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_maos_verts.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_maos_verts.c') diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 126d0727c6..2abf644ef2 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -310,7 +310,7 @@ static void init_tcl_verts( void ) void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) { - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + r100ContextPtr rmesa = R100_CONTEXT(ctx); struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; GLuint req = 0; GLuint unit; @@ -437,7 +437,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) { - radeonContextPtr rmesa = RADEON_CONTEXT( ctx ); + r100ContextPtr rmesa = R100_CONTEXT( ctx ); #if 0 if (RADEON_DEBUG & DEBUG_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_maos_verts.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_maos_verts.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 f3a1f321cfa275eba6186626ee691e8bc9ecfe8c Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Sat, 14 Feb 2009 12:01:54 +1000 Subject: radeon: check for valid bo in release arrays --- src/mesa/drivers/dri/radeon/radeon_maos_verts.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src/mesa/drivers/dri/radeon/radeon_maos_verts.c') diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index bb90a7b111..d468a97200 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -438,7 +438,9 @@ void radeonReleaseArrays( GLcontext *ctx, GLuint newinputs ) int i; for (i = 0; i < rmesa->tcl.nr_aos_components; i++) { - radeon_bo_unref(rmesa->tcl.aos[i].bo); - rmesa->tcl.aos[i].bo = NULL; + if (rmesa->tcl.aos[i].bo) { + radeon_bo_unref(rmesa->tcl.aos[i].bo); + rmesa->tcl.aos[i].bo = NULL; + } } } -- 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_maos_verts.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