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->Register.Index;
uint src = 0, sem_name, sem_ind;
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:
- if (sem_ind < 8)
- /* a texcoord */
- src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + sem_ind, D0_CHANNEL_ALL);
- else if ( (sem_ind >= 10) && (sem_ind < 18) )
- /* a varying */
- src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + sem_ind - 10, D0_CHANNEL_ALL);
- else
- debug_printf("%s: unhandled generic %d\n", __func__, sem_ind);
- 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;
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->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->Src[0]);
+ uint coord = src_vector( p, &inst->Src[0], fs);
i915_emit_texld( p,
get_result_vector( p, &inst->Dst[0] ),
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->Src[0] );
- arg2 = (numArgs < 2) ? 0 : src_vector( p, &inst->Src[1] );
- arg3 = (numArgs < 3) ? 0 : src_vector( p, &inst->Src[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,
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;
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->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_MAX,
get_result_vector(p, &inst->Dst[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->Src[0]);
- src1 = src_vector(p, &inst->Src[1]);
- src2 = src_vector(p, &inst->Src[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->Dst[0]),
get_result_flags(inst),
break;
case TGSI_OPCODE_COS:
- src0 = src_vector(p, &inst->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
i915_emit_arith(p,
break;
case TGSI_OPCODE_DP2:
- src0 = src_vector(p, &inst->Src[0]);
- src1 = src_vector(p, &inst->Src[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
i915_emit_arith(p,
A0_DP3,
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->Src[0]);
- src1 = src_vector(p, &inst->Src[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
i915_emit_arith(p,
A0_DP4,
break;
case TGSI_OPCODE_DST:
- src0 = src_vector(p, &inst->Src[0]);
- src1 = src_vector(p, &inst->Src[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];
break;
case TGSI_OPCODE_EX2:
- src0 = src_vector(p, &inst->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_EXP,
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->Src[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->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_LOG,
break;
case TGSI_OPCODE_LIT:
- src0 = src_vector(p, &inst->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
/* tmp = max( a.xyzw, a.00zw )
break;
case TGSI_OPCODE_LRP:
- src0 = src_vector(p, &inst->Src[0]);
- src1 = src_vector(p, &inst->Src[1]);
- src2 = src_vector(p, &inst->Src[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);
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->Src[0]);
- src1 = src_vector(p, &inst->Src[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);
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->Src[0]);
- src1 = src_vector(p, &inst->Src[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);
break;
case TGSI_OPCODE_RCP:
- src0 = src_vector(p, &inst->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_RCP,
break;
case TGSI_OPCODE_RSQ:
- src0 = src_vector(p, &inst->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
i915_emit_arith(p,
A0_RSQ,
break;
case TGSI_OPCODE_SCS:
- src0 = src_vector(p, &inst->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
/*
case TGSI_OPCODE_SEQ:
/* if we're both >= and <= then we're == */
- src0 = src_vector(p, &inst->Src[0]);
- src1 = src_vector(p, &inst->Src[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,
break;
case TGSI_OPCODE_SGE:
- emit_simple_arith(p, inst, A0_SGE, 2);
+ emit_simple_arith(p, inst, A0_SGE, 2, fs);
break;
case TGSI_OPCODE_SIN:
- src0 = src_vector(p, &inst->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
i915_emit_arith(p,
case TGSI_OPCODE_SLE:
/* like SGE, but swap reg0, reg1 */
- emit_simple_arith_swap2(p, inst, A0_SGE, 2);
+ 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 neither < nor > then we're != */
- src0 = src_vector(p, &inst->Src[0]);
- src1 = src_vector(p, &inst->Src[1]);
+ /* 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,
case TGSI_OPCODE_SSG:
/* compute (src>0) - (src<0) */
- src0 = src_vector(p, &inst->Src[0]);
+ src0 = src_vector(p, &inst->Src[0], fs);
tmp = i915_get_utemp(p);
i915_emit_arith(p,
break;
case TGSI_OPCODE_SUB:
- src0 = src_vector(p, &inst->Src[0]);
- src1 = src_vector(p, &inst->Src[1]);
+ src0 = src_vector(p, &inst->Src[0], fs);
+ src1 = src_vector(p, &inst->Src[1], fs);
i915_emit_arith(p,
A0_ADD,
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);
+ 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->Src[0]);
- src1 = src_vector(p, &inst->Src[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,
*/
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;
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
-}
-
}
p = i915_init_compile(i915, fs);
- i915_find_wpos_space(p);
- i915_translate_instructions(p, tokens);
+ i915_translate_instructions(p, tokens, fs);
i915_fixup_depth_write(p);
i915_fini_compile(i915, p);
projects / mesa.git / blobdiff