static void
create_new_program( const struct state_key *key,
struct gl_vertex_program *program,
GLboolean mvp_with_dp4,
GLuint max_temps)
{
struct tnl_program p;
memset(&p, 0, sizeof(p));
p.state = key;
p.program = program;
p.eye_position = undef;
p.eye_position_z = undef;
p.eye_position_normalized = undef;
p.transformed_normal = undef;
p.identity = undef;
p.temp_in_use = 0;
p.mvp_with_dp4 = mvp_with_dp4;
if (max_temps >= sizeof(int) * 8)
p.temp_reserved = 0;
else
p.temp_reserved = ~((1<<max_temps)-1);
/* Start by allocating 32 instructions.
* If we need more, we'll grow the instruction array as needed.
*/
p.max_inst = 32;
p.program->Base.Instructions = _mesa_alloc_instructions(p.max_inst);
p.program->Base.String = NULL;
p.program->Base.NumInstructions =
p.program->Base.NumTemporaries =
p.program->Base.NumParameters =
p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0;
p.program->Base.Parameters = _mesa_new_parameter_list();
p.program->Base.InputsRead = 0;
p.program->Base.OutputsWritten = 0;
build_tnl_program( &p );
}
/**
* Combine two programs into one. Fix instructions so the outputs of
* the first program go to the inputs of the second program.
*/
struct gl_program *
_mesa_combine_programs(GLcontext *ctx,
const struct gl_program *progA,
const struct gl_program *progB)
{
struct prog_instruction *newInst;
struct gl_program *newProg;
const GLuint lenA = progA->NumInstructions - 1; /* omit END instr */
const GLuint lenB = progB->NumInstructions;
const GLuint numParamsA = _mesa_num_parameters(progA->Parameters);
const GLuint newLength = lenA + lenB;
GLbitfield inputsB;
GLuint i;
ASSERT(progA->Target == progB->Target);
newInst = _mesa_alloc_instructions(newLength);
if (!newInst)
return GL_FALSE;
_mesa_copy_instructions(newInst, progA->Instructions, lenA);
_mesa_copy_instructions(newInst + lenA, progB->Instructions, lenB);
/* adjust branch / instruction addresses for B's instructions */
for (i = 0; i < lenB; i++) {
newInst[lenA + i].BranchTarget += lenA;
}
newProg = ctx->Driver.NewProgram(ctx, progA->Target, 0);
newProg->Instructions = newInst;
newProg->NumInstructions = newLength;
if (newProg->Target == GL_FRAGMENT_PROGRAM_ARB) {
struct gl_fragment_program *fprogA, *fprogB, *newFprog;
fprogA = (struct gl_fragment_program *) progA;
fprogB = (struct gl_fragment_program *) progB;
newFprog = (struct gl_fragment_program *) newProg;
newFprog->UsesKill = fprogA->UsesKill || fprogB->UsesKill;
/* Connect color outputs of fprogA to color inputs of fprogB, via a
* new temporary register.
*/
if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) &&
(progB->InputsRead & (1 << FRAG_ATTRIB_COL0))) {
GLint tempReg = _mesa_find_free_register(newProg, PROGRAM_TEMPORARY);
if (tempReg < 0) {
_mesa_problem(ctx, "No free temp regs found in "
"_mesa_combine_programs(), using 31");
tempReg = 31;
}
/* replace writes to result.color[0] with tempReg */
replace_registers(newInst, lenA,
PROGRAM_OUTPUT, FRAG_RESULT_COLR,
PROGRAM_TEMPORARY, tempReg);
/* replace reads from input.color[0] with tempReg */
replace_registers(newInst + lenA, lenB,
PROGRAM_INPUT, FRAG_ATTRIB_COL0,
PROGRAM_TEMPORARY, tempReg);
}
inputsB = progB->InputsRead;
if (progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) {
inputsB &= ~(1 << FRAG_ATTRIB_COL0);
}
newProg->InputsRead = progA->InputsRead | inputsB;
newProg->OutputsWritten = progB->OutputsWritten;
newProg->SamplersUsed = progA->SamplersUsed | progB->SamplersUsed;
}
else {
/* vertex program */
assert(0); /* XXX todo */
}
/*
* Merge parameters (uniforms, constants, etc)
*/
newProg->Parameters = _mesa_combine_parameter_lists(progA->Parameters,
progB->Parameters);
adjust_param_indexes(newInst + lenA, lenB, numParamsA);
return newProg;
}
/**
* Return a copy of a program.
* XXX Problem here if the program object is actually OO-derivation
* made by a device driver.
