From 27c07b6b28cb5d9f0b2ba446846670a234e48228 Mon Sep 17 00:00:00 2001 From: Richard Li Date: Tue, 21 Apr 2009 12:54:20 -0400 Subject: Initial pull of code from r6xx-r7xx-support branch Not functional yet. --- src/mesa/drivers/dri/r600/r700_clear.c | 267 +++++++++++++++++++++++++++++++++ 1 file changed, 267 insertions(+) create mode 100644 src/mesa/drivers/dri/r600/r700_clear.c (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c new file mode 100644 index 0000000000..5725360dac --- /dev/null +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -0,0 +1,267 @@ +/* + * Copyright (C) 2008-2009 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* + * Authors: + * Richard Li , + * CooperYuan , + */ + +#include "main/glheader.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" +#include "main/enums.h" + +#include "r600_context.h" +#include "r700_chip.h" + +#if 0 /* to be enabled */ +#include "r700_shaderinst.h" +#include "r700_emit.h" +#endif /* to be enabled */ + +extern void r700InitState (GLcontext * ctx); + +#if 0 /* to be enabled */ +extern void r700SetupVTXConstans(GLcontext * ctx, + unsigned int nStreamID, + unsigned int aos_offset, + unsigned int size, /* number of elements in vector */ + unsigned int stride, + unsigned int count); +extern GLboolean r700SendContextStates(context_t *context); +extern GLboolean r700SyncSurf(context_t *context); +#endif /* to be enabled */ + +static GLboolean r700ClearFast(context_t *context, GLbitfield mask) +{ + /* TODO, fast clear need implementation */ + return GL_FALSE; +} + +static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) +{ + GLcontext *ctx = GL_CONTEXT(context); + +#if 0 /* to be enabled */ + R700_CHIP_CONTEXT r700Saved; + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + struct r600_dma_region aos_vs; + struct r600_dma_region aos_fs; + struct r600_dma_region aos_vb; + aos_vs.buf = NULL; + aos_fs.buf = NULL; + aos_vb.buf = NULL; + + + unsigned int ui; + GLfloat fTemp; + GLfloat fVb[] = { 1.0f, 1.0f, 1.0f, 1.0f, + -1.0f, -1.0f, 1.0f, 1.0f, + 1.0f, -1.0f, 1.0f, 1.0f, + 1.0f, 1.0f, 1.0f, 1.0f, + -1.0f, 1.0f, 1.0f, 1.0f, + -1.0f, -1.0f, 1.0f, 1.0f}; /* TODO : Z set here */ + unsigned int uVs[] = { 0xC, 0x81000000, 0x4, 0xA01C0000, + 0xC001203C, 0x94000688, 0xC001C000, 0x94200688, + 0x10000001, 0x540C90, 0x10000401, 0x20540C90, + 0x10000801, 0x40540C90, 0x90000C01, 0x60400C90, + 0x10000100, 0x600C90, 0x10000500, 0x20600C90, + 0x10000900, 0x40600C90, 0x90000D00, 0x60680C90, + 0x7C000000, 0x2D1001, 0x80000, 0xBEADEAF }; + unsigned int uFs[] = { 0x2, 0xA00C0000, 0xC0008000, 0x94200688, + 0x10000000, 0x340C90, 0x10000400, 0x20340C90, + 0x10000800, 0x40340C90, 0x90000C00, 0x60200C90}; + + if (context->screen->chip.type <= CHIP_TYPE_RV670) + { + uVs[9] = 0x541910; + uVs[11] = 0x20541910; + uVs[13] = 0x40541910; + uVs[15] = 0x60401910; + uVs[17] = 0x601910; + uVs[19] = 0x20601910; + uVs[21] = 0x40601910; + uVs[23] = 0x60681910; + uFs[5] = 0x341910; + uFs[7] = 0x20341910; + uFs[9] = 0x40341910; + uFs[11] = 0x60201910; + } + + r700SyncSurf(context); + + /* Save current chip object. */ + memcpy(&r700Saved, r700, sizeof(R700_CHIP_CONTEXT)); + + r700InitState(ctx); + + r700->CB_COLOR0_BASE.u32All = context->target.rt.gpu >> 8; + + /* Turn off perspective divid. */ + SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit); + SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit); + SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit); + + if( (mask & BUFFER_BIT_FRONT_LEFT) || (mask & BUFFER_BIT_BACK_LEFT) ) + { /* Enable render target output. */ + SETfield(r700->CB_TARGET_MASK.u32All, 0xF, TARGET0_ENABLE_shift, TARGET0_ENABLE_mask); + } + else + { /* Disable render target output. */ + CLEARfield(r700->CB_TARGET_MASK.u32All, TARGET0_ENABLE_mask); /* TODO : OGL need 4 rt. */ + } + if (mask & BUFFER_BIT_DEPTH) + { + /* Set correct Z to clear. */ + SETbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit); + SETbit(r700->DB_DEPTH_CONTROL.u32All, Z_ENABLE_bit); + SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_ALWAYS, ZFUNC_shift, ZFUNC_mask); + fTemp = ctx->Depth.Clear; + for(ui=2; ui<24;) + { + fVb[ui] = fTemp; + ui += 4; + } + } + else + { + /* Disable Z write. */ + CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit); + CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_ENABLE_bit); + } + + /* Setup vb */ + R700_CMDBUF_CHECK_SPACE(6); + R700EP3 (context, IT_SET_CTL_CONST, 1); + R700E32 (context, mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); + R700E32 (context, 0); + R700EP3 (context, IT_SET_CTL_CONST, 1); + R700E32 (context, mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); + R700E32 (context, 0); + (context->chipobj.EmitVec)(ctx, &aos_vb, (GLvoid *)fVb, 4, 16, 6); + r700SetupVTXConstans(ctx, VERT_ATTRIB_POS, (unsigned int)aos_vb.aos_offset, 4, 16, 6); + + /* Setup shaders, copied from dump */ + r700->SQ_PGM_RESOURCES_PS.u32All = 0; + r700->SQ_PGM_RESOURCES_VS.u32All = 0; + SETbit(r700->SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); + SETbit(r700->SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); + /* vs */ + (context->chipobj.EmitShader)(ctx, &aos_vs, (GLvoid *)(&uVs[0]), 28); + r700->SQ_PGM_START_VS.u32All = (aos_vs.aos_offset >> 8) & 0x00FFFFFF; + r700->SQ_PGM_RESOURCES_VS.u32All = 0x00800004; + /* vs const */ /* TODO : Set color here */ + R700_CMDBUF_CHECK_SPACE(4 + 2); + R700EP3 (context, IT_SET_ALU_CONST, 4); + R700E32 (context, SQ_ALU_CONSTANT_VS_OFFSET * 4); + R700E32 (context, *((unsigned int*)&(ctx->Color.ClearColor[0]))); + R700E32 (context, *((unsigned int*)&(ctx->Color.ClearColor[1]))); + R700E32 (context, *((unsigned int*)&(ctx->Color.ClearColor[2]))); + R700E32 (context, *((unsigned int*)&(ctx->Color.ClearColor[3]))); + + r700->SPI_VS_OUT_CONFIG.u32All = 0x00000000; + r700->SPI_PS_IN_CONTROL_0.u32All = 0x20000001; + /* ps */ + (context->chipobj.EmitShader)(ctx, &aos_fs, (GLvoid *)(&uFs[0]), 12); + r700->SQ_PGM_START_PS.u32All = (aos_fs.aos_offset >> 8) & 0x00FFFFFF; + r700->SQ_PGM_RESOURCES_PS.u32All = 0x00800002; + r700->SQ_PGM_EXPORTS_PS.u32All = 0x00000002; + r700->DB_SHADER_CONTROL.u32All = 0x00000200; + + r700->CB_SHADER_CONTROL.u32All = 0x00000001; + + /* set a valid base address to make the command checker happy */ + r700->SQ_PGM_START_FS.u32All = (aos_fs.aos_offset >> 8) & 0x00FFFFFF; + r700->SQ_PGM_START_ES.u32All = (aos_fs.aos_offset >> 8) & 0x00FFFFFF; + r700->SQ_PGM_START_GS.u32All = (aos_fs.aos_offset >> 8) & 0x00FFFFFF; + + /* Now, send the states */ + r700SendContextStates(context); + + /* Draw */ + GLuint numEntires, j; + GLuint numIndices = 6; + unsigned int VGT_DRAW_INITIATOR = 0; + unsigned int VGT_INDEX_TYPE = 0; + unsigned int VGT_PRIMITIVE_TYPE = 0; + unsigned int VGT_NUM_INDICES = 0; + + numEntires = 2 /* VGT_INDEX_TYPE */ + + 3 /* VGT_PRIMITIVE_TYPE */ + + numIndices + 3; /* DRAW_INDEX_IMMD */ + + R700_CMDBUF_CHECK_SPACE(numEntires); + + SETfield(VGT_INDEX_TYPE, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); + + R700EP3(context, IT_INDEX_TYPE, 0); + R700E32(context, VGT_INDEX_TYPE); + + VGT_NUM_INDICES = numIndices; + + SETfield(VGT_PRIMITIVE_TYPE, DI_PT_TRILIST, VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask); + R700EP3(context, IT_SET_CONFIG_REG, 1); + R700E32(context, mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); + R700E32(context, VGT_PRIMITIVE_TYPE); + + SETfield(VGT_DRAW_INITIATOR, DI_SRC_SEL_IMMEDIATE, SOURCE_SELECT_shift, SOURCE_SELECT_mask); + SETfield(VGT_DRAW_INITIATOR, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask); + + R700EP3(context, IT_DRAW_INDEX_IMMD, (numIndices + 1)); + R700E32(context, VGT_NUM_INDICES); + R700E32(context, VGT_DRAW_INITIATOR); + + for (j=0; jchipobj.FlushCmdBuffer)(context); + + /* TODO : keep these in context, don't load and release every time. */ + (context->chipobj.FreeDmaRegion)(context, &aos_vs); + (context->chipobj.FreeDmaRegion)(context, &aos_fs); + (context->chipobj.FreeDmaRegion)(context, &aos_vb); + + /* Restore chip object. */ + memcpy(r700, &r700Saved, sizeof(R700_CHIP_CONTEXT)); + +#endif /* to be enabled */ + + return GL_TRUE; +} + +void r700Clear(GLcontext * ctx, GLbitfield mask) +{ + context_t *context = R700_CONTEXT(ctx); + + if( GL_TRUE == r700ClearFast(context, mask) ) + { + return; + } + + r700ClearWithDraw(context, mask); +} + + -- cgit v1.2.3 From e2dcebd2e6b2af6269a5ece6d6ced73ec8bb4a47 Mon Sep 17 00:00:00 2001 From: Richard Li Date: Fri, 8 May 2009 19:23:45 -0400 Subject: R6xx/R7xx: WIP r6xx-rewrite code --- src/mesa/drivers/dri/r600/Makefile | 3 + src/mesa/drivers/dri/r600/r600_cmdbuf.c | 589 ++++++++++++++++++++++++- src/mesa/drivers/dri/r600/r600_cmdbuf.h | 167 +++++++ src/mesa/drivers/dri/r600/r600_context.c | 69 +-- src/mesa/drivers/dri/r600/r600_context.h | 39 +- src/mesa/drivers/dri/r600/r600_emit.c | 224 +++++++++- src/mesa/drivers/dri/r600/r600_emit.h | 257 ++--------- src/mesa/drivers/dri/r600/r700_chip.c | 248 ++++++++++- src/mesa/drivers/dri/r600/r700_chip.h | 11 +- src/mesa/drivers/dri/r600/r700_clear.c | 117 +++-- src/mesa/drivers/dri/r600/r700_fragprog.c | 50 ++- src/mesa/drivers/dri/r600/r700_fragprog.h | 8 +- src/mesa/drivers/dri/r600/r700_ioctl.c | 28 +- src/mesa/drivers/dri/r600/r700_oglprog.c | 26 +- src/mesa/drivers/dri/r600/r700_render.c | 217 ++++----- src/mesa/drivers/dri/r600/r700_state.c | 265 ++++++----- src/mesa/drivers/dri/r600/r700_state.h | 1 + src/mesa/drivers/dri/r600/r700_tex.c | 24 +- src/mesa/drivers/dri/r600/r700_tex.h | 2 + src/mesa/drivers/dri/r600/r700_vertprog.c | 80 +--- src/mesa/drivers/dri/r600/r700_vertprog.h | 10 +- src/mesa/drivers/dri/radeon/radeon_cmdbuf.h | 124 +----- src/mesa/drivers/dri/radeon/radeon_common.c | 4 +- src/mesa/drivers/dri/radeon/radeon_cs_legacy.c | 138 +++--- src/mesa/drivers/dri/radeon/radeon_fbo.c | 4 +- src/mesa/drivers/dri/radeon/radeon_screen.c | 11 +- 26 files changed, 1788 insertions(+), 928 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/Makefile b/src/mesa/drivers/dri/r600/Makefile index 00c56c03cd..474a3167b4 100644 --- a/src/mesa/drivers/dri/r600/Makefile +++ b/src/mesa/drivers/dri/r600/Makefile @@ -48,9 +48,12 @@ DRIVER_SOURCES = \ r700_fragprog.c \ r700_vertprog.c \ r700_shader.c \ + r700_shaderinst.c \ r700_ioctl.c \ r700_oglprog.c \ r700_chip.c \ + r700_state.c \ + r700_clear.c \ r700_render.c \ r700_tex.c \ r700_debug.c \ diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index bc14f9a755..8550457566 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -28,11 +28,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ /** - * \file - * - * \author Nicolai Haehnle + * Mostly coppied from \radeon\radeon_cs_legacy.c */ +#include + #include "main/glheader.h" #include "main/state.h" #include "main/imports.h" @@ -52,11 +52,582 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_mipmap_tree.h" #include "radeon_reg.h" -/** - * Allocate memory for the command buffer and initialize the state atom - * list. Note that the initial hardware state is set by r600InitState(). - */ -void r600InitCmdBuf(r600ContextPtr r600) +struct r600_cs_manager_legacy { - rcommonInitCmdBuf(&r600->radeon); + struct radeon_cs_manager base; + struct radeon_context *ctx; + /* hack for scratch stuff */ + uint32_t pending_age; + uint32_t pending_count; +}; + +struct r600_cs_reloc_legacy { + struct radeon_cs_reloc base; + uint32_t cindices; + uint32_t *indices; + uint32_t *reloc_indices; + struct offset_modifiers offset_mod; +}; + +static struct radeon_cs * r600_cs_create(struct radeon_cs_manager *csm, + uint32_t ndw) +{ + struct radeon_cs *cs; + + cs = (struct radeon_cs*)calloc(1, sizeof(struct radeon_cs)); + if (cs == NULL) { + return NULL; + } + cs->csm = csm; + cs->ndw = (ndw + 0x3FF) & (~0x3FF); + cs->packets = (uint32_t*)malloc(4*cs->ndw); + if (cs->packets == NULL) { + free(cs); + return NULL; + } + cs->relocs_total_size = 0; + return cs; +} + +int r600_cs_write_reloc(struct radeon_cs *cs, + struct radeon_bo *bo, + uint32_t read_domain, + uint32_t write_domain, + uint32_t flags, + offset_modifiers* poffset_mod) +{ + struct r600_cs_reloc_legacy *relocs; + int i; + + relocs = (struct r600_cs_reloc_legacy *)cs->relocs; + /* check domains */ + if ((read_domain && write_domain) || (!read_domain && !write_domain)) { + /* in one CS a bo can only be in read or write domain but not + * in read & write domain at the same sime + */ + return -EINVAL; + } + if (read_domain == RADEON_GEM_DOMAIN_CPU) { + return -EINVAL; + } + if (write_domain == RADEON_GEM_DOMAIN_CPU) { + return -EINVAL; + } + /* check if bo is already referenced */ + for(i = 0; i < cs->crelocs; i++) { + uint32_t *indices; + uint32_t *reloc_indices; + + if (relocs[i].base.bo->handle == bo->handle) { + /* Check domains must be in read or write. As we check already + * checked that in argument one of the read or write domain was + * set we only need to check that if previous reloc as the read + * domain set then the read_domain should also be set for this + * new relocation. + */ + if (relocs[i].base.read_domain && !read_domain) { + return -EINVAL; + } + if (relocs[i].base.write_domain && !write_domain) { + return -EINVAL; + } + relocs[i].base.read_domain |= read_domain; + relocs[i].base.write_domain |= write_domain; + /* save indice */ + relocs[i].cindices++; + indices = (uint32_t*)realloc(relocs[i].indices, + relocs[i].cindices * 4); + reloc_indices = (uint32_t*)realloc(relocs[i].reloc_indices, + relocs[i].cindices * 4); + if ( (indices == NULL) || (reloc_indices == NULL) ) { + relocs[i].cindices -= 1; + return -ENOMEM; + } + relocs[i].indices = indices; + relocs[i].reloc_indices = reloc_indices; + relocs[i].indices[relocs[i].cindices - 1] = cs->cdw - 1; + relocs[i].reloc_indices[relocs[i].cindices - 1] = cs->section_cdw; + cs->section_ndw += 2; + cs->section_cdw += 2; + + relocs[i].offset_mod.shift = poffset_mod->shift; + relocs[i].offset_mod.shiftbits = poffset_mod->shiftbits; + relocs[i].offset_mod.mask = poffset_mod->mask; + + return 0; + } + } + /* add bo to reloc */ + relocs = (struct r600_cs_reloc_legacy*) + realloc(cs->relocs, + sizeof(struct r600_cs_reloc_legacy) * (cs->crelocs + 1)); + if (relocs == NULL) { + return -ENOMEM; + } + cs->relocs = relocs; + relocs[cs->crelocs].base.bo = bo; + relocs[cs->crelocs].base.read_domain = read_domain; + relocs[cs->crelocs].base.write_domain = write_domain; + relocs[cs->crelocs].base.flags = flags; + relocs[cs->crelocs].indices = (uint32_t*)malloc(4); + relocs[cs->crelocs].reloc_indices = (uint32_t*)malloc(4); + if ( (relocs[cs->crelocs].indices == NULL) || (relocs[cs->crelocs].reloc_indices == NULL) ) + { + return -ENOMEM; + } + relocs[cs->crelocs].offset_mod.shift = poffset_mod->shift; + relocs[cs->crelocs].offset_mod.shiftbits = poffset_mod->shiftbits; + relocs[cs->crelocs].offset_mod.mask = poffset_mod->mask; + + relocs[cs->crelocs].indices[0] = cs->cdw - 1; + relocs[cs->crelocs].reloc_indices[0] = cs->section_cdw; + cs->section_ndw += 2; + cs->section_cdw += 2; + relocs[cs->crelocs].cindices = 1; + cs->relocs_total_size += radeon_bo_legacy_relocs_size(bo); + cs->crelocs++; + + radeon_bo_ref(bo); + return 0; +} + +static int r600_cs_begin(struct radeon_cs *cs, + uint32_t ndw, + const char *file, + const char *func, + int line) +{ + if (cs->section) { + fprintf(stderr, "CS already in a section(%s,%s,%d)\n", + cs->section_file, cs->section_func, cs->section_line); + fprintf(stderr, "CS can't start section(%s,%s,%d)\n", + file, func, line); + return -EPIPE; + } + + if (cs->cdw + ndw + 32 > cs->ndw) { /* Left 32 DWORD (8 offset+pitch) spare room for reloc indices */ + uint32_t tmp, *ptr; + int num = (ndw > 0x3FF) ? ndw : 0x3FF; + + tmp = (cs->cdw + 1 + num) & (~num); + ptr = (uint32_t*)realloc(cs->packets, 4 * tmp); + if (ptr == NULL) { + return -ENOMEM; + } + cs->packets = ptr; + cs->ndw = tmp; + } + + cs->section = 1; + cs->section_ndw = 0; + cs->section_cdw = cs->cdw + ndw; /* start of reloc indices. */ + cs->section_file = file; + cs->section_func = func; + cs->section_line = line; + + return 0; +} + +static int r600_cs_end(struct radeon_cs *cs, + const char *file, + const char *func, + int line) + +{ + if (!cs->section) { + fprintf(stderr, "CS no section to end at (%s,%s,%d)\n", + file, func, line); + return -EPIPE; + } + cs->section = 0; + + if ( (cs->section_ndw + cs->cdw) != cs->section_cdw ) + { + fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n", + cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw); + fprintf(stderr, "CS section end at (%s,%s,%d)\n", + file, func, line); + return -EPIPE; + } + + cs->cdw = cs->section_cdw; + return 0; +} + +static int r600_cs_process_relocs(struct radeon_cs *cs, + uint32_t * reloc_chunk, + uint32_t * length_dw_reloc_chunk) +{ + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct r600_cs_reloc_legacy *relocs; + int i, j, r; + + uint32_t offset_dw = 0; + + csm = (struct r600_cs_manager_legacy*)cs->csm; + relocs = (struct r600_cs_reloc_legacy *)cs->relocs; +restart: + for (i = 0; i < cs->crelocs; i++) + { + for (j = 0; j < relocs[i].cindices; j++) + { + uint32_t soffset, eoffset, asicoffset; + + r = radeon_bo_legacy_validate(relocs[i].base.bo, + &soffset, &eoffset); + if (r == -EAGAIN) + { + goto restart; + } + if (r) + { + fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n", + relocs[i].base.bo, soffset, eoffset); + return r; + } + asicoffset = cs->packets[relocs[i].indices[j]] + soffset; + if (asicoffset >= eoffset) + { + /* radeon_bo_debug(relocs[i].base.bo, 12); */ + fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n", + relocs[i].base.bo, soffset, eoffset); + fprintf(stderr, "above end: %p 0x%08X 0x%08X\n", + relocs[i].base.bo, + cs->packets[relocs[i].indices[j]], + eoffset); + exit(0); + return -EINVAL; + } + /* apply offset operator */ + switch (relocs[i].offset_mod.shift) + { + case NO_SHIFT: + asicoffset = asicoffset & relocs[i].offset_mod.mask; + break; + case LEFT_SHIFT: + asicoffset = (asicoffset << relocs[i].offset_mod.shiftbits) & relocs[i].offset_mod.mask; + break; + case RIGHT_SHIFT: + asicoffset = (asicoffset >> relocs[i].offset_mod.shiftbits) & relocs[i].offset_mod.mask; + break; + default: + break; + }; + + /* pkt3 nop header in ib chunk */ + cs->packets[relocs[i].reloc_indices[j]] = 0xC0001000; + + /* reloc index in ib chunk */ + cs->packets[relocs[i].reloc_indices[j] + 1] = offset_dw; + + /* asic offset in reloc chunk */ /* see alex drm r600_nomm_relocate */ + reloc_chunk[offset_dw] = asicoffset; + reloc_chunk[offset_dw + 3] = 0; + + offset_dw += 4; + } + } + + *length_dw_reloc_chunk = offset_dw; + + return 0; +} + +static int r600_cs_set_age(struct radeon_cs *cs) /* -------------- */ +{ + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct r600_cs_reloc_legacy *relocs; + int i; + + relocs = (struct r600_cs_reloc_legacy *)cs->relocs; + for (i = 0; i < cs->crelocs; i++) { + radeon_bo_legacy_pending(relocs[i].base.bo, csm->pending_age); + radeon_bo_unref(relocs[i].base.bo); + } + return 0; +} + +static int r600_cs_emit(struct radeon_cs *cs) +{ + struct r600_cs_manager_legacy *csm = (struct r600_cs_manager_legacy*)cs->csm; + struct drm_radeon_cs cs_cmd; + struct drm_radeon_cs_chunk cs_chunk[2]; + drm_radeon_cmd_buffer_t cmd; + /* drm_r300_cmd_header_t age; */ + uint32_t length_dw_reloc_chunk; + uint64_t ull; + uint64_t * chunk_ptrs[2]; + uint32_t reloc_chunk[128]; + int r; + int retry = 0; + + /* TODO : put chip level things here if need. */ + /* csm->ctx->vtbl.emit_cs_header(cs, csm->ctx); */ + + /* TODO : append buffer age */ + + r = r600_cs_process_relocs(cs, &(reloc_chunk[0]), &length_dw_reloc_chunk); + if (r) { + return 0; + } + + /* raw ib chunk */ + cs_chunk[0].chunk_id = RADEON_CHUNK_ID_IB; + cs_chunk[0].length_dw = cs->cdw; + cs_chunk[0].chunk_data = (uint64_t)(cs->packets); + + /* reloc chaunk */ + cs_chunk[1].chunk_id = RADEON_CHUNK_ID_RELOCS; + cs_chunk[1].length_dw = length_dw_reloc_chunk; + cs_chunk[1].chunk_data = (uint64_t)&(reloc_chunk[0]); + + chunk_ptrs[0] = (uint64_t * )&(cs_chunk[0]); + chunk_ptrs[1] = (uint64_t * )&(cs_chunk[1]); + + cs_cmd.num_chunks = 2; + cs_cmd.cs_id = 0; + cs_cmd.chunks = (uint64_t)&(chunk_ptrs[0]); + + /* dump_cmdbuf(cs); */ + + do + { + r = drmCommandWriteRead(cs->csm->fd, DRM_RADEON_CS, &cs_cmd, sizeof(cs_cmd)); + retry++; + } while (r == -EAGAIN && retry < 1000); + + if (r) { + return r; + } + + r600_cs_set_age(cs); + + cs->csm->read_used = 0; + cs->csm->vram_write_used = 0; + cs->csm->gart_write_used = 0; + + return 0; +} + +static void inline r600_cs_free_reloc(void *relocs_p, int crelocs) +{ + struct r600_cs_reloc_legacy *relocs = relocs_p; + int i; + if (!relocs_p) + return; + for (i = 0; i < crelocs; i++) + { + free(relocs[i].indices); + free(relocs[i].reloc_indices); + } +} + +static int r600_cs_destroy(struct radeon_cs *cs) +{ + r600_cs_free_reloc(cs->relocs, cs->crelocs); + free(cs->relocs); + free(cs->packets); + free(cs); + return 0; +} + +static int r600_cs_erase(struct radeon_cs *cs) +{ + r600_cs_free_reloc(cs->relocs, cs->crelocs); + free(cs->relocs); + cs->relocs_total_size = 0; + cs->relocs = NULL; + cs->crelocs = 0; + cs->cdw = 0; + cs->section = 0; + return 0; +} + +static int r600_cs_need_flush(struct radeon_cs *cs) +{ + /* this function used to flush when the BO usage got to + * a certain size, now the higher levels handle this better */ + return 0; +} + +static void r600_cs_print(struct radeon_cs *cs, FILE *file) +{ +} + +static int r600_cs_check_space(struct radeon_cs *cs, struct radeon_cs_space_check *bos, int num_bo) +{ + struct radeon_cs_manager *csm = cs->csm; + int this_op_read = 0, this_op_gart_write = 0, this_op_vram_write = 0; + uint32_t read_domains, write_domain; + int i; + struct radeon_bo *bo; + + /* check the totals for this operation */ + + if (num_bo == 0) + return 0; + + /* prepare */ + for (i = 0; i < num_bo; i++) + { + bo = bos[i].bo; + + bos[i].new_accounted = 0; + read_domains = bos[i].read_domains; + write_domain = bos[i].write_domain; + + /* pinned bos don't count */ + if (radeon_legacy_bo_is_static(bo)) + continue; + + /* already accounted this bo */ + if (write_domain && (write_domain == bo->space_accounted)) + continue; + + if (read_domains && ((read_domains << 16) == bo->space_accounted)) + continue; + + if (bo->space_accounted == 0) + { + if (write_domain == RADEON_GEM_DOMAIN_VRAM) + this_op_vram_write += bo->size; + else if (write_domain == RADEON_GEM_DOMAIN_GTT) + this_op_gart_write += bo->size; + else + this_op_read += bo->size; + bos[i].new_accounted = (read_domains << 16) | write_domain; + } + else + { + uint16_t old_read, old_write; + + old_read = bo->space_accounted >> 16; + old_write = bo->space_accounted & 0xffff; + + if (write_domain && (old_read & write_domain)) + { + bos[i].new_accounted = write_domain; + /* moving from read to a write domain */ + if (write_domain == RADEON_GEM_DOMAIN_VRAM) + { + this_op_read -= bo->size; + this_op_vram_write += bo->size; + } + else if (write_domain == RADEON_GEM_DOMAIN_VRAM) + { + this_op_read -= bo->size; + this_op_gart_write += bo->size; + } + } + else if (read_domains & old_write) + { + bos[i].new_accounted = bo->space_accounted & 0xffff; + } + else + { + /* rewrite the domains */ + if (write_domain != old_write) + fprintf(stderr,"WRITE DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, write_domain, old_write); + if (read_domains != old_read) + fprintf(stderr,"READ DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, read_domains, old_read); + return RADEON_CS_SPACE_FLUSH; + } + } + } + + if (this_op_read < 0) + this_op_read = 0; + + /* check sizes - operation first */ + if ((this_op_read + this_op_gart_write > csm->gart_limit) || + (this_op_vram_write > csm->vram_limit)) { + return RADEON_CS_SPACE_OP_TO_BIG; + } + + if (((csm->vram_write_used + this_op_vram_write) > csm->vram_limit) || + ((csm->read_used + csm->gart_write_used + this_op_gart_write + this_op_read) > csm->gart_limit)) { + return RADEON_CS_SPACE_FLUSH; + } + + csm->gart_write_used += this_op_gart_write; + csm->vram_write_used += this_op_vram_write; + csm->read_used += this_op_read; + /* commit */ + for (i = 0; i < num_bo; i++) { + bo = bos[i].bo; + bo->space_accounted = bos[i].new_accounted; + } + + return RADEON_CS_SPACE_OK; +} + +static struct radeon_cs_funcs r600_cs_funcs = { + r600_cs_create, + r600_cs_write_reloc, + r600_cs_begin, + r600_cs_end, + r600_cs_emit, + r600_cs_destroy, + r600_cs_erase, + r600_cs_need_flush, + r600_cs_print, + r600_cs_check_space +}; + +struct radeon_cs_manager * r600_radeon_cs_manager_legacy_ctor(struct radeon_context *ctx) +{ + struct r600_cs_manager_legacy *csm; + + csm = (struct r600_cs_manager_legacy*) + calloc(1, sizeof(struct r600_cs_manager_legacy)); + if (csm == NULL) { + return NULL; + } + csm->base.funcs = &r600_cs_funcs; + csm->base.fd = ctx->dri.fd; + csm->ctx = ctx; + csm->pending_age = 1; + return (struct radeon_cs_manager*)csm; +} + +void r600InitCmdBuf(r600ContextPtr r600) /* from rcommonInitCmdBuf */ +{ + radeonContextPtr rmesa = &r600->radeon; + + GLuint size; + /* Initialize command buffer */ + size = 256 * driQueryOptioni(&rmesa->optionCache, + "command_buffer_size"); + if (size < 2 * rmesa->hw.max_state_size) { + size = 2 * rmesa->hw.max_state_size + 65535; + } + if (size > 64 * 256) + size = 64 * 256; + + if (rmesa->radeonScreen->kernel_mm) { + int fd = rmesa->radeonScreen->driScreen->fd; + rmesa->cmdbuf.csm = radeon_cs_manager_gem_ctor(fd); + } else { + rmesa->cmdbuf.csm = r600_radeon_cs_manager_legacy_ctor(rmesa); + } + if (rmesa->cmdbuf.csm == NULL) { + /* FIXME: fatal error */ + return; + } + rmesa->cmdbuf.cs = radeon_cs_create(rmesa->cmdbuf.csm, size); + assert(rmesa->cmdbuf.cs != NULL); + rmesa->cmdbuf.size = size; + + if (!rmesa->radeonScreen->kernel_mm) { + radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, rmesa->radeonScreen->texSize[0]); + radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, rmesa->radeonScreen->gartTextures.size); + } else { + struct drm_radeon_gem_info mminfo; + + if (!drmCommandWriteRead(rmesa->dri.fd, DRM_RADEON_GEM_INFO, &mminfo, sizeof(mminfo))) + { + radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_VRAM, mminfo.vram_visible); + radeon_cs_set_limit(rmesa->cmdbuf.cs, RADEON_GEM_DOMAIN_GTT, mminfo.gart_size); + } + } } diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.h b/src/mesa/drivers/dri/r600/r600_cmdbuf.h index 1bab9c5821..dbb723b2b5 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.h +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.h @@ -37,6 +37,173 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define __R600_CMDBUF_H__ #include "r600_context.h" +#include "r600_emit.h" + +#define RADEON_CP_PACKET3_NOP 0xC0001000 +#define RADEON_CP_PACKET3_NEXT_CHAR 0xC0001900 +#define RADEON_CP_PACKET3_PLY_NEXTSCAN 0xC0001D00 +#define RADEON_CP_PACKET3_SET_SCISSORS 0xC0001E00 +#define RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM 0xC0002300 +#define RADEON_CP_PACKET3_LOAD_MICROCODE 0xC0002400 +#define RADEON_CP_PACKET3_WAIT_FOR_IDLE 0xC0002600 +#define RADEON_CP_PACKET3_3D_DRAW_VBUF 0xC0002800 +#define RADEON_CP_PACKET3_3D_DRAW_IMMD 0xC0002900 +#define RADEON_CP_PACKET3_3D_DRAW_INDX 0xC0002A00 +#define RADEON_CP_PACKET3_LOAD_PALETTE 0xC0002C00 +#define RADEON_CP_PACKET3_3D_LOAD_VBPNTR 0xC0002F00 +#define RADEON_CP_PACKET3_CNTL_PAINT 0xC0009100 +#define RADEON_CP_PACKET3_CNTL_BITBLT 0xC0009200 +#define RADEON_CP_PACKET3_CNTL_SMALLTEXT 0xC0009300 +#define RADEON_CP_PACKET3_CNTL_HOSTDATA_BLT 0xC0009400 +#define RADEON_CP_PACKET3_CNTL_POLYLINE 0xC0009500 +#define RADEON_CP_PACKET3_CNTL_POLYSCANLINES 0xC0009800 +#define RADEON_CP_PACKET3_CNTL_PAINT_MULTI 0xC0009A00 +#define RADEON_CP_PACKET3_CNTL_BITBLT_MULTI 0xC0009B00 +#define RADEON_CP_PACKET3_CNTL_TRANS_BITBLT 0xC0009C00 + +/* r6xx/r7xx packet 3 type offsets */ +#define R600_SET_CONFIG_REG_OFFSET 0x00008000 +#define R600_SET_CONFIG_REG_END 0x0000ac00 +#define R600_SET_CONTEXT_REG_OFFSET 0x00028000 +#define R600_SET_CONTEXT_REG_END 0x00029000 +#define R600_SET_ALU_CONST_OFFSET 0x00030000 +#define R600_SET_ALU_CONST_END 0x00032000 +#define R600_SET_RESOURCE_OFFSET 0x00038000 +#define R600_SET_RESOURCE_END 0x0003c000 +#define R600_SET_SAMPLER_OFFSET 0x0003c000 +#define R600_SET_SAMPLER_END 0x0003cff0 +#define R600_SET_CTL_CONST_OFFSET 0x0003cff0 +#define R600_SET_CTL_CONST_END 0x0003e200 +#define R600_SET_LOOP_CONST_OFFSET 0x0003e200 +#define R600_SET_LOOP_CONST_END 0x0003e380 +#define R600_SET_BOOL_CONST_OFFSET 0x0003e380 +#define R600_SET_BOOL_CONST_END 0x00040000 + +/* r6xx/r7xx packet 3 types */ +#define R600_IT_INDIRECT_BUFFER_END 0x00001700 +#define R600_IT_SET_PREDICATION 0x00002000 +#define R600_IT_REG_RMW 0x00002100 +#define R600_IT_COND_EXEC 0x00002200 +#define R600_IT_PRED_EXEC 0x00002300 +#define R600_IT_START_3D_CMDBUF 0x00002400 +#define R600_IT_DRAW_INDEX_2 0x00002700 +#define R600_IT_CONTEXT_CONTROL 0x00002800 +#define R600_IT_DRAW_INDEX_IMMD_BE 0x00002900 +#define R600_IT_INDEX_TYPE 0x00002A00 +#define R600_IT_DRAW_INDEX 0x00002B00 +#define R600_IT_DRAW_INDEX_AUTO 0x00002D00 +#define R600_IT_DRAW_INDEX_IMMD 0x00002E00 +#define R600_IT_NUM_INSTANCES 0x00002F00 +#define R600_IT_STRMOUT_BUFFER_UPDATE 0x00003400 +#define R600_IT_INDIRECT_BUFFER_MP 0x00003800 +#define R600_IT_MEM_SEMAPHORE 0x00003900 +#define R600_IT_MPEG_INDEX 0x00003A00 +#define R600_IT_WAIT_REG_MEM 0x00003C00 +#define R600_IT_MEM_WRITE 0x00003D00 +#define R600_IT_INDIRECT_BUFFER 0x00003200 +#define R600_IT_CP_INTERRUPT 0x00004000 +#define R600_IT_SURFACE_SYNC 0x00004300 +#define R600_IT_ME_INITIALIZE 0x00004400 +#define R600_IT_COND_WRITE 0x00004500 +#define R600_IT_EVENT_WRITE 0x00004600 +#define R600_IT_EVENT_WRITE_EOP 0x00004700 +#define R600_IT_ONE_REG_WRITE 0x00005700 +#define R600_IT_SET_CONFIG_REG 0x00006800 +#define R600_IT_SET_CONTEXT_REG 0x00006900 +#define R600_IT_SET_ALU_CONST 0x00006A00 +#define R600_IT_SET_BOOL_CONST 0x00006B00 +#define R600_IT_SET_LOOP_CONST 0x00006C00 +#define R600_IT_SET_RESOURCE 0x00006D00 +#define R600_IT_SET_SAMPLER 0x00006E00 +#define R600_IT_SET_CTL_CONST 0x00006F00 +#define R600_IT_SURFACE_BASE_UPDATE 0x00007300 + +extern int r600_cs_write_reloc(struct radeon_cs *cs, + struct radeon_bo *bo, + uint32_t read_domain, + uint32_t write_domain, + uint32_t flags, + offset_modifiers* poffset_mod); + +static inline void r600_cs_write_dword(struct radeon_cs *cs, uint32_t dword) +{ + cs->packets[cs->cdw++] = dword; +} + +/** + * Write one dword to the command buffer. + */ +#define R600_OUT_BATCH(data) \ + do { \ + r600_cs_write_dword(b_l_rmesa->cmdbuf.cs, data);\ + } while(0) + +/** + * Write n dwords from ptr to the command buffer. + */ +#define R600_OUT_BATCH_TABLE(ptr,n) \ + do { \ + int _i; \ + for (_i=0; _i < n; _i++) {\ + r600_cs_write_dword(b_l_rmesa->cmdbuf.cs, ptr[_i]);\ + }\ + } while(0) + +/** + * Write a relocated dword to the command buffer. + */ +#define R600_OUT_BATCH_RELOC(data, bo, offset, rd, wd, flags, offset_mod) \ + do { \ + if (0 && offset) { \ + fprintf(stderr, "(%s:%s:%d) offset : %d\n", \ + __FILE__, __FUNCTION__, __LINE__, offset); \ + } \ + r600_cs_write_dword(b_l_rmesa->cmdbuf.cs, offset); \ + r600_cs_write_reloc(b_l_rmesa->cmdbuf.cs, \ + bo, rd, wd, flags, offset_mod); \ + } while(0) + +/* R600/R700 */ +#define R600_OUT_BATCH_REGS(reg, num) \ +do { \ + if ((reg) >= R600_SET_CONFIG_REG_OFFSET && (reg) < R600_SET_CONFIG_REG_END) { \ + OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, (num))); \ + OUT_BATCH(((reg) - R600_SET_CONFIG_REG_OFFSET) >> 2); \ + } else if ((reg) >= R600_SET_CONTEXT_REG_OFFSET && (reg) < R600_SET_CONTEXT_REG_END) { \ + OUT_BATCH(CP_PACKET3(R600_IT_SET_CONTEXT_REG, (num))); \ + OUT_BATCH(((reg) - R600_SET_CONTEXT_REG_OFFSET) >> 2); \ + } else if ((reg) >= R600_SET_ALU_CONST_OFFSET && (reg) < R600_SET_ALU_CONST_END) { \ + OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (num))); \ + OUT_BATCH(((reg) - R600_SET_ALU_CONST_OFFSET) >> 2); \ + } else if ((reg) >= R600_SET_RESOURCE_OFFSET && (reg) < R600_SET_RESOURCE_END) { \ + OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, (num))); \ + OUT_BATCH(((reg) - R600_SET_RESOURCE_OFFSET) >> 2); \ + } else if ((reg) >= R600_SET_SAMPLER_OFFSET && (reg) < R600_SET_SAMPLER_END) { \ + OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, (num))); \ + OUT_BATCH(((reg) - R600_SET_SAMPLER_OFFSET) >> 2); \ + } else if ((reg) >= R600_SET_CTL_CONST_OFFSET && (reg) < R600_SET_CTL_CONST_END) { \ + OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, (num))); \ + OUT_BATCH(((reg) - R600_SET_CTL_CONST_OFFSET) >> 2); \ + } else if ((reg) >= R600_SET_LOOP_CONST_OFFSET && (reg) < R600_SET_LOOP_CONST_END) { \ + OUT_BATCH(CP_PACKET3(R600_IT_SET_LOOP_CONST, (num))); \ + OUT_BATCH(((reg) - R600_SET_LOOP_CONST_OFFSET) >> 2); \ + } else if ((reg) >= R600_SET_BOOL_CONST_OFFSET && (reg) < R600_SET_BOOL_CONST_END) { \ + OUT_BATCH(CP_PACKET3(R600_IT_SET_BOOL_CONST, (num))); \ + OUT_BATCH(((reg) - R600_SET_BOOL_CONST_OFFSET) >> 2); \ + } else { \ + OUT_BATCH(CP_PACKET0((reg), (num))); \ + } \ +} while (0) + +/** Single register write to command buffer; requires 3 dwords for most things. */ +#define R600_OUT_BATCH_REGVAL(reg, val) \ + R600_OUT_BATCH_REGS((reg), 1); \ + OUT_BATCH((val)) + +/** Continuous register range write to command buffer; requires 1 dword, + * expects count dwords afterwards for register contents. */ +#define R600_OUT_BATCH_REGSEQ(reg, count) \ + R600_OUT_BATCH_REGS((reg), (count)) extern void r600InitCmdBuf(r600ContextPtr r600); diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index 1b43f68ac2..3c8ba003d4 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -156,42 +156,6 @@ const struct dri_extension gl_20_extension[] = { }; -extern struct tnl_pipeline_stage _r600_render_stage; -extern const struct tnl_pipeline_stage _r600_tcl_stage; - -static const struct tnl_pipeline_stage *r600_pipeline[] = { - - /* Try and go straight to t&l - */ - &_r600_tcl_stage, - - /* Catch any t&l fallbacks - */ - &_tnl_vertex_transform_stage, - &_tnl_normal_transform_stage, - &_tnl_lighting_stage, - &_tnl_fog_coordinate_stage, - &_tnl_texgen_stage, - &_tnl_texture_transform_stage, - &_tnl_vertex_program_stage, - - /* Try again to go to tcl? - * - no good for asymmetric-twoside (do with multipass) - * - no good for asymmetric-unfilled (do with multipass) - * - good for material - * - good for texgen - * - need to manipulate a bit of state - * - * - worth it/not worth it? - */ - - /* Else do them here. - */ - &_r600_render_stage, - &_tnl_render_stage, /* FALLBACK */ - 0, -}; - static void r600RunPipeline(GLcontext * ctx) { _mesa_lock_context_textures(ctx); @@ -244,31 +208,6 @@ static void r600_init_vtbl(radeonContextPtr radeon) radeon->vtbl.fallback = r600_fallback; } -/* to be enabled */ -static void r600EmitShader(GLcontext * ctx, - struct r600_dma_region *rvb, - GLvoid * data, - int sizeinDWORD) -{ -} -/* to be enabled */ -static void r600FreeDmaRegion(context_t *context, - struct r600_dma_region *region) -{ -} -/* to be enabled */ -static void r600EmitVec(GLcontext * ctx, - struct r600_dma_region *rvb, - GLvoid * data, - int size, - int stride, - int count) -{ -} -/* to be enabled */ -static void r600ReleaseArrays(GLcontext * ctx) -{ -} /* to be enabled */ static GLboolean r600LoadMemSurf(context_t *context, GLuint dst_offset, /* gpu addr */ @@ -287,6 +226,7 @@ static GLboolean r600AllocMemSurf(context_t *context, GLuint *prefered_heap, /* Now used RADEON_LOCAL_TEX_HEAP, return actual heap used. */ GLuint totalSize) { + return GL_TRUE; } /* to be enabled */ static int r600FlushCmdBuffer(context_t *context) @@ -332,6 +272,7 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, */ driParseConfigFiles(&r600->radeon.optionCache, &screen->optionCache, screen->driScreen->myNum, "r600"); + r600->radeon.initialMaxAnisotropy = driQueryOptionf(&r600->radeon.optionCache, "def_max_anisotropy"); @@ -344,9 +285,10 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, (r600->chipobj.InitFuncs)(&functions); r600->chipobj.EmitShader = r600EmitShader; + r600->chipobj.DeleteShader = r600DeleteShader; r600->chipobj.FreeDmaRegion = r600FreeDmaRegion; r600->chipobj.EmitVec = r600EmitVec; - r600->chipobj.ReleaseArrays = r600ReleaseArrays; + r600->chipobj.ReleaseArrays = r600ReleaseVec; r600->chipobj.LoadMemSurf = r600LoadMemSurf; r600->chipobj.AllocMemSurf = r600AllocMemSurf; r600->chipobj.FlushCmdBuffer = r600FlushCmdBuffer; @@ -476,8 +418,9 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, radeonInitSpanFuncs( ctx ); r600InitCmdBuf(r600); + + (r600->chipobj.InitState)(r600->radeon.glCtx); #if 0 /* to be enabled */ - r600InitState(r600); if (!