X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Ftnl%2Ft_vb_program.c;h=f240e98387f8ab6db3a385f6d5a3caac2a3983d3;hb=HEAD;hp=44b64b17d157044cc8e8b5ec5876c2a379074d30;hpb=c036d13d7d2cc905226fe53ebd86a18da808963f;p=mesa.git diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c index 44b64b17d15..f240e98387f 100644 --- a/src/mesa/tnl/t_vb_program.c +++ b/src/mesa/tnl/t_vb_program.c @@ -1,6 +1,5 @@ /* * Mesa 3-D graphics library - * Version: 7.6 * * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * Copyright (C) 2009 VMware, Inc. All Rights Reserved. @@ -18,9 +17,10 @@ * 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 - * BRIAN PAUL 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. + * THE AUTHORS OR COPYRIGHT HOLDERS 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. */ @@ -32,14 +32,16 @@ #include "main/glheader.h" -#include "main/colormac.h" -#include "main/context.h" #include "main/macros.h" -#include "main/imports.h" -#include "shader/prog_instruction.h" -#include "shader/prog_statevars.h" -#include "shader/prog_execute.h" +#include "main/samplerobj.h" +#include "main/state.h" +#include "math/m_xform.h" +#include "program/prog_instruction.h" +#include "program/prog_statevars.h" +#include "program/prog_execute.h" #include "swrast/s_context.h" +#include "util/bitscan.h" +#include "util/u_memory.h" #include "tnl/tnl.h" #include "tnl/t_context.h" @@ -48,10 +50,10 @@ #ifdef NAN_CHECK /** Check for NaNs and very large values */ -static INLINE void +static inline void check_float(float x) { - assert(!IS_INF_OR_NAN(x)); + assert(!util_is_inf_or_nan(x)); assert(1.0e-15 <= x && x <= 1.0e15); } #endif @@ -62,11 +64,15 @@ check_float(float x) */ struct vp_stage_data { /** The results of running the vertex program go into these arrays. */ - GLvector4f results[VERT_RESULT_MAX]; + GLvector4f results[VARYING_SLOT_MAX]; GLvector4f ndcCoords; /**< normalized device coords */ GLubyte *clipmask; /**< clip flags */ GLubyte ormask, andmask; /**< for clipping */ + + GLboolean vertex_textures; + + struct gl_program_machine machine; }; @@ -74,53 +80,51 @@ struct vp_stage_data { static void -userclip( GLcontext *ctx, +userclip( struct gl_context *ctx, GLvector4f *clip, GLubyte *clipmask, GLubyte *clipormask, GLubyte *clipandmask ) { - GLuint p; - - for (p = 0; p < ctx->Const.MaxClipPlanes; p++) { - if (ctx->Transform.ClipPlanesEnabled & (1 << p)) { - GLuint nr, i; - const GLfloat a = ctx->Transform._ClipUserPlane[p][0]; - const GLfloat b = ctx->Transform._ClipUserPlane[p][1]; - const GLfloat c = ctx->Transform._ClipUserPlane[p][2]; - const GLfloat d = ctx->Transform._ClipUserPlane[p][3]; - GLfloat *coord = (GLfloat *)clip->data; - GLuint stride = clip->stride; - GLuint count = clip->count; - - for (nr = 0, i = 0 ; i < count ; i++) { - GLfloat dp = (coord[0] * a + - coord[1] * b + - coord[2] * c + - coord[3] * d); - - if (dp < 0) { - nr++; - clipmask[i] |= CLIP_USER_BIT; - } - - STRIDE_F(coord, stride); - } + GLbitfield mask = ctx->Transform.ClipPlanesEnabled; + while (mask) { + const int p = u_bit_scan(&mask); + GLuint nr, i; + const GLfloat a = ctx->Transform._ClipUserPlane[p][0]; + const GLfloat b = ctx->Transform._ClipUserPlane[p][1]; + const GLfloat c = ctx->Transform._ClipUserPlane[p][2]; + const GLfloat d = ctx->Transform._ClipUserPlane[p][3]; + GLfloat *coord = (GLfloat *)clip->data; + GLuint stride = clip->stride; + GLuint count = clip->count; + + for (nr = 0, i = 0 ; i < count ; i++) { + GLfloat dp = (coord[0] * a + + coord[1] * b + + coord[2] * c + + coord[3] * d); + + if (dp < 