*/
struct gl_program *
_mesa_clone_program(GLcontext *ctx, const struct gl_program *prog)
{
struct gl_program *clone;
clone = ctx->Driver.NewProgram(ctx, prog->Target, prog->Id);
if (!clone)
return NULL;
assert(clone->Target == prog->Target);
assert(clone->RefCount == 1);
clone->String = (GLubyte *) _mesa_strdup((char *) prog->String);
clone->Format = prog->Format;
clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions);
if (!clone->Instructions) {
_mesa_reference_program(ctx, &clone, NULL);
return NULL;
}
_mesa_copy_instructions(clone->Instructions, prog->Instructions,
prog->NumInstructions);
clone->InputsRead = prog->InputsRead;
clone->OutputsWritten = prog->OutputsWritten;
clone->SamplersUsed = prog->SamplersUsed;
clone->ShadowSamplers = prog->ShadowSamplers;
memcpy(clone->TexturesUsed, prog->TexturesUsed, sizeof(prog->TexturesUsed));
if (prog->Parameters)
clone->Parameters = _mesa_clone_parameter_list(prog->Parameters);
memcpy(clone->LocalParams, prog->LocalParams, sizeof(clone->LocalParams));
if (prog->Varying)
clone->Varying = _mesa_clone_parameter_list(prog->Varying);
if (prog->Attributes)
clone->Attributes = _mesa_clone_parameter_list(prog->Attributes);
memcpy(clone->LocalParams, prog->LocalParams, sizeof(clone->LocalParams));
clone->NumInstructions = prog->NumInstructions;
clone->NumTemporaries = prog->NumTemporaries;
clone->NumParameters = prog->NumParameters;
clone->NumAttributes = prog->NumAttributes;
clone->NumAddressRegs = prog->NumAddressRegs;
clone->NumNativeInstructions = prog->NumNativeInstructions;
clone->NumNativeTemporaries = prog->NumNativeTemporaries;
clone->NumNativeParameters = prog->NumNativeParameters;
clone->NumNativeAttributes = prog->NumNativeAttributes;
clone->NumNativeAddressRegs = prog->NumNativeAddressRegs;
clone->NumAluInstructions = prog->NumAluInstructions;
clone->NumTexInstructions = prog->NumTexInstructions;
clone->NumTexIndirections = prog->NumTexIndirections;
clone->NumNativeAluInstructions = prog->NumNativeAluInstructions;
clone->NumNativeTexInstructions = prog->NumNativeTexInstructions;
clone->NumNativeTexIndirections = prog->NumNativeTexIndirections;
switch (prog->Target) {
case GL_VERTEX_PROGRAM_ARB:
{
const struct gl_vertex_program *vp
= (const struct gl_vertex_program *) prog;
struct gl_vertex_program *vpc = (struct gl_vertex_program *) clone;
vpc->IsPositionInvariant = vp->IsPositionInvariant;
}
break;
case GL_FRAGMENT_PROGRAM_ARB:
{
const struct gl_fragment_program *fp
= (const struct gl_fragment_program *) prog;
struct gl_fragment_program *fpc = (struct gl_fragment_program *) clone;
fpc->FogOption = fp->FogOption;
fpc->UsesKill = fp->UsesKill;
}
break;
default:
_mesa_problem(NULL, "Unexpected target in _mesa_clone_program");
}
return clone;
}
/**
* This function inserts instructions for coordinate modelview * projection
* into a vertex program.
* May be used to implement the position_invariant option.
*/
static void
_mesa_insert_mvp_dp4_code(struct gl_context *ctx, struct gl_vertex_program *vprog)
{
struct prog_instruction *newInst;
const GLuint origLen = vprog->Base.NumInstructions;
const GLuint newLen = origLen + 4;
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, 0 }, /* state.matrix.mvp.row[0] */
{ STATE_MVP_MATRIX, 0, 1, 1, 0 }, /* state.matrix.mvp.row[1] */
{ STATE_MVP_MATRIX, 0, 2, 2, 0 }, /* state.matrix.mvp.row[2] */
{ STATE_MVP_MATRIX, 0, 3, 3, 0 }, /* state.matrix.mvp.row[3] */
};
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;
}
/*
* Generated instructions:
* newInst[0] = DP4 result.position.x, mvp.row[0], vertex.position;
* newInst[1] = DP4 result.position.y, mvp.row[1], vertex.position;
* newInst[2] = DP4 result.position.z, mvp.row[2], vertex.position;
* newInst[3] = DP4 result.position.w, mvp.row[3], vertex.position;
*/
_mesa_init_instructions(newInst, 4);
for (i = 0; i < 4; i++) {
newInst[i].Opcode = OPCODE_DP4;
newInst[i].DstReg.File = PROGRAM_OUTPUT;
newInst[i].DstReg.Index = VARYING_SLOT_POS;
newInst[i].DstReg.WriteMask = (WRITEMASK_X << i);
newInst[i].SrcReg[0].File = PROGRAM_STATE_VAR;
newInst[i].SrcReg[0].Index = mvpRef[i];
newInst[i].SrcReg[0].Swizzle = SWIZZLE_NOOP;
newInst[i].SrcReg[1].File = PROGRAM_INPUT;
newInst[i].SrcReg[1].Index = VERT_ATTRIB_POS;
newInst[i].SrcReg[1].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 |= BITFIELD64_BIT(VARYING_SLOT_POS);
}
/**
* Append instructions to implement fog
*
* The \c fragment.fogcoord input is used to compute the fog blend factor.
*
* \param ctx The GL context
* \param fprog Fragment program that fog instructions will be appended to.
* \param fog_mode Fog mode. One of \c GL_EXP, \c GL_EXP2, or \c GL_LINEAR.
* \param saturate True if writes to color outputs should be clamped to [0, 1]
*
* \note
* This function sets \c VARYING_BIT_FOGC in \c fprog->Base.InputsRead.
*
* \todo With a little work, this function could be adapted to add fog code
* to vertex programs too.
