p->error = 1;
}
-
+static uint get_mapping(struct i915_fragment_shader* fs, int unit)
+{
+ int i;
+ for (i = 0; i < I915_TEX_UNITS; i++)
+ {
+ if (fs->generic_mapping[i] == -1) {
+ fs->generic_mapping[i] = unit;
+ return i;
+ }
+ if (fs->generic_mapping[i] == unit)
+ return i;
+ }
+ debug_printf("Exceeded max generics\n");
+ return 0;
+}
/**
* Construct a ureg for the given source register. Will emit
*/
static uint
src_vector(struct i915_fp_compile *p,
- const struct tgsi_full_src_register *source)
+ const struct tgsi_full_src_register *source,
+ struct i915_fragment_shader* fs)
{
- uint index = source->SrcRegister.Index;
+ uint index = source->Register.Index;
uint src = 0, sem_name, sem_ind;
- switch (source->SrcRegister.File) {
+ switch (source->Register.File) {
case TGSI_FILE_TEMPORARY:
- if (source->SrcRegister.Index >= I915_MAX_TEMPORARY) {
+ if (source->Register.Index >= I915_MAX_TEMPORARY) {
i915_program_error(p, "Exceeded max temporary reg");
return 0;
}
switch (sem_name) {
case TGSI_SEMANTIC_POSITION:
- debug_printf("SKIP SEM POS\n");
- /*
- assert(p->wpos_tex != -1);
- src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL);
- */
- break;
+ {
+ /* for fragcoord */
+ int real_tex_unit = get_mapping(fs, I915_SEMANTIC_POS);
+ src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + real_tex_unit, D0_CHANNEL_ALL);
+ break;
+ }
case TGSI_SEMANTIC_COLOR:
if (sem_ind == 0) {
src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL);
src = swizzle(src, W, W, W, W);
break;
case TGSI_SEMANTIC_GENERIC:
- /* usually a texcoord */
- src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + sem_ind, D0_CHANNEL_ALL);
- break;
+ {
+ int real_tex_unit = get_mapping(fs, sem_ind);
+ src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + real_tex_unit, D0_CHANNEL_ALL);
+ break;
+ }
default:
i915_program_error(p, "Bad source->Index");
return 0;
}
src = swizzle(src,
- source->SrcRegister.SwizzleX,
- source->SrcRegister.SwizzleY,
- source->SrcRegister.SwizzleZ,
- source->SrcRegister.SwizzleW);
+ source->Register.SwizzleX,
+ source->Register.SwizzleY,
+ source->Register.SwizzleZ,
+ source->Register.SwizzleW);
/* There's both negate-all-components and per-component negation.
* Try to handle both here.
*/
{
- int nx = source->SrcRegisterExtSwz.NegateX;
- int ny = source->SrcRegisterExtSwz.NegateY;
- int nz = source->SrcRegisterExtSwz.NegateZ;
- int nw = source->SrcRegisterExtSwz.NegateW;
- if (source->SrcRegister.Negate) {
- nx = !nx;
- ny = !ny;
- nz = !nz;
- nw = !nw;
- }
- src = negate(src, nx, ny, nz, nw);
+ int n = source->Register.Negate;
+ src = negate(src, n, n, n, n);
}
- /* no abs() or post-abs negation */
+ /* no abs() */
#if 0
/* XXX assertions disabled to allow arbfplight.c to run */
/* XXX enable these assertions, or fix things */
- assert(!source->SrcRegisterExtMod.Absolute);
- assert(!source->SrcRegisterExtMod.Negate);
+ assert(!source->Register.Absolute);
#endif
return src;
}
get_result_vector(struct i915_fp_compile *p,
const struct tgsi_full_dst_register *dest)
{
- switch (dest->DstRegister.File) {
+ switch (dest->Register.File) {
case TGSI_FILE_OUTPUT:
{
- uint sem_name = p->shader->info.output_semantic_name[dest->DstRegister.Index];