(screen->chip_flags & RADEON_CHIPSET_TCL)) r600InitSwtcl(ctx); #endif diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index 74e34e3684..70108b74c4 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -140,26 +140,19 @@ struct r600_swtcl_info { GLubyte vertex_attr_count; }; -/* to be enabled */ /* DELETE FOLLOWING 2 SRTUCTS */ -struct r600_dma_buffer +enum { - int refcount; /* the number of retained regions in buf */ - drmBufPtr buf; - int id; + NO_SHIFT = 0, + LEFT_SHIFT = 1, + RIGHT_SHIFT = 2, }; -struct r600_dma_region + +typedef struct offset_modifiers { - struct r600_dma_buffer *buf; - char *address; /* == buf->address */ - int start; - int end; - int ptr; /* offsets from start of buf */ - - int aos_offset; /* address in GART memory */ - int aos_stride; /* distance between elements, in dwords */ - int aos_size; /* number of components (1-4) */ -}; -/* ----------------------- */ + GLuint shift; + GLuint shiftbits; + GLuint mask; +} offset_modifiers; typedef struct chip_object { @@ -175,14 +168,16 @@ typedef struct chip_object GLuint (*GetTexObjSize)(void); /* ------------ IN ------------------- */ - void (*EmitShader)( GLcontext * ctx, - struct r600_dma_region *rvb, + GLboolean (*EmitShader)( GLcontext * ctx, + void ** shaderbo, GLvoid * data, int sizeinDWORD); + GLboolean (*DeleteShader)(GLcontext * ctx, + void * shaderbo); void (*FreeDmaRegion)( GLcontext * ctx, - struct r600_dma_region *region); - void (*EmitVec)(GLcontext * ctx, - struct r600_dma_region *rvb, + void * shaderbo); + GLboolean (*EmitVec)(GLcontext * ctx, + struct radeon_aos *aos, GLvoid * data, int size, int stride, diff --git a/src/mesa/drivers/dri/r600/r600_emit.c b/src/mesa/drivers/dri/r600/r600_emit.c index 9d43873c4d..fc8f987cab 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.c +++ b/src/mesa/drivers/dri/r600/r600_emit.c @@ -1,9 +1,8 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. +/************************************************************************** + +Copyright 2008, 2009 Advanced Micro Devices Inc. (AMD) -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. +Copyright (C) Advanced Micro Devices Inc. (AMD) 2009. All Rights Reserved. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the @@ -27,10 +26,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/** - * \file - * - * \author Keith Whitwell +/* + * Authors: + * Richard Li , + * CooperYuan , */ #include "main/glheader.h" @@ -48,6 +47,105 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r600_context.h" #include "r600_emit.h" +#if defined(USE_X86_ASM) +#define COPY_DWORDS( dst, src, nr ) \ +do { \ + int __tmp; \ + __asm__ __volatile__( "rep ; movsl" \ + : "=%c" (__tmp), "=D" (dst), "=S" (__tmp) \ + : "0" (nr), \ + "D" ((long)dst), \ + "S" ((long)src) ); \ +} while (0) +#else +#define COPY_DWORDS( dst, src, nr ) \ +do { \ + int j; \ + for ( j = 0 ; j < nr ; j++ ) \ + dst[j] = ((int *)src)[j]; \ + dst += nr; \ +} while (0) +#endif + +static void r600EmitVec4(uint32_t *out, GLvoid * data, int stride, int count) +{ + int i; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s count %d stride %d out %p data %p\n", + __FUNCTION__, count, stride, (void *)out, (void *)data); + + if (stride == 4) + COPY_DWORDS(out, data, count); + else + for (i = 0; i < count; i++) { + out[0] = *(int *)data; + out++; + data += stride; + } +} + +static void r600EmitVec8(uint32_t *out, GLvoid * data, int stride, int count) +{ + int i; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s count %d stride %d out %p data %p\n", + __FUNCTION__, count, stride, (void *)out, (void *)data); + + if (stride == 8) + COPY_DWORDS(out, data, count * 2); + else + for (i = 0; i < count; i++) { + out[0] = *(int *)data; + out[1] = *(int *)(data + 4); + out += 2; + data += stride; + } +} + +static void r600EmitVec12(uint32_t *out, GLvoid * data, int stride, int count) +{ + int i; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s count %d stride %d out %p data %p\n", + __FUNCTION__, count, stride, (void *)out, (void *)data); + + if (stride == 12) { + COPY_DWORDS(out, data, count * 3); + } + else + for (i = 0; i < count; i++) { + out[0] = *(int *)data; + out[1] = *(int *)(data + 4); + out[2] = *(int *)(data + 8); + out += 3; + data += stride; + } +} + +static void r600EmitVec16(uint32_t *out, GLvoid * data, int stride, int count) +{ + int i; + + if (RADEON_DEBUG & DEBUG_VERTS) + fprintf(stderr, "%s count %d stride %d out %p data %p\n", + __FUNCTION__, count, stride, (void *)out, (void *)data); + + if (stride == 16) + COPY_DWORDS(out, data, count * 4); + else + for (i = 0; i < count; i++) { + out[0] = *(int *)data; + out[1] = *(int *)(data + 4); + out[2] = *(int *)(data + 8); + out[3] = *(int *)(data + 12); + out += 4; + data += stride; + } +} + /* Emit vertex data to GART memory * Route inputs to the vertex processor * This function should never return R600_FALLBACK_TCL when using software tcl. @@ -73,3 +171,111 @@ void r600EmitCacheFlush(r600ContextPtr rmesa) COMMIT_BATCH(); */ } + +GLboolean r600EmitShader(GLcontext * ctx, + void ** shaderbo, + GLvoid * data, + int sizeinDWORD) +{ + radeonContextPtr radeonctx = RADEON_CONTEXT(ctx); + + struct radeon_bo * pbo; + uint32_t *out; + +shader_again_alloc: + pbo = radeon_bo_open(radeonctx->radeonScreen->bom, + 0, + sizeinDWORD * 4, + 256, + RADEON_GEM_DOMAIN_GTT, + 0); + + if (!pbo) + { + rcommonFlushCmdBuf(radeonctx, __FUNCTION__); + goto shader_again_alloc; + } + + radeon_validate_bo(radeonctx, pbo, RADEON_GEM_DOMAIN_GTT, 0); + + if (radeon_revalidate_bos(radeonctx->glCtx) == GL_FALSE) + { + fprintf(stderr,"failure to revalidate BOs - badness\n"); + } + + radeon_bo_map(pbo, 1); + + radeon_bo_ref(pbo); + + out = (uint32_t*)(pbo->ptr); + + memcpy(out, data, sizeinDWORD * 4); + + *shaderbo = (void*)pbo; + + return GL_TRUE; +} + +GLboolean r600DeleteShader(GLcontext * ctx, + void * shaderbo) +{ + struct radeon_bo * pbo = (struct radeon_bo *)shaderbo; + + radeon_bo_unmap(pbo); + radeon_bo_unref(pbo); /* when bo->cref <= 0, bo will be bo_free */ + + return GL_TRUE; +} + +GLboolean r600EmitVec(GLcontext * ctx, + struct radeon_aos *aos, + GLvoid * data, + int size, + int stride, + int count) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + uint32_t *out; + + if (stride == 0) + { + radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * count * 4, 32); + aos->stride = 0; + } + else + { + radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * count * 4, 32); + aos->stride = size; + } + + aos->components = size; + aos->count = count; + + out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); + switch (size) { + case 1: r600EmitVec4(out, data, stride, count); break; + case 2: r600EmitVec8(out, data, stride, count); break; + case 3: r600EmitVec12(out, data, stride, count); break; + case 4: r600EmitVec16(out, data, stride, count); break; + default: + assert(0); + break; + } + + return GL_TRUE; +} + +void r600ReleaseVec(GLcontext * ctx) +{ + radeonReleaseArrays(ctx, ~0); +} + +void r600FreeDmaRegion(context_t *context, + void * shaderbo) +{ + struct radeon_bo *pbo = (struct radeon_bo *)shaderbo; + if(pbo) + { + radeon_bo_unref(pbo); + } +} diff --git a/src/mesa/drivers/dri/r600/r600_emit.h b/src/mesa/drivers/dri/r600/r600_emit.h index 3e1cf1ff5f..11765c868c 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.h +++ b/src/mesa/drivers/dri/r600/r600_emit.h @@ -1,41 +1,38 @@ -/* - * Copyright (C) 2005 Vladimir Dergachev. - * - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining - * a copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial - * portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - */ +/************************************************************************** + +Copyright 2008, 2009 Advanced Micro Devices Inc. (AMD) + +Copyright (C) Advanced Micro Devices Inc. (AMD) 2009. All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ /* * Authors: - * Vladimir Dergachev - * Nicolai Haehnle - * Aapo Tahkola - * Ben Skeggs - * Jerome Glisse + * Richard Li , + * CooperYuan , */ -/* This files defines functions for accessing R600 hardware. - */ + #ifndef __R600_EMIT_H__ #define __R600_EMIT_H__ @@ -44,180 +41,24 @@ #include "r600_cmdbuf.h" #include "radeon_reg.h" -static INLINE uint32_t cmdpacket0(struct radeon_screen *rscrn, - int reg, int count) -{ - if (count) { - return CP_PACKET0(reg, count - 1); - } - return CP_PACKET2; -} - -static INLINE uint32_t cmdvpu(struct radeon_screen *rscrn, int addr, int count) -{ - drm_r300_cmd_header_t cmd; - - cmd.u = 0; - cmd.vpu.cmd_type = R300_CMD_VPU; - cmd.vpu.count = count; - cmd.vpu.adrhi = ((unsigned int)addr & 0xFF00) >> 8; - cmd.vpu.adrlo = ((unsigned int)addr & 0x00FF); - - return cmd.u; -} - -static INLINE uint32_t cmdr500fp(struct radeon_screen *rscrn, - int addr, int count, int type, int clamp) -{ - drm_r300_cmd_header_t cmd; - - cmd.u = 0; - cmd.r500fp.cmd_type = R300_CMD_R500FP; - cmd.r500fp.count = count; - cmd.r500fp.adrhi_flags = ((unsigned int)addr & 0x100) >> 8; - cmd.r500fp.adrhi_flags |= type ? R500FP_CONSTANT_TYPE : 0; - cmd.r500fp.adrhi_flags |= clamp ? R500FP_CONSTANT_CLAMP : 0; - cmd.r500fp.adrlo = ((unsigned int)addr & 0x00FF); - - return cmd.u; -} - -static INLINE uint32_t cmdpacket3(struct radeon_screen *rscrn, int packet) -{ - drm_r300_cmd_header_t cmd; - - cmd.u = 0; - cmd.packet3.cmd_type = R300_CMD_PACKET3; - cmd.packet3.packet = packet; - - return cmd.u; -} - -static INLINE uint32_t cmdcpdelay(struct radeon_screen *rscrn, - unsigned short count) -{ - drm_r300_cmd_header_t cmd; - - cmd.u = 0; - - cmd.delay.cmd_type = R300_CMD_CP_DELAY; - cmd.delay.count = count; - - return cmd.u; -} - -static INLINE uint32_t cmdwait(struct radeon_screen *rscrn, - unsigned char flags) -{ - drm_r300_cmd_header_t cmd; - - cmd.u = 0; - cmd.wait.cmd_type = R300_CMD_WAIT; - cmd.wait.flags = flags; - - return cmd.u; -} - -static INLINE uint32_t cmdpacify(struct radeon_screen *rscrn) -{ - drm_r300_cmd_header_t cmd; - - cmd.u = 0; - cmd.header.cmd_type = R300_CMD_END3D; - - return cmd.u; -} - -/** - * Write the header of a packet3 to the command buffer. - * Outputs 2 dwords and expects (num_extra+1) additional dwords afterwards. - */ -#define OUT_BATCH_PACKET3(packet, num_extra) do {\ - if (!b_l_rmesa->radeonScreen->kernel_mm) { \ - OUT_BATCH(cmdpacket3(b_l_rmesa->radeonScreen,\ - R300_CMD_PACKET3_RAW)); \ - } else b_l_rmesa->cmdbuf.cs->section_cdw++;\ - OUT_BATCH(CP_PACKET3((packet), (num_extra))); \ - } while(0) - -/** - * Must be sent to switch to 2d commands - */ -void static INLINE end_3d(radeonContextPtr radeon) -{ - BATCH_LOCALS(radeon); - - if (!radeon->radeonScreen->kernel_mm) { - BEGIN_BATCH_NO_AUTOSTATE(1); - OUT_BATCH(cmdpacify(radeon->radeonScreen)); - END_BATCH(); - } -} - -void static INLINE cp_delay(r600ContextPtr rmesa, unsigned short count) -{ - BATCH_LOCALS(&rmesa->radeon); - - if (!rmesa->radeon.radeonScreen->kernel_mm) { - BEGIN_BATCH_NO_AUTOSTATE(1); - OUT_BATCH(cmdcpdelay(rmesa->radeon.radeonScreen, count)); - END_BATCH(); - } -} - -void static INLINE cp_wait(radeonContextPtr radeon, unsigned char flags) -{ - BATCH_LOCALS(radeon); - uint32_t wait_until; - - if (!radeon->radeonScreen->kernel_mm) { - BEGIN_BATCH_NO_AUTOSTATE(1); - OUT_BATCH(cmdwait(radeon->radeonScreen, flags)); - END_BATCH(); - } else { - switch(flags) { - case R300_WAIT_2D: - wait_until = (1 << 14); - break; - case R300_WAIT_3D: - wait_until = (1 << 15); - break; - case R300_NEW_WAIT_2D_3D: - wait_until = (1 << 14) | (1 << 15); - break; - case R300_NEW_WAIT_2D_2D_CLEAN: - wait_until = (1 << 14) | (1 << 16) | (1 << 18); - break; - case R300_NEW_WAIT_3D_3D_CLEAN: - wait_until = (1 << 15) | (1 << 17) | (1 << 18); - break; - case R300_NEW_WAIT_2D_2D_CLEAN_3D_3D_CLEAN: - wait_until = (1 << 14) | (1 << 16) | (1 << 18); - wait_until |= (1 << 15) | (1 << 17) | (1 << 18); - break; - default: - return; - } - BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH(CP_PACKET0(RADEON_WAIT_UNTIL, 0)); - OUT_BATCH(wait_until); - END_BATCH(); - } -} - -extern int r600EmitArrays(GLcontext * ctx); - -extern int r600PrimitiveType(r600ContextPtr rmesa, int prim); -extern int r600NumVerts(r600ContextPtr rmesa, int num_verts, int prim); - -extern void r600EmitCacheFlush(r600ContextPtr rmesa); - -extern GLuint r600VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr, - int *inputs, GLint * tab, GLuint nr); -extern GLuint r600VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr); -extern GLuint r600VAPInputCntl0(GLcontext * ctx, GLuint InputsRead); -extern GLuint r600VAPInputCntl1(GLcontext * ctx, GLuint InputsRead); -extern GLuint r600VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten); -extern GLuint r600VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten); +extern GLboolean r600EmitShader(GLcontext * ctx, + void ** shaderbo, + GLvoid * data, + int sizeinDWORD); + +extern GLboolean r600DeleteShader(GLcontext * ctx, + void * shaderbo); + +extern GLboolean r600EmitVec(GLcontext * ctx, + struct radeon_aos *aos, + GLvoid * data, + int size, + int stride, + int count); + +extern void r600ReleaseVec(GLcontext * ctx); + +extern void r600FreeDmaRegion(context_t *context, + void * shaderbo); #endif diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index d0371cdda0..e534bb11b2 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -29,15 +29,15 @@ #include "main/glheader.h" #include "r600_context.h" +#include "r600_cmdbuf.h" #include "r700_chip.h" #include "r700_state.h" #include "r700_tex.h" #include "r700_oglprog.h" +#include "r700_fragprog.h" +#include "r700_vertprog.h" #include "r700_ioctl.h" -/* to be enable -#include "r700_emit.h" -*/ extern const struct tnl_pipeline_stage *r700_pipeline[]; @@ -370,15 +370,235 @@ GLboolean r700InitChipObject(context_t *context) return GL_TRUE; } -GLboolean r700SendContextStates(context_t *context) +void r700SetupVTXConstans(GLcontext * ctx, + unsigned int nStreamID, + void * pAos, + unsigned int size, /* number of elements in vector */ + unsigned int stride, + unsigned int count) /* number of vectors in stream */ { -#if 0 //to be enable + context_t *context = R700_CONTEXT(ctx); + uint32_t *dest; + struct radeon_aos * paos = (struct radeon_aos *)pAos; + offset_modifiers offset_mod = {NO_SHIFT, 0, 0xFFFFFFFF}; + + BATCH_LOCALS(&context->radeon); + + unsigned int uSQ_VTX_CONSTANT_WORD0_0; + unsigned int uSQ_VTX_CONSTANT_WORD1_0; + unsigned int uSQ_VTX_CONSTANT_WORD2_0 = 0; + unsigned int uSQ_VTX_CONSTANT_WORD3_0 = 0; + unsigned int uSQ_VTX_CONSTANT_WORD6_0 = 0; + + uSQ_VTX_CONSTANT_WORD0_0 = paos->offset; + uSQ_VTX_CONSTANT_WORD1_0 = count * stride - 1; + + uSQ_VTX_CONSTANT_WORD2_0 |= 0 << BASE_ADDRESS_HI_shift /* TODO */ + |stride << SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift + |GetSurfaceFormat(GL_FLOAT, size, NULL) << SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_shift /* TODO : trace back api for initial data type, not only GL_FLOAT */ + |SQ_NUM_FORMAT_SCALED << SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift + |SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit; + + uSQ_VTX_CONSTANT_WORD3_0 |= 1 << MEM_REQUEST_SIZE_shift; + + uSQ_VTX_CONSTANT_WORD6_0 |= SQ_TEX_VTX_VALID_BUFFER << SQ_TEX_RESOURCE_WORD6_0__TYPE_shift; + + BEGIN_BATCH_NO_AUTOSTATE(9); + + OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); + OUT_BATCH((nStreamID + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); + + R600_OUT_BATCH_RELOC(uSQ_VTX_CONSTANT_WORD1_0, + paos->bo, + uSQ_VTX_CONSTANT_WORD1_0, + RADEON_GEM_DOMAIN_GTT, 0, 0, &offset_mod); + OUT_BATCH(uSQ_VTX_CONSTANT_WORD1_0); + OUT_BATCH(uSQ_VTX_CONSTANT_WORD2_0); + OUT_BATCH(uSQ_VTX_CONSTANT_WORD3_0); + OUT_BATCH(0); + OUT_BATCH(0); + OUT_BATCH(uSQ_VTX_CONSTANT_WORD6_0); + + END_BATCH(); + COMMIT_BATCH(); + +} + +int r700SetupStreams(GLcontext * ctx) +{ + context_t *context = R700_CONTEXT(ctx); + + BATCH_LOCALS(&context->radeon); + + struct r700_vertex_program *vpc + = (struct r700_vertex_program *)ctx->VertexProgram._Current; + + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *vb = &tnl->vb; + + unsigned int unBit; + unsigned int i; + + BEGIN_BATCH_NO_AUTOSTATE(6); + OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); + OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); + OUT_BATCH(0); + + OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); + OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); + OUT_BATCH(0); + END_BATCH(); + COMMIT_BATCH(); + + //context->aos_count = 0; + for(i=0; imesa_program.Base.InputsRead & unBit) + { + (context->chipobj.EmitVec)(ctx, + &context->radeon.tcl.aos[i], + vb->AttribPtr[i]->data, + vb->AttribPtr[i]->size, + vb->AttribPtr[i]->stride, + vb->Count); + + /* currently aos are packed */ + r700SetupVTXConstans(ctx, + i, + (void*)(&context->radeon.tcl.aos[i]), + (unsigned int)vb->AttribPtr[i]->size, + (unsigned int)(vb->AttribPtr[i]->size * 4), + (unsigned int)vb->Count); + } + } + + return R600_FALLBACK_NONE; +} + +inline GLboolean setRelocReg(context_t *context, unsigned int reg, + void * pbo_vs, void * pbo_fs) +{ + BATCH_LOCALS(&context->radeon); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + + struct radeon_bo * pbo; + uint32_t voffset; + offset_modifiers offset_mod; + + switch (reg + ASIC_CONTEXT_BASE_INDEX) + { + case mmCB_COLOR0_BASE: + case mmCB_COLOR1_BASE: + case mmCB_COLOR2_BASE: + case mmCB_COLOR3_BASE: + case mmCB_COLOR4_BASE: + case mmCB_COLOR5_BASE: + case mmCB_COLOR6_BASE: + case mmCB_COLOR7_BASE: + { + GLcontext *ctx = GL_CONTEXT(context); + struct radeon_renderbuffer *rrb; + + rrb = radeon_get_colorbuffer(&context->radeon); + if (!rrb || !rrb->bo) + { + fprintf(stderr, "no rrb\n"); + return GL_FALSE; + } + + offset_mod.shift = RIGHT_SHIFT; + offset_mod.shiftbits = 8; + offset_mod.mask = 0x00FFFFFF; + + R600_OUT_BATCH_RELOC(r700->CB_COLOR0_BASE.u32All, + rrb->bo, + r700->CB_COLOR0_BASE.u32All, + 0, RADEON_GEM_DOMAIN_VRAM, 0, &offset_mod); + return GL_TRUE; + } + break; + case mmDB_DEPTH_BASE: + { + GLcontext *ctx = GL_CONTEXT(context); + struct radeon_renderbuffer *rrb; + rrb = radeon_get_depthbuffer(&context->radeon); + + offset_mod.shift = RIGHT_SHIFT; + offset_mod.shiftbits = 8; + offset_mod.mask = 0x00FFFFFF; + + R600_OUT_BATCH_RELOC(r700->DB_DEPTH_BASE.u32All, + rrb->bo, + r700->DB_DEPTH_BASE.u32All, + 0, RADEON_GEM_DOMAIN_VRAM, 0, &offset_mod); + + return GL_TRUE; + } + break; + case mmSQ_PGM_START_VS: + { + if(NULL != pbo_vs) + { + pbo = (struct radeon_bo *)pbo_vs; + } + else + { + pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); + } + + offset_mod.shift = RIGHT_SHIFT; + offset_mod.shiftbits = 8; + offset_mod.mask = 0x00FFFFFF; + + R600_OUT_BATCH_RELOC(r700->SQ_PGM_START_VS.u32All, + pbo, + r700->SQ_PGM_START_VS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0, &offset_mod); + return GL_TRUE; + } + break; + case mmSQ_PGM_START_FS: + case mmSQ_PGM_START_ES: + case mmSQ_PGM_START_GS: + case mmSQ_PGM_START_PS: + { + if(NULL != pbo_fs) + { + pbo = (struct radeon_bo *)pbo_fs; + } + else + { + pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); + } + + offset_mod.shift = RIGHT_SHIFT; + offset_mod.shiftbits = 8; + offset_mod.mask = 0x00FFFFFF; + + voffset = 0; + R600_OUT_BATCH_RELOC(r700->SQ_PGM_START_PS.u32All, + pbo, + r700->SQ_PGM_START_PS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0, &offset_mod); + return GL_TRUE; + } + break; + } + + return GL_FALSE; +} + +GLboolean r700SendContextStates(context_t *context, void * pbo_vs, void * pbo_fs) +{ + BATCH_LOCALS(&context->radeon); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); ContextState * pState = r700->pStateList; ContextState * pInit; unsigned int toSend; - unsigned int ui; + unsigned int ui; while(NULL != pState) { @@ -401,17 +621,21 @@ GLboolean r700SendContextStates(context_t *context) pState = pState->pNext; - R700_CMDBUF_CHECK_SPACE(toSend + 2); - R700EP3(context, IT_SET_CONTEXT_REG, toSend); - R700E32(context, pInit->unOffset); - + BEGIN_BATCH_NO_AUTOSTATE(toSend + 2); + R600_OUT_BATCH_REGSEQ(((pInit->unOffset + ASIC_CONTEXT_BASE_INDEX)<<2), toSend); for(ui=0; uipuiValue)); + if( GL_FALSE == setRelocReg(context, (pInit->unOffset+ui), pbo_vs, pbo_fs) ) + { + /* for not reloc reg. */ + OUT_BATCH(*(pInit->puiValue)); + } pInit = pInit->pNext; }; + END_BATCH(); }; -#endif //to be enable + COMMIT_BATCH(); + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_chip.h b/src/mesa/drivers/dri/r600/r700_chip.h index f08190027c..30ca2cab84 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.h +++ b/src/mesa/drivers/dri/r600/r700_chip.h @@ -446,13 +446,22 @@ typedef struct _R700_CHIP_CONTEXT ContextState * pStateList; R700_TEXTURE_STATES texture_states; + + GLboolean bEnablePerspective; } R700_CHIP_CONTEXT; #define R700_CONTEXT_STATES(context) ((R700_CHIP_CONTEXT *)(context->chipobj.pvChipObj)) extern GLboolean r700InitChipObject(context_t *context); -extern GLboolean r700SendContextStates(context_t *context); +extern GLboolean r700SendContextStates(context_t *context, void * pbo_vs, void * pbo_fs); +extern int r700SetupStreams(GLcontext * ctx); +extern void r700SetupVTXConstans(GLcontext * ctx, + unsigned int nStreamID, + void * pAos, + unsigned int size, /* number of elements in vector */ + unsigned int stride, + unsigned int Count); /* number of vectors in stream */ #endif /* _R700_CHIP_H_ */ diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index 5725360dac..743875dfe2 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -35,23 +35,12 @@ #include "r600_context.h" #include "r700_chip.h" -#if 0 /* to be enabled */ #include "r700_shaderinst.h" -#include "r700_emit.h" -#endif /* to be enabled */ +#include "r600_emit.h" extern void r700InitState (GLcontext * ctx); -#if 0 /* to be enabled */ -extern void r700SetupVTXConstans(GLcontext * ctx, - unsigned int nStreamID, - unsigned int aos_offset, - unsigned int size, /* number of elements in vector */ - unsigned int stride, - unsigned int count); -extern GLboolean r700SendContextStates(context_t *context); extern GLboolean r700SyncSurf(context_t *context); -#endif /* to be enabled */ static GLboolean r700ClearFast(context_t *context, GLbitfield mask) { @@ -63,17 +52,15 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) { GLcontext *ctx = GL_CONTEXT(context); -#if 0 /* to be enabled */ + BATCH_LOCALS(&context->radeon); + R700_CHIP_CONTEXT r700Saved; R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); - struct r600_dma_region aos_vs; - struct r600_dma_region aos_fs; - struct r600_dma_region aos_vb; - aos_vs.buf = NULL; - aos_fs.buf = NULL; - aos_vb.buf = NULL; - + void * pbo_vs; + void * pbo_fs; + struct radeon_aos aos_vb; + unsigned int ui; GLfloat fTemp; GLfloat fVb[] = { 1.0f, 1.0f, 1.0f, 1.0f, @@ -93,7 +80,7 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) 0x10000000, 0x340C90, 0x10000400, 0x20340C90, 0x10000800, 0x40340C90, 0x90000C00, 0x60200C90}; - if (context->screen->chip.type <= CHIP_TYPE_RV670) + if (context->radeon.radeonScreen->chip_family <= CHIP_FAMILY_RV670) { uVs[9] = 0x541910; uVs[11] = 0x20541910; @@ -116,7 +103,7 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) r700InitState(ctx); - r700->CB_COLOR0_BASE.u32All = context->target.rt.gpu >> 8; + r700SetRenderTarget(context); /* Turn off perspective divid. */ SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit); @@ -152,15 +139,19 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) } /* Setup vb */ - R700_CMDBUF_CHECK_SPACE(6); - R700EP3 (context, IT_SET_CTL_CONST, 1); - R700E32 (context, mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); - R700E32 (context, 0); - R700EP3 (context, IT_SET_CTL_CONST, 1); - R700E32 (context, mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); - R700E32 (context, 0); + BEGIN_BATCH_NO_AUTOSTATE(6); + OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); + OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); + OUT_BATCH(0); + + OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); + OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); + OUT_BATCH(0); + END_BATCH(); + COMMIT_BATCH(); + (context->chipobj.EmitVec)(ctx, &aos_vb, (GLvoid *)fVb, 4, 16, 6); - r700SetupVTXConstans(ctx, VERT_ATTRIB_POS, (unsigned int)aos_vb.aos_offset, 4, 16, 6); + r700SetupVTXConstans(ctx, VERT_ATTRIB_POS, &aos_vb, 4, 16, 6); /* Setup shaders, copied from dump */ r700->SQ_PGM_RESOURCES_PS.u32All = 0; @@ -168,23 +159,26 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) SETbit(r700->SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); SETbit(r700->SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); /* vs */ - (context->chipobj.EmitShader)(ctx, &aos_vs, (GLvoid *)(&uVs[0]), 28); - r700->SQ_PGM_START_VS.u32All = (aos_vs.aos_offset >> 8) & 0x00FFFFFF; + (context->chipobj.EmitShader)(ctx, &pbo_vs, (GLvoid *)(&uVs[0]), 28); + r700->SQ_PGM_START_VS.u32All = 0; r700->SQ_PGM_RESOURCES_VS.u32All = 0x00800004; + /* vs const */ /* TODO : Set color here */ - R700_CMDBUF_CHECK_SPACE(4 + 2); - R700EP3 (context, IT_SET_ALU_CONST, 4); - R700E32 (context, SQ_ALU_CONSTANT_VS_OFFSET * 4); - R700E32 (context, *((unsigned int*)&(ctx->Color.ClearColor[0]))); - R700E32 (context, *((unsigned int*)&(ctx->Color.ClearColor[1]))); - R700E32 (context, *((unsigned int*)&(ctx->Color.ClearColor[2]))); - R700E32 (context, *((unsigned int*)&(ctx->Color.ClearColor[3]))); + BEGIN_BATCH_NO_AUTOSTATE(4 + 2); + OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, 4)); + OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); + OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[0]))); + OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[1]))); + OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[2]))); + OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[3]))); + END_BATCH(); + COMMIT_BATCH(); r700->SPI_VS_OUT_CONFIG.u32All = 0x00000000; r700->SPI_PS_IN_CONTROL_0.u32All = 0x20000001; /* ps */ - (context->chipobj.EmitShader)(ctx, &aos_fs, (GLvoid *)(&uFs[0]), 12); - r700->SQ_PGM_START_PS.u32All = (aos_fs.aos_offset >> 8) & 0x00FFFFFF; + (context->chipobj.EmitShader)(ctx, &pbo_fs, (GLvoid *)(&uFs[0]), 12); + r700->SQ_PGM_START_PS.u32All = 0; r700->SQ_PGM_RESOURCES_PS.u32All = 0x00800002; r700->SQ_PGM_EXPORTS_PS.u32All = 0x00000002; r700->DB_SHADER_CONTROL.u32All = 0x00000200; @@ -192,12 +186,12 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) r700->CB_SHADER_CONTROL.u32All = 0x00000001; /* set a valid base address to make the command checker happy */ - r700->SQ_PGM_START_FS.u32All = (aos_fs.aos_offset >> 8) & 0x00FFFFFF; - r700->SQ_PGM_START_ES.u32All = (aos_fs.aos_offset >> 8) & 0x00FFFFFF; - r700->SQ_PGM_START_GS.u32All = (aos_fs.aos_offset >> 8) & 0x00FFFFFF; + r700->SQ_PGM_START_FS.u32All = 0; + r700->SQ_PGM_START_ES.u32All = 0; + r700->SQ_PGM_START_GS.u32All = 0; /* Now, send the states */ - r700SendContextStates(context); + r700SendContextStates(context, pbo_vs, pbo_fs); /* Draw */ GLuint numEntires, j; @@ -211,44 +205,47 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) + 3 /* VGT_PRIMITIVE_TYPE */ + numIndices + 3; /* DRAW_INDEX_IMMD */ - R700_CMDBUF_CHECK_SPACE(numEntires); + BEGIN_BATCH_NO_AUTOSTATE(numEntires); SETfield(VGT_INDEX_TYPE, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); - R700EP3(context, IT_INDEX_TYPE, 0); - R700E32(context, VGT_INDEX_TYPE); + OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); + OUT_BATCH(VGT_INDEX_TYPE); VGT_NUM_INDICES = numIndices; SETfield(VGT_PRIMITIVE_TYPE, DI_PT_TRILIST, VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask); - R700EP3(context, IT_SET_CONFIG_REG, 1); - R700E32(context, mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); - R700E32(context, VGT_PRIMITIVE_TYPE); + + OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); + OUT_BATCH(VGT_PRIMITIVE_TYPE); SETfield(VGT_DRAW_INITIATOR, DI_SRC_SEL_IMMEDIATE, SOURCE_SELECT_shift, SOURCE_SELECT_mask); SETfield(VGT_DRAW_INITIATOR, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask); - R700EP3(context, IT_DRAW_INDEX_IMMD, (numIndices + 1)); - R700E32(context, VGT_NUM_INDICES); - R700E32(context, VGT_DRAW_INITIATOR); + OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1))); + OUT_BATCH(VGT_NUM_INDICES); + OUT_BATCH(VGT_DRAW_INITIATOR); for (j=0; jchipobj.FlushCmdBuffer)(context); /* TODO : keep these in context, don't load and release every time. */ - (context->chipobj.FreeDmaRegion)(context, &aos_vs); - (context->chipobj.FreeDmaRegion)(context, &aos_fs); - (context->chipobj.FreeDmaRegion)(context, &aos_vb); + (context->chipobj.DeleteShader)(context, &pbo_vs); + + (context->chipobj.DeleteShader)(context, &pbo_fs); + + (context->chipobj.FreeDmaRegion)(context, aos_vb.bo); /* Restore chip object. */ memcpy(r700, &r700Saved, sizeof(R700_CHIP_CONTEXT)); -#endif /* to be enabled */ - return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index 2b813828a5..d06af109e5 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -36,15 +36,11 @@ #include "shader/prog_statevars.h" #include "r600_context.h" +#include "r600_cmdbuf.h" #include "r700_chip.h" #include "r700_fragprog.h" -/* to be enabled */ -#if 0 -#include "r700_emit.h" -#endif - #include "r700_debug.h" //TODO : Validate FP input with VP output. @@ -250,9 +246,18 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, return GL_TRUE; } +void * r700GetActiveFpShaderBo(GLcontext * ctx) +{ + struct r700_fragment_program *fp = (struct r700_fragment_program *) + (ctx->FragmentProgram._Current); + + return fp->shaderbo; +} + GLboolean r700SetupFragmentProgram(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); + BATCH_LOCALS(&context->radeon); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); @@ -271,14 +276,13 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) { Assemble( &(fp->r700Shader) ); } -/* to be enabled */ -#if 0 + /* Load fp to gpu */ (context->chipobj.EmitShader)(ctx, - &(fp->shadercode), + &(fp->shaderbo), (GLvoid *)(fp->r700Shader.pProgram), fp->r700Shader.uShaderBinaryDWORDSize); -#endif + fp->loaded = GL_TRUE; } @@ -288,10 +292,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) /* TODO : enable this after MemUse fixed *= (context->chipobj.MemUse)(context, fp->shadercode.buf->id); */ -/* to be enabled */ -#if 0 - r700->SQ_PGM_START_PS.u32All = (fp->shadercode.aos_offset >> 8) & 0x00FFFFFF; -#endif + + r700->SQ_PGM_START_PS.u32All = 0; /* set from buffer obj */ unNumOfReg = fp->r700Shader.nRegs + 1; @@ -337,25 +339,27 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) if(NULL != paramList) { _mesa_load_state_parameters(ctx, paramList); -/* to be enabled */ -#if 0 + unNumParamData = paramList->NumParameters * 4; - R700_CMDBUF_CHECK_SPACE(2 + unNumParamData); - R700EP3(context, IT_SET_ALU_CONST, unNumParamData); + BEGIN_BATCH_NO_AUTOSTATE(2 + unNumParamData); + + OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData)); + /* assembler map const from very beginning. */ - R700E32(context, SQ_ALU_CONSTANT_PS_OFFSET * 4); + OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4); unNumParamData = paramList->NumParameters; for(ui=0; uiParameterValues[ui][0]))); - R700E32(context, *((unsigned int*)&(paramList->ParameterValues[ui][1]))); - R700E32(context, *((unsigned int*)&(paramList->ParameterValues[ui][2]))); - R700E32(context, *((unsigned int*)&(paramList->ParameterValues[ui][3]))); + OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][0]))); + OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1]))); + OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2]))); + OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3]))); } -#endif + END_BATCH(); + COMMIT_BATCH(); } return GL_TRUE; diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h index 649ee4822d..9c7813e908 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.h +++ b/src/mesa/drivers/dri/r600/r700_fragprog.h @@ -40,10 +40,8 @@ struct r700_fragment_program GLboolean translated; GLboolean loaded; GLboolean error; -/* to be enabled */ -#if 0 - struct r600_dma_region shadercode; -#endif + + void * shaderbo; GLboolean WritesDepth; GLuint optimization; @@ -60,4 +58,6 @@ extern GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, struct gl_fragment_program *mesa_vp); extern GLboolean r700SetupFragmentProgram(GLcontext * ctx); +extern void * r700GetActiveFpShaderBo(GLcontext * ctx); + #endif /*_R700_FRAGPROG_H_*/ diff --git a/src/mesa/drivers/dri/r600/r700_ioctl.c b/src/mesa/drivers/dri/r600/r700_ioctl.c index 0541dc76e4..7493efab0f 100644 --- a/src/mesa/drivers/dri/r600/r700_ioctl.c +++ b/src/mesa/drivers/dri/r600/r700_ioctl.c @@ -37,12 +37,38 @@ #include "radeon_lock.h" #include "r600_context.h" +#include "r700_chip.h" #include "r700_ioctl.h" #include "r700_clear.h" +static void r700Flush(GLcontext *ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + context_t * context = R700_CONTEXT(ctx); + + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s %d\n", __FUNCTION__, radeon->cmdbuf.cs->cdw); + + /* okay if we have no cmds in the buffer && + we have no DMA flush && + we have no DMA buffer allocated. + then no point flushing anything at all. + */ + if (!radeon->dma.flush && !radeon->cmdbuf.cs->cdw && !radeon->dma.current) + return; + + if (radeon->dma.flush) + radeon->dma.flush( ctx ); + + r700SendContextStates(context, NULL, NULL); + + if (radeon->cmdbuf.cs->cdw) + rcommonFlushCmdBuf(radeon, __FUNCTION__); +} + void r700InitIoctlFuncs(struct dd_function_table *functions) { functions->Clear = r700Clear; functions->Finish = radeonFinish; - functions->Flush = radeonFlush; + functions->Flush = r700Flush; } diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.c b/src/mesa/drivers/dri/r600/r700_oglprog.c index ee6377632b..dc8c936165 100644 --- a/src/mesa/drivers/dri/r600/r700_oglprog.c +++ b/src/mesa/drivers/dri/r600/r700_oglprog.c @@ -36,18 +36,16 @@ #include "r700_chip.h" #include "r700_oglprog.h" - -#if 0 /* to be enabled */ #include "r700_fragprog.h" #include "r700_vertprog.h" -#endif /* to be enabled */ + static struct gl_program *r700NewProgram(GLcontext * ctx, GLenum target, GLuint id) { struct gl_program *pProgram = NULL; -#if 0 /* to be enabled */ + struct r700_vertex_program *vp; struct r700_fragment_program *fp; @@ -62,7 +60,9 @@ static struct gl_program *r700NewProgram(GLcontext * ctx, id); vp->translated = GL_FALSE; vp->loaded = GL_FALSE; - vp->shadercode.buf = NULL; + + vp->shaderbo = NULL; + break; case GL_FRAGMENT_PROGRAM_NV: case GL_FRAGMENT_PROGRAM_ARB: @@ -73,18 +73,19 @@ static struct gl_program *r700NewProgram(GLcontext * ctx, id); fp->translated = GL_FALSE; fp->loaded = GL_FALSE; - fp->shadercode.buf = NULL; + + fp->shaderbo = NULL; + break; default: _mesa_problem(ctx, "Bad target in r700NewProgram"); } -#endif /* to be enabled */ + return pProgram; } static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) { -#if 0 /* to be enabled */ struct r700_vertex_program * vp; struct r700_fragment_program * fp; context_t *context = R700_CONTEXT(ctx); @@ -95,7 +96,9 @@ static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) case GL_VERTEX_PROGRAM_ARB: vp = (struct r700_vertex_program*)prog; /* Release DMA region */ - (context->chipobj.FreeDmaRegion)(context, &(vp->shadercode)); + + (context->chipobj.DeleteShader)(ctx, vp->shaderbo); + /* Clean up */ Clean_Up_Assembler(&(vp->r700AsmCode)); Clean_Up_Shader(&(vp->r700Shader)); @@ -104,7 +107,9 @@ static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) case GL_FRAGMENT_PROGRAM_ARB: fp = (struct r700_fragment_program*)prog; /* Release DMA region */ - (context->chipobj.FreeDmaRegion)(context, &(fp->shadercode)); + + (context->chipobj.DeleteShader)(ctx, fp->shaderbo); + /* Clean up */ Clean_Up_Assembler(&(fp->r700AsmCode)); Clean_Up_Shader(&(fp->r700Shader)); @@ -114,7 +119,6 @@ static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) } _mesa_delete_program(ctx, prog); -#endif /* to be enabled */ } static void diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index fce99ada54..f338cf823c 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -45,107 +45,64 @@ #include "tnl/t_pipeline.h" #include "r600_context.h" +#include "r600_cmdbuf.h" #include "r700_chip.h" +#include "r700_tex.h" -/* to be enable #include "r700_vertprog.h" #include "r700_fragprog.h" #include "r700_state.h" -#include "r700_tex.h" -#include "r700_emit.h" -*/ -#if 0 // to be enable void r700WaitForIdle(context_t *context) { - R700EP3 (context, IT_SET_CONFIG_REG, 1); - R700E32 (context, mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX); - R700E32 (context, 1 << 15); + BATCH_LOCALS(&context->radeon); + BEGIN_BATCH_NO_AUTOSTATE(3); + + OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX); + OUT_BATCH(1 << 15); + + END_BATCH(); + COMMIT_BATCH(); } void r700WaitForIdleClean(context_t *context) { - R700EP3 (context, IT_EVENT_WRITE, 0); - R700E32 (context, 0x16); + BATCH_LOCALS(&context->radeon); + BEGIN_BATCH_NO_AUTOSTATE(5); + + OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 0)); + OUT_BATCH(0x16); - R700EP3 (context, IT_SET_CONFIG_REG, 1); - R700E32 (context, mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX); - R700E32 (context, 1 << 17); + OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX); + OUT_BATCH(1 << 17); + + END_BATCH(); + COMMIT_BATCH(); } static void r700Start3D(context_t *context) { - if (context->screen->chip.type <= CHIP_TYPE_RV670) + BATCH_LOCALS(&context->radeon); + if (context->radeon.radeonScreen->chip_family <= CHIP_FAMILY_RV670) { - R700EP3 (context, IT_START_3D_CMDBUF, 1); - R700E32 (context, 0); + BEGIN_BATCH_NO_AUTOSTATE(2); + OUT_BATCH(CP_PACKET3(R600_IT_START_3D_CMDBUF, 1)); + OUT_BATCH(0); + END_BATCH(); } - R700EP3 (context, IT_CONTEXT_CONTROL, 1); - R700E32 (context, 0x80000000); - R700E32 (context, 0x80000000); - r700WaitForIdleClean(context); -} + BEGIN_BATCH_NO_AUTOSTATE(3); + OUT_BATCH(CP_PACKET3(R600_IT_CONTEXT_CONTROL, 1)); + OUT_BATCH(0x80000000); + OUT_BATCH(0x80000000); + END_BATCH(); + COMMIT_BATCH(); -static int r700SetupStreams(GLcontext * ctx) -{ - context_t *context = R700_CONTEXT(ctx); - - struct r700_vertex_program *vpc - = (struct r700_vertex_program *)ctx->VertexProgram._Current; - - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *vb = &tnl->vb; - - unsigned int unBit; - unsigned int i; - - R700_CMDBUF_CHECK_SPACE(6); - R700EP3 (context, IT_SET_CTL_CONST, 1); - R700E32 (context, mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); - R700E32 (context, 0); - - R700EP3 (context, IT_SET_CTL_CONST, 1); - R700E32 (context, mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); - R700E32 (context, 0); - - context->aos_count = 0; - for(i=0; imesa_program.Base.InputsRead & unBit) - { - (context->chipobj.EmitVec)(ctx, - &(context->aos[context->aos_count]), - vb->AttribPtr[i]->data, - vb->AttribPtr[i]->size, - vb->AttribPtr[i]->stride, - vb->Count); - - context->aos[context->aos_count].aos_size = vb->AttribPtr[i]->size; - - /* currently aos are packed */ - r700SetupVTXConstans(ctx, - i, - (unsigned int)context->aos[context->aos_count].aos_offset, - (unsigned int)vb->AttribPtr[i]->size, - (unsigned int)(vb->AttribPtr[i]->size * 4), - (unsigned int)vb->Count); - /* TODO : enable this after MemUse fixed *= - (context->chipobj.MemUse)(context, context->aos[context->aos_count].buf->id); - */ - - context->aos_count++; - } - } - for(i=context->aos_count; iaos[i].buf = NULL; - } - - return R600_FALLBACK_NONE; + r700WaitForIdleClean(context); } static GLboolean r700SetupShaders(GLcontext * ctx) @@ -177,7 +134,7 @@ GLboolean r700SendTextureState(context_t *context) unsigned int i; R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); - +#if 0 /* to be enabled */ for(i=0; itexture_states.textures[i] != 0) @@ -204,12 +161,15 @@ GLboolean r700SendTextureState(context_t *context) R700E32 (context, r700->texture_states.samplers[i]->SQ_TEX_SAMPLER2.u32All); } } - +#endif return GL_TRUE; } GLboolean r700SyncSurf(context_t *context) { +#if 0 //to be enabled + BATCH_LOCALS(&context->radeon); + /* TODO : too heavy? */ unsigned int CP_COHER_CNTL = 0; @@ -221,23 +181,20 @@ GLboolean r700SyncSurf(context_t *context) |SMX_ACTION_ENA_bit; - R700_CMDBUF_CHECK_SPACE(5); - R700EP3(context, IT_SURFACE_SYNC, 3); - R700E32(context, CP_COHER_CNTL); - R700E32(context, 0xFFFFFFFF); - R700E32(context, 0x00000000); - R700E32(context, 10); + BEGIN_BATCH_NO_AUTOSTATE(5); + + OUT_BATCH(CP_PACKET3((IT_SURFACE_SYNC << 8), 3))); + OUT_BATCH(CP_COHER_CNTL); + OUT_BATCH(0xFFFFFFFF); + OUT_BATCH(0x00000000); + OUT_BATCH(10); + END_BATCH(); + COMMIT_BATCH(); +#endif return GL_TRUE; } -static void r700SetRenderTarget(context_t *context) -{ - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); - - r700->CB_COLOR0_BASE.u32All = context->target.rt.gpu >> 8; -} - unsigned int r700PrimitiveType(int prim) { switch (prim & PRIM_MODE_MASK) @@ -279,21 +236,21 @@ unsigned int r700PrimitiveType(int prim) } } -#endif // to be enable - static GLboolean r700RunRender(GLcontext * ctx, struct tnl_pipeline_stage *stage) { -#if 0 // to be enable context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + + BATCH_LOCALS(&context->radeon); + unsigned int i, j; TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *vb = &tnl->vb; struct r700_fragment_program *fp = (struct r700_fragment_program *) (ctx->FragmentProgram._Current); - if (context->screen->chip.type <= CHIP_TYPE_RV670) + if (context->radeon.radeonScreen->chip_family <= CHIP_FAMILY_RV670) { fp->r700AsmCode.bR6xx = 1; } @@ -329,7 +286,7 @@ static GLboolean r700RunRender(GLcontext * ctx, r700->SQ_PGM_START_ES.u32All = r700->SQ_PGM_START_PS.u32All; r700->SQ_PGM_START_GS.u32All = r700->SQ_PGM_START_PS.u32All; - r700SendContextStates(context); + r700SendContextStates(context, NULL, NULL); /* richard test code */ for (i = 0; i < vb->PrimitiveCount; i++) @@ -351,51 +308,55 @@ static GLboolean r700RunRender(GLcontext * ctx, + numIndices + 3 /* DRAW_INDEX_IMMD */ + 2; /* test stamp */ - R700_CMDBUF_CHECK_SPACE(numEntires); + BEGIN_BATCH_NO_AUTOSTATE(numEntires); VGT_INDEX_TYPE |= DI_INDEX_SIZE_32_BIT << INDEX_TYPE_shift; - R700EP3(context, IT_INDEX_TYPE, 0); - R700E32(context, VGT_INDEX_TYPE); + OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); + OUT_BATCH(VGT_INDEX_TYPE); VGT_NUM_INDICES = numIndices; VGT_PRIMITIVE_TYPE |= r700PrimitiveType(prim) << VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift; - R700EP3(context, IT_SET_CONFIG_REG, 1); - R700E32(context, mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); - R700E32(context, VGT_PRIMITIVE_TYPE); + OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); + OUT_BATCH(VGT_PRIMITIVE_TYPE); VGT_DRAW_INITIATOR |= DI_SRC_SEL_IMMEDIATE << SOURCE_SELECT_shift; VGT_DRAW_INITIATOR |= DI_MAJOR_MODE_0 << MAJOR_MODE_shift; - R700EP3(context, IT_DRAW_INDEX_IMMD, (numIndices + 1)); - R700E32(context, VGT_NUM_INDICES); - R700E32(context, VGT_DRAW_INITIATOR); + OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1))); + OUT_BATCH(VGT_NUM_INDICES); + OUT_BATCH(VGT_DRAW_INITIATOR); for (j=0; jchipobj.FlushCmdBuffer)(context); - /* free aos => TODO : cache mgr */ - for (i = 0; i < context->aos_count; i++) - { - (context->chipobj.FreeDmaRegion)(context, &(context->aos[i])); - } -#endif // to be enable + (context->chipobj.ReleaseArrays)(ctx); + + //richard test + /* test stamp, write a number to mmSCRATCH4 */ + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL((0x2144 << 2), 0x56785678); + END_BATCH(); + COMMIT_BATCH(); + + rcommonFlushCmdBuf( &context->radeon, __FUNCTION__ ); + return GL_FALSE; } @@ -412,11 +373,19 @@ static GLboolean r700RunTCLRender(GLcontext * ctx, /*----------------------*/ { GLboolean bRet = GL_FALSE; -#if 0 // to be enable + /* TODO : sw fallback */ + + /** + * Ensure all enabled and complete textures are uploaded along with any buffers being used. + */ + if(!r700ValidateBuffers(ctx)) + { + return GL_TRUE; + } + context_t *context = R700_CONTEXT(ctx); r700UpdateShaders(ctx); -#endif // to be enable bRet = r700RunRender(ctx, stage); diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index 4341f33665..a31a89627c 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -52,10 +52,9 @@ #include "r700_chip.h" #include "r700_state.h" -#if 0 /* to be enabled */ #include "r700_fragprog.h" #include "r700_vertprog.h" -#endif /* to be enabled */ + void r700SetDefaultStates(context_t *context) //-------------------- { @@ -68,13 +67,13 @@ void r700UpdateShaders (GLcontext * ctx) //---------------------------------- GLvector4f dummy_attrib[_TNL_ATTRIB_MAX]; GLvector4f *temp_attrib[_TNL_ATTRIB_MAX]; -#if 0 /* to be enabled */ + struct r700_vertex_program *vp; int i; - if (context->NewGLState) + if (context->radeon.NewGLState) { - context->NewGLState = 0; + context->radeon.NewGLState = 0; for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) { @@ -108,7 +107,6 @@ void r700UpdateShaders (GLcontext * ctx) //---------------------------------- } r700UpdateStateParameters(ctx, _NEW_PROGRAM); -#endif /* to be enabled */ } /* @@ -116,6 +114,7 @@ void r700UpdateShaders (GLcontext * ctx) //---------------------------------- */ void r700UpdateViewportOffset(GLcontext * ctx) //------------------ { + return; } /** @@ -152,7 +151,6 @@ static void r700FetchStateParameter(GLcontext * ctx, void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state) //-------------------- { -#if 0 /* to be enabled */ struct r700_fragment_program *fp; struct gl_program_parameter_list *paramList; GLuint i; @@ -183,7 +181,6 @@ void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state) //------------ paramList->ParameterValues[i]); } } -#endif /* to be enabled */ } /** @@ -191,7 +188,6 @@ void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state) //------------ */ static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //------------------- { -#if 0 /* to be enabled */ context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); @@ -204,12 +200,16 @@ static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //----------- if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { + _mesa_update_framebuffer(ctx); + /* this updates the DrawBuffer's Width/Height if it's a FBO */ + _mesa_update_draw_buffer_bounds(ctx); + r700UpdateDrawBuffer(ctx); } r700UpdateStateParameters(ctx, new_state); - if(GL_TRUE == context->bEnablePerspective) + if(GL_TRUE == r700->bEnablePerspective) { /* Do scale XY and Z by 1/W0 for perspective correction on pos. For orthogonal case, set both to one. */ CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit); @@ -232,8 +232,7 @@ static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //----------- CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit); } - context->NewGLState |= new_state; -#endif /* to be enabled */ + context->radeon.NewGLState |= new_state; } static void r700SetDepthState(GLcontext * ctx) @@ -492,17 +491,11 @@ static void r700Viewport(GLcontext * ctx, GLsizei width, GLsizei height) //-------------------- { -#if 0 /* to be enabled */ context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); - context->vport_x = x; - context->vport_y = y; - context->vport_width = width; - context->vport_height= height; - - __DRIdrawablePrivate *dPriv = context->currentDraw; + __DRIdrawablePrivate *dPriv = context->radeon.dri.drawable; GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0; GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0; @@ -534,12 +527,12 @@ static void r700Viewport(GLcontext * ctx, tz = v[MAT_TZ] * scale; /* TODO : Need DMA flush as well. */ - +#if 0 /* to be enabled */ if(context->cmdbuf.count_used > 0) { (context->chipobj.FlushCmdBuffer)(context); } - +#endif /* to be enabled */ r700->PA_CL_VPORT_XSCALE.u32All = *((unsigned int*)(&sx)); r700->PA_CL_VPORT_XOFFSET.u32All = *((unsigned int*)(&tx)); @@ -548,7 +541,6 @@ static void r700Viewport(GLcontext * ctx, r700->PA_CL_VPORT_ZSCALE.u32All = *((unsigned int*)(&sz)); r700->PA_CL_VPORT_ZOFFSET.u32All = *((unsigned int*)(&tz)); -#endif /* to be enabled */ } @@ -600,60 +592,47 @@ static void r700Scissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h) } } - -/** - * Calculate initial hardware state and register state functions. - * Assumes that the command buffer and state atoms have been - * initialized already. - */ -void r700InitState(GLcontext * ctx) //------------------- +void r700SetRenderTarget(context_t *context) { -#if 0 /* to be enabled */ - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); - - if(context->ctx->Visual.doubleBufferMode && context->sarea->pfCurrentPage == 0) - { - context->target.rt = context->screen->backBuffer; - } - else - { - context->target.rt = context->screen->frontBuffer; - } + struct radeon_renderbuffer *rrb; + unsigned int nPitchInPixel; + + /* screen/window/view */ SETfield(r700->CB_TARGET_MASK.u32All, 0xF, TARGET0_ENABLE_shift, TARGET0_ENABLE_mask); SETfield(r700->CB_SHADER_MASK.u32All, 0xF, OUTPUT0_ENABLE_shift, OUTPUT0_ENABLE_mask); /* screen */ r700->PA_SC_SCREEN_SCISSOR_TL.u32All = 0x0; - SETfield(r700->PA_SC_SCREEN_SCISSOR_BR.u32All, context->screen->width, + + SETfield(r700->PA_SC_SCREEN_SCISSOR_BR.u32All, ((RADEONDRIPtr)(context->radeon.radeonScreen->driScreen->pDevPriv))->width, PA_SC_SCREEN_SCISSOR_BR__BR_X_shift, PA_SC_SCREEN_SCISSOR_BR__BR_X_mask); - SETfield(r700->PA_SC_SCREEN_SCISSOR_BR.u32All, context->screen->height, + SETfield(r700->PA_SC_SCREEN_SCISSOR_BR.u32All, ((RADEONDRIPtr)(context->radeon.radeonScreen->driScreen->pDevPriv))->height, PA_SC_SCREEN_SCISSOR_BR__BR_Y_shift, PA_SC_SCREEN_SCISSOR_BR__BR_Y_mask); /* window */ SETbit(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit); - SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, context->currentDraw->x, + SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, context->radeon.dri.drawable->x, PA_SC_WINDOW_SCISSOR_TL__TL_X_shift, PA_SC_WINDOW_SCISSOR_TL__TL_X_mask); - SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, context->currentDraw->y, + SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, context->radeon.dri.drawable->y, PA_SC_WINDOW_SCISSOR_TL__TL_Y_shift, PA_SC_WINDOW_SCISSOR_TL__TL_Y_mask); - SETfield(r700->PA_SC_WINDOW_SCISSOR_BR.u32All, context->currentDraw->x + context->currentDraw->w, + SETfield(r700->PA_SC_WINDOW_SCISSOR_BR.u32All, context->radeon.dri.drawable->x + context->radeon.dri.drawable->w, PA_SC_WINDOW_SCISSOR_BR__BR_X_shift, PA_SC_WINDOW_SCISSOR_BR__BR_X_mask); - SETfield(r700->PA_SC_WINDOW_SCISSOR_BR.u32All, context->currentDraw->y + context->currentDraw->h, + SETfield(r700->PA_SC_WINDOW_SCISSOR_BR.u32All, context->radeon.dri.drawable->y + context->radeon.dri.drawable->h, PA_SC_WINDOW_SCISSOR_BR__BR_Y_shift, PA_SC_WINDOW_SCISSOR_BR__BR_Y_mask); /* 4 clip rectangles */ /* TODO : set these clip rects according to context->currentDraw->numClipRects */ r700->PA_SC_CLIPRECT_RULE.u32All = 0x0000FFFF; - SETfield(r700->PA_SC_CLIPRECT_0_TL.u32All, context->currentDraw->x, + SETfield(r700->PA_SC_CLIPRECT_0_TL.u32All, context->radeon.dri.drawable->x, PA_SC_CLIPRECT_0_TL__TL_X_shift, PA_SC_CLIPRECT_0_TL__TL_X_mask); - SETfield(r700->PA_SC_CLIPRECT_0_TL.u32All, context->currentDraw->y, + SETfield(r700->PA_SC_CLIPRECT_0_TL.u32All, context->radeon.dri.drawable->y, PA_SC_CLIPRECT_0_TL__TL_Y_shift, PA_SC_CLIPRECT_0_TL__TL_Y_mask); - SETfield(r700->PA_SC_CLIPRECT_0_BR.u32All, context->currentDraw->x + context->currentDraw->w, + SETfield(r700->PA_SC_CLIPRECT_0_BR.u32All, context->radeon.dri.drawable->x + context->radeon.dri.drawable->w, PA_SC_CLIPRECT_0_BR__BR_X_shift, PA_SC_CLIPRECT_0_BR__BR_X_mask); - SETfield(r700->PA_SC_CLIPRECT_0_BR.u32All, context->currentDraw->y + context->currentDraw->h, + SETfield(r700->PA_SC_CLIPRECT_0_BR.u32All, context->radeon.dri.drawable->y + context->radeon.dri.drawable->h, PA_SC_CLIPRECT_0_BR__BR_Y_shift, PA_SC_CLIPRECT_0_BR__BR_Y_mask); r700->PA_SC_CLIPRECT_1_TL.u32All = r700->PA_SC_CLIPRECT_0_TL.u32All; @@ -665,42 +644,138 @@ void r700InitState(GLcontext * ctx) //------------------- /* more....2d clip */ SETbit(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit); - SETfield(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, context->currentDraw->x, + SETfield(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, context->radeon.dri.drawable->x, PA_SC_GENERIC_SCISSOR_TL__TL_X_shift, PA_SC_GENERIC_SCISSOR_TL__TL_X_mask); - SETfield(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, context->currentDraw->y, + SETfield(r700->PA_SC_GENERIC_SCISSOR_TL.u32All, context->radeon.dri.drawable->y, PA_SC_GENERIC_SCISSOR_TL__TL_Y_shift, PA_SC_GENERIC_SCISSOR_TL__TL_Y_mask); - SETfield(r700->PA_SC_GENERIC_SCISSOR_BR.u32All, context->currentDraw->x + context->currentDraw->w, + SETfield(r700->PA_SC_GENERIC_SCISSOR_BR.u32All, context->radeon.dri.drawable->x + context->radeon.dri.drawable->w, PA_SC_GENERIC_SCISSOR_BR__BR_X_shift, PA_SC_GENERIC_SCISSOR_BR__BR_X_mask); - SETfield(r700->PA_SC_GENERIC_SCISSOR_BR.u32All, context->currentDraw->y + context->currentDraw->h, + SETfield(r700->PA_SC_GENERIC_SCISSOR_BR.u32All, context->radeon.dri.drawable->y + context->radeon.dri.drawable->h, PA_SC_GENERIC_SCISSOR_BR__BR_Y_shift, PA_SC_GENERIC_SCISSOR_BR__BR_Y_mask); SETbit(r700->PA_SC_VPORT_SCISSOR_0_TL.u32All, WINDOW_OFFSET_DISABLE_bit); - SETfield(r700->PA_SC_VPORT_SCISSOR_0_TL.u32All, context->currentDraw->x, + SETfield(r700->PA_SC_VPORT_SCISSOR_0_TL.u32All, context->radeon.dri.drawable->x, PA_SC_VPORT_SCISSOR_0_TL__TL_X_shift, PA_SC_VPORT_SCISSOR_0_TL__TL_X_mask); - SETfield(r700->PA_SC_VPORT_SCISSOR_0_TL.u32All, context->currentDraw->y, + SETfield(r700->PA_SC_VPORT_SCISSOR_0_TL.u32All, context->radeon.dri.drawable->y, PA_SC_VPORT_SCISSOR_0_TL__TL_Y_shift, PA_SC_VPORT_SCISSOR_0_TL__TL_Y_mask); - SETfield(r700->PA_SC_VPORT_SCISSOR_0_BR.u32All, context->currentDraw->x + context->currentDraw->w, + SETfield(r700->PA_SC_VPORT_SCISSOR_0_BR.u32All, context->radeon.dri.drawable->x + context->radeon.dri.drawable->w, PA_SC_VPORT_SCISSOR_0_BR__BR_X_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_X_mask); - SETfield(r700->PA_SC_VPORT_SCISSOR_0_BR.u32All, context->currentDraw->y + context->currentDraw->h, + SETfield(r700->PA_SC_VPORT_SCISSOR_0_BR.u32All, context->radeon.dri.drawable->y + context->radeon.dri.drawable->h, PA_SC_VPORT_SCISSOR_0_BR__BR_Y_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_Y_mask); SETbit(r700->PA_SC_VPORT_SCISSOR_1_TL.u32All, WINDOW_OFFSET_DISABLE_bit); - SETfield(r700->PA_SC_VPORT_SCISSOR_1_TL.u32All, context->currentDraw->x, + SETfield(r700->PA_SC_VPORT_SCISSOR_1_TL.u32All, context->radeon.dri.drawable->x, PA_SC_VPORT_SCISSOR_0_TL__TL_X_shift, PA_SC_VPORT_SCISSOR_0_TL__TL_X_mask); - SETfield(r700->PA_SC_VPORT_SCISSOR_1_TL.u32All, context->currentDraw->y, + SETfield(r700->PA_SC_VPORT_SCISSOR_1_TL.u32All, context->radeon.dri.drawable->y, PA_SC_VPORT_SCISSOR_0_TL__TL_Y_shift, PA_SC_VPORT_SCISSOR_0_TL__TL_Y_mask); - SETfield(r700->PA_SC_VPORT_SCISSOR_1_BR.u32All, context->currentDraw->x + context->currentDraw->w, + SETfield(r700->PA_SC_VPORT_SCISSOR_1_BR.u32All, context->radeon.dri.drawable->x + context->radeon.dri.drawable->w, PA_SC_VPORT_SCISSOR_0_BR__BR_X_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_X_mask); - SETfield(r700->PA_SC_VPORT_SCISSOR_1_BR.u32All, context->currentDraw->y + context->currentDraw->h, + SETfield(r700->PA_SC_VPORT_SCISSOR_1_BR.u32All, context->radeon.dri.drawable->y + context->radeon.dri.drawable->h, PA_SC_VPORT_SCISSOR_0_BR__BR_Y_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_Y_mask); /* setup viewport */ - r700Viewport(ctx, + r700Viewport(GL_CONTEXT(context), 0, 0, - context->currentDraw->w, - context->currentDraw->h); + context->radeon.dri.drawable->w, + context->radeon.dri.drawable->h); + + rrb = radeon_get_colorbuffer(&context->radeon); + if (!rrb || !rrb->bo) { + fprintf(stderr, "no rrb\n"); + return; + } + + /* color buffer */ + r700->CB_COLOR0_BASE.u32All = context->radeon.state.color.draw_offset; + nPitchInPixel = rrb->pitch/rrb->cpp; + SETfield(r700->CB_COLOR0_SIZE.u32All, (nPitchInPixel/8)-1, + PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask); + SETfield(r700->CB_COLOR0_SIZE.u32All, ( (nPitchInPixel * rrb->base.Height)/64 )-1, + SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask); + r700->CB_COLOR0_BASE.u32All = 0; + SETfield(r700->CB_COLOR0_INFO.u32All, ENDIAN_NONE, ENDIAN_shift, ENDIAN_mask); + SETfield(r700->CB_COLOR0_INFO.u32All, ARRAY_LINEAR_GENERAL, + CB_COLOR0_INFO__ARRAY_MODE_shift, CB_COLOR0_INFO__ARRAY_MODE_mask); + if(4 == rrb->cpp) + { + SETfield(r700->CB_COLOR0_INFO.u32All, COLOR_8_8_8_8, + CB_COLOR0_INFO__FORMAT_shift, CB_COLOR0_INFO__FORMAT_mask); + SETfield(r700->CB_COLOR0_INFO.u32All, SWAP_ALT, COMP_SWAP_shift, COMP_SWAP_mask); + } + else + { + SETfield(r700->CB_COLOR0_INFO.u32All, COLOR_5_6_5, + CB_COLOR0_INFO__FORMAT_shift, CB_COLOR0_INFO__FORMAT_mask); + SETfield(r700->CB_COLOR0_INFO.u32All, SWAP_ALT_REV, + COMP_SWAP_shift, COMP_SWAP_mask); + } + SETbit(r700->CB_COLOR0_INFO.u32All, SOURCE_FORMAT_bit); + SETbit(r700->CB_COLOR0_INFO.u32All, BLEND_CLAMP_bit); + SETfield(r700->CB_COLOR0_INFO.u32All, NUMBER_UNORM, NUMBER_TYPE_shift, NUMBER_TYPE_mask); + + /* depth buf */ + r700->DB_DEPTH_SIZE.u32All = 0; + r700->DB_DEPTH_BASE.u32All = 0; + r700->DB_DEPTH_INFO.u32All = 0; + + r700->DB_DEPTH_CONTROL.u32All = 0; + r700->DB_DEPTH_CLEAR.u32All = 0x3F800000; + r700->DB_DEPTH_VIEW.u32All = 0; + r700->DB_RENDER_CONTROL.u32All = 0; + r700->DB_RENDER_OVERRIDE.u32All = 0; + SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIZ_ENABLE_shift, FORCE_HIZ_ENABLE_mask); + SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE0_shift, FORCE_HIS_ENABLE0_mask); + SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE1_shift, FORCE_HIS_ENABLE1_mask); + + rrb = radeon_get_depthbuffer(&context->radeon); + if (!rrb) + return; + + nPitchInPixel = rrb->pitch/rrb->cpp; + + SETfield(r700->DB_DEPTH_SIZE.u32All, (nPitchInPixel/8)-1, + PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask); + SETfield(r700->DB_DEPTH_SIZE.u32All, ( (nPitchInPixel * rrb->base.Height)/64 )-1, + SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask); /* size in pixel / 64 - 1 */ + + if(4 == rrb->cpp) + { + switch (GL_CONTEXT(context)->Visual.depthBits) + { + case 16: + case 24: + SETfield(r700->DB_DEPTH_INFO.u32All, DEPTH_8_24, + DB_DEPTH_INFO__FORMAT_shift, DB_DEPTH_INFO__FORMAT_mask); + break; + default: + fprintf(stderr, "Error: Unsupported depth %d... exiting\n", + GL_CONTEXT(context)->Visual.depthBits); + _mesa_exit(-1); + } + } + else + { + SETfield(r700->DB_DEPTH_INFO.u32All, DEPTH_16, + DB_DEPTH_INFO__FORMAT_shift, DB_DEPTH_INFO__FORMAT_mask); + } + SETfield(r700->DB_DEPTH_INFO.u32All, ARRAY_2D_TILED_THIN1, + DB_DEPTH_INFO__ARRAY_MODE_shift, DB_DEPTH_INFO__ARRAY_MODE_mask); + /* r700->DB_PREFETCH_LIMIT.bits.DEPTH_HEIGHT_TILE_MAX = (context->currentDraw->h >> 3) - 1; */ /* z buffer sie may much bigger than what need, so use actual used h. */ +} + +/** + * Calculate initial hardware state and register state functions. + * Assumes that