_STATE_SIZE, "UCP/userclip-2", 1 ); - ALLOC_STATE( ucp[3], tcl_ucp3, UCP_STATE_SIZE, "UCP/userclip-3", 1 ); - ALLOC_STATE( ucp[4], tcl_ucp4, UCP_STATE_SIZE, "UCP/userclip-4", 1 ); - ALLOC_STATE( ucp[5], tcl_ucp5, UCP_STATE_SIZE, "UCP/userclip-5", 1 ); - ALLOC_STATE( lit[0], tcl_lit0, LIT_STATE_SIZE, "LIT/light-0", 1 ); - ALLOC_STATE( lit[1], tcl_lit1, LIT_STATE_SIZE, "LIT/light-1", 1 ); - ALLOC_STATE( lit[2], tcl_lit2, LIT_STATE_SIZE, "LIT/light-2", 1 ); - ALLOC_STATE( lit[3], tcl_lit3, LIT_STATE_SIZE, "LIT/light-3", 1 ); - ALLOC_STATE( lit[4], tcl_lit4, LIT_STATE_SIZE, "LIT/light-4", 1 ); - ALLOC_STATE( lit[5], tcl_lit5, LIT_STATE_SIZE, "LIT/light-5", 1 ); - ALLOC_STATE( lit[6], tcl_lit6, LIT_STATE_SIZE, "LIT/light-6", 1 ); - ALLOC_STATE( lit[7], tcl_lit7, LIT_STATE_SIZE, "LIT/light-7", 1 ); ALLOC_STATE_IDX( txr[0], txr0, TXR_STATE_SIZE, "TXR/txr-0", 0, 0 ); ALLOC_STATE_IDX( txr[1], txr1, TXR_STATE_SIZE, "TXR/txr-1", 0, 1 ); ALLOC_STATE_IDX( txr[2], txr2, TXR_STATE_SIZE, "TXR/txr-2", 0, 2 ); diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 58b3be9391..a1b398e560 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -285,7 +285,7 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r100ContextPtr rmesa = R100_CONTEXT(ctx); rcommonEnsureCmdBufSpace(&rmesa->radeon, - rmesa->radeon.hw.max_state_size + (12*sizeof(int)), + radeonCountEmitSize( &rmesa->radeon ) + (12*sizeof(int)), __FUNCTION__); diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index df6708f05e..c5a73a2818 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -149,9 +149,6 @@ static GLushort *radeonAllocElts( r100ContextPtr rmesa, GLuint nr ) if (rmesa->radeon.dma.flush) rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); - rcommonEnsureCmdBufSpace(&rmesa->radeon, rmesa->radeon.hw.max_state_size + ELTS_BUFSZ(nr) + - AOS_BUFSZ(rmesa->radeon.tcl.aos_count), __FUNCTION__); - radeonEmitAOS( rmesa, rmesa->radeon.tcl.aos_count, 0 ); @@ -176,10 +173,6 @@ static void radeonEmitPrim( GLcontext *ctx, r100ContextPtr rmesa = R100_CONTEXT( ctx ); radeonTclPrimitive( ctx, prim, hwprim ); - rcommonEnsureCmdBufSpace( &rmesa->radeon, - AOS_BUFSZ(rmesa->radeon.tcl.aos_count) + - rmesa->radeon.hw.max_state_size + VBUF_BUFSZ, __FUNCTION__ ); - radeonEmitAOS( rmesa, rmesa->radeon.tcl.aos_count, start ); @@ -196,6 +189,8 @@ static void radeonEmitPrim( GLcontext *ctx, radeonEmitPrim( ctx, prim, hwprim, start, count ); \ (void) rmesa; } while (0) +#define MAX_CONVERSION_SIZE 40 + /* Try & join small primitives */ #if 0 @@ -360,6 +355,68 @@ radeonComputeFogBlendFactor( GLcontext *ctx, GLfloat fogcoord ) } } +/** + * Predict total emit size for next rendering operation so there is no flush in middle of rendering + * Prediction has to aim towards the best possible value that is worse than worst case scenario + */ +static void radeonEnsureEmitSize( GLcontext * ctx , GLuint inputs ) +{ + r100ContextPtr rmesa = R100_CONTEXT(ctx); + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *VB = &tnl->vb; + GLuint space_required; + GLuint nr_aos = 1; /* radeonEmitArrays does always emit one */ + int i; + /* list of flags that are allocating aos object */ + const GLuint flags_to_check[] = { + VERT_BIT_NORMAL, + VERT_BIT_COLOR0, + VERT_BIT_COLOR1, + VERT_BIT_FOG + }; + /* predict number of aos to emit */ + for (i=0; i < sizeof(flags_to_check)/sizeof(flags_to_check[0]); ++i) + { + if (inputs & flags_to_check[i]) + ++nr_aos; + } + for (i = 0; i < ctx->Const.MaxTextureUnits; ++i) + { + if (inputs & VERT_BIT_TEX(i)) + ++nr_aos; + } + + { + /* count the prediction for state size */ + space_required = radeonCountEmitSize( &rmesa->radeon ); + /* tcl may be changed in radeonEmitArrays so account for it if not dirty */ + if (!rmesa->hw.tcl.dirty) + space_required += rmesa->hw.tcl.check( rmesa->radeon.glCtx, &rmesa->hw.tcl ); + /* predict size for elements */ + for (i = 0; i < VB->PrimitiveCount; ++i) + { + if (!VB->Primitive[i].count) + continue; + /* If primitive.count is less than MAX_CONVERSION_SIZE + rendering code may decide convert to elts. + In that case we have to make pessimistic prediction. + and use larger of 2 paths. */ + const GLuint elts = ELTS_BUFSZ(nr_aos); + const GLuint index = INDEX_BUFSZ; + const GLuint vbuf = VBUF_BUFSZ; + if ( (!VB->Elts && VB->Primitive[i].count >= MAX_CONVERSION_SIZE) + || vbuf > index + elts) + space_required += vbuf; + else + space_required += index + elts; + space_required += AOS_BUFSZ(nr_aos); + } + space_required += SCISSOR_BUFSZ; + } + /* flush the buffer in case we need more than is left. */ + rcommonEnsureCmdBufSpace(&rmesa->radeon, space_required, __FUNCTION__); +} + /**********************************************************************/ /* Render pipeline stage */ /**********************************************************************/ @@ -410,6 +467,7 @@ static GLboolean radeon_run_tcl_render( GLcontext *ctx, } radeonReleaseArrays( ctx, ~0 ); + radeonEnsureEmitSize( ctx, inputs ); radeonEmitArrays( ctx, inputs ); rmesa->tcl.Elts = VB->Elts; -- cgit v1.2.3 From b7ec2ebe33608f9107ff7cae6b974edc8097cecd Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Fri, 21 Aug 2009 20:44:54 +0300 Subject: radeon: Improve state emit code. Trying to make understanding code easier with small refactoring and renaming. --- src/mesa/drivers/dri/r200/r200_swtcl.c | 2 +- src/mesa/drivers/dri/r200/r200_tcl.c | 2 +- src/mesa/drivers/dri/r300/r300_draw.c | 4 +- src/mesa/drivers/dri/radeon/radeon_common.c | 122 ++++++++++++++++------------ src/mesa/drivers/dri/radeon/radeon_common.h | 2 +- src/mesa/drivers/dri/radeon/radeon_dma.c | 2 +- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 2 +- src/mesa/drivers/dri/radeon/radeon_tcl.c | 2 +- 8 files changed, 78 insertions(+), 60 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 8dd2e69d33..56930c7863 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -269,7 +269,7 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) { r200ContextPtr rmesa = R200_CONTEXT(ctx); rcommonEnsureCmdBufSpace(&rmesa->radeon, - radeonCountEmitSize(&rmesa->radeon) + (12*sizeof(int)), + radeonCountStateEmitSize(&rmesa->radeon) + (12*sizeof(int)), __FUNCTION__); diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c index d8020df97c..0f5e501c1a 100644 --- a/src/mesa/drivers/dri/r200/r200_tcl.c +++ b/src/mesa/drivers/dri/r200/r200_tcl.c @@ -387,7 +387,7 @@ static void r200EnsureEmitSize( GLcontext * ctx , GLubyte* vimap_rev ) { /* count the prediction for state size */ - space_required = radeonCountEmitSize( &rmesa->radeon ); + space_required = radeonCountStateEmitSize( &rmesa->radeon ); /* vtx may be changed in r200EmitArrays so account for it if not dirty */ if (!rmesa->hw.vtx.dirty) space_required += rmesa->hw.vtx.check(rmesa->radeon.glCtx, &rmesa->hw.vtx); diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c index 1437535760..c8d2f44fa1 100644 --- a/src/mesa/drivers/dri/r300/r300_draw.c +++ b/src/mesa/drivers/dri/r300/r300_draw.c @@ -582,12 +582,12 @@ static GLuint r300PredictTryDrawPrimsSize(GLcontext *ctx, GLuint nr_prims) + SCISSORS_BUFSZ + FIREAOS_BUFSZ )*nr_prims; - state_size= radeonCountEmitSize(&r300->radeon); + state_size = radeonCountStateEmitSize(&r300->radeon); flushed = rcommonEnsureCmdBufSpace(&r300->radeon, dwords + state_size, __FUNCTION__); if (flushed) - dwords += radeonCountEmitSize(&r300->radeon); + dwords += radeonCountStateEmitSize(&r300->radeon); else dwords += state_size; diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index 6c3f7a57e5..4088ef303c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -895,7 +895,7 @@ void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei he ctx->Driver.Viewport = old_viewport; } -static void radeon_print_state_atom(radeonContextPtr radeon, struct radeon_state_atom *state) +static void radeon_print_state_atom_prekmm(radeonContextPtr radeon, struct radeon_state_atom *state) { int i, j, reg; int dwords = (*state->check) (radeon->glCtx, state); @@ -920,13 +920,22 @@ static void radeon_print_state_atom(radeonContextPtr radeon, struct radeon_state } } -static void radeon_print_state_atom_kmm(radeonContextPtr radeon, struct radeon_state_atom *state) +static void radeon_print_state_atom(radeonContextPtr radeon, struct radeon_state_atom *state) { int i, j, reg, count; - int dwords = (*state->check) (radeon->glCtx, state); + int dwords; uint32_t packet0; + if (! (DEBUG_CMDBUF || RADEON_DEBUG & DEBUG_STATE)) + return; - fprintf(stderr, " emit %s %d/%d\n", state->name, state->cmd_size, dwords); + if (!radeon->radeonScreen->kernel_mm) { + radeon_print_state_atom_prekmm(radeon, state); + return; + } + + dwords = (*state->check) (radeon->glCtx, state); + + fprintf(stderr, " emit %s %d/%d\n", state->name, dwords, state->cmd_size); if (RADEON_DEBUG & DEBUG_VERBOSE) { for (i = 0; i < state->cmd_size;) { @@ -949,60 +958,68 @@ static void radeon_print_state_atom_kmm(radeonContextPtr radeon, struct radeon_s /** * Count total size for next state emit. **/ -GLuint radeonCountEmitSize(radeonContextPtr radeon) +GLuint radeonCountStateEmitSize(radeonContextPtr radeon) { - struct radeon_state_atom *atom; - int dwords = 0; - /* check if we are going to emit full state */ - if (radeon->cmdbuf.cs->cdw && !radeon->hw.all_dirty) { - if (!radeon->hw.is_dirty) - return dwords; - foreach(atom, &radeon->hw.atomlist) { - if (atom->dirty) - dwords += atom->check(radeon->glCtx, atom); - } - } else { - foreach(atom, &radeon->hw.atomlist) { - dwords += atom->check(radeon->glCtx, atom); - } - } - return dwords; + struct radeon_state_atom *atom; + int dwords = 0; + /* check if we are going to emit full state */ + if (radeon->cmdbuf.cs->cdw && !radeon->hw.all_dirty) { + if (!radeon->hw.is_dirty) + return dwords; + foreach(atom, &radeon->hw.atomlist) { + if (atom->dirty) + dwords += atom->check(radeon->glCtx, atom); + } + } else { + foreach(atom, &radeon->hw.atomlist) { + dwords += atom->check(radeon->glCtx, atom); + } + } + return dwords; } -static INLINE void radeonEmitAtoms(radeonContextPtr radeon, GLboolean dirty) +static INLINE void radeon_emit_atom(radeonContextPtr radeon, struct radeon_state_atom *atom) { BATCH_LOCALS(radeon); - struct radeon_state_atom *atom; int dwords; + dwords = (*atom->check) (radeon->glCtx, atom); + if (dwords) { + + radeon_print_state_atom(radeon, atom); + + if (atom->emit) { + (*atom->emit)(radeon->glCtx, atom); + } else { + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_BATCH_TABLE(atom->cmd, dwords); + END_BATCH(); + } + } else { + if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { + fprintf(stderr, " skip state %s\n", + atom->name); + } + } + atom->dirty = GL_FALSE; + +} + +static INLINE void radeonEmitAtoms(radeonContextPtr radeon, GLboolean emitAll) +{ + struct radeon_state_atom *atom; + if (radeon->vtbl.pre_emit_atoms) radeon->vtbl.pre_emit_atoms(radeon); /* Emit actual atoms */ - foreach(atom, &radeon->hw.atomlist) { - if ((atom->dirty || radeon->hw.all_dirty) == dirty) { - dwords = (*atom->check) (radeon->glCtx, atom); - if (dwords) { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - if (radeon->radeonScreen->kernel_mm) - radeon_print_state_atom_kmm(radeon, atom); - else - radeon_print_state_atom(radeon, atom); - } - if (atom->emit) { - (*atom->emit)(radeon->glCtx, atom); - } else { - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(atom->cmd, dwords); - END_BATCH(); - } - atom->dirty = GL_FALSE; - } else { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - fprintf(stderr, " skip state %s\n", - atom->name); - } - } + if (radeon->hw.all_dirty || emitAll) { + foreach(atom, &radeon->hw.atomlist) + radeon_emit_atom( radeon, atom ); + } else { + foreach(atom, &radeon->hw.atomlist) { + if ( atom->dirty ) + radeon_emit_atom( radeon, atom ); } } @@ -1036,16 +1053,17 @@ void radeonEmitState(radeonContextPtr radeon) if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "Begin reemit state\n"); + radeonEmitAtoms(radeon, GL_TRUE); + } else { + + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin dirty state\n"); + radeonEmitAtoms(radeon, GL_FALSE); } - if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "Begin dirty state\n"); - - radeonEmitAtoms(radeon, GL_TRUE); radeon->hw.is_dirty = GL_FALSE; radeon->hw.all_dirty = GL_FALSE; - } diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h index 6e81100d66..a9e1ca49eb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.h +++ b/src/mesa/drivers/dri/radeon/radeon_common.h @@ -24,7 +24,7 @@ void radeonUpdatePageFlipping(radeonContextPtr rmesa); void radeonFlush(GLcontext *ctx); void radeonFinish(GLcontext * ctx); void radeonEmitState(radeonContextPtr radeon); -GLuint radeonCountEmitSize(radeonContextPtr radeon); +GLuint radeonCountStateEmitSize(radeonContextPtr radeon); void radeon_clear_tris(GLcontext *ctx, GLbitfield mask); diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 79b2b77d30..7e8f50e3df 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -431,7 +431,7 @@ restart: if (!rmesa->dma.flush) { /* make sure we have enough space to use this in cmdbuf */ rcommonEnsureCmdBufSpace(rmesa, - radeonCountEmitSize( rmesa ) + (20*sizeof(int)), + radeonCountStateEmitSize( rmesa ) + (20*sizeof(int)), __FUNCTION__); /* if cmdbuf flushed DMA restart */ if (is_empty_list(&rmesa->dma.reserved)) diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index a1b398e560..20ce6c470b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -285,7 +285,7 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r100ContextPtr rmesa = R100_CONTEXT(ctx); rcommonEnsureCmdBufSpace(&rmesa->radeon, - radeonCountEmitSize( &rmesa->radeon ) + (12*sizeof(int)), + radeonCountStateEmitSize( &rmesa->radeon ) + (12*sizeof(int)), __FUNCTION__); diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index c5a73a2818..908b3c9f06 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -388,7 +388,7 @@ static void radeonEnsureEmitSize( GLcontext * ctx , GLuint inputs ) { /* count the prediction for state size */ - space_required = radeonCountEmitSize( &rmesa->radeon ); + space_required = radeonCountStateEmitSize( &rmesa->radeon ); /* tcl may be changed in radeonEmitArrays so account for it if not dirty */ if (!rmesa->hw.tcl.dirty) space_required += rmesa->hw.tcl.check( rmesa->radeon.glCtx, &rmesa->hw.tcl ); -- cgit v1.2.3 From c3374bf97ecd82b915fb29c7c04951e2b75d4dbc Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Tue, 25 Aug 2009 19:28:00 +0300 Subject: radeon: Fix all compiler warnings. --- src/mesa/drivers/dri/r200/r200_cmdbuf.c | 3 +-- src/mesa/drivers/dri/r200/r200_pixel.c | 14 ++++++----- src/mesa/drivers/dri/r200/r200_swtcl.c | 2 +- .../drivers/dri/r300/compiler/radeon_program_alu.c | 4 ++- src/mesa/drivers/dri/r600/r600_tex.c | 4 +-- src/mesa/drivers/dri/r600/r700_debug.c | 2 +- src/mesa/drivers/dri/radeon/radeon_context.c | 1 + src/mesa/drivers/dri/radeon/radeon_cs_legacy.c | 2 +- src/mesa/drivers/dri/radeon/radeon_dma.c | 7 +++--- src/mesa/drivers/dri/radeon/radeon_ioctl.c | 2 +- src/mesa/drivers/dri/radeon/radeon_screen.c | 29 ++++------------------ src/mesa/drivers/dri/radeon/radeon_span.c | 2 ++ src/mesa/drivers/dri/radeon/radeon_swtcl.c | 2 +- 13 files changed, 31 insertions(+), 43 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c index e63935378e..5f10279e56 100644 --- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c +++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c @@ -105,7 +105,7 @@ void r200SetUpAtomList( r200ContextPtr rmesa ) insert_at_tail( &rmesa->radeon.hw.atomlist, &rmesa->hw.vpi[1] ); } -void r200EmitScissor(r200ContextPtr rmesa) +static void r200EmitScissor(r200ContextPtr rmesa) { unsigned x1, y1, x2, y2; struct radeon_renderbuffer *rrb; @@ -234,7 +234,6 @@ GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa, GLuint min_nr ) { GLushort *retval; - int ret; if (R200_DEBUG & DEBUG_IOCTL) fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive); diff --git a/src/mesa/drivers/dri/r200/r200_pixel.c b/src/mesa/drivers/dri/r200/r200_pixel.c index 654f2c6ae9..7201ddad2d 100644 --- a/src/mesa/drivers/dri/r200/r200_pixel.c +++ b/src/mesa/drivers/dri/r200/r200_pixel.c @@ -101,7 +101,7 @@ check_color_per_fragment_ops( const GLcontext *ctx ) } - +#if 0 static GLboolean clip_pixelrect( const GLcontext *ctx, const GLframebuffer *buffer, @@ -142,6 +142,7 @@ clip_pixelrect( const GLcontext *ctx, return GL_TRUE; } +#endif static GLboolean r200TryReadPixels( GLcontext *ctx, @@ -150,14 +151,14 @@ r200TryReadPixels( GLcontext *ctx, const struct gl_pixelstore_attrib *pack, GLvoid *pixels ) { + return GL_FALSE; +#if 0 r200ContextPtr rmesa = R200_CONTEXT(ctx); GLint pitch = pack->RowLength ? pack->RowLength : width; GLint blit_format; GLuint cpp = rmesa->radeon.radeonScreen->cpp; GLint size = width * height * cpp; - return GL_FALSE; -#if 0 if (R200_DEBUG & DEBUG_PIXEL) fprintf(stderr, "%s\n", __FUNCTION__); @@ -292,6 +293,10 @@ static void do_draw_pix( GLcontext *ctx, const void *pixels, GLuint planemask) { + if (R200_DEBUG & DEBUG_PIXEL) + fprintf(stderr, "%s\n", __FUNCTION__); + +#if 0 r200ContextPtr rmesa = R200_CONTEXT(ctx); __DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon); drm_clip_rect_t *box = dPriv->pClipRects; @@ -304,9 +309,6 @@ static void do_draw_pix( GLcontext *ctx, int src_offset = r200GartOffsetFromVirtual( rmesa, pixels ); int src_pitch = pitch * rmesa->radeon.radeonScreen->cpp; - if (R200_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); -#if 0 switch ( rmesa->radeon.radeonScreen->cpp ) { case 2: blit_format = R200_CP_COLOR_FORMAT_RGB565; diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 56930c7863..13bd6ac4e2 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -444,7 +444,7 @@ do { \ #define LOCAL_VARS(n) \ r200ContextPtr rmesa = R200_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = {0}, spec[n] = {0}; \ GLuint coloroffset = rmesa->swtcl.coloroffset; \ GLuint specoffset = rmesa->swtcl.specoffset; \ (void) color; (void) spec; (void) coloroffset; (void) specoffset; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c index 609e510ff2..8071899eaa 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c @@ -91,6 +91,7 @@ static struct prog_dst_register dstreg(int file, int index) dst.Index = index; dst.WriteMask = WRITEMASK_XYZW; dst.CondMask = COND_TR; + dst.RelAddr = 0; dst.CondSwizzle = SWIZZLE_NOOP; dst.CondSrc = 0; dst.pad = 0; @@ -99,10 +100,11 @@ static struct prog_dst_register dstreg(int file, int index) static struct prog_dst_register dstregtmpmask(int index, int mask) { - struct prog_dst_register dst; + struct prog_dst_register dst = {0}; dst.File = PROGRAM_TEMPORARY; dst.Index = index; dst.WriteMask = mask; + dst.RelAddr = 0; dst.CondMask = COND_TR; dst.CondSwizzle = SWIZZLE_NOOP; dst.CondSrc = 0; diff --git a/src/mesa/drivers/dri/r600/r600_tex.c b/src/mesa/drivers/dri/r600/r600_tex.c index 6d531bf0f9..a06a1fa921 100644 --- a/src/mesa/drivers/dri/r600/r600_tex.c +++ b/src/mesa/drivers/dri/r600/r600_tex.c @@ -167,9 +167,9 @@ static void r600SetTexDefaultState(radeonTexObjPtr t) } +#if 0 static GLuint aniso_filter(GLfloat anisotropy) { -#if 0 if (anisotropy >= 16.0) { return R300_TX_MAX_ANISO_16_TO_1; } else if (anisotropy >= 8.0) { @@ -181,9 +181,9 @@ static GLuint aniso_filter(GLfloat anisotropy) } else { return R300_TX_MAX_ANISO_1_TO_1; } -#endif return 0; } +#endif /** * Set the texture magnification and minification modes. diff --git a/src/mesa/drivers/dri/r600/r700_debug.c b/src/mesa/drivers/dri/r600/r700_debug.c index ecdb75ad48..dabd0d2c25 100644 --- a/src/mesa/drivers/dri/r600/r700_debug.c +++ b/src/mesa/drivers/dri/r600/r700_debug.c @@ -75,7 +75,7 @@ void DumpHwBinary(int type, void *addr, int size) { DEBUGP("0x%08x,\t", *pHw); if (i%4 == 3) - DEBUGP("\n", *pHw); + DEBUGP("0x%08x\n", *pHw); pHw++; } diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index c457fb654e..12e4ed0a95 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -202,6 +202,7 @@ static void r100_init_vtbl(radeonContextPtr radeon) radeon->vtbl.swtcl_flush = r100_swtcl_flush; radeon->vtbl.pre_emit_state = r100_vtbl_pre_emit_state; radeon->vtbl.fallback = radeonFallback; + radeon->vtbl.free_context = r100_vtbl_free_context; } /* Create the device specific context. diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c index 4f1065ebcf..587e2acf91 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c @@ -317,7 +317,7 @@ static int cs_emit(struct radeon_cs *cs) if ((!IS_R300_CLASS(csm->ctx->radeonScreen)) && (!IS_R600_CLASS(csm->ctx->radeonScreen))) { /* +r6/r7 : No irq for r6/r7 yet. */ drm_radeon_irq_emit_t emit_cmd; - emit_cmd.irq_seq = &csm->pending_age; + emit_cmd.irq_seq = (int*)&csm->pending_age; r = drmCommandWrite(cs->csm->fd, DRM_RADEON_IRQ_EMIT, &emit_cmd, sizeof(emit_cmd)); if (r) { return r; diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index f306befec4..f15013c33b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -173,6 +173,7 @@ void radeon_init_dma(radeonContextPtr rmesa) void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) { + struct radeon_dma_bo *dma_bo = NULL; /* we set minimum sizes to at least requested size aligned to next 16 bytes. */ if (size > rmesa->dma.minimum_size) @@ -191,7 +192,7 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) if (is_empty_list(&rmesa->dma.free) || last_elem(&rmesa->dma.free)->bo->size < size) { - struct radeon_dma_bo *dma_bo = CALLOC(sizeof(struct radeon_dma_bo)); + dma_bo = CALLOC_STRUCT(radeon_dma_bo); assert(dma_bo); again_alloc: @@ -208,7 +209,7 @@ again_alloc: /* We push and pop buffers from end of list so we can keep counter on unused buffers for later freeing them from begin of list */ - struct radeon_dma_bo *dma_bo = last_elem(&rmesa->dma.free); + dma_bo = last_elem(&rmesa->dma.free); assert(dma_bo->bo->cref == 1); remove_from_list(dma_bo); insert_at_head(&rmesa->dma.reserved, dma_bo); @@ -265,7 +266,7 @@ void radeonAllocDmaRegion(radeonContextPtr rmesa, void radeonFreeDmaRegions(radeonContextPtr rmesa) { - struct radeon_dma_bo *dma_bo; + struct radeon_dma_bo *dma_bo = CALLOC_STRUCT(radeon_dma_bo); struct radeon_dma_bo *temp; if (RADEON_DEBUG & DEBUG_DMA) fprintf(stderr, "%s\n", __FUNCTION__); diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index a5e4df7941..afdb5e723e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -113,7 +113,7 @@ void