*/
void
_mesa_append_fog_code(struct gl_context *ctx,
struct gl_fragment_program *fprog, GLenum fog_mode,
GLboolean saturate)
{
static const gl_state_index fogPStateOpt[STATE_LENGTH]
= { STATE_INTERNAL, STATE_FOG_PARAMS_OPTIMIZED, 0, 0, 0 };
static const gl_state_index fogColorState[STATE_LENGTH]
= { STATE_FOG_COLOR, 0, 0, 0, 0};
struct prog_instruction *newInst, *inst;
const GLuint origLen = fprog->Base.NumInstructions;
const GLuint newLen = origLen + 5;
GLuint i;
GLint fogPRefOpt, fogColorRef; /* state references */
GLuint colorTemp, fogFactorTemp; /* temporary registerss */
if (fog_mode == GL_NONE) {
_mesa_problem(ctx, "_mesa_append_fog_code() called for fragment program"
" with fog_mode == GL_NONE");
return;
}
if (!(fprog->Base.OutputsWritten & (1 << FRAG_RESULT_COLOR))) {
/* program doesn't output color, so nothing to do */
return;
}
/* Alloc storage for new instructions */
newInst = _mesa_alloc_instructions(newLen);
if (!newInst) {
_mesa_error(ctx, GL_OUT_OF_MEMORY,
"glProgramString(inserting fog_option code)");
return;
}
/* Copy orig instructions into new instruction buffer */
_mesa_copy_instructions(newInst, fprog->Base.Instructions, origLen);
/* PARAM fogParamsRefOpt = internal optimized fog params; */
fogPRefOpt
= _mesa_add_state_reference(fprog->Base.Parameters, fogPStateOpt);
/* PARAM fogColorRef = state.fog.color; */
fogColorRef
= _mesa_add_state_reference(fprog->Base.Parameters, fogColorState);
/* TEMP colorTemp; */
colorTemp = fprog->Base.NumTemporaries++;
/* TEMP fogFactorTemp; */
fogFactorTemp = fprog->Base.NumTemporaries++;
/* Scan program to find where result.color is written */
inst = newInst;
for (i = 0; i < fprog->Base.NumInstructions; i++) {
if (inst->Opcode == OPCODE_END)
break;
if (inst->DstReg.File == PROGRAM_OUTPUT &&
inst->DstReg.Index == FRAG_RESULT_COLOR) {
/* change the instruction to write to colorTemp w/ clamping */
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = colorTemp;
inst->SaturateMode = saturate;
/* don't break (may be several writes to result.color) */
}
inst++;
}
assert(inst->Opcode == OPCODE_END); /* we'll overwrite this inst */
_mesa_init_instructions(inst, 5);
/* emit instructions to compute fog blending factor */
/* this is always clamped to [0, 1] regardless of fragment clamping */
if (fog_mode == GL_LINEAR) {
/* MAD fogFactorTemp.x, fragment.fogcoord.x, fogPRefOpt.x, fogPRefOpt.y; */
inst->Opcode = OPCODE_MAD;
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = fogFactorTemp;
inst->DstReg.WriteMask = WRITEMASK_X;
inst->SrcReg[0].File = PROGRAM_INPUT;
inst->SrcReg[0].Index = VARYING_SLOT_FOGC;
inst->SrcReg[0].Swizzle = SWIZZLE_XXXX;
inst->SrcReg[1].File = PROGRAM_STATE_VAR;
inst->SrcReg[1].Index = fogPRefOpt;
inst->SrcReg[1].Swizzle = SWIZZLE_XXXX;
inst->SrcReg[2].File = PROGRAM_STATE_VAR;
inst->SrcReg[2].Index = fogPRefOpt;
inst->SrcReg[2].Swizzle = SWIZZLE_YYYY;
inst->SaturateMode = SATURATE_ZERO_ONE;
inst++;
}
else {
ASSERT(fog_mode == GL_EXP || fog_mode == GL_EXP2);
/* fogPRefOpt.z = d/ln(2), fogPRefOpt.w = d/sqrt(ln(2) */
/* EXP: MUL fogFactorTemp.x, fogPRefOpt.z, fragment.fogcoord.x; */
/* EXP2: MUL fogFactorTemp.x, fogPRefOpt.w, fragment.fogcoord.x; */
inst->Opcode = OPCODE_MUL;
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = fogFactorTemp;
inst->DstReg.WriteMask = WRITEMASK_X;
inst->SrcReg[0].File = PROGRAM_STATE_VAR;
inst->SrcReg[0].Index = fogPRefOpt;
inst->SrcReg[0].Swizzle
= (fog_mode == GL_EXP) ? SWIZZLE_ZZZZ : SWIZZLE_WWWW;
inst->SrcReg[1].File = PROGRAM_INPUT;
inst->SrcReg[1].Index = VARYING_SLOT_FOGC;
inst->SrcReg[1].Swizzle = SWIZZLE_XXXX;
inst++;
if (fog_mode == GL_EXP2) {
/* MUL fogFactorTemp.x, fogFactorTemp.x, fogFactorTemp.x; */
inst->Opcode = OPCODE_MUL;
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = fogFactorTemp;
inst->DstReg.WriteMask = WRITEMASK_X;
inst->SrcReg[0].File = PROGRAM_TEMPORARY;
inst->SrcReg[0].Index = fogFactorTemp;
inst->SrcReg[0].Swizzle = SWIZZLE_XXXX;
inst->SrcReg[1].File = PROGRAM_TEMPORARY;
inst->SrcReg[1].Index = fogFactorTemp;
inst->SrcReg[1].Swizzle = SWIZZLE_XXXX;
inst++;
}
/* EX2_SAT fogFactorTemp.x, -fogFactorTemp.x; */
inst->Opcode = OPCODE_EX2;
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = fogFactorTemp;
inst->DstReg.WriteMask = WRITEMASK_X;
inst->SrcReg[0].File = PROGRAM_TEMPORARY;
inst->SrcReg[0].Index = fogFactorTemp;
inst->SrcReg[0].Negate = NEGATE_XYZW;
inst->SrcReg[0].Swizzle = SWIZZLE_XXXX;
inst->SaturateMode = SATURATE_ZERO_ONE;
inst++;
}
/* LRP result.color.xyz, fogFactorTemp.xxxx, colorTemp, fogColorRef; */
inst->Opcode = OPCODE_LRP;
inst->DstReg.File = PROGRAM_OUTPUT;
inst->DstReg.Index = FRAG_RESULT_COLOR;
inst->DstReg.WriteMask = WRITEMASK_XYZ;
inst->SrcReg[0].File = PROGRAM_TEMPORARY;
inst->SrcReg[0].Index = fogFactorTemp;
inst->SrcReg[0].Swizzle = SWIZZLE_XXXX;
inst->SrcReg[1].File = PROGRAM_TEMPORARY;
inst->SrcReg[1].Index = colorTemp;
inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
inst->SrcReg[2].File = PROGRAM_STATE_VAR;
inst->SrcReg[2].Index = fogColorRef;
inst->SrcReg[2].Swizzle = SWIZZLE_NOOP;
inst++;
/* MOV result.color.w, colorTemp.x; # copy alpha */
inst->Opcode = OPCODE_MOV;
inst->DstReg.File = PROGRAM_OUTPUT;
inst->DstReg.Index = FRAG_RESULT_COLOR;
inst->DstReg.WriteMask = WRITEMASK_W;
inst->SrcReg[0].File = PROGRAM_TEMPORARY;
inst->SrcReg[0].Index = colorTemp;
inst->SrcReg[0].Swizzle = SWIZZLE_NOOP;
inst++;
/* END; */
inst->Opcode = OPCODE_END;
inst++;
/* free old instructions */
_mesa_free_instructions(fprog->Base.Instructions, origLen);
/* install new instructions */
fprog->Base.Instructions = newInst;
fprog->Base.NumInstructions = inst - newInst;
fprog->Base.InputsRead |= VARYING_BIT_FOGC;
assert(fprog->Base.OutputsWritten & (1 << FRAG_RESULT_COLOR));
}
static void
_mesa_insert_mvp_mad_code(struct gl_context *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 = VARYING_SLOT_POS;
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 |= BITFIELD64_BIT(VARYING_SLOT_POS);
}
/**
* Combine two programs into one. Fix instructions so the outputs of
* the first program go to the inputs of the second program.