the command buffer and state atoms have been + * initialized already. + */ +void r700InitState(GLcontext * ctx) //------------------- +{ + context_t *context = R700_CONTEXT(ctx); + + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + /* Turn off vgt reuse */ r700->VGT_REUSE_OFF.u32All = 0; SETbit(r700->VGT_REUSE_OFF.u32All, REUSE_OFF_bit); @@ -740,7 +815,7 @@ void r700InitState(GLcontext * ctx) //------------------- POLYMODE_BACK_PTYPE_shift, POLYMODE_BACK_PTYPE_mask); /* Do scale XY and Z by 1/W0. */ - context->bEnablePerspective = GL_TRUE; + r700->bEnablePerspective = GL_TRUE; CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit); CLEARbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit); SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit); @@ -811,35 +886,8 @@ void r700InitState(GLcontext * ctx) //------------------- /* depth buf */ r700->DB_DEPTH_SIZE.u32All = 0; - SETfield(r700->DB_DEPTH_SIZE.u32All, (context->screen->depthBuffer.pitch/8)-1, - PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask); - SETfield(r700->DB_DEPTH_SIZE.u32All, ( (context->screen->depthBuffer.size / context->screen->cpp)/64 )-1, - SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask); /* size in pixel / 64 - 1 */ - r700->DB_DEPTH_BASE.u32All = context->screen->depthBuffer.gpu >> 8; + r700->DB_DEPTH_BASE.u32All = 0; r700->DB_DEPTH_INFO.u32All = 0; - if(4 == context->screen->cpp) /* TODO : in scrren create, gives z its own format alloc. */ - { - switch (ctx->Visual.depthBits) - { - case 16: - case 24: - SETfield(r700->DB_DEPTH_INFO.u32All, DEPTH_8_24, - DB_DEPTH_INFO__FORMAT_shift, DB_DEPTH_INFO__FORMAT_mask); - break; - default: - fprintf(stderr, "Error: Unsupported depth %d... exiting\n", - ctx->Visual.depthBits); - _mesa_exit(-1); - } - } - else - { - SETfield(r700->DB_DEPTH_INFO.u32All, DEPTH_16, - DB_DEPTH_INFO__FORMAT_shift, DB_DEPTH_INFO__FORMAT_mask); - } - SETfield(r700->DB_DEPTH_INFO.u32All, ARRAY_2D_TILED_THIN1, - DB_DEPTH_INFO__ARRAY_MODE_shift, DB_DEPTH_INFO__ARRAY_MODE_mask); - /* r700->DB_PREFETCH_LIMIT.bits.DEPTH_HEIGHT_TILE_MAX = (context->currentDraw->h >> 3) - 1; */ /* z buffer sie may much bigger than what need, so use actual used h. */ r700->DB_DEPTH_CONTROL.u32All = 0; r700->DB_DEPTH_CLEAR.u32All = 0x3F800000; r700->DB_DEPTH_VIEW.u32All = 0; @@ -850,27 +898,9 @@ void r700InitState(GLcontext * ctx) //------------------- SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE1_shift, FORCE_HIS_ENABLE1_mask); /* color buffer */ - SETfield(r700->CB_COLOR0_SIZE.u32All, (context->screen->frontBuffer.pitch/8)-1, - PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask); - SETfield(r700->CB_COLOR0_SIZE.u32All, ( (context->screen->frontBuffer.size / context->screen->cpp)/64 )-1, - SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask); - r700->CB_COLOR0_BASE.u32All = context->screen->frontBuffer.gpu >> 8; - SETfield(r700->CB_COLOR0_INFO.u32All, ENDIAN_NONE, ENDIAN_shift, ENDIAN_mask); - SETfield(r700->CB_COLOR0_INFO.u32All, ARRAY_LINEAR_GENERAL, - CB_COLOR0_INFO__ARRAY_MODE_shift, CB_COLOR0_INFO__ARRAY_MODE_mask); - if(4 == context->screen->cpp) - { - SETfield(r700->CB_COLOR0_INFO.u32All, COLOR_8_8_8_8, - CB_COLOR0_INFO__FORMAT_shift, CB_COLOR0_INFO__FORMAT_mask); - SETfield(r700->CB_COLOR0_INFO.u32All, SWAP_ALT, COMP_SWAP_shift, COMP_SWAP_mask); - } - else - { - SETfield(r700->CB_COLOR0_INFO.u32All, COLOR_5_6_5, - CB_COLOR0_INFO__FORMAT_shift, CB_COLOR0_INFO__FORMAT_mask); - SETfield(r700->CB_COLOR0_INFO.u32All, SWAP_ALT_REV, - COMP_SWAP_shift, COMP_SWAP_mask); - } + r700->CB_COLOR0_SIZE.u32All = 0; + r700->CB_COLOR0_BASE.u32All = 0; + r700->CB_COLOR0_INFO.u32All = 0; SETbit(r700->CB_COLOR0_INFO.u32All, SOURCE_FORMAT_bit); SETbit(r700->CB_COLOR0_INFO.u32All, BLEND_CLAMP_bit); SETfield(r700->CB_COLOR0_INFO.u32All, NUMBER_UNORM, NUMBER_TYPE_shift, NUMBER_TYPE_mask); @@ -879,8 +909,7 @@ void r700InitState(GLcontext * ctx) //------------------- r700->CB_COLOR0_FRAG.u32All = 0; r700->CB_COLOR0_MASK.u32All = 0; - r700->PA_SC_VPORT_ZMAX_0.u32All = 0x3F800000; -#endif /* to be enabled */ + r700->PA_SC_VPORT_ZMAX_0.u32All = 0x3F800000; } void r700InitStateFuncs(struct dd_function_table *functions) //----------------- diff --git a/src/mesa/drivers/dri/r600/r700_state.h b/src/mesa/drivers/dri/r600/r700_state.h index b9542f4dd2..a75c7f0188 100644 --- a/src/mesa/drivers/dri/r600/r700_state.h +++ b/src/mesa/drivers/dri/r600/r700_state.h @@ -42,6 +42,7 @@ extern void r700UpdateDrawBuffer (GLcontext * ctx); extern void r700InitState (GLcontext * ctx); extern void r700InitStateFuncs (struct dd_function_table *functions); +extern void r700SetRenderTarget(context_t *context); extern void r700SetDefaultStates(context_t * context); #endif /* _R600_SCREEN_H */ diff --git a/src/mesa/drivers/dri/r600/r700_tex.c b/src/mesa/drivers/dri/r600/r700_tex.c index 75300d2c5f..85352022bc 100644 --- a/src/mesa/drivers/dri/r600/r700_tex.c +++ b/src/mesa/drivers/dri/r600/r700_tex.c @@ -42,9 +42,7 @@ #include "r600_context.h" #include "r700_chip.h" -#if 0 /* to be enabled */ #include "r700_state.h" -#endif /* to be enabled */ #include "r700_tex.h" @@ -72,7 +70,6 @@ void r700SetTexOffset(__DRIcontext *pDRICtx, GLint texname, { } -#if 0 /* to be enabled */ static GLboolean r700GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_format) { r700TexObjPtr t = (r700TexObjPtr) tObj->DriverData; @@ -548,7 +545,6 @@ static void compute_tex_image_offset( *curOffset += size; } -#endif /* to be enabled */ void r700DestroyTexObj(context_t context, r700TexObjPtr t) { @@ -563,9 +559,9 @@ void r700DestroyTexObj(context_t context, r700TexObjPtr t) //} } -#if 0 /* to be enabled */ static void r700SetTexImages(context_t *context, struct gl_texture_object *tObj) { +#if 0 /* to be enabled */ r700TexObjPtr t = (r700TexObjPtr) tObj->DriverData; const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel]; GLint curOffset; @@ -668,6 +664,7 @@ static void r700SetTexImages(context_t *context, struct gl_texture_object *tObj) TEX_WIDTH_shift, TEX_WIDTH_mask); SETfield(t->texture_state.SQ_TEX_RESOURCE1.u32All, tObj->Image[0][t->base.firstLevel]->Height - 1, TEX_HEIGHT_shift, TEX_HEIGHT_mask); +#endif /* to be enabled */ } static void r700UploadSubImage(context_t *context, @@ -677,6 +674,7 @@ static void r700UploadSubImage(context_t *context, GLint y, GLuint face) { +#if 0 /* to be enabled */ struct gl_texture_image *texImage = NULL; GLuint offset; GLint imageWidth, imageHeight; @@ -739,10 +737,12 @@ static void r700UploadSubImage(context_t *context, t->src_hight_in_pexel[face][hwlevel], /* height */ t->byte_per_texel, /* byte_per_pixel */ pSrc); /* source data */ +#endif /* to be enabled */ } int r700UploadTexImages(GLcontext * ctx, struct gl_texture_object *tObj, GLuint face) { +#if 0 /* to be enabled */ context_t *context = R700_CONTEXT(ctx); r700TexObjPtr t = (r700TexObjPtr) tObj->DriverData; @@ -823,7 +823,7 @@ int r700UploadTexImages(GLcontext * ctx, struct gl_texture_object *tObj, GLuint SETfield(t->texture_state.SQ_TEX_RESOURCE4.u32All, t->base.firstLevel, BASE_LEVEL_shift, BASE_LEVEL_mask); SETfield(t->texture_state.SQ_TEX_RESOURCE5.u32All, t->base.lastLevel, LAST_LEVEL_shift, LAST_LEVEL_mask); } - +#endif /* to be enabled */ return 0; } @@ -889,8 +889,6 @@ static const struct gl_texture_format *r700Choose8888TexFormat(GLenum srcFormat, return gtfRet; } -#endif /* to be enabled */ - static r700TexObjPtr r700AllocTexObj(struct gl_texture_object *texObj) { r700TexObjPtr t; @@ -1527,6 +1525,16 @@ void r700UpdateTextureState(context_t * context) #endif /* to be enabled */ } +/** + * Ensure all enabled and complete textures are uploaded along with any buffers being used. + */ +GLboolean r700ValidateBuffers(GLcontext * ctx) +{ + /* TODO */ + + return radeon_revalidate_bos(ctx); +} + void r700InitTextureFuncs(struct dd_function_table *functions) { /* Note: we only plug in the functions we implement in the driver diff --git a/src/mesa/drivers/dri/r600/r700_tex.h b/src/mesa/drivers/dri/r600/r700_tex.h index 13290578a1..e322bbc47d 100644 --- a/src/mesa/drivers/dri/r600/r700_tex.h +++ b/src/mesa/drivers/dri/r600/r700_tex.h @@ -97,6 +97,8 @@ extern void r700SetTexOffset(__DRIcontext *pDRICtx, extern void r700DestroyTexObj(context_t rmesa, r700TexObjPtr t); +extern GLboolean r700ValidateBuffers(GLcontext * ctx); + extern void r700InitTextureFuncs(struct dd_function_table *functions); #endif /* __r700_TEX_H__ */ diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index 22a12a00d1..f0dfbe20f0 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -39,15 +39,12 @@ #include "shader/prog_statevars.h" #include "r600_context.h" +#include "r600_cmdbuf.h" #include "r700_chip.h" #include "r700_debug.h" #include "r700_vertprog.h" -#if 0 /* to be enabled */ -#include "r700_emit.h" -#endif - unsigned int Map_Vertex_Output(r700_AssemblerBase *pAsm, struct gl_vertex_program *mesa_vp, unsigned int unStart) @@ -297,11 +294,10 @@ GLboolean r700TranslateVertexShader(struct r700_vertex_program *vp, void r700SelectVertexShader(GLcontext *ctx) { -#if 0 /* to be enabled */ context_t *context = R700_CONTEXT(ctx); struct r700_vertex_program *vpc = (struct r700_vertex_program *)ctx->VertexProgram._Current; - if (context->screen->chip.type <= CHIP_TYPE_RV670) + if (context->radeon.radeonScreen->chip_family <= CHIP_FAMILY_RV670) { vpc->r700AsmCode.bR6xx = 1; } @@ -327,56 +323,22 @@ void r700SelectVertexShader(GLcontext *ctx) r700TranslateVertexShader(vpc, &(vpc->mesa_program) ); } -#endif /* to be enabled */ } -void r700SetupVTXConstans(GLcontext * ctx, - unsigned int nStreamID, - unsigned int aos_offset, - unsigned int size, /* number of elements in vector */ - unsigned int stride, - unsigned int count) /* number of vectors in stream */ +void * r700GetActiveVpShaderBo(GLcontext * ctx) { - context_t *context = R700_CONTEXT(ctx); - uint32_t *dest; - - unsigned int uSQ_VTX_CONSTANT_WORD0_0; - unsigned int uSQ_VTX_CONSTANT_WORD1_0; - unsigned int uSQ_VTX_CONSTANT_WORD2_0 = 0; - unsigned int uSQ_VTX_CONSTANT_WORD3_0 = 0; - unsigned int uSQ_VTX_CONSTANT_WORD6_0 = 0; - - uSQ_VTX_CONSTANT_WORD0_0 = aos_offset; - uSQ_VTX_CONSTANT_WORD1_0 = count * stride - 1; - - uSQ_VTX_CONSTANT_WORD2_0 |= 0 << BASE_ADDRESS_HI_shift /* TODO */ - |stride << SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift - |GetSurfaceFormat(GL_FLOAT, size, NULL) << SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_shift /* TODO : trace back api for initial data type, not only GL_FLOAT */ - |SQ_NUM_FORMAT_SCALED << SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift - |SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit; - - uSQ_VTX_CONSTANT_WORD3_0 |= 1 << MEM_REQUEST_SIZE_shift; - - uSQ_VTX_CONSTANT_WORD6_0 |= SQ_TEX_VTX_VALID_BUFFER << SQ_TEX_RESOURCE_WORD6_0__TYPE_shift; -#if 0 /* to be enabled */ - R700_CMDBUF_CHECK_SPACE(9); - R700EP3 (context, IT_SET_RESOURCE, 7); - R700E32 (context, (nStreamID + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); - - R700E32 (context, uSQ_VTX_CONSTANT_WORD0_0); - R700E32 (context, uSQ_VTX_CONSTANT_WORD1_0); - R700E32 (context, uSQ_VTX_CONSTANT_WORD2_0); - R700E32 (context, uSQ_VTX_CONSTANT_WORD3_0); - R700E32 (context, 0); - R700E32 (context, 0); - R700E32 (context, uSQ_VTX_CONSTANT_WORD6_0); -#endif /* to be enabled */ + struct r700_vertex_program *vp + = (struct r700_vertex_program *)ctx->VertexProgram._Current; + + return vp->shaderbo; } GLboolean r700SetupVertexProgram(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); + BATCH_LOCALS(&context->radeon); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); struct r700_vertex_program *vp @@ -396,7 +358,7 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) /* Load vp to gpu */ (context->chipobj.EmitShader)(ctx, - &(vp->shadercode), + &(vp->shaderbo), (GLvoid *)(vp->r700Shader.pProgram), vp->r700Shader.uShaderBinaryDWORDSize); @@ -410,7 +372,7 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) (context->chipobj.MemUse)(context, vp->shadercode.buf->id); */ - r700->SQ_PGM_START_VS.u32All = (vp->shadercode.aos_offset >> 8) & 0x00FFFFFF; + r700->SQ_PGM_START_VS.u32All = 0; /* set from buffer object. */ SETfield(r700->SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.nRegs + 1, NUM_GPRS_shift, NUM_GPRS_mask); @@ -438,24 +400,26 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) if(NULL != paramList) { _mesa_load_state_parameters(ctx, paramList); -#if 0 /* to be enabled */ + unNumParamData = paramList->NumParameters * 4; - R700_CMDBUF_CHECK_SPACE(unNumParamData + 2); - R700EP3 (context, IT_SET_ALU_CONST, unNumParamData); + BEGIN_BATCH_NO_AUTOSTATE(unNumParamData + 2); + + OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData)); /* assembler map const from very beginning. */ - R700E32 (context, SQ_ALU_CONSTANT_VS_OFFSET * 4); + OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); unNumParamData = paramList->NumParameters; for(ui=0; uiParameterValues[ui][0]))); - R700E32 (context, *((unsigned int*)&(paramList->ParameterValues[ui][1]))); - R700E32 (context, *((unsigned int*)&(paramList->ParameterValues[ui][2]))); - R700E32 (context, *((unsigned int*)&(paramList->ParameterValues[ui][3]))); + OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][0]))); + OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1]))); + OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2]))); + OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3]))); } -#endif /* to be enabled */ + END_BATCH(); + COMMIT_BATCH(); } return GL_TRUE; diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.h b/src/mesa/drivers/dri/r600/r700_vertprog.h index 0822335c10..6a9726a3d0 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.h +++ b/src/mesa/drivers/dri/r600/r700_vertprog.h @@ -55,7 +55,8 @@ struct r700_vertex_program /* ... */ - struct r600_dma_region shadercode; + void * shaderbo; + ArrayDesc aos_desc[VERT_ATTRIB_MAX]; }; @@ -82,11 +83,6 @@ extern void r700SelectVertexShader(GLcontext *ctx); extern GLboolean r700SetupVertexProgram(GLcontext * ctx); -extern void r700SetupVTXConstans(GLcontext * ctx, - unsigned int nStreamID, - unsigned int aos_offset, - unsigned int size, /* number of elements in vector */ - unsigned int stride, - unsigned int Count); /* number of vectors in stream */ +extern void * r700GetActiveVpShaderBo(GLcontext * ctx); #endif /* _R700_VERTPROG_H_ */ diff --git a/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h b/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h index 851b488433..abb023c7de 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h +++ b/src/mesa/drivers/dri/radeon/radeon_cmdbuf.h @@ -16,85 +16,7 @@ void rcommonBeginBatch(radeonContextPtr rmesa, const char *function, int line); -#define RADEON_CP_PACKET3_NOP 0xC0001000 -#define RADEON_CP_PACKET3_NEXT_CHAR 0xC0001900 -#define RADEON_CP_PACKET3_PLY_NEXTSCAN 0xC0001D00 -#define RADEON_CP_PACKET3_SET_SCISSORS 0xC0001E00 -#define RADEON_CP_PACKET3_3D_RNDR_GEN_INDX_PRIM 0xC0002300 -#define RADEON_CP_PACKET3_LOAD_MICROCODE 0xC0002400 -#define RADEON_CP_PACKET3_WAIT_FOR_IDLE 0xC0002600 -#define RADEON_CP_PACKET3_3D_DRAW_VBUF 0xC0002800 -#define RADEON_CP_PACKET3_3D_DRAW_IMMD 0xC0002900 -#define RADEON_CP_PACKET3_3D_DRAW_INDX 0xC0002A00 -#define RADEON_CP_PACKET3_LOAD_PALETTE 0xC0002C00 -#define RADEON_CP_PACKET3_3D_LOAD_VBPNTR 0xC0002F00 -#define RADEON_CP_PACKET3_CNTL_PAINT 0xC0009100 -#define RADEON_CP_PACKET3_CNTL_BITBLT 0xC0009200 -#define RADEON_CP_PACKET3_CNTL_SMALLTEXT 0xC0009300 -#define RADEON_CP_PACKET3_CNTL_HOSTDATA_BLT 0xC0009400 -#define RADEON_CP_PACKET3_CNTL_POLYLINE 0xC0009500 -#define RADEON_CP_PACKET3_CNTL_POLYSCANLINES 0xC0009800 -#define RADEON_CP_PACKET3_CNTL_PAINT_MULTI 0xC0009A00 -#define RADEON_CP_PACKET3_CNTL_BITBLT_MULTI 0xC0009B00 -#define RADEON_CP_PACKET3_CNTL_TRANS_BITBLT 0xC0009C00 - -/* r6xx/r7xx packet 3 type offsets */ -#define R600_SET_CONFIG_REG_OFFSET 0x00008000 -#define R600_SET_CONFIG_REG_END 0x0000ac00 -#define R600_SET_CONTEXT_REG_OFFSET 0x00028000 -#define R600_SET_CONTEXT_REG_END 0x00029000 -#define R600_SET_ALU_CONST_OFFSET 0x00030000 -#define R600_SET_ALU_CONST_END 0x00032000 -#define R600_SET_RESOURCE_OFFSET 0x00038000 -#define R600_SET_RESOURCE_END 0x0003c000 -#define R600_SET_SAMPLER_OFFSET 0x0003c000 -#define R600_SET_SAMPLER_END 0x0003cff0 -#define R600_SET_CTL_CONST_OFFSET 0x0003cff0 -#define R600_SET_CTL_CONST_END 0x0003e200 -#define R600_SET_LOOP_CONST_OFFSET 0x0003e200 -#define R600_SET_LOOP_CONST_END 0x0003e380 -#define R600_SET_BOOL_CONST_OFFSET 0x0003e380 -#define R600_SET_BOOL_CONST_END 0x00040000 - -/* r6xx/r7xx packet 3 types */ -#define R600_IT_INDIRECT_BUFFER_END 0x00001700 -#define R600_IT_SET_PREDICATION 0x00002000 -#define R600_IT_REG_RMW 0x00002100 -#define R600_IT_COND_EXEC 0x00002200 -#define R600_IT_PRED_EXEC 0x00002300 -#define R600_IT_START_3D_CMDBUF 0x00002400 -#define R600_IT_DRAW_INDEX_2 0x00002700 -#define R600_IT_CONTEXT_CONTROL 0x00002800 -#define R600_IT_DRAW_INDEX_IMMD_BE 0x00002900 -#define R600_IT_INDEX_TYPE 0x00002A00 -#define R600_IT_DRAW_INDEX 0x00002B00 -#define R600_IT_DRAW_INDEX_AUTO 0x00002D00 -#define R600_IT_DRAW_INDEX_IMMD 0x00002E00 -#define R600_IT_NUM_INSTANCES 0x00002F00 -#define R600_IT_STRMOUT_BUFFER_UPDATE 0x00003400 -#define R600_IT_INDIRECT_BUFFER_MP 0x00003800 -#define R600_IT_MEM_SEMAPHORE 0x00003900 -#define R600_IT_MPEG_INDEX 0x00003A00 -#define R600_IT_WAIT_REG_MEM 0x00003C00 -#define R600_IT_MEM_WRITE 0x00003D00 -#define R600_IT_INDIRECT_BUFFER 0x00003200 -#define R600_IT_CP_INTERRUPT 0x00004000 -#define R600_IT_SURFACE_SYNC 0x00004300 -#define R600_IT_ME_INITIALIZE 0x00004400 -#define R600_IT_COND_WRITE 0x00004500 -#define R600_IT_EVENT_WRITE 0x00004600 -#define R600_IT_EVENT_WRITE_EOP 0x00004700 -#define R600_IT_ONE_REG_WRITE 0x00005700 -#define R600_IT_SET_CONFIG_REG 0x00006800 -#define R600_IT_SET_CONTEXT_REG 0x00006900 -#define R600_IT_SET_ALU_CONST 0x00006A00 -#define R600_IT_SET_BOOL_CONST 0x00006B00 -#define R600_IT_SET_LOOP_CONST 0x00006C00 -#define R600_IT_SET_RESOURCE 0x00006D00 -#define R600_IT_SET_SAMPLER 0x00006E00 -#define R600_IT_SET_CTL_CONST 0x00006F00 -#define R600_IT_SURFACE_BASE_UPDATE 0x00007300 - +/* +r6/r7 : code here moved */ #define CP_PACKET2 (2 << 30) #define CP_PACKET0(reg, n) (RADEON_CP_PACKET0 | ((n)<<16) | ((reg)>>2)) @@ -188,54 +110,14 @@ void rcommonBeginBatch(radeonContextPtr rmesa, #define OUT_BATCH_FLOAT32(f) \ OUT_BATCH(radeonPackFloat32((f))) -/* R600/R700 */ -#define R600_OUT_BATCH_REGS(reg, num) \ -do { \ - if ((reg) >= R600_SET_CONFIG_REG_OFFSET && (reg) < R600_SET_CONFIG_REG_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, (num))); \ - OUT_BATCH(((reg) - R600_SET_CONFIG_REG_OFFSET) >> 2); \ - } else if ((reg) >= R600_SET_CONTEXT_REG_OFFSET && (reg) < R600_SET_CONTEXT_REG_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_CONTEXT_REG, (num))); \ - OUT_BATCH(((reg) - R600_SET_CONTEXT_REG_OFFSET) >> 2); \ - } else if ((reg) >= R600_SET_ALU_CONST_OFFSET && (reg) < R600_SET_ALU_CONST_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (num))); \ - OUT_BATCH(((reg) - R600_SET_ALU_CONST_OFFSET) >> 2); \ - } else if ((reg) >= R600_SET_RESOURCE_OFFSET && (reg) < R600_SET_RESOURCE_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, (num))); \ - OUT_BATCH(((reg) - R600_SET_RESOURCE_OFFSET) >> 2); \ - } else if ((reg) >= R600_SET_SAMPLER_OFFSET && (reg) < R600_SET_SAMPLER_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, (num))); \ - OUT_BATCH(((reg) - R600_SET_SAMPLER_OFFSET) >> 2); \ - } else if ((reg) >= R600_SET_CTL_CONST_OFFSET && (reg) < R600_SET_CTL_CONST_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, (num))); \ - OUT_BATCH(((reg) - R600_SET_CTL_CONST_OFFSET) >> 2); \ - } else if ((reg) >= R600_SET_LOOP_CONST_OFFSET && (reg) < R600_SET_LOOP_CONST_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_LOOP_CONST, (num))); \ - OUT_BATCH(((reg) - R600_SET_LOOP_CONST_OFFSET) >> 2); \ - } else if ((reg) >= R600_SET_BOOL_CONST_OFFSET && (reg) < R600_SET_BOOL_CONST_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_BOOL_CONST, (num))); \ - OUT_BATCH(((reg) - R600_SET_BOOL_CONST_OFFSET) >> 2); \ - } else { \ - OUT_BATCH(cmdpacket0(b_l_rmesa->radeonScreen, (reg), (num))); \ - } \ -} while (0) - -/** Single register write to command buffer; requires 3 dwords for most things. */ -#define R600_OUT_BATCH_REGVAL(reg, val) \ - R600_OUT_BATCH_REGS((reg), 1); \ - OUT_BATCH((val)) - -/** Continuous register range write to command buffer; requires 1 dword, - * expects count dwords afterwards for register contents. */ -#define R600_OUT_BATCH_REGSEQ(reg, count) \ - R600_OUT_BATCH_REGS((reg), (count)) +/* +r6/r7 : code here moved */ /* Fire the buffered vertices no matter what. */ static INLINE void radeon_firevertices(radeonContextPtr radeon) { if (radeon->cmdbuf.cs->cdw || radeon->dma.flush ) - radeonFlush(radeon->glCtx); + radeon->glCtx->Driver.Flush(radeon->glCtx); /* +r6/r7 */ } #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index 756c09fff3..691615938f 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -914,7 +914,7 @@ again: if (ret == RADEON_CS_SPACE_OP_TO_BIG) return GL_FALSE; if (ret == RADEON_CS_SPACE_FLUSH) { - radeonFlush(ctx); + ctx->Driver.Flush(ctx); /* +r6/r7 */ if (flushed) return GL_FALSE; flushed = 1; @@ -1014,7 +1014,7 @@ void radeonFinish(GLcontext * ctx) struct gl_framebuffer *fb = ctx->DrawBuffer; int i; - radeonFlush(ctx); + ctx->Driver.Flush(ctx); /* +r6/r7 */ if (radeon->radeonScreen->kernel_mm) { for (i = 0; i < fb->_NumColorDrawBuffers; i++) { diff --git a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c index b47b095cf2..b33417e93e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_cs_legacy.c @@ -216,22 +216,28 @@ static int cs_process_relocs(struct radeon_cs *cs) csm = (struct cs_manager_legacy*)cs->csm; relocs = (struct cs_reloc_legacy *)cs->relocs; - restart: - for (i = 0; i < cs->crelocs; i++) { - for (j = 0; j < relocs[i].cindices; j++) { +restart: + for (i = 0; i < cs->crelocs; i++) + { + for (j = 0; j < relocs[i].cindices; j++) + { uint32_t soffset, eoffset; r = radeon_bo_legacy_validate(relocs[i].base.bo, &soffset, &eoffset); - if (r == -EAGAIN) - goto restart; - if (r) { + if (r == -EAGAIN) + { + goto restart; + } + if (r) + { fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n", relocs[i].base.bo, soffset, eoffset); return r; } cs->packets[relocs[i].indices[j]] += soffset; - if (cs->packets[relocs[i].indices[j]] >= eoffset) { + if (cs->packets[relocs[i].indices[j]] >= eoffset) + { /* radeon_bo_debug(relocs[i].base.bo, 12); */ fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n", relocs[i].base.bo, soffset, eoffset); @@ -280,7 +286,8 @@ static int cs_emit(struct radeon_cs *cs) csm->ctx->vtbl.emit_cs_header(cs, csm->ctx); /* append buffer age */ - if (IS_R300_CLASS(csm->ctx->radeonScreen)) { + if ( IS_R300_CLASS(csm->ctx->radeonScreen) ) + { age.scratch.cmd_type = R300_CMD_SCRATCH; /* Scratch register 2 corresponds to what radeonGetAge polls */ csm->pending_age = 0; @@ -315,7 +322,8 @@ static int cs_emit(struct radeon_cs *cs) if (r) { return r; } - if (!IS_R300_CLASS(csm->ctx->radeonScreen)) { + if ((!IS_R300_CLASS(csm->ctx->radeonScreen)) && + (!IS_R600_CLASS(csm->ctx->radeonScreen))) { /* +r6/r7 : No irq for r6/r7 yet. */ drm_radeon_irq_emit_t emit_cmd; emit_cmd.irq_seq = &csm->pending_age; r = drmCommandWrite(cs->csm->fd, DRM_RADEON_IRQ_EMIT, &emit_cmd, sizeof(emit_cmd)); @@ -387,59 +395,71 @@ static int cs_check_space(struct radeon_cs *cs, struct radeon_cs_space_check *bo return 0; /* prepare */ - for (i = 0; i < num_bo; i++) { - bo = bos[i].bo; - - bos[i].new_accounted = 0; - read_domains = bos[i].read_domains; - write_domain = bos[i].write_domain; - - /* pinned bos don't count */ - if (radeon_legacy_bo_is_static(bo)) - continue; + for (i = 0; i < num_bo; i++) + { + bo = bos[i].bo; + + bos[i].new_accounted = 0; + read_domains = bos[i].read_domains; + write_domain = bos[i].write_domain; + + /* pinned bos don't count */ + if (radeon_legacy_bo_is_static(bo)) + continue; - /* already accounted this bo */ - if (write_domain && (write_domain == bo->space_accounted)) - continue; + /* already accounted this bo */ + if (write_domain && (write_domain == bo->space_accounted)) + continue; - if (read_domains && ((read_domains << 16) == bo->space_accounted)) - continue; + if (read_domains && ((read_domains << 16) == bo->space_accounted)) + continue; - if (bo->space_accounted == 0) { - if (write_domain == RADEON_GEM_DOMAIN_VRAM) - this_op_vram_write += bo->size; - else if (write_domain == RADEON_GEM_DOMAIN_GTT) - this_op_gart_write += bo->size; - else - this_op_read += bo->size; - bos[i].new_accounted = (read_domains << 16) | write_domain; - } else { - uint16_t old_read, old_write; - - old_read = bo->space_accounted >> 16; - old_write = bo->space_accounted & 0xffff; - - if (write_domain && (old_read & write_domain)) { - bos[i].new_accounted = write_domain; - /* moving from read to a write domain */ - if (write_domain == RADEON_GEM_DOMAIN_VRAM) { - this_op_read -= bo->size; - this_op_vram_write += bo->size; - } else if (write_domain == RADEON_GEM_DOMAIN_VRAM) { - this_op_read -= bo->size; - this_op_gart_write += bo->size; - } - } else if (read_domains & old_write) { - bos[i].new_accounted = bo->space_accounted & 0xffff; - } else { - /* rewrite the domains */ - if (write_domain != old_write) - fprintf(stderr,"WRITE DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, write_domain, old_write); - if (read_domains != old_read) - fprintf(stderr,"READ DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, read_domains, old_read); - return RADEON_CS_SPACE_FLUSH; - } - } + if (bo->space_accounted == 0) + { + if (write_domain == RADEON_GEM_DOMAIN_VRAM) + this_op_vram_write += bo->size; + else if (write_domain == RADEON_GEM_DOMAIN_GTT) + this_op_gart_write += bo->size; + else + this_op_read += bo->size; + bos[i].new_accounted = (read_domains << 16) | write_domain; + } + else + { + uint16_t old_read, old_write; + + old_read = bo->space_accounted >> 16; + old_write = bo->space_accounted & 0xffff; + + if (write_domain && (old_read & write_domain)) + { + bos[i].new_accounted = write_domain; + /* moving from read to a write domain */ + if (write_domain == RADEON_GEM_DOMAIN_VRAM) + { + this_op_read -= bo->size; + this_op_vram_write += bo->size; + } + else if (write_domain == RADEON_GEM_DOMAIN_VRAM) + { + this_op_read -= bo->size; + this_op_gart_write += bo->size; + } + } + else if (read_domains & old_write) + { + bos[i].new_accounted = bo->space_accounted & 0xffff; + } + else + { + /* rewrite the domains */ + if (write_domain != old_write) + fprintf(stderr,"WRITE DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, write_domain, old_write); + if (read_domains != old_read) + fprintf(stderr,"READ DOMAIN RELOC FAILURE 0x%x %d %d\n", bo->handle, read_domains, old_read); + return RADEON_CS_SPACE_FLUSH; + } + } } if (this_op_read < 0) diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c index f62ca7f9eb..b2585eae77 100644 --- a/src/mesa/drivers/dri/radeon/radeon_fbo.c +++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c @@ -165,7 +165,7 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, return GL_FALSE; } - radeonFlush(ctx); + ctx->Driver.Flush(ctx); /* +r6/r7 */ if (rrb->bo) radeon_bo_unref(rrb->bo); @@ -371,7 +371,7 @@ radeon_framebuffer_renderbuffer(GLcontext * ctx, GLenum attachment, struct gl_renderbuffer *rb) { - radeonFlush(ctx); + ctx->Driver.Flush(ctx); /* +r6/r7 */ _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb); radeon_draw_buffer(ctx, fb); diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index b75db5b80c..be3b816b86 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -61,8 +61,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_tex.h" #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) #include "r600_context.h" -#include "r700_fragprog.h" -#include "r700_tex.h" +//#include "r700_driconf.h" /* +r6/r7 */ +#include "r700_tex.h" /* +r6/r7 */ #endif #include "utils.h" @@ -405,13 +405,13 @@ static const __DRItexBufferExtension r300TexBufferExtension = { #if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) static const __DRItexOffsetExtension r600texOffsetExtension = { { __DRI_TEX_OFFSET, __DRI_TEX_OFFSET_VERSION }, - r700SetTexOffset, + r700SetTexOffset, /* +r6/r7 */ }; static const __DRItexBufferExtension r600TexBufferExtension = { { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION }, - r700SetTexBuffer, - r700SetTexBuffer2, + r700SetTexBuffer, /* +r6/r7 */ + r700SetTexBuffer2, /* +r6/r7 */ }; #endif @@ -941,7 +941,6 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv ) #if DO_DEBUG && RADEON_COMMON && defined(RADEON_COMMON_FOR_R300) RADEON_DEBUG = driParseDebugString(getenv("RADEON_DEBUG"), debug_control); #endif - /* parse information in __driConfigOptions */ driParseOptionInfo (&screen->optionCache, __driConfigOptions, __driNConfigOptions); -- cgit v1.2.3 From 2a7a2c6b77116fd475e1d05fb0adbd1702d11f77 Mon Sep 17 00:00:00 2001 From: Richard Li Date: Fri, 15 May 2009 19:37:09 -0400 Subject: Fix r6 code bugs. --- src/mesa/drivers/dri/r600/r600_cmdbuf.c | 19 +-- src/mesa/drivers/dri/r600/r600_cmdbuf.h | 36 ++--- src/mesa/drivers/dri/r600/r600_context.c | 12 ++ src/mesa/drivers/dri/r600/r600_context.h | 5 +- src/mesa/drivers/dri/r600/r600_emit.c | 15 +- src/mesa/drivers/dri/r600/r600_emit.h | 3 +- src/mesa/drivers/dri/r600/r600_swtcl.c | 6 +- src/mesa/drivers/dri/r600/r700_chip.c | 153 ++++++++++++++------- src/mesa/drivers/dri/r600/r700_chip.h | 4 +- src/mesa/drivers/dri/r600/r700_clear.c | 62 +++++---- src/mesa/drivers/dri/r600/r700_fragprog.c | 15 +- src/mesa/drivers/dri/r600/r700_ioctl.c | 2 +- src/mesa/drivers/dri/r600/r700_render.c | 67 +++++---- src/mesa/drivers/dri/r600/r700_state.c | 6 +- src/mesa/drivers/dri/r600/r700_vertprog.c | 15 +- .../drivers/dri/radeon/radeon_common_context.c | 12 ++ src/mesa/drivers/dri/radeon/radeon_screen.c | 40 ++++-- 17 files changed, 299 insertions(+), 173 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index 8550457566..b1f78388e4 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -188,6 +188,7 @@ int r600_cs_write_reloc(struct radeon_cs *cs, cs->crelocs++; radeon_bo_ref(bo); + return 0; } @@ -245,6 +246,8 @@ static int r600_cs_end(struct radeon_cs *cs, { fprintf(stderr, "CS section size missmatch start at (%s,%s,%d) %d vs %d\n", cs->section_file, cs->section_func, cs->section_line, cs->section_ndw, cs->section_cdw); + fprintf(stderr, "cs->section_ndw = %d, cs->cdw = %d, cs->section_cdw = %d \n", + cs->section_ndw, cs->cdw, cs->section_cdw); fprintf(stderr, "CS section end at (%s,%s,%d)\n", file, func, line); return -EPIPE; @@ -356,7 +359,7 @@ static int r600_cs_emit(struct radeon_cs *cs) /* drm_r300_cmd_header_t age; */ uint32_t length_dw_reloc_chunk; uint64_t ull; - uint64_t * chunk_ptrs[2]; + uint64_t chunk_ptrs[2]; uint32_t reloc_chunk[128]; int r; int retry = 0; @@ -370,23 +373,23 @@ static int r600_cs_emit(struct radeon_cs *cs) if (r) { return 0; } - + /* raw ib chunk */ cs_chunk[0].chunk_id = RADEON_CHUNK_ID_IB; cs_chunk[0].length_dw = cs->cdw; - cs_chunk[0].chunk_data = (uint64_t)(cs->packets); + cs_chunk[0].chunk_data = (unsigned long)(cs->packets); /* reloc chaunk */ cs_chunk[1].chunk_id = RADEON_CHUNK_ID_RELOCS; cs_chunk[1].length_dw = length_dw_reloc_chunk; - cs_chunk[1].chunk_data = (uint64_t)&(reloc_chunk[0]); + cs_chunk[1].chunk_data = (unsigned long)&(reloc_chunk[0]); - chunk_ptrs[0] = (uint64_t * )&(cs_chunk[0]); - chunk_ptrs[1] = (uint64_t * )&(cs_chunk[1]); + chunk_ptrs[0] = (uint64_t)(unsigned long)&(cs_chunk[0]); + chunk_ptrs[1] = (uint64_t)(unsigned long)&(cs_chunk[1]); cs_cmd.num_chunks = 2; - cs_cmd.cs_id = 0; - cs_cmd.chunks = (uint64_t)&(chunk_ptrs[0]); + /* cs_cmd.cs_id = 0; */ + cs_cmd.chunks = (uint64_t)(unsigned