radeonSetUpAtomList( r100ContextPtr rmesa ) insert_at_tail(&rmesa->radeon.hw.atomlist, &rmesa->hw.glt); } -void radeonEmitScissor(r100ContextPtr rmesa) +static void radeonEmitScissor(r100ContextPtr rmesa) { BATCH_LOCALS(&rmesa->radeon); if (!rmesa->radeon.radeonScreen->kernel_mm) { diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index bdcfd10c06..e8b2dc89fe 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -259,7 +259,7 @@ radeonGetParam(__DRIscreenPrivate *sPriv, int param, void *value) struct drm_radeon_info info = { 0 }; if (sPriv->drm_version.major >= 2) { - info.value = (uint64_t)value; + info.value = (uint64_t)(uintptr_t)value; switch (param) { case RADEON_PARAM_DEVICE_ID: info.request = RADEON_INFO_DEVICE_ID; @@ -1604,28 +1604,6 @@ radeonDestroyBuffer(__DRIdrawablePrivate *driDrawPriv) _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL); } -/** - * Choose the appropriate CreateContext function based on the chipset. - * Eventually, all drivers will go through this process. - */ -static GLboolean radeonCreateContext(const __GLcontextModes * glVisual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPriv) -{ - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); -#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) - if (IS_R300_CLASS(screen)) - return r300CreateContext(glVisual, driContextPriv, sharedContextPriv); -#endif - -#if !RADEON_COMMON - (void)screen; - return r100CreateContext(glVisual, driContextPriv, sharedContextPriv); -#endif - return GL_FALSE; -} - /** * This is the driver specific part of the createNewScreen entry point. @@ -1824,8 +1802,11 @@ const struct __DriverAPIRec driDriverAPI = { #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) .CreateContext = r600CreateContext, .DestroyContext = radeonDestroyContext, +#elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) + .CreateContext = r300CreateContext, + .DestroyContext = radeonDestroyContext, #else - .CreateContext = radeonCreateContext, + .CreateContext = r100CreateContext, .DestroyContext = radeonDestroyContext, #endif .CreateBuffer = radeonCreateBuffer, diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index 5e4bf00d7a..4e100d854e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -55,6 +55,7 @@ static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb); /* r200 depth buffer is always tiled - this is the formula according to the docs unless I typo'ed in it */ +#if defined(RADEON_COMMON_FOR_R200) static GLubyte *r200_depth_2byte(const struct radeon_renderbuffer * rrb, GLint x, GLint y) { @@ -103,6 +104,7 @@ static GLubyte *r200_depth_4byte(const struct radeon_renderbuffer * rrb, } return &ptr[offset]; } +#endif /* radeon tiling on r300-r500 has 4 states, macro-linear/micro-linear diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 20ce6c470b..14d5b5a2af 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -552,7 +552,7 @@ do { \ #define LOCAL_VARS(n) \ r100ContextPtr rmesa = R100_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ + GLuint color[n] = {0}, spec[n] = {0}; \ GLuint coloroffset = rmesa->swtcl.coloroffset; \ GLuint specoffset = rmesa->swtcl.specoffset; \ (void) color; (void) spec; (void) coloroffset; (void) specoffset; -- cgit v1.2.3 From 97029c1860e8a99448ac157408e71bea3aa2559f Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Wed, 26 Aug 2009 22:24:25 +0300 Subject: radeon/r200/r300: Fix swtcl flushing not to invalidate dma region. We were check command buffer sizes too alte so allocated dma regions were freed before relocations so space checking failed. --- src/mesa/drivers/dri/r200/r200_swtcl.c | 22 +++++++++++-- src/mesa/drivers/dri/r300/r300_swtcl.c | 37 ++++++++++++++++++++-- .../drivers/dri/radeon/radeon_common_context.h | 1 + src/mesa/drivers/dri/radeon/radeon_dma.c | 7 ---- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 31 ++++++++++++------ 5 files changed, 75 insertions(+), 23 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 13bd6ac4e2..6b7279e8db 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -204,7 +204,21 @@ static void r200SetVertexFormat( GLcontext *ctx ) static void r200RenderStart( GLcontext *ctx ) { + const int vertex_array_size = 7; + const int prim_size = 3; + r200ContextPtr rmesa = R200_CONTEXT( ctx ); r200SetVertexFormat( ctx ); + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); + if (!rmesa->radeon.swtcl.primitive_counter) { + if (rcommonEnsureCmdBufSpace(&rmesa->radeon, + radeonCountStateEmitSize(&rmesa->radeon) + + vertex_array_size + prim_size, + __FUNCTION__)) + rmesa->radeon.swtcl.primitive_counter = 0; + else + rmesa->radeon.swtcl.primitive_counter = 1; + } } @@ -268,9 +282,8 @@ void r200ChooseVertexState( GLcontext *ctx ) void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - rcommonEnsureCmdBufSpace(&rmesa->radeon, - radeonCountStateEmitSize(&rmesa->radeon) + (12*sizeof(int)), - __FUNCTION__); + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); radeonEmitState(&rmesa->radeon); @@ -284,6 +297,8 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) rmesa->radeon.swtcl.hw_primitive, rmesa->radeon.swtcl.numverts); + rmesa->radeon.swtcl.primitive_counter = 0; + } /**************************************************************************/ @@ -890,6 +905,7 @@ void r200InitSwtcl( GLcontext *ctx ) init_rast_tab(); firsttime = 0; } + rmesa->radeon.swtcl.primitive_counter = 0; tnl->Driver.Render.Start = r200RenderStart; tnl->Driver.Render.Finish = r200RenderFinish; diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index 9d6f756879..18af51e666 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -83,6 +83,9 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ rmesa->swtcl.coloroffset = rmesa->swtcl.specoffset = 0; rmesa->radeon.swtcl.vertex_attr_count = 0; + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); + /* We always want non Ndc coords format */ VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr; @@ -462,6 +465,8 @@ static void r300ChooseRenderState( GLcontext *ctx ) r300ContextPtr rmesa = R300_CONTEXT(ctx); GLuint index = 0; GLuint flags = ctx->_TriangleCaps; + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); if (flags & DD_TRI_UNFILLED) index |= R300_UNFILLED_BIT; @@ -489,6 +494,8 @@ static void r300ChooseRenderState( GLcontext *ctx ) void r300RenderStart(GLcontext *ctx) { + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); r300ContextPtr rmesa = R300_CONTEXT( ctx ); r300ChooseRenderState(ctx); @@ -501,6 +508,19 @@ void r300RenderStart(GLcontext *ctx) r300UpdateShaderStates(rmesa); + const int vertex_size = 7; + const int prim_size = 3; + + if (!rmesa->radeon.swtcl.primitive_counter) { + if (rcommonEnsureCmdBufSpace(&rmesa->radeon, + radeonCountStateEmitSize(&rmesa->radeon) + + + vertex_size + prim_size, + __FUNCTION__)) + rmesa->radeon.swtcl.primitive_counter = 0; + else + rmesa->radeon.swtcl.primitive_counter = 1; + } + r300EmitCacheFlush(rmesa); /* investigate if we can put back flush optimisation if needed */ @@ -516,6 +536,8 @@ void r300RenderFinish(GLcontext *ctx) static void r300RasterPrimitive( GLcontext *ctx, GLuint hwprim ) { r300ContextPtr rmesa = R300_CONTEXT(ctx); + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); if (rmesa->radeon.swtcl.hw_primitive != hwprim) { R300_NEWPRIM( rmesa ); @@ -528,6 +550,8 @@ void r300RenderPrimitive(GLcontext *ctx, GLenum prim) r300ContextPtr rmesa = R300_CONTEXT(ctx); rmesa->radeon.swtcl.render_primitive = prim; + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); if ((prim == GL_TRIANGLES) && (ctx->_TriangleCaps & DD_TRI_UNFILLED)) return; @@ -537,6 +561,8 @@ void r300RenderPrimitive(GLcontext *ctx, GLenum prim) void r300ResetLineStipple(GLcontext *ctx) { + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); } void r300InitSwtcl(GLcontext *ctx) @@ -544,11 +570,14 @@ void r300InitSwtcl(GLcontext *ctx) TNLcontext *tnl = TNL_CONTEXT(ctx); r300ContextPtr rmesa = R300_CONTEXT(ctx); static int firsttime = 1; + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); if (firsttime) { init_rast_tab(); firsttime = 0; } + rmesa->radeon.swtcl.primitive_counter = 0; tnl->Driver.Render.Start = r300RenderStart; tnl->Driver.Render.Finish = r300RenderFinish; @@ -597,6 +626,8 @@ static void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vert { BATCH_LOCALS(&rmesa->radeon); int type, num_verts; + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); type = r300PrimitiveType(rmesa, primitive); num_verts = r300NumVerts(rmesa, vertex_nr, primitive); @@ -609,11 +640,10 @@ static void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vert void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) { + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); r300ContextPtr rmesa = R300_CONTEXT(ctx); - rcommonEnsureCmdBufSpace(&rmesa->radeon, - rmesa->radeon.hw.max_state_size + (12*sizeof(int)), - __FUNCTION__); radeonEmitState(&rmesa->radeon); r300_emit_scissor(ctx); r300EmitVertexAOS(rmesa, @@ -625,5 +655,6 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) rmesa->radeon.swtcl.hw_primitive, rmesa->radeon.swtcl.numverts); r300EmitCacheFlush(rmesa); + rmesa->radeon.swtcl.primitive_counter = 0; COMMIT_BATCH(); } diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index cb47484de1..5c4b4abaca 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -342,6 +342,7 @@ struct radeon_swtcl_info { struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; GLuint vertex_attr_count; + GLuint primitive_counter; }; #define RADEON_MAX_AOS_ARRAYS 16 diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 51f764cf47..7aa59675b7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -428,7 +428,6 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) { GLuint bytes = vsize * nverts; void *head; -restart: if (RADEON_DEBUG & DEBUG_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); if (is_empty_list(&rmesa->dma.reserved) @@ -437,13 +436,7 @@ restart: } if (!rmesa->dma.flush) { - /* make sure we have enough space to use this in cmdbuf */ - rcommonEnsureCmdBufSpace(rmesa, - radeonCountStateEmitSize( rmesa ) + (20*sizeof(int)), - __FUNCTION__); /* if cmdbuf flushed DMA restart */ - if (is_empty_list(&rmesa->dma.reserved)) - goto restart; rmesa->glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; rmesa->dma.flush = rcommon_flush_last_swtcl_prim; } diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 14d5b5a2af..c9e399f9b6 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -223,13 +223,23 @@ static void radeonSetVertexFormat( GLcontext *ctx ) static void radeonRenderStart( GLcontext *ctx ) { - r100ContextPtr rmesa = R100_CONTEXT( ctx ); - - radeonSetVertexFormat( ctx ); - - if (rmesa->radeon.dma.flush != 0 && - rmesa->radeon.dma.flush != rcommon_flush_last_swtcl_prim) - rmesa->radeon.dma.flush( ctx ); + r100ContextPtr rmesa = R100_CONTEXT( ctx ); + + radeonSetVertexFormat( ctx ); + + if (rmesa->radeon.dma.flush != 0 && + rmesa->radeon.dma.flush != rcommon_flush_last_swtcl_prim) + rmesa->radeon.dma.flush( ctx ); + + if (!rmesa->radeon.swtcl.primitive_counter) { + if (rcommonEnsureCmdBufSpace(&rmesa->radeon, + radeonCountStateEmitSize( &rmesa->radeon ) + + (8 + 8 + 7), /* scissor + primis + VertexAOS */ + __FUNCTION__)) + rmesa->radeon.swtcl.primitive_counter = 0; + else + rmesa->radeon.swtcl.primitive_counter = 1; + } } @@ -284,9 +294,6 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) { r100ContextPtr rmesa = R100_CONTEXT(ctx); - rcommonEnsureCmdBufSpace(&rmesa->radeon, - radeonCountStateEmitSize( &rmesa->radeon ) + (12*sizeof(int)), - __FUNCTION__); radeonEmitState(&rmesa->radeon); @@ -301,6 +308,9 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) rmesa->radeon.swtcl.hw_primitive, rmesa->radeon.swtcl.numverts); + + rmesa->radeon.swtcl.primitive_counter = 0; + } /* @@ -804,6 +814,7 @@ void radeonInitSwtcl( GLcontext *ctx ) if (firsttime) { init_rast_tab(); firsttime = 0; + rmesa->radeon.swtcl.primitive_counter = 0; } tnl->Driver.Render.Start = radeonRenderStart; -- cgit v1.2.3 From 570d4e375a327787441c2c7c4ae698e8993a5d6b Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Thu, 27 Aug 2009 14:21:13 +0300 Subject: radeon/r200/r300: Fix swtcl prediction to work after primitie change. Swtcl calls flush everytime primitive changes so prediction has to made again after flushing. --- src/mesa/drivers/dri/r200/r200_swtcl.c | 37 +++++++++++------ src/mesa/drivers/dri/r300/r300_swtcl.c | 47 ++++++++++++++-------- .../drivers/dri/radeon/radeon_common_context.h | 2 +- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 41 +++++++++++++------ 4 files changed, 86 insertions(+), 41 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 6b7279e8db..3d4e701155 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -201,27 +201,35 @@ static void r200SetVertexFormat( GLcontext *ctx ) } } - -static void r200RenderStart( GLcontext *ctx ) +static void r200_predict_emit_size( GLcontext *ctx ) { + r200ContextPtr rmesa = R200_CONTEXT( ctx ); const int vertex_array_size = 7; const int prim_size = 3; - r200ContextPtr rmesa = R200_CONTEXT( ctx ); - r200SetVertexFormat( ctx ); - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s\n", __func__); - if (!rmesa->radeon.swtcl.primitive_counter) { + if (!rmesa->radeon.swtcl.emit_prediction) { + const int state_size = radeonCountStateEmitSize(&rmesa->radeon); if (rcommonEnsureCmdBufSpace(&rmesa->radeon, - radeonCountStateEmitSize(&rmesa->radeon) + + state_size + vertex_array_size + prim_size, __FUNCTION__)) - rmesa->radeon.swtcl.primitive_counter = 0; + rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize(&rmesa->radeon); else - rmesa->radeon.swtcl.primitive_counter = 1; + rmesa->radeon.swtcl.emit_prediction = state_size; + rmesa->radeon.swtcl.emit_prediction += vertex_array_size + prim_size + + rmesa->radeon.cmdbuf.cs->cdw; } } +static void r200RenderStart( GLcontext *ctx ) +{ + r200SetVertexFormat( ctx ); + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); + r200_predict_emit_size( ctx ); +} + + /** * Set vertex state for SW TCL. The primary purpose of this function is to * determine in advance whether or not the hardware can / should do the @@ -296,8 +304,13 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r200EmitVbufPrim( rmesa, rmesa->radeon.swtcl.hw_primitive, rmesa->radeon.swtcl.numverts); + if ( rmesa->radeon.swtcl.emit_prediction < rmesa->radeon.cmdbuf.cs->cdw ) + WARN_ONCE("Rendering was %d commands larger than predicted size." + " We might overflow command buffer.\n", + rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction ); - rmesa->radeon.swtcl.primitive_counter = 0; + rmesa->radeon.swtcl.emit_prediction = 0; + r200_predict_emit_size( ctx ); } @@ -905,7 +918,7 @@ void r200InitSwtcl( GLcontext *ctx ) init_rast_tab(); firsttime = 0; } - rmesa->radeon.swtcl.primitive_counter = 0; + rmesa->radeon.swtcl.emit_prediction = 0; tnl->Driver.Render.Start = r200RenderStart; tnl->Driver.Render.Finish = r200RenderFinish; diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index 18af51e666..ca17f306d1 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -490,6 +490,27 @@ static void r300ChooseRenderState( GLcontext *ctx ) rmesa->radeon.swtcl.RenderIndex = index; } } +static void r300_predict_emit_size( GLcontext *ctx ) +{ + r300ContextPtr rmesa = R300_CONTEXT( ctx ); + if (!rmesa->radeon.swtcl.emit_prediction) { + const int vertex_size = 7; + const int prim_size = 3; + const int cache_flush_size = 4; + const int state_size = radeonCountStateEmitSize(&rmesa->radeon); + + if (rcommonEnsureCmdBufSpace(&rmesa->radeon, + state_size + + + vertex_size + prim_size, + __FUNCTION__)) + rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize(&rmesa->radeon); + else + rmesa->radeon.swtcl.emit_prediction = state_size; + + rmesa->radeon.swtcl.emit_prediction += rmesa->radeon.cmdbuf.cs->cdw + + vertex_size + prim_size + cache_flush_size * 2; + } +} void r300RenderStart(GLcontext *ctx) @@ -508,20 +529,7 @@ void r300RenderStart(GLcontext *ctx) r300UpdateShaderStates(rmesa); - const int vertex_size = 7; - const int prim_size = 3; - - if (!rmesa->radeon.swtcl.primitive_counter) { - if (rcommonEnsureCmdBufSpace(&rmesa->radeon, - radeonCountStateEmitSize(&rmesa->radeon) + - + vertex_size + prim_size, - __FUNCTION__)) - rmesa->radeon.swtcl.primitive_counter = 0; - else - rmesa->radeon.swtcl.primitive_counter = 1; - } - - r300EmitCacheFlush(rmesa); + r300_predict_emit_size( ctx ); /* investigate if we can put back flush optimisation if needed */ if (rmesa->radeon.dma.flush != NULL) { @@ -577,7 +585,7 @@ void r300InitSwtcl(GLcontext *ctx) init_rast_tab(); firsttime = 0; } - rmesa->radeon.swtcl.primitive_counter = 0; + rmesa->radeon.swtcl.emit_prediction = 0; tnl->Driver.Render.Start = r300RenderStart; tnl->Driver.Render.Finish = r300RenderFinish; @@ -644,6 +652,8 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) fprintf(stderr, "%s\n", __func__); r300ContextPtr rmesa = R300_CONTEXT(ctx); + r300EmitCacheFlush(rmesa); + radeonEmitState(&rmesa->radeon); r300_emit_scissor(ctx); r300EmitVertexAOS(rmesa, @@ -655,6 +665,11 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) rmesa->radeon.swtcl.hw_primitive, rmesa->radeon.swtcl.numverts); r300EmitCacheFlush(rmesa); - rmesa->radeon.swtcl.primitive_counter = 0; + if ( rmesa->radeon.swtcl.emit_prediction < rmesa->radeon.cmdbuf.cs->cdw ) + WARN_ONCE("Rendering was %d commands larger than predicted size." + " We might overflow command buffer.\n", + rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction ); + rmesa->radeon.swtcl.emit_prediction = 0; + r300_predict_emit_size( ctx ); COMMIT_BATCH(); } diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index 07a0c4c055..427eb946ff 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -342,7 +342,7 @@ struct radeon_swtcl_info { struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; GLuint vertex_attr_count; - GLuint primitive_counter; + GLuint emit_prediction; }; #define RADEON_MAX_AOS_ARRAYS 16 diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index c9e399f9b6..7358e22255 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -220,6 +220,27 @@ static void radeonSetVertexFormat( GLcontext *ctx ) } } +static void radeon_predict_emit_size( GLcontext* ctx ) +{ + r100ContextPtr rmesa = R100_CONTEXT( ctx ); + + if (!rmesa->radeon.swtcl.emit_prediction) { + const int state_size = radeonCountStateEmitSize( &rmesa->radeon ); + const int scissor_size = 8; + const int prims_size = 8; + const int vertex_size = 7; + + if (rcommonEnsureCmdBufSpace(&rmesa->radeon, + state_size + + (scissor_size + prims_size + vertex_size), + __FUNCTION__)) + rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize( &rmesa->radeon ); + else + rmesa->radeon.swtcl.emit_prediction = state_size; + rmesa->radeon.swtcl.emit_prediction += scissor_size + prims_size + vertex_size + + rmesa->radeon.cmdbuf.cs->cdw; + } +} static void radeonRenderStart( GLcontext *ctx ) { @@ -230,16 +251,7 @@ static void radeonRenderStart( GLcontext *ctx ) if (rmesa->radeon.dma.flush != 0 && rmesa->radeon.dma.flush != rcommon_flush_last_swtcl_prim) rmesa->radeon.dma.flush( ctx ); - - if (!rmesa->radeon.swtcl.primitive_counter) { - if (rcommonEnsureCmdBufSpace(&rmesa->radeon, - radeonCountStateEmitSize( &rmesa->radeon ) + - (8 + 8 + 7), /* scissor + primis + VertexAOS */ - __FUNCTION__)) - rmesa->radeon.swtcl.primitive_counter = 0; - else - rmesa->radeon.swtcl.primitive_counter = 1; - } + radeon_predict_emit_size( ctx ); } @@ -307,9 +319,14 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) rmesa->swtcl.vertex_format, rmesa->radeon.swtcl.hw_primitive, rmesa->radeon.swtcl.numverts); + if ( rmesa->radeon.swtcl.emit_prediction < rmesa->radeon.cmdbuf.cs->cdw ) + WARN_ONCE("Rendering was %d commands larger than predicted size." + " We might overflow command buffer.\n", + rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction ); + radeon_predict_emit_size( ctx ); - rmesa->radeon.swtcl.primitive_counter = 0; + rmesa->radeon.swtcl.emit_prediction = 0; } @@ -814,8 +831,8 @@ void radeonInitSwtcl( GLcontext *ctx ) if (firsttime) { init_rast_tab(); firsttime = 0; - rmesa->radeon.swtcl.primitive_counter = 0; } + rmesa->radeon.swtcl.emit_prediction = 0; tnl->Driver.Render.Start = radeonRenderStart; tnl->Driver.Render.Finish = radeonRenderFinish; -- cgit v1.2.3 From 0926a5f4a110926711d0f8a4eb297dc772b09aa2 Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Sat, 29 Aug 2009 00:46:08 +0300 Subject: radeon: Fix swtcl emit pediction. Problem was to find the correct place to run prediction. Only place that is called for every primitive is ALLOC_VERTS so we have to do prediction there before allocation. --- src/mesa/drivers/dri/r200/r200_swtcl.c | 19 ++++++--- src/mesa/drivers/dri/r300/r300_swtcl.c | 63 +++++++++++++++++------------- src/mesa/drivers/dri/radeon/radeon_dma.c | 13 +++--- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 20 ++++++---- 4 files changed, 70 insertions(+), 45 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 3d4e701155..0e7d24e7e9 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -201,9 +201,10 @@ static void r200SetVertexFormat( GLcontext *ctx ) } } -static void r200_predict_emit_size( GLcontext *ctx ) +static void r200_predict_emit_size( r200ContextPtr rmesa ) { - r200ContextPtr rmesa = R200_CONTEXT( ctx ); + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s\n", __func__); const int vertex_array_size = 7; const int prim_size = 3; if (!rmesa->radeon.swtcl.emit_prediction) { @@ -226,7 +227,6 @@ static void r200RenderStart( GLcontext *ctx ) r200SetVertexFormat( ctx ); if (RADEON_DEBUG & DEBUG_VERTS) fprintf(stderr, "%s\n", __func__); - r200_predict_emit_size( ctx ); } @@ -310,7 +310,6 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction ); rmesa->radeon.swtcl.emit_prediction = 0; - r200_predict_emit_size( ctx ); } @@ -358,11 +357,21 @@ static void r200ResetLineStipple( GLcontext *ctx ); #define HAVE_POLYGONS 1 #define HAVE_ELTS 0 +static void* r200_alloc_verts( r200ContextPtr rmesa, GLuint n, GLuint size) +{ + void *rv; + do { + r200_predict_emit_size( rmesa ); + rv = rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 ); + } while(!rv); + return rv; +} + #undef LOCAL_VARS #undef ALLOC_VERTS #define CTX_ARG r200ContextPtr rmesa #define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size -#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 ) +#define ALLOC_VERTS( n, size ) r200_alloc_verts(rmesa, n, size) #define LOCAL_VARS \ r200ContextPtr rmesa = R200_CONTEXT(ctx); \ const char *r200verts = (char *)rmesa->radeon.swtcl.verts; diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index ca17f306d1..e4a56cadc6 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -43,9 +43,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define EMIT_ATTR( ATTR, STYLE ) \ do { \ - rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR); \ - rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE); \ - rmesa->radeon.swtcl.vertex_attr_count++; \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].attrib = (ATTR); \ + rmesa->radeon.swtcl.vertex_attrs[rmesa->radeon.swtcl.vertex_attr_count].format = (STYLE); \ + rmesa->radeon.swtcl.vertex_attr_count++; \ } while (0) #define EMIT_PAD( N ) \ @@ -242,6 +242,27 @@ static void r300PrepareVertices(GLcontext *ctx) rmesa->radeon.swtcl.vertex_size /= 4; } +static void r300_predict_emit_size( r300ContextPtr rmesa ) +{ + if (!rmesa->radeon.swtcl.emit_prediction) { + const int vertex_size = 7; + const int prim_size = 3; + const int cache_flush_size = 4; + const int state_size = radeonCountStateEmitSize(&rmesa->radeon); + + if (rcommonEnsureCmdBufSpace(&rmesa->radeon, + state_size + + + vertex_size + prim_size, + __FUNCTION__)) + rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize(&rmesa->radeon); + else + rmesa->radeon.swtcl.emit_prediction = state_size; + + rmesa->radeon.swtcl.emit_prediction += rmesa->radeon.cmdbuf.cs->cdw + + vertex_size + prim_size + cache_flush_size * 2; + } +} + static GLuint reduced_prim[] = { GL_POINTS, @@ -275,11 +296,21 @@ static void r300RasterPrimitive( GLcontext *ctx, GLuint prim ); #define HAVE_POLYGONS 1 #define HAVE_ELTS 1 +static void* r300_alloc_verts(r300ContextPtr rmesa, GLuint n, GLuint size) +{ + void *rv; + do { + r300_predict_emit_size( rmesa ); + rv = rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 ); + } while (!rv); + return rv; +} + #undef LOCAL_VARS #undef ALLOC_VERTS #define CTX_ARG r300ContextPtr rmesa #define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size -#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, size * 4 ) +#define ALLOC_VERTS( n, size ) r300_alloc_verts(rmesa, n, size); #define LOCAL_VARS \ r300ContextPtr rmesa = R300_CONTEXT(ctx); \ const char *r300verts = (char *)rmesa->radeon.swtcl.verts; @@ -490,28 +521,6 @@ static void r300ChooseRenderState( GLcontext *ctx ) rmesa->radeon.swtcl.RenderIndex = index; } } -static void r300_predict_emit_size( GLcontext *ctx ) -{ - r300ContextPtr rmesa = R300_CONTEXT( ctx ); - if (!rmesa->radeon.swtcl.emit_prediction) { - const int vertex_size = 7; - const int prim_size = 3; - const int cache_flush_size = 4; - const int state_size = radeonCountStateEmitSize(&rmesa->radeon); - - if (rcommonEnsureCmdBufSpace(&rmesa->radeon, - state_size + - + vertex_size + prim_size, - __FUNCTION__)) - rmesa->radeon.swtcl.emit_prediction = radeonCountStateEmitSize(&rmesa->radeon); - else - rmesa->radeon.swtcl.emit_prediction = state_size; - - rmesa->radeon.swtcl.emit_prediction += rmesa->radeon.cmdbuf.cs->cdw - + vertex_size + prim_size + cache_flush_size * 2; - } -} - void r300RenderStart(GLcontext *ctx) { @@ -529,7 +538,6 @@ void r300RenderStart(GLcontext *ctx) r300UpdateShaderStates(rmesa); - r300_predict_emit_size( ctx ); /* investigate if we can put back flush optimisation if needed */ if (rmesa->radeon.dma.flush != NULL) { @@ -670,6 +678,5 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) " We might overflow command buffer.\n", rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction ); rmesa->radeon.swtcl.emit_prediction = 0; - r300_predict_emit_size( ctx ); COMMIT_BATCH(); } diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 7aa59675b7..386262b126 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -182,9 +182,6 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) fprintf(stderr, "%s\n", __FUNCTION__); - if (rmesa->dma.flush) { - rmesa->dma.flush(rmesa->glCtx); - } /* unmap old reserved bo */ if (!is_empty_list(&rmesa->dma.reserved)) @@ -430,9 +427,15 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) void *head; if (RADEON_DEBUG & DEBUG_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); - if (is_empty_list(&rmesa->dma.reserved) - || rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) { + if(is_empty_list(&rmesa->dma.reserved) + ||rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) { + if (rmesa->dma.flush) { + rmesa->dma.flush(rmesa->glCtx); + } + radeonRefillCurrentDmaRegion(rmesa, bytes); + + return NULL; } if (!rmesa->dma.flush) { diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 7358e22255..32df569257 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -220,9 +220,8 @@ static void radeonSetVertexFormat( GLcontext *ctx ) } } -static void radeon_predict_emit_size( GLcontext* ctx ) +static void radeon_predict_emit_size( r100ContextPtr rmesa ) { - r100ContextPtr rmesa = R100_CONTEXT( ctx ); if (!rmesa->radeon.swtcl.emit_prediction) { const int state_size = radeonCountStateEmitSize( &rmesa->radeon ); @@ -251,7 +250,6 @@ static void radeonRenderStart( GLcontext *ctx ) if (rmesa->radeon.dma.flush != 0 && rmesa->radeon.dma.flush != rcommon_flush_last_swtcl_prim) rmesa->radeon.dma.flush( ctx ); - radeon_predict_emit_size( ctx ); } @@ -324,7 +322,6 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) " We might overflow command buffer.\n", rmesa->radeon.cmdbuf.cs->cdw - rmesa->radeon.swtcl.emit_prediction ); - radeon_predict_emit_size( ctx ); rmesa->radeon.swtcl.emit_prediction = 0; @@ -369,6 +366,16 @@ radeonDmaPrimitive( r100ContextPtr rmesa, GLenum prim ) // assert(rmesa->radeon.dma.current.ptr == rmesa->radeon.dma.current.start); } +static void* radeon_alloc_verts( r100ContextPtr rmesa , GLuint nr, GLuint size ) +{ + void *rv; + do { + radeon_predict_emit_size( rmesa ); + rv = rcommonAllocDmaLowVerts( &rmesa->radeon, nr, size ); + } while (!rv); + return rv; +} + #define LOCAL_VARS r100ContextPtr rmesa = R100_CONTEXT(ctx) #define INIT( prim ) radeonDmaPrimitive( rmesa, prim ) #define FLUSH() RADEON_NEWPRIM( rmesa ) @@ -376,8 +383,7 @@ radeonDmaPrimitive( r100ContextPtr rmesa, GLenum prim ) // (((int)rmesa->radeon.dma.current.end - (int)rmesa->radeon.dma.current.ptr) / (rmesa->radeon.swtcl.vertex_size*4)) #define GET_SUBSEQUENT_VB_MAX_VERTS() \ ((RADEON_BUFFER_SIZE) / (rmesa->radeon.swtcl.vertex_size*4)) -#define ALLOC_VERTS( nr ) \ - rcommonAllocDmaLowVerts( &rmesa->radeon, nr, rmesa->radeon.swtcl.vertex_size * 4 ) +#define ALLOC_VERTS( nr ) radeon_alloc_verts( rmesa, nr, rmesa->radeon.swtcl.vertex_size * 4 ) #define EMIT_VERTS( ctx, j, nr, buf ) \ _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf) @@ -470,7 +476,7 @@ static void radeonResetLineStipple( GLcontext *ctx ); #undef ALLOC_VERTS #define CTX_ARG r100ContextPtr rmesa #define GET_VERTEX_DWORDS() rmesa->radeon.swtcl.vertex_size -#define ALLOC_VERTS( n, size ) rcommonAllocDmaLowVerts( &rmesa->radeon, n, (size) * 4 ) +#define ALLOC_VERTS( n, size ) radeon_alloc_verts( rmesa, n, (size) * 4 ) #undef LOCAL_VARS #define LOCAL_VARS \ r100ContextPtr rmesa = R100_CONTEXT(ctx); \ -- cgit v1.2.3 From 4e0d99a63588c67a955f797733da32d04e6f4ee6 Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Sun, 30 Aug 2009 19:44:16 +0300 Subject: r100: Use shared debug code. Converted r100 to use shared debug code with sed and fast compile check. New code has compability layer so old debugging code doesn't have to be changed all immidiatly. --- src/mesa/drivers/dri/radeon/Makefile | 11 ++-- src/mesa/drivers/dri/radeon/radeon_bo_legacy.c | 4 +- src/mesa/drivers/dri/radeon/radeon_common.c | 59 ++++++++++------------ .../drivers/dri/radeon/radeon_common_context.c | 14 ++--- .../drivers/dri/radeon/radeon_common_context.h | 45 +---------------- src/mesa/drivers/dri/radeon/radeon_context.c | 19 ------- src/mesa/drivers/dri/radeon/radeon_debug.c | 1 + src/mesa/drivers/dri/radeon/radeon_debug.h | 33 ++++++------ src/mesa/drivers/dri/radeon/radeon_dma.c | 26 +++++----- src/mesa/drivers/dri/radeon/radeon_fbo.c | 2 +- src/mesa/drivers/dri/radeon/radeon_ioctl.c | 16 +++--- src/mesa/drivers/dri/radeon/radeon_lighting.c | 6 +-- src/mesa/drivers/dri/radeon/radeon_lock.c | 2 +- src/mesa/drivers/dri/radeon/radeon_maos_arrays.c | 10 ++-- src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h | 3 +- src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c | 2 +- src/mesa/drivers/dri/radeon/radeon_sanity.c | 4 +- src/mesa/drivers/dri/radeon/radeon_screen.c | 33 ++---------- src/mesa/drivers/dri/radeon/radeon_state.c | 8 +-- src/mesa/drivers/dri/radeon/radeon_swtcl.c | 14 ++--- src/mesa/drivers/dri/radeon/radeon_tcl.c | 6 +-- src/mesa/drivers/dri/radeon/radeon_tex.c | 11 ++-- src/mesa/drivers/dri/radeon/radeon_texstate.c | 8 +-- src/mesa/drivers/dri/radeon/radeon_texture.c | 10 ++-- 24 files changed, 125 insertions(+), 222 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_swtcl.c') diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile index 8fe9d98a0b..1f286776b5 100644 --- a/src/mesa/drivers/dri/radeon/Makefile +++ b/src/mesa/drivers/dri/radeon/Makefile @@ -15,17 +15,18 @@ CS_SOURCES = radeon_cs_space_drm.c endif RADEON_COMMON_SOURCES = \ - radeon_texture.c \ + radeon_bo_legacy.c \ radeon_common_context.c \ radeon_common.c \ + radeon_cs_legacy.c \ radeon_dma.c \ + radeon_debug.c \ + radeon_fbo.c \ radeon_lock.c \ - radeon_bo_legacy.c \ - radeon_cs_legacy.c \ radeon_mipmap_tree.c \ + radeon_queryobj.c \ radeon_span.c \ - radeon_fbo.c \ - radeon_queryobj.c + radeon_texture.c DRIVER_SOURCES = \ radeon_context.c \ diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c index a10c6b73ab..3e7547d2f9 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c @@ -640,7 +640,7 @@ static int bo_vram_validate(struct radeon_bo *bo, &tex, sizeof(drm_radeon_texture_t)); if (ret) { - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "DRM_RADEON_TEXTURE: again!\n"); usleep(1); } @@ -677,7 +677,7 @@ static int bo_vram_validate(struct radeon_bo *bo, &tex, sizeof(drm_radeon_texture_t)); if (ret) { - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "DRM_RADEON_TEXTURE: again!