+ uint sem_name = p->shader->info.output_semantic_name[dest->Register.Index];
switch (sem_name) {
case TGSI_SEMANTIC_POSITION:
return UREG(REG_TYPE_OD, 0);
}
}
case TGSI_FILE_TEMPORARY:
- return UREG(REG_TYPE_R, dest->DstRegister.Index);
+ return UREG(REG_TYPE_R, dest->Register.Index);
default:
i915_program_error(p, "Bad inst->DstReg.File");
return 0;
get_result_flags(const struct tgsi_full_instruction *inst)
{
const uint writeMask
- = inst->FullDstRegisters[0].DstRegister.WriteMask;
+ = inst->Dst[0].Register.WriteMask;
uint flags = 0x0;
if (inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE)
static void
emit_tex(struct i915_fp_compile *p,
const struct tgsi_full_instruction *inst,
- uint opcode)
+ uint opcode,
+ struct i915_fragment_shader* fs)
{
- uint texture = inst->InstructionExtTexture.Texture;
- uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+ uint texture = inst->Texture.Texture;
+ uint unit = inst->Src[1].Register.Index;
uint tex = translate_tex_src_target( p, texture );
uint sampler = i915_emit_decl(p, REG_TYPE_S, unit, tex);
- uint coord = src_vector( p, &inst->FullSrcRegisters[0]);
+ uint coord = src_vector( p, &inst->Src[0], fs);
i915_emit_texld( p,
- get_result_vector( p, &inst->FullDstRegisters[0] ),
+ get_result_vector( p, &inst->Dst[0] ),
get_result_flags( inst ),
sampler,
coord,
static void
emit_simple_arith(struct i915_fp_compile *p,
const struct tgsi_full_instruction *inst,
- uint opcode, uint numArgs)
+ uint opcode, uint numArgs,
+ struct i915_fragment_shader* fs)
{
uint arg1, arg2, arg3;
assert(numArgs <= 3);
- arg1 = (numArgs < 1) ? 0 : src_vector( p, &inst->FullSrcRegisters[0] );
- arg2 = (numArgs < 2) ? 0 : src_vector( p, &inst->FullSrcRegisters[1] );
- arg3 = (numArgs < 3) ? 0 : src_vector( p, &inst->FullSrcRegisters[2] );
+ arg1 = (numArgs < 1) ? 0 : src_vector( p, &inst->Src[0], fs );
+ arg2 = (numArgs < 2) ? 0 : src_vector( p, &inst->Src[1], fs );
+ arg3 = (numArgs < 3) ? 0 : src_vector( p, &inst->Src[2], fs );
i915_emit_arith( p,
opcode,
- get_result_vector( p, &inst->FullDstRegisters[0]),
+ get_result_vector( p, &inst->Dst[0]),
get_result_flags( inst ), 0,
arg1,
arg2,
static void
emit_simple_arith_swap2(struct i915_fp_compile *p,
const struct tgsi_full_instruction *inst,
- uint opcode, uint numArgs)
+ uint opcode, uint numArgs,
+ struct i915_fragment_shader* fs)
{
struct tgsi_full_instruction inst2;
/* transpose first two registers */
inst2 = *inst;
- inst2.FullSrcRegisters[0] = inst->FullSrcRegisters[1];
- inst2.FullSrcRegisters[1] = inst->FullSrcRegisters[0];
+ inst2.Src[0] = inst->Src[1];
+ inst2.Src[1] = inst->Src[0];
- emit_simple_arith(p, &inst2, opcode, numArgs);
+ emit_simple_arith(p, &inst2, opcode, numArgs, fs);
}
*/
static void
i915_translate_instruction(struct i915_fp_compile *p,
- const struct tgsi_full_instruction *inst)
+ const struct tgsi_full_instruction *inst,
+ struct i915_fragment_shader *fs)
{
uint writemask;
uint src0, src1, src2, flags;
switch (inst->Instruction.Opcode) {
case TGSI_OPCODE_ABS:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_MAX,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
src0, negate(src0, 1, 1, 1, 1), 0);
break;
case TGSI_OPCODE_ADD:
- emit_simple_arith(p, inst, A0_ADD, 2);