long)chunk_ptrs; /* dump_cmdbuf(cs); */ diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.h b/src/mesa/drivers/dri/r600/r600_cmdbuf.h index dbb723b2b5..9ad73d9d3a 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.h +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.h @@ -167,38 +167,38 @@ static inline void r600_cs_write_dword(struct radeon_cs *cs, uint32_t dword) #define R600_OUT_BATCH_REGS(reg, num) \ do { \ if ((reg) >= R600_SET_CONFIG_REG_OFFSET && (reg) < R600_SET_CONFIG_REG_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, (num))); \ - OUT_BATCH(((reg) - R600_SET_CONFIG_REG_OFFSET) >> 2); \ + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, (num))); \ + R600_OUT_BATCH(((reg) - R600_SET_CONFIG_REG_OFFSET) >> 2); \ } else if ((reg) >= R600_SET_CONTEXT_REG_OFFSET && (reg) < R600_SET_CONTEXT_REG_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_CONTEXT_REG, (num))); \ - OUT_BATCH(((reg) - R600_SET_CONTEXT_REG_OFFSET) >> 2); \ + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONTEXT_REG, (num))); \ + R600_OUT_BATCH(((reg) - R600_SET_CONTEXT_REG_OFFSET) >> 2); \ } else if ((reg) >= R600_SET_ALU_CONST_OFFSET && (reg) < R600_SET_ALU_CONST_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (num))); \ - OUT_BATCH(((reg) - R600_SET_ALU_CONST_OFFSET) >> 2); \ + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (num))); \ + R600_OUT_BATCH(((reg) - R600_SET_ALU_CONST_OFFSET) >> 2); \ } else if ((reg) >= R600_SET_RESOURCE_OFFSET && (reg) < R600_SET_RESOURCE_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, (num))); \ - OUT_BATCH(((reg) - R600_SET_RESOURCE_OFFSET) >> 2); \ + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, (num))); \ + R600_OUT_BATCH(((reg) - R600_SET_RESOURCE_OFFSET) >> 2); \ } else if ((reg) >= R600_SET_SAMPLER_OFFSET && (reg) < R600_SET_SAMPLER_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, (num))); \ - OUT_BATCH(((reg) - R600_SET_SAMPLER_OFFSET) >> 2); \ + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_SAMPLER, (num))); \ + R600_OUT_BATCH(((reg) - R600_SET_SAMPLER_OFFSET) >> 2); \ } else if ((reg) >= R600_SET_CTL_CONST_OFFSET && (reg) < R600_SET_CTL_CONST_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, (num))); \ - OUT_BATCH(((reg) - R600_SET_CTL_CONST_OFFSET) >> 2); \ + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, (num))); \ + R600_OUT_BATCH(((reg) - R600_SET_CTL_CONST_OFFSET) >> 2); \ } else if ((reg) >= R600_SET_LOOP_CONST_OFFSET && (reg) < R600_SET_LOOP_CONST_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_LOOP_CONST, (num))); \ - OUT_BATCH(((reg) - R600_SET_LOOP_CONST_OFFSET) >> 2); \ + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_LOOP_CONST, (num))); \ + R600_OUT_BATCH(((reg) - R600_SET_LOOP_CONST_OFFSET) >> 2); \ } else if ((reg) >= R600_SET_BOOL_CONST_OFFSET && (reg) < R600_SET_BOOL_CONST_END) { \ - OUT_BATCH(CP_PACKET3(R600_IT_SET_BOOL_CONST, (num))); \ - OUT_BATCH(((reg) - R600_SET_BOOL_CONST_OFFSET) >> 2); \ + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_BOOL_CONST, (num))); \ + R600_OUT_BATCH(((reg) - R600_SET_BOOL_CONST_OFFSET) >> 2); \ } else { \ - OUT_BATCH(CP_PACKET0((reg), (num))); \ + R600_OUT_BATCH(CP_PACKET0((reg), (num))); \ } \ } while (0) /** Single register write to command buffer; requires 3 dwords for most things. */ #define R600_OUT_BATCH_REGVAL(reg, val) \ R600_OUT_BATCH_REGS((reg), 1); \ - OUT_BATCH((val)) + R600_OUT_BATCH((val)) /** Continuous register range write to command buffer; requires 1 dword, * expects count dwords afterwards for register contents. */ diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index 3c8ba003d4..e6c03ecfd5 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -448,3 +448,15 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, return GL_TRUE; } +/* Clean our own things only, radeonDestroyContext will do every thing else. */ +void +r600DestroyContext (__DRIcontextPrivate * driContextPriv) +{ + GET_CURRENT_CONTEXT (ctx); + context_t *context = R700_CONTEXT(ctx); + + (context->chipobj.DestroyChipObj)(context->chipobj.pvChipObj); +} + + + diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index 70108b74c4..ae80037602 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -159,7 +159,7 @@ typedef struct chip_object void *pvChipObj; /* ------------ OUT ------------------- */ - GLboolean (*DestroyChipObj)(void* pvChipObj); + GLboolean (*DestroyChipObj)(GLcontext * ctx); void (*InitFuncs)(struct dd_function_table *functions); @@ -171,7 +171,8 @@ typedef struct chip_object GLboolean (*EmitShader)( GLcontext * ctx, void ** shaderbo, GLvoid * data, - int sizeinDWORD); + int sizeinDWORD, + char * szShaderUsage); GLboolean (*DeleteShader)(GLcontext * ctx, void * shaderbo); void (*FreeDmaRegion)( GLcontext * ctx, diff --git a/src/mesa/drivers/dri/r600/r600_emit.c b/src/mesa/drivers/dri/r600/r600_emit.c index fc8f987cab..dee4cd3435 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.c +++ b/src/mesa/drivers/dri/r600/r600_emit.c @@ -175,7 +175,8 @@ void r600EmitCacheFlush(r600ContextPtr rmesa) GLboolean r600EmitShader(GLcontext * ctx, void ** shaderbo, GLvoid * data, - int sizeinDWORD) + int sizeinDWORD, + char * szShaderUsage) { radeonContextPtr radeonctx = RADEON_CONTEXT(ctx); @@ -183,12 +184,22 @@ GLboolean r600EmitShader(GLcontext * ctx, uint32_t *out; shader_again_alloc: +#ifdef RADEON_DEBUG_BO pbo = radeon_bo_open(radeonctx->radeonScreen->bom, 0, sizeinDWORD * 4, 256, + RADEON_GEM_DOMAIN_GTT, + 0, + szShaderUsage); +#else + pbo = radeon_bo_open(radeonctx->radeonScreen->bom, + 0, + sizeinDWORD * 4, + 256, RADEON_GEM_DOMAIN_GTT, 0); +#endif /* RADEON_DEBUG_BO */ if (!pbo) { @@ -211,6 +222,8 @@ shader_again_alloc: memcpy(out, data, sizeinDWORD * 4); + radeon_bo_unmap(pbo); + *shaderbo = (void*)pbo; return GL_TRUE; diff --git a/src/mesa/drivers/dri/r600/r600_emit.h b/src/mesa/drivers/dri/r600/r600_emit.h index 11765c868c..a2d8312287 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.h +++ b/src/mesa/drivers/dri/r600/r600_emit.h @@ -44,7 +44,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. extern GLboolean r600EmitShader(GLcontext * ctx, void ** shaderbo, GLvoid * data, - int sizeinDWORD); + int sizeinDWORD, + char * szShaderUsage); extern GLboolean r600DeleteShader(GLcontext * ctx, void * shaderbo); diff --git a/src/mesa/drivers/dri/r600/r600_swtcl.c b/src/mesa/drivers/dri/r600/r600_swtcl.c index 4d134070d6..a60e7c942c 100644 --- a/src/mesa/drivers/dri/r600/r600_swtcl.c +++ b/src/mesa/drivers/dri/r600/r600_swtcl.c @@ -594,8 +594,8 @@ static void r600EmitVertexAOS(r600ContextPtr rmesa, GLuint vertex_size, struct r BEGIN_BATCH(7); OUT_BATCH_PACKET3(R600_PACKET3_3D_LOAD_VBPNTR, 2); - OUT_BATCH(1); - OUT_BATCH(vertex_size | (vertex_size << 8)); + R600_OUT_BATCH(1); + R600_OUT_BATCH(vertex_size | (vertex_size << 8)); OUT_BATCH_RELOC(offset, bo, offset, RADEON_GEM_DOMAIN_GTT, 0, 0); END_BATCH(); #endif /* to be enabled */ @@ -612,7 +612,7 @@ static void r600EmitVbufPrim(r600ContextPtr rmesa, GLuint primitive, GLuint vert BEGIN_BATCH(3); OUT_BATCH_PACKET3(R600_PACKET3_3D_DRAW_VBUF_2, 0); - OUT_BATCH(R600_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type); + R600_OUT_BATCH(R600_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (num_verts << 16) | type); END_BATCH(); #endif /* to be enabled */ } diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index e534bb11b2..1f160364fa 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -41,16 +41,27 @@ extern const struct tnl_pipeline_stage *r700_pipeline[]; -static GLboolean r700DestroyChipObj(void* pvChipObj) +static GLboolean r700DestroyChipObj(GLcontext * ctx) { + context_t * context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700; - if(NULL == pvChipObj) + if(NULL == context->chipobj.pvChipObj) { return GL_TRUE; } - r700 = (R700_CHIP_CONTEXT *)pvChipObj; + r700 = (R700_CHIP_CONTEXT *)(context->chipobj.pvChipObj); + + if(0 != r700->pbo_vs_clear) + { + (context->chipobj.DeleteShader)(context, r700->pbo_vs_clear); + } + + if(0 != r700->pbo_fs_clear) + { + (context->chipobj.DeleteShader)(context, r700->pbo_fs_clear); + } FREE(r700->pStateList); @@ -405,19 +416,19 @@ void r700SetupVTXConstans(GLcontext * ctx, BEGIN_BATCH_NO_AUTOSTATE(9); - OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); - OUT_BATCH((nStreamID + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); + R600_OUT_BATCH((nStreamID + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); - R600_OUT_BATCH_RELOC(uSQ_VTX_CONSTANT_WORD1_0, + R600_OUT_BATCH_RELOC(uSQ_VTX_CONSTANT_WORD0_0, paos->bo, - uSQ_VTX_CONSTANT_WORD1_0, + uSQ_VTX_CONSTANT_WORD0_0, RADEON_GEM_DOMAIN_GTT, 0, 0, &offset_mod); - OUT_BATCH(uSQ_VTX_CONSTANT_WORD1_0); - OUT_BATCH(uSQ_VTX_CONSTANT_WORD2_0); - OUT_BATCH(uSQ_VTX_CONSTANT_WORD3_0); - OUT_BATCH(0); - OUT_BATCH(0); - OUT_BATCH(uSQ_VTX_CONSTANT_WORD6_0); + R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD1_0); + R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD2_0); + R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD3_0); + R600_OUT_BATCH(0); + R600_OUT_BATCH(0); + R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD6_0); END_BATCH(); COMMIT_BATCH(); @@ -440,13 +451,13 @@ int r700SetupStreams(GLcontext * ctx) unsigned int i; BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); - OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); - OUT_BATCH(0); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); + R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); + R600_OUT_BATCH(0); - OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); - OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); - OUT_BATCH(0); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); + R600_OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); + R600_OUT_BATCH(0); END_BATCH(); COMMIT_BATCH(); @@ -476,8 +487,33 @@ int r700SetupStreams(GLcontext * ctx) return R600_FALLBACK_NONE; } +inline GLboolean needRelocReg(context_t *context, unsigned int reg) +{ + switch (reg + ASIC_CONTEXT_BASE_INDEX) + { + case mmCB_COLOR0_BASE: + case mmCB_COLOR1_BASE: + case mmCB_COLOR2_BASE: + case mmCB_COLOR3_BASE: + case mmCB_COLOR4_BASE: + case mmCB_COLOR5_BASE: + case mmCB_COLOR6_BASE: + case mmCB_COLOR7_BASE: + case mmDB_DEPTH_BASE: + case mmSQ_PGM_START_VS: + case mmSQ_PGM_START_FS: + case mmSQ_PGM_START_ES: + case mmSQ_PGM_START_GS: + case mmSQ_PGM_START_PS: + return GL_TRUE; + break; + } + + return GL_FALSE; +} + inline GLboolean setRelocReg(context_t *context, unsigned int reg, - void * pbo_vs, void * pbo_fs) + GLboolean bUseStockShader) { BATCH_LOCALS(&context->radeon); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); @@ -507,9 +543,10 @@ inline GLboolean setRelocReg(context_t *context, unsigned int reg, return GL_FALSE; } - offset_mod.shift = RIGHT_SHIFT; - offset_mod.shiftbits = 8; - offset_mod.mask = 0x00FFFFFF; + /* refer to radeonCreateScreen : screen->fbLocation = (temp & 0xffff) << 16; */ + offset_mod.shift = NO_SHIFT; + offset_mod.shiftbits = 0; + offset_mod.mask = 0xFFFFFFFF; R600_OUT_BATCH_RELOC(r700->CB_COLOR0_BASE.u32All, rrb->bo, @@ -524,9 +561,9 @@ inline GLboolean setRelocReg(context_t *context, unsigned int reg, struct radeon_renderbuffer *rrb; rrb = radeon_get_depthbuffer(&context->radeon); - offset_mod.shift = RIGHT_SHIFT; - offset_mod.shiftbits = 8; - offset_mod.mask = 0x00FFFFFF; + offset_mod.shift = NO_SHIFT; + offset_mod.shiftbits = 0; + offset_mod.mask = 0xFFFFFFFF; R600_OUT_BATCH_RELOC(r700->DB_DEPTH_BASE.u32All, rrb->bo, @@ -538,18 +575,25 @@ inline GLboolean setRelocReg(context_t *context, unsigned int reg, break; case mmSQ_PGM_START_VS: { - if(NULL != pbo_vs) + if(GL_TRUE == bUseStockShader) { - pbo = (struct radeon_bo *)pbo_vs; + if(NULL != r700->pbo_vs_clear) + { + pbo = (struct radeon_bo *)(r700->pbo_vs_clear); + } + else + { + return GL_FALSE; + } } else { pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); } - offset_mod.shift = RIGHT_SHIFT; - offset_mod.shiftbits = 8; - offset_mod.mask = 0x00FFFFFF; + offset_mod.shift = NO_SHIFT; + offset_mod.shiftbits = 0; + offset_mod.mask = 0xFFFFFFFF; R600_OUT_BATCH_RELOC(r700->SQ_PGM_START_VS.u32All, pbo, @@ -563,18 +607,25 @@ inline GLboolean setRelocReg(context_t *context, unsigned int reg, case mmSQ_PGM_START_GS: case mmSQ_PGM_START_PS: { - if(NULL != pbo_fs) + if(GL_TRUE == bUseStockShader) { - pbo = (struct radeon_bo *)pbo_fs; + if(NULL != r700->pbo_fs_clear) + { + pbo = (struct radeon_bo *)(r700->pbo_fs_clear); + } + else + { + return GL_FALSE; + } } else { pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); } - offset_mod.shift = RIGHT_SHIFT; - offset_mod.shiftbits = 8; - offset_mod.mask = 0x00FFFFFF; + offset_mod.shift = NO_SHIFT; + offset_mod.shiftbits = 0; + offset_mod.mask = 0xFFFFFFFF; voffset = 0; R600_OUT_BATCH_RELOC(r700->SQ_PGM_START_PS.u32All, @@ -589,7 +640,7 @@ inline GLboolean setRelocReg(context_t *context, unsigned int reg, return GL_FALSE; } -GLboolean r700SendContextStates(context_t *context, void * pbo_vs, void * pbo_fs) +GLboolean r700SendContextStates(context_t *context, GLboolean bUseStockShader) { BATCH_LOCALS(&context->radeon); @@ -606,18 +657,22 @@ GLboolean r700SendContextStates(context_t *context, void * pbo_vs, void * pbo_fs pInit = pState; - while(NULL != pState->pNext) + if(GL_FALSE == needRelocReg(context, pState->unOffset)) { - if( (pState->pNext->unOffset - pState->unOffset) > 1 ) + while(NULL != pState->pNext) { - break; - } - else - { - pState = pState->pNext; - toSend++; - } - }; + if( ((pState->pNext->unOffset - pState->unOffset) > 1) + || (GL_TRUE == needRelocReg(context, pState->pNext->unOffset)) ) + { + break; + } + else + { + pState = pState->pNext; + toSend++; + } + }; + } pState = pState->pNext; @@ -625,10 +680,10 @@ GLboolean r700SendContextStates(context_t *context, void * pbo_vs, void * pbo_fs R600_OUT_BATCH_REGSEQ(((pInit->unOffset + ASIC_CONTEXT_BASE_INDEX)<<2), toSend); for(ui=0; uiunOffset+ui), pbo_vs, pbo_fs) ) + if( GL_FALSE == setRelocReg(context, (pInit->unOffset+ui), bUseStockShader) ) { /* for not reloc reg. */ - OUT_BATCH(*(pInit->puiValue)); + R600_OUT_BATCH(*(pInit->puiValue)); } pInit = pInit->pNext; }; diff --git a/src/mesa/drivers/dri/r600/r700_chip.h b/src/mesa/drivers/dri/r600/r700_chip.h index 30ca2cab84..42a72b7019 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.h +++ b/src/mesa/drivers/dri/r600/r700_chip.h @@ -447,6 +447,8 @@ typedef struct _R700_CHIP_CONTEXT R700_TEXTURE_STATES texture_states; + void * pbo_vs_clear; + void * pbo_fs_clear; GLboolean bEnablePerspective; } R700_CHIP_CONTEXT; @@ -454,7 +456,7 @@ typedef struct _R700_CHIP_CONTEXT #define R700_CONTEXT_STATES(context) ((R700_CHIP_CONTEXT *)(context->chipobj.pvChipObj)) extern GLboolean r700InitChipObject(context_t *context); -extern GLboolean r700SendContextStates(context_t *context, void * pbo_vs, void * pbo_fs); +extern GLboolean r700SendContextStates(context_t *context, GLboolean bUseStockShader); extern int r700SetupStreams(GLcontext * ctx); extern void r700SetupVTXConstans(GLcontext * ctx, unsigned int nStreamID, diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index 743875dfe2..9a7cc000a6 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -140,17 +140,18 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) /* Setup vb */ BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); - OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); - OUT_BATCH(0); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); + R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); + R600_OUT_BATCH(0); - OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); - OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); - OUT_BATCH(0); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); + R600_OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); + R600_OUT_BATCH(0); END_BATCH(); COMMIT_BATCH(); (context->chipobj.EmitVec)(ctx, &aos_vb, (GLvoid *)fVb, 4, 16, 6); + r700SetupVTXConstans(ctx, VERT_ATTRIB_POS, &aos_vb, 4, 16, 6); /* Setup shaders, copied from dump */ @@ -159,25 +160,33 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) SETbit(r700->SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); SETbit(r700->SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); /* vs */ - (context->chipobj.EmitShader)(ctx, &pbo_vs, (GLvoid *)(&uVs[0]), 28); + if(0 == r700->pbo_vs_clear) + { + (context->chipobj.EmitShader)(ctx, &(r700->pbo_vs_clear), (GLvoid *)(&uVs[0]), 28, "Clr VS"); + } + r700->SQ_PGM_START_VS.u32All = 0; r700->SQ_PGM_RESOURCES_VS.u32All = 0x00800004; /* vs const */ /* TODO : Set color here */ BEGIN_BATCH_NO_AUTOSTATE(4 + 2); - OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, 4)); - OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); - OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[0]))); - OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[1]))); - OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[2]))); - OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[3]))); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, 4)); + R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); + R600_OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[0]))); + R600_OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[1]))); + R600_OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[2]))); + R600_OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[3]))); END_BATCH(); COMMIT_BATCH(); r700->SPI_VS_OUT_CONFIG.u32All = 0x00000000; r700->SPI_PS_IN_CONTROL_0.u32All = 0x20000001; /* ps */ - (context->chipobj.EmitShader)(ctx, &pbo_fs, (GLvoid *)(&uFs[0]), 12); + if(0 == r700->pbo_fs_clear) + { + (context->chipobj.EmitShader)(ctx, &(r700->pbo_fs_clear), (GLvoid *)(&uFs[0]), 12, "Clr PS"); + } + r700->SQ_PGM_START_PS.u32All = 0; r700->SQ_PGM_RESOURCES_PS.u32All = 0x00800002; r700->SQ_PGM_EXPORTS_PS.u32All = 0x00000002; @@ -191,7 +200,7 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) r700->SQ_PGM_START_GS.u32All = 0; /* Now, send the states */ - r700SendContextStates(context, pbo_vs, pbo_fs); + r700SendContextStates(context, GL_TRUE); /* Draw */ GLuint numEntires, j; @@ -209,38 +218,33 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) SETfield(VGT_INDEX_TYPE, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); - OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); - OUT_BATCH(VGT_INDEX_TYPE); + R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); + R600_OUT_BATCH(VGT_INDEX_TYPE); VGT_NUM_INDICES = numIndices; SETfield(VGT_PRIMITIVE_TYPE, DI_PT_TRILIST, VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask); - OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); - OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); - OUT_BATCH(VGT_PRIMITIVE_TYPE); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + R600_OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); + R600_OUT_BATCH(VGT_PRIMITIVE_TYPE); SETfield(VGT_DRAW_INITIATOR, DI_SRC_SEL_IMMEDIATE, SOURCE_SELECT_shift, SOURCE_SELECT_mask); SETfield(VGT_DRAW_INITIATOR, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask); - OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1))); - OUT_BATCH(VGT_NUM_INDICES); - OUT_BATCH(VGT_DRAW_INITIATOR); + R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1))); + R600_OUT_BATCH(VGT_NUM_INDICES); + R600_OUT_BATCH(VGT_DRAW_INITIATOR); for (j=0; jchipobj.FlushCmdBuffer)(context); - /* TODO : keep these in context, don't load and release every time. */ - (context->chipobj.DeleteShader)(context, &pbo_vs); - - (context->chipobj.DeleteShader)(context, &pbo_fs); - (context->chipobj.FreeDmaRegion)(context, aos_vb.bo); /* Restore chip object. */ diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index d06af109e5..3aacad9919 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -281,7 +281,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) (context->chipobj.EmitShader)(ctx, &(fp->shaderbo), (GLvoid *)(fp->r700Shader.pProgram), - fp->r700Shader.uShaderBinaryDWORDSize); + fp->r700Shader.uShaderBinaryDWORDSize, + "FS"); fp->loaded = GL_TRUE; } @@ -344,19 +345,19 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) BEGIN_BATCH_NO_AUTOSTATE(2 + unNumParamData); - OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData)); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData)); /* assembler map const from very beginning. */ - OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4); + R600_OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4); unNumParamData = paramList->NumParameters; for(ui=0; uiParameterValues[ui][0]))); - OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1]))); - OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2]))); - OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3]))); + R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][0]))); + R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1]))); + R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2]))); + R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3]))); } END_BATCH(); COMMIT_BATCH(); diff --git a/src/mesa/drivers/dri/r600/r700_ioctl.c b/src/mesa/drivers/dri/r600/r700_ioctl.c index 7493efab0f..1ab3ee7de5 100644 --- a/src/mesa/drivers/dri/r600/r700_ioctl.c +++ b/src/mesa/drivers/dri/r600/r700_ioctl.c @@ -60,7 +60,7 @@ static void r700Flush(GLcontext *ctx) if (radeon->dma.flush) radeon->dma.flush( ctx ); - r700SendContextStates(context, NULL, NULL); + r700SendContextStates(context, GL_FALSE); if (radeon->cmdbuf.cs->cdw) rcommonFlushCmdBuf(radeon, __FUNCTION__); diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index f338cf823c..aee6a68565 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -59,9 +59,9 @@ void r700WaitForIdle(context_t *context) BATCH_LOCALS(&context->radeon); BEGIN_BATCH_NO_AUTOSTATE(3); - OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); - OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX); - OUT_BATCH(1 << 15); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX); + R600_OUT_BATCH(1 << 15); END_BATCH(); COMMIT_BATCH(); @@ -72,12 +72,12 @@ void r700WaitForIdleClean(context_t *context) BATCH_LOCALS(&context->radeon); BEGIN_BATCH_NO_AUTOSTATE(5); - OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 0)); - OUT_BATCH(0x16); + R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 0)); + R600_OUT_BATCH(0x16); - OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); - OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX); - OUT_BATCH(1 << 17); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + R600_OUT_BATCH(mmWAIT_UNTIL - ASIC_CONFIG_BASE_INDEX); + R600_OUT_BATCH(1 << 17); END_BATCH(); COMMIT_BATCH(); @@ -89,15 +89,15 @@ static void r700Start3D(context_t *context) if (context->radeon.radeonScreen->chip_family <= CHIP_FAMILY_RV670) { BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH(CP_PACKET3(R600_IT_START_3D_CMDBUF, 1)); - OUT_BATCH(0); + R600_OUT_BATCH(CP_PACKET3(R600_IT_START_3D_CMDBUF, 1)); + R600_OUT_BATCH(0); END_BATCH(); } BEGIN_BATCH_NO_AUTOSTATE(3); - OUT_BATCH(CP_PACKET3(R600_IT_CONTEXT_CONTROL, 1)); - OUT_BATCH(0x80000000); - OUT_BATCH(0x80000000); + R600_OUT_BATCH(CP_PACKET3(R600_IT_CONTEXT_CONTROL, 1)); + R600_OUT_BATCH(0x80000000); + R600_OUT_BATCH(0x80000000); END_BATCH(); COMMIT_BATCH(); @@ -183,11 +183,11 @@ GLboolean r700SyncSurf(context_t *context) BEGIN_BATCH_NO_AUTOSTATE(5); - OUT_BATCH(CP_PACKET3((IT_SURFACE_SYNC << 8), 3))); - OUT_BATCH(CP_COHER_CNTL); - OUT_BATCH(0xFFFFFFFF); - OUT_BATCH(0x00000000); - OUT_BATCH(10); + R600_OUT_BATCH(CP_PACKET3((IT_SURFACE_SYNC << 8), 3))); + R600_OUT_BATCH(CP_COHER_CNTL); + R600_OUT_BATCH(0xFFFFFFFF); + R600_OUT_BATCH(0x00000000); + R600_OUT_BATCH(10); END_BATCH(); COMMIT_BATCH(); @@ -241,7 +241,7 @@ static GLboolean r700RunRender(GLcontext * ctx, { context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); - +#if 1 BATCH_LOCALS(&context->radeon); unsigned int i, j; @@ -286,7 +286,7 @@ static GLboolean r700RunRender(GLcontext * ctx, r700->SQ_PGM_START_ES.u32All = r700->SQ_PGM_START_PS.u32All; r700->SQ_PGM_START_GS.u32All = r700->SQ_PGM_START_PS.u32All; - r700SendContextStates(context, NULL, NULL); + r700SendContextStates(context, GL_FALSE); /* richard test code */ for (i = 0; i < vb->PrimitiveCount; i++) @@ -305,33 +305,32 @@ static GLboolean r700RunRender(GLcontext * ctx, numEntires = 2 /* VGT_INDEX_TYPE */ + 3 /* VGT_PRIMITIVE_TYPE */ - + numIndices + 3 /* DRAW_INDEX_IMMD */ - + 2; /* test stamp */ + + numIndices + 3; /* DRAW_INDEX_IMMD */ BEGIN_BATCH_NO_AUTOSTATE(numEntires); VGT_INDEX_TYPE |= DI_INDEX_SIZE_32_BIT << INDEX_TYPE_shift; - OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); - OUT_BATCH(VGT_INDEX_TYPE); + R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); + R600_OUT_BATCH(VGT_INDEX_TYPE); VGT_NUM_INDICES = numIndices; VGT_PRIMITIVE_TYPE |= r700PrimitiveType(prim) << VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift; - OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); - OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); - OUT_BATCH(VGT_PRIMITIVE_TYPE); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); + R600_OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); + R600_OUT_BATCH(VGT_PRIMITIVE_TYPE); VGT_DRAW_INITIATOR |= DI_SRC_SEL_IMMEDIATE << SOURCE_SELECT_shift; VGT_DRAW_INITIATOR |= DI_MAJOR_MODE_0 << MAJOR_MODE_shift; - OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1))); - OUT_BATCH(VGT_NUM_INDICES); - OUT_BATCH(VGT_DRAW_INITIATOR); + R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1))); + R600_OUT_BATCH(VGT_NUM_INDICES); + R600_OUT_BATCH(VGT_DRAW_INITIATOR); for (j=0; jradeon, __FUNCTION__ ); return GL_FALSE; diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index a31a89627c..117d0add8b 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -692,7 +692,7 @@ void r700SetRenderTarget(context_t *context) nPitchInPixel = rrb->pitch/rrb->cpp; SETfield(r700->CB_COLOR0_SIZE.u32All, (nPitchInPixel/8)-1, PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask); - SETfield(r700->CB_COLOR0_SIZE.u32All, ( (nPitchInPixel * rrb->base.Height)/64 )-1, + SETfield(r700->CB_COLOR0_SIZE.u32All, ( (nPitchInPixel * context->radeon.radeonScreen->driScreen->fbHeight)/64 )-1, SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask); r700->CB_COLOR0_BASE.u32All = 0; SETfield(r700->CB_COLOR0_INFO.u32All, ENDIAN_NONE, ENDIAN_shift, ENDIAN_mask); @@ -737,7 +737,7 @@ void r700SetRenderTarget(context_t *context) SETfield(r700->DB_DEPTH_SIZE.u32All, (nPitchInPixel/8)-1, PITCH_TILE_MAX_shift, PITCH_TILE_MAX_mask); - SETfield(r700->DB_DEPTH_SIZE.u32All, ( (nPitchInPixel * rrb->base.Height)/64 )-1, + SETfield(r700->DB_DEPTH_SIZE.u32All, ( (nPitchInPixel * context->radeon.radeonScreen->driScreen->fbHeight)/64 )-1, SLICE_TILE_MAX_shift, SLICE_TILE_MAX_mask); /* size in pixel / 64 - 1 */ if(4 == rrb->cpp) @@ -909,7 +909,7 @@ void r700InitState(GLcontext * ctx) //------------------- r700->CB_COLOR0_FRAG.u32All = 0; r700->CB_COLOR0_MASK.u32All = 0; - r700->PA_SC_VPORT_ZMAX_0.u32All = 0x3F800000; + r700->PA_SC_VPORT_ZMAX_0.u32All = 0x3F800000; } void r700InitStateFuncs(struct dd_function_table *functions) //----------------- diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index f0dfbe20f0..4d113e2e0b 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -360,7 +360,8 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) (context->chipobj.EmitShader)(ctx, &(vp->shaderbo), (GLvoid *)(vp->r700Shader.pProgram), - vp->r700Shader.uShaderBinaryDWORDSize); + vp->r700Shader.uShaderBinaryDWORDSize, + "VS"); vp->loaded = GL_TRUE; } @@ -405,18 +406,18 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) BEGIN_BATCH_NO_AUTOSTATE(unNumParamData + 2); - OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData)); + R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData)); /* assembler map const from very beginning. */ - OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); + R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); unNumParamData = paramList->NumParameters; for(ui=0; uiParameterValues[ui][0]))); - OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1]))); - OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2]))); - OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3]))); + R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][0]))); + R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1]))); + R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2]))); + R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3]))); } END_BATCH(); COMMIT_BATCH(); diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index f8a29fdba0..30c858fce0 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -216,6 +216,11 @@ void radeonDestroyContext(__DRIcontextPrivate *driContextPriv ) radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL; + /* +r6/r7 */ + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); + /* --------- */ + if (radeon == current) { radeon_firevertices(radeon); _mesa_make_current(NULL, NULL, NULL); @@ -224,6 +229,13 @@ void radeonDestroyContext(__DRIcontextPrivate *driContextPriv ) assert(radeon); if (radeon) { +#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) /* +r6/r7 */ + if (IS_R600_CLASS(screen)) + { + r600DestroyContext(driContextPriv); + } +#endif + if (radeon->dma.current) { rcommonFlushCmdBuf( radeon, __FUNCTION__ ); } diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index be3b816b86..1d26e2c40a 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -1086,16 +1086,38 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv ) screen->AGPMode = dri_priv->AGPMode; ret = radeonGetParam(sPriv, RADEON_PARAM_FB_LOCATION, &temp); - if (ret) { - if (screen->chip_family < CHIP_FAMILY_RS600 && !screen->kernel_mm) - screen->fbLocation = ( INREG( RADEON_MC_FB_LOCATION ) & 0xffff) << 16; - else { - FREE( screen ); - fprintf(stderr, "Unable to get fb location need newer drm\n"); - return NULL; + + /* +r6/r7 */ + if(screen->chip_family >= CHIP_FAMILY_R600) + { + if (ret) + { + FREE( screen ); + fprintf(stderr, "Unable to get fb location need newer drm\n"); + return NULL; } - } else { - screen->fbLocation = (temp & 0xffff) << 16; + else + { + screen->fbLocation = (temp & 0xffff) << 24; + } + } + else + { + if (ret) + { + if (screen->chip_family < CHIP_FAMILY_RS600 && !screen->kernel_mm) + screen->fbLocation = ( INREG( RADEON_MC_FB_LOCATION ) & 0xffff) << 16; + else + { + FREE( screen ); + fprintf(stderr, "Unable to get fb location need newer drm\n"); + return NULL; + } + } + else + { + screen->fbLocation = (temp & 0xffff) << 16; + } } if (IS_R300_CLASS(screen)) { -- cgit v1.2.3 From 465588cd0e6451c758a12108787331d03ed52780 Mon Sep 17 00:00:00 2001 From: Richard Li Date: Tue, 19 May 2009 10:28:40 -0400 Subject: Fix nop insertion bug. redbook hello can run. --- src/mesa/drivers/dri/r600/r700_chip.c | 2 +- src/mesa/drivers/dri/r600/r700_clear.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index 1f160364fa..9395bae296 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -680,7 +680,7 @@ GLboolean r700SendContextStates(context_t *context, GLboolean bUseStockShader) R600_OUT_BATCH_REGSEQ(((pInit->unOffset + ASIC_CONTEXT_BASE_INDEX)<<2), toSend); for(ui=0; uiunOffset+ui), bUseStockShader) ) + if( GL_FALSE == setRelocReg(context, pInit->unOffset, bUseStockShader) ) { /* for not reloc reg. */ R600_OUT_BATCH(*(pInit->puiValue)); diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index 9a7cc000a6..e2571626dc 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -262,7 +262,7 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) return; } - r700ClearWithDraw(context, mask); + //r700ClearWithDraw(context, mask); } -- cgit v1.2.3 From f3eaba017c77f4dcf6cdecba7c19c9ea4fd3ce63 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 26 May 2009 12:21:54 -0400 Subject: r6xx: fix more r6xx specific cases --- src/mesa/drivers/dri/r600/r700_clear.c | 2 +- src/mesa/drivers/dri/r600/r700_render.