*/
struct gl_program *
_mesa_combine_programs(struct gl_context *ctx,
const struct gl_program *progA,
const struct gl_program *progB)
{
struct prog_instruction *newInst;
struct gl_program *newProg;
const GLuint lenA = progA->NumInstructions - 1; /* omit END instr */
const GLuint lenB = progB->NumInstructions;
const GLuint numParamsA = _mesa_num_parameters(progA->Parameters);
const GLuint newLength = lenA + lenB;
GLboolean usedTemps[MAX_PROGRAM_TEMPS];
GLuint firstTemp = 0;
GLbitfield inputsB;
GLuint i;
ASSERT(progA->Target == progB->Target);
newInst = _mesa_alloc_instructions(newLength);
if (!newInst)
return GL_FALSE;
_mesa_copy_instructions(newInst, progA->Instructions, lenA);
_mesa_copy_instructions(newInst + lenA, progB->Instructions, lenB);
/* adjust branch / instruction addresses for B's instructions */
for (i = 0; i < lenB; i++) {
newInst[lenA + i].BranchTarget += lenA;
}
newProg = ctx->Driver.NewProgram(ctx, progA->Target, 0);
newProg->Instructions = newInst;
newProg->NumInstructions = newLength;
/* find used temp regs (we may need new temps below) */
_mesa_find_used_registers(newProg, PROGRAM_TEMPORARY,
usedTemps, MAX_PROGRAM_TEMPS);
if (newProg->Target == GL_FRAGMENT_PROGRAM_ARB) {
struct gl_fragment_program *fprogA, *fprogB, *newFprog;
GLbitfield progB_inputsRead = progB->InputsRead;
GLint progB_colorFile, progB_colorIndex;
fprogA = (struct gl_fragment_program *) progA;
fprogB = (struct gl_fragment_program *) progB;
newFprog = (struct gl_fragment_program *) newProg;
newFprog->UsesKill = fprogA->UsesKill || fprogB->UsesKill;
/* We'll do a search and replace for instances
* of progB_colorFile/progB_colorIndex below...
*/
progB_colorFile = PROGRAM_INPUT;
progB_colorIndex = FRAG_ATTRIB_COL0;
/*
* The fragment program may get color from a state var rather than
* a fragment input (vertex output) if it's constant.
* See the texenvprogram.c code.
* So, search the program's parameter list now to see if the program
* gets color from a state var instead of a conventional fragment
* input register.
*/
for (i = 0; i < progB->Parameters->NumParameters; i++) {
struct gl_program_parameter *p = &progB->Parameters->Parameters[i];
if (p->Type == PROGRAM_STATE_VAR &&
p->StateIndexes[0] == STATE_INTERNAL &&
p->StateIndexes[1] == STATE_CURRENT_ATTRIB &&
(int) p->StateIndexes[2] == (int) VERT_ATTRIB_COLOR0) {
progB_inputsRead |= FRAG_BIT_COL0;
progB_colorFile = PROGRAM_STATE_VAR;
progB_colorIndex = i;
break;
}
}
/* Connect color outputs of fprogA to color inputs of fprogB, via a
* new temporary register.
*/
if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLOR)) &&
(progB_inputsRead & FRAG_BIT_COL0)) {
GLint tempReg = _mesa_find_free_register(usedTemps, MAX_PROGRAM_TEMPS,
firstTemp);
if (tempReg < 0) {
_mesa_problem(ctx, "No free temp regs found in "
"_mesa_combine_programs(), using 31");
tempReg = 31;
}
firstTemp = tempReg + 1;
/* replace writes to result.color[0] with tempReg */
replace_registers(newInst, lenA,
PROGRAM_OUTPUT, FRAG_RESULT_COLOR,
PROGRAM_TEMPORARY, tempReg);
/* replace reads from the input color with tempReg */
replace_registers(newInst + lenA, lenB,
progB_colorFile, progB_colorIndex, /* search for */
PROGRAM_TEMPORARY, tempReg /* replace with */ );
}
/* compute combined program's InputsRead */
inputsB = progB_inputsRead;
if (progA->OutputsWritten & (1 << FRAG_RESULT_COLOR)) {
inputsB &= ~(1 << FRAG_ATTRIB_COL0);
}
newProg->InputsRead = progA->InputsRead | inputsB;
newProg->OutputsWritten = progB->OutputsWritten;
newProg->SamplersUsed = progA->SamplersUsed | progB->SamplersUsed;
}
else {
/* vertex program */
assert(0); /* XXX todo */
}
/*
* Merge parameters (uniforms, constants, etc)
*/
newProg->Parameters = _mesa_combine_parameter_lists(progA->Parameters,
progB->Parameters);
adjust_param_indexes(newInst + lenA, lenB, numParamsA);
return newProg;
}
/**
* Parse/compile the 'str' returning the compiled 'program'.
* ctx->Program.ErrorPos will be -1 if successful. Otherwise, ErrorPos
* indicates the position of the error in 'str'.
