/* See comments below for info about this */
#define LAMBDA_ZERO 1
-/* if 1, print some debugging info */
+/* debug predicate */
#define DEBUG_FRAG 0
GLfloat Inputs[MAX_NV_FRAGMENT_PROGRAM_INPUTS][4];
GLfloat Outputs[MAX_NV_FRAGMENT_PROGRAM_OUTPUTS][4];
GLuint CondCodes[4]; /**< COND_* value for x/y/z/w */
+
+ GLuint CallStack[MAX_PROGRAM_CALL_DEPTH]; /**< For CAL/RET instructions */
+ GLuint StackDepth; /**< Index/ptr to top of CallStack[] */
};
return;
}
-#if DEBUG_FRAG
+#if 0
if (value[0] > 1.0e10 ||
IS_INF_OR_NAN(value[0]) ||
IS_INF_OR_NAN(value[1]) ||
{
GLuint pc;
-#if DEBUG_FRAG
- printf("execute fragment program --------------------\n");
-#endif
+ if (DEBUG_FRAG) {
+ printf("execute fragment program --------------------\n");
+ }
for (pc = 0; pc < maxInst; pc++) {
const struct prog_instruction *inst = program->Base.Instructions + pc;
ctx->FragmentProgram.CallbackData);
}
-#if DEBUG_FRAG
- _mesa_print_instruction(inst);
-#endif
+ if (DEBUG_FRAG) {
+ _mesa_print_instruction(inst);
+ }
+
switch (inst->Opcode) {
case OPCODE_ABS:
{
result[2] = a[2] + b[2];
result[3] = a[3] + b[3];
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
- result[0], result[1], result[2], result[3],
- a[0], a[1], a[2], a[3],
- b[0], b[1], b[2], b[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
+ result[0], result[1], result[2], result[3],
+ a[0], a[1], a[2], a[3],
+ b[0], b[1], b[2], b[3]);
+ }
+ }
+ break;
+ case OPCODE_BRA: /* conditional branch */
+ {
+ /* NOTE: The return is conditional! */
+ const GLuint swizzle = inst->DstReg.CondSwizzle;
+ const GLuint condMask = inst->DstReg.CondMask;
+ if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
+ /* take branch */
+ pc = inst->BranchTarget;
+ }
+ }
+ break;
+ case OPCODE_CAL: /* Call subroutine */
+ {
+ /* NOTE: The call is conditional! */
+ const GLuint swizzle = inst->DstReg.CondSwizzle;
+ const GLuint condMask = inst->DstReg.CondMask;
+ if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
+ if (machine->StackDepth >= MAX_PROGRAM_CALL_DEPTH) {
+ return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
+ }
+ machine->CallStack[machine->StackDepth++] = pc + 1;
+ pc = inst->BranchTarget;
+ }
}
break;
case OPCODE_CMP:
fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
result[0] = result[1] = result[2] = result[3] = DOT3(a, b);
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
- result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
+ result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
+ }
}
break;
case OPCODE_DP4:
fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
result[0] = result[1] = result[2] = result[3] = DOT4(a,b);
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
- result[0], a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
+ result[0], a[0], a[1], a[2], a[3],
+ b[0], b[1], b[2], b[3]);
+ }
}
break;
case OPCODE_DPH:
}
result[3] = 1.0F;
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
- result[0], result[1], result[2], result[3],
- a[0], a[1], a[2], a[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
+ result[0], result[1], result[2], result[3],
+ a[0], a[1], a[2], a[3]);
+ }
}
break;
case OPCODE_LRP:
result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("LRP (%g %g %g %g) = (%g %g %g %g), "
- "(%g %g %g %g), (%g %g %g %g)\n",
- result[0], result[1], result[2], result[3],
- a[0], a[1], a[2], a[3],
- b[0], b[1], b[2], b[3],
- c[0], c[1], c[2], c[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("LRP (%g %g %g %g) = (%g %g %g %g), "
+ "(%g %g %g %g), (%g %g %g %g)\n",
+ result[0], result[1], result[2], result[3],
+ a[0], a[1], a[2], a[3],
+ b[0], b[1], b[2], b[3],
+ c[0], c[1], c[2], c[3]);
+ }
}
break;
case OPCODE_MAD:
result[2] = a[2] * b[2] + c[2];
result[3] = a[3] * b[3] + c[3];
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
- "(%g %g %g %g) + (%g %g %g %g)\n",
- result[0], result[1], result[2], result[3],
- a[0], a[1], a[2], a[3],
- b[0], b[1], b[2], b[3],
