case TGSI_OPCODE_DP2A: return SVGA3DOP_DP2ADD;
case TGSI_OPCODE_DP3: return SVGA3DOP_DP3;
case TGSI_OPCODE_DP4: return SVGA3DOP_DP4;
- case TGSI_OPCODE_ENDFOR: return SVGA3DOP_ENDLOOP;
case TGSI_OPCODE_FRC: return SVGA3DOP_FRC;
- case TGSI_OPCODE_BGNFOR: return SVGA3DOP_LOOP;
case TGSI_OPCODE_MAD: return SVGA3DOP_MAD;
case TGSI_OPCODE_MAX: return SVGA3DOP_MAX;
case TGSI_OPCODE_MIN: return SVGA3DOP_MIN;
}
dest.mask = reg->Register.WriteMask;
+ assert(dest.mask);
if (insn->Instruction.Saturate)
dest.dstMod = SVGA3DDSTMOD_SATURATE;
}
+
+
+/* SVGA shaders may not refer to >1 constant register in a single
+ * instruction. This function checks for that usage and inserts a
+ * move to temporary if detected.
+ */
+static boolean submit_op4( struct svga_shader_emitter *emit,
+ SVGA3dShaderInstToken inst,
+ SVGA3dShaderDestToken dest,
+ struct src_register src0,
+ struct src_register src1,
+ struct src_register src2,
+ struct src_register src3)
+{
+ SVGA3dShaderDestToken temp0;
+ SVGA3dShaderDestToken temp3;
+ boolean need_temp0 = FALSE;
+ boolean need_temp3 = FALSE;
+ SVGA3dShaderRegType type0, type1, type2, type3;
+
+ temp0.value = 0;
+ temp3.value = 0;
+ type0 = SVGA3dShaderGetRegType( src0.base.value );
+ type1 = SVGA3dShaderGetRegType( src1.base.value );
+ type2 = SVGA3dShaderGetRegType( src2.base.value );
+ type3 = SVGA3dShaderGetRegType( src2.base.value );
+
+ /* Make life a little easier - this is only used by the TXD
+ * instruction which is guaranteed not to have a constant/input reg
+ * in one slot at least:
+ */
+ assert(type1 == SVGA3DREG_SAMPLER);
+
+ if (type0 == SVGA3DREG_CONST &&
+ ((type3 == SVGA3DREG_CONST && src0.base.num != src3.base.num) ||
+ (type2 == SVGA3DREG_CONST && src0.base.num != src2.base.num)))
+ need_temp0 = TRUE;
+
+ if (type3 == SVGA3DREG_CONST &&
+ (type2 == SVGA3DREG_CONST && src3.base.num != src2.base.num))
+ need_temp3 = TRUE;
+
+ if (type0 == SVGA3DREG_INPUT &&
+ ((type3 == SVGA3DREG_INPUT && src0.base.num != src3.base.num) ||
+ (type2 == SVGA3DREG_INPUT && src0.base.num != src2.base.num)))
+ need_temp0 = TRUE;
+
+ if (type3 == SVGA3DREG_INPUT &&
+ (type2 == SVGA3DREG_INPUT && src3.base.num != src2.base.num))
+ need_temp3 = TRUE;
+
+ if (need_temp0)
+ {
+ temp0 = get_temp( emit );
+
+ if (!emit_op1( emit, inst_token( SVGA3DOP_MOV ), temp0, src0 ))
+ return FALSE;
+
+ src0 = src( temp0 );
+ }
+
+ if (need_temp3)
+ {
+ temp3 = get_temp( emit );
+
+ if (!emit_op1( emit, inst_token( SVGA3DOP_MOV ), temp3, src3 ))
+ return FALSE;
+
+ src3 = src( temp3 );
+ }
+
+ if (!emit_op4( emit, inst, dest, src0, src1, src2, src3 ))
+ return FALSE;
+
+ if (need_temp3)
+ release_temp( emit, temp3 );
+ if (need_temp0)
+ release_temp( emit, temp0 );
+ return TRUE;
+}
+
+
static boolean emit_def_const( struct svga_shader_emitter *emit,
SVGA3dShaderConstType type,
unsigned idx,
break;
default:
assert(0);
+ opcode = inst_token( SVGA3DOP_NOP );
break;
}
SVGA3dShaderDestToken tmp;
inst.value = 0;
- inst.op = SVGA3DOP_TEX;
switch (insn->Instruction.Opcode) {
case TGSI_OPCODE_TEX:
+ inst.op = SVGA3DOP_TEX;
break;
case TGSI_OPCODE_TXP:
+ inst.op = SVGA3DOP_TEX;
inst.control = SVGA3DOPCONT_PROJECT;
break;
case TGSI_OPCODE_TXB:
+ inst.op = SVGA3DOP_TEX;
inst.control = SVGA3DOPCONT_BIAS;
break;
+ case TGSI_OPCODE_TXL:
+ inst.op = SVGA3DOP_TEXLDL;
+ break;
default:
assert(0);
return FALSE;
* zero in that case.