*/
void
_mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum dstTarget,
const GLubyte *str, GLsizei len,
struct gl_vertex_program *program)
{
struct parse_state parseState;
struct prog_instruction instBuffer[MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS];
struct prog_instruction *newInst;
GLenum target;
GLubyte *programString;
/* Make a null-terminated copy of the program string */
programString = (GLubyte *) MALLOC(len + 1);
if (!programString) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glLoadProgramNV");
return;
}
MEMCPY(programString, str, len);
programString[len] = 0;
/* Get ready to parse */
parseState.ctx = ctx;
parseState.start = programString;
parseState.isPositionInvariant = GL_FALSE;
parseState.isVersion1_1 = GL_FALSE;
parseState.numInst = 0;
parseState.inputsRead = 0;
parseState.outputsWritten = 0;
parseState.anyProgRegsWritten = GL_FALSE;
/* Reset error state */
_mesa_set_program_error(ctx, -1, NULL);
/* check the program header */
if (_mesa_strncmp((const char *) programString, "!!VP1.0", 7) == 0) {
target = GL_VERTEX_PROGRAM_NV;
parseState.pos = programString + 7;
parseState.isStateProgram = GL_FALSE;
}
else if (_mesa_strncmp((const char *) programString, "!!VP1.1", 7) == 0) {
target = GL_VERTEX_PROGRAM_NV;
parseState.pos = programString + 7;
parseState.isStateProgram = GL_FALSE;
parseState.isVersion1_1 = GL_TRUE;
}
else if (_mesa_strncmp((const char *) programString, "!!VSP1.0", 8) == 0) {
target = GL_VERTEX_STATE_PROGRAM_NV;
parseState.pos = programString + 8;
parseState.isStateProgram = GL_TRUE;
}
else {
/* invalid header */
ctx->Program.ErrorPos = 0;
_mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV(bad header)");
return;
}
/* make sure target and header match */
if (target != dstTarget) {
_mesa_error(ctx, GL_INVALID_OPERATION,
"glLoadProgramNV(target mismatch)");
return;
}
if (Parse_Program(&parseState, instBuffer)) {
/* successful parse! */
if (parseState.isStateProgram) {
if (!parseState.anyProgRegsWritten) {
_mesa_error(ctx, GL_INVALID_OPERATION,
"glLoadProgramNV(c[#] not written)");
return;
}
}
else {
if (!parseState.isPositionInvariant &&
!(parseState.outputsWritten & (1 << VERT_RESULT_HPOS))) {
/* bit 1 = HPOS register */
_mesa_error(ctx, GL_INVALID_OPERATION,
"glLoadProgramNV(HPOS not written)");
return;
}
}
/* copy the compiled instructions */
assert(parseState.numInst <= MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS);
newInst = _mesa_alloc_instructions(parseState.numInst);
if (!newInst) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glLoadProgramNV");
_mesa_free(programString);
return; /* out of memory */
}
_mesa_copy_instructions(newInst, instBuffer, parseState.numInst);
/* install the program */
program->Base.Target = target;
if (program->Base.String) {
_mesa_free(program->Base.String);
}
program->Base.String = programString;
program->Base.Format = GL_PROGRAM_FORMAT_ASCII_ARB;
if (program->Base.Instructions) {
_mesa_free(program->Base.Instructions);
}
program->Base.Instructions = newInst;
program->Base.InputsRead = parseState.inputsRead;
if (parseState.isPositionInvariant)
program->Base.InputsRead |= VERT_BIT_POS;
program->Base.NumInstructions = parseState.numInst;
program->Base.OutputsWritten = parseState.outputsWritten;
program->IsPositionInvariant = parseState.isPositionInvariant;
program->IsNVProgram = GL_TRUE;
#ifdef DEBUG_foo
_mesa_printf("--- glLoadProgramNV result ---\n");
_mesa_print_nv_vertex_program(program);
_mesa_printf("------------------------------\n");
#endif
}
else {
/* Error! */
_mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV");
/* NOTE: _mesa_set_program_error would have been called already */
/* GL_NV_vertex_program isn't supposed to set the error string
* so we reset it here.
*/
_mesa_set_program_error(ctx, ctx->Program.ErrorPos, NULL);
}
}
/**
* Parse/compile the 'str' returning the compiled 'program'.
* ctx->Program.ErrorPos will be -1 if successful. Otherwise, ErrorPos
* indicates the position of the error in 'str'.