+ emit_simple_arith(p, inst, A0_ADD, 2, fs);
break;
case TGSI_OPCODE_CMP:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
- src1 = src_vector(p, &inst->FullSrcRegisters[1]);
- src2 = src_vector(p, &inst->FullSrcRegisters[2]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
+ src2 = src_vector(p, &inst->Src[2], fs);
i915_emit_arith(p, A0_CMP,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst),
0, src0, src2, src1); /* NOTE: order of src2, src1 */
break;
case TGSI_OPCODE_COS:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
i915_emit_arith(p,
i915_emit_arith(p,
A0_DP4,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
swizzle(tmp, ONE, Z, Y, X),
i915_emit_const4fv(p, cos_constants), 0);
break;
+ case TGSI_OPCODE_DP2:
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
+
+ i915_emit_arith(p,
+ A0_DP3,
+ get_result_vector(p, &inst->Dst[0]),
+ get_result_flags(inst), 0,
+ swizzle(src0, X, Y, ZERO, ZERO), src1, 0);
+ break;
+
case TGSI_OPCODE_DP3:
- emit_simple_arith(p, inst, A0_DP3, 2);
+ emit_simple_arith(p, inst, A0_DP3, 2, fs);
break;
case TGSI_OPCODE_DP4:
- emit_simple_arith(p, inst, A0_DP4, 2);
+ emit_simple_arith(p, inst, A0_DP4, 2, fs);
break;
case TGSI_OPCODE_DPH:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
- src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
i915_emit_arith(p,
A0_DP4,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
swizzle(src0, X, Y, Z, ONE), src1, 0);
break;
case TGSI_OPCODE_DST:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
- src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
/* result[0] = 1 * 1;
* result[1] = a[1] * b[1];
*/
i915_emit_arith(p,
A0_MUL,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
swizzle(src0, ONE, Y, Z, ONE),
swizzle(src1, ONE, Y, ONE, W), 0);
break;
case TGSI_OPCODE_EX2:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_EXP,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
swizzle(src0, X, X, X, X), 0, 0);
break;
case TGSI_OPCODE_FLR:
- emit_simple_arith(p, inst, A0_FLR, 1);
+ emit_simple_arith(p, inst, A0_FLR, 1, fs);
break;
case TGSI_OPCODE_FRC:
- emit_simple_arith(p, inst, A0_FRC, 1);
+ emit_simple_arith(p, inst, A0_FRC, 1, fs);
break;
case TGSI_OPCODE_KIL:
/* kill if src[0].x < 0 || src[0].y < 0 ... */
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
i915_emit_texld(p,
break;
case TGSI_OPCODE_LG2:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_LOG,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
swizzle(src0, X, X, X, X), 0, 0);
break;
case TGSI_OPCODE_LIT:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
/* tmp = max( a.xyzw, a.00zw )
swizzle(tmp, Y, Y, Y, Y), 0, 0);
i915_emit_arith(p, A0_CMP,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
negate(swizzle(tmp, ONE, ONE, X, ONE), 0, 0, 1, 0),
swizzle(tmp, ONE, X, ZERO, ONE),
break;
case TGSI_OPCODE_LRP:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
- src1 = src_vector(p, &inst->FullSrcRegisters[1]);
- src2 = src_vector(p, &inst->FullSrcRegisters[2]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
+ src2 = src_vector(p, &inst->Src[2], fs);
flags = get_result_flags(inst);
tmp = i915_get_utemp(p);
flags & A0_DEST_CHANNEL_ALL, 0, src1, src0, src2);