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index e2571626dc..c011b863fe 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -80,7 +80,7 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) 0x10000000, 0x340C90, 0x10000400, 0x20340C90, 0x10000800, 0x40340C90, 0x90000C00, 0x60200C90}; - if (context->radeon.radeonScreen->chip_family <= CHIP_FAMILY_RV670) + if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) { uVs[9] = 0x541910; uVs[11] = 0x20541910; diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index 464023e96f..fccf416c8c 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -86,7 +86,7 @@ void r700WaitForIdleClean(context_t *context) static void r700Start3D(context_t *context) { BATCH_LOCALS(&context->radeon); - if (context->radeon.radeonScreen->chip_family <= CHIP_FAMILY_RV670) + if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) { BEGIN_BATCH_NO_AUTOSTATE(2); R600_OUT_BATCH(CP_PACKET3(R600_IT_START_3D_CMDBUF, 0)); -- cgit v1.2.3 From 3f1474594697fd77211851d699710299721b4618 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 28 May 2009 17:09:19 -0400 Subject: R6xx/r7xx: re-enable clears flush cache and wait for idle after drawing. Probably just need the cache flush. This gets redbook hello working properly. --- src/mesa/drivers/dri/r600/r700_clear.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index c011b863fe..fa0d81b725 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -96,6 +96,8 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) uFs[11] = 0x60201910; } + r700Start3D(context); + r700SyncSurf(context); /* Save current chip object. */ @@ -243,6 +245,8 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) END_BATCH(); COMMIT_BATCH(); + r700WaitForIdleClean(context); + (context->chipobj.FlushCmdBuffer)(context); (context->chipobj.FreeDmaRegion)(context, aos_vb.bo); @@ -262,7 +266,7 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) return; } - //r700ClearWithDraw(context, mask); + r700ClearWithDraw(context, mask); } -- cgit v1.2.3 From 7f022fb5cd282826dac37b7ac0c02f551b6f5f98 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 28 May 2009 17:18:17 -0400 Subject: Argh. fix last commit. clears are still broken --- src/mesa/drivers/dri/r600/r700_clear.c | 2 ++ src/mesa/drivers/dri/r600/r700_render.c | 2 +- 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index fa0d81b725..7ebab8daa1 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -41,6 +41,8 @@ extern void r700InitState (GLcontext * ctx); extern GLboolean r700SyncSurf(context_t *context); +extern void r700Start3D(context_t *context); +extern void r700WaitForIdleClean(context_t *context); static GLboolean r700ClearFast(context_t *context, GLbitfield mask) { diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index 43b279da5c..a6f937ddeb 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -83,7 +83,7 @@ void r700WaitForIdleClean(context_t *context) COMMIT_BATCH(); } -static void r700Start3D(context_t *context) +void r700Start3D(context_t *context) { BATCH_LOCALS(&context->radeon); if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) -- cgit v1.2.3 From 846b24ea82cfb8a4867b6da1345584379f4aaa04 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 29 May 2009 12:49:45 -0400 Subject: r6xx/r7xx: switch to common dma functions for vecs --- src/mesa/drivers/dri/r600/r600_context.c | 4 - src/mesa/drivers/dri/r600/r600_context.h | 10 -- src/mesa/drivers/dri/r600/r600_emit.c | 162 ------------------------------- src/mesa/drivers/dri/r600/r600_emit.h | 12 --- src/mesa/drivers/dri/r600/r700_chip.c | 12 +-- src/mesa/drivers/dri/r600/r700_clear.c | 8 +- src/mesa/drivers/dri/r600/r700_render.c | 4 +- 7 files changed, 9 insertions(+), 203 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index 5be486b176..3ea8f00086 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -286,12 +286,8 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, (r600->chipobj.InitFuncs)(&functions); r600->chipobj.EmitShader = r600EmitShader; r600->chipobj.DeleteShader = r600DeleteShader; - r600->chipobj.FreeDmaRegion = r600FreeDmaRegion; - r600->chipobj.EmitVec = r600EmitVec; - r600->chipobj.ReleaseArrays = r600ReleaseVec; r600->chipobj.LoadMemSurf = r600LoadMemSurf; r600->chipobj.AllocMemSurf = r600AllocMemSurf; - r600->chipobj.FlushCmdBuffer = r600FlushCmdBuffer; r600->chipobj.MemUse = r600MemUse; if (!radeonInitContext(&r600->radeon, &functions, diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index ae80037602..2740c0a1f8 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -175,17 +175,7 @@ typedef struct chip_object char * szShaderUsage); GLboolean (*DeleteShader)(GLcontext * ctx, void * shaderbo); - void (*FreeDmaRegion)( GLcontext * ctx, - void * shaderbo); - GLboolean (*EmitVec)(GLcontext * ctx, - struct radeon_aos *aos, - GLvoid * data, - int size, - int stride, - int count); void (*MemUse)(context_t *context, int id); - void (*ReleaseArrays)(GLcontext * ctx); - int (*FlushCmdBuffer)(GLcontext * ctx); GLboolean (*LoadMemSurf)(context_t *context, GLuint dst_offset, /* gpu addr */ GLuint dst_pitch_in_pixel, diff --git a/src/mesa/drivers/dri/r600/r600_emit.c b/src/mesa/drivers/dri/r600/r600_emit.c index e4ba656ee8..711dfefd73 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.c +++ b/src/mesa/drivers/dri/r600/r600_emit.c @@ -47,115 +47,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r600_context.h" #include "r600_emit.h" -#if defined(USE_X86_ASM) -#define COPY_DWORDS( dst, src, nr ) \ -do { \ - int __tmp; \ - __asm__ __volatile__( "rep ; movsl" \ - : "=%c" (__tmp), "=D" (dst), "=S" (__tmp) \ - : "0" (nr), \ - "D" ((long)dst), \ - "S" ((long)src) ); \ -} while (0) -#else -#define COPY_DWORDS( dst, src, nr ) \ -do { \ - int j; \ - for ( j = 0 ; j < nr ; j++ ) \ - dst[j] = ((int *)src)[j]; \ - dst += nr; \ -} while (0) -#endif - -static void r600EmitVec4(uint32_t *out, GLvoid * data, int stride, int count) -{ - int i; - - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 4) - COPY_DWORDS(out, data, count); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out++; - data += stride; - } -} - -static void r600EmitVec8(uint32_t *out, GLvoid * data, int stride, int count) -{ - int i; - - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 8) - COPY_DWORDS(out, data, count * 2); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data + 4); - out += 2; - data += stride; - } -} - -static void r600EmitVec12(uint32_t *out, GLvoid * data, int stride, int count) -{ - int i; - - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 12) { - COPY_DWORDS(out, data, count * 3); - } - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data + 4); - out[2] = *(int *)(data + 8); - out += 3; - data += stride; - } -} - -static void r600EmitVec16(uint32_t *out, GLvoid * data, int stride, int count) -{ - int i; - - if (RADEON_DEBUG & DEBUG_VERTS) - fprintf(stderr, "%s count %d stride %d out %p data %p\n", - __FUNCTION__, count, stride, (void *)out, (void *)data); - - if (stride == 16) - COPY_DWORDS(out, data, count * 4); - else - for (i = 0; i < count; i++) { - out[0] = *(int *)data; - out[1] = *(int *)(data + 4); - out[2] = *(int *)(data + 8); - out[3] = *(int *)(data + 12); - out += 4; - data += stride; - } -} - -/* Emit vertex data to GART memory - * Route inputs to the vertex processor - * This function should never return R600_FALLBACK_TCL when using software tcl. - */ -int r600EmitArrays(GLcontext * ctx) -{ - - return R600_FALLBACK_NONE; -} - void r600EmitCacheFlush(r600ContextPtr rmesa) { BATCH_LOCALS(&rmesa->radeon); @@ -241,56 +132,3 @@ GLboolean r600DeleteShader(GLcontext * ctx, return GL_TRUE; } - -GLboolean r600EmitVec(GLcontext * ctx, - struct radeon_aos *aos, - GLvoid * data, - int size, - int stride, - int count) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - uint32_t *out; - - if (stride == 0) - { - radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * count * 4, 32); - aos->stride = 0; - } - else - { - radeonAllocDmaRegion(rmesa, &aos->bo, &aos->offset, size * count * 4, 32); - aos->stride = size; - } - - aos->components = size; - aos->count = count; - - out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); - switch (size) { - case 1: r600EmitVec4(out, data, stride, count); break; - case 2: r600EmitVec8(out, data, stride, count); break; - case 3: r600EmitVec12(out, data, stride, count); break; - case 4: r600EmitVec16(out, data, stride, count); break; - default: - assert(0); - break; - } - - return GL_TRUE; -} - -void r600ReleaseVec(GLcontext * ctx) -{ - radeonReleaseArrays(ctx, ~0); -} - -void r600FreeDmaRegion(context_t *context, - void * shaderbo) -{ - struct radeon_bo *pbo = (struct radeon_bo *)shaderbo; - if(pbo) - { - radeon_bo_unref(pbo); - } -} diff --git a/src/mesa/drivers/dri/r600/r600_emit.h b/src/mesa/drivers/dri/r600/r600_emit.h index a2d8312287..d3c9ef59eb 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.h +++ b/src/mesa/drivers/dri/r600/r600_emit.h @@ -50,16 +50,4 @@ extern GLboolean r600EmitShader(GLcontext * ctx, extern GLboolean r600DeleteShader(GLcontext * ctx, void * shaderbo); -extern GLboolean r600EmitVec(GLcontext * ctx, - struct radeon_aos *aos, - GLvoid * data, - int size, - int stride, - int count); - -extern void r600ReleaseVec(GLcontext * ctx); - -extern void r600FreeDmaRegion(context_t *context, - void * shaderbo); - #endif diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index 9395bae296..12e9b474e1 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -467,12 +467,12 @@ int r700SetupStreams(GLcontext * ctx) unBit = 1 << i; if(vpc->mesa_program.Base.InputsRead & unBit) { - (context->chipobj.EmitVec)(ctx, - &context->radeon.tcl.aos[i], - vb->AttribPtr[i]->data, - vb->AttribPtr[i]->size, - vb->AttribPtr[i]->stride, - vb->Count); + rcommon_emit_vector(ctx, + &context->radeon.tcl.aos[i], + vb->AttribPtr[i]->data, + vb->AttribPtr[i]->size, + vb->AttribPtr[i]->stride, + vb->Count); /* currently aos are packed */ r700SetupVTXConstans(ctx, diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index 7ebab8daa1..ffe236131c 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -154,7 +154,7 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) END_BATCH(); COMMIT_BATCH(); - (context->chipobj.EmitVec)(ctx, &aos_vb, (GLvoid *)fVb, 4, 16, 6); + rcommon_emit_vector(ctx, &aos_vb, (GLvoid *)fVb, 4, 16, 6); r700SetupVTXConstans(ctx, VERT_ATTRIB_POS, &aos_vb, 4, 16, 6); @@ -249,10 +249,6 @@ static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) r700WaitForIdleClean(context); - (context->chipobj.FlushCmdBuffer)(context); - - (context->chipobj.FreeDmaRegion)(context, aos_vb.bo); - /* Restore chip object. */ memcpy(r700, &r700Saved, sizeof(R700_CHIP_CONTEXT)); @@ -268,7 +264,7 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) return; } - r700ClearWithDraw(context, mask); + //r700ClearWithDraw(context, mask); } diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index a6f937ddeb..35f30605ce 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -350,9 +350,7 @@ static GLboolean r700RunRender(GLcontext * ctx, /* flush dst */ //r700SyncSurf(context); /* */ - (context->chipobj.FlushCmdBuffer)(context); - - (context->chipobj.ReleaseArrays)(ctx); + radeonReleaseArrays(ctx, 0); //richard test /* test stamp, write a number to mmSCRATCH4 */ -- cgit v1.2.3 From 41a44ff8919f8e16ffb61fd3a76a82864b86f50a Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 29 May 2009 13:07:05 -0400 Subject: R6xx/R7xx: switch to common clear with draw code --- src/mesa/drivers/dri/r600/r700_clear.c | 273 ++++++++------------------------- 1 file changed, 64 insertions(+), 209 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index ffe236131c..e610514e01 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -38,233 +38,88 @@ #include "r700_shaderinst.h" #include "r600_emit.h" -extern void r700InitState (GLcontext * ctx); - -extern GLboolean r700SyncSurf(context_t *context); -extern void r700Start3D(context_t *context); -extern void r700WaitForIdleClean(context_t *context); - static GLboolean r700ClearFast(context_t *context, GLbitfield mask) { /* TODO, fast clear need implementation */ return GL_FALSE; } -static GLboolean r700ClearWithDraw(context_t *context, GLbitfield mask) +static void r700UserClear(GLcontext *ctx, GLuint mask) { - GLcontext *ctx = GL_CONTEXT(context); - - BATCH_LOCALS(&context->radeon); - - R700_CHIP_CONTEXT r700Saved; - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); - - void * pbo_vs; - void * pbo_fs; - struct radeon_aos aos_vb; - - unsigned int ui; - GLfloat fTemp; - GLfloat fVb[] = { 1.0f, 1.0f, 1.0f, 1.0f, - -1.0f, -1.0f, 1.0f, 1.0f, - 1.0f, -1.0f, 1.0f, 1.0f, - 1.0f, 1.0f, 1.0f, 1.0f, - -1.0f, 1.0f, 1.0f, 1.0f, - -1.0f, -1.0f, 1.0f, 1.0f}; /* TODO : Z set here */ - unsigned int uVs[] = { 0xC, 0x81000000, 0x4, 0xA01C0000, - 0xC001203C, 0x94000688, 0xC001C000, 0x94200688, - 0x10000001, 0x540C90, 0x10000401, 0x20540C90, - 0x10000801, 0x40540C90, 0x90000C01, 0x60400C90, - 0x10000100, 0x600C90, 0x10000500, 0x20600C90, - 0x10000900, 0x40600C90, 0x90000D00, 0x60680C90, - 0x7C000000, 0x2D1001, 0x80000, 0xBEADEAF }; - unsigned int uFs[] = { 0x2, 0xA00C0000, 0xC0008000, 0x94200688, - 0x10000000, 0x340C90, 0x10000400, 0x20340C90, - 0x10000800, 0x40340C90, 0x90000C00, 0x60200C90}; - - if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) - { - uVs[9] = 0x541910; - uVs[11] = 0x20541910; - uVs[13] = 0x40541910; - uVs[15] = 0x60401910; - uVs[17] = 0x601910; - uVs[19] = 0x20601910; - uVs[21] = 0x40601910; - uVs[23] = 0x60681910; - uFs[5] = 0x341910; - uFs[7] = 0x20341910; - uFs[9] = 0x40341910; - uFs[11] = 0x60201910; - } - - r700Start3D(context); - - r700SyncSurf(context); - - /* Save current chip object. */ - memcpy(&r700Saved, r700, sizeof(R700_CHIP_CONTEXT)); - - r700InitState(ctx); - - r700SetRenderTarget(context); - - /* Turn off perspective divid. */ - SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_XY_FMT_bit); - SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_Z_FMT_bit); - SETbit(r700->PA_CL_VTE_CNTL.u32All, VTX_W0_FMT_bit); - - if( (mask & BUFFER_BIT_FRONT_LEFT) || (mask & BUFFER_BIT_BACK_LEFT) ) - { /* Enable render target output. */ - SETfield(r700->CB_TARGET_MASK.u32All, 0xF, TARGET0_ENABLE_shift, TARGET0_ENABLE_mask); - } - else - { /* Disable render target output. */ - CLEARfield(r700->CB_TARGET_MASK.u32All, TARGET0_ENABLE_mask); /* TODO : OGL need 4 rt. */ - } - if (mask & BUFFER_BIT_DEPTH) - { - /* Set correct Z to clear. */ - SETbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit); - SETbit(r700->DB_DEPTH_CONTROL.u32All, Z_ENABLE_bit); - SETfield(r700->DB_DEPTH_CONTROL.u32All, FRAG_ALWAYS, ZFUNC_shift, ZFUNC_mask); - fTemp = ctx->Depth.Clear; - for(ui=2; ui<24;) - { - fVb[ui] = fTemp; - ui += 4; - } - } - else - { - /* Disable Z write. */ - CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_WRITE_ENABLE_bit); - CLEARbit(r700->DB_DEPTH_CONTROL.u32All, Z_ENABLE_bit); - } - - /* Setup vb */ - BEGIN_BATCH_NO_AUTOSTATE(6); - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); - R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX); - R600_OUT_BATCH(0); - - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1)); - R600_OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX); - R600_OUT_BATCH(0); - END_BATCH(); - COMMIT_BATCH(); - - rcommon_emit_vector(ctx, &aos_vb, (GLvoid *)fVb, 4, 16, 6); - - r700SetupVTXConstans(ctx, VERT_ATTRIB_POS, &aos_vb, 4, 16, 6); - - /* Setup shaders, copied from dump */ - r700->SQ_PGM_RESOURCES_PS.u32All = 0; - r700->SQ_PGM_RESOURCES_VS.u32All = 0; - SETbit(r700->SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); - SETbit(r700->SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit); - /* vs */ - if(0 == r700->pbo_vs_clear) - { - (context->chipobj.EmitShader)(ctx, &(r700->pbo_vs_clear), (GLvoid *)(&uVs[0]), 28, "Clr VS"); - } - - r700->SQ_PGM_START_VS.u32All = 0; - r700->SQ_PGM_RESOURCES_VS.u32All = 0x00800004; - - /* vs const */ /* TODO : Set color here */ - BEGIN_BATCH_NO_AUTOSTATE(4 + 2); - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, 4)); - R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4); - R600_OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[0]))); - R600_OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[1]))); - R600_OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[2]))); - R600_OUT_BATCH(*((unsigned int*)&(ctx->Color.ClearColor[3]))); - END_BATCH(); - COMMIT_BATCH(); - - r700->SPI_VS_OUT_CONFIG.u32All = 0x00000000; - r700->SPI_PS_IN_CONTROL_0.u32All = 0x20000001; - /* ps */ - if(0 == r700->pbo_fs_clear) - { - (context->chipobj.EmitShader)(ctx, &(r700->pbo_fs_clear), (GLvoid *)(&uFs[0]), 12, "Clr PS"); - } - - r700->SQ_PGM_START_PS.u32All = 0; - r700->SQ_PGM_RESOURCES_PS.u32All = 0x00800002; - r700->SQ_PGM_EXPORTS_PS.u32All = 0x00000002; - r700->DB_SHADER_CONTROL.u32All = 0x00000200; - - r700->CB_SHADER_CONTROL.u32All = 0x00000001; - - /* set a valid base address to make the command checker happy */ - r700->SQ_PGM_START_FS.u32All = 0; - r700->SQ_PGM_START_ES.u32All = 0; - r700->SQ_PGM_START_GS.u32All = 0; - - /* Now, send the states */ - r700SendContextStates(context, GL_TRUE); - - /* Draw */ - GLuint numEntires, j; - GLuint numIndices = 6; - unsigned int VGT_DRAW_INITIATOR = 0; - unsigned int VGT_INDEX_TYPE = 0; - unsigned int VGT_PRIMITIVE_TYPE = 0; - unsigned int VGT_NUM_INDICES = 0; - - numEntires = 2 /* VGT_INDEX_TYPE */ - + 3 /* VGT_PRIMITIVE_TYPE */ - + numIndices + 3; /* DRAW_INDEX_IMMD */ - - BEGIN_BATCH_NO_AUTOSTATE(numEntires); - - SETfield(VGT_INDEX_TYPE, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); - - R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); - R600_OUT_BATCH(VGT_INDEX_TYPE); - - VGT_NUM_INDICES = numIndices; - - SETfield(VGT_PRIMITIVE_TYPE, DI_PT_TRILIST, VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask); - - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); - R600_OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); - R600_OUT_BATCH(VGT_PRIMITIVE_TYPE); - - SETfield(VGT_DRAW_INITIATOR, DI_SRC_SEL_IMMEDIATE, SOURCE_SELECT_shift, SOURCE_SELECT_mask); - SETfield(VGT_DRAW_INITIATOR, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask); - - R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (numIndices + 1))); - R600_OUT_BATCH(VGT_NUM_INDICES); - R600_OUT_BATCH(VGT_DRAW_INITIATOR); - - for (j=0; jradeon.dma.flush ) \ + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); \ + } while (0) + void r700Clear(GLcontext * ctx, GLbitfield mask) { context_t *context = R700_CONTEXT(ctx); + __DRIdrawablePrivate *dPriv = context->radeon.dri.drawable; + const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); + GLbitfield swrast_mask = 0, tri_mask = 0; + int i; + struct gl_framebuffer *fb = ctx->DrawBuffer; if( GL_TRUE == r700ClearFast(context, mask) ) { return; } - //r700ClearWithDraw(context, mask); +#if 0 + if (!context->radeon.radeonScreen->driScreen->dri2.enabled) { + LOCK_HARDWARE(&context->radeon); + UNLOCK_HARDWARE(&context->radeon); + if (dPriv->numClipRects == 0) + return; + } +#endif + + R600_NEWPRIM(context); + + if (colorMask == ~0) + tri_mask |= (mask & BUFFER_BITS_COLOR); + + + /* HW stencil */ + if (mask & BUFFER_BIT_STENCIL) { + tri_mask |= BUFFER_BIT_STENCIL; + } + + /* HW depth */ + if (mask & BUFFER_BIT_DEPTH) { + tri_mask |= BUFFER_BIT_DEPTH; + } + + /* If we're doing a tri pass for depth/stencil, include a likely color + * buffer with it. + */ + + for (i = 0; i < BUFFER_COUNT; i++) { + GLuint bufBit = 1 << i; + if ((tri_mask) & bufBit) { + if (!fb->Attachment[i].Renderbuffer->ClassID) { + tri_mask &= ~bufBit; + swrast_mask |= bufBit; + } + } + } + + /* SW fallback clearing */ + swrast_mask = mask & ~tri_mask; + + if (tri_mask) + r700UserClear(ctx, tri_mask); + if (swrast_mask) { + if (RADEON_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s: swrast clear, mask: %x\n", + __FUNCTION__, swrast_mask); + _swrast_Clear(ctx, swrast_mask); + } + } -- cgit v1.2.3 From d8694cd3f69dfc40509544293aa35f297571e878 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 29 May 2009 15:37:04 -0400 Subject: get rid of chip_object struct --- src/mesa/drivers/dri/r600/r600_context.c | 6 +----- src/mesa/drivers/dri/r600/r600_context.h | 30 ++++++++++++++++-------------- src/mesa/drivers/dri/r600/r700_assembler.c | 1 - src/mesa/drivers/dri/r600/r700_chip.c | 26 +------------------------- src/mesa/drivers/dri/r600/r700_chip.h | 12 ------------ src/mesa/drivers/dri/r600/r700_clear.c | 1 - src/mesa/drivers/dri/r600/r700_fragprog.c | 3 +-- src/mesa/drivers/dri/r600/r700_ioctl.c | 1 - src/mesa/drivers/dri/r600/r700_oglprog.c | 1 - src/mesa/drivers/dri/r600/r700_render.c | 7 +++---- src/mesa/drivers/dri/r600/r700_shader.c | 1 - src/mesa/drivers/dri/r600/r700_state.c | 20 +++++++------------- src/mesa/drivers/dri/r600/r700_tex.c | 1 - src/mesa/drivers/dri/r600/r700_vertprog.c | 3 +-- 14 files changed, 30 insertions(+), 83 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index c54dae0c53..526d02ed57 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -63,7 +63,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r600_emit.h" #include "radeon_bocs_wrapper.h" -#include "r700_chip.h" #include "r700_state.h" #include "r700_ioctl.h" @@ -369,9 +368,6 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } - r600->disable_lowimpact_fallback = - driQueryOptionb(&r600->radeon.optionCache, - "disable_lowimpact_fallback"); radeon_fbo_init(&r600->radeon); radeonInitSpanFuncs( ctx ); @@ -410,7 +406,7 @@ r600DestroyContext (__DRIcontextPrivate * driContextPriv) context_t *context = ctx ? R700_CONTEXT(ctx) : NULL; if (context) - (context->chipobj.DestroyChipObj)(context->chipobj.pvChipObj); + FREE(context->hw.pStateList); } diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index 321b05d930..52512aacc0 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -48,6 +48,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/mtypes.h" #include "main/colormac.h" +#include "r700_chip.h" + struct r600_context; typedef struct r600_context context_t; @@ -127,32 +129,20 @@ typedef struct offset_modifiers GLuint mask; } offset_modifiers; -typedef struct chip_object -{ - void *pvChipObj; - - /* ------------ OUT ------------------- */ - GLboolean (*DestroyChipObj)(GLcontext * ctx); - -} chip_object; - /** * \brief R600 context structure. */ struct r600_context { struct radeon_context radeon; /* parent class, must be first */ - /* ------ */ - chip_object chipobj; + /* ------ */ + R700_CHIP_CONTEXT hw; /* Vertex buffers */ GLvector4f dummy_attrib[_TNL_ATTRIB_MAX]; GLvector4f *temp_attrib[_TNL_ATTRIB_MAX]; - GLboolean disable_lowimpact_fallback; - - GLboolean vap_flush_needed; }; #define R700_CONTEXT(ctx) ((context_t *)(ctx->DriverCtx)) @@ -163,6 +153,18 @@ extern GLboolean r600CreateContext(const __GLcontextModes * glVisual, __DRIcontextPrivate * driContextPriv, void *sharedContextPrivate); +#define R700_CONTEXT_STATES(context) ((R700_CHIP_CONTEXT *)(&context->hw)) + +extern GLboolean r700InitChipObject(context_t *context); +extern GLboolean r700SendContextStates(context_t *context); +extern int r700SetupStreams(GLcontext * ctx); +extern void r700SetupVTXConstants(GLcontext * ctx, + unsigned int nStreamID, + void * pAos, + unsigned int size, /* number of elements in vector */ + unsigned int stride, + unsigned int Count); /* number of vectors in stream */ + #define RADEON_D_CAPTURE 0 #define RADEON_D_PLAYBACK 1 #define RADEON_D_PLAYBACK_RAW 2 diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index 1610020ee8..a564d6f099 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -34,7 +34,6 @@ #include "main/imports.h" #include "r600_context.h" -#include "r700_chip.h" #include "r700_debug.h" #include "r700_assembler.h" diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index 9b30249231..5ae70c99c7 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -31,7 +31,6 @@ #include "r600_context.h" #include "r600_cmdbuf.h" -#include "r700_chip.h" #include "r700_state.h" #include "r700_tex.h" #include "r700_oglprog.h" @@ -39,25 +38,6 @@ #include "r700_vertprog.h" #include "r700_ioctl.h" -static GLboolean r700DestroyChipObj(GLcontext * ctx) -{ - context_t * context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700; - - if(NULL == context->chipobj.pvChipObj) - { - return GL_TRUE; - } - - r700 = (R700_CHIP_CONTEXT *)(context->chipobj.pvChipObj); - - FREE(r700->pStateList); - - FREE(r700); - - return GL_TRUE; -} - #define LINK_STATES(reg) \ do \ { \ @@ -71,11 +51,7 @@ GLboolean r700InitChipObject(context_t *context) { ContextState * pStateListWork; - R700_CHIP_CONTEXT *r700 = CALLOC( sizeof(R700_CHIP_CONTEXT) ); - - context->chipobj.pvChipObj = (void*)r700; - - context->chipobj.DestroyChipObj = r700DestroyChipObj; + R700_CHIP_CONTEXT *r700 = &context->hw; /* init state list */ r700->pStateList = (ContextState*) MALLOC (sizeof(ContextState)*sizeof(R700_CHIP_CONTEXT)/sizeof(unsigned int)); diff --git a/src/mesa/drivers/dri/r600/r700_chip.h b/src/mesa/drivers/dri/r600/r700_chip.h index 88b5075479..918cb222d4 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.h +++ b/src/mesa/drivers/dri/r600/r700_chip.h @@ -451,17 +451,5 @@ typedef struct _R700_CHIP_CONTEXT } R700_CHIP_CONTEXT; -#define R700_CONTEXT_STATES(context) ((R700_CHIP_CONTEXT *)(context->chipobj.pvChipObj)) - -extern GLboolean r700InitChipObject(context_t *context); -extern GLboolean r700SendContextStates(context_t *context); -extern int r700SetupStreams(GLcontext * ctx); -extern void r700SetupVTXConstants(GLcontext * ctx, - unsigned int nStreamID, - void * pAos, - unsigned int size, /* number of elements in vector */ - unsigned int stride, - unsigned int Count); /* number of vectors in stream */ - #endif /* _R700_CHIP_H_ */ diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index e610514e01..e1164a09d7 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -33,7 +33,6 @@ #include "main/enums.h" #include "r600_context.h" -#include "r700_chip.h" #include "r700_shaderinst.h" #include "r600_emit.h" diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index c46094dcd7..cebb82d853 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -38,7 +38,6 @@ #include "r600_context.h" #include "r600_cmdbuf.h" -#include "r700_chip.h" #include "r700_fragprog.h" #include "r700_debug.h" @@ -259,7 +258,7 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) context_t *context = R700_CONTEXT(ctx); BATCH_LOCALS(&context->radeon); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); struct r700_fragment_program *fp = (struct r700_fragment_program *) (ctx->FragmentProgram._Current); diff --git a/src/mesa/drivers/dri/r600/r700_ioctl.c b/src/mesa/drivers/dri/r600/r700_ioctl.c index f8c3359a47..c479532001 100644 --- a/src/mesa/drivers/dri/r600/r700_ioctl.c +++ b/src/mesa/drivers/dri/r600/r700_ioctl.c @@ -37,7 +37,6 @@ #include "radeon_lock.h" #include "r600_context.h" -#include "r700_chip.h" #include "r700_ioctl.h" #include "r700_clear.h" diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.c b/src/mesa/drivers/dri/r600/r700_oglprog.c index ba49367dc4..36de143b1a 100644 --- a/src/mesa/drivers/dri/r600/r700_oglprog.c +++ b/src/mesa/drivers/dri/r600/r700_oglprog.c @@ -34,7 +34,6 @@ #include "r600_context.h" -#include "r700_chip.h" #include "r700_oglprog.h" #include "r700_fragprog.h" #include "r700_vertprog.h" diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index ded5f51a6a..cc90e8dfb7 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -47,7 +47,6 @@ #include "r600_context.h" #include "r600_cmdbuf.h" -#include "r700_chip.h" #include "r700_tex.h" #include "r700_vertprog.h" @@ -109,7 +108,7 @@ static GLboolean r700SetupShaders(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); GLuint exportCount; @@ -133,7 +132,7 @@ GLboolean r700SendTextureState(context_t *context) { unsigned int i; - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); #if 0 /* to be enabled */ for(i=0; ichipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); #if 1 BATCH_LOCALS(&context->radeon); diff --git a/src/mesa/drivers/dri/r600/r700_shader.c b/src/mesa/drivers/dri/r600/r700_shader.c index 446f9f959b..b4fd51c137 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.c +++ b/src/mesa/drivers/dri/r600/r700_shader.c @@ -35,7 +35,6 @@ #include "main/glheader.h" #include "r600_context.h" -#include "r700_chip.h" #include "r700_debug.h" #include "r700_shader.h" diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index e3e78df16c..768e184623 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -49,7 +49,6 @@ #include "r600_context.h" -#include "r700_chip.h" #include "r700_state.h" #include "r700_fragprog.h" @@ -190,7 +189,7 @@ static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //----------- { context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); _swrast_InvalidateState(ctx, new_state); _swsetup_InvalidateState(ctx, new_state); @@ -239,7 +238,7 @@ static void r700SetDepthState(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); if (ctx->Depth.Test) { @@ -322,7 +321,7 @@ static void r700BlendFuncSeparate(GLcontext * ctx, static void r700UpdateCulling(GLcontext * ctx) { - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(R700_CONTEXT(ctx)->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw); CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit); CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit); @@ -490,7 +489,7 @@ static void r700Viewport(GLcontext * ctx, { context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); __DRIdrawablePrivate *dPriv = context->radeon.dri.drawable; @@ -524,12 +523,7 @@ static void r700Viewport(GLcontext * ctx, tz = v[MAT_TZ] * scale; /* TODO : Need DMA flush as well. */ -#if 0 /* to be enabled */ - if(context->cmdbuf.count_used > 0) - { - (context->chipobj.FlushCmdBuffer)(context); - } -#endif /* to be enabled */ + r700->PA_CL_VPORT_XSCALE.u32All = *((unsigned int*)(&sx)); r700->PA_CL_VPORT_XOFFSET.u32All = *((unsigned int*)(&tx)); @@ -591,7 +585,7 @@ static void r700Scissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h) void r700SetRenderTarget(context_t *context) { - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); struct radeon_renderbuffer *rrb; unsigned int nPitchInPixel; @@ -774,7 +768,7 @@ void r700InitState(GLcontext * ctx) //------------------- { context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); /* Turn off vgt reuse */ r700->VGT_REUSE_OFF.u32All = 0; diff --git a/src/mesa/drivers/dri/r600/r700_tex.c b/src/mesa/drivers/dri/r600/r700_tex.c index e2de8faf77..21c9379729 100644 --- a/src/mesa/drivers/dri/r600/r700_tex.c +++ b/src/mesa/drivers/dri/r600/r700_tex.c @@ -40,7 +40,6 @@ #include "radeon_common.h" #include "r600_context.h" -#include "r700_chip.h" #include "r700_state.h" diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index 2861918c42..a827de197b 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -41,7 +41,6 @@ #include "r600_context.h" #include "r600_cmdbuf.h" -#include "r700_chip.h" #include "r700_debug.h" #include "r700_vertprog.h" @@ -339,7 +338,7 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) BATCH_LOCALS(&context->radeon); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(context->chipobj.pvChipObj); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); struct r700_vertex_program *vp = (struct r700_vertex_program *)ctx->VertexProgram._Current; -- cgit v1.2.3 From 116f40a1c4544242040afc7c6cf8621667a058c7 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 2 Jun 2009 19:23:16 -0400 Subject: Disable clear code for now --- src/mesa/drivers/dri/r600/r700_clear.