X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fdrivers%2Fdri%2Fr300%2Fr300_swtcl.c;h=ef65fbb127a683c86926ae068e0041d4fff0366b;hb=e45213d89bf26c68c9f4c9074eaec9ab3311de7d;hp=3b4e92cd042aff0102df8fa77c87896f30af8fbd;hpb=ad8abf71472ac7d8f25764e7a235ce97cf2fa700;p=mesa.git diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index 3b4e92cd042..ef65fbb127a 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -34,13 +34,14 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. -#include "glheader.h" -#include "mtypes.h" -#include "colormac.h" -#include "enums.h" -#include "image.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/mtypes.h" +#include "main/colormac.h" +#include "main/enums.h" +#include "main/image.h" +#include "main/imports.h" +#include "main/light.h" +#include "main/macros.h" #include "swrast/s_context.h" #include "swrast/s_fog.h" @@ -55,18 +56,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_state.h" #include "r300_ioctl.h" #include "r300_emit.h" -#include "r300_mem.h" -#define R300_NEWPRIM( rmesa ) \ - do { \ - if ( rmesa->dma.flush ) \ - rmesa->dma.flush( rmesa ); \ - } while (0) +static void flush_last_swtcl_prim( GLcontext *ctx); -static void flush_last_swtcl_prim( r300ContextPtr rmesa ); - -void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, GLuint offset); +void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, struct radeon_bo *bo, GLuint offset); void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr); #define EMIT_ATTR( ATTR, STYLE ) \ do { \ @@ -83,362 +77,209 @@ do { \ rmesa->swtcl.vertex_attr_count++; \ } while (0) -static GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr, - int *inputs, GLint * tab, GLuint nr) +static void r300SetVertexFormat( GLcontext *ctx ) { - GLuint i, dw; - - /* type, inputs, stop bit, size */ - for (i = 0; i + 1 < nr; i += 2) { - dw = (inputs[tab[i]] << 8) | 0x3; - dw |= ((inputs[tab[i + 1]] << 8) | 0x3) << 16; - if (i + 2 == nr) { - dw |= (R300_VAP_INPUT_ROUTE_END << 16); - } - dst[i >> 1] = dw; + r300ContextPtr rmesa = R300_CONTEXT( ctx ); + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *VB = &tnl->vb; + DECLARE_RENDERINPUTS(index_bitset); + GLuint InputsRead = 0, OutputsWritten = 0; + int vap_fmt_0 = 0; + int offset = 0; + int vte = 0; + GLint inputs[VERT_ATTRIB_MAX]; + GLint tab[VERT_ATTRIB_MAX]; + int swizzle[VERT_ATTRIB_MAX][4]; + GLuint i, nr; + GLuint sz, vap_fmt_1 = 0; + + DECLARE_RENDERINPUTS(render_inputs_bitset); + RENDERINPUTS_COPY(render_inputs_bitset, tnl->render_inputs_bitset); + RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); + RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset); + + vte = rmesa->hw.vte.cmd[1]; + vte &= ~(R300_VTX_XY_FMT | R300_VTX_Z_FMT | R300_VTX_W0_FMT); + /* Important: + */ + if ( VB->NdcPtr != NULL ) { + VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr; + vte |= R300_VTX_XY_FMT | R300_VTX_Z_FMT; } - - if (nr & 1) { - dw = (inputs[tab[nr - 1]] << 8) | 0x3; - dw |= R300_VAP_INPUT_ROUTE_END; - dst[nr >> 1] = dw; + else { + VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr; + vte |= R300_VTX_W0_FMT; } - return (nr + 1) >> 1; -} - -static GLuint r300VAPInputRoute1Swizzle(int swizzle[4]) -{ - return (swizzle[0] << R300_INPUT_ROUTE_X_SHIFT) | - (swizzle[1] << R300_INPUT_ROUTE_Y_SHIFT) | - (swizzle[2] << R300_INPUT_ROUTE_Z_SHIFT) | - (swizzle[3] << R300_INPUT_ROUTE_W_SHIFT); -} - -static GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr) -{ - GLuint i; + assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL ); + rmesa->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. + */ + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POS)) { + sz = VB->AttribPtr[VERT_ATTRIB_POS]->size; + InputsRead |= 1 << VERT_ATTRIB_POS; + OutputsWritten |= 1 << VERT_RESULT_HPOS; + EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_1F + sz - 1 ); + offset = sz; + } else { + offset = 4; + EMIT_PAD(4 * sizeof(float)); + } - for (i = 0; i + 1 < nr; i += 2) { - dst[i >> 1] = r300VAPInputRoute1Swizzle(swizzle[i]) | R300_INPUT_ROUTE_ENABLE; - dst[i >> 1] |= (r300VAPInputRoute1Swizzle(swizzle[i + 1]) | R300_INPUT_ROUTE_ENABLE) << 16; + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POINTSIZE )) { + EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F ); + vap_fmt_0 |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT; + offset += 1; } - if (nr & 1) { - dst[nr >> 1] = r300VAPInputRoute1Swizzle(swizzle[nr - 1]) | R300_INPUT_ROUTE_ENABLE; + if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_COLOR0)) { + sz = VB->AttribPtr[VERT_ATTRIB_COLOR0]->size; + rmesa->swtcl.coloroffset = offset; + InputsRead |= 1 << VERT_ATTRIB_COLOR0; + OutputsWritten |= 1 << VERT_RESULT_COL0; + EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_1F + sz - 1 ); + offset += sz; } - return (nr + 1) >> 1; -} + rmesa->swtcl.specoffset = 0; + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) { + sz = VB->AttribPtr[VERT_ATTRIB_COLOR1]->size; + rmesa->swtcl.specoffset = offset; + EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_1F + sz - 1 ); + InputsRead |= 1 << VERT_ATTRIB_COLOR1; + OutputsWritten |= 1 << VERT_RESULT_COL1; + } -static GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead) -{ - /* No idea what this value means. I have seen other values written to - * this register... */ - return 0x5555; -} + if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) { + int i; + + for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { + sz = VB->TexCoordPtr[i]->size; + InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i); + OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i); + EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_1F + sz - 1 ); + vap_fmt_1 |= sz << (3 * i); + } + } + } -static GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead) -{ - r300ContextPtr rmesa = R300_CONTEXT(ctx); - GLuint i, vic_1 = 0; + for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) { + if (InputsRead & (1 << i)) { + inputs[i] = nr++; + } else { + inputs[i] = -1; + } + } + /* Fixed, apply to vir0 only */ if (InputsRead & (1 << VERT_ATTRIB_POS)) - vic_1 |= R300_INPUT_CNTL_POS; - - if (InputsRead & (1 << VERT_ATTRIB_NORMAL)) - vic_1 |= R300_INPUT_CNTL_NORMAL; - + inputs[VERT_ATTRIB_POS] = 0; if (InputsRead & (1 << VERT_ATTRIB_COLOR0)) - vic_1 |= R300_INPUT_CNTL_COLOR; - - rmesa->state.texture.tc_count = 0; - for (i = 0; i < ctx->Const.MaxTextureUnits; i++) - if (InputsRead & (1 << (VERT_ATTRIB_TEX0 + i))) { - rmesa->state.texture.tc_count++; - vic_1 |= R300_INPUT_CNTL_TC0 << i; + inputs[VERT_ATTRIB_COLOR0] = 2; + if (InputsRead & (1 << VERT_ATTRIB_COLOR1)) + inputs[VERT_ATTRIB_COLOR1] = 3; + for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++) + if (InputsRead & (1 << i)) + inputs[i] = 6 + (i - VERT_ATTRIB_TEX0); + + for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) { + if (InputsRead & (1 << i)) { + tab[nr++] = i; } + } - return vic_1; -} - -static GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten) -{ - GLuint ret = 0; - - if (OutputsWritten & (1 << VERT_RESULT_HPOS)) - ret |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT; - - if (OutputsWritten & (1 << VERT_RESULT_COL0)) - ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT; + for (i = 0; i < nr; i++) { + int ci; - if (OutputsWritten & (1 << VERT_RESULT_COL1)) - ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT; + swizzle[i][0] = SWIZZLE_ZERO; + swizzle[i][1] = SWIZZLE_ZERO; + swizzle[i][2] = SWIZZLE_ZERO; + swizzle[i][3] = SWIZZLE_ONE; -#if 0 - if (OutputsWritten & (1 << VERT_RESULT_BFC0)) - ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT; + for (ci = 0; ci < VB->AttribPtr[tab[i]]->size; ci++) { + swizzle[i][ci] = ci; + } + } - if (OutputsWritten & (1 << VERT_RESULT_BFC1)) - ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT; + R300_NEWPRIM(rmesa); + R300_STATECHANGE(rmesa, vir[0]); + ((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count = + r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0], + VB->AttribPtr, inputs, tab, nr); + R300_STATECHANGE(rmesa, vir[1]); + ((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count = + r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle, + nr); - if (OutputsWritten & (1 << VERT_RESULT_FOGC)) ; -#endif + R300_STATECHANGE(rmesa, vic); + rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead); + rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead); - if (OutputsWritten & (1 << VERT_RESULT_PSIZ)) - ret |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT; + R300_STATECHANGE(rmesa, vof); + rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten); + rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = vap_fmt_1; - return ret; -} + rmesa->swtcl.vertex_size = + _tnl_install_attrs( ctx, + rmesa->swtcl.vertex_attrs, + rmesa->swtcl.vertex_attr_count, + NULL, 0 ); -static GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten) -{ - GLuint i, ret = 0; + rmesa->swtcl.vertex_size /= 4; - for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { - if (OutputsWritten & (1 << (VERT_RESULT_TEX0 + i))) { - ret |= (4 << (3 * i)); - } - } + RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset ); - return ret; -} -static void r300SetVertexFormat( GLcontext *ctx ) -{ - r300ContextPtr rmesa = R300_CONTEXT( ctx ); - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &tnl->vb; - DECLARE_RENDERINPUTS(index_bitset); - GLuint InputsRead = 0, OutputsWritten = 0; - int vap_fmt_0 = 0; - int vap_vte_cntl = 0; - int offset = 0; - int vte = 0; - GLuint inputs[VERT_ATTRIB_MAX]; - GLint tab[VERT_ATTRIB_MAX]; - int swizzle[VERT_ATTRIB_MAX][4]; - GLuint i, nr; - - DECLARE_RENDERINPUTS(render_inputs_bitset); - - RENDERINPUTS_COPY(render_inputs_bitset, tnl->render_inputs_bitset); - - RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - - RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset); - - /* Important: - */ - if ( VB->NdcPtr != NULL ) { - VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr; - } - else { - VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr; - } - - assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL ); - rmesa->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. - */ - EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F ); - vap_vte_cntl |= R300_VTX_W0_FMT; - InputsRead |= 1 << VERT_ATTRIB_POS; - OutputsWritten |= 1 << VERT_RESULT_HPOS; - offset = 4; - - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_POINTSIZE )) { - EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F ); - vap_fmt_0 |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT; - offset += 1; - } - - rmesa->swtcl.coloroffset = offset; -#if MESA_LITTLE_ENDIAN - EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4F ); -#else - EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4F ); -#endif - - InputsRead |= 1 << VERT_ATTRIB_COLOR0; - OutputsWritten |= 1 << VERT_RESULT_COL0; - offset += 4; - - rmesa->swtcl.specoffset = 0; - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 ) || - RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) { - -#if MESA_LITTLE_ENDIAN - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) { - rmesa->swtcl.specoffset = offset; - EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3F ); - InputsRead |= 1 << VERT_ATTRIB_COLOR1; - OutputsWritten |= 1 << VERT_RESULT_COL1; - } - else { - EMIT_PAD( 3 ); - } - - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) { - EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F ); - InputsRead |= 1 << VERT_ATTRIB_COLOR1; - OutputsWritten |= 1 << VERT_RESULT_COL1; - } - else { - EMIT_PAD( 1 ); - } -#else - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) { - EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F ); - InputsRead |= 1 << VERT_ATTRIB_COLOR1; - OutputsWritten |= 1 << VERT_RESULT_COL1; - } - else { - EMIT_PAD( 1 ); - } - - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) { - rmesa->swtcl.specoffset = offset; - EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR ); - InputsRead |= 1 << VERT_ATTRIB_COLOR1; - OutputsWritten |= 1 << VERT_RESULT_COL1; - } - else { - EMIT_PAD( 3 ); - } -#endif - } - - if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) { - int i; - - for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; - - InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i); - OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i); - EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_4F ); - } - } - } - -#if 0 - if ( (rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] & R200_FOG_USE_MASK) - != R200_FOG_USE_SPEC_ALPHA ) { - R200_STATECHANGE( rmesa, ctx ); - rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] &= ~R200_FOG_USE_MASK; - rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] |= R200_FOG_USE_SPEC_ALPHA; - } -#endif - - for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) { - if (InputsRead & (1 << i)) { - inputs[i] = nr++; - } else { - inputs[i] = -1; - } - } - - /* Fixed, apply to vir0 only */ - if (InputsRead & VERT_ATTRIB_POS) - inputs[VERT_ATTRIB_POS] = 0; - if (InputsRead & (1 << VERT_ATTRIB_COLOR0)) - inputs[VERT_ATTRIB_COLOR0] = 2; - if (InputsRead & (1 << VERT_ATTRIB_COLOR1)) - inputs[VERT_ATTRIB_COLOR1] = 3; - for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++) - if (InputsRead & (1 << i)) - inputs[i] = 6 + (i - VERT_ATTRIB_TEX0); - - for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) { - if (InputsRead & (1 << i)) { - tab[nr++] = i; - } - } - - for (i = 0; i < nr; i++) { - int ci, fix, found = 0; - - swizzle[i][0] = SWIZZLE_ZERO; - swizzle[i][1] = SWIZZLE_ZERO; - swizzle[i][2] = SWIZZLE_ZERO; - swizzle[i][3] = SWIZZLE_ONE; - - for (ci = 0; ci < VB->AttribPtr[tab[i]]->size; ci++) { - swizzle[i][ci] = ci; - } - } - - R300_STATECHANGE(rmesa, vir[0]); - ((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count = - r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0], - VB->AttribPtr, inputs, tab, nr); - R300_STATECHANGE(rmesa, vir[1]); - ((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count = - r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle, - nr); - - R300_STATECHANGE(rmesa, vic); - rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead); - rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead); - - R300_STATECHANGE(rmesa, vof); - rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten); - rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten); - - if (!RENDERINPUTS_EQUAL( rmesa->tnl_index_bitset, index_bitset)) { - - rmesa->swtcl.vertex_size = - _tnl_install_attrs( ctx, - rmesa->swtcl.vertex_attrs, - rmesa->swtcl.vertex_attr_count, - NULL, 0 ); - - rmesa->swtcl.vertex_size /= 4; - RENDERINPUTS_COPY( rmesa->tnl_index_bitset, index_bitset ); - - vte = rmesa->hw.vte.cmd[1]; - R300_STATECHANGE(rmesa, vte); - rmesa->hw.vte.cmd[1] = vte; - rmesa->hw.vte.cmd[2] = rmesa->swtcl.vertex_size; - } + R300_STATECHANGE(rmesa, vte); + rmesa->hw.vte.cmd[1] = vte; + rmesa->hw.vte.cmd[2] = rmesa->swtcl.vertex_size; } /* Flush vertices in the current dma region. */ -static void flush_last_swtcl_prim( r300ContextPtr rmesa ) +static void flush_last_swtcl_prim( GLcontext *ctx ) { - if (RADEON_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - - rmesa->dma.flush = NULL; - - if (rmesa->dma.current.buf) { - struct r300_dma_region *current = &rmesa->dma.current; - GLuint current_offset = GET_START(current); - - assert (current->start + - rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == - current->ptr); - - if (rmesa->dma.current.start != rmesa->dma.current.ptr) { - - r300EmitVertexAOS( rmesa, - rmesa->swtcl.vertex_size, - current_offset); - - r300EmitVbufPrim( rmesa, - rmesa->swtcl.hw_primitive, - rmesa->swtcl.numverts); - } - - rmesa->swtcl.numverts = 0; - current->start = current->ptr; - } + r300ContextPtr rmesa = R300_CONTEXT(ctx); + struct radeon_dma *dma = &rmesa->radeon.dma; + + + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + dma->flush = NULL; + + if (dma->current) { + GLuint current_offset = dma->current_used; + + assert (dma->current_used + + rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == + dma->current_vertexptr); + + radeon_bo_unmap(dma->current); + if (dma->current_used != dma->current_vertexptr) { + dma->current_used = dma->current_vertexptr; + + rcommonEnsureCmdBufSpace(rmesa, + rmesa->hw.max_state_size + (12*sizeof(int)), + __FUNCTION__); + r300EmitState(rmesa); + r300EmitVertexAOS(rmesa, + rmesa->swtcl.vertex_size, + dma->current, + current_offset); + + r300EmitVbufPrim(rmesa, + rmesa->swtcl.hw_primitive, + rmesa->swtcl.numverts); + r300EmitCacheFlush(rmesa); + COMMIT_BATCH(); + } + radeonReleaseDmaRegion(&rmesa->radeon); + rmesa->swtcl.numverts = 0; + } } /* Alloc space in the current dma region. @@ -446,30 +287,30 @@ static void flush_last_swtcl_prim( r300ContextPtr rmesa ) static void * r300AllocDmaLowVerts( r300ContextPtr rmesa, int nverts, int vsize ) { - GLuint bytes = vsize * nverts; - - if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) - r300RefillCurrentDmaRegion( rmesa, bytes); - - 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 ); - + GLuint bytes = vsize * nverts; + void *head; - { - GLubyte *head = (GLubyte *) (rmesa->dma.current.address + rmesa->dma.current.ptr); - rmesa->dma.current.ptr += bytes; - rmesa->swtcl.numverts += nverts; - return head; - } + if (!rmesa->radeon.dma.current || rmesa->radeon.dma.current_vertexptr + bytes > rmesa->radeon.dma.current->size) { + radeonRefillCurrentDmaRegion( &rmesa->radeon, bytes); + } + if (!rmesa->radeon.dma.flush) { + rmesa->radeon.glCtx->Driver.NeedFlush |= FLUSH_STORED_VERTICES; + rmesa->radeon.dma.flush = flush_last_swtcl_prim; + } + + ASSERT( vsize == rmesa->swtcl.vertex_size * 4 ); + ASSERT( rmesa->radeon.dma.flush == flush_last_swtcl_prim ); + ASSERT( rmesa->radeon.dma.current_used + + rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == + rmesa->radeon.dma.current_vertexptr ); + +// fprintf(stderr,"current %p %x\n", rmesa->radeon.dma.current->ptr, +// rmesa->radeon.dma.current_vertexptr); + head = (rmesa->radeon.dma.current->ptr + rmesa->radeon.dma.current_vertexptr); + rmesa->radeon.dma.current_vertexptr += bytes; + rmesa->swtcl.numverts += nverts; + return head; } static GLuint reduced_prim[] = { @@ -515,7 +356,7 @@ static void r300RenderPrimitive( GLcontext *ctx, GLenum prim ); r300ContextPtr rmesa = R300_CONTEXT(ctx); \ const char *r300verts = (char *)rmesa->swtcl.verts; #define VERT(x) (r300Vertex *)(r300verts + ((x) * vertsize * sizeof(int))) -#define VERTEX r300Vertex +#define VERTEX r300Vertex #define DO_DEBUG_VERTS (1 && (RADEON_DEBUG & DEBUG_VERTS)) #define PRINT_VERTEX(x) #undef TAG @@ -700,82 +541,63 @@ static void init_rast_tab( void ) /**********************************************************************/ static void r300ChooseRenderState( GLcontext *ctx ) { - TNLcontext *tnl = TNL_CONTEXT(ctx); - r300ContextPtr rmesa = R300_CONTEXT(ctx); - GLuint index = 0; - GLuint flags = ctx->_TriangleCaps; - - // if (!rmesa->TclFallback || rmesa->Fallback) -// return; - - if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R300_TWOSIDE_BIT; - if (flags & DD_TRI_UNFILLED) index |= R300_UNFILLED_BIT; - - if (index != rmesa->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; - tnl->Driver.Render.Triangle = rast_tab[index].triangle; - tnl->Driver.Render.Quad = rast_tab[index].quad; - - if (index == 0) { - tnl->Driver.Render.PrimTabVerts = r300_render_tab_verts; - tnl->Driver.Render.PrimTabElts = r300_render_tab_elts; - tnl->Driver.Render.ClippedPolygon = r300_fast_clipped_poly; - } else { - tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts; - tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts; - tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon; - } - - rmesa->swtcl.RenderIndex = index; - } + TNLcontext *tnl = TNL_CONTEXT(ctx); + r300ContextPtr rmesa = R300_CONTEXT(ctx); + GLuint index = 0; + GLuint flags = ctx->_TriangleCaps; + + if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R300_TWOSIDE_BIT; + if (flags & DD_TRI_UNFILLED) index |= R300_UNFILLED_BIT; + + if (index != rmesa->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; + tnl->Driver.Render.Triangle = rast_tab[index].triangle; + tnl->Driver.Render.Quad = rast_tab[index].quad; + + if (index == 0) { + tnl->Driver.Render.PrimTabVerts = r300_render_tab_verts; + tnl->Driver.Render.PrimTabElts = r300_render_tab_elts; + tnl->Driver.Render.ClippedPolygon = r300_fast_clipped_poly; + } else { + tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts; + tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts; + tnl->Driver.Render.ClippedPolygon = _tnl_RenderClippedPolygon; + } + + rmesa->swtcl.RenderIndex = index; + } } static void r300RenderStart(GLcontext *ctx) { - r300ContextPtr rmesa = R300_CONTEXT( ctx ); - int cmd_reserved = 0; - int cmd_written = 0; - drm_radeon_cmd_header_t *cmd = NULL; - + r300ContextPtr rmesa = R300_CONTEXT( ctx ); // fprintf(stderr, "%s\n", __FUNCTION__); - + + r300ChooseRenderState(ctx); r300SetVertexFormat(ctx); - r300UpdateShaderStates(rmesa); + r300ValidateTextures(ctx); - reg_start(R300_RB3D_DSTCACHE_CTLSTAT, 0); - e32(R300_RB3D_DSTCACHE_UNKNOWN_0A); - - reg_start(R300_RB3D_ZCACHE_CTLSTAT, 0); - e32(R300_RB3D_ZCACHE_UNKNOWN_03); - + r300UpdateShaders(rmesa); + r300UpdateShaderStates(rmesa); - if (rmesa->dma.flush != 0 && - rmesa->dma.flush != flush_last_swtcl_prim) - rmesa->dma.flush( rmesa ); + r300EmitCacheFlush(rmesa); + if (rmesa->radeon.dma.flush != NULL) { + rmesa->radeon.dma.flush(ctx); + } } static void r300RenderFinish(GLcontext *ctx) { - r300ContextPtr rmesa = R300_CONTEXT( ctx ); - int cmd_reserved = 0; - int cmd_written = 0; - drm_radeon_cmd_header_t *cmd = NULL; - - reg_start(R300_RB3D_DSTCACHE_CTLSTAT, 0); - e32(R300_RB3D_DSTCACHE_UNKNOWN_0A); - - reg_start(R300_RB3D_ZCACHE_CTLSTAT, 0); - e32(R300_RB3D_ZCACHE_UNKNOWN_03); } static void r300RasterPrimitive( GLcontext *ctx, GLuint hwprim ) { r300ContextPtr rmesa = R300_CONTEXT(ctx); - + if (rmesa->swtcl.hw_primitive != hwprim) { R300_NEWPRIM( rmesa ); rmesa->swtcl.hw_primitive = hwprim; @@ -787,10 +609,13 @@ static void r300RenderPrimitive(GLcontext *ctx, GLenum prim) r300ContextPtr rmesa = R300_CONTEXT(ctx); rmesa->swtcl.render_primitive = prim; - if (prim < GL_TRIANGLES || !