i915_emit_arith(p, A0_MAD,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
flags, 0, negate(src2, 1, 1, 1, 1), src0, tmp);
break;
case TGSI_OPCODE_MAD:
- emit_simple_arith(p, inst, A0_MAD, 3);
+ emit_simple_arith(p, inst, A0_MAD, 3, fs);
break;
case TGSI_OPCODE_MAX:
- emit_simple_arith(p, inst, A0_MAX, 2);
+ emit_simple_arith(p, inst, A0_MAX, 2, fs);
break;
case TGSI_OPCODE_MIN:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
- src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
tmp = i915_get_utemp(p);
flags = get_result_flags(inst);
i915_emit_arith(p,
A0_MOV,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
flags, 0, negate(tmp, 1, 1, 1, 1), 0, 0);
break;
case TGSI_OPCODE_MOV:
- emit_simple_arith(p, inst, A0_MOV, 1);
+ emit_simple_arith(p, inst, A0_MOV, 1, fs);
break;
case TGSI_OPCODE_MUL:
- emit_simple_arith(p, inst, A0_MUL, 2);
+ emit_simple_arith(p, inst, A0_MUL, 2, fs);
break;
case TGSI_OPCODE_POW:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
- src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
tmp = i915_get_utemp(p);
flags = get_result_flags(inst);
i915_emit_arith(p,
A0_EXP,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
flags, 0, swizzle(tmp, X, X, X, X), 0, 0);
break;
break;
case TGSI_OPCODE_RCP:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_RCP,
- get_result_vector(p, &inst->FullDstRegisters[0]),
- get_result_flags(inst), 0,
+ get_result_vector(p, &inst->Dst[0]),
+ get_result_flags(inst), 0,
swizzle(src0, X, X, X, X), 0, 0);
break;
case TGSI_OPCODE_RSQ:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_RSQ,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
swizzle(src0, X, X, X, X), 0, 0);
break;
case TGSI_OPCODE_SCS:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
/*
swizzle(tmp, X, Y, X, Y),
swizzle(tmp, X, X, ONE, ONE), 0);
- writemask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+ writemask = inst->Dst[0].Register.WriteMask;
if (writemask & TGSI_WRITEMASK_Y) {
uint tmp1;
i915_emit_arith(p,
A0_DP4,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
A0_DEST_CHANNEL_Y, 0,
swizzle(tmp1, W, Z, Y, X),
i915_emit_const4fv(p, sin_constants), 0);
i915_emit_arith(p,
A0_DP4,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
A0_DEST_CHANNEL_X, 0,
swizzle(tmp, ONE, Z, Y, X),
i915_emit_const4fv(p, cos_constants), 0);
}
break;
- case TGSI_OPCODE_SGE:
- emit_simple_arith(p, inst, A0_SGE, 2);
+ case TGSI_OPCODE_SEQ:
+ /* if we're both >= and <= then we're == */
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
+ tmp = i915_get_utemp(p);
+
+ i915_emit_arith(p,
+ A0_SGE,
+ tmp, A0_DEST_CHANNEL_ALL, 0,
+ src0,
+ src1, 0);
+
+ i915_emit_arith(p,
+ A0_SGE,
+ get_result_vector(p, &inst->Dst[0]),
+ A0_DEST_CHANNEL_ALL, 0,
+ src1,
+ src0, 0);
+
+ i915_emit_arith(p,
+ A0_MUL,
+ get_result_vector(p, &inst->Dst[0]),
+ A0_DEST_CHANNEL_ALL, 0,
+ get_result_vector(p, &inst->Dst[0]),
+ tmp, 0);
+
break;
- case TGSI_OPCODE_SLE:
- /* like SGE, but swap reg0, reg1 */
- emit_simple_arith_swap2(p, inst, A0_SGE, 2);
+ case TGSI_OPCODE_SGE:
+ emit_simple_arith(p, inst, A0_SGE, 2, fs);
break;
case TGSI_OPCODE_SIN:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
i915_emit_arith(p,
i915_emit_arith(p,