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index e1164a09d7..b62aa66eb2 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -63,6 +63,9 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) int i; struct gl_framebuffer *fb = ctx->DrawBuffer; + /* XXX FIXME */ + return; + if( GL_TRUE == r700ClearFast(context, mask) ) { return; -- cgit v1.2.3 From eaadba6eee05cd725295821e0c1f5e1a01e99e89 Mon Sep 17 00:00:00 2001 From: Cooper Yuan Date: Thu, 25 Jun 2009 19:48:43 +0800 Subject: Fix color data can't be fetched issue --- src/mesa/drivers/dri/r600/r600_cmdbuf.c | 2 +- src/mesa/drivers/dri/r600/r700_clear.c | 3 --- 2 files changed, 1 insertion(+), 4 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index 5edc42b532..90f546edf1 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -288,7 +288,7 @@ restart: relocs[i].base.bo, soffset, eoffset); return r; } - asicoffset = cs->packets[relocs[i].indices[j]] + soffset; + asicoffset = soffset; if (asicoffset >= eoffset) { /* radeon_bo_debug(relocs[i].base.bo, 12); */ diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index f142aafd03..b04441da28 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -63,9 +63,6 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) int i; struct gl_framebuffer *fb = ctx->DrawBuffer; - /* XXX FIXME */ - return; - if( GL_TRUE == r700ClearFast(context, mask) ) { return; -- cgit v1.2.3 From d10006e67fb070f846213ef38e07df21699df5a1 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 30 Jun 2009 18:40:48 -0400 Subject: R6xx/r7xx: fix locking for clear this allows redbook hello to render correctly mostly. --- src/mesa/drivers/dri/r600/r700_clear.c | 3 +-- src/mesa/drivers/dri/r600/r700_render.c | 4 ---- 2 files changed, 1 insertion(+), 6 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index b04441da28..7696795676 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -32,6 +32,7 @@ #include "main/mtypes.h" #include "main/enums.h" +#include "radeon_lock.h" #include "r600_context.h" #include "r700_shaderinst.h" @@ -68,14 +69,12 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) return; } -#if 0 if (!context->radeon.radeonScreen->driScreen->dri2.enabled) { LOCK_HARDWARE(&context->radeon); UNLOCK_HARDWARE(&context->radeon); if (dPriv->numClipRects == 0) return; } -#endif R600_NEWPRIM(context); diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index b496543e1d..e46b21c85c 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -44,7 +44,6 @@ #include "tnl/t_vertex.h" #include "tnl/t_pipeline.h" -#include "radeon_lock.h" #include "r600_context.h" #include "r600_cmdbuf.h" @@ -296,9 +295,6 @@ static GLboolean r700RunRender(GLcontext * ctx, /* flush vtx */ //r700SyncSurf(context); /* */ - LOCK_HARDWARE( &context->radeon ); - UNLOCK_HARDWARE( &context->radeon ); - r700SendContextStates(context); r700SendViewportState(context, 0); r700SendRenderTargetState(context, 0); -- cgit v1.2.3 From c42a6163144aa5441262235818b59b3108833a61 Mon Sep 17 00:00:00 2001 From: Kevin DeKorte Date: Thu, 16 Jul 2009 11:57:14 -0400 Subject: Warning fixes --- src/mesa/drivers/dri/r600/r600_cmdbuf.c | 15 --------------- src/mesa/drivers/dri/r600/r600_cmdbuf.h | 19 +++++++++++++++++++ src/mesa/drivers/dri/r600/r700_clear.c | 1 - 3 files changed, 19 insertions(+), 16 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index ca9dab8639..74fec02584 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -52,22 +52,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_mipmap_tree.h" #include "radeon_reg.h" -struct r600_cs_manager_legacy -{ - struct radeon_cs_manager base; - struct radeon_context *ctx; - /* hack for scratch stuff */ - uint32_t pending_age; - uint32_t pending_count; -}; -struct r600_cs_reloc_legacy { - struct radeon_cs_reloc base; - uint32_t cindices; - uint32_t *indices; - uint32_t *reloc_indices; - struct offset_modifiers offset_mod; -}; static struct radeon_cs * r600_cs_create(struct radeon_cs_manager *csm, uint32_t ndw) diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.h b/src/mesa/drivers/dri/r600/r600_cmdbuf.h index 6c3ceb38fa..bd1ed7fdff 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.h +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.h @@ -118,6 +118,23 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define R600_IT_SET_CTL_CONST 0x00006F00 #define R600_IT_SURFACE_BASE_UPDATE 0x00007300 +struct r600_cs_manager_legacy +{ + struct radeon_cs_manager base; + struct radeon_context *ctx; + /* hack for scratch stuff */ + uint32_t pending_age; + uint32_t pending_count; +}; + +struct r600_cs_reloc_legacy { + struct radeon_cs_reloc base; + uint32_t cindices; + uint32_t *indices; + uint32_t *reloc_indices; + struct offset_modifiers offset_mod; +}; + extern int r600_cs_write_reloc(struct radeon_cs *cs, struct radeon_bo *bo, uint32_t read_domain, @@ -130,6 +147,8 @@ static inline void r600_cs_write_dword(struct radeon_cs *cs, uint32_t dword) cs->packets[cs->cdw++] = dword; } +struct radeon_cs_manager * r600_radeon_cs_manager_legacy_ctor(struct radeon_context *ctx); + /** * Write one dword to the command buffer. */ diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index e484845cc6..46a7010363 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -63,7 +63,6 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) { return; } - if (!context->radeon.radeonScreen->driScreen->dri2.enabled) { LOCK_HARDWARE(&context->radeon); UNLOCK_HARDWARE(&context->radeon); -- cgit v1.2.3 From 79a170108f1d9cad949cd2a34e4de597beae72db Mon Sep 17 00:00:00 2001 From: Kevin DeKorte Date: Fri, 17 Jul 2009 14:43:42 -0400 Subject: R6xx/r7xx: warning fixes patch from Kevin DeKorte with some minor fixes from me. --- src/mesa/drivers/dri/r600/r600_context.h | 2 ++ src/mesa/drivers/dri/r600/r600_emit.h | 2 ++ src/mesa/drivers/dri/r600/r600_texstate.c | 2 ++ src/mesa/drivers/dri/r600/r700_clear.c | 1 + src/mesa/drivers/dri/radeon/radeon_common_context.c | 4 ++++ 5 files changed, 11 insertions(+) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index 7957be6e0d..bcb33e1386 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -49,6 +49,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/colormac.h" #include "r700_chip.h" +#include "r600_tex.h" +#include "r700_oglprog.h" struct r600_context; typedef struct r600_context context_t; diff --git a/src/mesa/drivers/dri/r600/r600_emit.h b/src/mesa/drivers/dri/r600/r600_emit.h index d3c9ef59eb..661774d11e 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.h +++ b/src/mesa/drivers/dri/r600/r600_emit.h @@ -41,6 +41,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r600_cmdbuf.h" #include "radeon_reg.h" +void r600EmitCacheFlush(context_t *rmesa); + extern GLboolean r600EmitShader(GLcontext * ctx, void ** shaderbo, GLvoid * data, diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c index f56cc970d4..2466aa9595 100644 --- a/src/mesa/drivers/dri/r600/r600_texstate.c +++ b/src/mesa/drivers/dri/r600/r600_texstate.c @@ -49,6 +49,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_mipmap_tree.h" #include "r600_tex.h" +void r600UpdateTextureState(GLcontext * ctx); + void r600UpdateTextureState(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index 46a7010363..e84be38622 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -37,6 +37,7 @@ #include "r700_shaderinst.h" #include "r600_emit.h" +#include "r700_clear.h" static GLboolean r700ClearFast(context_t *context, GLbitfield mask) { diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index bbe7225ee6..285e015c92 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -46,6 +46,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "swrast_setup/swrast_setup.h" #include "tnl/tnl.h" +#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) /* +r6/r7 */ +#include "r600_context.h" +#endif + #define DRIVER_DATE "20090101" #ifndef RADEON_DEBUG -- cgit v1.2.3 From ef3ad412c746203727324edbd4cbe04079332d7c Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 12 Aug 2009 11:12:46 -0700 Subject: radeon: Minor warnings cleanup. --- src/mesa/drivers/dri/r600/r600_context.h | 11 +++++++++++ src/mesa/drivers/dri/r600/r700_clear.c | 1 + src/mesa/drivers/dri/r600/r700_oglprog.c | 2 +- src/mesa/drivers/dri/r600/r700_state.c | 2 -- src/mesa/drivers/dri/radeon/radeon_common_context.c | 7 +++---- 5 files changed, 16 insertions(+), 7 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index fbb8164af5..30ddce682c 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -158,6 +158,17 @@ extern GLboolean r700InitChipObject(context_t *context); extern GLboolean r700SendContextStates(context_t *context); extern GLboolean r700SendViewportState(context_t *context, int id); extern GLboolean r700SendRenderTargetState(context_t *context, int id); +extern GLboolean r700SendTextureState(context_t *context); +extern GLboolean r700SendDepthTargetState(context_t *context); +extern GLboolean r700SendUCPState(context_t *context); +extern GLboolean r700SendFSState(context_t *context); +extern void r700EmitState(GLcontext * ctx); + +extern GLboolean r700SyncSurf(context_t *context, + struct radeon_bo *pbo, + uint32_t read_domain, + uint32_t write_domain, + uint32_t sync_type); extern int r700SetupStreams(GLcontext * ctx); extern void r700SetupVTXConstants(GLcontext * ctx, diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index e84be38622..05d4af331e 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -31,6 +31,7 @@ #include "main/imports.h" #include "main/mtypes.h" #include "main/enums.h" +#include "swrast/swrast.h" #include "radeon_lock.h" #include "r600_context.h" diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.c b/src/mesa/drivers/dri/r600/r700_oglprog.c index 36de143b1a..c49b90c1cc 100644 --- a/src/mesa/drivers/dri/r600/r700_oglprog.c +++ b/src/mesa/drivers/dri/r600/r700_oglprog.c @@ -33,6 +33,7 @@ #include "tnl/tnl.h" #include "r600_context.h" +#include "r600_emit.h" #include "r700_oglprog.h" #include "r700_fragprog.h" @@ -87,7 +88,6 @@ static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) { struct r700_vertex_program * vp; struct r700_fragment_program * fp; - context_t *context = R700_CONTEXT(ctx); switch (prog->Target) { diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index e95f52400a..7f54cf9f56 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -179,8 +179,6 @@ static void r700FetchStateParameter(GLcontext * ctx, const gl_state_index state[STATE_LENGTH], GLfloat * value) { - context_t *context = R700_CONTEXT(ctx); - /* TODO */ } diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index d68aa2bd62..3a9468e88d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -295,11 +295,10 @@ void radeonDestroyContext(__DRIcontextPrivate *driContextPriv ) GET_CURRENT_CONTEXT(ctx); radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL; - - /* +r6/r7 */ - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; +#if RADEON_COMMON && defined(RADEON_COMMON_FOR_R600) /* +r6/r7 */ + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); - /* --------- */ +#endif if (radeon == current) { radeon_firevertices(radeon); -- cgit v1.2.3 From a47bb56d45af07b8cf1461c4c00620ec0b6aca41 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 20 Aug 2009 17:20:27 -0400 Subject: r600: improve state emission Slowing migrating to atom based system like the other radeon drivers. --- src/mesa/drivers/dri/r600/r600_context.h | 13 +++++ src/mesa/drivers/dri/r600/r700_chip.c | 33 +++++++++++- src/mesa/drivers/dri/r600/r700_chip.h | 1 + src/mesa/drivers/dri/r600/r700_clear.c | 6 --- src/mesa/drivers/dri/r600/r700_render.c | 66 ++++++++++++++++++----- src/mesa/drivers/dri/r600/r700_state.c | 93 ++++++++++++++++++++++++++++---- 6 files changed, 182 insertions(+), 30 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index 17ac088901..0cf2eb2daa 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -137,6 +137,19 @@ extern GLboolean r600CreateContext(const __GLcontextModes * glVisual, #define R700_CONTEXT_STATES(context) ((R700_CHIP_CONTEXT *)(&context->hw)) +#define R600_NEWPRIM( rmesa ) \ +do { \ + if ( rmesa->radeon.dma.flush ) \ + rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); \ +} while (0) + +#define R600_STATECHANGE(r600, atom) \ +do { \ + R600_NEWPRIM(r600); \ + (atom) = GL_TRUE; \ + r600->radeon.hw.is_dirty = GL_TRUE; \ +} while(0) + extern GLboolean r700SendSPIState(context_t *context); extern GLboolean r700SendVGTState(context_t *context); extern GLboolean r700SendSXState(context_t *context); diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index e67e544d53..7f46b0c468 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -258,6 +258,8 @@ GLboolean r700SendDepthTargetState(context_t *context) r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM, DB_ACTION_ENA_bit | DB_DEST_BASE_ENA_bit); + r700->db_target_dirty = GL_FALSE; + return GL_TRUE; } @@ -310,6 +312,8 @@ GLboolean r700SendRenderTargetState(context_t *context, int id) r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM, CB_ACTION_ENA_bit | (1 << (id + 6))); + r700->render_target[id].dirty = GL_FALSE; + return GL_TRUE; } @@ -343,6 +347,8 @@ GLboolean r700SendPSState(context_t *context) COMMIT_BATCH(); + r700->ps.dirty = GL_FALSE; + return GL_TRUE; } @@ -375,6 +381,8 @@ GLboolean r700SendVSState(context_t *context) COMMIT_BATCH(); + r700->vs.dirty = GL_FALSE; + return GL_TRUE; } @@ -416,6 +424,8 @@ GLboolean r700SendFSState(context_t *context) COMMIT_BATCH(); + r700->fs.dirty = GL_FALSE; + return GL_TRUE; } @@ -448,6 +458,8 @@ GLboolean r700SendViewportState(context_t *context, int id) COMMIT_BATCH(); + r700->viewport[id].dirty = GL_FALSE; + return GL_TRUE; } @@ -485,6 +497,8 @@ GLboolean r700SendSQConfig(context_t *context) COMMIT_BATCH(); + r700->sq_dirty = GL_FALSE; + return GL_TRUE; } @@ -495,7 +509,7 @@ GLboolean r700SendUCPState(context_t *context) int i; for (i = 0; i < R700_MAX_UCP; i++) { - if (r700->ucp[i].enabled) { + if (r700->ucp[i].enabled && r700->ucp[i].dirty) { BEGIN_BATCH_NO_AUTOSTATE(6); R600_OUT_BATCH_REGSEQ(PA_CL_UCP_0_X + (16 * i), 4); R600_OUT_BATCH(r700->ucp[i].PA_CL_UCP_0_X.u32All); @@ -504,6 +518,7 @@ GLboolean r700SendUCPState(context_t *context) R600_OUT_BATCH(r700->ucp[i].PA_CL_UCP_0_W.u32All); END_BATCH(); COMMIT_BATCH(); + r700->ucp[i].dirty = GL_FALSE; } } @@ -582,6 +597,8 @@ GLboolean r700SendSPIState(context_t *context) END_BATCH(); COMMIT_BATCH(); + r700->spi_dirty = GL_FALSE; + return GL_TRUE; } @@ -628,6 +645,8 @@ GLboolean r700SendVGTState(context_t *context) END_BATCH(); COMMIT_BATCH(); + r700->vgt_dirty = GL_FALSE; + return GL_TRUE; } @@ -643,6 +662,8 @@ GLboolean r700SendSXState(context_t *context) END_BATCH(); COMMIT_BATCH(); + r700->sx_dirty = GL_FALSE; + return GL_TRUE; } @@ -675,6 +696,8 @@ GLboolean r700SendDBState(context_t *context) END_BATCH(); COMMIT_BATCH(); + r700->db_dirty = GL_FALSE; + return GL_TRUE; } @@ -734,6 +757,8 @@ GLboolean r700SendCBState(context_t *context) COMMIT_BATCH(); + r700->cb_dirty = GL_FALSE; + return GL_TRUE; } @@ -764,6 +789,8 @@ GLboolean r700SendSUState(context_t *context) END_BATCH(); COMMIT_BATCH(); + r700->su_dirty = GL_FALSE; + return GL_TRUE; } @@ -787,6 +814,8 @@ GLboolean r700SendCLState(context_t *context) END_BATCH(); COMMIT_BATCH(); + r700->cl_dirty = GL_FALSE; + return GL_TRUE; } @@ -832,5 +861,7 @@ GLboolean r700SendSCState(context_t *context) END_BATCH(); COMMIT_BATCH(); + r700->sc_dirty = GL_FALSE; + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_chip.h b/src/mesa/drivers/dri/r600/r700_chip.h index 545e4a188a..c0ec4b0dd5 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.h +++ b/src/mesa/drivers/dri/r600/r700_chip.h @@ -275,6 +275,7 @@ typedef struct _R700_CHIP_CONTEXT union UINT_FLOAT DB_DEPTH_VIEW ; /* 0xA001 */ union UINT_FLOAT DB_DEPTH_BASE ; /* 0xA003 */ union UINT_FLOAT DB_DEPTH_INFO ; /* 0xA004 */ + GLboolean db_target_dirty; union UINT_FLOAT DB_HTILE_DATA_BASE ; /* 0xA005 */ union UINT_FLOAT DB_STENCIL_CLEAR ; /* 0xA00A */ union UINT_FLOAT DB_DEPTH_CLEAR ; /* 0xA00B */ diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index 05d4af331e..6d4ea90ccc 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -46,12 +46,6 @@ static GLboolean r700ClearFast(context_t *context, GLbitfield mask) return GL_FALSE; } -#define R600_NEWPRIM( rmesa ) \ - do { \ - if ( rmesa->radeon.dma.flush ) \ - rmesa->radeon.dma.flush( rmesa->radeon.glCtx ); \ - } while (0) - void r700Clear(GLcontext * ctx, GLbitfield mask) { context_t *context = R700_CONTEXT(ctx); diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index 1cc886d5e9..52f8f90dad 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -332,30 +332,70 @@ static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim } +static void r700EmitAtoms(GLcontext * ctx, GLboolean dirty) +{ + context_t *context = R700_CONTEXT(ctx); + radeonContextPtr radeon = &context->radeon; + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + + if ((r700->sq_dirty || radeon->hw.all_dirty) == dirty) + r700SendSQConfig(context); + r700SendUCPState(context); + if ((r700->sc_dirty || radeon->hw.all_dirty) == dirty) + r700SendSCState(context); + if ((r700->su_dirty || radeon->hw.all_dirty) == dirty) + r700SendSUState(context); + if ((r700->cl_dirty || radeon->hw.all_dirty) == dirty) + r700SendCLState(context); + if ((r700->cb_dirty || radeon->hw.all_dirty) == dirty) + r700SendCBState(context); + if ((r700->db_dirty || radeon->hw.all_dirty) == dirty) + r700SendDBState(context); + if ((r700->sx_dirty || radeon->hw.all_dirty) == dirty) + r700SendSXState(context); + if ((r700->vgt_dirty || radeon->hw.all_dirty) == dirty) + r700SendVGTState(context); + if ((r700->spi_dirty || radeon->hw.all_dirty) == dirty) + r700SendSPIState(context); + if ((r700->viewport[0].dirty || radeon->hw.all_dirty) == dirty) + r700SendViewportState(context, 0); + if ((r700->render_target[0].dirty || radeon->hw.all_dirty) == dirty) + r700SendRenderTargetState(context, 0); + if ((r700->db_target_dirty || radeon->hw.all_dirty) == dirty) + r700SendDepthTargetState(context); + +} + void r700EmitState(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); radeonContextPtr radeon = &context->radeon; + if (RADEON_DEBUG & (DEBUG_STATE|DEBUG_PRIMS)) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (radeon->vtbl.pre_emit_state) + radeon->vtbl.pre_emit_state(radeon); + if (radeon->cmdbuf.cs->cdw && !radeon->hw.is_dirty && !radeon->hw.all_dirty) return; rcommonEnsureCmdBufSpace(&context->radeon, 652, __FUNCTION__); - r700SendSQConfig(context); - r700SendUCPState(context); - r700SendSCState(context); - r700SendSUState(context); - r700SendCLState(context); - r700SendCBState(context); - r700SendDBState(context); - r700SendSXState(context); - r700SendVGTState(context); - r700SendSPIState(context); - r700SendViewportState(context, 0); - r700SendRenderTargetState(context, 0); - r700SendDepthTargetState(context); + if (!radeon->cmdbuf.cs->cdw) { + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin reemit state\n"); + + r700EmitAtoms(ctx, GL_FALSE); + } + + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin dirty state\n"); + + r700EmitAtoms(ctx, GL_TRUE); + radeon->hw.is_dirty = GL_FALSE; + radeon->hw.all_dirty = GL_FALSE; } diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index 835b5e18c2..ada394ec61 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -148,8 +148,15 @@ void r700UpdateViewportOffset(GLcontext * ctx) //------------------ GLfloat tx = v[MAT_TX] + xoffset; GLfloat ty = (-v[MAT_TY]) + yoffset; - r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All = tx; - r700->viewport[id].PA_CL_VPORT_YOFFSET.f32All = ty; + if (r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All != tx || + r700->viewport[id].PA_CL_VPORT_YOFFSET.f32All != ty) { + /* Note: this should also modify whatever data the context reset + * code uses... + */ + R600_STATECHANGE(context, r700->viewport[id].dirty); + r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All = tx; + r700->viewport[id].PA_CL_VPORT_YOFFSET.f32All = ty; + } radeonUpdateScissor(ctx); } @@ -161,6 +168,10 @@ void r700UpdateViewportOffset(GLcontext * ctx) //------------------ void r700UpdateDrawBuffer(GLcontext * ctx) /* TODO */ //--------------------- { context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + + R600_STATECHANGE(context, r700->render_target[0].dirty); + R600_STATECHANGE(context, r700->db_target_dirty); r700SetRenderTarget(context, 0); r700SetDepthTarget(context); @@ -233,6 +244,9 @@ static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //----------- r700UpdateStateParameters(ctx, new_state); + R600_STATECHANGE(context, r700->cl_dirty); + R600_STATECHANGE(context, r700->spi_dirty); + if(GL_TRUE == r700->bEnablePerspective) { /* Do scale XY and Z by 1/W0 for perspective correction on pos. For orthogonal case, set both to one. */ @@ -256,14 +270,15 @@ static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //----------- SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, LINEAR_GRADIENT_ENA_bit); } - context->radeon.NewGLState |= new_state; + context->radeon.NewGLState |= new_state; } static void r700SetDepthState(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + R600_STATECHANGE(context, r700->db_dirty); if (ctx->Depth.Test) { @@ -331,6 +346,8 @@ static void r700SetAlphaState(GLcontext * ctx) uint32_t alpha_func = REF_ALWAYS; GLboolean really_enabled = ctx->Color.AlphaEnabled; + R600_STATECHANGE(context, r700->sx_dirty); + switch (ctx->Color.AlphaFunc) { case GL_NEVER: alpha_func = REF_NEVER; @@ -383,6 +400,8 @@ static void r700BlendColor(GLcontext * ctx, const GLfloat cf[4]) //------------- context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + R600_STATECHANGE(context, r700->cb_dirty); + r700->CB_BLEND_RED.f32All = cf[0]; r700->CB_BLEND_GREEN.f32All = cf[1]; r700->CB_BLEND_BLUE.f32All = cf[2]; @@ -451,6 +470,8 @@ static void r700SetBlendState(GLcontext * ctx) int id = 0; uint32_t blend_reg = 0, eqn, eqnA; + R600_STATECHANGE(context, r700->cb_dirty); + if (RGBA_LOGICOP_ENABLED(ctx) || !ctx->Color.BlendEnabled) { SETfield(blend_reg, BLEND_ONE, COLOR_SRCBLEND_shift, COLOR_SRCBLEND_mask); @@ -637,8 +658,11 @@ static GLuint translate_logicop(GLenum logicop) */ static void r700SetLogicOpState(GLcontext *ctx) { + context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw); + R600_STATECHANGE(context, r700->cb_dirty); + if (RGBA_LOGICOP_ENABLED(ctx)) SETfield(r700->CB_COLOR_CONTROL.u32All, translate_logicop(ctx->Color.LogicOp), ROP3_shift, ROP3_mask); @@ -658,7 +682,10 @@ static void r700LogicOpcode(GLcontext *ctx, GLenum logicop) static void r700UpdateCulling(GLcontext * ctx) { - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw); + + R600_STATECHANGE(context, r700->su_dirty); CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, FACE_bit); CLEARbit(r700->PA_SU_SC_MODE_CNTL.u32All, CULL_FRONT_bit); @@ -703,7 +730,11 @@ static void r700UpdateCulling(GLcontext * ctx) static void r700UpdateLineStipple(GLcontext * ctx) { - R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw); + + R600_STATECHANGE(context, r700->sc_dirty); + if (ctx->Line.StippleFlag) { SETbit(r700->PA_SC_MODE_CNTL.u32All, LINE_STIPPLE_ENABLE_bit); @@ -778,14 +809,17 @@ static void r700Enable(GLcontext * ctx, GLenum cap, GLboolean state) //--------- static void r700ColorMask(GLcontext * ctx, GLboolean r, GLboolean g, GLboolean b, GLboolean a) //------------------ { + context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&R700_CONTEXT(ctx)->hw); unsigned int mask = ((r ? 1 : 0) | (g ? 2 : 0) | (b ? 4 : 0) | (a ? 8 : 0)); - if (mask != r700->CB_SHADER_MASK.u32All) + if (mask != r700->CB_SHADER_MASK.u32All) { + R600_STATECHANGE(context, r700->cb_dirty); SETfield(r700->CB_SHADER_MASK.u32All, mask, OUTPUT0_ENABLE_shift, OUTPUT0_ENABLE_mask); + } } /** @@ -841,6 +875,8 @@ static void r700ShadeModel(GLcontext * ctx, GLenum mode) //-------------------- context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + R600_STATECHANGE(context, r700->spi_dirty); + /* also need to set/clear FLAT_SHADE bit per param in SPI_PS_INPUT_CNTL_[0-31] */ switch (mode) { case GL_FLAT: @@ -862,6 +898,8 @@ static void r700PointSize(GLcontext * ctx, GLfloat size) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + R600_STATECHANGE(context, r700->su_dirty); + /* We need to clamp to user defined range here, because * the HW clamping happens only for per vertex point size. */ size = CLAMP(size, ctx->Point.MinSize, ctx->Point.MaxSize); @@ -882,6 +920,8 @@ static void r700PointParameter(GLcontext * ctx, GLenum pname, const GLfloat * pa context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + R600_STATECHANGE(context, r700->su_dirty); + /* format is 12.4 fixed point */ switch (pname) { case GL_POINT_SIZE_MIN: @@ -966,6 +1006,7 @@ static void r700SetStencilState(GLcontext * ctx, GLboolean state) } if (hw_stencil) { + R600_STATECHANGE(context, r700->db_dirty); if (state) SETbit(r700->DB_DEPTH_CONTROL.u32All, STENCIL_ENABLE_bit); else @@ -983,6 +1024,8 @@ static void r700StencilFuncSeparate(GLcontext * ctx, GLenum face, //fixme //r300CatchStencilFallback(ctx); + R600_STATECHANGE(context, r700->db_dirty); + //front SETfield(r700->DB_STENCILREFMASK.u32All, ctx->Stencil.Ref[0], STENCILREF_shift, STENCILREF_mask); @@ -1012,6 +1055,8 @@ static void r700StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) / //fixme //r300CatchStencilFallback(ctx); + R600_STATECHANGE(context, r700->db_dirty); + // front SETfield(r700->DB_STENCILREFMASK.u32All, ctx->Stencil.WriteMask[0], STENCILWRITEMASK_shift, STENCILWRITEMASK_mask); @@ -1032,6 +1077,8 @@ static void r700StencilOpSeparate(GLcontext * ctx, GLenum face, //fixme //r300CatchStencilFallback(ctx); + R600_STATECHANGE(context, r700->db_dirty); + SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_op(ctx->Stencil.FailFunc[0]), STENCILFAIL_shift, STENCILFAIL_mask); SETfield(r700->DB_DEPTH_CONTROL.u32All, translate_stencil_op(ctx->Stencil.ZFailFunc[0]), @@ -1074,7 +1121,7 @@ static void r700UpdateWindow(GLcontext * ctx, int id) //-------------------- GLfloat sz = v[MAT_SZ] * depthScale; GLfloat tz = v[MAT_TZ] * depthScale; - /* TODO : Need DMA flush as well. */ + R600_STATECHANGE(context, r700->viewport[id].dirty); r700->viewport[id].PA_CL_VPORT_XSCALE.f32All = sx; r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All = tx; @@ -1112,10 +1159,13 @@ static void r700LineWidth(GLcontext * ctx, GLfloat widthf) //--------------- context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); uint32_t lineWidth = (uint32_t)((widthf * 0.5) * (1 << 4)); + + R600_STATECHANGE(context, r700->su_dirty); + if (lineWidth > 0xFFFF) - lineWidth = 0xFFFF; + lineWidth = 0xFFFF; SETfield(r700->PA_SU_LINE_CNTL.u32All,(uint16_t)lineWidth, - PA_SU_LINE_CNTL__WIDTH_shift, PA_SU_LINE_CNTL__WIDTH_mask); + PA_SU_LINE_CNTL__WIDTH_shift, PA_SU_LINE_CNTL__WIDTH_mask); } static void r700LineStipple(GLcontext *ctx, GLint factor, GLushort pattern) @@ -1123,6 +1173,8 @@ static void r700LineStipple(GLcontext *ctx, GLint factor, GLushort pattern) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + R600_STATECHANGE(context, r700->sc_dirty); + SETfield(r700->PA_SC_LINE_STIPPLE.u32All, pattern, LINE_PATTERN_shift, LINE_PATTERN_mask); SETfield(r700->PA_SC_LINE_STIPPLE.u32All, (factor-1), REPEAT_COUNT_shift, REPEAT_COUNT_mask); SETfield(r700->PA_SC_LINE_STIPPLE.u32All, 1, AUTO_RESET_CNTL_shift, AUTO_RESET_CNTL_mask); @@ -1133,6 +1185,8 @@ static void r700SetPolygonOffsetState(GLcontext * ctx, GLboolean state) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + R600_STATECHANGE(context, r700->su_dirty); + if (state) { SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, POLY_OFFSET_FRONT_ENABLE_bit); SETbit(r700->PA_SU_SC_MODE_CNTL.u32All, POLY_OFFSET_BACK_ENABLE_bit); @@ -1161,6 +1215,8 @@ static void r700PolygonOffset(GLcontext * ctx, GLfloat factor, GLfloat units) // factor *= 12.0; + R600_STATECHANGE(context, r700->su_dirty); + r700->PA_SU_POLY_OFFSET_FRONT_SCALE.f32All = factor; r700->PA_SU_POLY_OFFSET_FRONT_OFFSET.f32All = constant; r700->PA_SU_POLY_OFFSET_BACK_SCALE.f32All = factor; @@ -1172,6 +1228,8 @@ static void r700UpdatePolygonMode(GLcontext * ctx) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + R600_STATECHANGE(context, r700->su_dirty); + SETfield(r700->PA_SU_SC_MODE_CNTL.u32All, X_DISABLE_POLY_MODE, POLY_MODE_shift, POLY_MODE_mask); /* Only do something if a polygon mode is wanted, default is GL_FILL */ @@ -1247,6 +1305,8 @@ static void r700ClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq ) p = (GLint) plane - (GLint) GL_CLIP_PLANE0; ip = (GLint *)ctx->Transform._ClipUserPlane[p]; + R600_STATECHANGE(context, r700->ucp[p].dirty); + r700->ucp[p].PA_CL_UCP_0_X.u32All = ip[0]; r700->ucp[p].PA_CL_UCP_0_Y.u32All = ip[1]; r700->ucp[p].PA_CL_UCP_0_Z.u32All = ip[2]; @@ -1260,6 +1320,9 @@ static void r700SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state) GLuint p; p = cap - GL_CLIP_PLANE0; + + R600_STATECHANGE(context, r700->cl_dirty); + if (state) { r700->PA_CL_CLIP_CNTL.u32All |= (UCP_ENA_0_bit << p); r700->ucp[p].enabled = GL_TRUE; @@ -1293,6 +1356,8 @@ void r700SetScissor(context_t *context) //--------------- y2 = rrb->dPriv->y + rrb->dPriv->h; } + R600_STATECHANGE(context, r700->sc_dirty); + /* window */ SETbit(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, WINDOW_OFFSET_DISABLE_bit); SETfield(r700->PA_SC_WINDOW_SCISSOR_TL.u32All, x1, @@ -1361,6 +1426,9 @@ static void r700SetRenderTarget(context_t *context, int id) return; } + R600_STATECHANGE(context, r700->render_target[id].dirty); + R600_STATECHANGE(context, r700->cb_dirty); + /* screen/window/view */ SETfield(r700->CB_TARGET_MASK.u32All, 0xF, (4 * id), TARGET0_ENABLE_mask); @@ -1407,6 +1475,9 @@ static void r700SetDepthTarget(context_t *context) if (!rrb) return; + R600_STATECHANGE(context, r700->db_target_dirty); + R600_STATECHANGE(context, r700->db_dirty); + /* depth buf */ r700->DB_DEPTH_SIZE.u32All = 0; r700->DB_DEPTH_BASE.u32All = 0; @@ -1467,6 +1538,8 @@ static void r700InitSQConfig(GLcontext * ctx) int num_gs_stack_entries; int num_es_stack_entries; + R600_STATECHANGE(context, r700->sq_dirty); + // SQ ps_prio = 0; vs_prio = 1; -- cgit v1.2.3 From aabb36fdd58ff891e3a0946e2326d458f5a4d832 Mon Sep 17 00:00:00 2001 From: Pauli Nieminen Date: Sun, 30 Aug 2009 22:08:38 +0300 Subject: r600: Convert to shared debug code and add a few new debug messages. There is only a few functions that have debugging enabled now. --- src/mesa/drivers/dri/r600/Makefile | 9 ++-- src/mesa/drivers/dri/r600/r600_context.c | 10 +++- src/mesa/drivers/dri/r600/r600_emit.c | 28 ++++++----- src/mesa/drivers/dri/r600/r600_tex.c | 20 ++++---- src/mesa/drivers/dri/r600/r600_texstate.c | 8 ++-- src/mesa/drivers/dri/r600/r700_assembler.c | 72 ++++++++++++++--------------- src/mesa/drivers/dri/r600/r700_chip.c | 34 ++++++++++++++ src/mesa/drivers/dri/r600/r700_clear.c | 5 +- src/mesa/drivers/dri/r600/r700_debug.c | 35 +++----------- src/mesa/drivers/dri/r600/r700_debug.h | 60 ++---------------------- src/mesa/drivers/dri/r600/r700_oglprog.c | 6 +++ src/mesa/drivers/dri/r600/r700_render.c | 10 ++-- src/mesa/drivers/dri/r600/r700_shaderinst.c | 14 +++--- src/mesa/drivers/dri/r600/r700_vertprog.c | 3 +- 14 files changed, 141 insertions(+), 173 deletions(-) (limited to 'src/mesa/drivers/dri/r600/r700_clear.c') diff --git a/src/mesa/drivers/dri/r600/Makefile b/src/mesa/drivers/dri/r600/Makefile index 3c3100ab91..d925a2dfe3 100644 --- a/src/mesa/drivers/dri/r600/Makefile +++ b/src/mesa/drivers/dri/r600/Makefile @@ -27,16 +27,17 @@ COMMON_SOURCES = \ ../common/dri_util.c RADEON_COMMON_SOURCES = \ - radeon_texture.c \ + radeon_bo_legacy.c \ radeon_common_context.c \ radeon_common.c \ + radeon_cs_legacy.c \ radeon_dma.c \ + radeon_debug.c \ + radeon_fbo.c \ radeon_lock.c \ - radeon_bo_legacy.c \ - radeon_cs_legacy.c \ radeon_mipmap_tree.c \ radeon_span.c \ - radeon_fbo.c \ + radeon_texture.c \ radeon_queryobj.c DRIVER_SOURCES = \ diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index 6a066f3510..251c124cbf 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -56,6 +56,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "drivers/common/driverfuncs.h" +#include "radeon_debug.h" #include "r600_context.h" #include "radeon_common_context.h" #include "radeon_span.h" @@ -225,8 +226,10 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, /* Allocate the R600 context */ r600 = (context_t*) CALLOC(sizeof(*r600)); - if (!r600) + if (!r600) { + radeon_error("Failed to allocate memory for context.