diff options
author | Alex Deucher <alexdeucher@gmail.com> | 2009-06-12 12:09:34 -0400 |
---|---|---|
committer | Alex Deucher <alexdeucher@gmail.com> | 2009-06-12 12:09:34 -0400 |
commit | 1036ef2bf468611d37b5df06fc4424f2002e3837 (patch) | |
tree | f0859a6d903c2570a0a00c918da88139f8f7d065 /src/mesa/shader | |
parent | 917f8bc1a85e61311cef6478127b387df70fba14 (diff) | |
parent | 1cd0afffc9edbcac690f8ab436aecfced26b0aba (diff) |
Merge master and fix conflicts
Diffstat (limited to 'src/mesa/shader')
-rw-r--r-- | src/mesa/shader/arbprogparse.c | 136 | ||||
-rw-r--r-- | src/mesa/shader/arbprogram.c | 44 | ||||
-rw-r--r-- | src/mesa/shader/nvfragparse.c | 277 | ||||
-rw-r--r-- | src/mesa/shader/nvfragparse.h | 4 | ||||
-rw-r--r-- | src/mesa/shader/nvprogram.c | 31 | ||||
-rw-r--r-- | src/mesa/shader/nvvertparse.c | 168 | ||||
-rw-r--r-- | src/mesa/shader/nvvertparse.h | 5 | ||||
-rw-r--r-- | src/mesa/shader/prog_execute.c | 62 | ||||
-rw-r--r-- | src/mesa/shader/prog_instruction.c | 57 | ||||
-rw-r--r-- | src/mesa/shader/prog_instruction.h | 35 | ||||
-rw-r--r-- | src/mesa/shader/prog_optimize.c | 79 | ||||
-rw-r--r-- | src/mesa/shader/prog_optimize.h | 12 | ||||
-rw-r--r-- | src/mesa/shader/prog_parameter.c | 7 | ||||
-rw-r--r-- | src/mesa/shader/prog_print.c | 36 | ||||
-rw-r--r-- | src/mesa/shader/prog_statevars.c | 9 | ||||
-rw-r--r-- | src/mesa/shader/program.c | 1 | ||||
-rw-r--r-- | src/mesa/shader/programopt.c | 121 | ||||
-rw-r--r-- | src/mesa/shader/shader_api.c | 17 | ||||
-rw-r--r-- | src/mesa/shader/slang/slang_compile.c | 6 | ||||
-rw-r--r-- | src/mesa/shader/slang/slang_emit.c | 2 | ||||
-rw-r--r-- | src/mesa/shader/slang/slang_link.c | 25 | ||||
-rw-r--r-- | src/mesa/shader/slang/slang_preprocess.c | 94 |
22 files changed, 609 insertions, 619 deletions
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 35253daa2e..7e166830fd 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -563,6 +563,7 @@ struct var_cache * we take up with our state tokens or constants. Note that * this is _not_ the same as the number of param registers * we eventually use */ + GLuint swizzle; /**< swizzle to access this variable */ struct var_cache *next; }; @@ -581,6 +582,7 @@ var_cache_create (struct var_cache **va) (**va).param_binding_begin = ~0; (**va).param_binding_length = ~0; (**va).alias_binding = NULL; + (**va).swizzle = SWIZZLE_XYZW; (**va).next = NULL; } } @@ -872,15 +874,16 @@ parse_signed_float (const GLubyte ** inst, struct arb_program *Program) * This picks out a constant value from the parsed array. The constant vector is r * returned in the *values array, which should be of length 4. * - * \param values - The 4 component vector with the constant value in it + * \param values - return the vector constant values. + * \param size - returns the number elements in valuesOut [1..4] */ static GLvoid -parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Program, - GLboolean use) +parse_constant(const GLubyte ** inst, GLfloat *values, GLint *size, + struct arb_program *Program, + GLboolean use) { GLuint components, i; - switch (*(*inst)++) { case CONSTANT_SCALAR: if (use == GL_TRUE) { @@ -893,7 +896,7 @@ parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Prog values[1] = values[2] = values[3] = parse_signed_float (inst, Program); } - + *size = 1; break; case CONSTANT_VECTOR: values[0] = values[1] = values[2] = 0; @@ -902,7 +905,12 @@ parse_constant (const GLubyte ** inst, GLfloat *values, struct arb_program *Prog for (i = 0; i < components; i++) { values[i] = parse_signed_float (inst, Program); } + *size = 4; break; + default: + _mesa_problem(NULL, "unexpected case in parse_constant()"); + values[0] = 0.0F; + *size = 0; } } @@ -945,7 +953,7 @@ parse_generic_attrib_num(GLcontext *ctx, const GLubyte ** inst, { GLint i = parse_integer(inst, Program); - if ((i < 0) || (i >= MAX_VERTEX_PROGRAM_ATTRIBS)) + if ((i < 0) || (i >= MAX_VERTEX_GENERIC_ATTRIBS)) { program_error(ctx, Program->Position, "Invalid generic vertex attribute index"); @@ -1014,7 +1022,10 @@ parse_teximage_num (GLcontext * ctx, const GLubyte ** inst, GLint i = parse_integer (inst, Program); if ((i < 0) || (i >= (int)ctx->Const.MaxTextureImageUnits)) { - program_error(ctx, Program->Position, "Invalid texture image index"); + char s[100]; + _mesa_snprintf(s, sizeof(s), "Invalid texture image index %d (%u is max)", + i, ctx->Const.MaxTextureImageUnits); + program_error(ctx, Program->Position, s); return 1; } @@ -1502,10 +1513,10 @@ generic_attrib_check(struct var_cache *vc_head) { int a; struct var_cache *curr; - GLboolean explicitAttrib[MAX_VERTEX_PROGRAM_ATTRIBS], - genericAttrib[MAX_VERTEX_PROGRAM_ATTRIBS]; + GLboolean explicitAttrib[MAX_VERTEX_GENERIC_ATTRIBS], + genericAttrib[MAX_VERTEX_GENERIC_ATTRIBS]; - for (a=0; a<MAX_VERTEX_PROGRAM_ATTRIBS; a++) { + for (a=0; a<MAX_VERTEX_GENERIC_ATTRIBS; a++) { explicitAttrib[a] = GL_FALSE; genericAttrib[a] = GL_FALSE; } @@ -1513,16 +1524,22 @@ generic_attrib_check(struct var_cache *vc_head) curr = vc_head; while (curr) { if (curr->type == vt_attrib) { - if (curr->attrib_is_generic) - genericAttrib[ curr->attrib_binding ] = GL_TRUE; - else + if (curr->attrib_is_generic) { + GLuint attr = (curr->attrib_binding == 0) + ? 0 : (curr->attrib_binding - VERT_ATTRIB_GENERIC0); + assert(attr < MAX_VERTEX_GENERIC_ATTRIBS); + genericAttrib[attr] = GL_TRUE; + } + else { + assert(curr->attrib_binding < MAX_VERTEX_GENERIC_ATTRIBS); explicitAttrib[ curr->attrib_binding ] = GL_TRUE; + } } curr = curr->next; } - for (a=0; a<MAX_VERTEX_PROGRAM_ATTRIBS; a++) { + for (a=0; a<MAX_VERTEX_GENERIC_ATTRIBS; a++) { if ((explicitAttrib[a]) && (genericAttrib[a])) return 1; } @@ -1810,7 +1827,6 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst, GLint idx; GLuint err = 0; gl_state_index state_tokens[STATE_LENGTH] = {0, 0, 0, 0, 0}; - GLfloat const_values[4]; GLubyte token = *(*inst)++; @@ -1902,18 +1918,31 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst, case PARAM_CONSTANT: /* parsing something like {1.0, 2.0, 3.0, 4.0} */ - parse_constant (inst, const_values, Program, use); - idx = _mesa_add_named_constant(Program->Base.Parameters, - (char *) param_var->name, - const_values, 4); - if (param_var->param_binding_begin == ~0U) - param_var->param_binding_begin = idx; - param_var->param_binding_type = PROGRAM_STATE_VAR; - /* Note: when we reference this parameter in an instruction later, - * we'll check if it's really a constant/immediate and set the - * instruction register type appropriately. - */ - param_var->param_binding_length++; + { + GLfloat const_values[4]; + GLint size; + parse_constant(inst, const_values, &size, Program, use); + if (param_var->name[0] == ' ') { + /* this is an unnamed constant */ + idx = _mesa_add_unnamed_constant(Program->Base.Parameters, + const_values, size, + ¶m_var->swizzle); + } + else { + /* named parameter/constant */ + idx = _mesa_add_named_constant(Program->Base.Parameters, + (char *) param_var->name, + const_values, size); + } + if (param_var->param_binding_begin == ~0U) + param_var->param_binding_begin = idx; + param_var->param_binding_type = PROGRAM_STATE_VAR; + /* Note: when we reference this parameter in an instruction later, + * we'll check if it's really a constant/immediate and set the + * instruction register type appropriately. + */ + param_var->param_binding_length++; + } break; default: @@ -2422,6 +2451,9 @@ parse_swizzle_mask(const GLubyte ** inst, GLubyte *swizzle, GLint len) return; } } + + if (len == 1) + swizzle[1] = swizzle[2] = swizzle[3] = swizzle[0]; } @@ -2476,7 +2508,7 @@ static GLuint parse_src_reg (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, struct arb_program *Program, - gl_register_file * File, GLint * Index, + gl_register_file * File, GLint * Index, GLuint *swizzle, GLboolean *IsRelOffset ) { struct var_cache *src; @@ -2485,6 +2517,8 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, *IsRelOffset = 0; + *swizzle = SWIZZLE_XYZW; /* default */ + /* And the binding for the src */ switch (*(*inst)++) { case REGISTER_ATTRIB: @@ -2540,6 +2574,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, } *Index = src->param_binding_begin + offset; + *swizzle = src->swizzle; break; case ARRAY_INDEX_RELATIVE: @@ -2562,6 +2597,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, /* And store it properly */ *Index = src->param_binding_begin + rel_off; *IsRelOffset = 1; + *swizzle = src->swizzle; } break; } @@ -2573,6 +2609,7 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, *File = (gl_register_file) src->param_binding_type; *Index = src->param_binding_begin; + *swizzle = src->swizzle; break; } break; @@ -2641,6 +2678,21 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, } +static GLuint +swizzle_swizzle(GLuint baseSwizzle, const GLubyte swizzle[4]) +{ + GLuint i, swz, s[4]; + for (i = 0; i < 4; i++) { + GLuint c = swizzle[i]; + if (c <= SWIZZLE_W) + s[i] = GET_SWZ(baseSwizzle, c); + else + s[i] = c; + } + swz = MAKE_SWIZZLE4(s[0], s[1], s[2], s[3]); + return swz; +} + /** * Parse vertex/fragment program vector source register. */ @@ -2655,12 +2707,14 @@ parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst, GLubyte negateMask; GLubyte swizzle[4]; GLboolean isRelOffset; + GLuint baseSwizzle; /* Grab the sign */ negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; /* And the src reg */ - if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset)) + if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &baseSwizzle, + &isRelOffset)) return 1; /* finally, the swizzle */ @@ -2668,8 +2722,8 @@ parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst, reg->File = file; reg->Index = index; - reg->Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], swizzle[3]); - reg->NegateBase = negateMask; + reg->Swizzle = swizzle_swizzle(baseSwizzle, swizzle); + reg->Negate = negateMask; reg->RelAddr = isRelOffset; return 0; } @@ -2689,12 +2743,14 @@ parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst, GLubyte negateMask; GLubyte swizzle[4]; GLboolean isRelOffset; + GLuint baseSwizzle; /* Grab the sign */ negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; /* And the src reg */ - if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset)) + if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &baseSwizzle, + &isRelOffset)) return 1; /* finally, the swizzle */ @@ -2702,8 +2758,8 @@ parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst, reg->File = file; reg->Index = index; - reg->Swizzle = (swizzle[0] << 0); - reg->NegateBase = negateMask; + reg->Swizzle = swizzle_swizzle(baseSwizzle, swizzle); + reg->Negate = negateMask; reg->RelAddr = isRelOffset; return 0; } @@ -3013,13 +3069,15 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, GLubyte negateMask; gl_register_file file; GLint index; + GLuint baseSwizzle; - if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &rel)) + if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, + &baseSwizzle, &rel)) return 1; parse_extended_swizzle_mask(inst, swizzle, &negateMask); fp->SrcReg[0].File = file; fp->SrcReg[0].Index = index; - fp->SrcReg[0].NegateBase = negateMask; + fp->SrcReg[0].Negate = negateMask; fp->SrcReg[0].Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], @@ -3354,16 +3412,18 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, GLboolean relAddr; gl_register_file file; GLint index; + GLuint baseSwizzle; if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) return 1; - if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &relAddr)) + if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, + &baseSwizzle, &relAddr)) return 1; parse_extended_swizzle_mask (inst, swizzle, &negateMask); vp->SrcReg[0].File = file; vp->SrcReg[0].Index = index; - vp->SrcReg[0].NegateBase = negateMask; + vp->SrcReg[0].Negate = negateMask; vp->SrcReg[0].Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c index 329c0ea0b0..39136efada 100644 --- a/src/mesa/shader/arbprogram.c +++ b/src/mesa/shader/arbprogram.c @@ -74,8 +74,6 @@ _mesa_BindProgram(GLenum target, GLuint id) GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - /* Error-check target and get curProg */ if ((target == GL_VERTEX_PROGRAM_ARB) && /* == GL_VERTEX_PROGRAM_NV */ (ctx->Extensions.NV_vertex_program || @@ -132,6 +130,9 @@ _mesa_BindProgram(GLenum target, GLuint id) return; } + /* signal new program (and its new constants) */ + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); + /* bind newProg */ if (target == GL_VERTEX_PROGRAM_ARB) { /* == GL_VERTEX_PROGRAM_NV */ _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, @@ -253,6 +254,8 @@ _mesa_EnableVertexAttribArrayARB(GLuint index) return; } + ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib)); + FLUSH_VERTICES(ctx, _NEW_ARRAY); ctx->Array.ArrayObj->VertexAttrib[index].Enabled = GL_TRUE; ctx->Array.ArrayObj->_Enabled |= _NEW_ARRAY_ATTRIB(index); @@ -272,6 +275,8 @@ _mesa_DisableVertexAttribArrayARB(GLuint index) return; } + ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib)); + FLUSH_VERTICES(ctx, _NEW_ARRAY); ctx->Array.ArrayObj->VertexAttrib[index].Enabled = GL_FALSE; ctx->Array.ArrayObj->_Enabled &= ~_NEW_ARRAY_ATTRIB(index); @@ -298,32 +303,41 @@ _mesa_GetVertexAttribdvARB(GLuint index, GLenum pname, GLdouble *params) } +/** + * Return info for a generic vertex attribute array (no alias with + * legacy vertex attributes (pos, normal, color, etc)). + */ void GLAPIENTRY _mesa_GetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat *params) { + const struct gl_client_array *array; GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - if (index >= MAX_VERTEX_PROGRAM_ATTRIBS) { + if (index >= MAX_VERTEX_GENERIC_ATTRIBS) { _mesa_error(ctx, GL_INVALID_VALUE, "glGetVertexAttribfvARB(index)"); return; } + ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib)); + + array = &ctx->Array.ArrayObj->VertexAttrib[index]; + switch (pname) { case GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB: - params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Enabled; + params[0] = (GLfloat) array->Enabled; break; case GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB: - params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Size; + params[0] = (GLfloat) array->Size; break; case GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB: - params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Stride; + params[0] = (GLfloat) array->Stride; break; case GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB: - params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Type; + params[0] = (GLfloat) array->Type; break; case GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB: - params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Normalized; + params[0] = array->Normalized; break; case GL_CURRENT_VERTEX_ATTRIB_ARB: if (index == 0) { @@ -335,7 +349,7 @@ _mesa_GetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat *params) COPY_4V(params, ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + index]); break; case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB: - params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].BufferObj->Name; + params[0] = (GLfloat) array->BufferObj->Name; break; default: _mesa_error(ctx, GL_INVALID_ENUM, "glGetVertexAttribfvARB(pname)"); @@ -379,6 +393,8 @@ _mesa_GetVertexAttribPointervARB(GLuint index, GLenum pname, GLvoid **pointer) return; } + ASSERT(index < Elements(ctx->Array.ArrayObj->VertexAttrib)); + *pointer = (GLvoid *) ctx->Array.ArrayObj->VertexAttrib[index].Ptr; } @@ -489,7 +505,7 @@ _mesa_ProgramEnvParameter4fARB(GLenum target, GLuint index, GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS); if (target == GL_FRAGMENT_PROGRAM_ARB && ctx->Extensions.ARB_fragment_program) { @@ -537,7 +553,7 @@ _mesa_ProgramEnvParameters4fvEXT(GLenum target, GLuint index, GLsizei count, GLfloat * dest; ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS); if (count <= 0) { _mesa_error(ctx, GL_INVALID_VALUE, "glProgramEnvParameters4fv(count)"); @@ -595,8 +611,6 @@ _mesa_GetProgramEnvParameterfvARB(GLenum target, GLuint index, { GET_CURRENT_CONTEXT(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - ASSERT_OUTSIDE_BEGIN_END(ctx); if (target == GL_FRAGMENT_PROGRAM_ARB @@ -633,7 +647,7 @@ _mesa_ProgramLocalParameter4fARB(GLenum target, GLuint index, struct gl_program *prog; ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS); if ((target == GL_FRAGMENT_PROGRAM_NV && ctx->Extensions.NV_fragment_program) || @@ -687,7 +701,7 @@ _mesa_ProgramLocalParameters4fvEXT(GLenum target, GLuint index, GLsizei count, GLint i; ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS); if (count <= 0) { _mesa_error(ctx, GL_INVALID_VALUE, "glProgramLocalParameters4fv(count)"); diff --git a/src/mesa/shader/nvfragparse.c b/src/mesa/shader/nvfragparse.c index b935cb562a..0fd55524ab 100644 --- a/src/mesa/shader/nvfragparse.c +++ b/src/mesa/shader/nvfragparse.c @@ -43,6 +43,7 @@ #include "main/macros.h" #include "program.h" #include "prog_parameter.h" +#include "prog_print.h" #include "prog_instruction.h" #include "nvfragparse.h" @@ -385,10 +386,6 @@ static const char *InputRegisters[MAX_NV_FRAGMENT_PROGRAM_INPUTS + 1] = { }; -static const char *OutputRegisters[MAX_NV_FRAGMENT_PROGRAM_OUTPUTS + 1] = { - "DEPR", "COLR", "DATA0", NULL -}; - /**********************************************************************/ @@ -960,6 +957,7 @@ Parse_VectorSrc(struct parse_state *parseState, GLfloat sign = 1.0F; GLubyte token[100]; GLint idx; + GLuint negateBase, negateAbs; /* * First, take care of +/- and absolute value stuff. @@ -971,21 +969,23 @@ Parse_VectorSrc(struct parse_state *parseState, if (Parse_String(parseState, "|")) { srcReg->Abs = GL_TRUE; - srcReg->NegateAbs = (sign < 0.0F) ? GL_TRUE : GL_FALSE; + negateAbs = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; if (Parse_String(parseState, "-")) - srcReg->NegateBase = NEGATE_XYZW; + negateBase = NEGATE_XYZW; else if (Parse_String(parseState, "+")) - srcReg->NegateBase = NEGATE_NONE; + negateBase = NEGATE_NONE; else - srcReg->NegateBase = NEGATE_NONE; + negateBase = NEGATE_NONE; } else { srcReg->Abs = GL_FALSE; - srcReg->NegateAbs = GL_FALSE; - srcReg->NegateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; + negateAbs = NEGATE_NONE; + negateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; } + srcReg->Negate = srcReg->Abs ? negateAbs : negateBase; + /* This should be the real src vector/register name */ if (!Peek_Token(parseState, token)) RETURN_ERROR; @@ -1086,6 +1086,7 @@ Parse_ScalarSrcReg(struct parse_state *parseState, GLfloat sign = 1.0F; GLboolean needSuffix = GL_TRUE; GLint idx; + GLuint negateBase, negateAbs; /* * First, take care of +/- and absolute value stuff. @@ -1097,21 +1098,23 @@ Parse_ScalarSrcReg(struct parse_state *parseState, if (Parse_String(parseState, "|")) { srcReg->Abs = GL_TRUE; - srcReg->NegateAbs = (sign < 0.0F) ? GL_TRUE : GL_FALSE; + negateAbs = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; if (Parse_String(parseState, "-")) - srcReg->NegateBase = NEGATE_XYZW; + negateBase = NEGATE_XYZW; else if (Parse_String(parseState, "+")) - srcReg->NegateBase = NEGATE_NONE; + negateBase = NEGATE_NONE; else - srcReg->NegateBase = NEGATE_NONE; + negateBase = NEGATE_NONE; } else { srcReg->Abs = GL_FALSE; - srcReg->NegateAbs = GL_FALSE; - srcReg->NegateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; + negateAbs = NEGATE_NONE; + negateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE; } + srcReg->Negate = srcReg->Abs ? negateAbs : negateBase; + if (!Peek_Token(parseState, token)) RETURN_ERROR; @@ -1250,9 +1253,8 @@ Parse_PrintInstruction(struct parse_state *parseState, } inst->SrcReg[0].Swizzle = SWIZZLE_NOOP; - inst->SrcReg[0].NegateBase = NEGATE_NONE; inst->SrcReg[0].Abs = GL_FALSE; - inst->SrcReg[0].NegateAbs = GL_FALSE; + inst->SrcReg[0].Negate = NEGATE_NONE; return GL_TRUE; } @@ -1559,7 +1561,7 @@ _mesa_parse_nv_fragment_program(GLcontext *ctx, GLenum dstTarget, #ifdef DEBUG_foo _mesa_printf("--- glLoadProgramNV(%d) result ---\n", program->Base.Id); - _mesa_print_nv_fragment_program(program); + _mesa_fprint_program_opt(stdout, &program->Base, PROG_PRINT_NV, 0); _mesa_printf("----------------------------------\n"); #endif } @@ -1571,243 +1573,6 @@ _mesa_parse_nv_fragment_program(GLcontext *ctx, GLenum dstTarget, } -static void -PrintSrcReg(const struct gl_fragment_program *program, - const struct prog_src_register *src) -{ - static const char comps[5] = "xyzw"; - - if (src->NegateAbs) { - _mesa_printf("-"); - } - if (src->Abs) { - _mesa_printf("|"); - } - if (src->NegateBase) { - _mesa_printf("-"); - } - if (src->File == PROGRAM_NAMED_PARAM) { - if (program->Base.Parameters->Parameters[src->Index].Type - == PROGRAM_CONSTANT) { - const GLfloat *v; - v = program->Base.Parameters->ParameterValues[src->Index]; - _mesa_printf("{%g, %g, %g, %g}", v[0], v[1], v[2], v[3]); - } - else { - ASSERT(program->Base.Parameters->Parameters[src->Index].Type - == PROGRAM_NAMED_PARAM); - _mesa_printf("%s", program->Base.Parameters->Parameters[src->Index].Name); - } - } - else if (src->File == PROGRAM_OUTPUT) { - _mesa_printf("o[%s]", OutputRegisters[src->Index]); - } - else if (src->File == PROGRAM_INPUT) { - _mesa_printf("f[%s]", InputRegisters[src->Index]); - } - else if (src->File == PROGRAM_LOCAL_PARAM) { - _mesa_printf("p[%d]", src->Index); - } - else if (src->File == PROGRAM_TEMPORARY) { - if (src->Index >= 32) - _mesa_printf("H%d", src->Index); - else - _mesa_printf("R%d", src->Index); - } - else if (src->File == PROGRAM_WRITE_ONLY) { - _mesa_printf("%cC", "HR"[src->Index]); - } - else { - _mesa_problem(NULL, "Invalid fragment register %d", src->Index); - return; - } - if (GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 1) && - GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 2) && - GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 3)) { - _mesa_printf(".%c", comps[GET_SWZ(src->Swizzle, 0)]); - } - else if (src->Swizzle != SWIZZLE_NOOP) { - _mesa_printf(".%c%c%c%c", - comps[GET_SWZ(src->Swizzle, 0)], - comps[GET_SWZ(src->Swizzle, 1)], - comps[GET_SWZ(src->Swizzle, 2)], - comps[GET_SWZ(src->Swizzle, 3)]); - } - if (src->Abs) { - _mesa_printf("|"); - } -} - -static void -PrintTextureSrc(const struct prog_instruction *inst) -{ - _mesa_printf("TEX%d, ", inst->TexSrcUnit); - switch (inst->TexSrcTarget) { - case TEXTURE_1D_INDEX: - _mesa_printf("1D"); - break; - case TEXTURE_2D_INDEX: - _mesa_printf("2D"); - break; - case TEXTURE_3D_INDEX: - _mesa_printf("3D"); - break; - case TEXTURE_RECT_INDEX: - _mesa_printf("RECT"); - break; - case TEXTURE_CUBE_INDEX: - _mesa_printf("CUBE"); - break; - default: - _mesa_problem(NULL, "Invalid textue target in PrintTextureSrc"); - } -} - -static void -PrintCondCode(const struct prog_dst_register *dst) -{ - static const char *comps = "xyzw"; - static const char *ccString[] = { - "??", "GT", "EQ", "LT", "UN", "GE", "LE", "NE", "TR", "FL", "??" - }; - - _mesa_printf("%s", ccString[dst->CondMask]); - if (GET_SWZ(dst->CondSwizzle, 0) == GET_SWZ(dst->CondSwizzle, 1) && - GET_SWZ(dst->CondSwizzle, 0) == GET_SWZ(dst->CondSwizzle, 2) && - GET_SWZ(dst->CondSwizzle, 0) == GET_SWZ(dst->CondSwizzle, 3)) { - _mesa_printf(".%c", comps[GET_SWZ(dst->CondSwizzle, 0)]); - } - else if (dst->CondSwizzle != SWIZZLE_NOOP) { - _mesa_printf(".%c%c%c%c", - comps[GET_SWZ(dst->CondSwizzle, 0)], - comps[GET_SWZ(dst->CondSwizzle, 1)], - comps[GET_SWZ(dst->CondSwizzle, 2)], - comps[GET_SWZ(dst->CondSwizzle, 3)]); - } -} - - -static void -PrintDstReg(const struct prog_dst_register *dst) -{ - if (dst->File == PROGRAM_OUTPUT) { - _mesa_printf("o[%s]", OutputRegisters[dst->Index]); - } - else if (dst->File == PROGRAM_TEMPORARY) { - if (dst->Index >= 32) - _mesa_printf("H%d", dst->Index); - else - _mesa_printf("R%d", dst->Index); - } - else if (dst->File == PROGRAM_LOCAL_PARAM) { - _mesa_printf("p[%d]", dst->Index); - } - else if (dst->File == PROGRAM_WRITE_ONLY) { - _mesa_printf("%cC", "HR"[dst->Index]); - } - else { - _mesa_printf("???"); - } - - if (dst->WriteMask != 0 && dst->WriteMask != WRITEMASK_XYZW) { - _mesa_printf("."); - if (dst->WriteMask & WRITEMASK_X) - _mesa_printf("x"); - if (dst->WriteMask & WRITEMASK_Y) - _mesa_printf("y"); - if (dst->WriteMask & WRITEMASK_Z) - _mesa_printf("z"); - if (dst->WriteMask & WRITEMASK_W) - _mesa_printf("w"); - } - - if (dst->CondMask != COND_TR || - dst->CondSwizzle != SWIZZLE_NOOP) { - _mesa_printf(" ("); - PrintCondCode(dst); - _mesa_printf(")"); - } -} - - -/** - * Print (unparse) the given vertex program. Just for debugging. - */ -void -_mesa_print_nv_fragment_program(const struct gl_fragment_program *program) -{ - const struct prog_instruction *inst; - - for (inst = program->Base.Instructions; inst->Opcode != OPCODE_END; inst++) { - int i; - for (i = 0; Instructions[i].name; i++) { - if (inst->Opcode == Instructions[i].opcode) { - /* print instruction name */ - _mesa_printf("%s", Instructions[i].name); - if (inst->Precision == FLOAT16) - _mesa_printf("H"); - else if (inst->Precision == FIXED12) - _mesa_printf("X"); - if (inst->CondUpdate) - _mesa_printf("C"); - if (inst->SaturateMode == SATURATE_ZERO_ONE) - _mesa_printf("_SAT"); - _mesa_printf(" "); - - if (Instructions[i].inputs == INPUT_CC) { - PrintCondCode(&inst->DstReg); - } - else if (Instructions[i].outputs == OUTPUT_V || - Instructions[i].outputs == OUTPUT_S) { - /* print dest register */ - PrintDstReg(&inst->DstReg); - _mesa_printf(", "); - } - - /* print source register(s) */ - if (Instructions[i].inputs == INPUT_1V || - Instructions[i].inputs == INPUT_1S) { - PrintSrcReg(program, &inst->SrcReg[0]); - } - else if (Instructions[i].inputs == INPUT_2V || - Instructions[i].inputs == INPUT_2S) { - PrintSrcReg(program, &inst->SrcReg[0]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[1]); - } - else if (Instructions[i].inputs == INPUT_3V) { - PrintSrcReg(program, &inst->SrcReg[0]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[1]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[2]); - } - else if (Instructions[i].inputs == INPUT_1V_T) { - PrintSrcReg(program, &inst->SrcReg[0]); - _mesa_printf(", "); - PrintTextureSrc(inst); - } - else if (Instructions[i].inputs == INPUT_3V_T) { - PrintSrcReg(program, &inst->SrcReg[0]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[1]); - _mesa_printf(", "); - PrintSrcReg(program, &inst->SrcReg[2]); - _mesa_printf(", "); - PrintTextureSrc(inst); - } - _mesa_printf(";\n"); - break; - } - } - if (!Instructions[i].name) { - _mesa_printf("Invalid opcode %d\n", inst->Opcode); - } - } - _mesa_printf("END\n"); -} - - const char * _mesa_nv_fragment_input_register_name(GLuint i) { diff --git a/src/mesa/shader/nvfragparse.h b/src/mesa/shader/nvfragparse.h index ac97921080..544ab80c56 100644 --- a/src/mesa/shader/nvfragparse.h +++ b/src/mesa/shader/nvfragparse.h @@ -37,10 +37,6 @@ _mesa_parse_nv_fragment_program(GLcontext *ctx, GLenum target, struct gl_fragment_program *program); -extern void -_mesa_print_nv_fragment_program(const struct gl_fragment_program *program); - - extern const char * _mesa_nv_fragment_input_register_name(GLuint i); diff --git a/src/mesa/shader/nvprogram.c b/src/mesa/shader/nvprogram.c index 50358cf107..d6469b17be 100644 --- a/src/mesa/shader/nvprogram.c +++ b/src/mesa/shader/nvprogram.c @@ -354,6 +354,7 @@ _mesa_GetTrackMatrixivNV(GLenum target, GLuint address, void GLAPIENTRY _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params) { + const struct gl_client_array *array; GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); @@ -362,15 +363,17 @@ _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params) return; } + array = &ctx->Array.ArrayObj->VertexAttrib[index]; + switch (pname) { case GL_ATTRIB_ARRAY_SIZE_NV: - params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Size; + params[0] = array->Size; break; case GL_ATTRIB_ARRAY_STRIDE_NV: - params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Stride; + params[0] = array->Stride; break; case GL_ATTRIB_ARRAY_TYPE_NV: - params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Type; + params[0] = array->Type; break; case GL_CURRENT_ATTRIB_NV: if (index == 0) { @@ -395,6 +398,7 @@ _mesa_GetVertexAttribdvNV(GLuint index, GLenum pname, GLdouble *params) void GLAPIENTRY _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params) { + const struct gl_client_array *array; GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); @@ -403,15 +407,17 @@ _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params) return; } + array = &ctx->Array.ArrayObj->VertexAttrib[index]; + switch (pname) { case GL_ATTRIB_ARRAY_SIZE_NV: - params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Size; + params[0] = (GLfloat) array->Size; break; case GL_ATTRIB_ARRAY_STRIDE_NV: - params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Stride; + params[0] = (GLfloat) array->Stride; break; case GL_ATTRIB_ARRAY_TYPE_NV: - params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].Type; + params[0] = (GLfloat) array->Type; break; case GL_CURRENT_ATTRIB_NV: if (index == 0) { @@ -436,6 +442,7 @@ _mesa_GetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat *params) void GLAPIENTRY _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params) { + const struct gl_client_array *array; GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); @@ -444,15 +451,17 @@ _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params) return; } + array = &ctx->Array.ArrayObj->VertexAttrib[index]; + switch (pname) { case GL_ATTRIB_ARRAY_SIZE_NV: - params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Size; + params[0] = array->Size; break; case GL_ATTRIB_ARRAY_STRIDE_NV: - params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Stride; + params[0] = array->Stride; break; case GL_ATTRIB_ARRAY_TYPE_NV: - params[0] = ctx->Array.ArrayObj->VertexAttrib[index].Type; + params[0] = array->Type; break; case GL_CURRENT_ATTRIB_NV: if (index == 0) { @@ -467,7 +476,7 @@ _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params) params[3] = (GLint) ctx->Current.Attrib[index][3]; break; case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB: - params[0] = ctx->Array.ArrayObj->VertexAttrib[index].BufferObj->Name; + params[0] = array->BufferObj->Name; break; default: _mesa_error(ctx, GL_INVALID_ENUM, "glGetVertexAttribdvNV"); @@ -706,7 +715,7 @@ _mesa_ProgramNamedParameter4fNV(GLuint id, GLsizei len, const GLubyte *name, GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS); prog = _mesa_lookup_program(ctx, id); if (!prog || prog->Target != GL_FRAGMENT_PROGRAM_NV) { diff --git a/src/mesa/shader/nvvertparse.c b/src/mesa/shader/nvvertparse.c index 268b577aec..f5e2df2670 100644 --- a/src/mesa/shader/nvvertparse.c +++ b/src/mesa/shader/nvvertparse.c @@ -44,6 +44,7 @@ #include "nvprogram.h" #include "nvvertparse.h" #include "prog_instruction.h" +#include "prog_print.h" #include "program.h" @@ -640,12 +641,12 @@ Parse_SwizzleSrcReg(struct parse_state *parseState, struct prog_src_register *sr RETURN_ERROR; if (token[0] == '-') { (void) Parse_String(parseState, "-"); - srcReg->NegateBase = NEGATE_XYZW; + srcReg->Negate = NEGATE_XYZW; if (!Peek_Token(parseState, token)) RETURN_ERROR; } else { - srcReg->NegateBase = NEGATE_NONE; + srcReg->Negate = NEGATE_NONE; } /* Src reg can be R<n>, c[n], c[n +/- offset], or a named vertex attrib */ @@ -733,13 +734,13 @@ Parse_ScalarSrcReg(struct parse_state *parseState, struct prog_src_register *src if (!Peek_Token(parseState, token)) RETURN_ERROR; if (token[0] == '-') { - srcReg->NegateBase = NEGATE_XYZW; + srcReg->Negate = NEGATE_XYZW; (void) Parse_String(parseState, "-"); /* consume '-' */ if (!Peek_Token(parseState, token)) RETURN_ERROR; } else { - srcReg->NegateBase = NEGATE_NONE; + srcReg->Negate = NEGATE_NONE; } /* Src reg can be R<n>, c[n], c[n +/- offset], or a named vertex attrib */ @@ -1061,7 +1062,7 @@ Parse_PrintInstruction(struct parse_state *parseState, struct prog_instruction * RETURN_ERROR; srcReg->RelAddr = GL_FALSE; - srcReg->NegateBase = NEGATE_NONE; + srcReg->Negate = NEGATE_NONE; srcReg->Swizzle = SWIZZLE_NOOP; /* Register can be R<n>, c[n], c[n +/- offset], a named vertex attrib, @@ -1394,7 +1395,7 @@ _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum dstTarget, #ifdef DEBUG_foo _mesa_printf("--- glLoadProgramNV result ---\n"); - _mesa_print_nv_vertex_program(program); + _mesa_fprint_program_opt(stdout, &program->Base, PROG_PRINT_NV, 0); _mesa_printf("------------------------------\n"); #endif } @@ -1410,161 +1411,6 @@ _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum dstTarget, } -static void -PrintSrcReg(const struct prog_src_register *src) -{ - static const char comps[5] = "xyzw"; - if (src->NegateBase) - _mesa_printf("-"); - if (src->RelAddr) { - if (src->Index > 0) - _mesa_printf("c[A0.x + %d]", src->Index); - else if (src->Index < 0) - _mesa_printf("c[A0.x - %d]", -src->Index); - else - _mesa_printf("c[A0.x]"); - } - else if (src->File == PROGRAM_OUTPUT) { - _mesa_printf("o[%s]", OutputRegisters[src->Index]); - } - else if (src->File == PROGRAM_INPUT) { - _mesa_printf("v[%s]", InputRegisters[src->Index]); - } - else if (src->File == PROGRAM_ENV_PARAM) { - _mesa_printf("c[%d]", src->Index); - } - else { - ASSERT(src->File == PROGRAM_TEMPORARY); - _mesa_printf("R%d", src->Index); - } - - if (GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 1) && - GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 2) && - GET_SWZ(src->Swizzle, 0) == GET_SWZ(src->Swizzle, 3)) { - _mesa_printf(".%c", comps[GET_SWZ(src->Swizzle, 0)]); - } - else if (src->Swizzle != SWIZZLE_NOOP) { - _mesa_printf(".