A0_DP4,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
swizzle(tmp, W, Z, Y, X),
i915_emit_const4fv(p, sin_constants), 0);
break;
+ case TGSI_OPCODE_SLE:
+ /* like SGE, but swap reg0, reg1 */
+ emit_simple_arith_swap2(p, inst, A0_SGE, 2, fs);
+ break;
+
case TGSI_OPCODE_SLT:
- emit_simple_arith(p, inst, A0_SLT, 2);
+ emit_simple_arith(p, inst, A0_SLT, 2, fs);
break;
case TGSI_OPCODE_SGT:
/* like SLT, but swap reg0, reg1 */
- emit_simple_arith_swap2(p, inst, A0_SLT, 2);
+ emit_simple_arith_swap2(p, inst, A0_SLT, 2, fs);
+ break;
+
+ case TGSI_OPCODE_SNE:
+ /* if we're < or > then we're != */
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
+ tmp = i915_get_utemp(p);
+
+ i915_emit_arith(p,
+ A0_SLT,
+ tmp,
+ A0_DEST_CHANNEL_ALL, 0,
+ src0,
+ src1, 0);
+
+ i915_emit_arith(p,
+ A0_SLT,
+ get_result_vector(p, &inst->Dst[0]),
+ A0_DEST_CHANNEL_ALL, 0,
+ src1,
+ src0, 0);
+
+ i915_emit_arith(p,
+ A0_ADD,
+ get_result_vector(p, &inst->Dst[0]),
+ A0_DEST_CHANNEL_ALL, 0,
+ get_result_vector(p, &inst->Dst[0]),
+ tmp, 0);
+ break;
+
+ case TGSI_OPCODE_SSG:
+ /* compute (src>0) - (src<0) */
+ src0 = src_vector(p, &inst->Src[0], fs);
+ tmp = i915_get_utemp(p);
+
+ i915_emit_arith(p,
+ A0_SLT,
+ tmp,
+ A0_DEST_CHANNEL_ALL, 0,
+ src0,
+ swizzle(src0, ZERO, ZERO, ZERO, ZERO), 0);
+
+ i915_emit_arith(p,
+ A0_SLT,
+ get_result_vector(p, &inst->Dst[0]),
+ A0_DEST_CHANNEL_ALL, 0,
+ swizzle(src0, ZERO, ZERO, ZERO, ZERO),
+ src0, 0);
+
+ i915_emit_arith(p,
+ A0_ADD,
+ get_result_vector(p, &inst->Dst[0]),
+ A0_DEST_CHANNEL_ALL, 0,
+ get_result_vector(p, &inst->Dst[0]),
+ negate(tmp, 1, 1, 1, 1), 0);
break;
case TGSI_OPCODE_SUB:
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
- src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
i915_emit_arith(p,
A0_ADD,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
src0, negate(src1, 1, 1, 1, 1), 0);
break;
case TGSI_OPCODE_TEX:
- emit_tex(p, inst, T0_TEXLD);
+ emit_tex(p, inst, T0_TEXLD, fs);
+ break;
+
+ case TGSI_OPCODE_TRUNC:
+ emit_simple_arith(p, inst, A0_TRC, 1, fs);
break;
case TGSI_OPCODE_TXB:
- emit_tex(p, inst, T0_TEXLDB);
+ emit_tex(p, inst, T0_TEXLDB, fs);
break;
case TGSI_OPCODE_TXP:
- emit_tex(p, inst, T0_TEXLDP);
+ emit_tex(p, inst, T0_TEXLDP, fs);
break;
case TGSI_OPCODE_XPD:
* result.z = src0.x * src1.y - src0.y * src1.x;
* result.w = undef;
*/
- src0 = src_vector(p, &inst->FullSrcRegisters[0]);
- src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
tmp = i915_get_utemp(p);
i915_emit_arith(p,
i915_emit_arith(p,
A0_MAD,
- get_result_vector(p, &inst->FullDstRegisters[0]),
+ get_result_vector(p, &inst->Dst[0]),
get_result_flags(inst), 0,
swizzle(src0, Y, Z, X, ONE),
swizzle(src1, Z, X, Y, ONE),
*/
static void
i915_translate_instructions(struct i915_fp_compile *p,
- const struct tgsi_token *tokens)
+ const struct tgsi_token *tokens,
+ struct i915_fragment_shader *fs)
{
struct i915_fragment_shader *ifs = p->shader;
struct tgsi_parse_context parse;
tgsi_parse_token( &parse );
switch( parse.FullToken.Token.Type ) {
+ case TGSI_TOKEN_TYPE_PROPERTY:
+ /*
+ * We only support one cbuf, but we still need to ignore the property
+ * correctly so we don't hit the assert at the end of the switch case.