(ctx->_TriangleCaps & DD_TRI_UNFILLED)) - r300RasterPrimitive( ctx, reduced_prim[prim] ); + + if ((prim == GL_TRIANGLES) && (ctx->_TriangleCaps & DD_TRI_UNFILLED)) + return; + + r300RasterPrimitive( ctx, reduced_prim[prim] ); // fprintf(stderr, "%s\n", __FUNCTION__); - + } static void r300ResetLineStipple(GLcontext *ctx) @@ -804,12 +629,12 @@ void r300InitSwtcl(GLcontext *ctx) TNLcontext *tnl = TNL_CONTEXT(ctx); r300ContextPtr rmesa = R300_CONTEXT(ctx); static int firsttime = 1; - + if (firsttime) { init_rast_tab(); firsttime = 0; } - + tnl->Driver.Render.Start = r300RenderStart; tnl->Driver.Render.Finish = r300RenderFinish; tnl->Driver.Render.PrimitiveNotify = r300RenderPrimitive; @@ -817,15 +642,15 @@ void r300InitSwtcl(GLcontext *ctx) tnl->Driver.Render.BuildVertices = _tnl_build_vertices; tnl->Driver.Render.CopyPV = _tnl_copy_pv; tnl->Driver.Render.Interp = _tnl_interp; - + /* FIXME: what are these numbers? */ - _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, + _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 48 * sizeof(GLfloat) ); - + rmesa->swtcl.verts = (GLubyte *)tnl->clipspace.vertex_buf; rmesa->swtcl.RenderIndex = ~0; rmesa->swtcl.render_primitive = GL_TRIANGLES; - rmesa->swtcl.hw_primitive = 0; + rmesa->swtcl.hw_primitive = 0; _tnl_invalidate_vertex_state( ctx, ~0 ); _tnl_invalidate_vertices( ctx, ~0 ); @@ -833,45 +658,43 @@ void r300InitSwtcl(GLcontext *ctx) _tnl_need_projected_coords( ctx, GL_FALSE ); r300ChooseRenderState(ctx); + + _mesa_validate_all_lighting_tables( ctx ); + + tnl->Driver.NotifyMaterialChange = + _mesa_validate_all_lighting_tables; } void r300DestroySwtcl(GLcontext *ctx) { } -void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, GLuint offset) +void r300EmitVertexAOS(r300ContextPtr rmesa, GLuint vertex_size, struct radeon_bo *bo, GLuint offset) { - int cmd_reserved = 0; - int cmd_written = 0; - int vte; - int route0; + BATCH_LOCALS(&rmesa->radeon); - drm_radeon_cmd_header_t *cmd = NULL; if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s: vertex_size %d, offset 0x%x \n", - __FUNCTION__, vertex_size, offset); - - /* emit vte */ - - start_packet3(CP_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, 2), 2); - e32(1); - e32(vertex_size | (vertex_size << 8)); - e32(offset); + fprintf(stderr, "%s: vertex_size %d, offset 0x%x \n", + __FUNCTION__, vertex_size, offset); + + BEGIN_BATCH(5); + OUT_BATCH_PACKET3(R300_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(); } void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vertex_nr) { - - int cmd_reserved = 0; - int cmd_written = 0; + BATCH_LOCALS(&rmesa->radeon); int type, num_verts; - drm_radeon_cmd_header_t *cmd = NULL; type = r300PrimitiveType(rmesa, primitive); num_verts = r300NumVerts(rmesa, vertex_nr, primitive); - - r300EmitState(rmesa); - - start_packet3(CP_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0), 0); - e32(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type); + + BEGIN_BATCH(3); + OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0); + OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type); + END_BATCH(); }