0) { + nr++; + clipmask[i] |= CLIP_USER_BIT; + } - if (nr > 0) { - *clipormask |= CLIP_USER_BIT; - if (nr == count) { - *clipandmask |= CLIP_USER_BIT; - return; - } - } + STRIDE_F(coord, stride); + } + + if (nr > 0) { + *clipormask |= CLIP_USER_BIT; + if (nr == count) { + *clipandmask |= CLIP_USER_BIT; + return; + } } } } static GLboolean -do_ndc_cliptest(GLcontext *ctx, struct vp_stage_data *store) +do_ndc_cliptest(struct gl_context *ctx, struct vp_stage_data *store) { TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; @@ -139,7 +143,8 @@ do_ndc_cliptest(GLcontext *ctx, struct vp_stage_data *store) store->clipmask, &store->ormask, &store->andmask, - !ctx->Transform.DepthClamp ); + !(ctx->Transform.DepthClampNear && + ctx->Transform.DepthClampFar) ); } else { VB->NdcPtr = NULL; @@ -148,7 +153,8 @@ do_ndc_cliptest(GLcontext *ctx, struct vp_stage_data *store) store->clipmask, &store->ormask, &store->andmask, - !ctx->Transform.DepthClamp ); + !(ctx->Transform.DepthClampNear && + ctx->Transform.DepthClampFar) ); } if (store->andmask) { @@ -159,8 +165,9 @@ do_ndc_cliptest(GLcontext *ctx, struct vp_stage_data *store) /* Test userclip planes. This contributes to VB->ClipMask. */ /** XXX NEW_SLANG _Enabled ??? */ - if (ctx->Transform.ClipPlanesEnabled && (!ctx->VertexProgram._Enabled || - ctx->VertexProgram.Current->IsPositionInvariant)) { + if (ctx->Transform.ClipPlanesEnabled && + (!_mesa_arb_vertex_program_enabled(ctx) || + ctx->VertexProgram.Current->arb.IsPositionInvariant)) { userclip( ctx, VB->ClipPtr, store->clipmask, @@ -187,13 +194,14 @@ do_ndc_cliptest(GLcontext *ctx, struct vp_stage_data *store) * moved into main/ someday. */ static void -vp_fetch_texel(GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda, +vp_fetch_texel(struct gl_context *ctx, const GLfloat texcoord[4], GLfloat lambda, GLuint unit, GLfloat color[4]) { SWcontext *swrast = SWRAST_CONTEXT(ctx); /* XXX use a float-valued TextureSample routine here!!! */ - swrast->TextureSample[unit](ctx, ctx->Texture.Unit[unit]._Current, + swrast->TextureSample[unit](ctx, _mesa_get_samplerobj(ctx, unit), + ctx->Texture.Unit[unit]._Current, 1, (const GLfloat (*)[4]) texcoord, &lambda, (GLfloat (*)[4]) color); } @@ -204,7 +212,7 @@ vp_fetch_texel(GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda, * string has been parsed. */ GLboolean -_tnl_program_string(GLcontext *ctx, GLenum target, struct gl_program *program) +_tnl_program_string(struct gl_context *ctx, GLenum target, struct gl_program *program) { /* No-op. * If we had derived anything from the program that was private to this @@ -218,42 +226,24 @@ _tnl_program_string(GLcontext *ctx, GLenum target, struct gl_program *program) * Initialize virtual machine state prior to executing vertex program. */ static void -init_machine(GLcontext *ctx, struct gl_program_machine *machine) +init_machine(struct gl_context *ctx, struct gl_program_machine *machine, + GLuint instID) { /* Input registers get initialized from the current vertex attribs */ - MEMCPY(machine->VertAttribs, ctx->Current.Attrib, + memcpy(machine->VertAttribs, ctx->Current.Attrib, MAX_VERTEX_GENERIC_ATTRIBS * 4 * sizeof(GLfloat)); - if (ctx->VertexProgram._Current->IsNVProgram) { - GLuint i; - /* Output/result regs are initialized to [0,0,0,1] */ - for (i = 0; i < MAX_NV_VERTEX_PROGRAM_OUTPUTS; i++) { - ASSIGN_4V(machine->Outputs[i], 0.0F, 0.0F, 0.0F, 1.0F); - } - /* Temp regs are initialized to [0,0,0,0] */ - for (i = 0; i < MAX_NV_VERTEX_PROGRAM_TEMPS; i++) { - ASSIGN_4V(machine->Temporaries[i], 0.0F, 0.0F, 0.0F, 0.0F); - } - for (i = 0; i < MAX_VERTEX_PROGRAM_ADDRESS_REGS; i++) { - ASSIGN_4V(machine->AddressReg[i], 0, 0, 0, 0); - } - } - machine->NumDeriv = 0; - /* init condition codes */ - machine->CondCodes[0] = COND_EQ; - machine->CondCodes[1] = COND_EQ; - machine->CondCodes[2] = COND_EQ; - machine->CondCodes[3] = COND_EQ; - /* init call stack */ machine->StackDepth = 0; machine->FetchTexelLod = vp_fetch_texel; machine->FetchTexelDeriv = NULL; /* not used by vertex programs */ - machine->Samplers = ctx->VertexProgram._Current->Base.SamplerUnits; + machine->Samplers = ctx->VertexProgram._Current->SamplerUnits; + + machine->SystemValues[SYSTEM_VALUE_INSTANCE_ID][0] = (GLfloat) instID; } @@ -261,19 +251,16 @@ init_machine(GLcontext *ctx, struct gl_program_machine *machine) * Map the texture images which the vertex program will access (if any). */ static void -map_textures(GLcontext *ctx, const struct gl_vertex_program *vp) +map_textures(struct gl_context *ctx, const struct gl_program *vp) { GLuint u; - if (!ctx->Driver.MapTexture) - return; - - for (u = 0; u < ctx->Const.MaxVertexTextureImageUnits; u++) { - if (vp->Base.TexturesUsed[u]) { + for (u = 0; u < ctx->Const.Program[MESA_SHADER_VERTEX].MaxTextureImageUnits; u++) { + if (vp->TexturesUsed[u]) { /* Note: _Current *should* correspond to the target indicated * in TexturesUsed[u]. */ - ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[u]._Current); + _swrast_map_texture(ctx, ctx->Texture.Unit[u]._Current); } } } @@ -283,19 +270,16 @@ map_textures(GLcontext *ctx, const struct gl_vertex_program *vp) * Unmap the texture images which were used by the vertex program (if any). */ static void -unmap_textures(GLcontext *ctx, const struct gl_vertex_program *vp) +unmap_textures(struct gl_context *ctx, const struct gl_program *vp) { GLuint u; - if (!ctx->Driver.MapTexture) - return; - - for (u = 0; u < ctx->Const.MaxVertexTextureImageUnits; u++) { - if (vp->Base.TexturesUsed[u]) { + for (u = 0; u < ctx->Const.Program[MESA_SHADER_VERTEX].MaxTextureImageUnits; u++) { + if (vp->TexturesUsed[u]) { /* Note: _Current *should* correspond to the target indicated * in TexturesUsed[u]. */ - ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[u]._Current); + _swrast_unmap_texture(ctx, ctx->Texture.Unit[u]._Current); } } } @@ -305,41 +289,47 @@ unmap_textures(GLcontext *ctx, const struct gl_vertex_program *vp) * This function executes vertex programs */ static GLboolean -run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) +run_vp( struct gl_context *ctx, struct tnl_pipeline_stage *stage ) { TNLcontext *tnl = TNL_CONTEXT(ctx); struct vp_stage_data *store = VP_STAGE_DATA(stage); struct vertex_buffer *VB = &tnl->vb; - struct gl_vertex_program *program = ctx->VertexProgram._Current; - struct gl_program_machine machine; - GLuint outputs[VERT_RESULT_MAX], numOutputs; + struct gl_program *program = ctx->VertexProgram._Current; + struct gl_program_machine *machine = &store->machine; + GLuint outputs[VARYING_SLOT_MAX], numOutputs; GLuint i, j; if (!program) return GL_TRUE; - if (program->IsNVProgram) { - _mesa_load_tracked_matrices(ctx); - } - else { - /* ARB program or vertex shader */ - _mesa_load_state_parameters(ctx, program->Base.Parameters); - } + /* ARB program or vertex shader */ + _mesa_load_state_parameters(ctx, program->Parameters); /* make list of outputs to save some time below */ numOutputs = 0; - for (i = 0; i < VERT_RESULT_MAX; i++) { - if (program->Base.OutputsWritten & BITFIELD64_BIT(i)) { + for (i = 0; i < VARYING_SLOT_MAX; i++) { + if (program->info.outputs_written & BITFIELD64_BIT(i)) { outputs[numOutputs++] = i; } } + /* Allocate