\n"); return GL_FALSE; + } if (!(screen->chip_flags & RADEON_CHIPSET_TCL)) hw_tcl_on = future_hw_tcl_on = 0; @@ -255,6 +258,7 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, if (!radeonInitContext(&r600->radeon, &functions, glVisual, driContextPriv, sharedContextPrivate)) { + radeon_error("Initializing context failed.\n"); FREE(r600); return GL_FALSE; } @@ -347,6 +351,8 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, ctx->VertexProgram._MaintainTnlProgram = GL_TRUE; ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; + radeon_init_debug(); + driInitExtensions(ctx, card_extensions, GL_TRUE); if (r600->radeon.radeonScreen->kernel_mm) driInitExtensions(ctx, mm_extensions, GL_FALSE); @@ -375,7 +381,7 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, TNL_CONTEXT(ctx)->Driver.RunPipeline = r600RunPipeline; if (driQueryOptionb(&r600->radeon.optionCache, "no_rast")) { - fprintf(stderr, "disabling 3D acceleration\n"); + radeon_warning("disabling 3D acceleration\n"); #if R200_MERGED FALLBACK(&r600->radeon, RADEON_FALLBACK_DISABLE, 1); #endif diff --git a/src/mesa/drivers/dri/r600/r600_emit.c b/src/mesa/drivers/dri/r600/r600_emit.c index f8c56e7774..5c250c2418 100644 --- a/src/mesa/drivers/dri/r600/r600_emit.c +++ b/src/mesa/drivers/dri/r600/r600_emit.c @@ -60,28 +60,30 @@ GLboolean r600EmitShader(GLcontext * ctx, radeonContextPtr radeonctx = RADEON_CONTEXT(ctx); struct radeon_bo * pbo; uint32_t *out; - shader_again_alloc: pbo = radeon_bo_open(radeonctx->radeonScreen->bom, - 0, - sizeinDWORD * 4, - 256, - RADEON_GEM_DOMAIN_GTT, - 0); + 0, + sizeinDWORD * 4, + 256, + RADEON_GEM_DOMAIN_GTT, + 0); + + radeon_print(RADEON_SHADER, RADEON_NORMAL, "%s %p size %d: %s\n", __func__, pbo, sizeinDWORD, szShaderUsage); if (!pbo) { + radeon_print(RADEON_MEMORY | RADEON_CS, RADEON_IMPORTANT, "No memory for buffer object. Flushing command buffer.\n"); rcommonFlushCmdBuf(radeonctx, __FUNCTION__); goto shader_again_alloc; } radeon_cs_space_add_persistent_bo(radeonctx->cmdbuf.cs, - pbo, - RADEON_GEM_DOMAIN_GTT, 0); + pbo, + RADEON_GEM_DOMAIN_GTT, 0); - if (radeon_cs_space_check_with_bo(radeonctx->cmdbuf.cs, - pbo, - RADEON_GEM_DOMAIN_GTT, 0)) { - fprintf(stderr,"failure to revalidate BOs - badness\n"); + if (radeon_cs_space_check_with_bo(radeonctx->cmdbuf.cs, + pbo, + RADEON_GEM_DOMAIN_GTT, 0)) { + radeon_error("failure to revalidate BOs - badness\n"); return GL_FALSE; } @@ -103,6 +105,8 @@ GLboolean r600DeleteShader(GLcontext * ctx, { struct radeon_bo * pbo = (struct radeon_bo *)shaderbo; + radeon_print(RADEON_SHADER, RADEON_NORMAL, "%s: %p\n", __func__, pbo); + if (pbo) { if (pbo->ptr) radeon_bo_unmap(pbo); diff --git a/src/mesa/drivers/dri/r600/r600_tex.c b/src/mesa/drivers/dri/r600/r600_tex.c index a06a1fa921..d105b90cd1 100644 --- a/src/mesa/drivers/dri/r600/r600_tex.c +++ b/src/mesa/drivers/dri/r600/r600_tex.c @@ -67,7 +67,7 @@ static unsigned int translate_wrap_mode(GLenum wrapmode) case GL_MIRROR_CLAMP_TO_EDGE_EXT: return SQ_TEX_MIRROR_ONCE_LAST_TEXEL; case GL_MIRROR_CLAMP_TO_BORDER_EXT: return SQ_TEX_MIRROR_ONCE_BORDER; default: - _mesa_problem(NULL, "bad wrap mode in %s", __FUNCTION__); + radeon_error("bad wrap mode in %s", __FUNCTION__); return 0; } } @@ -208,8 +208,7 @@ static void r600SetTexFilter(radeonTexObjPtr t, GLenum minf, GLenum magf, GLfloa | R300_TX_MIN_FILTER_ANISO | R300_TX_MIN_FILTER_MIP_LINEAR | aniso_filter(anisotropy);*/ - if (RADEON_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "Using maximum anisotropy of %f\n", anisotropy); + radeon_print(RADEON_TEXTURE, RADEON_NORMAL, "Using maximum anisotropy of %f\n", anisotropy); return; } @@ -288,10 +287,9 @@ static void r600TexParameter(GLcontext * ctx, GLenum target, { radeonTexObj* t = radeon_tex_obj(texObj); - if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) { - fprintf(stderr, "%s( %s )\n", __FUNCTION__, + radeon_print(RADEON_STATE | RADEON_TEXTURE, RADEON_VERBOSE, + "%s( %s )\n", __FUNCTION__, _mesa_lookup_enum_by_nr(pname)); - } switch (pname) { case GL_TEXTURE_MIN_FILTER: @@ -351,11 +349,10 @@ static void r600DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) context_t* rmesa = R700_CONTEXT(ctx); radeonTexObj* t = radeon_tex_obj(texObj); - if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) { - fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__, + radeon_print(RADEON_STATE | RADEON_TEXTURE, RADEON_NORMAL, + "%s( %p (target = %s) )\n", __FUNCTION__, (void *)texObj, _mesa_lookup_enum_by_nr(texObj->Target)); - } if (rmesa) { int i; @@ -393,10 +390,9 @@ static struct gl_texture_object *r600NewTextureObject(GLcontext * ctx, radeonTexObj* t = CALLOC_STRUCT(radeon_tex_obj); - if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_TEXTURE)) { - fprintf(stderr, "%s( %p (target = %s) )\n", __FUNCTION__, + radeon_print(RADEON_STATE | RADEON_TEXTURE, RADEON_NORMAL, + "%s( %p (target = %s) )\n", __FUNCTION__, t, _mesa_lookup_enum_by_nr(target)); - } _mesa_initialize_texture_object(&t->base, name, target); t->base.MaxAnisotropy = rmesa->radeon.initialMaxAnisotropy; diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c index 9f1bf45246..237eaa8249 100644 --- a/src/mesa/drivers/dri/r600/r600_texstate.c +++ b/src/mesa/drivers/dri/r600/r600_texstate.c @@ -567,7 +567,7 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex if (!t->image_override) { if (!r600GetTexFormat(texObj, firstImage->TexFormat->MesaFormat)) { - _mesa_problem(NULL, "unexpected texture format in %s", + radeon_error("unexpected texture format in %s\n", __FUNCTION__); return; } @@ -593,7 +593,7 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex SETfield(t->SQ_TEX_RESOURCE1, 0, TEX_DEPTH_shift, TEX_DEPTH_mask); break; default: - _mesa_problem(NULL, "unexpected texture target type in %s", __FUNCTION__); + radeon_error("unexpected texture target type in %s\n", __FUNCTION__); return; } @@ -675,9 +675,7 @@ GLboolean r600ValidateBuffers(GLcontext * ctx) continue; if (!r600_validate_texture(ctx, ctx->Texture.Unit[i]._Current)) { - _mesa_warning(ctx, - "failed to validate texture for unit %d.\n", - i); + radeon_warning("failed to validate texture for unit %d.\n", i); } t = radeon_tex_obj(ctx->Texture.Unit[i]._Current); if (t->image_override && t->bo) diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index 834c23593b..2d8480daaf 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -33,8 +33,8 @@ #include "main/mtypes.h" #include "main/imports.h" +#include "radeon_debug.h" #include "r600_context.h" -#include "r700_debug.h" #include "r700_assembler.h" @@ -366,8 +366,8 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) case SQ_OP2_INST_COS: return 1; - default: r700_error(TODO_ASM_NEEDIMPINST, - "Need instruction operand number. \n");; + default: radeon_error( + "Need instruction operand number for %x.\n", pAsm->D.dst.opcode); }; return 3; @@ -531,7 +531,7 @@ int check_current_clause(r700_AssemblerBase* pAsm, case CF_EMPTY_CLAUSE: break; default: - r700_error(ERROR_ASM_VTX_CLAUSE, + radeon_error( "Unknown CF_CLAUSE_TYPE (%d) in check_current_clause. \n", (int) new_clause_type); return GL_FALSE; } @@ -565,7 +565,7 @@ int check_current_clause(r700_AssemblerBase* pAsm, } else { - r700_error(ERROR_ASM_ALLOCEXPORTCF, + radeon_error( "Error allocating new EXPORT CF instruction in check_current_clause. \n"); return GL_FALSE; } @@ -578,7 +578,7 @@ int check_current_clause(r700_AssemblerBase* pAsm, pAsm->cf_current_clause_type = CF_OTHER_CLAUSE; break; default: - r700_error(ERROR_ASM_UNKOWNCLAUSE, + radeon_error( "Unknown CF_CLAUSE_TYPE (%d) in check_current_clause. \n", (int) new_clause_type); return GL_FALSE; } @@ -611,7 +611,7 @@ GLboolean add_vfetch_instruction(r700_AssemblerBase* pAsm, } else { - r700_error(ERROR_ASM_ALLOCVTXCF, "Could not allocate a new VFetch CF instruction."); + radeon_error("Could not allocate a new VFetch CF instruction.\n"); return GL_FALSE; } @@ -661,7 +661,7 @@ GLboolean add_tex_instruction(r700_AssemblerBase* pAsm, } else { - r700_error(ERROR_ASM_ALLOCTEXCF, "Could not allocate a new TEX CF instruction."); + radeon_error("Could not allocate a new TEX CF instruction.\n"); return GL_FALSE; } @@ -1047,7 +1047,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, } break; default: - r700_error(ERROR_ASM_SRCARGUMENT, "Invalid source argument type"); + radeon_error("Invalid source argument type\n"); return GL_FALSE; } } @@ -1094,7 +1094,7 @@ GLboolean assemble_dst(r700_AssemblerBase *pAsm) } break; default: - r700_error(ERROR_ASM_DSTARGUMENT, "Invalid destination output argument type"); + radeon_error("Invalid destination output argument type\n"); return GL_FALSE; } @@ -1134,7 +1134,7 @@ GLboolean tex_dst(r700_AssemblerBase *pAsm) } else { - r700_error(ERROR_ASM_DSTARGUMENT, "Invalid destination output argument type"); + radeon_error("Invalid destination output argument type\n"); return GL_FALSE; } @@ -1188,7 +1188,7 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) } else { - r700_error(ERROR_ASM_BADTEXSRC, "Invalid source texcoord for TEX instruction"); + radeon_error("Invalid source texcoord for TEX instruction\n"); return GL_FALSE; } @@ -1269,7 +1269,7 @@ GLboolean assemble_tex_instruction(r700_AssemblerBase *pAsm, GLboolean normalize } else { - r700_error(ERROR_ASM_TEXDSTBADTYPE, "Only temp destination registers supported for TEX dest regs."); + radeon_error("Only temp destination registers supported for TEX dest regs.\n"); return GL_FALSE; } @@ -1362,7 +1362,7 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, } else { - r700_error(ERROR_ASM_ALUSRCBADTYPE, "Source (%d) register type (%d) not one of TEMP, INPUT, or CONSTANT.", + radeon_error("Source (%d) register type (%d) not one of TEMP, INPUT, or CONSTANT.\n", source_index, pSource->rtype); return GL_FALSE; } @@ -1397,7 +1397,7 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, src_chan = SQ_CHAN_X; break; default: - r700_error(ERROR_ASM_ALUSRCSELECT, "Unknown source select value (%d) in assemble_alu_src()."); + radeon_error("Unknown source select value (%d) in assemble_alu_src().\n", channel_swizzle); return GL_FALSE; break; } @@ -1432,7 +1432,7 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, alu_instruction_ptr->m_Word1_OP3.f.src2_neg = src_neg; break; default: - r700_error(ERROR_ASM_ALUSRCNUMBER, "Only three sources allowed in ALU opcodes."); + radeon_error("Only three sources allowed in ALU opcodes.\n"); return GL_FALSE; break; } @@ -1467,7 +1467,7 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, } else { - r700_error(ERROR_ASM_ALLOCALUCF, "Could not allocate a new ALU CF instruction."); + radeon_error("Could not allocate a new ALU CF instruction.\n"); return GL_FALSE; } @@ -1635,7 +1635,7 @@ GLboolean reserve_cfile(r700_AssemblerBase* pAsm, } else { - r700_error(ERROR_ASM_CONSTCHANNEL, "All cfile read ports are used, cannot reference C$sel, channel $chan."); + radeon_error("All cfile read ports are used, cannot reference C$sel, channel $chan.\n"); return GL_FALSE; } return GL_TRUE; @@ -1649,7 +1649,7 @@ GLboolean reserve_gpr(r700_AssemblerBase* pAsm, GLuint sel, GLuint chan, GLuint } else if(pAsm->hw_gpr[cycle][chan] != (int)sel) { - r700_error(ERROR_ASM_BADGPRRESERVE, "Another scalar operation has already used GPR read port for given channel"); + radeon_error("Another scalar operation has already used GPR read port for given channel\n"); return GL_FALSE; } @@ -1689,7 +1689,7 @@ GLboolean cycle_for_scalar_bank_swizzle(const int swiz, const int sel, GLuint* p } break; default: - r700_error(ERROR_ASM_BADSCALARBZ, "Bad Scalar bank swizzle value"); + radeon_error("Bad Scalar bank swizzle value\n"); break; } @@ -1737,7 +1737,7 @@ GLboolean cycle_for_vector_bank_swizzle(const int swiz, const int sel, GLuint* p } break; default: - r700_error(ERROR_ASM_BADVECTORBZ, "Bad Vec bank swizzle value"); + radeon_error("Bad Vec bank swizzle value\n"); return GL_FALSE; break; } @@ -2056,7 +2056,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) } else { - r700_error(ERROR_ASM_ALUDSTBADTYPE, "Only temp destination registers supported for ALU dest regs."); + radeon_error("Only temp destination registers supported for ALU dest regs.\n"); return GL_FALSE; } @@ -2207,13 +2207,13 @@ GLboolean next_ins(r700_AssemblerBase *pAsm) if (pILInst->TexSrcTarget == TEXTURE_RECT_INDEX) { if( GL_FALSE == assemble_tex_instruction(pAsm, GL_FALSE) ) { - r700_error(ERROR_ASM_TEXINSTRUCTION, "Error assembling TEX instruction"); + radeon_error("Error assembling TEX instruction\n"); return GL_FALSE; } } else { if( GL_FALSE == assemble_tex_instruction(pAsm, GL_TRUE) ) { - r700_error(ERROR_ASM_TEXINSTRUCTION, "Error assembling TEX instruction"); + radeon_error("Error assembling TEX instruction\n"); return GL_FALSE; } } @@ -2222,7 +2222,7 @@ GLboolean next_ins(r700_AssemblerBase *pAsm) { //ALU if( GL_FALSE == assemble_alu_instruction(pAsm) ) { - r700_error(ERROR_ASM_TEXINSTRUCTION, "Error assembling ALU instruction"); + radeon_error("Error assembling ALU instruction\n"); return GL_FALSE; } } @@ -2367,7 +2367,7 @@ GLboolean assemble_ADD(r700_AssemblerBase *pAsm) GLboolean assemble_BAD(char *opcode_str) { - r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction (%s)", opcode_str); + radeon_error("Not yet implemented instruction (%s)\n", opcode_str); return GL_FALSE; } @@ -3381,7 +3381,7 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) if (GL_TRUE == src_const) { - r700_error(TODO_ASM_CONSTTEXADDR, "TODO: Texture coordinates from a constant register not supported."); + radeon_error("TODO: Texture coordinates from a constant register not supported.\n"); return GL_FALSE; } @@ -3391,7 +3391,7 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; break; case OPCODE_TXB: - r700_error(TODO_ASM_TXB, "do not support TXB yet"); + radeon_error("do not support TXB yet\n"); return GL_FALSE; break; case OPCODE_TXP: @@ -3399,7 +3399,7 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; break; default: - r700_error(ERROR_ASM_BADTEXINST, "Internal error: bad texture op (not TEX)"); + radeon_error("Internal error: bad texture op (not TEX)\n"); return GL_FALSE; break; } @@ -3581,12 +3581,12 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, break; case OPCODE_ARL: - r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_ARL "); + radeon_error("Not yet implemented instruction OPCODE_ARL \n"); //if ( GL_FALSE == assemble_BAD("ARL") ) return GL_FALSE; break; case OPCODE_ARR: - r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_ARR "); + radeon_error("Not yet implemented instruction OPCODE_ARR \n"); //if ( GL_FALSE == assemble_BAD("ARR") ) return GL_FALSE; break; @@ -3617,7 +3617,7 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_EXP: - r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_EXP "); + radeon_error("Not yet implemented instruction OPCODE_EXP \n"); //if ( GL_FALSE == assemble_BAD("EXP") ) return GL_FALSE; break; // approx of EX2 @@ -3653,7 +3653,7 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_LOG: - r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_LOG "); + radeon_error("Not yet implemented instruction OPCODE_LOG \n"); //if ( GL_FALSE == assemble_BAD("LOG") ) return GL_FALSE; break; // approx of LG2 @@ -3752,7 +3752,7 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_ELSE : - r700_error(TODO_ASM_NEEDIMPINST, "Not yet implemented instruction OPCODE_ELSE "); + radeon_error("Not yet implemented instruction OPCODE_ELSE \n"); //if ( GL_FALSE == assemble_BAD("ELSE") ) return GL_FALSE; break; @@ -3774,7 +3774,7 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_TRUE; default: - r700_error(ERROR_ASM_UNKNOWNILINST, "internal: unknown instruction"); + radeon_error("internal: unknown instruction\n"); return GL_FALSE; } } @@ -3818,7 +3818,7 @@ GLboolean Process_Export(r700_AssemblerBase* pAsm, break; default: - r700_error(ERROR_ASM_BADEXPORTTYPE, "Unknown export type: %d", type); + radeon_error("Unknown export type: %d\n", type); return GL_FALSE; break; } diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index 16f1a3df36..4fc624ed0b 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -49,6 +49,8 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) unsigned int i; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { radeonTexObj *t = r700->textures[i]; if (t) { @@ -93,6 +95,7 @@ static void r700SendTexSamplerState(GLcontext *ctx, struct radeon_state_atom *at R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); unsigned int i; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { radeonTexObj *t = r700->textures[i]; @@ -115,6 +118,7 @@ static void r700SendTexBorderColorState(GLcontext *ctx, struct radeon_state_atom R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); unsigned int i; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { radeonTexObj *t = r700->textures[i]; @@ -141,6 +145,7 @@ static void r700SetupVTXConstants(GLcontext * ctx, context_t *context = R700_CONTEXT(ctx); struct radeon_aos * paos = (struct radeon_aos *)pAos; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); unsigned int uSQ_VTX_CONSTANT_WORD0_0; unsigned int uSQ_VTX_CONSTANT_WORD1_0; @@ -205,6 +210,7 @@ void r700SetupStreams(GLcontext *ctx) TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *vb = &tnl->vb; unsigned int i, j = 0; + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); R600_STATECHANGE(context, vtx); @@ -229,6 +235,7 @@ static void r700SendVTXState(GLcontext *ctx, struct radeon_state_atom *atom) = (struct r700_vertex_program *)ctx->VertexProgram._Current; unsigned int i, j = 0; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); if (context->radeon.tcl.aos_count == 0) return; @@ -264,6 +271,7 @@ static void r700SendDepthTargetState(GLcontext *ctx, struct radeon_state_atom *a R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); struct radeon_renderbuffer *rrb; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); rrb = radeon_get_depthbuffer(&context->radeon); if (!rrb || !rrb->bo) { @@ -303,6 +311,7 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom * struct radeon_renderbuffer *rrb; BATCH_LOCALS(&context->radeon); int id = 0; + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); rrb = radeon_get_colorbuffer(&context->radeon); if (!rrb || !rrb->bo) { @@ -352,6 +361,7 @@ static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); struct radeon_bo * pbo; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); @@ -385,6 +395,7 @@ static void r700SendVSState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); struct radeon_bo * pbo; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); @@ -416,6 +427,7 @@ static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); struct radeon_bo * pbo; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); /* XXX fixme * R6xx chips require a FS be emitted, even if it's not used. @@ -457,6 +469,7 @@ static void r700SendViewportState(GLcontext *ctx, struct radeon_state_atom *atom R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); int id = 0; + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); if (id > R700_MAX_VIEWPORTS) return; @@ -489,6 +502,7 @@ static void r700SendSQConfig(GLcontext *ctx, struct radeon_state_atom *atom) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(34); R600_OUT_BATCH_REGSEQ(SQ_CONFIG, 6); @@ -526,6 +540,7 @@ static void r700SendUCPState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); int i; + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); for (i = 0; i < R700_MAX_UCP; i++) { if (r700->ucp[i].enabled) { @@ -547,6 +562,7 @@ static void r700SendSPIState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); unsigned int ui; + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(59 + R700_MAX_SHADER_EXPORTS); @@ -620,6 +636,7 @@ static void r700SendVGTState(GLcontext *ctx, struct radeon_state_atom *atom) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(41); @@ -665,6 +682,7 @@ static void r700SendSXState(GLcontext *ctx, struct radeon_state_atom *atom) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(9); R600_OUT_BATCH_REGVAL(SX_MISC, r700->SX_MISC.u32All); @@ -679,6 +697,7 @@ static void r700SendDBState(GLcontext *ctx, struct radeon_state_atom *atom) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(23); R600_OUT_BATCH_REGVAL(DB_HTILE_DATA_BASE, r700->DB_HTILE_DATA_BASE.u32All); @@ -720,6 +739,7 @@ static void r700SendCBState(GLcontext *ctx, struct radeon_state_atom *atom) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) { BEGIN_BATCH_NO_AUTOSTATE(11); @@ -766,6 +786,7 @@ static void r700SendCBBlendState(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); unsigned int ui; + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) { BEGIN_BATCH_NO_AUTOSTATE(3); @@ -796,6 +817,7 @@ static void r700SendCBBlendColorState(GLcontext *ctx, struct radeon_state_atom * context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(6); R600_OUT_BATCH_REGSEQ(CB_BLEND_RED, 4); @@ -850,6 +872,7 @@ static void r700SendCLState(GLcontext *ctx, struct radeon_state_atom *atom) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(12); R600_OUT_BATCH_REGVAL(PA_CL_CLIP_CNTL, r700->PA_CL_CLIP_CNTL.u32All); @@ -881,6 +904,7 @@ static void r700SendScissorState(GLcontext *ctx, struct radeon_state_atom *atom) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(22); R600_OUT_BATCH_REGSEQ(PA_SC_SCREEN_SCISSOR_TL, 2); @@ -913,6 +937,7 @@ static void r700SendSCState(GLcontext *ctx, struct radeon_state_atom *atom) context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); BEGIN_BATCH_NO_AUTOSTATE(15); R600_OUT_BATCH_REGVAL(R7xx_PA_SC_EDGERULE, r700->PA_SC_EDGERULE.u32All); @@ -969,6 +994,7 @@ static void r700SendVSConsts(GLcontext *ctx, struct radeon_state_atom *atom) R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); int i; BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); if (r700->vs.num_consts == 0) return; @@ -999,6 +1025,7 @@ static int check_cb(GLcontext *ctx, struct radeon_state_atom *atom) if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) count += 11; + radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count); return count; } @@ -1019,6 +1046,7 @@ static int check_blnd(GLcontext *ctx, struct radeon_state_atom *atom) count += 3; } } + radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count); return count; } @@ -1034,6 +1062,7 @@ static int check_ucp(GLcontext *ctx, struct radeon_state_atom *atom) if (r700->ucp[i].enabled) count += 6; } + radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count); return count; } @@ -1045,6 +1074,7 @@ static int check_vtx(GLcontext *ctx, struct radeon_state_atom *atom) if (count) count += 6; + radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count); return count; } @@ -1059,6 +1089,7 @@ static int check_tx(GLcontext *ctx, struct radeon_state_atom *atom) if (t) count++; } + radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count); return count * 31; } @@ -1070,6 +1101,7 @@ static int check_ps_consts(GLcontext *ctx, struct radeon_state_atom *atom) if (count) count += 2; + radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count); return count; } @@ -1082,6 +1114,7 @@ static int check_vs_consts(GLcontext *ctx, struct radeon_state_atom *atom) if (count) count += 2; + radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count); return count; } @@ -1101,6 +1134,7 @@ do { \ void r600InitAtoms(context_t *context) { + radeon_print(RADEON_STATE, RADEON_NORMAL, "%s %p\n", __func__, context); context->radeon.hw.max_state_size = 10 + 5 + 14; /* start 3d, idle, cb/db flush */ /* Setup the atom linked list */ diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c index 6d4ea90ccc..c6546ab00c 100644 --- a/src/mesa/drivers/dri/r600/r700_clear.c +++ b/src/mesa/drivers/dri/r600/r700_clear.c @@ -55,6 +55,8 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) int i; struct gl_framebuffer *fb = ctx->DrawBuffer; + radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s %x\n", __func__, mask); + if( GL_TRUE == r700ClearFast(context, mask) ) { return; @@ -106,8 +108,7 @@ void r700Clear(GLcontext * ctx, GLbitfield mask) } if (swrast_mask) { - if (RADEON_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "%s: swrast clear, mask: %x\n", + radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT, "%s: swrast clear, mask: %x\n", __FUNCTION__, swrast_mask); _swrast_Clear(ctx, swrast_mask); } diff --git a/src/mesa/drivers/dri/r600/r700_debug.c b/src/mesa/drivers/dri/r600/r700_debug.c index dabd0d2c25..cd1ba9eca3 100644 --- a/src/mesa/drivers/dri/r600/r700_debug.c +++ b/src/mesa/drivers/dri/r600/r700_debug.c @@ -25,31 +25,8 @@ * CooperYuan , */ -#include -#include -#include -#include - -#include "main/glheader.h" - #include "r700_debug.h" -#include "r600_context.h" - -void NormalizeLogErrorCode(int nError) -{ - //TODO -} - -void r700_error(int nLocalError, char* fmt, ...) -{ - va_list args; - - NormalizeLogErrorCode(nLocalError); - - va_start(args, fmt); - fprintf(stderr, fmt, args); - va_end(args); -} +#include "radeon_debug.h" void DumpHwBinary(int type, void *addr, int size) { @@ -61,21 +38,21 @@ void DumpHwBinary(int type, void *addr, int size) switch (type) { case DUMP_PIXEL_SHADER: - DEBUGF("Pixel Shader\n"); + radeon_print(RADEON_SHADER, RADEON_TRACE, "Pixel Shader\n"); break; case DUMP_VERTEX_SHADER: - DEBUGF("Vertex Shader\n"); + radeon_print(RADEON_SHADER, RADEON_TRACE, "Vertex Shader\n"); break; case DUMP_FETCH_SHADER: - DEBUGF("Fetch Shader\n"); + radeon_print(RADEON_SHADER, RADEON_TRACE, "Fetch Shader\n"); break; } for (i = 0; i < size; i++) { - DEBUGP("0x%08x,\t", *pHw); + radeon_print(RADEON_SHADER, RADEON_TRACE, "0x%08x,\t", *pHw); if (i%4 == 3) - DEBUGP("0x%08x\n", *pHw); + radeon_print(RADEON_SHADER, RADEON_TRACE, "0x%08x\n", *pHw); pHw++; } diff --git a/src/mesa/drivers/dri/r600/r700_debug.h b/src/mesa/drivers/dri/r600/r700_debug.h index e810e6da08..c0921bf610 100644 --- a/src/mesa/drivers/dri/r600/r700_debug.h +++ b/src/mesa/drivers/dri/r600/r700_debug.h @@ -27,67 +27,13 @@ #ifndef _R700_DEBUG_H_ #define _R700_DEBUG_H_ - -enum R700_ERROR -{ - ERROR_ASM_VTX_CLAUSE = 0x1000, - ERROR_ASM_UNKOWNCLAUSE = 0x1001, - ERROR_ASM_ALLOCEXPORTCF = 0x1002, - ERROR_ASM_ALLOCVTXCF = 0x1003, - ERROR_ASM_ALLOCTEXCF = 0x1004, - ERROR_ASM_ALLOCALUCF = 0x1005, - ERROR_ASM_UNKNOWNILINST = 0x1006, - ERROR_ASM_SRCARGUMENT = 0x1007, - ERROR_ASM_DSTARGUMENT = 0x1008, - ERROR_ASM_TEXINSTRUCTION = 0x1009, - ERROR_ASM_ALUINSTRUCTION = 0x100A, - ERROR_ASM_INSTDSTTRACK = 0x100B, - ERROR_ASM_TEXDSTBADTYPE = 0x100C, - ERROR_ASM_ALUSRCBADTYPE = 0x100D, - ERROR_ASM_ALUSRCSELECT = 0x100E, - ERROR_ASM_ALUSRCNUMBER = 0x100F, - ERROR_ASM_ALUDSTBADTYPE = 0x1010, - ERROR_ASM_CONSTCHANNEL = 0x1011, - ERROR_ASM_BADSCALARBZ = 0x1012, - ERROR_ASM_BADGPRRESERVE = 0x1013, - ERROR_ASM_BADVECTORBZ = 0x1014, - ERROR_ASM_BADTEXINST = 0x1015, - ERROR_ASM_BADTEXSRC = 0x1016, - ERROR_ASM_BADEXPORTTYPE = 0x1017, - - - TODO_ASM_CONSTTEXADDR = 0x8000, - TODO_ASM_NEEDIMPINST = 0x8001, - TODO_ASM_TXB = 0x8002, - TODO_ASM_TXP = 0x8003 -}; - enum R700_DUMP_TYPE { - DUMP_VERTEX_SHADER = 0x1, - DUMP_PIXEL_SHADER = 0x2, - DUMP_FETCH_SHADER = 0x4, + DUMP_VERTEX_SHADER = 0x1, + DUMP_PIXEL_SHADER = 0x2, + DUMP_FETCH_SHADER = 0x4, }; -#define DEBUGF printf -#define DEBUGP printf - -void NormalizeLogErrorCode(int nError); -/*NormalizeLogErrorCode(nLocalError); */ -void r700_error(int nLocalError, char *fmt, ...); extern void DumpHwBinary(int, void *, int); -#ifdef STANDALONE_COMPILER -#ifdef __cplusplus -extern "C" -{ -#endif //__cplusplus - -void LogString(char* szStr); - -#ifdef __cplusplus -} -#endif //__cplusplus -#endif /*STANDALONE_COMPILER*/ - #endif /*_R700_DEBUG_H_*/ diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.c b/src/mesa/drivers/dri/r600/r700_oglprog.c index c49b90c1cc..3c8c1fd7a3 100644 --- a/src/mesa/drivers/dri/r600/r700_oglprog.c +++ b/src/mesa/drivers/dri/r600/r700_oglprog.c @@ -49,6 +49,9 @@ static struct gl_program *r700NewProgram(GLcontext * ctx, struct r700_vertex_program *vp; struct r700_fragment_program *fp; + radeon_print(RADEON_SHADER, RADEON_VERBOSE, + "%s %u, %u\n", __func__, target, id); + switch (target) { case GL_VERTEX_STATE_PROGRAM_NV: @@ -89,6 +92,9 @@ static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) struct r700_vertex_program * vp; struct r700_fragment_program * fp; + radeon_print(RADEON_SHADER, RADEON_VERBOSE, + "%s %p\n", __func__, prog); + switch (prog->Target) { case GL_VERTEX_STATE_PROGRAM_NV: diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index 0b3ceb01ba..4753c757a1 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -327,8 +327,8 @@ static GLuint r700PredictRenderSize(GLcontext* ctx) else dwords += state_size; - if (RADEON_DEBUG & DEBUG_PRIMS) - fprintf(stderr, "%s: total prediction size is %d.\n", __FUNCTION__, dwords); + radeon_print(RADEON_RENDER, RADEON_VERBOSE, + "%s: total prediction size is %d.\n", __FUNCTION__, dwords); return dwords; } @@ -342,8 +342,7 @@ static GLboolean r700RunRender(GLcontext * ctx, struct vertex_buffer *vb = &tnl->vb; struct radeon_renderbuffer *rrb; - if (RADEON_DEBUG & DEBUG_PRIMS) - fprintf(stderr, "%s: cs begin at %d\n", + radeon_print(RADEON_RENDER, RADEON_NORMAL, "%s: cs begin at %d\n", __func__, context->radeon.cmdbuf.cs->cdw); /* always emit CB base to prevent @@ -388,8 +387,7 @@ static GLboolean r700RunRender(GLcontext * ctx, radeonReleaseArrays(ctx, ~0); - if (RADEON_DEBUG & DEBUG_PRIMS) - fprintf(stderr, "%s: cs end at %d\n", + radeon_print(RADEON_RENDER, RADEON_TRACE, "%s: cs end at %d\n", __func__, context->radeon.cmdbuf.cs->cdw); if ( emit_end < context->radeon.cmdbuf.cs->cdw ) diff --git a/src/mesa/drivers/dri/r600/r700_shaderinst.c b/src/mesa/drivers/dri/r600/r700_shaderinst.c index c1bffee91f..f120d9f941 100644 --- a/src/mesa/drivers/dri/r600/r700_shaderinst.c +++ b/src/mesa/drivers/dri/r600/r700_shaderinst.c @@ -27,7 +27,7 @@ #include "main/mtypes.h" -#include "r700_debug.h" +#include "radeon_debug.h" #include "r700_shaderinst.h" void Init_R700ControlFlowGenericClause(R700ControlFlowGenericClause* pInst) @@ -208,16 +208,16 @@ unsigned int GetCFMaxInstructions(ShaderInstType instType) GLboolean LinkVertexInstruction(R700ControlFlowGenericClause *pCFGeneric, R700VertexInstruction *pVTXInstruction) { - if (pCFGeneric->m_pLinkedTEXInstruction != 0) - { - r700_error(ERROR_ASM_VTX_CLAUSE, "This instruction is already linked to a texture instruction"); - return GL_FALSE; + if (pCFGeneric->m_pLinkedTEXInstruction != 0) + { + radeon_error("This instruction is already linked to a texture instruction.\n"); + return GL_FALSE; } pCFGeneric->m_pLinkedVTXInstruction = pVTXInstruction; - pVTXInstruction->m_pLinkedGenericClause = pCFGeneric; + pVTXInstruction->m_pLinkedGenericClause = pCFGeneric; - return GL_TRUE; + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index 86a67ab766..d107f99e7b 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -38,6 +38,7 @@ #include "shader/prog_parameter.h" #include "shader/prog_statevars.h" +#include "radeon_debug.h" #include "r600_context.h" #include "r600_cmdbuf.h" @@ -174,7 +175,7 @@ void Map_Vertex_Program(struct r700_vertex_program *vp, // Create VFETCH instructions for inputs if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions(vp, mesa_vp) ) { - r700_error(ERROR_ASM_VTX_CLAUSE, "Calling Process_Vertex_Program_Vfetch_Instructions return error. \n"); + radeon_error("Calling Process_Vertex_Program_Vfetch_Instructions return error. \n"); return; //error } -- cgit v1.2.3