/*
* Mesa 3-D graphics library
- * Version: 5.1
+ * Version: 6.1
*
- * Copyright (C) 1999-2003 Brian Paul All Rights Reserved.
+ * Copyright (C) 1999-2004 Brian Paul All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
+/*
+ * Regarding GL_NV_fragment_program:
+ *
+ * Portions of this software may use or implement intellectual
+ * property owned and licensed by NVIDIA Corporation. NVIDIA disclaims
+ * any and all warranties with respect to such intellectual property,
+ * including any use thereof or modifications thereto.
+ */
#include "glheader.h"
#include "colormac.h"
GLchan rgba[4];
SWcontext *swrast = SWRAST_CONTEXT(ctx);
+ /* XXX use a float-valued TextureSample routine here!!! */
swrast->TextureSample[unit](ctx, unit, ctx->Texture.Unit[unit]._Current,
1, (const GLfloat (*)[4]) texcoord,
&lambda, &rgba);
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
- const struct gl_texture_image *texImg = texObj->Image[texObj->BaseLevel];
+ const struct gl_texture_image *texImg = texObj->Image[0][texObj->BaseLevel];
const GLfloat texW = (GLfloat) texImg->WidthScale;
const GLfloat texH = (GLfloat) texImg->HeightScale;
GLchan rgba[4];
* need to execute another instance of the program (ugh)!
*/
static GLboolean
-fetch_vector4_deriv( const struct fp_src_register *source,
+fetch_vector4_deriv( GLcontext *ctx,
+ const struct fp_src_register *source,
const struct sw_span *span,
char xOrY, GLint column, GLfloat result[4] )
{
const GLboolean clamp = inst->Saturate;
const GLboolean updateCC = inst->UpdateCondRegister;
GLfloat *dstReg;
+ GLfloat dummyReg[4];
GLfloat clampedValue[4];
const GLboolean *writeMask = dest->WriteMask;
GLboolean condWriteMask[4];
case PROGRAM_TEMPORARY:
dstReg = machine->Temporaries[dest->Index];
break;
+ case PROGRAM_WRITE_ONLY:
+ dstReg = dummyReg;
+ return;
default:
_mesa_problem(NULL, "bad register file in store_vector4(fp)");
return;
/* copy existing machine */
_mesa_memcpy(dMachine, machine, sizeof(struct fp_machine));
- /* Clear temporary registers */
- _mesa_bzero( (void*) machine->Temporaries,
- MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
+ if (program->Base.Target == GL_FRAGMENT_PROGRAM_NV) {
+ /* Clear temporary registers (undefined for ARB_f_p) */
+ _mesa_bzero( (void*) machine->Temporaries,
+ MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
+ }
/* Add derivatives */
if (program->InputsRead & (1 << FRAG_ATTRIB_WPOS)) {
printf("execute fragment program --------------------\n");
#endif
- /* XXX: This should go someplace else, but it is safe here (and slow!)
- * - karl
- */
- _mesa_load_state_parameters(ctx, program->Parameters);
-
for (pc = 0; pc < maxInst; pc++) {
const struct fp_instruction *inst = program->Instructions + pc;
{
GLfloat a[4], aNext[4], result[4];
struct fp_machine dMachine;
- if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'X', column, result)) {
+ if (!fetch_vector4_deriv(ctx, &inst->SrcReg[0], span, 'X',
+ column, result)) {
/* This is tricky. Make a copy of the current machine state,
* increment the input registers by the dx or dy partial
* derivatives, then re-execute the program up to the
{
GLfloat a[4], aNext[4], result[4];
struct fp_machine dMachine;
- if (!fetch_vector4_deriv(&inst->SrcReg[0], span, 'Y', column, result)) {
+ if (!fetch_vector4_deriv(ctx, &inst->SrcReg[0], span, 'Y',
+ column, result)) {
init_machine_deriv(ctx, machine, program, span,
'Y', &dMachine);
fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a);
result[0] = result[1] = result[2] = result[3] =
a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3];
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
}
break;
case FP_OPCODE_DPH:
store_vector4( inst, machine, result );
}
break;
- case FP_OPCODE_KIL:
+ case FP_OPCODE_KIL_NV: /* NV_f_p only */
{
const GLuint *swizzle = inst->DstReg.CondSwizzle;
const GLuint condMask = inst->DstReg.CondMask;
}
}
break;
+ case FP_OPCODE_KIL: /* ARB_f_p only */
+ {
+ GLfloat a[4];
+ fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
+ if (a[0] < 0.0F || a[1] < 0.0F || a[2] < 0.0F || a[3] < 0.0F) {
+ return GL_FALSE;
