diff options
Diffstat (limited to 'src/mesa/drivers/dri/radeon/radeon_state_init.c')
| -rw-r--r-- | src/mesa/drivers/dri/radeon/radeon_state_init.c | 521 |
1 files changed, 420 insertions, 101 deletions
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index 57dc380050..fc42318017 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -38,39 +38,156 @@ #include "swrast_setup/swrast_setup.h" #include "radeon_context.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 */ -void radeonPrintDirty( radeonContextPtr rmesa, const char *msg ) +void radeonPrintDirty( r100ContextPtr rmesa, const char *msg ) { struct radeon_state_atom *l; fprintf(stderr, msg); fprintf(stderr, ": "); - foreach(l, &rmesa->hw.atomlist) { - if (l->dirty || rmesa->hw.all_dirty) + foreach(l, &rmesa->radeon.hw.atomlist) { + if (l->dirty || rmesa->radeon.hw.all_dirty) fprintf(stderr, "%s, ", l->name); } 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; } @@ -96,17 +213,17 @@ static int cmdscl( int offset, int stride, int count ) return h.i; } -#define CHECK( NM, FLAG ) \ -static GLboolean check_##NM( GLcontext *ctx ) \ -{ \ - return FLAG; \ +#define CHECK( NM, FLAG ) \ +static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \ +{ \ + return FLAG ? atom->cmd_size : 0; \ } #define TCL_CHECK( NM, FLAG ) \ -static GLboolean check_##NM( GLcontext *ctx ) \ +static int check_##NM( GLcontext *ctx, struct radeon_state_atom *atom ) \ { \ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \ - return !rmesa->TclFallback && (FLAG); \ + r100ContextPtr rmesa = R100_CONTEXT(ctx); \ + return (!rmesa->radeon.TclFallback && (FLAG)) ? atom->cmd_size : 0; \ } @@ -146,17 +263,207 @@ 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 && t->mt && !t->image_override) + dwords += 2; + BEGIN_BATCH_NO_AUTOSTATE(dwords); + OUT_BATCH_TABLE(atom->cmd, 3); + if (t && t->mt && !t->image_override) { + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + RADEON_GEM_DOMAIN_VRAM, 0, 0); + } else { + /* workaround for old CS mechanism */ + OUT_BATCH(r100->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]); + // OUT_BATCH(r100->radeon.radeonScreen); + } + + OUT_BATCH_TABLE((atom->cmd+4), 5); + END_BATCH(); +} /* Initialize the context's hardware state. */ -void radeonInitState( radeonContextPtr rmesa ) +void radeonInitState( r100ContextPtr rmesa ) { - GLcontext *ctx = rmesa->glCtx; + GLcontext *ctx = rmesa->radeon.glCtx; GLuint color_fmt, depth_fmt, i; GLint drawPitch, drawOffset; - switch ( rmesa->radeonScreen->cpp ) { + switch ( rmesa->radeon.radeonScreen->cpp ) { case 2: color_fmt = RADEON_COLOR_FORMAT_RGB565; break; @@ -168,20 +475,20 @@ void radeonInitState( radeonContextPtr rmesa ) exit( -1 ); } - rmesa->state.color.clear = 0x00000000; + rmesa->radeon.state.color.clear = 0x00000000; switch ( ctx->Visual.depthBits ) { case 16: - rmesa->state.depth.clear = 0x0000ffff; - rmesa->state.depth.scale = 1.0 / (GLfloat)0xffff; + rmesa->radeon.state.depth.clear = 0x0000ffff; + rmesa->radeon.state.depth.scale = 1.0 / (GLfloat)0xffff; depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z; - rmesa->state.stencil.clear = 0x00000000; + rmesa->radeon.state.stencil.clear = 0x00000000; break; case 24: - rmesa->state.depth.clear = 0x00ffffff; - rmesa->state.depth.scale = 1.0 / (GLfloat)0xffffff; + rmesa->radeon.state.depth.clear = 0x00ffffff; + rmesa->radeon.state.depth.scale = 1.0 / (GLfloat)0xffffff; depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z; - rmesa->state.stencil.clear = 0xffff0000; + rmesa->radeon.state.stencil.clear = 0xffff0000; break; default: fprintf( stderr, "Error: Unsupported depth %d... exiting\n", @@ -190,37 +497,44 @@ void radeonInitState( radeonContextPtr rmesa ) } /* Only have hw stencil when depth buffer is 24 bits deep */ - rmesa->state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 && + rmesa->radeon.state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 && ctx->Visual.depthBits == 24 ); - rmesa->Fallback = 0; + rmesa->radeon.Fallback = 0; - if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) { - drawOffset = rmesa->radeonScreen->backOffset; - drawPitch = rmesa->radeonScreen->backPitch; + if ( ctx->Visual.doubleBufferMode && rmesa->radeon.sarea->pfCurrentPage == 0 ) { + drawOffset = rmesa->radeon.radeonScreen->backOffset; + drawPitch = rmesa->radeon.radeonScreen->backPitch; } else { - drawOffset = rmesa->radeonScreen->frontOffset; - drawPitch = rmesa->radeonScreen->frontPitch; + drawOffset = rmesa->radeon.radeonScreen->frontOffset; + drawPitch = rmesa->radeon.radeonScreen->frontPitch; } - rmesa->hw.max_state_size = 0; + rmesa->radeon.hw.max_state_size = 0; -#define ALLOC_STATE( ATOM, CHK, SZ, NM, FLAG ) \ +#define ALLOC_STATE_IDX( ATOM, CHK, SZ, NM, FLAG, IDX ) \ do { \ rmesa->hw.ATOM.cmd_size = SZ; \ - rmesa->hw.ATOM.cmd = (int *)CALLOC(SZ * sizeof(int)); \ - rmesa->hw.ATOM.lastcmd = (int *)CALLOC(SZ * sizeof(int)); \ - rmesa->hw.ATOM.name = NM; \ + rmesa->hw.ATOM.cmd = (GLuint *)CALLOC(SZ * sizeof(int)); \ + rmesa->hw.ATOM.lastcmd = (GLuint *)CALLOC(SZ * sizeof(int)); \ + rmesa->hw.ATOM.name = NM; \ rmesa->hw.ATOM.is_tcl = FLAG; \ rmesa->hw.ATOM.check = check_##CHK; \ - rmesa->hw.ATOM.dirty = GL_TRUE; \ - rmesa->hw.max_state_size += SZ * sizeof(int); \ + rmesa->hw.ATOM.dirty = GL_TRUE; \ + rmesa->hw.ATOM.idx = IDX; \ + rmesa->radeon.hw.max_state_size += SZ * sizeof(int); \ } while (0) - - + +#define ALLOC_STATE( ATOM, CHK, SZ, NM, FLAG ) \ + ALLOC_STATE_IDX(ATOM, CHK, SZ, NM, FLAG, 0) + /* 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 ); @@ -233,20 +547,23 @@ void radeonInitState( radeonContextPtr rmesa ) ALLOC_STATE( fog, fog, FOG_STATE_SIZE, "FOG/fog", 1 ); ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 1 ); ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 1 ); - ALLOC_STATE( 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 ); - if (rmesa->radeonScreen->drmSupportsCubeMapsR100) + ALLOC_STATE_IDX( tex[0], tex0, TEX_STATE_SIZE, "TEX/tex-0", 0, 0); + ALLOC_STATE_IDX( tex[1], tex1, TEX_STATE_SIZE, "TEX/tex-1", 0, 1); + ALLOC_STATE_IDX( tex[2], tex2, TEX_STATE_SIZE, "TEX/tex-2", 0, 2 ); + + 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 ); - ALLOC_STATE( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0 ); - ALLOC_STATE( cube[2], cube2, CUBE_STATE_SIZE, "CUBE/cube-2", 0 ); + ALLOC_STATE_IDX( cube[0], cube0, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 ); + ALLOC_STATE_IDX( cube[1], cube1, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 ); + ALLOC_STATE_IDX( cube[2], cube2, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 ); } else { - ALLOC_STATE( cube[0], never, CUBE_STATE_SIZE, "CUBE/cube-0", 0 ); - ALLOC_STATE( cube[1], never, CUBE_STATE_SIZE, "CUBE/cube-1", 0 ); - ALLOC_STATE( cube[2], never, CUBE_STATE_SIZE, "CUBE/cube-2", 0 ); + ALLOC_STATE_IDX( cube[0], never, CUBE_STATE_SIZE, "CUBE/cube-0", 0, 0 ); + ALLOC_STATE_IDX( cube[1], never, CUBE_STATE_SIZE, "CUBE/cube-1", 0, 1 ); + ALLOC_STATE_IDX( cube[2], never, CUBE_STATE_SIZE, "CUBE/cube-2", 0, 2 ); } ALLOC_STATE( mat[0], tcl, MAT_STATE_SIZE, "MAT/modelproject", 1 ); ALLOC_STATE( mat[1], tcl_eyespace_or_fog, MAT_STATE_SIZE, "MAT/modelview", 1 ); @@ -268,43 +585,43 @@ void radeonInitState( radeonContextPtr rmesa ) ALLOC_STATE( lit[5], tcl_lit5, LIT_STATE_SIZE, "LIT/light-5", 1 ); ALLOC_STATE( lit[6], tcl_lit6, LIT_STATE_SIZE, "LIT/light-6", 1 ); ALLOC_STATE( lit[7], tcl_lit7, LIT_STATE_SIZE, "LIT/light-7", 1 ); - ALLOC_STATE( txr[0], txr0, TXR_STATE_SIZE, "TXR/txr-0", 0 ); - ALLOC_STATE( txr[1], txr1, TXR_STATE_SIZE, "TXR/txr-1", 0 ); - ALLOC_STATE( txr[2], txr2, TXR_STATE_SIZE, "TXR/txr-2", 0 ); + ALLOC_STATE_IDX( txr[0], txr0, TXR_STATE_SIZE, "TXR/txr-0", 0, 0 ); + ALLOC_STATE_IDX( txr[1], txr1, TXR_STATE_SIZE, "TXR/txr-1", 0, 1 ); + ALLOC_STATE_IDX( txr[2], txr2, TXR_STATE_SIZE, "TXR/txr-2", 0, 2 ); radeonSetUpAtomList( 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] = @@ -353,10 +670,10 @@ void radeonInitState( radeonContextPtr rmesa ) RADEON_DST_BLEND_GL_ZERO ); rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHOFFSET] = - rmesa->radeonScreen->depthOffset + rmesa->radeonScreen->fbLocation; + rmesa->radeon.radeonScreen->depthOffset + rmesa->radeon.radeonScreen->fbLocation; rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] = - ((rmesa->radeonScreen->depthPitch & + ((rmesa->radeon.radeonScreen->depthPitch & RADEON_DEPTHPITCH_MASK) | RADEON_DEPTH_ENDIAN_NO_SWAP); @@ -374,7 +691,7 @@ void radeonInitState( radeonContextPtr rmesa ) if (rmesa->using_hyperz) { rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_COMPRESSION_ENABLE | RADEON_Z_DECOMPRESSION_ENABLE; - if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) { + if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) { /* works for q3, but slight rendering errors with glxgears ? */ /* rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] |= RADEON_Z_HIERARCHY_ENABLE;*/ /* need this otherwise get lots of lockups with q3 ??? */ @@ -389,7 +706,7 @@ void radeonInitState( radeonContextPtr rmesa ) color_fmt | RADEON_ZBLOCK16); - switch ( driQueryOptioni( &rmesa->optionCache, "dither_mode" ) ) { + switch ( driQueryOptioni( &rmesa->radeon.optionCache, "dither_mode" ) ) { case DRI_CONF_DITHER_XERRORDIFFRESET: rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_DITHER_INIT; break; @@ -397,19 +714,19 @@ void radeonInitState( radeonContextPtr rmesa ) rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_SCALE_DITHER_ENABLE; break; } - if ( driQueryOptioni( &rmesa->optionCache, "round_mode" ) == + if ( driQueryOptioni( &rmesa->radeon.optionCache, "round_mode" ) == DRI_CONF_ROUND_ROUND ) - rmesa->state.color.roundEnable = RADEON_ROUND_ENABLE; + rmesa->radeon.state.color.roundEnable = RADEON_ROUND_ENABLE; else - rmesa->state.color.roundEnable = 0; - if ( driQueryOptioni (&rmesa->optionCache, "color_reduction" ) == + rmesa->radeon.state.color.roundEnable = 0; + if ( driQueryOptioni (&rmesa->radeon.optionCache, "color_reduction" ) == DRI_CONF_COLOR_REDUCTION_DITHER ) rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= RADEON_DITHER_ENABLE; else - rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->state.color.roundEnable; + rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |= rmesa->radeon.state.color.roundEnable; rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = ((drawOffset + - rmesa->radeonScreen->fbLocation) + rmesa->radeon.radeonScreen->fbLocation) & RADEON_COLOROFFSET_MASK); rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((drawPitch & @@ -418,7 +735,7 @@ void radeonInitState( radeonContextPtr rmesa ) /* (fixed size) sarea is initialized to zero afaics so can omit version check. Phew! */ - if (rmesa->sarea->tiling_enabled) { + if (rmesa->radeon.sarea->tiling_enabled) { rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= RADEON_COLOR_TILE_ENABLE; } @@ -444,7 +761,7 @@ void radeonInitState( radeonContextPtr rmesa ) RADEON_VC_NO_SWAP; #endif - if (!(rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) { + if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) { rmesa->hw.set.cmd[SET_SE_CNTL_STATUS] |= RADEON_TCL_BYPASS; } @@ -491,8 +808,8 @@ void radeonInitState( radeonContextPtr rmesa ) (2 << RADEON_TXFORMAT_HEIGHT_SHIFT)); /* Initialize the texture offset to the start of the card texture heap */ - rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET] = - rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; + // rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET] = + // rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.tex[i].cmd[TEX_PP_BORDER_COLOR] = 0; rmesa->hw.tex[i].cmd[TEX_PP_TXCBLEND] = @@ -513,15 +830,15 @@ void radeonInitState( radeonContextPtr rmesa ) rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_FACES] = 0; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_0] = - rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_1] = - rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_2] = - rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_3] = - rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_4] = - rmesa->radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; + rmesa->radeon.radeonScreen->texOffset[RADEON_LOCAL_TEX_HEAP]; } /* Can only add ST1 at the time of doing some multitex but can keep @@ -613,5 +930,7 @@ void radeonInitState( radeonContextPtr rmesa ) rmesa->hw.eye.cmd[EYE_Z] = IEEE_ONE; rmesa->hw.eye.cmd[EYE_RESCALE_FACTOR] = IEEE_ONE; - rmesa->hw.all_dirty = GL_TRUE; + rmesa->radeon.hw.all_dirty = GL_TRUE; + + rcommonInitCmdBuf(&rmesa->radeon); } |