*/
void
_mesa_parse_nv_fragment_program(struct gl_context *ctx, GLenum dstTarget,
const GLubyte *str, GLsizei len,
struct gl_fragment_program *program)
{
struct parse_state parseState;
struct prog_instruction instBuffer[MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS];
struct prog_instruction *newInst;
GLenum target;
GLubyte *programString;
/* Make a null-terminated copy of the program string */
programString = (GLubyte *) MALLOC(len + 1);
if (!programString) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glLoadProgramNV");
return;
}
memcpy(programString, str, len);
programString[len] = 0;
/* Get ready to parse */
memset(&parseState, 0, sizeof(struct parse_state));
parseState.ctx = ctx;
parseState.start = programString;
parseState.program = program;
parseState.numInst = 0;
parseState.curLine = programString;
parseState.parameters = _mesa_new_parameter_list();
/* Reset error state */
_mesa_set_program_error(ctx, -1, NULL);
/* check the program header */
if (strncmp((const char *) programString, "!!FP1.0", 7) == 0) {
target = GL_FRAGMENT_PROGRAM_NV;
parseState.pos = programString + 7;
}
else if (strncmp((const char *) programString, "!!FCP1.0", 8) == 0) {
/* fragment / register combiner program - not supported */
_mesa_set_program_error(ctx, 0, "Invalid fragment program header");
_mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV(bad header)");
return;
}
else {
/* invalid header */
_mesa_set_program_error(ctx, 0, "Invalid fragment program header");
_mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV(bad header)");
return;
}
/* make sure target and header match */
if (target != dstTarget) {
_mesa_error(ctx, GL_INVALID_OPERATION,
"glLoadProgramNV(target mismatch 0x%x != 0x%x)",
target, dstTarget);
return;
}
if (Parse_InstructionSequence(&parseState, instBuffer)) {
GLuint u;
/* successful parse! */
if (parseState.outputsWritten == 0) {
/* must write at least one output! */
_mesa_error(ctx, GL_INVALID_OPERATION,
"Invalid fragment program - no outputs written.");
return;
}
/* copy the compiled instructions */
assert(parseState.numInst <= MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS);
newInst = _mesa_alloc_instructions(parseState.numInst);
if (!newInst) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glLoadProgramNV");
return; /* out of memory */
}
_mesa_copy_instructions(newInst, instBuffer, parseState.numInst);
/* install the program */
program->Base.Target = target;
if (program->Base.String) {
FREE(program->Base.String);
}
program->Base.String = programString;
program->Base.Format = GL_PROGRAM_FORMAT_ASCII_ARB;
if (program->Base.Instructions) {
free(program->Base.Instructions);
}
program->Base.Instructions = newInst;
program->Base.NumInstructions = parseState.numInst;
program->Base.InputsRead = parseState.inputsRead;
program->Base.OutputsWritten = parseState.outputsWritten;
for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++)
program->Base.TexturesUsed[u] = parseState.texturesUsed[u];
/* save program parameters */
program->Base.Parameters = parseState.parameters;
/* allocate registers for declared program parameters */
#if 00
_mesa_assign_program_registers(&(program->SymbolTable));
#endif
#ifdef DEBUG_foo
printf("--- glLoadProgramNV(%d) result ---\n", program->Base.Id);
_mesa_fprint_program_opt(stdout, &program->Base, PROG_PRINT_NV, 0);
printf("----------------------------------\n");
#endif
}
else {
/* Error! */
_mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV");
/* NOTE: _mesa_set_program_error would have been called already */
}
}
/**
* Create fragment program that does a TEX() instruction to get a Z and/or
* stencil value value, then writes to FRAG_RESULT_DEPTH/FRAG_RESULT_STENCIL.
* Used for glDrawPixels(GL_DEPTH_COMPONENT / GL_STENCIL_INDEX).
* Pass fragment color through as-is.
* \return pointer to the gl_fragment program
*/
struct gl_fragment_program *
st_make_drawpix_z_stencil_program(struct st_context *st,
GLboolean write_depth,
GLboolean write_stencil)
{
struct gl_context *ctx = st->ctx;
struct gl_program *p;
struct gl_fragment_program *fp;
GLuint ic = 0;
const GLuint shaderIndex = write_depth * 2 + write_stencil;
assert(shaderIndex < Elements(st->drawpix.shaders));
if (st->drawpix.shaders[shaderIndex]) {
/* already have the proper shader */
return st->drawpix.shaders[shaderIndex];
}
/*
* Create shader now
*/
p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
if (!p)
return NULL;
p->NumInstructions = write_depth ? 2 : 1;
p->NumInstructions += write_stencil ? 1 : 0;
p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
if (!p->Instructions) {
ctx->Driver.DeleteProgram(ctx, p);
return NULL;
}
_mesa_init_instructions(p->Instructions, p->NumInstructions);
if (write_depth) {
/* TEX result.depth, fragment.texcoord[0], texture[0], 2D; */
p->Instructions[ic].Opcode = OPCODE_TEX;
p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
p->Instructions[ic].DstReg.Index = FRAG_RESULT_DEPTH;
p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Z;
p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
p->Instructions[ic].TexSrcUnit = 0;
p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
ic++;
}
if (write_stencil) {
/* TEX result.stencil, fragment.texcoord[0], texture[0], 2D; */
p->Instructions[ic].Opcode = OPCODE_TEX;
p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
p->Instructions[ic].DstReg.Index = FRAG_RESULT_STENCIL;
p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Y;
p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
p->Instructions[ic].TexSrcUnit = 1;
p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
ic++;
}
/* END; */
p->Instructions[ic++].Opcode = OPCODE_END;
assert(ic == p->NumInstructions);
p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0;
p->OutputsWritten = 0;
if (write_depth)
p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_DEPTH);
if (write_stencil)
p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_STENCIL);
p->SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */
if (write_stencil)
p->SamplersUsed |= 1 << 1;
fp = (struct gl_fragment_program *) p;
/* save the new shader */
st->drawpix.shaders[shaderIndex] = fp;
return fp;
}
/**
* Returns a fragment program which implements the current pixel transfer ops.
*/
static struct gl_fragment_program *
get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key)
{
struct st_context *st = ctx->st;
struct prog_instruction inst[MAX_INST];
struct gl_program_parameter_list *params;
struct gl_fragment_program *fp;
GLuint ic = 0;
const GLuint colorTemp = 0;
fp = (struct gl_fragment_program *)
ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
if (!fp)
return NULL;
params = _mesa_new_parameter_list();
/*
* Get initial pixel color from the texture.