- c[0], c[1], c[2], c[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
+ "(%g %g %g %g) + (%g %g %g %g)\n",
+ result[0], result[1], result[2], result[3],
+ a[0], a[1], a[2], a[3],
+ b[0], b[1], b[2], b[3],
+ c[0], c[1], c[2], c[3]);
+ }
}
break;
case OPCODE_MAX:
result[2] = MAX2(a[2], b[2]);
result[3] = MAX2(a[3], b[3]);
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
- result[0], result[1], result[2], result[3],
- a[0], a[1], a[2], a[3],
- b[0], b[1], b[2], b[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
+ result[0], result[1], result[2], result[3],
+ a[0], a[1], a[2], a[3],
+ b[0], b[1], b[2], b[3]);
+ }
}
break;
case OPCODE_MIN:
GLfloat result[4];
fetch_vector4( ctx, &inst->SrcReg[0], machine, program, result );
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("MOV (%g %g %g %g)\n",
- result[0], result[1], result[2], result[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("MOV (%g %g %g %g)\n",
+ result[0], result[1], result[2], result[3]);
+ }
}
break;
case OPCODE_MUL:
result[2] = a[2] * b[2];
result[3] = a[3] * b[3];
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
- result[0], result[1], result[2], result[3],
- a[0], a[1], a[2], a[3],
- b[0], b[1], b[2], b[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
+ result[0], result[1], result[2], result[3],
+ a[0], a[1], a[2], a[3],
+ b[0], b[1], b[2], b[3]);
+ }
}
break;
case OPCODE_PK2H: /* pack two 16-bit floats in one 32-bit float */
{
GLfloat a[4], result[4];
fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
-#if DEBUG_FRAG
- if (a[0] == 0)
- printf("RCP(0)\n");
- else if (IS_INF_OR_NAN(a[0]))
- printf("RCP(inf)\n");
-#endif
+ if (DEBUG_FRAG) {
+ if (a[0] == 0)
+ printf("RCP(0)\n");
+ else if (IS_INF_OR_NAN(a[0]))
+ printf("RCP(inf)\n");
+ }
result[0] = result[1] = result[2] = result[3] = 1.0F / a[0];
store_vector4( inst, machine, result );
}
break;
+ case OPCODE_RET: /* return from subroutine */
+ {
+ /* NOTE: The return is conditional! */
+ const GLuint swizzle = inst->DstReg.CondSwizzle;
+ const GLuint condMask = inst->DstReg.CondMask;
+ if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
+ test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
+ if (machine->StackDepth == 0) {
+ return GL_TRUE; /* Per GL_NV_vertex_program2 spec */
+ }
+ pc = machine->CallStack[--machine->StackDepth];
+ }
+ }
+ break;
case OPCODE_RFL: /* reflection vector */
{
GLfloat axis[4], dir[4], result[4], tmpX, tmpW;
a[0] = FABSF(a[0]);
result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
+ }
}
break;
case OPCODE_SCS: /* sine and cos */
result[2] = a[2] - b[2];
result[3] = a[3] - b[3];
store_vector4( inst, machine, result );
-#if DEBUG_FRAG
- printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
- result[0], result[1], result[2], result[3],
- a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
-#endif
+ if (DEBUG_FRAG) {
+ printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
+ result[0], result[1], result[2], result[3],
+ a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
+ }
}
break;
case OPCODE_SWZ: /* extended swizzle */
case OPCODE_TEX: /* Both ARB and NV frag prog */
/* Texel lookup */
{
- /* Note: we're using zero instead of lambda for the LOD.
- * The problem is we didn't necessarily use the right partial
- * derivatives when we called _swrast_compute_lambda() earlier.
- * A texture can be sampled with any coordinate, not just
- * fragment.texcoord[n].
- * Use zero for now because that's better than using a bad
- * lambda value (I've seen a few test programs fail otherwise).
- * We need to overhaul this stuff someday.
+ /* Note: only use the precomputed lambda value when we're
+ * sampling texture unit [K] with texcoord[K].
+ * Otherwise, the lambda value may have no relation to the
+ * instruction's texcoord or texture image. Using the wrong
+ * lambda is usually bad news.
+ * The rest of the time, just use zero (until we get a more
+ * sophisticated way of computing lambda).