+ */
+ assert(parse.FullToken.FullProperty.Property.PropertyName ==
+ TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS);
+ break;
case TGSI_TOKEN_TYPE_DECLARATION:
if (parse.FullToken.FullDeclaration.Declaration.File
== TGSI_FILE_CONSTANT) {
uint i;
- for (i = parse.FullToken.FullDeclaration.DeclarationRange.First;
- i <= parse.FullToken.FullDeclaration.DeclarationRange.Last;
+ for (i = parse.FullToken.FullDeclaration.Range.First;
+ i <= parse.FullToken.FullDeclaration.Range.Last;
i++) {
assert(ifs->constant_flags[i] == 0x0);
ifs->constant_flags[i] = I915_CONSTFLAG_USER;
else if (parse.FullToken.FullDeclaration.Declaration.File
== TGSI_FILE_TEMPORARY) {
uint i;
- for (i = parse.FullToken.FullDeclaration.DeclarationRange.First;
- i <= parse.FullToken.FullDeclaration.DeclarationRange.Last;
+ for (i = parse.FullToken.FullDeclaration.Range.First;
+ i <= parse.FullToken.FullDeclaration.Range.Last;
i++) {
assert(i < I915_MAX_TEMPORARY);
/* XXX just use shader->info->file_mask[TGSI_FILE_TEMPORARY] */
p->first_instruction = FALSE;
}
- i915_translate_instruction(p, &parse.FullToken.FullInstruction);
+ i915_translate_instruction(p, &parse.FullToken.FullInstruction, fs);
break;
default:
struct i915_fragment_shader *ifs)
{
struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile);
+ int i;
p->shader = ifs;
ifs->num_constants = 0;
memset(ifs->constant_flags, 0, sizeof(ifs->constant_flags));
+ for (i = 0; i < I915_TEX_UNITS; i++)
+ ifs->generic_mapping[i] = -1;
+
p->first_instruction = TRUE;
p->nr_tex_indirect = 1; /* correct? */
p->temp_flag = ~0x0 << I915_MAX_TEMPORARY;
p->utemp_flag = ~0x7;
- p->wpos_tex = -1;
-
/* initialize the first program word */
*(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM;
}
-/**
- * Find an unused texture coordinate slot to use for fragment WPOS.
- * Update p->fp->wpos_tex with the result (-1 if no used texcoord slot is found).
- */
-static void
-i915_find_wpos_space(struct i915_fp_compile *p)
-{
-#if 0
- const uint inputs
- = p->shader->inputs_read | (1 << TGSI_ATTRIB_POS); /*XXX hack*/
- uint i;
-
- p->wpos_tex = -1;
-
- if (inputs & (1 << TGSI_ATTRIB_POS)) {
- for (i = 0; i < I915_TEX_UNITS; i++) {
- if ((inputs & (1 << (TGSI_ATTRIB_TEX0 + i))) == 0) {
- p->wpos_tex = i;
- return;
- }
- }
-
- i915_program_error(p, "No free texcoord for wpos value");
- }
-#else
- if (p->shader->info.input_semantic_name[0] == TGSI_SEMANTIC_POSITION) {
- /* frag shader using the fragment position input */
-#if 0
- assert(0);
-#endif
- }
-#endif
-}
-
i915_translate_fragment_program( struct i915_context *i915,
struct i915_fragment_shader *fs)
{
- struct i915_fp_compile *p = i915_init_compile(i915, fs);
+ struct i915_fp_compile *p;
const struct tgsi_token *tokens = fs->state.tokens;
- i915_find_wpos_space(p);
-
#if 0
tgsi_dump(tokens, 0);
#endif
- i915_translate_instructions(p, tokens);
+ /* hw doesn't seem to like empty frag programs, even when the depth write
+ * fixup gets emitted below - may that one is fishy, too? */
+ if (fs->info.num_instructions == 1) {
+ i915_use_passthrough_shader(fs);
+
+ return;
+ }
+
+ p = i915_init_compile(i915, fs);
+
+ i915_translate_instructions(p, tokens, fs);
i915_fixup_depth_write(p);
i915_fini_compile(i915, p);