return GL_FALSE;
}
- /* dst.z = exp(tmp.x) */
- pAsm->D.dst.opcode = SQ_OP2_INST_EXP_IEEE;
- pAsm->D.dst.math = 1;
- pAsm->D.dst.rtype = dstType;
- pAsm->D.dst.reg = dstReg;
- pAsm->D.dst.writex = 0;
- pAsm->D.dst.writey = 0;
- pAsm->D.dst.writez = 1;
- pAsm->D.dst.writew = 0;
-
- pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
- pAsm->S[0].src.reg = tmp;
- setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
- noneg_PVSSRC(&(pAsm->S[0].src));
- pAsm->S[0].src.swizzlex = SQ_SEL_X;
- pAsm->S[0].src.swizzley = SQ_SEL_X;
- pAsm->S[0].src.swizzlez = SQ_SEL_X;
- pAsm->S[0].src.swizzlew = SQ_SEL_X;
-
- if( GL_FALSE == next_ins(pAsm) )
+ /* dst.z = exp(tmp.x) */
+ pAsm->D.dst.opcode = SQ_OP2_INST_EXP_IEEE;
+ pAsm->D.dst.math = 1;
+ pAsm->D.dst.rtype = dstType;
+ pAsm->D.dst.reg = dstReg;
+ pAsm->D.dst.writex = 0;
+ pAsm->D.dst.writey = 0;
+ pAsm->D.dst.writez = 1;
+ pAsm->D.dst.writew = 0;
+
+ pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp;
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ noneg_PVSSRC(&(pAsm->S[0].src));
+ pAsm->S[0].src.swizzlex = SQ_SEL_X;
+ pAsm->S[0].src.swizzley = SQ_SEL_X;
+ pAsm->S[0].src.swizzlez = SQ_SEL_X;
+ pAsm->S[0].src.swizzlew = SQ_SEL_X;
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_MAX(r700_AssemblerBase *pAsm)
+{
+ if( GL_FALSE == checkop2(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->D.dst.opcode = SQ_OP2_INST_MAX;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_MIN(r700_AssemblerBase *pAsm)
+{
+ if( GL_FALSE == checkop2(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->D.dst.opcode = SQ_OP2_INST_MIN;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_MOV(r700_AssemblerBase *pAsm)
+{
+ checkop1(pAsm);
+
+ pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+ if (GL_FALSE == assemble_dst(pAsm))
+ {
+ return GL_FALSE;
+ }
+
+ if (GL_FALSE == assemble_src(pAsm, 0, -1))
+ {
+ return GL_FALSE;
+ }
+
+ if ( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_MUL(r700_AssemblerBase *pAsm)
+{
+ if( GL_FALSE == checkop2(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_POW(r700_AssemblerBase *pAsm)
+{
+ BITS tmp;
+
+ checkop1(pAsm);
+
+ tmp = gethelpr(pAsm);
+
+ // LG2 tmp.x, a.swizzle
+ pAsm->D.dst.opcode = SQ_OP2_INST_LOG_IEEE;
+ pAsm->D.dst.math = 1;
+
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp;
+ nomask_PVSDST(&(pAsm->D.dst));
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ // MUL tmp.x, tmp.x, b.swizzle
+ pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp;
+ nomask_PVSDST(&(pAsm->D.dst));
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp;
+ setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+ noneg_PVSSRC(&(pAsm->S[0].src));
+
+ if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ // EX2 dst.mask, tmp.x
+ // EX2 tmp.x, tmp.x
+ pAsm->D.dst.opcode = SQ_OP2_INST_EXP_IEEE;
+ pAsm->D.dst.math = 1;
+
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp;
+ nomask_PVSDST(&(pAsm->D.dst));
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp;
+ setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+ noneg_PVSSRC(&(pAsm->S[0].src));
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ // Now replicate result to all necessary channels in destination
+ pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp;
+
+ setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+ noneg_PVSSRC(&(pAsm->S[0].src));
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_RCP(r700_AssemblerBase *pAsm)
+{
+ return assemble_math_function(pAsm, SQ_OP2_INST_RECIP_IEEE);
+}
+
+GLboolean assemble_RSQ(r700_AssemblerBase *pAsm)
+{
+ return assemble_math_function(pAsm, SQ_OP2_INST_RECIPSQRT_IEEE);
+}
+
+GLboolean assemble_SIN(r700_AssemblerBase *pAsm)
+{
+ return assemble_math_function(pAsm, SQ_OP2_INST_SIN);
+}
+
+GLboolean assemble_SCS(r700_AssemblerBase *pAsm)
+{
+ BITS tmp;
+
+ checkop1(pAsm);
+
+ tmp = gethelpr(pAsm);
+
+ // COS tmp.x, a.x
+ pAsm->D.dst.opcode = SQ_OP2_INST_COS;
+ pAsm->D.dst.math = 1;
+
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp;
+ pAsm->D.dst.writex = 1;
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if ( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ // SIN tmp.y, a.x