*/
-#ifdef LAMBDA_ZERO
- GLfloat lambda = 0.0;
-#else
- GLfloat lambda = span->array->lambda[inst->TexSrcUnit][column];
-#endif
- GLfloat coord[4], color[4];
+ GLfloat coord[4], color[4], lambda;
+ if (inst->SrcReg[0].File == PROGRAM_INPUT &&
+ inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0+inst->TexSrcUnit)
+ lambda = span->array->lambda[inst->TexSrcUnit][column];
+ else
+ lambda = 0.0;
fetch_vector4(ctx, &inst->SrcReg[0], machine, program, coord);
fetch_texel( ctx, coord, lambda, inst->TexSrcUnit, color );
-#if DEBUG_FRAG
- printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g], "
- "lod %f\n",
- color[0], color[1], color[2], color[3], inst->TexSrcUnit,
- coord[0], coord[1], coord[2], coord[3], lambda);
-#endif
+ if (DEBUG_FRAG) {
+ printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g], "
+ "lod %f\n",
+ color[0], color[1], color[2], color[3],
+ inst->TexSrcUnit,
+ coord[0], coord[1], coord[2], coord[3], lambda);
+ }
store_vector4( inst, machine, color );
}
break;
case OPCODE_TXB: /* GL_ARB_fragment_program only */
/* Texel lookup with LOD bias */
{
-#ifdef LAMBDA_ZERO
- GLfloat lambda = 0.0;
-#else
- GLfloat lambda = span->array->lambda[inst->TexSrcUnit][column];
-#endif
- GLfloat coord[4], color[4], bias;
+ GLfloat coord[4], color[4], lambda, bias;
+ if (inst->SrcReg[0].File == PROGRAM_INPUT &&
+ inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0+inst->TexSrcUnit)
+ lambda = span->array->lambda[inst->TexSrcUnit][column];
+ else
+ lambda = 0.0;
fetch_vector4(ctx, &inst->SrcReg[0], machine, program, coord);
/* coord[3] is the bias to add to lambda */
bias = ctx->Texture.Unit[inst->TexSrcUnit].LodBias
case OPCODE_TXP: /* GL_ARB_fragment_program only */
/* Texture lookup w/ projective divide */
{
-#ifdef LAMBDA_ZERO
- GLfloat lambda = 0.0;
-#else
- GLfloat lambda = span->array->lambda[inst->TexSrcUnit][column];
-#endif
- GLfloat texcoord[4], color[4];
+ GLfloat texcoord[4], color[4], lambda;
+ if (inst->SrcReg[0].File == PROGRAM_INPUT &&
+ inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0+inst->TexSrcUnit)
+ lambda = span->array->lambda[inst->TexSrcUnit][column];
+ else
+ lambda = 0.0;
fetch_vector4(ctx, &inst->SrcReg[0], machine, program,texcoord);
/* Not so sure about this test - if texcoord[3] is
* zero, we'd probably be fine except for an ASSERT in
case OPCODE_TXP_NV: /* GL_NV_fragment_program only */
/* Texture lookup w/ projective divide */
{
-#ifdef LAMBDA_ZERO
- GLfloat lambda = 0.0;
-#else
- GLfloat lambda = span->array->lambda[inst->TexSrcUnit][column];
-#endif
- GLfloat texcoord[4], color[4];
- fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
+ GLfloat texcoord[4], color[4], lambda;
+ if (inst->SrcReg[0].File == PROGRAM_INPUT &&
+ inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0+inst->TexSrcUnit)
+ lambda = span->array->lambda[inst->TexSrcUnit][column];
+ else
+ lambda = 0.0;
+ fetch_vector4(ctx, &inst->SrcReg[0], machine, program,texcoord);
if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX &&
texcoord[3] != 0.0) {
texcoord[0] /= texcoord[3];
machine->CondCodes[1] = COND_EQ;
machine->CondCodes[2] = COND_EQ;
machine->CondCodes[3] = COND_EQ;
+
+ /* init call stack */
+ machine->StackDepth = 0;
}
ctx->_CurrentProgram = GL_FRAGMENT_PROGRAM_ARB; /* or NV, doesn't matter */
-#if 1 /* we really shouldn't need this here... */
- if (program->Base.Parameters) {
- _mesa_load_state_parameters(ctx, program->Base.Parameters);
- }
-#endif
run_program(ctx, span, 0, span->end);
if (program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) {
projects / mesa.git / blobdiff