* TEX colorTemp, fragment.texcoord[0], texture[0], 2D;
*/
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_TEX;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = colorTemp;
inst[ic].SrcReg[0].File = PROGRAM_INPUT;
inst[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
inst[ic].TexSrcUnit = 0;
inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
ic++;
fp->Base.InputsRead = (1 << FRAG_ATTRIB_TEX0);
fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR);
fp->Base.SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */
if (key->scaleAndBias) {
static const gl_state_index scale_state[STATE_LENGTH] =
{ STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 };
static const gl_state_index bias_state[STATE_LENGTH] =
{ STATE_INTERNAL, STATE_PT_BIAS, 0, 0, 0 };
GLfloat scale[4], bias[4];
GLint scale_p, bias_p;
scale[0] = ctx->Pixel.RedScale;
scale[1] = ctx->Pixel.GreenScale;
scale[2] = ctx->Pixel.BlueScale;
scale[3] = ctx->Pixel.AlphaScale;
bias[0] = ctx->Pixel.RedBias;
bias[1] = ctx->Pixel.GreenBias;
bias[2] = ctx->Pixel.BlueBias;
bias[3] = ctx->Pixel.AlphaBias;
scale_p = _mesa_add_state_reference(params, scale_state);
bias_p = _mesa_add_state_reference(params, bias_state);
/* MAD colorTemp, colorTemp, scale, bias; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_MAD;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = colorTemp;
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = colorTemp;
inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
inst[ic].SrcReg[1].Index = scale_p;
inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR;
inst[ic].SrcReg[2].Index = bias_p;
ic++;
}
if (key->pixelMaps) {
const GLuint temp = 1;
/* create the colormap/texture now if not already done */
if (!st->pixel_xfer.pixelmap_texture) {
st->pixel_xfer.pixelmap_texture = create_color_map_texture(ctx);
}
/* with a little effort, we can do four pixel map look-ups with
* two TEX instructions:
*/
/* TEX temp.rg, colorTemp.rgba, texture[1], 2D; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_TEX;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = temp;
inst[ic].DstReg.WriteMask = WRITEMASK_XY; /* write R,G */
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = colorTemp;
inst[ic].TexSrcUnit = 1;
inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
ic++;
/* TEX temp.ba, colorTemp.baba, texture[1], 2D; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_TEX;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = temp;
inst[ic].DstReg.WriteMask = WRITEMASK_ZW; /* write B,A */
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = colorTemp;
inst[ic].SrcReg[0].Swizzle = MAKE_SWIZZLE4(SWIZZLE_Z, SWIZZLE_W,
SWIZZLE_Z, SWIZZLE_W);
inst[ic].TexSrcUnit = 1;
inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
ic++;
/* MOV colorTemp, temp; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_MOV;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = colorTemp;
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = temp;
ic++;
fp->Base.SamplersUsed |= (1 << 1); /* sampler 1 is used */
}
if (key->colorMatrix) {
static const gl_state_index row0_state[STATE_LENGTH] =
{ STATE_COLOR_MATRIX, 0, 0, 0, 0 };
static const gl_state_index row1_state[STATE_LENGTH] =
{ STATE_COLOR_MATRIX, 0, 1, 1, 0 };
static const gl_state_index row2_state[STATE_LENGTH] =
{ STATE_COLOR_MATRIX, 0, 2, 2, 0 };
static const gl_state_index row3_state[STATE_LENGTH] =
{ STATE_COLOR_MATRIX, 0, 3, 3, 0 };
GLint row0_p = _mesa_add_state_reference(params, row0_state);
GLint row1_p = _mesa_add_state_reference(params, row1_state);
GLint row2_p = _mesa_add_state_reference(params, row2_state);
GLint row3_p = _mesa_add_state_reference(params, row3_state);
const GLuint temp = 1;
/* DP4 temp.x, colorTemp, matrow0; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_DP4;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = temp;
inst[ic].DstReg.WriteMask = WRITEMASK_X;
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = colorTemp;
inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
inst[ic].SrcReg[1].Index = row0_p;
ic++;
/* DP4 temp.y, colorTemp, matrow1; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_DP4;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = temp;
inst[ic].DstReg.WriteMask = WRITEMASK_Y;
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = colorTemp;
inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
inst[ic].SrcReg[1].Index = row1_p;
ic++;
/* DP4 temp.z, colorTemp, matrow2; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_DP4;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = temp;
inst[ic].DstReg.WriteMask = WRITEMASK_Z;
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = colorTemp;
inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
inst[ic].SrcReg[1].Index = row2_p;
ic++;
/* DP4 temp.w, colorTemp, matrow3; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_DP4;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = temp;
inst[ic].DstReg.WriteMask = WRITEMASK_W;
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = colorTemp;
inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
inst[ic].SrcReg[1].Index = row3_p;
ic++;
/* MOV colorTemp, temp; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_MOV;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = colorTemp;
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = temp;
ic++;
}
if (key->colorMatrixPostScaleBias) {
static const gl_state_index scale_state[STATE_LENGTH] =
{ STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 };
static const gl_state_index bias_state[STATE_LENGTH] =
{ STATE_INTERNAL, STATE_PT_BIAS, 0, 0, 0 };
GLint scale_param, bias_param;
scale_param = _mesa_add_state_reference(params, scale_state);
bias_param = _mesa_add_state_reference(params, bias_state);
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_MAD;
inst[ic].DstReg.File = PROGRAM_TEMPORARY;
inst[ic].DstReg.Index = colorTemp;
inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
inst[ic].SrcReg[0].Index = colorTemp;
inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
inst[ic].SrcReg[1].Index = scale_param;
inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR;
inst[ic].SrcReg[2].Index = bias_param;
ic++;
}
/* Modify last instruction's dst reg to write to result.color */
{
struct prog_instruction *last = &inst[ic - 1];
last->DstReg.File = PROGRAM_OUTPUT;
last->DstReg.Index = FRAG_RESULT_COLR;
}
/* END; */
_mesa_init_instructions(inst + ic, 1);
inst[ic].Opcode = OPCODE_END;
ic++;
assert(ic <= MAX_INST);
fp->Base.Instructions = _mesa_alloc_instructions(ic);
if (!fp->Base.Instructions) {
_mesa_error(ctx, GL_OUT_OF_MEMORY,
"generating pixel transfer program");
return NULL;
}
_mesa_copy_instructions(fp->Base.Instructions, inst, ic);
fp->Base.NumInstructions = ic;
fp->Base.Parameters = params;
#if 0
printf("========= pixel transfer prog\n");
_mesa_print_program(&fp->Base);
_mesa_print_parameter_list(fp->Base.Parameters);
#endif
return fp;
}
/**
* Return a copy of a program.
* XXX Problem here if the program object is actually OO-derivation
* made by a device driver.