%c%c%c%c", - comps[GET_SWZ(src->Swizzle, 0)], - comps[GET_SWZ(src->Swizzle, 1)], - comps[GET_SWZ(src->Swizzle, 2)], - comps[GET_SWZ(src->Swizzle, 3)]); - } -} - - -static void -PrintDstReg(const struct prog_dst_register *dst) -{ - if (dst->File == PROGRAM_OUTPUT) { - _mesa_printf("o[%s]", OutputRegisters[dst->Index]); - } - else if (dst->File == PROGRAM_INPUT) { - _mesa_printf("v[%s]", InputRegisters[dst->Index]); - } - else if (dst->File == PROGRAM_ENV_PARAM) { - _mesa_printf("c[%d]", dst->Index); - } - else { - ASSERT(dst->File == PROGRAM_TEMPORARY); - _mesa_printf("R%d", dst->Index); - } - - if (dst->WriteMask != 0 && dst->WriteMask != WRITEMASK_XYZW) { - _mesa_printf("."); - if (dst->WriteMask & WRITEMASK_X) - _mesa_printf("x"); - if (dst->WriteMask & WRITEMASK_Y) - _mesa_printf("y"); - if (dst->WriteMask & WRITEMASK_Z) - _mesa_printf("z"); - if (dst->WriteMask & WRITEMASK_W) - _mesa_printf("w"); - } -} - - -/** - * Print a single NVIDIA vertex program instruction. - */ -void -_mesa_print_nv_vertex_instruction(const struct prog_instruction *inst) -{ - GLuint i, n; - - switch (inst->Opcode) { - case OPCODE_MOV: - case OPCODE_LIT: - case OPCODE_RCP: - case OPCODE_RSQ: - case OPCODE_EXP: - case OPCODE_LOG: - case OPCODE_RCC: - case OPCODE_ABS: - case OPCODE_MUL: - case OPCODE_ADD: - case OPCODE_DP3: - case OPCODE_DP4: - case OPCODE_DST: - case OPCODE_MIN: - case OPCODE_MAX: - case OPCODE_SLT: - case OPCODE_SGE: - case OPCODE_DPH: - case OPCODE_SUB: - case OPCODE_MAD: - _mesa_printf("%s ", _mesa_opcode_string(inst->Opcode)); - PrintDstReg(&inst->DstReg); - _mesa_printf(", "); - n = _mesa_num_inst_src_regs(inst->Opcode); - for (i = 0; i < n; i++) { - PrintSrcReg(&inst->SrcReg[i]); - if (i + 1 < n) - _mesa_printf(", "); - } - _mesa_printf(";\n"); - break; - case OPCODE_ARL: - _mesa_printf("ARL A0.x, "); - PrintSrcReg(&inst->SrcReg[0]); - _mesa_printf(";\n"); - break; - case OPCODE_PRINT: - _mesa_printf("PRINT '%s'", inst->Data); - if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) { - _mesa_printf(", "); - PrintSrcReg(&inst->SrcReg[0]); - _mesa_printf(";\n"); - } - else { - _mesa_printf("\n"); - } - break; - case OPCODE_END: - _mesa_printf("END\n"); - break; - default: - _mesa_printf("BAD INSTRUCTION\n"); - } -} - - -/** - * Print (unparse) the given vertex program. Just for debugging. - */ -void -_mesa_print_nv_vertex_program(const struct gl_vertex_program *program) -{ - const struct prog_instruction *inst; - - for (inst = program->Base.Instructions; ; inst++) { - _mesa_print_nv_vertex_instruction(inst); - if (inst->Opcode == OPCODE_END) - return; - } -} - - const char * _mesa_nv_vertex_input_register_name(GLuint i) { diff --git a/src/mesa/shader/nvvertparse.h b/src/mesa/shader/nvvertparse.h index 15fb03cd4e..9919e22388 100644 --- a/src/mesa/shader/nvvertparse.h +++ b/src/mesa/shader/nvvertparse.h @@ -35,11 +35,6 @@ _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum target, const GLubyte *str, GLsizei len, struct gl_vertex_program *program); -extern void -_mesa_print_nv_vertex_instruction(const struct prog_instruction *inst); - -extern void -_mesa_print_nv_vertex_program(const struct gl_vertex_program *program); extern const char * _mesa_nv_vertex_input_register_name(GLuint i); diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c index bdac1d4f8a..f4beb9a78b 100644 --- a/src/mesa/shader/prog_execute.c +++ b/src/mesa/shader/prog_execute.c @@ -212,24 +212,26 @@ fetch_vector4(const struct prog_src_register *source, result[3] = src[GET_SWZ(source->Swizzle, 3)]; } - if (source->NegateBase) { - result[0] = -result[0]; - result[1] = -result[1]; - result[2] = -result[2]; - result[3] = -result[3]; - } if (source->Abs) { result[0] = FABSF(result[0]); result[1] = FABSF(result[1]); result[2] = FABSF(result[2]); result[3] = FABSF(result[3]); } - if (source->NegateAbs) { + if (source->Negate) { + ASSERT(source->Negate == NEGATE_XYZW); result[0] = -result[0]; result[1] = -result[1]; result[2] = -result[2]; result[3] = -result[3]; } + +#ifdef NAN_CHECK + assert(!IS_INF_OR_NAN(result[0])); + assert(!IS_INF_OR_NAN(result[0])); + assert(!IS_INF_OR_NAN(result[0])); + assert(!IS_INF_OR_NAN(result[0])); +#endif } @@ -259,7 +261,7 @@ fetch_vector4ui(const struct prog_src_register *source, result[3] = src[GET_SWZ(source->Swizzle, 3)]; } - /* Note: no NegateBase, Abs, NegateAbs here */ + /* Note: no Negate or Abs here */ } @@ -299,19 +301,14 @@ fetch_vector4_deriv(GLcontext * ctx, result[2] = deriv[GET_SWZ(source->Swizzle, 2)]; result[3] = deriv[GET_SWZ(source->Swizzle, 3)]; - if (source->NegateBase) { - result[0] = -result[0]; - result[1] = -result[1]; - result[2] = -result[2]; - result[3] = -result[3]; - } if (source->Abs) { result[0] = FABSF(result[0]); result[1] = FABSF(result[1]); result[2] = FABSF(result[2]); result[3] = FABSF(result[3]); } - if (source->NegateAbs) { + if (source->Negate) { + ASSERT(source->Negate == NEGATE_XYZW); result[0] = -result[0]; result[1] = -result[1]; result[2] = -result[2]; @@ -336,13 +333,10 @@ fetch_vector1(const struct prog_src_register *source, result[0] = src[GET_SWZ(source->Swizzle, 0)]; - if (source->NegateBase) { - result[0] = -result[0]; - } if (source->Abs) { result[0] = FABSF(result[0]); } - if (source->NegateAbs) { + if (source->Negate) { result[0] = -result[0]; } } @@ -492,6 +486,13 @@ store_vector4(const struct prog_instruction *inst, } } +#ifdef NAN_CHECK + assert(!IS_INF_OR_NAN(value[0])); + assert(!IS_INF_OR_NAN(value[0])); + assert(!IS_INF_OR_NAN(value[0])); + assert(!IS_INF_OR_NAN(value[0])); +#endif + if (writeMask & WRITEMASK_X) dst[0] = value[0]; if (writeMask & WRITEMASK_Y) @@ -845,10 +846,14 @@ _mesa_execute_program(GLcontext * ctx, break; case OPCODE_EX2: /* Exponential base 2 */ { - GLfloat a[4], result[4]; + GLfloat a[4], result[4], val; fetch_vector1(&inst->SrcReg[0], machine, a); - result[0] = result[1] = result[2] = result[3] = - (GLfloat) _mesa_pow(2.0, a[0]); + val = (GLfloat) _mesa_pow(2.0, a[0]); + /* + if (IS_INF_OR_NAN(val)) + val = 1.0e10; + */ + result[0] = result[1] = result[2] = result[3] = val; store_vector4(inst, machine, result); } break; @@ -924,12 +929,17 @@ _mesa_execute_program(GLcontext * ctx, break; case OPCODE_LG2: /* log base 2 */ { - GLfloat a[4], result[4]; + GLfloat a[4], result[4], val; fetch_vector1(&inst->SrcReg[0], machine, a); /* The fast LOG2 macro doesn't meet the precision requirements. */ - result[0] = result[1] = result[2] = result[3] = - (log(a[0]) * 1.442695F); + if (a[0] == 0.0F) { + val = 0.0F; + } + else { + val = log(a[0]) * 1.442695F; + } + result[0] = result[1] = result[2] = result[3] = val; store_vector4(inst, machine, result); } break; @@ -1514,7 +1524,7 @@ _mesa_execute_program(GLcontext * ctx, ASSERT(swz <= 3); result[i] = src[swz]; } - if (source->NegateBase & (1 << i)) + if (source->Negate & (1 << i)) result[i] = -result[i]; } store_vector4(inst, machine, result); diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c index ca7565c091..44c961927a 100644 --- a/src/mesa/shader/prog_instruction.c +++ b/src/mesa/shader/prog_instruction.c @@ -286,6 +286,56 @@ _mesa_is_tex_instruction(gl_inst_opcode opcode) /** + * Check if there's a potential src/dst register data dependency when + * using SOA execution. + * Example: + * MOV T, T.yxwz; + * This would expand into: + * MOV t0, t1; + * MOV t1, t0; + * MOV t2, t3; + * MOV t3, t2; + * The second instruction will have the wrong value for t0 if executed as-is. + */ +GLboolean +_mesa_check_soa_dependencies(const struct prog_instruction *inst) +{ + GLuint i, chan; + + if (inst->DstReg.WriteMask == WRITEMASK_X || + inst->DstReg.WriteMask == WRITEMASK_Y || + inst->DstReg.WriteMask == WRITEMASK_Z || + inst->DstReg.WriteMask == WRITEMASK_W || + inst->DstReg.WriteMask == 0x0) { + /* no chance of data dependency */ + return GL_FALSE; + } + + /* loop over src regs */ + for (i = 0; i < 3; i++) { + if (inst->SrcReg[i].File == inst->DstReg.File && + inst->SrcReg[i].Index == inst->DstReg.Index) { + /* loop over dest channels */ + GLuint channelsWritten = 0x0; + for (chan = 0; chan < 4; chan++) { + if (inst->DstReg.WriteMask & (1 << chan)) { + /* check if we're reading a channel that's been written */ + GLuint swizzle = GET_SWZ(inst->SrcReg[i].Swizzle, chan); + if (swizzle <= SWIZZLE_W && + (channelsWritten & (1 << swizzle))) { + return GL_TRUE; + } + + channelsWritten |= (1 << chan); + } + } + } + } + return GL_FALSE; +} + + +/** * Return string name for given program opcode. */ const char * @@ -293,7 +343,10 @@ _mesa_opcode_string(gl_inst_opcode opcode) { if (opcode < MAX_OPCODE) return InstInfo[opcode].Name; - else - return "OP?"; + else { + static char s[20]; + _mesa_snprintf(s, sizeof(s), "OP%u", opcode); + return s; + } } diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h index 4adce11f95..40ad998f79 100644 --- a/src/mesa/shader/prog_instruction.h +++ b/src/mesa/shader/prog_instruction.h @@ -261,37 +261,15 @@ struct prog_src_register GLuint Swizzle:12; GLuint RelAddr:1; - /** - * \name Source register "sign" control. - * - * The ARB and NV extensions allow varrying degrees of control over the - * sign of the source vector components. These values allow enough control - * for all flavors of the extensions. - */ - /*@{*/ - /** - * Per-component negation for the SWZ instruction. For non-SWZ - * instructions the only possible values are NEGATE_XYZW and NEGATE_NONE. - * - * \since - * ARB_vertex_program, ARB_fragment_program - */ - GLuint NegateBase:4; - - /** - * Take the component-wise absolute value. - * - * \since - * NV_fragment_program, NV_fragment_program_option, NV_vertex_program2, - * NV_vertex_program2_option. - */ + /** Take the component-wise absolute value */ GLuint Abs:1; /** - * Post-absolute value negation (all components). + * Post-Abs negation. + * This will either be NEGATE_NONE or NEGATE_XYZW, except for the SWZ + * instruction which allows per-component negation. */ - GLuint NegateAbs:1; - /*@}*/ + GLuint Negate:4; }; @@ -450,6 +428,9 @@ _mesa_num_inst_dst_regs(gl_inst_opcode opcode); extern GLboolean _mesa_is_tex_instruction(gl_inst_opcode opcode); +extern GLboolean +_mesa_check_soa_dependencies(const struct prog_instruction *inst); + extern const char * _mesa_opcode_string(gl_inst_opcode opcode); diff --git a/src/mesa/shader/prog_optimize.c b/src/mesa/shader/prog_optimize.c index 5f35dbf128..be903106a0 100644 --- a/src/mesa/shader/prog_optimize.c +++ b/src/mesa/shader/prog_optimize.c @@ -547,15 +547,13 @@ update_interval(GLint intBegin[], GLint intEnd[], GLuint index, GLuint ic) /** - * Find the live intervals for each temporary register in the program. - * For register R, the interval [A,B] indicates that R is referenced - * from instruction A through instruction B. - * Special consideration is needed for loops and subroutines. - * \return GL_TRUE if success, GL_FALSE if we cannot proceed for some reason + * Find first/last instruction that references each temporary register. */ -static GLboolean -find_live_intervals(struct gl_program *prog, - struct interval_list *liveIntervals) +GLboolean +_mesa_find_temp_intervals(const struct prog_instruction *instructions, + GLuint numInstructions, + GLint intBegin[MAX_PROGRAM_TEMPS], + GLint intEnd[MAX_PROGRAM_TEMPS]) { struct loop_info { @@ -563,26 +561,15 @@ find_live_intervals(struct gl_program *prog, }; struct loop_info loopStack[MAX_LOOP_NESTING]; GLuint loopStackDepth = 0; - GLint intBegin[MAX_PROGRAM_TEMPS], intEnd[MAX_PROGRAM_TEMPS]; GLuint i; - /* - * Note: we'll return GL_FALSE below if we find relative indexing - * into the TEMP register file. We can't handle that yet. - * We also give up on subroutines for now. - */ - - if (dbg) { - _mesa_printf("Optimize: Begin find intervals\n"); - } - for (i = 0; i < MAX_PROGRAM_TEMPS; i++){ intBegin[i] = intEnd[i] = -1; } /* Scan instructions looking for temporary registers */ - for (i = 0; i < prog->NumInstructions; i++) { - const struct prog_instruction *inst = prog->Instructions + i; + for (i = 0; i < numInstructions; i++) { + const struct prog_instruction *inst = instructions + i; if (inst->Opcode == OPCODE_BGNLOOP) { loopStack[loopStackDepth].Start = i; loopStack[loopStackDepth].End = inst->BranchTarget; @@ -595,7 +582,7 @@ find_live_intervals(struct gl_program *prog, return GL_FALSE; } else { - const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode); + const GLuint numSrc = 3;/*_mesa_num_inst_src_regs(inst->Opcode);*/ GLuint j; for (j = 0; j < numSrc; j++) { if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) { @@ -624,6 +611,39 @@ find_live_intervals(struct gl_program *prog, } } + return GL_TRUE; +} + + +/** + * Find the live intervals for each temporary register in the program. + * For register R, the interval [A,B] indicates that R is referenced + * from instruction A through instruction B. + * Special consideration is needed for loops and subroutines. + * \return GL_TRUE if success, GL_FALSE if we cannot proceed for some reason + */ +static GLboolean +find_live_intervals(struct gl_program *prog, + struct interval_list *liveIntervals) +{ + GLint intBegin[MAX_PROGRAM_TEMPS], intEnd[MAX_PROGRAM_TEMPS]; + GLuint i; + + /* + * Note: we'll return GL_FALSE below if we find relative indexing + * into the TEMP register file. We can't handle that yet. + * We also give up on subroutines for now. + */ + + if (dbg) { + _mesa_printf("Optimize: Begin find intervals\n"); + } + + /* build intermediate arrays */ + if (!_mesa_find_temp_intervals(prog->Instructions, prog->NumInstructions, + intBegin, intEnd)) + return GL_FALSE; + /* Build live intervals list from intermediate arrays */ liveIntervals->Num = 0; for (i = 0; i < MAX_PROGRAM_TEMPS; i++) { @@ -660,7 +680,8 @@ find_live_intervals(struct gl_program *prog, } -static GLuint +/** Scan the array of used register flags to find free entry */ +static GLint alloc_register(GLboolean usedRegs[MAX_PROGRAM_TEMPS]) { GLuint k; @@ -670,7 +691,7 @@ alloc_register(GLboolean usedRegs[MAX_PROGRAM_TEMPS]) return k; } } - return MAX_PROGRAM_TEMPS; + return -1; } @@ -689,7 +710,7 @@ _mesa_reallocate_registers(struct gl_program *prog) GLint registerMap[MAX_PROGRAM_TEMPS]; GLboolean usedRegs[MAX_PROGRAM_TEMPS]; GLuint i; - GLuint maxTemp = 0; + GLint maxTemp = -1; if (dbg) { _mesa_printf("Optimize: Begin live-interval register reallocation\n"); @@ -754,15 +775,15 @@ _mesa_reallocate_registers(struct gl_program *prog) /* find a free register for this live interval */ { - const GLuint k = alloc_register(usedRegs); - if (k == MAX_PROGRAM_TEMPS) { + const GLint k = alloc_register(usedRegs); + if (k < 0) { /* out of registers, give up */ return; } registerMap[live->Reg] = k; maxTemp = MAX2(maxTemp, k); if (dbg) - _mesa_printf(" remap register %d -> %d\n", live->Reg, k); + _mesa_printf(" remap register %u -> %d\n", live->Reg, k); } /* Insert this live interval into the active list which is sorted @@ -791,8 +812,6 @@ _mesa_reallocate_registers(struct gl_program *prog) } - - /** * Apply optimizations to the given program to eliminate unnecessary * instructions, temp regs, etc. diff --git a/src/mesa/shader/prog_optimize.h b/src/mesa/shader/prog_optimize.h index d102cfd9fc..43894a2723 100644 --- a/src/mesa/shader/prog_optimize.h +++ b/src/mesa/shader/prog_optimize.h @@ -25,7 +25,19 @@ #ifndef PROG_OPT_H #define PROG_OPT_H + +#include "main/config.h" + + struct gl_program; +struct prog_instruction; + + +extern GLboolean +_mesa_find_temp_intervals(const struct prog_instruction *instructions, + GLuint numInstructions, + GLint intBegin[MAX_PROGRAM_TEMPS], + GLint intEnd[MAX_PROGRAM_TEMPS]); extern void _mesa_optimize_program(GLcontext *ctx, struct gl_program *program); diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index e9ed3985ee..bcd8c5d9dc 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -327,15 +327,16 @@ _mesa_add_sampler(struct gl_program_parameter_list *paramList, else { GLuint i; const GLint size = 1; /* a sampler is basically a texture unit number */ - GLfloat value; + GLfloat value[4]; GLint numSamplers = 0; for (i = 0; i < paramList->NumParameters; i++) { if (paramList->Parameters[i].Type == PROGRAM_SAMPLER) numSamplers++; } - value = (GLfloat) numSamplers; + value[0] = (GLfloat) numSamplers; + value[1] = value[2] = value[3] = 0.0F; (void) _mesa_add_parameter(paramList, PROGRAM_SAMPLER, name, - size, datatype, &value, NULL, 0x0); + size, datatype, value, NULL, 0x0); return numSamplers; } } diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index b832ddb477..de7fef1f86 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -75,7 +75,11 @@ file_string(gl_register_file f, gl_prog_print_mode mode) case PROGRAM_UNDEFINED: return "UNDEFINED"; default: - return "Unknown program file!"; + { + static char s[20]; + _mesa_snprintf(s, sizeof(s), "FILE%u", f); + return s; + } } } @@ -325,19 +329,19 @@ reg_string(gl_register_file f, GLint index, gl_prog_print_mode mode, * \param extended if true, also allow 0, 1 values */ const char * -_mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) +_mesa_swizzle_string(GLuint swizzle, GLuint negateMask, GLboolean extended) { static const char swz[] = "xyzw01!?"; /* See SWIZZLE_x definitions */ static char s[20]; GLuint i = 0; - if (!extended && swizzle == SWIZZLE_NOOP && negateBase == 0) + if (!extended && swizzle == SWIZZLE_NOOP && negateMask == 0) return ""; /* no swizzle/negation */ if (!extended) s[i++] = '.'; - if (negateBase & NEGATE_X) + if (negateMask & NEGATE_X) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 0)]; @@ -345,7 +349,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & NEGATE_Y) + if (negateMask & NEGATE_Y) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 1)]; @@ -353,7 +357,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & NEGATE_Z) + if (negateMask & NEGATE_Z) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 2)]; @@ -361,7 +365,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & NEGATE_W) + if (negateMask & NEGATE_W) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 3)]; @@ -465,14 +469,14 @@ fprint_src_reg(FILE *f, reg_string((gl_register_file) srcReg->File, srcReg->Index, mode, srcReg->RelAddr, prog), _mesa_swizzle_string(srcReg->Swizzle, - srcReg->NegateBase, GL_FALSE), + srcReg->Negate, GL_FALSE), abs); #if 0 _mesa_fprintf(f, "%s[%d]%s", file_string((gl_register_file) srcReg->File, mode), srcReg->Index, _mesa_swizzle_string(srcReg->Swizzle, - srcReg->NegateBase, GL_FALSE)); + srcReg->Negate, GL_FALSE)); #endif } @@ -566,7 +570,7 @@ _mesa_fprint_instruction_opt(FILE *f, mode), inst->SrcReg[0].Index, _mesa_swizzle_string(inst->SrcReg[0].Swizzle, - inst->SrcReg[0].NegateBase, GL_FALSE)); + inst->SrcReg[0].Negate, GL_FALSE)); } if (inst->Comment) _mesa_fprintf(f, " # %s", inst->Comment); @@ -583,7 +587,7 @@ _mesa_fprint_instruction_opt(FILE *f, mode), inst->SrcReg[0].Index, _mesa_swizzle_string(inst->SrcReg[0].Swizzle, - inst->SrcReg[0].NegateBase, GL_TRUE)); + inst->SrcReg[0].Negate, GL_TRUE)); fprint_comment(f, inst); break; case OPCODE_TEX: @@ -736,7 +740,10 @@ _mesa_fprint_instruction_opt(FILE *f, mode, prog); } else { - _mesa_fprintf(f, "Other opcode %d\n", inst->Opcode); + fprint_alu_instruction(f, inst, + _mesa_opcode_string(inst->Opcode), + 3/*_mesa_num_inst_src_regs(inst->Opcode)*/, + mode, prog); } break; } @@ -873,6 +880,7 @@ _mesa_fprint_parameter_list(FILE *f, return; _mesa_fprintf(f, "param list %p\n", (void *) list); + _mesa_fprintf(f, "dirty state flags: 0x%x\n", list->StateFlags); for (i = 0; i < list->NumParameters; i++){ struct gl_program_parameter *param = list->Parameters + i; const GLfloat *v = list->ParameterValues[i]; @@ -940,6 +948,10 @@ _mesa_write_shader_to_file(const struct gl_shader *shader) fprintf(f, "/*\n"); _mesa_fprint_program_opt(f, shader->Program, PROG_PRINT_DEBUG, GL_TRUE); fprintf(f, "*/\n"); + fprintf(f, "/* Parameters / constants */\n"); + fprintf(f, "/*\n"); + _mesa_fprint_parameter_list(f, shader->Program->Parameters); + fprintf(f, "*/\n"); } fclose(f); diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 37a3f1fc8c..058d4bbafb 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -816,7 +816,10 @@ append_token(char *dst, gl_state_index k) break; /* BEGIN internal state vars */ case STATE_INTERNAL: - append(dst, "(internal)"); + append(dst, ".internal."); + break; + case STATE_CURRENT_ATTRIB: + append(dst, "current"); break; case STATE_NORMAL_SCALE: append(dst, "normalScale"); @@ -986,7 +989,9 @@ _mesa_program_state_string(const gl_state_index state[STATE_LENGTH]) break; case STATE_INTERNAL: append_token(str, state[1]); - break; + if (state[1] == STATE_CURRENT_ATTRIB) + append_index(str, state[2]); + break; default: _mesa_problem(NULL, "Invalid state in _mesa_program_state_string"); break; diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 2e5632710e..d270bf9e1c 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -296,6 +296,7 @@ _mesa_new_program(GLcontext *ctx, GLenum target, GLuint id) struct gl_program *prog; switch (target) { case GL_VERTEX_PROGRAM_ARB: /* == GL_VERTEX_PROGRAM_NV */ + case GL_VERTEX_STATE_PROGRAM_NV: prog = _mesa_init_vertex_program(ctx, CALLOC_STRUCT(gl_vertex_program), target, id ); break; diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index e283f8933b..f70c75cec8 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -45,8 +45,8 @@ * into a vertex program. * May be used to implement the position_invariant option. */ -void -_mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog) +static void +_mesa_insert_mvp_dp4_code(GLcontext *ctx, struct gl_vertex_program *vprog) { struct prog_instruction *newInst; const GLuint origLen = vprog->Base.NumInstructions; @@ -113,6 +113,121 @@ _mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog) } +static void +_mesa_insert_mvp_mad_code(GLcontext *ctx, struct gl_vertex_program *vprog) +{ + struct prog_instruction *newInst; + const GLuint origLen = vprog->Base.NumInstructions; + const GLuint newLen = origLen + 4; + GLuint hposTemp; + GLuint i; + + /* + * Setup state references for the modelview/projection matrix. + * XXX we should check if these state vars are already declared. + */ + static const gl_state_index mvpState[4][STATE_LENGTH] = { + { STATE_MVP_MATRIX, 0, 0, 0, STATE_MATRIX_TRANSPOSE }, + { STATE_MVP_MATRIX, 0, 1, 1, STATE_MATRIX_TRANSPOSE }, + { STATE_MVP_MATRIX, 0, 2, 2, STATE_MATRIX_TRANSPOSE }, + { STATE_MVP_MATRIX, 0, 3, 3, STATE_MATRIX_TRANSPOSE }, + }; + GLint mvpRef[4]; + + for (i = 0; i < 4; i++) { + mvpRef[i] = _mesa_add_state_reference(vprog->Base.Parameters, + mvpState[i]); + } + + /* Alloc storage for new instructions */ + newInst = _mesa_alloc_instructions(newLen); + if (!newInst) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, + "glProgramString(inserting position_invariant code)"); + return; + } + + /* TEMP hposTemp; */ + hposTemp = vprog->Base.NumTemporaries++; + + /* + * Generated instructions: + * emit_op2(p, OPCODE_MUL, tmp, 0, swizzle1(src,X), mat[0]); + * emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Y), mat[1], tmp); + * emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Z), mat[2], tmp); + * emit_op3(p, OPCODE_MAD, dest, 0, swizzle1(src,W), mat[3], tmp); + */ + _mesa_init_instructions(newInst, 4); + + newInst[0].Opcode = OPCODE_MUL; + newInst[0].DstReg.File = PROGRAM_TEMPORARY; + newInst[0].DstReg.Index = hposTemp; + newInst[0].DstReg.WriteMask = WRITEMASK_XYZW; + newInst[0].SrcReg[0].File = PROGRAM_INPUT; + newInst[0].SrcReg[0].Index = VERT_ATTRIB_POS; + newInst[0].SrcReg[0].Swizzle = SWIZZLE_XXXX; + newInst[0].SrcReg[1].File = PROGRAM_STATE_VAR; + newInst[0].SrcReg[1].Index = mvpRef[0]; + newInst[0].SrcReg[1].Swizzle = SWIZZLE_NOOP; + + for (i = 1; i <= 2; i++) { + newInst[i].Opcode = OPCODE_MAD; + newInst[i].DstReg.File = PROGRAM_TEMPORARY; + newInst[i].DstReg.Index = hposTemp; + newInst[i].DstReg.WriteMask = WRITEMASK_XYZW; + newInst[i].SrcReg[0].File = PROGRAM_INPUT; + newInst[i].SrcReg[0].Index = VERT_ATTRIB_POS; + newInst[i].SrcReg[0].Swizzle = MAKE_SWIZZLE4(i,i,i,i); + newInst[i].SrcReg[1].File = PROGRAM_STATE_VAR; + newInst[i].SrcReg[1].Index = mvpRef[i]; + newInst[i].SrcReg[1].Swizzle = SWIZZLE_NOOP; + newInst[i].SrcReg[2].File = PROGRAM_TEMPORARY; + newInst[i].SrcReg[2].Index = hposTemp; + newInst[1].SrcReg[2].Swizzle = SWIZZLE_NOOP; + } + + newInst[3].Opcode = OPCODE_MAD; + newInst[3].DstReg.File = PROGRAM_OUTPUT; + newInst[3].DstReg.Index = VERT_RESULT_HPOS; + newInst[3].DstReg.WriteMask = WRITEMASK_XYZW; + newInst[3].SrcReg[0].File = PROGRAM_INPUT; + newInst[3].SrcReg[0].Index = VERT_ATTRIB_POS; + newInst[3].SrcReg[0].Swizzle = SWIZZLE_WWWW; + newInst[3].SrcReg[1].File = PROGRAM_STATE_VAR; + newInst[3].SrcReg[1].Index = mvpRef[3]; + newInst[3].SrcReg[1].Swizzle = SWIZZLE_NOOP; + newInst[3].SrcReg[2].File = PROGRAM_TEMPORARY; + newInst[3].SrcReg[2].Index = hposTemp; + newInst[3].SrcReg[2].Swizzle = SWIZZLE_NOOP; + + + /* Append original instructions after new instructions */ + _mesa_copy_instructions (newInst + 4, vprog->Base.Instructions, origLen); + + /* free old instructions */ + _mesa_free_instructions(vprog->Base.Instructions, origLen); + + /* install new instructions */ + vprog->Base.Instructions = newInst; + vprog->Base.NumInstructions = newLen; + vprog->Base.InputsRead |= VERT_BIT_POS; + vprog->Base.OutputsWritten |= (1 << VERT_RESULT_HPOS); +} + + +void +_mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog) +{ + if (ctx->mvp_with_dp4) + _mesa_insert_mvp_dp4_code( ctx, vprog ); + else + _mesa_insert_mvp_mad_code( ctx, vprog ); +} + + + + + /** * Append extra instructions onto the given fragment program to implement @@ -241,7 +356,7 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog) inst->DstReg.WriteMask = WRITEMASK_X; inst->SrcReg[0].File = PROGRAM_TEMPORARY; inst->SrcReg[0].Index = fogFactorTemp; - inst->SrcReg[0].NegateBase = NEGATE_XYZW; + inst->SrcReg[0].Negate = NEGATE_XYZW; inst->SrcReg[0].Swizzle = SWIZZLE_XXXX; inst->SaturateMode = SATURATE_ZERO_ONE; inst++; diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 61289db2d2..a8390d3094 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -51,11 +51,6 @@ #include "glapi/dispatch.h" -#ifndef GL_PROGRAM_BINARY_LENGTH_OES -#define GL_PROGRAM_BINARY_LENGTH_OES 0x8741 -#endif - - /** * Allocate a new gl_shader_program object, initialize it. */ @@ -1492,7 +1487,7 @@ _mesa_use_program(GLcontext *ctx, GLuint program) return; } - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); if (program) { shProg = _mesa_lookup_shader_program_err(ctx, program, "glUseProgram"); @@ -1514,6 +1509,10 @@ _mesa_use_program(GLcontext *ctx, GLuint program) shProg->Shaders[i]->Name, shProg->Shaders[i]->Type); } + if (shProg->VertexProgram) + printf(" vert prog %u\n", shProg->VertexProgram->Base.Id); + if (shProg->FragmentProgram) + printf(" frag prog %u\n", shProg->FragmentProgram->Base.Id); } } else { @@ -1794,7 +1793,7 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, return; } - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS); uniform = &shProg->Uniforms->Uniforms[location]; @@ -1902,7 +1901,7 @@ set_program_uniform_matrix(GLcontext *ctx, struct gl_program *program, */ static void _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, - GLenum matrixType, GLint location, GLsizei count, + GLint location, GLsizei count, GLboolean transpose, const GLfloat *values) { struct gl_shader_program *shProg = ctx->Shader.CurrentProgram; @@ -1934,7 +1933,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, return; } - FLUSH_VERTICES(ctx, _NEW_PROGRAM); + FLUSH_VERTICES(ctx, _NEW_PROGRAM_CONSTANTS); uniform = &shProg->Uniforms->Uniforms[location]; diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index ba2fc4f85c..d7ad879e97 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2161,6 +2161,12 @@ parse_function(slang_parse_ctx * C, slang_output_ctx * O, int definition, (O->funs->num_functions + 1) * sizeof(slang_function)); if (O->funs->functions == NULL) { + /* Make sure that there are no functions marked, as the + * allocation is currently NULL, in order to avoid + * a potental segfault as we clean up later. + */ + O->funs->num_functions = 0; + slang_info_log_memory(C->L); slang_function_destruct(&parsed_func); return GL_FALSE; diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 8493c490fb..3f455e0640 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1135,7 +1135,7 @@ emit_negation(slang_emit_info *emitInfo, slang_ir_node *n) n->Children[0]->Store, NULL, NULL); - inst->SrcReg[0].NegateBase = NEGATE_XYZW; + inst->SrcReg[0].Negate = NEGATE_XYZW; return inst; } diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 1fdf4db054..5ea89d2ff3 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -97,7 +97,8 @@ bits_agree(GLbitfield flags1, GLbitfield flags2, GLbitfield bit) * which inputs are centroid-sampled, invariant, etc. */ static GLboolean -link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) +link_varying_vars(GLcontext *ctx, + struct gl_shader_program *shProg, struct gl_program *prog) { GLuint *map, i, firstVarying, newFile; GLbitfield *inOutFlags; @@ -156,8 +157,12 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) var->Flags); } + if (shProg->Varying->NumParameters > ctx->Const.MaxVarying) { + link_error(shProg, "Too many varying variables"); + return GL_FALSE; + } + /* Map varying[i] to varying[j]. - * Plus, set prog->Input/OutputFlags[] as described above. * Note: the loop here takes care of arrays or large (sz>4) vars. */ { @@ -319,7 +324,7 @@ _slang_resolve_attributes(struct gl_shader_program *shProg, const struct gl_program *origProg, struct gl_program *linkedProg) { - GLint attribMap[MAX_VERTEX_ATTRIBS]; + GLint attribMap[MAX_VERTEX_GENERIC_ATTRIBS]; GLuint i, j; GLbitfield usedAttributes; /* generics only, not legacy attributes */ @@ -355,7 +360,7 @@ _slang_resolve_attributes(struct gl_shader_program *shProg, } /* initialize the generic attribute map entries to -1 */ - for (i = 0; i < MAX_VERTEX_ATTRIBS; i++) { + for (i = 0; i < MAX_VERTEX_GENERIC_ATTRIBS; i++) { attribMap[i] = -1; } @@ -396,11 +401,11 @@ _slang_resolve_attributes(struct gl_shader_program *shProg, * Start at 1 since generic attribute 0 always aliases * glVertex/position. */ - for (attr = 0; attr < MAX_VERTEX_ATTRIBS; attr++) { + for (attr = 0; attr < MAX_VERTEX_GENERIC_ATTRIBS; attr++) { if (((1 << attr) & usedAttributes) == 0) break; } - if (attr == MAX_VERTEX_ATTRIBS) { + if (attr == MAX_VERTEX_GENERIC_ATTRIBS) { link_error(shProg, "Too many vertex attributes"); return GL_FALSE; } @@ -712,6 +717,8 @@ _slang_link(GLcontext *ctx, struct gl_vertex_program *linked_vprog = vertex_program(_mesa_clone_program(ctx, &vertProg->Base)); shProg->VertexProgram = linked_vprog; /* refcount OK */ + /* vertex program ID not significant; just set Id for debugging purposes */ + shProg->VertexProgram->Base.Id = shProg->Name; ASSERT(shProg->VertexProgram->Base.RefCount == 1); } @@ -720,16 +727,18 @@ _slang_link(GLcontext *ctx, struct gl_fragment_program *linked_fprog = fragment_program(_mesa_clone_program(ctx, &fragProg->Base)); shProg->FragmentProgram = linked_fprog; /* refcount OK */ + /* vertex program ID not significant; just set Id for debugging purposes */ + shProg->FragmentProgram->Base.Id = shProg->Name; ASSERT(shProg->FragmentProgram->Base.RefCount == 1); } /* link varying vars */ if (shProg->VertexProgram) { - if (!link_varying_vars(shProg, &shProg->VertexProgram->Base)) + if (!link_varying_vars(ctx, shProg, &shProg->VertexProgram->Base)) return; } if (shProg->FragmentProgram) { - if (!link_varying_vars(shProg, &shProg->FragmentProgram->Base)) + if (!link_varying_vars(ctx, shProg, &shProg->FragmentProgram->Base)) return; } diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c index ff913ad883..e9a24cc009 100644 --- a/src/mesa/shader/slang/slang_preprocess.c +++ b/src/mesa/shader/slang/slang_preprocess.c @@ -1035,11 +1035,11 @@ preprocess_source (slang_string *output, const char *source, /* Parse optional macro parameters. */ while (prod[i++] != PARAM_END) { - if (state.cond.top->effective) { - pp_symbol *param; + pp_symbol *param; - id = (const char *) (&prod[i]); - idlen = _mesa_strlen (id); + id = (const char *) (&prod[i]); + idlen = _mesa_strlen (id); + if (state.cond.top->effective) { pp_annotate (output, "%s, ", id); param = pp_symbols_push (&symbol->parameters); if (param == NULL) @@ -1053,8 +1053,23 @@ preprocess_source (slang_string *output, const char *source, id = (const char *) (&prod[i]); idlen = _mesa_strlen (id); if (state.cond.top->effective) { + slang_string replacement; + expand_state es; + pp_annotate (output, ") %s", id); - slang_string_pushs (&symbol->replacement, id, idlen); + + slang_string_init(&replacement); + slang_string_pushs(&replacement, id, idlen); + + /* Expand macro replacement. */ + es.output = &symbol->replacement; + es.input = slang_string_cstr(&replacement); + es.state = &state; + if (!expand(&es, &state.symbols)) { + slang_string_free(&replacement); + goto error; + } + slang_string_free(&replacement); } i += idlen + 1; } @@ -1286,6 +1301,45 @@ error: /** + * Remove the continuation characters from the input string. + * This is the very first step in preprocessing and is effective + * even inside comment blocks. + * If there is a whitespace between a backslash and a newline, + * this is not considered as a line continuation. + * \return GL_TRUE for success, GL_FALSE otherwise. + */ +static GLboolean +_slang_preprocess_backslashes(slang_string *output, + const char *input) +{ + while (*input) { + if (input[0] == '\\') { + /* If a newline follows, eat the backslash and the newline. */ + if (input[1] == '\r') { + if (input[2] == '\n') { + input += 3; + } else { + input += 2; + } + } else if (input[1] == '\n') { + if (input[2] == '\r') { + input += 3; + } else { + input += 2; + } + } else { + /* Leave the backslash alone. */ + slang_string_pushc(output, *input++); + } + } else { + slang_string_pushc(output, *input++); + } + } + return GL_TRUE; +} + + +/** * Run preprocessor on source code. * \param extensions indicates which GL extensions are enabled * \param output the post-process results @@ -1304,6 +1358,7 @@ _slang_preprocess_directives(slang_string *output, { grammar pid, eid; GLboolean success; + slang_string without_backslashes; pid = grammar_load_from_text ((const byte *) (slang_pp_directives_syn)); if (pid == 0) { @@ -1316,9 +1371,36 @@ _slang_preprocess_directives(slang_string *output, grammar_destroy (pid); return GL_FALSE; } - success = preprocess_source (output, input, pid, eid, elog, extensions, pragmas); + + slang_string_init(&without_backslashes); + success = _slang_preprocess_backslashes(&without_backslashes, input); + + if (0) { + _mesa_printf("Pre-processed shader:\n"); + _mesa_printf("%s", slang_string_cstr(&without_backslashes)); + _mesa_printf("----------------------\n"); + } + + if (success) { + success = preprocess_source(output, + slang_string_cstr(&without_backslashes), + pid, + eid, + elog, + extensions, + pragmas); + } + + slang_string_free(&without_backslashes); grammar_destroy (eid); grammar_destroy (pid); + + if (0) { + _mesa_printf("Post-processed shader:\n"); + _mesa_printf("%s", slang_string_cstr(output)); + _mesa_printf("----------------------\n"); + } + return success; } |