+ pAsm->D.dst.opcode = SQ_OP2_INST_SIN;
+ pAsm->D.dst.math = 1;
+
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp;
+ pAsm->D.dst.writey = 1;
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ // MOV dst.mask, tmp
+ pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = DST_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp;
+
+ noswizzle_PVSSRC(&(pAsm->S[0].src));
+ pAsm->S[0].src.swizzlez = SQ_SEL_0;
+ pAsm->S[0].src.swizzlew = SQ_SEL_0;
+
+ if ( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode)
+{
+ if( GL_FALSE == checkop2(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->D.dst.opcode = opcode;
+ pAsm->D.dst.math = 1;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode)
+{
+ if( GL_FALSE == checkop2(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->D.dst.opcode = opcode;
+ pAsm->D.dst.math = 1;
+ pAsm->D.dst.predicated = 1;
+ pAsm->D2.dst2.SaturateMode = pAsm->pILInst[pAsm->uiCurInst].SaturateMode;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins2(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_SGE(r700_AssemblerBase *pAsm)
+{
+ if( GL_FALSE == checkop2(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->D.dst.opcode = SQ_OP2_INST_SETGE;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 1, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_SLT(r700_AssemblerBase *pAsm)
+{
+ if( GL_FALSE == checkop2(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->D.dst.opcode = SQ_OP2_INST_SETGT;
+
+ if( GL_FALSE == assemble_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 0, 1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 1, 0) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ return GL_TRUE;
+}
+
+GLboolean assemble_STP(r700_AssemblerBase *pAsm)
+{
+ return GL_TRUE;
+}
+
+GLboolean assemble_TEX(r700_AssemblerBase *pAsm)
+{
+ GLboolean src_const;
+ GLboolean need_barrier = GL_FALSE;
+
+ checkop1(pAsm);
+
+ switch (pAsm->pILInst[pAsm->uiCurInst].SrcReg[0].File)
+ {
+ case PROGRAM_UNIFORM:
+ case PROGRAM_CONSTANT:
+ case PROGRAM_LOCAL_PARAM:
+ case PROGRAM_ENV_PARAM:
+ case PROGRAM_STATE_VAR:
+ src_const = GL_TRUE;
+ break;
+ case PROGRAM_TEMPORARY:
+ case PROGRAM_INPUT:
+ default:
+ src_const = GL_FALSE;
+ break;
+ }
+
+ if (GL_TRUE == src_const)
+ {
+ if ( GL_FALSE == mov_temp(pAsm, 0) )
+ return GL_FALSE;
+ need_barrier = GL_TRUE;
+ }
+
- switch (pAsm->pILInst[pAsm->uiCurInst].Opcode)
- {
- case OPCODE_TEX:
- break;
- case OPCODE_TXB:
- radeon_error("do not support TXB yet\n");
- return GL_FALSE;
- break;
- case OPCODE_TXP:
- break;
- default:
- radeon_error("Internal error: bad texture op (not TEX)\n");
- return GL_FALSE;
- break;
- }
-
+ if (pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXP)
+ {
+ GLuint tmp = gethelpr(pAsm);
+ pAsm->D.dst.opcode = SQ_OP2_INST_RECIP_IEEE;
+ pAsm->D.dst.math = 1;
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp;
+ pAsm->D.dst.writew = 1;
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+ swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_W, SQ_SEL_W, SQ_SEL_W, SQ_SEL_W);
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp;
+ pAsm->D.dst.writex = 1;
+ pAsm->D.dst.writey = 1;
+ pAsm->D.dst.writez = 1;
+ pAsm->D.dst.writew = 0;
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+ setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+ pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[1].src.reg = tmp;
+ setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_W);
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ pAsm->aArgSubst[1] = tmp;
+ need_barrier = GL_TRUE;
+ }
+
+ if (pAsm->pILInst[pAsm->uiCurInst].TexSrcTarget == TEXTURE_CUBE_INDEX )
+ {
+ GLuint tmp1 = gethelpr(pAsm);
+ GLuint tmp2 = gethelpr(pAsm);
+
+ /* tmp1.xyzw = CUBE(R0.zzxy, R0.yxzz) */
+ pAsm->D.dst.opcode = SQ_OP2_INST_CUBE;
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp1;
+ nomask_PVSDST(&(pAsm->D.dst));
+
+ if( GL_FALSE == assemble_src(pAsm, 0, -1) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == assemble_src(pAsm, 0, 1) )
+ {
+ return GL_FALSE;
+ }
+
+ swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_Z, SQ_SEL_Z, SQ_SEL_X, SQ_SEL_Y);