\n"); usleep(1); } diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index ee0cbf65ca..aaa5165bc8 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -91,7 +91,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * 1 most output * 2 also print state alues */ -#define DEBUG_CMDBUF 0 +#define RADEON_CMDBUF 0 /* ============================================================= * Scissoring @@ -521,7 +521,7 @@ void radeonCopyBuffer( __DRIdrawablePrivate *dPriv, rfb = dPriv->driverPrivate; - if ( RADEON_DEBUG & DEBUG_IOCTL ) { + if ( RADEON_DEBUG & RADEON_IOCTL ) { fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, (void *) rmesa->glCtx ); } @@ -615,7 +615,7 @@ static GLboolean radeonPageFlip( __DRIdrawablePrivate *dPriv ) LOCK_HARDWARE(radeon); - if ( RADEON_DEBUG & DEBUG_IOCTL ) { + if ( RADEON_DEBUG & RADEON_IOCTL ) { fprintf(stderr, "%s: pfCurrentPage: %d %d\n", __FUNCTION__, radeon->sarea->pfCurrentPage, radeon->sarea->pfState); } @@ -875,7 +875,7 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb) */ void radeonDrawBuffer( GLcontext *ctx, GLenum mode ) { - if (RADEON_DEBUG & DEBUG_DRI) + if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr( mode )); @@ -981,7 +981,7 @@ static void radeon_print_state_atom_prekmm(radeonContextPtr radeon, struct radeo fprintf(stderr, " emit %s %d/%d\n", state->name, dwords, state->cmd_size); - if (DEBUG_CMDBUF > 1 && RADEON_DEBUG & DEBUG_VERBOSE) { + if (radeon_is_debug_enabled(RADEON_STATE, RADEON_TRACE)) { if (dwords > state->cmd_size) dwords = state->cmd_size; @@ -1006,7 +1006,7 @@ static void radeon_print_state_atom(radeonContextPtr radeon, struct radeon_state int i, j, reg, count; int dwords; uint32_t packet0; - if (! (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) ) + if (!radeon_is_debug_enabled(RADEON_STATE, RADEON_VERBOSE) ) return; if (!radeon->radeonScreen->kernel_mm) { @@ -1018,7 +1018,7 @@ static void radeon_print_state_atom(radeonContextPtr radeon, struct radeon_state fprintf(stderr, " emit %s %d/%d\n", state->name, dwords, state->cmd_size); - if (DEBUG_CMDBUF > 1 && RADEON_DEBUG & DEBUG_VERBOSE) { + if (radeon_is_debug_enabled(RADEON_STATE, RADEON_TRACE)) { if (dwords > state->cmd_size) dwords = state->cmd_size; for (i = 0; i < dwords;) { @@ -1046,17 +1046,15 @@ GLuint radeonCountStateEmitSize(radeonContextPtr radeon) struct radeon_state_atom *atom; GLuint dwords = 0; /* check if we are going to emit full state */ - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_VERBOSE) - fprintf(stderr, "%s\n", __func__); if (radeon->cmdbuf.cs->cdw && !radeon->hw.all_dirty) { if (!radeon->hw.is_dirty) - return dwords; + goto out; foreach(atom, &radeon->hw.atomlist) { if (atom->dirty) { const GLuint atom_size = atom->check(radeon->glCtx, atom); dwords += atom_size; - if (DEBUG_CMDBUF && atom_size) { + if (RADEON_CMDBUF && atom_size) { radeon_print_state_atom(radeon, atom); } } @@ -1065,12 +1063,14 @@ GLuint radeonCountStateEmitSize(radeonContextPtr radeon) foreach(atom, &radeon->hw.atomlist) { const GLuint atom_size = atom->check(radeon->glCtx, atom); dwords += atom_size; - if (DEBUG_CMDBUF && atom_size) { + if (RADEON_CMDBUF && atom_size) { radeon_print_state_atom(radeon, atom); } } } +out: + radeon_print(RADEON_STATE, RADEON_NORMAL, "%s %u\n", __func__, dwords); return dwords; } @@ -1092,10 +1092,7 @@ static INLINE void radeon_emit_atom(radeonContextPtr radeon, struct radeon_state END_BATCH(); } } else { - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_STATE) { - fprintf(stderr, " skip state %s\n", - atom->name); - } + radeon_print(RADEON_STATE, RADEON_VERBOSE, " skip state %s\n", atom->name); } atom->dirty = GL_FALSE; @@ -1135,8 +1132,7 @@ static GLboolean radeon_revalidate_bos(GLcontext *ctx) void radeonEmitState(radeonContextPtr radeon) { - if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) - fprintf(stderr, "%s\n", __FUNCTION__); + radeon_print(RADEON_STATE, RADEON_NORMAL, "%s\n", __FUNCTION__); if (radeon->vtbl.pre_emit_state) radeon->vtbl.pre_emit_state(radeon); @@ -1146,13 +1142,13 @@ void radeonEmitState(radeonContextPtr radeon) return; if (!radeon->cmdbuf.cs->cdw) { - if (RADEON_DEBUG & DEBUG_STATE) + if (RADEON_DEBUG & RADEON_STATE) fprintf(stderr, "Begin reemit state\n"); radeonEmitAtoms(radeon, GL_TRUE); } else { - if (RADEON_DEBUG & DEBUG_STATE) + if (RADEON_DEBUG & RADEON_STATE) fprintf(stderr, "Begin dirty state\n"); radeonEmitAtoms(radeon, GL_FALSE); @@ -1166,7 +1162,7 @@ void radeonEmitState(radeonContextPtr radeon) void radeonFlush(GLcontext *ctx) { radeonContextPtr radeon = RADEON_CONTEXT(ctx); - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s %d\n", __FUNCTION__, radeon->cmdbuf.cs->cdw); /* okay if we have no cmds in the buffer && @@ -1258,7 +1254,7 @@ int rcommonFlushCmdBufLocked(radeonContextPtr rmesa, const char *caller) } rmesa->cmdbuf.flushing = 1; - if (RADEON_DEBUG & DEBUG_IOCTL) { + if (RADEON_DEBUG & RADEON_IOCTL) { fprintf(stderr, "%s from %s - %i cliprects\n", __FUNCTION__, caller, rmesa->numClipRects); } @@ -1327,15 +1323,13 @@ void rcommonInitCmdBuf(radeonContextPtr rmesa) if (size > 64 * 256) size = 64 * 256; - if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) { - fprintf(stderr, "sizeof(drm_r300_cmd_header_t)=%zd\n", - sizeof(drm_r300_cmd_header_t)); - fprintf(stderr, "sizeof(drm_radeon_cmd_buffer_t)=%zd\n", - sizeof(drm_radeon_cmd_buffer_t)); - fprintf(stderr, + radeon_print(RADEON_CS, RADEON_VERBOSE, + "sizeof(drm_r300_cmd_header_t)=%zd\n", sizeof(drm_r300_cmd_header_t)); + radeon_print(RADEON_CS, RADEON_VERBOSE, + "sizeof(drm_radeon_cmd_buffer_t)=%zd\n", sizeof(drm_radeon_cmd_buffer_t)); + radeon_print(RADEON_CS, RADEON_VERBOSE, "Allocating %d bytes command buffer (max state is %d bytes)\n", size * 4, rmesa->hw.max_state_size * 4); - } if (rmesa->radeonScreen->kernel_mm) { int fd = rmesa->radeonScreen->driScreen->fd; @@ -1388,14 +1382,13 @@ void rcommonBeginBatch(radeonContextPtr rmesa, int n, int line) { if (!rmesa->cmdbuf.cs->cdw && dostate) { - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "Reemit state after flush (from %s)\n", function); + radeon_print(RADEON_STATE, RADEON_NORMAL, + "Reemit state after flush (from %s)\n", function); radeonEmitState(rmesa); } radeon_cs_begin(rmesa->cmdbuf.cs, n, file, function, line); - if (DEBUG_CMDBUF && RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "BEGIN_BATCH(%d) at %d, from %s:%i\n", + radeon_print(RADEON_CS, RADEON_VERBOSE, "BEGIN_BATCH(%d) at %d, from %s:%i\n", n, rmesa->cmdbuf.cs->cdw, function, line); } diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index b76efa8eaa..71ee06d9a7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -363,7 +363,7 @@ GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv) { radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; - if (RADEON_DEBUG & DEBUG_DRI) + if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "%s ctx %p\n", __FUNCTION__, radeon->glCtx); @@ -527,7 +527,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) char *regname; struct radeon_bo *depth_bo = NULL, *bo; - if (RADEON_DEBUG & DEBUG_DRI) + if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable); draw = drawable->driverPrivate; @@ -654,7 +654,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) continue; } - if (RADEON_DEBUG & DEBUG_DRI) + if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "attaching buffer %s, %d, at %d, cpp %d, pitch %d\n", regname, buffers[i].name, buffers[i].attachment, @@ -667,7 +667,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) rb->has_surface = 0; if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_bo) { - if (RADEON_DEBUG & DEBUG_DRI) + if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "(reusing depth buffer as stencil)\n"); bo = depth_bo; radeon_bo_ref(bo); @@ -740,7 +740,7 @@ GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, struct gl_framebuffer *readfb; if (!driContextPriv) { - if (RADEON_DEBUG & DEBUG_DRI) + if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "%s ctx is null\n", __FUNCTION__); _mesa_make_current(NULL, NULL, NULL); return GL_TRUE; @@ -762,7 +762,7 @@ GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, radeon_make_renderbuffer_current(radeon, drfb); } - if (RADEON_DEBUG & DEBUG_DRI) + if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "%s ctx %p dfb %p rfb %p\n", __FUNCTION__, radeon->glCtx, drfb, readfb); driUpdateFramebufferSize(radeon->glCtx, driDrawPriv); @@ -797,7 +797,7 @@ GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, } - if (RADEON_DEBUG & DEBUG_DRI) + if (RADEON_DEBUG & RADEON_DRI) fprintf(stderr, "End %s\n", __FUNCTION__); return GL_TRUE; diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index 3463b4d264..d17060d5cc 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -8,6 +8,7 @@ #include "tnl/t_context.h" #include "main/colormac.h" +#include "radeon_debug.h" #include "radeon_screen.h" #include "radeon_drm.h" #include "dri_util.h" @@ -18,22 +19,6 @@ struct radeon_context; #include "radeon_bocs_wrapper.h" -/* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html . - I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble - with other compilers ... GLUE! -*/ -#define WARN_ONCE(a, ...) { \ - static int warn##__LINE__=1; \ - if(warn##__LINE__){ \ - fprintf(stderr, "*********************************WARN_ONCE*********************************\n"); \ - fprintf(stderr, "File %s function %s line %d\n", \ - __FILE__, __FUNCTION__, __LINE__); \ - fprintf(stderr, a, ## __VA_ARGS__);\ - fprintf(stderr, "***************************************************************************\n"); \ - warn##__LINE__=0;\ - } \ - } - /* This union is used to avoid warnings/miscompilation with float to uint32_t casts due to strict-aliasing */ typedef union { GLfloat f; uint32_t ui32; } float_ui32_type; @@ -401,23 +386,6 @@ struct radeon_dri_mirror { int drmMinor; }; -#define DEBUG_TEXTURE 0x001 -#define DEBUG_STATE 0x002 -#define DEBUG_IOCTL 0x004 -#define DEBUG_PRIMS 0x008 -#define DEBUG_VERTS 0x010 -#define DEBUG_FALLBACKS 0x020 -#define DEBUG_VFMT 0x040 -#define DEBUG_CODEGEN 0x080 -#define DEBUG_VERBOSE 0x100 -#define DEBUG_DRI 0x200 -#define DEBUG_DMA 0x400 -#define DEBUG_SANITY 0x800 -#define DEBUG_SYNC 0x1000 -#define DEBUG_PIXEL 0x2000 -#define DEBUG_MEMORY 0x4000 - - typedef void (*radeon_tri_func) (radeonContextPtr, radeonVertex *, radeonVertex *, radeonVertex *); @@ -621,15 +589,4 @@ GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driReadPriv); extern void radeonDestroyContext(__DRIcontextPrivate * driContextPriv); -/* ================================================================ - * Debugging: - */ -#define DO_DEBUG 1 - -#if DO_DEBUG -extern int RADEON_DEBUG; -#else -#define RADEON_DEBUG 0 -#endif - #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index 32485a7270..8f4485aee7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -137,25 +137,6 @@ static const struct tnl_pipeline_stage *radeon_pipeline[] = { NULL, }; -static const struct dri_debug_control debug_control[] = -{ - { "fall", DEBUG_FALLBACKS }, - { "tex", DEBUG_TEXTURE }, - { "ioctl", DEBUG_IOCTL }, - { "prim", DEBUG_PRIMS }, - { "vert", DEBUG_VERTS }, - { "state", DEBUG_STATE }, - { "code", DEBUG_CODEGEN }, - { "vfmt", DEBUG_VFMT }, - { "vtxf", DEBUG_VFMT }, - { "verb", DEBUG_VERBOSE }, - { "dri", DEBUG_DRI }, - { "dma", DEBUG_DMA }, - { "san", DEBUG_SANITY }, - { "sync", DEBUG_SYNC }, - { NULL, 0 } -}; - static void r100_get_lock(radeonContextPtr radeon) { r100ContextPtr rmesa = (r100ContextPtr)radeon; diff --git a/src/mesa/drivers/dri/radeon/radeon_debug.c b/src/mesa/drivers/dri/radeon/radeon_debug.c index b5e7b668ce..cb1fd63cf7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_debug.c +++ b/src/mesa/drivers/dri/radeon/radeon_debug.c @@ -35,6 +35,7 @@ static const struct dri_debug_control debug_control[] = { {"fall", RADEON_FALLBACKS}, {"tex", RADEON_TEXTURE}, {"ioctl", RADEON_IOCTL}, + {"verts", RADEON_RENDER}, {"render", RADEON_RENDER}, {"swrender", RADEON_SWRENDER}, {"state", RADEON_STATE}, diff --git a/src/mesa/drivers/dri/radeon/radeon_debug.h