+ }
+ }
+ break;
case FP_OPCODE_LG2: /* log base 2 */
{
GLfloat a[4], result[4];
a[1] = 0.0F;
result[0] = 1.0F;
result[1] = a[0];
- result[2] = (a[0] > 0.0F) ? (GLfloat)_mesa_pow(2.0, a[3]) : 0.0F;
+ result[2] = (a[0] > 0.0F) ? (GLfloat) exp(a[3] * log(a[1])) : 0.0F;
result[3] = 1.0F;
store_vector4( inst, machine, result );
}
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
}
break;
case FP_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
}
break;
case FP_OPCODE_MUL:
#endif
}
break;
- case FP_OPCODE_PK2H: /* pack two 16-bit floats */
- /* XXX this is probably wrong */
+ case FP_OPCODE_PK2H: /* pack two 16-bit floats in one 32-bit float */
{
GLfloat a[4], result[4];
- const GLuint *rawBits = (const GLuint *) a;
+ GLhalfNV hx, hy;
GLuint *rawResult = (GLuint *) result;
+ GLuint twoHalves;
fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
+ hx = _mesa_float_to_half(a[0]);
+ hy = _mesa_float_to_half(a[1]);
+ twoHalves = hx | (hy << 16);
rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
- = rawBits[0] | (rawBits[1] << 16);
+ = twoHalves;
store_vector4( inst, machine, result );
}
break;
- case FP_OPCODE_PK2US: /* pack two GLushorts */
+ case FP_OPCODE_PK2US: /* pack two GLushorts into one 32-bit float */
{
GLfloat a[4], result[4];
GLuint usx, usy, *rawResult = (GLuint *) result;
fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
a[0] = CLAMP(a[0], 0.0F, 1.0F);
- a[1] = CLAMP(a[0], 0.0F, 1.0F);
+ a[1] = CLAMP(a[1], 0.0F, 1.0F);
usx = IROUND(a[0] * 65535.0F);
usy = IROUND(a[1] * 65535.0F);
rawResult[0] = rawResult[1] = rawResult[2] = rawResult[3]
store_vector4( inst, machine, result );
}
break;
- case FP_OPCODE_PK4B: /* pack four GLbytes */
+ case FP_OPCODE_PK4B: /* pack four GLbytes into one 32-bit float */
{
GLfloat a[4], result[4];
GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
store_vector4( inst, machine, result );
}
break;
- case FP_OPCODE_PK4UB: /* pack four GLubytes */
+ case FP_OPCODE_PK4UB: /* pack four GLubytes into one 32-bit float */
{
GLfloat a[4], result[4];
GLuint ubx, uby, ubz, ubw, *rawResult = (GLuint *) result;
store_vector4( inst, machine, result );
}
break;
- case FP_OPCODE_TEX:
+ case FP_OPCODE_TEX: /* Both ARB and NV frag prog */
/* Texel lookup */
{
GLfloat texcoord[4], color[4];
fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
- /* XXX: Undo perspective divide from interpolate_texcoords() */
- fetch_texel( ctx, texcoord,
- span->array->lambda[inst->TexSrcUnit][column],
- inst->TexSrcUnit, color );
+ /* Note: we pass 0 for LOD. The ARB extension requires it
+ * while the NV extension says it's implementation dependant.
+ */
+ fetch_texel( ctx, texcoord, 0.0F, inst->TexSrcUnit, color );
store_vector4( inst, machine, color );
}
break;
- case FP_OPCODE_TXB:
+ case FP_OPCODE_TXB: /* GL_ARB_fragment_program only */
/* Texel lookup with LOD bias */
{
GLfloat texcoord[4], color[4], bias, lambda;
store_vector4( inst, machine, color );
}
break;
- case FP_OPCODE_TXD:
+ case FP_OPCODE_TXD: /* GL_NV_fragment_program only */
/* Texture lookup w/ partial derivatives for LOD */
{
GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
store_vector4( inst, machine, color );
}
break;
- case FP_OPCODE_TXP:
- /* Texture lookup w/ perspective divide */
+ case FP_OPCODE_TXP: /* GL_ARB_fragment_program only */
+ /* Texture lookup w/ projective divide */
+ {
+ GLfloat texcoord[4], color[4];
+ fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
+ texcoord[0] /= texcoord[3];
+ texcoord[1] /= texcoord[3];
+ texcoord[2] /= texcoord[3];
+ /* Note: LOD=0 */
+ fetch_texel( ctx, texcoord, 0.0F, inst->TexSrcUnit, color );
+ store_vector4( inst, machine, color );
+ }
+ break;
+ case FP_OPCODE_TXP_NV: /* GL_NV_fragment_program only */
+ /* Texture lookup w/ projective divide */
{
GLfloat texcoord[4], color[4];
fetch_vector4( ctx, &inst->SrcReg[0], machine, program, texcoord );
- /* Already did perspective divide in interpolate_texcoords() */
+ if (inst->TexSrcBit != TEXTURE_CUBE_BIT) {
+ texcoord[0] /= texcoord[3];
+ texcoord[1] /= texcoord[3];
+ texcoord[2] /= texcoord[3];