*/
if (emit->dynamic_branching_level > 0 &&
+ inst.op == SVGA3DOP_TEX &&
SVGA3dShaderGetRegType(texcoord.base.value) == SVGA3DREG_TEMP) {
struct src_register zero = get_zero_immediate( emit );
/* Translate texture instructions to SVGA3D representation.
*/
-static boolean emit_tex3(struct svga_shader_emitter *emit,
+static boolean emit_tex4(struct svga_shader_emitter *emit,
const struct tgsi_full_instruction *insn,
SVGA3dShaderDestToken dst )
{
SVGA3dShaderInstToken inst;
- struct src_register src0;
- struct src_register src1;
- struct src_register src2;
+ struct src_register texcoord;
+ struct src_register ddx;
+ struct src_register ddy;
+ struct src_register sampler;
+
+ texcoord = translate_src_register( emit, &insn->Src[0] );
+ ddx = translate_src_register( emit, &insn->Src[1] );
+ ddy = translate_src_register( emit, &insn->Src[2] );
+ sampler = translate_src_register( emit, &insn->Src[3] );
inst.value = 0;
case TGSI_OPCODE_TXD:
inst.op = SVGA3DOP_TEXLDD; /* 4 args! */
break;
- case TGSI_OPCODE_TXL:
- inst.op = SVGA3DOP_TEXLDL; /* 2 args! */
- break;
+ default:
+ assert(0);
+ return FALSE;
}
- src0 = translate_src_register( emit, &insn->Src[0] );
- src1 = translate_src_register( emit, &insn->Src[1] );
- src2 = translate_src_register( emit, &insn->Src[2] );
-
- return submit_op3( emit, inst, dst, src0, src1, src2 );
+ return submit_op4( emit, inst, dst, texcoord, sampler, ddx, ddy );
}
case TGSI_OPCODE_TEX:
case TGSI_OPCODE_TXB:
case TGSI_OPCODE_TXP:
+ case TGSI_OPCODE_TXL:
if (!emit_tex2( emit, insn, tex_result ))
return FALSE;
break;
- case TGSI_OPCODE_TXL:
case TGSI_OPCODE_TXD:
- if (!emit_tex3( emit, insn, tex_result ))
+ if (!emit_tex4( emit, insn, tex_result ))
return FALSE;
break;
default:
if (compare) {
- SVGA3dShaderDestToken src0_zdivw = get_temp( emit );
- struct src_register tex_src_x = scalar(src(tex_result), TGSI_SWIZZLE_Y);
- struct src_register one =
- scalar( get_zero_immediate( emit ), TGSI_SWIZZLE_W );
-
- /* Divide texcoord R by Q */
- if (!submit_op1( emit, inst_token( SVGA3DOP_RCP ),
- src0_zdivw,
- scalar(src0, TGSI_SWIZZLE_W) ))
- return FALSE;
+ if (dst.mask & TGSI_WRITEMASK_XYZ) {
+ SVGA3dShaderDestToken src0_zdivw = get_temp( emit );
+ struct src_register tex_src_x = scalar(src(tex_result), TGSI_SWIZZLE_Y);
+
+ /* Divide texcoord R by Q */
+ if (!submit_op1( emit, inst_token( SVGA3DOP_RCP ),
+ writemask(src0_zdivw, TGSI_WRITEMASK_X),
+ scalar(src0, TGSI_SWIZZLE_W) ))
+ return FALSE;
- if (!submit_op2( emit, inst_token( SVGA3DOP_MUL ),
- src0_zdivw,
- scalar(src0, TGSI_SWIZZLE_Z),
- src(src0_zdivw) ))
- return FALSE;
+ if (!submit_op2( emit, inst_token( SVGA3DOP_MUL ),
+ writemask(src0_zdivw, TGSI_WRITEMASK_X),
+ scalar(src0, TGSI_SWIZZLE_Z),
+ scalar(src(src0_zdivw), TGSI_SWIZZLE_X) ))
+ return FALSE;
- if (!emit_select(
- emit,
- emit->key.fkey.tex[src1.base.num].compare_func,
- dst,
- src(src0_zdivw),
- tex_src_x))
- return FALSE;
+ if (!emit_select(
+ emit,
+ emit->key.fkey.tex[src1.base.num].compare_func,
+ writemask( dst, TGSI_WRITEMASK_XYZ ),
+ scalar(src(src0_zdivw), TGSI_SWIZZLE_X),
+ tex_src_x))
+ return FALSE;
+ }
- return submit_op1( emit, inst_token( SVGA3DOP_MOV ),
- writemask( dst, TGSI_WRITEMASK_W),
- one );
+ if (dst.mask & TGSI_WRITEMASK_W) {
+ struct src_register one =
+ scalar( get_zero_immediate( emit ), TGSI_SWIZZLE_W );
+
+ if (!submit_op1( emit, inst_token( SVGA3DOP_MOV ),
+ writemask( dst, TGSI_WRITEMASK_W ),
+ one ))
+ return FALSE;
+ }
+
+ return TRUE;
}
else if (!emit->use_sm30 && dst.mask != TGSI_WRITEMASK_XYZW)
{
fraction = dst;
else if (dst.mask & TGSI_WRITEMASK_X)
fraction = get_temp( emit );
+ else
+ fraction.value = 0;
/* If y is being written, fill it with src0 - floor(src0).