result vectors. We delay this until now to avoid allocating + * memory that would never be used if we don't run the software tnl pipeline. + */ + if (!store->results[0].storage) { + for (i = 0; i < VARYING_SLOT_MAX; i++) { + assert(!store->results[i].storage); + _mesa_vector4f_alloc( &store->results[i], 0, VB->Size, 32 ); + store->results[i].size = 4; + } + } + map_textures(ctx, program); for (i = 0; i < VB->Count; i++) { GLuint attr; - init_machine(ctx, &machine); + init_machine(ctx, machine, tnl->CurInstance); #if 0 printf("Input %d: %f, %f, %f, %f\n", i, @@ -361,7 +351,7 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) /* the vertex array case */ for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { - if (program->Base.InputsRead & (1 << attr)) { + if (program->info.inputs_read & BITFIELD64_BIT(attr)) { const GLubyte *ptr = (const GLubyte*) VB->AttribPtr[attr]->data; const GLuint size = VB->AttribPtr[attr]->size; const GLuint stride = VB->AttribPtr[attr]->stride; @@ -372,63 +362,46 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) check_float(data[2]); check_float(data[3]); #endif - COPY_CLEAN_4V(machine.VertAttribs[attr], size, data); + COPY_CLEAN_4V(machine->VertAttribs[attr], size, data); } } /* execute the program */ - _mesa_execute_program(ctx, &program->Base, &machine); + _mesa_execute_program(ctx, program, machine); /* copy the output registers into the VB->attribs arrays */ for (j = 0; j < numOutputs; j++) { const GLuint attr = outputs[j]; #ifdef NAN_CHECK - check_float(machine.Outputs[attr][0]); - check_float(machine.Outputs[attr][1]); - check_float(machine.Outputs[attr][2]); - check_float(machine.Outputs[attr][3]); + check_float(machine->Outputs[attr][0]); + check_float(machine->Outputs[attr][1]); + check_float(machine->Outputs[attr][2]); + check_float(machine->Outputs[attr][3]); #endif - COPY_4V(store->results[attr].data[i], machine.Outputs[attr]); + COPY_4V(store->results[attr].data[i], machine->Outputs[attr]); } /* FOGC is a special case. Fragment shader expects (f,0,0,1) */ - if (program->Base.OutputsWritten & BITFIELD64_BIT(VERT_RESULT_FOGC)) { - store->results[VERT_RESULT_FOGC].data[i][1] = 0.0; - store->results[VERT_RESULT_FOGC].data[i][2] = 0.0; - store->results[VERT_RESULT_FOGC].data[i][3] = 1.0; + if (program->info.outputs_written & BITFIELD64_BIT(VARYING_SLOT_FOGC)) { + store->results[VARYING_SLOT_FOGC].data[i][1] = 0.0; + store->results[VARYING_SLOT_FOGC].data[i][2] = 0.0; + store->results[VARYING_SLOT_FOGC].data[i][3] = 1.0; } #ifdef NAN_CHECK - ASSERT(machine.Outputs[0][3] != 0.0F); + assert(machine->Outputs[0][3] != 0.0F); #endif #if 0 printf("HPOS: %f %f %f %f\n", - machine.Outputs[0][0], - machine.Outputs[0][1], - machine.Outputs[0][2], - machine.Outputs[0][3]); + machine->Outputs[0][0], + machine->Outputs[0][1], + machine->Outputs[0][2], + machine->Outputs[0][3]); #endif } unmap_textures(ctx, program); - /* Fixup fog and point size results if needed */ - if (program->IsNVProgram) { - if (ctx->Fog.Enabled && - (program->Base.OutputsWritten & BITFIELD64_BIT(VERT_RESULT_FOGC)) == 0) { - for (i = 0; i < VB->Count; i++) { - store->results[VERT_RESULT_FOGC].data[i][0] = 1.0; - } - } - - if (ctx->VertexProgram.PointSizeEnabled && - (program->Base.OutputsWritten & BITFIELD64_BIT(VERT_RESULT_PSIZ)) == 0) { - for (i = 0; i < VB->Count; i++) { - store->results[VERT_RESULT_PSIZ].data[i][0] = ctx->Point.Size; - } - } - } - - if (program->IsPositionInvariant) { + if (program->arb.IsPositionInvariant) { /* We need the exact same transform as in the fixed function path here * to guarantee invariance, depending on compiler optimization flags * results could be different otherwise. @@ -456,28 +429,29 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) /* Setup the VB pointers so that the next pipeline stages get * their data from the right place (the program output arrays). */ - VB->ClipPtr = &store->results[VERT_RESULT_HPOS]; + VB->ClipPtr = &store->results[VARYING_SLOT_POS]; VB->ClipPtr->size = 4; VB->ClipPtr->count = VB->Count; } - VB->AttribPtr[VERT_ATTRIB_COLOR0] = &store->results[VERT_RESULT_COL0]; - VB->AttribPtr[VERT_ATTRIB_COLOR1] = &store->results[VERT_RESULT_COL1]; - VB->AttribPtr[VERT_ATTRIB_FOG] = &store->results[VERT_RESULT_FOGC]; - VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->results[VERT_RESULT_PSIZ]; - VB->BackfaceColorPtr = &store->results[VERT_RESULT_BFC0]; - VB->BackfaceSecondaryColorPtr = &store->results[VERT_RESULT_BFC1]; + VB->AttribPtr[VERT_ATTRIB_COLOR0] = &store->results[VARYING_SLOT_COL0]; + VB->AttribPtr[VERT_ATTRIB_COLOR1] = &store->results[VARYING_SLOT_COL1]; + VB->AttribPtr[VERT_ATTRIB_FOG] = &store->results[VARYING_SLOT_FOGC]; + VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->results[VARYING_SLOT_PSIZ]; + VB->BackfaceColorPtr = &store->results[VARYING_SLOT_BFC0]; + VB->BackfaceSecondaryColorPtr = &store->results[VARYING_SLOT_BFC1]; for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { VB->AttribPtr[_TNL_ATTRIB_TEX0 + i] - = &store->results[VERT_RESULT_TEX0 + i]; + = &store->results[VARYING_SLOT_TEX0 + i]; } for (i = 0; i < ctx->Const.MaxVarying; i++) { - if (program->Base.OutputsWritten & BITFIELD64_BIT(VERT_RESULT_VAR0 + i)) { + if (program->info.outputs_written & + BITFIELD64_BIT(VARYING_SLOT_VAR0 + i)) { /* Note: varying results get put into the generic attributes */ VB->AttribPtr[VERT_ATTRIB_GENERIC0+i] - = &store->results[VERT_RESULT_VAR0 + i]; + = &store->results[VARYING_SLOT_VAR0 + i]; } } @@ -493,28 +467,21 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) * allocate data until the first time the stage is run. */ static GLboolean -init_vp(GLcontext *ctx, struct tnl_pipeline_stage *stage) +init_vp(struct gl_context *ctx, struct tnl_pipeline_stage *stage) { TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &(tnl->vb); struct vp_stage_data *store; const GLuint size = VB->Size; - GLuint i; - stage->privatePtr = MALLOC(sizeof(*store)); + stage->privatePtr = calloc(1, sizeof(*store)); store = VP_STAGE_DATA(stage); if (!store) return GL_FALSE; - /* Allocate arrays of vertex output values */ - for (i = 0; i < VERT_RESULT_MAX; i++) { - _mesa_vector4f_alloc( &store->results[i], 0, size, 32 ); - store->results[i].size = 4; - } - /* a few other misc allocations */ _mesa_vector4f_alloc( &store->ndcCoords, 0, size, 32 ); - store->clipmask = (GLubyte *) ALIGN_MALLOC(sizeof(GLubyte)*size, 32 ); + store->clipmask = align_malloc(sizeof(GLubyte)*size, 32 ); return GL_TRUE; } @@ -532,21 +499,21 @@ dtr(struct tnl_pipeline_stage *stage) GLuint i; /* free the vertex program result arrays */ - for (i = 0; i < VERT_RESULT_MAX; i++) + for (i = 0; i < VARYING_SLOT_MAX; i++) _mesa_vector4f_free( &store->results[i] ); /* free misc arrays */ _mesa_vector4f_free( &store->ndcCoords ); - ALIGN_FREE( store->clipmask ); + align_free( store->clipmask ); - FREE( store ); + free( store ); stage->privatePtr = NULL; } } static void -validate_vp_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage) +validate_vp_stage(struct gl_context *ctx, struct tnl_pipeline_stage *stage) { if (ctx->VertexProgram._Current) { _swrast_update_texture_samplers(ctx);