+ }
fetch_texel( ctx, texcoord,
span->array->lambda[inst->TexSrcUnit][column],
inst->TexSrcUnit, color );
}
break;
case FP_OPCODE_UP2H: /* unpack two 16-bit floats */
- /* XXX this is probably wrong */
{
GLfloat a[4], result[4];
const GLuint *rawBits = (const GLuint *) a;
- GLuint *rawResult = (GLuint *) result;
+ GLhalfNV hx, hy;
fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
- rawResult[0] = rawBits[0] & 0xffff;
- rawResult[1] = (rawBits[0] >> 16) & 0xffff;
- rawResult[2] = rawBits[0] & 0xffff;
- rawResult[3] = (rawBits[0] >> 16) & 0xffff;
+ hx = rawBits[0] & 0xffff;
+ hy = rawBits[0] >> 16;
+ result[0] = result[2] = _mesa_half_to_float(hx);
+ result[1] = result[3] = _mesa_half_to_float(hy);
store_vector4( inst, machine, result );
}
break;
{
GLfloat a[4], result[4];
const GLuint *rawBits = (const GLuint *) a;
+ GLushort usx, usy;
fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
- result[0] = (GLfloat) ((rawBits[0] >> 0) & 0xffff) / 65535.0F;
- result[1] = (GLfloat) ((rawBits[0] >> 16) & 0xffff) / 65535.0F;
- result[2] = result[0];
- result[3] = result[1];
+ usx = rawBits[0] & 0xffff;
+ usy = rawBits[0] >> 16;
+ result[0] = result[2] = usx * (1.0f / 65535.0f);
+ result[1] = result[3] = usy * (1.0f / 65535.0f);
store_vector4( inst, machine, result );
}
break;
const GLuint *rawBits = (const GLuint *) a;
fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
result[0] = (((rawBits[0] >> 0) & 0xff) - 128) / 127.0F;
- result[0] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
- result[0] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
- result[0] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
+ result[1] = (((rawBits[0] >> 8) & 0xff) - 128) / 127.0F;
+ result[2] = (((rawBits[0] >> 16) & 0xff) - 128) / 127.0F;
+ result[3] = (((rawBits[0] >> 24) & 0xff) - 128) / 127.0F;
store_vector4( inst, machine, result );
}
break;
const GLuint *rawBits = (const GLuint *) a;
fetch_vector1( ctx, &inst->SrcReg[0], machine, program, a );
result[0] = ((rawBits[0] >> 0) & 0xff) / 255.0F;
- result[0] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
- result[0] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
- result[0] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
+ result[1] = ((rawBits[0] >> 8) & 0xff) / 255.0F;
+ result[2] = ((rawBits[0] >> 16) & 0xff) / 255.0F;
+ result[3] = ((rawBits[0] >> 24) & 0xff) / 255.0F;
+ store_vector4( inst, machine, result );
+ }
+ break;
+ case FP_OPCODE_XPD: /* cross product */
+ {
+ GLfloat a[4], b[4], result[4];
+ fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
+ fetch_vector4( ctx, &inst->SrcReg[1], machine, program, b );
+ result[0] = a[1] * b[2] - a[2] * b[1];
+ result[1] = a[2] * b[0] - a[0] * b[2];
+ result[2] = a[0] * b[1] - a[1] * b[0];
+ result[3] = 1.0;
store_vector4( inst, machine, result );
}
break;
if (ctx->FragmentProgram.CallbackEnabled)
inputsRead = ~0;
- /* Clear temporary registers */
- _mesa_bzero(machine->Temporaries,
- MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
+ if (program->Base.Target == GL_FRAGMENT_PROGRAM_NV) {
+ /* Clear temporary registers (undefined for ARB_f_p) */
+ _mesa_bzero(machine->Temporaries,
+ MAX_NV_FRAGMENT_PROGRAM_TEMPS * 4 * sizeof(GLfloat));
+ }
/* Load input registers */
if (inputsRead & (1 << FRAG_ATTRIB_WPOS)) {
}
+
+/**
+ * Execute the current fragment program, operating on the given span.
+ */
void
-_swrast_exec_nv_fragment_program( GLcontext *ctx, struct sw_span *span )
+_swrast_exec_fragment_program( GLcontext *ctx, struct sw_span *span )
{
const struct fragment_program *program = ctx->FragmentProgram.Current;
GLuint i;
init_machine(ctx, &ctx->FragmentProgram.Machine,
ctx->FragmentProgram.Current, span, i);
+#ifdef USE_TCC
+ if (!_swrast_execute_codegen_program(ctx, program, ~0,
+ &ctx->FragmentProgram.Machine,
+ span, i)) {
+ span->array->mask[i] = GL_FALSE; /* killed fragment */
+ }
+#else
if (!execute_program(ctx, program, ~0,
&ctx->FragmentProgram.Machine, span, i)) {
span->array->mask[i] = GL_FALSE; /* killed fragment */
}
+#endif
/* Store output registers */
{