b/src/mesa/drivers/dri/radeon/radeon_debug.h index 63827103d5..ace8a99e26 100644 --- a/src/mesa/drivers/dri/radeon/radeon_debug.h +++ b/src/mesa/drivers/dri/radeon/radeon_debug.h @@ -49,22 +49,23 @@ typedef enum radeon_debug_levels { #endif typedef enum radeon_debug_types { - RADEON_TEXTURE = 0x0001, - RADEON_STATE = 0x0002, - RADEON_IOCTL = 0x0004, - RADEON_RENDER = 0x0008, - RADEON_SWRENDER = 0x0010, - RADEON_FALLBACKS = 0x0020, - RADEON_VFMT = 0x0040, - RADEON_SHADER = 0x0080, - RADEON_CS = 0x0100, - RADEON_DRI = 0x0200, - RADEON_DMA = 0x0400, - RADEON_SANITY = 0x0800, - RADEON_SYNC = 0x1000, - RADEON_PIXEL = 0x2000, - RADEON_MEMORY = 0x4000, - RADEON_GENERAL = 0x8000 /* Used for errors and warnings */ + RADEON_TEXTURE = 0x00001, + RADEON_STATE = 0x00002, + RADEON_IOCTL = 0x00004, + RADEON_RENDER = 0x00008, + RADEON_SWRENDER = 0x00010, + RADEON_FALLBACKS = 0x00020, + RADEON_VFMT = 0x00040, + RADEON_SHADER = 0x00080, + RADEON_CS = 0x00100, + RADEON_DRI = 0x00200, + RADEON_DMA = 0x00400, + RADEON_SANITY = 0x00800, + RADEON_SYNC = 0x01000, + RADEON_PIXEL = 0x02000, + RADEON_MEMORY = 0x04000, + RADEON_VERTS = 0x08000, + RADEON_GENERAL = 0x10000 /* Used for errors and warnings */ } radeon_debug_type_t; extern radeon_debug_type_t radeon_enabled_debug_types; diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index 386262b126..2eefa3f2b1 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -58,7 +58,7 @@ void radeonEmitVec4(uint32_t *out, const GLvoid * data, int stride, int count) { int i; - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s count %d stride %d out %p data %p\n", __FUNCTION__, count, stride, (void *)out, (void *)data); @@ -76,7 +76,7 @@ void radeonEmitVec8(uint32_t *out, const GLvoid * data, int stride, int count) { int i; - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s count %d stride %d out %p data %p\n", __FUNCTION__, count, stride, (void *)out, (void *)data); @@ -95,7 +95,7 @@ void radeonEmitVec12(uint32_t *out, const GLvoid * data, int stride, int count) { int i; - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s count %d stride %d out %p data %p\n", __FUNCTION__, count, stride, (void *)out, (void *)data); @@ -116,7 +116,7 @@ void radeonEmitVec16(uint32_t *out, const GLvoid * data, int stride, int count) { int i; - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s count %d stride %d out %p data %p\n", __FUNCTION__, count, stride, (void *)out, (void *)data); @@ -179,8 +179,8 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) if (size > rmesa->dma.minimum_size) rmesa->dma.minimum_size = (size + 15) & (~15); - if (RADEON_DEBUG & (DEBUG_IOCTL | DEBUG_DMA)) - fprintf(stderr, "%s\n", __FUNCTION__); + radeon_print(RADEON_DMA, RADEON_NORMAL, "%s size %d minimum_size %d\n", + __FUNCTION__, size, rmesa->dma.minimum_size); /* unmap old reserved bo */ @@ -235,7 +235,7 @@ void radeonAllocDmaRegion(radeonContextPtr rmesa, struct radeon_bo **pbo, int *poffset, int bytes, int alignment) { - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s %d\n", __FUNCTION__, bytes); if (rmesa->dma.flush) @@ -265,7 +265,7 @@ void radeonFreeDmaRegions(radeonContextPtr rmesa) { struct radeon_dma_bo *dma_bo = CALLOC_STRUCT(radeon_dma_bo); struct radeon_dma_bo *temp; - if (RADEON_DEBUG & DEBUG_DMA) + if (RADEON_DEBUG & RADEON_DMA) fprintf(stderr, "%s\n", __FUNCTION__); foreach_s(dma_bo, temp, &rmesa->dma.free) { @@ -293,7 +293,7 @@ void radeonReturnDmaRegion(radeonContextPtr rmesa, int return_bytes) if (is_empty_list(&rmesa->dma.reserved)) return; - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s %d\n", __FUNCTION__, return_bytes); rmesa->dma.current_used -= return_bytes; rmesa->dma.current_vertexptr = rmesa->dma.current_used; @@ -317,7 +317,7 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa) const int expire_at = ++rmesa->dma.free.expire_counter + DMA_BO_FREE_TIME; const int time = rmesa->dma.free.expire_counter; - if (RADEON_DEBUG & DEBUG_DMA) { + if (RADEON_DEBUG & RADEON_DMA) { size_t free = 0, wait = 0, reserved = 0; @@ -399,7 +399,7 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) struct radeon_dma *dma = &rmesa->dma; - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); dma->flush = NULL; @@ -425,7 +425,7 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) { GLuint bytes = vsize * nverts; void *head; - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); if(is_empty_list(&rmesa->dma.reserved) ||rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) { @@ -460,7 +460,7 @@ void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) { radeonContextPtr radeon = RADEON_CONTEXT( ctx ); int i; - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); if (radeon->dma.flush) { diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c index 6f0cc08770..d83b166742 100644 --- a/src/mesa/drivers/dri/radeon/radeon_fbo.c +++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c @@ -40,7 +40,7 @@ #include "radeon_common.h" #include "radeon_mipmap_tree.h" -#define FILE_DEBUG_FLAG DEBUG_TEXTURE +#define FILE_DEBUG_FLAG RADEON_TEXTURE #define DBG(...) do { \ if (RADEON_DEBUG & FILE_DEBUG_FLAG) \ _mesa_printf(__VA_ARGS__); \ diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index 9ced62418b..a0106d00fa 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -200,7 +200,7 @@ void radeonFlushElts( GLcontext *ctx ) uint32_t *cmd = (uint32_t *)(rmesa->radeon.cmdbuf.cs->packets + rmesa->tcl.elt_cmd_start); int dwords = (rmesa->radeon.cmdbuf.cs->section_ndw - rmesa->radeon.cmdbuf.cs->section_cdw); - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); assert( rmesa->radeon.dma.flush == radeonFlushElts ); @@ -236,7 +236,7 @@ void radeonFlushElts( GLcontext *ctx ) END_BATCH(); - if (RADEON_DEBUG & DEBUG_SYNC) { + if (RADEON_DEBUG & RADEON_SYNC) { fprintf(stderr, "%s: Syncing\n", __FUNCTION__); radeonFinish( rmesa->radeon.glCtx ); } @@ -252,7 +252,7 @@ GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa, int align_min_nr; BATCH_LOCALS(&rmesa->radeon); - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive); assert((primitive & RADEON_CP_VC_CNTL_PRIM_WALK_IND)); @@ -296,7 +296,7 @@ GLushort *radeonAllocEltsOpenEnded( r100ContextPtr rmesa, retval = (GLushort *)(rmesa->radeon.cmdbuf.cs->packets + rmesa->tcl.elt_cmd_offset); - if (RADEON_DEBUG & DEBUG_PRIMS) + if (RADEON_DEBUG & RADEON_RENDER) fprintf(stderr, "%s: header prim %x \n", __FUNCTION__, primitive); @@ -318,7 +318,7 @@ void radeonEmitVertexAOS( r100ContextPtr rmesa, #else BATCH_LOCALS(&rmesa->radeon); - if (RADEON_DEBUG & (DEBUG_PRIMS|DEBUG_IOCTL)) + if (RADEON_DEBUG & (RADEON_PRIMS|DEBUG_IOCTL)) fprintf(stderr, "%s: vertex_size 0x%x offset 0x%x \n", __FUNCTION__, vertex_size, offset); @@ -350,7 +350,7 @@ void radeonEmitAOS( r100ContextPtr rmesa, int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2; int i; - if (RADEON_DEBUG & DEBUG_IOCTL) + if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); BEGIN_BATCH(sz+2+(nr * 2)); @@ -575,7 +575,7 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask ) GLuint color_mask = 0; GLuint orig_mask = mask; - if ( RADEON_DEBUG & DEBUG_IOCTL ) { + if ( RADEON_DEBUG & RADEON_IOCTL ) { fprintf( stderr, "radeonClear\n"); } @@ -611,7 +611,7 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask ) } if ( mask ) { - if (RADEON_DEBUG & DEBUG_FALLBACKS) + if (RADEON_DEBUG & RADEON_FALLBACKS) fprintf(stderr, "%s: swrast clear, mask: %x\n", __FUNCTION__, mask); _swrast_Clear( ctx, mask ); } diff --git a/src/mesa/drivers/dri/radeon/radeon_lighting.c b/src/mesa/drivers/dri/radeon/radeon_lighting.c index ac3b94e4a6..ba444f2b10 100644 --- a/src/mesa/drivers/dri/radeon/radeon_lighting.c +++ b/src/mesa/drivers/dri/radeon/radeon_lighting.c @@ -195,7 +195,7 @@ void radeonUpdateMaterial( GLcontext *ctx ) if (ctx->Light.ColorMaterialEnabled) mask &= ~ctx->Light.ColorMaterialBitmask; - if (RADEON_DEBUG & DEBUG_STATE) + if (RADEON_DEBUG & RADEON_STATE) fprintf(stderr, "%s\n", __FUNCTION__); @@ -234,7 +234,7 @@ void radeonUpdateMaterial( GLcontext *ctx ) check_twoside_fallback( ctx ); update_global_ambient( ctx ); } - else if (RADEON_DEBUG & (DEBUG_PRIMS|DEBUG_STATE)) + else if (RADEON_DEBUG & (RADEON_PRIMS|DEBUG_STATE)) fprintf(stderr, "%s: Elided noop material call\n", __FUNCTION__); } @@ -624,7 +624,7 @@ static void radeonLightingSpaceChange( GLcontext *ctx ) GLboolean tmp; RADEON_STATECHANGE( rmesa, tcl ); - if (RADEON_DEBUG & DEBUG_STATE) + if (RADEON_DEBUG & RADEON_STATE) fprintf(stderr, "%s %d\n", __FUNCTION__, ctx->_NeedEyeCoords); if (ctx->_NeedEyeCoords) diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.c b/src/mesa/drivers/dri/radeon/radeon_lock.c index 6294b7e42b..02de8e5fd1 100644 --- a/src/mesa/drivers/dri/radeon/radeon_lock.c +++ b/src/mesa/drivers/dri/radeon/radeon_lock.c @@ -131,7 +131,7 @@ void radeon_lock_hardware(radeonContextPtr radeon if (ATOMIC_INC_AND_FETCH(radeon->dri.hwLockCount) > 1) { #ifndef NDEBUG - if ( RADEON_DEBUG & DEBUG_SANITY ) + if ( RADEON_DEBUG & RADEON_SANITY ) fprintf(stderr, "*** %d times of recursive call to %s ***\n" "Original call was from %s (file: %s line: %d)\n" "Now call is coming from %s (file: %s line: %d)\n" diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 7c6ea0530e..08e1c5d00d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -56,7 +56,7 @@ static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, int size = 1; radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s count %d stride %d\n", __FUNCTION__, count, stride); @@ -87,7 +87,7 @@ static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, static void emit_s0_vec(uint32_t *out, GLvoid *data, int stride, int count) { int i; - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s count %d stride %d\n", __FUNCTION__, count, stride); @@ -103,7 +103,7 @@ static void emit_stq_vec(uint32_t *out, GLvoid *data, int stride, int count) { int i; - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s count %d stride %d\n", __FUNCTION__, count, stride); @@ -126,7 +126,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, int emitsize; uint32_t *out; - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size); switch (size) { @@ -188,7 +188,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) GLuint vtx, unit; #if 0 - if (RADEON_DEBUG & DEBUG_VERTS) + if (RADEON_DEBUG & RADEON_VERTS) _tnl_print_vert_flags( __FUNCTION__, inputs ); #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h index 034cda8a65..515783135d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h @@ -54,8 +54,7 @@ static void TAG(emit)( GLcontext *ctx, union emit_union *v = (union emit_union *)dest; - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s\n", __FUNCTION__); + radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __FUNCTION__); coord = (GLuint (*)[4])VB->ObjPtr->data; coord_stride = VB->ObjPtr->stride; diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c index eba9f5857f..d9f17cfc58 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c @@ -121,7 +121,7 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree lvl->faces[face].offset = *curOffset; *curOffset += lvl->size; - if (RADEON_DEBUG & DEBUG_TEXTURE) + if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, "level %d, face %d: rs:%d %dx%d at %d\n", level, face, lvl->rowstride, lvl->width, lvl->height, lvl->faces[face].offset); diff --git a/src/mesa/drivers/dri/radeon/radeon_sanity.c b/src/mesa/drivers/dri/radeon/radeon_sanity.c index bbed838b59..1ab570f507 100644 --- a/src/mesa/drivers/dri/radeon/radeon_sanity.c +++ b/src/mesa/drivers/dri/radeon/radeon_sanity.c @@ -44,11 +44,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define MORE_VERBOSE 1 #if MORE_VERBOSE -#define VERBOSE (RADEON_DEBUG & DEBUG_VERBOSE) +#define VERBOSE (RADEON_DEBUG & RADEON_VERBOSE) #define NORMAL (1) #else #define VERBOSE 0 -#define NORMAL (RADEON_DEBUG & DEBUG_VERBOSE) +#define NORMAL (RADEON_DEBUG & RADEON_VERBOSE) #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index 3d06c95a18..daee3b85d9 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -220,30 +220,6 @@ static const GLuint __driNConfigOptions = 17; extern const struct dri_extension gl_20_extension[]; -#ifndef RADEON_DEBUG - -static const struct dri_debug_control debug_control[] = { - {"fall", DEBUG_FALLBACKS}, - {"tex", DEBUG_TEXTURE}, - {"ioctl", DEBUG_IOCTL}, - {"prim", DEBUG_PRIMS}, - {"vert", DEBUG_VERTS}, - {"state", DEBUG_STATE}, - {"code", DEBUG_CODEGEN}, - {"vfmt", DEBUG_VFMT}, - {"vtxf", DEBUG_VFMT}, - {"verb", DEBUG_VERBOSE}, - {"dri", DEBUG_DRI}, - {"dma", DEBUG_DMA}, - {"san", DEBUG_SANITY}, - {"sync", DEBUG_SYNC}, - {"pix", DEBUG_PIXEL}, - {"mem", DEBUG_MEMORY}, - {"allmsg", ~DEBUG_SYNC}, /* avoid the term "sync" because the parser uses strstr */ - {NULL, 0} -}; -#endif /* RADEON_DEBUG */ - #endif /* RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) */ extern const struct dri_extension card_extensions[]; @@ -966,9 +942,8 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv ) return NULL; } -#if DO_DEBUG && RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) - RADEON_DEBUG = driParseDebugString(getenv("RADEON_DEBUG"), debug_control); -#endif + radeon_init_debug(); + /* parse information in __driConfigOptions */ driParseOptionInfo (&screen->optionCache, __driConfigOptions, __driNConfigOptions); @@ -1301,9 +1276,7 @@ radeonCreateScreen2(__DRIscreenPrivate *sPriv) return NULL; } -#if DO_DEBUG && RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) - RADEON_DEBUG = driParseDebugString(getenv("RADEON_DEBUG"), debug_control); -#endif + radeon_init_debug(); /* parse information in __driConfigOptions */ driParseOptionInfo (&screen->optionCache, diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c index e03551d666..c8cc7f9bcf 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.c +++ b/src/mesa/drivers/dri/radeon/radeon_state.c @@ -814,7 +814,7 @@ void radeonUpdateMaterial( GLcontext *ctx ) if (ctx->Light.ColorMaterialEnabled) mask &= ~ctx->Light.ColorMaterialBitmask; - if (RADEON_DEBUG & DEBUG_STATE) + if (RADEON_DEBUG & RADEON_STATE) fprintf(stderr, "%s\n", __FUNCTION__); @@ -1548,7 +1548,7 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state ) r100ContextPtr rmesa = R100_CONTEXT(ctx); GLuint p, flag; - if ( RADEON_DEBUG & DEBUG_STATE ) + if ( RADEON_DEBUG & RADEON_STATE ) fprintf( stderr, "%s( %s = %s )\n", __FUNCTION__, _mesa_lookup_enum_by_nr( cap ), state ? "GL_TRUE" : "GL_FALSE" ); @@ -1842,7 +1842,7 @@ static void radeonLightingSpaceChange( GLcontext *ctx ) GLboolean tmp; RADEON_STATECHANGE( rmesa, tcl ); - if (RADEON_DEBUG & DEBUG_STATE) + if (RADEON_DEBUG & RADEON_STATE) fprintf(stderr, "%s %d BEFORE %x\n", __FUNCTION__, ctx->_NeedEyeCoords, rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]); @@ -1857,7 +1857,7 @@ static void radeonLightingSpaceChange( GLcontext *ctx ) rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL] &= ~RADEON_RESCALE_NORMALS; } - if (RADEON_DEBUG & DEBUG_STATE) + if (RADEON_DEBUG & RADEON_STATE) fprintf(stderr, "%s %d AFTER %x\n", __FUNCTION__, ctx->_NeedEyeCoords, rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL]); } diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 32df569257..e61f59eaea 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -51,6 +51,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_state.h" #include "radeon_swtcl.h" #include "radeon_tcl.h" +#include "radeon_debug.h" /* R100: xyzw, c0, c1/fog, stq[0..2] = 4+1+1+3*3 = 15 right? */ @@ -214,9 +215,8 @@ static void radeonSetVertexFormat( GLcontext *ctx ) NULL, 0 ); rmesa->radeon.swtcl.vertex_size /= 4; RENDERINPUTS_COPY( rmesa->radeon.tnl_index_bitset, index_bitset ); - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf( stderr, "%s: vertex_size= %d floats\n", - __FUNCTION__, rmesa->radeon.swtcl.vertex_size); + radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, + "%s: vertex_size= %d floats\n", __FUNCTION__, rmesa->radeon.swtcl.vertex_size); } } @@ -420,8 +420,8 @@ static GLboolean radeon_run_render( GLcontext *ctx, if (!length) continue; - if (RADEON_DEBUG & DEBUG_PRIMS) - fprintf(stderr, "radeon_render.c: prim %s %d..%d\n", + radeon_print(RADEON_SWRENDER, RADEON_NORMAL, + "radeon_render.c: prim %s %d..%d\n", _mesa_lookup_enum_by_nr(prim & PRIM_MODE_MASK), start, start+length); @@ -784,7 +784,7 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) TCL_FALLBACK( ctx, RADEON_TCL_FALLBACK_RASTER, GL_TRUE ); _swsetup_Wakeup( ctx ); rmesa->radeon.swtcl.RenderIndex = ~0; - if (RADEON_DEBUG & DEBUG_FALLBACKS) { + if (RADEON_DEBUG & RADEON_FALLBACKS) { fprintf(stderr, "Radeon begin rasterization fallback: 0x%x %s\n", bit, getFallbackString(bit)); } @@ -815,7 +815,7 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) radeonChooseVertexState( ctx ); radeonChooseRenderState( ctx ); } - if (RADEON_DEBUG & DEBUG_FALLBACKS) { + if (RADEON_DEBUG & RADEON_FALLBACKS) { fprintf(stderr, "Radeon end rasterization fallback: 0x%x %s\n", bit, getFallbackString(bit)); } diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index 2404f28450..b334ea05e5 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -587,7 +587,7 @@ static void transition_to_hwtnl( GLcontext *ctx ) // radeonReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, // __FUNCTION__ ); - if (RADEON_DEBUG & DEBUG_FALLBACKS) + if (RADEON_DEBUG & RADEON_FALLBACKS) fprintf(stderr, "Radeon end tcl fallback\n"); } @@ -624,7 +624,7 @@ void radeonTclFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) if (mode) { rmesa->radeon.TclFallback |= bit; if (oldfallback == 0) { - if (RADEON_DEBUG & DEBUG_FALLBACKS) + if (RADEON_DEBUG & RADEON_FALLBACKS) fprintf(stderr, "Radeon begin tcl fallback %s\n", getFallbackString( bit )); transition_to_swtnl( ctx ); @@ -633,7 +633,7 @@ void radeonTclFallback( GLcontext *ctx, GLuint bit, GLboolean mode ) else { rmesa->radeon.TclFallback &= ~bit; if (oldfallback == bit) { - if (RADEON_DEBUG & DEBUG_FALLBACKS) + if (RADEON_DEBUG & RADEON_FALLBACKS) fprintf(stderr, "Radeon end tcl fallback %s\n", getFallbackString( bit )); transition_to_hwtnl( ctx ); diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c index 2549d5cb5c..99865fff27 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex.c @@ -263,7 +263,7 @@ static void radeonTexEnv( GLcontext *ctx, GLenum target, GLuint unit = ctx->Texture.CurrentUnit; struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - if ( RADEON_DEBUG & DEBUG_STATE ) { + if ( RADEON_DEBUG & RADEON_STATE ) { fprintf( stderr, "%s( %s )\n", __FUNCTION__, _mesa_lookup_enum_by_nr( pname ) ); } @@ -325,10 +325,8 @@ static void radeonTexParameter( GLcontext *ctx, GLenum target, { radeonTexObj* t = radeon_tex_obj(texObj); - if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) { - fprintf( stderr, "%s( %s )\n", __FUNCTION__, + radeon_print(RADEON_TEXTURE, RADEON_VERBOSE, "%s( %s )\n", __FUNCTION__, _mesa_lookup_enum_by_nr( pname ) ); - } switch ( pname ) { case GL_TEXTURE_MIN_FILTER: @@ -376,10 +374,9 @@ static void radeonDeleteTexture( GLcontext *ctx, radeonTexObj* t = radeon_tex_obj(texObj); int i; - if ( RADEON_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) { - fprintf( stderr, "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj, + radeon_print(RADEON_TEXTURE, RADEON_NORMAL, + "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj, _mesa_lookup_enum_by_nr( texObj->Target ) ); - } if ( rmesa ) { radeon_firevertices(&rmesa->radeon); diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index c29105d7b8..9d252aa74c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -277,7 +277,7 @@ static GLboolean radeonUpdateTextureEnv( GLcontext *ctx, int unit ) assert( (texUnit->_ReallyEnabled == 0) || (texUnit->_Current != NULL) ); - if ( RADEON_DEBUG & DEBUG_TEXTURE ) { + if ( RADEON_DEBUG & RADEON_TEXTURE ) { fprintf( stderr, "%s( %p, %d )\n", __FUNCTION__, (void *)ctx, unit ); } @@ -933,7 +933,7 @@ static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit ) (texUnit->GenS.Mode != texUnit->GenQ.Mode)) ) { /* Mixed modes, fallback: */ - if (RADEON_DEBUG & DEBUG_FALLBACKS) + if (RADEON_DEBUG & RADEON_FALLBACKS) fprintf(stderr, "fallback mixed texgen\n"); return GL_FALSE; } @@ -941,7 +941,7 @@ static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit ) } else { /* some texgen mode not including both S and T bits */ - if (RADEON_DEBUG & DEBUG_FALLBACKS) + if (RADEON_DEBUG & RADEON_FALLBACKS) fprintf(stderr, "fallback mixed texgen/nontexgen\n"); return GL_FALSE; } @@ -991,7 +991,7 @@ static GLboolean radeon_validate_texgen( GLcontext *ctx, GLuint unit ) default: /* Unsupported mode, fallback: */ - if (RADEON_DEBUG & DEBUG_FALLBACKS) + if (RADEON_DEBUG & RADEON_FALLBACKS) fprintf(stderr, "fallback GL_SPHERE_MAP\n"); return GL_FALSE; } diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c index fa16f44c18..181b23c396 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texture.c +++ b/src/mesa/drivers/dri/radeon/radeon_texture.c @@ -936,7 +936,7 @@ int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *t if (t->validated || t->image_override) return GL_TRUE; - if (RADEON_DEBUG & DEBUG_TEXTURE) + if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, "%s: Validating texture %p now\n", __FUNCTION__, texObj); if (baseimage->base.Border > 0) @@ -964,7 +964,7 @@ int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *t } if (!t->mt) { - if (RADEON_DEBUG & DEBUG_TEXTURE) + if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, " Allocate new miptree\n"); radeon_try_alloc_miptree(rmesa, t, &baseimage->base, 0, texObj->BaseLevel); if (!t->mt) { @@ -977,16 +977,16 @@ int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *t for(face = 0; face < t->mt->faces; ++face) { for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) { radeon_texture_image *image = get_radeon_texture_image(texObj->Image[face][level]); - if (RADEON_DEBUG & DEBUG_TEXTURE) + if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, " face %i, level %i... %p vs %p ", face, level, t->mt, image->mt); if (t->mt == image->mt) { - if (RADEON_DEBUG & DEBUG_TEXTURE) + if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, "OK\n"); continue; } - if (RADEON_DEBUG & DEBUG_TEXTURE) + if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, "migrating\n"); migrate_image_to_miptree(t->mt, image, face, level); } -- cgit v1.2.3