*/
*/
if (dst.mask & TGSI_WRITEMASK_X) {
if (!submit_op2( emit, inst_token( SVGA3DOP_ADD ),
- writemask( dst, dst.mask & TGSI_WRITEMASK_X ),
+ writemask( dst, TGSI_WRITEMASK_X ),
src0,
scalar( negate( src( fraction ) ), TGSI_SWIZZLE_Y ) ) )
return FALSE;
if (!submit_op1( emit, inst_token( SVGA3DOP_EXP ),
- writemask( dst, dst.mask & TGSI_WRITEMASK_X ),
+ writemask( dst, TGSI_WRITEMASK_X ),
scalar( src( dst ), TGSI_SWIZZLE_X ) ) )
return FALSE;
*/
if (dst.mask & TGSI_WRITEMASK_Z) {
if (!submit_op1( emit, inst_token( SVGA3DOP_EXPP ),
- writemask( dst, dst.mask & TGSI_WRITEMASK_Z ),
+ writemask( dst, TGSI_WRITEMASK_Z ),
src0 ) )
return FALSE;
}
struct src_register abs_src0;
SVGA3dShaderDestToken log2_abs;
+ abs_tmp.value = 0;
+
if (dst.mask & TGSI_WRITEMASK_Z)
log2_abs = dst;
else if (dst.mask & TGSI_WRITEMASK_XY)
log2_abs = get_temp( emit );
+ else
+ log2_abs.value = 0;
/* If z is being written, fill it with log2( abs( src0 ) ).
*/
if_token = inst_token( SVGA3DOP_IFC );
- if (emit->key.fkey.front_cw)
- if_token.control = SVGA3DOPCOMP_GT;
- else
+ if (emit->key.fkey.front_ccw)
if_token.control = SVGA3DOPCOMP_LT;
+ else
+ if_token.control = SVGA3DOPCOMP_GT;
zero = scalar(zero, TGSI_SWIZZLE_X);
temp = dst_register( SVGA3DREG_TEMP,
emit->nr_hw_temp++ );
- if (emit->key.fkey.front_cw) {
- pass = scalar( zero, TGSI_SWIZZLE_W );
- fail = scalar( zero, TGSI_SWIZZLE_X );
- } else {
+ if (emit->key.fkey.front_ccw) {
pass = scalar( zero, TGSI_SWIZZLE_X );
fail = scalar( zero, TGSI_SWIZZLE_W );
+ } else {
+ pass = scalar( zero, TGSI_SWIZZLE_W );
+ fail = scalar( zero, TGSI_SWIZZLE_X );
}
if (!emit_conditional(emit, PIPE_FUNC_GREATER,
if (emit->info.opcode_count[TGSI_OPCODE_IF] >= 1 ||
emit->info.opcode_count[TGSI_OPCODE_BGNLOOP] >= 1 ||
- emit->info.opcode_count[TGSI_OPCODE_BGNFOR] >= 1 ||
emit->info.opcode_count[TGSI_OPCODE_DDX] >= 1 ||
emit->info.opcode_count[TGSI_OPCODE_DDY] >= 1 ||
emit->info.opcode_count[TGSI_OPCODE_SGE] >= 1 ||
projects / mesa.git / blobdiff