*/
struct gl_program *
_mesa_clone_program(struct gl_context *ctx, const struct gl_program *prog)
{
struct gl_program *clone;
clone = ctx->Driver.NewProgram(ctx, prog->Target, prog->Id);
if (!clone)
return NULL;
assert(clone->Target == prog->Target);
assert(clone->RefCount == 1);
clone->String = (GLubyte *) _mesa_strdup((char *) prog->String);
clone->Format = prog->Format;
clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions);
if (!clone->Instructions) {
_mesa_reference_program(ctx, &clone, NULL);
return NULL;
}
_mesa_copy_instructions(clone->Instructions, prog->Instructions,
prog->NumInstructions);
clone->InputsRead = prog->InputsRead;
clone->OutputsWritten = prog->OutputsWritten;
clone->SamplersUsed = prog->SamplersUsed;
clone->ShadowSamplers = prog->ShadowSamplers;
memcpy(clone->TexturesUsed, prog->TexturesUsed, sizeof(prog->TexturesUsed));
if (prog->Parameters)
clone->Parameters = _mesa_clone_parameter_list(prog->Parameters);
memcpy(clone->LocalParams, prog->LocalParams, sizeof(clone->LocalParams));
memcpy(clone->LocalParams, prog->LocalParams, sizeof(clone->LocalParams));
clone->IndirectRegisterFiles = prog->IndirectRegisterFiles;
clone->NumInstructions = prog->NumInstructions;
clone->NumTemporaries = prog->NumTemporaries;
clone->NumParameters = prog->NumParameters;
clone->NumAttributes = prog->NumAttributes;
clone->NumAddressRegs = prog->NumAddressRegs;
clone->NumNativeInstructions = prog->NumNativeInstructions;
clone->NumNativeTemporaries = prog->NumNativeTemporaries;
clone->NumNativeParameters = prog->NumNativeParameters;
clone->NumNativeAttributes = prog->NumNativeAttributes;
clone->NumNativeAddressRegs = prog->NumNativeAddressRegs;
clone->NumAluInstructions = prog->NumAluInstructions;
clone->NumTexInstructions = prog->NumTexInstructions;
clone->NumTexIndirections = prog->NumTexIndirections;
clone->NumNativeAluInstructions = prog->NumNativeAluInstructions;
clone->NumNativeTexInstructions = prog->NumNativeTexInstructions;
clone->NumNativeTexIndirections = prog->NumNativeTexIndirections;
switch (prog->Target) {
case GL_VERTEX_PROGRAM_ARB:
{
const struct gl_vertex_program *vp = gl_vertex_program_const(prog);
struct gl_vertex_program *vpc = gl_vertex_program(clone);
vpc->IsPositionInvariant = vp->IsPositionInvariant;
vpc->IsNVProgram = vp->IsNVProgram;
}
break;
case GL_FRAGMENT_PROGRAM_ARB:
{
const struct gl_fragment_program *fp = gl_fragment_program_const(prog);
struct gl_fragment_program *fpc = gl_fragment_program(clone);
fpc->UsesKill = fp->UsesKill;
fpc->UsesDFdy = fp->UsesDFdy;
fpc->OriginUpperLeft = fp->OriginUpperLeft;
fpc->PixelCenterInteger = fp->PixelCenterInteger;
}
break;
case MESA_GEOMETRY_PROGRAM:
{
const struct gl_geometry_program *gp = gl_geometry_program_const(prog);
struct gl_geometry_program *gpc = gl_geometry_program(clone);
gpc->VerticesOut = gp->VerticesOut;
gpc->InputType = gp->InputType;
gpc->OutputType = gp->OutputType;
}
break;
default:
_mesa_problem(NULL, "Unexpected target in _mesa_clone_program");
}
return clone;
}
/**
* Create a simple vertex shader that just passes through the
* vertex position and texcoord (and optionally, color).
*/
static struct st_vertex_program *
st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor)
{
GLcontext *ctx = st->ctx;
struct st_vertex_program *stvp;
struct gl_program *p;
GLuint ic = 0;
if (st->drawpix.vert_shaders[passColor])
return st->drawpix.vert_shaders[passColor];
/*
* Create shader now
*/
p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0);
if (!p)
return NULL;
if (passColor)
p->NumInstructions = 4;
else
p->NumInstructions = 3;
p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
if (!p->Instructions) {
ctx->Driver.DeleteProgram(ctx, p);
return NULL;
}
_mesa_init_instructions(p->Instructions, p->NumInstructions);
/* MOV result.pos, vertex.pos; */
p->Instructions[0].Opcode = OPCODE_MOV;
p->Instructions[0].DstReg.File = PROGRAM_OUTPUT;
p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS;
p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT;
p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS;
/* MOV result.texcoord0, vertex.texcoord0; */
p->Instructions[1].Opcode = OPCODE_MOV;
p->Instructions[1].DstReg.File = PROGRAM_OUTPUT;
p->Instructions[1].DstReg.Index = VERT_RESULT_TEX0;
p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT;
p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_TEX0;
ic = 2;
if (passColor) {
/* MOV result.color0, vertex.color0; */
p->Instructions[ic].Opcode = OPCODE_MOV;
p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
p->Instructions[ic].DstReg.Index = VERT_RESULT_COL0;
p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
p->Instructions[ic].SrcReg[0].Index = VERT_ATTRIB_COLOR0;
ic++;
}
/* END; */
p->Instructions[ic].Opcode = OPCODE_END;
ic++;
assert(ic == p->NumInstructions);
p->InputsRead = VERT_BIT_POS | VERT_BIT_TEX0;
p->OutputsWritten = ((1 << VERT_RESULT_TEX0) |
(1 << VERT_RESULT_HPOS));
if (passColor) {
p->InputsRead |= VERT_BIT_COLOR0;
p->OutputsWritten |= (1 << VERT_RESULT_COL0);
}
stvp = (struct st_vertex_program *) p;
st_translate_vertex_program(st, stvp, NULL, NULL, NULL);
st->drawpix.vert_shaders[passColor] = stvp;
return stvp;
}