+ swizzleagain_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Y, SQ_SEL_X, SQ_SEL_Z, SQ_SEL_Z);
+
+ if( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ /* tmp1.z = ABS(tmp1.z) dont have abs support in assembler currently
+ * have to do explicit instruction
+ */
+ pAsm->D.dst.opcode = SQ_OP2_INST_MAX;
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp1;
+ pAsm->D.dst.writez = 1;
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp1;
+ noswizzle_PVSSRC(&(pAsm->S[0].src));
+ pAsm->S[1].bits = pAsm->S[0].bits;
+ flipneg_PVSSRC(&(pAsm->S[1].src));
+
+ next_ins(pAsm);
+
+ /* tmp1.z = RCP_e(|tmp1.z|) */
+ pAsm->D.dst.opcode = SQ_OP2_INST_RECIP_IEEE;
+ pAsm->D.dst.math = 1;
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp1;
+ pAsm->D.dst.writez = 1;
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp1;
+ pAsm->S[0].src.swizzlex = SQ_SEL_Z;
+
+ next_ins(pAsm);
+
+ /* MULADD R0.x, R0.x, PS1, (0x3FC00000, 1.5f).x
+ * MULADD R0.y, R0.y, PS1, (0x3FC00000, 1.5f).x
+ * muladd has no writemask, have to use another temp
+ * also no support for imm constants, so add 1 here
+ */
+ pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+ pAsm->D.dst.op3 = 1;
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp2;
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp1;
+ noswizzle_PVSSRC(&(pAsm->S[0].src));
+ setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+ pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[1].src.reg = tmp1;
+ setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Z);
+ setaddrmode_PVSSRC(&(pAsm->S[2].src), ADDR_ABSOLUTE);
+ pAsm->S[2].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[2].src.reg = tmp1;
+ setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_1);
+
+ next_ins(pAsm);
+
+ /* ADD the remaining .5 */
+ pAsm->D.dst.opcode = SQ_OP2_INST_ADD;
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp2;
+ pAsm->D.dst.writex = 1;
+ pAsm->D.dst.writey = 1;
+ pAsm->D.dst.writez = 0;
+ pAsm->D.dst.writew = 0;
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp2;
+ noswizzle_PVSSRC(&(pAsm->S[0].src));
+ setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE);
+ pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[1].src.reg = 252; // SQ_ALU_SRC_0_5
+ noswizzle_PVSSRC(&(pAsm->S[1].src));
+
+ next_ins(pAsm);
+
+ /* tmp1.xy = temp2.xy */
+ pAsm->D.dst.opcode = SQ_OP2_INST_MOV;
+ setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+ pAsm->D.dst.rtype = DST_REG_TEMPORARY;
+ pAsm->D.dst.reg = tmp1;
+ pAsm->D.dst.writex = 1;
+ pAsm->D.dst.writey = 1;
+ pAsm->D.dst.writez = 0;
+ pAsm->D.dst.writew = 0;
+
+ setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+ pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+ pAsm->S[0].src.reg = tmp2;
+ noswizzle_PVSSRC(&(pAsm->S[0].src));
+
+ next_ins(pAsm);
+ pAsm->aArgSubst[1] = tmp1;
+ need_barrier = GL_TRUE;
+
+ }
+
- pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE;
++ if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXB)
++ {
++ pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L;
++ }
++ else
++ {
++ pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE;
++ }
++
++ pAsm->is_tex = GL_TRUE;
++ if ( GL_TRUE == need_barrier )
++
+ pAsm->is_tex = GL_TRUE;
+ if ( GL_TRUE == need_barrier )
+ {
+ pAsm->need_tex_barrier = GL_TRUE;
+ }
+ // Set src1 to tex unit id
+ pAsm->S[1].src.reg = pAsm->pILInst[pAsm->uiCurInst].TexSrcUnit;
+ pAsm->S[1].src.rtype = SRC_REG_TEMPORARY;
+
+ //No sw info from mesa compiler, so hard code here.
+ pAsm->S[1].src.swizzlex = SQ_SEL_X;
+ pAsm->S[1].src.swizzley = SQ_SEL_Y;
+ pAsm->S[1].src.swizzlez = SQ_SEL_Z;
+ pAsm->S[1].src.swizzlew = SQ_SEL_W;
+
+ if( GL_FALSE == tex_dst(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if( GL_FALSE == tex_src(pAsm) )
+ {
+ return GL_FALSE;
+ }
+
+ if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXP)
{
- return GL_FALSE;
+ /* hopefully did swizzles before */
+ noswizzle_PVSSRC(&(pAsm->S[0].src));
+ }
+
+ if(pAsm->pILInst[pAsm->uiCurInst].TexSrcTarget == TEXTURE_CUBE_INDEX)
+ {
+ /* SAMPLE dst, tmp.yxwy, CUBE */
+ pAsm->S[0].src.swizzlex = SQ_SEL_Y;
+ pAsm->S[0].src.swizzley = SQ_SEL_X;
+ pAsm->S[0].src.swizzlez = SQ_SEL_W;
+ pAsm->S[0].src.swizzlew = SQ_SEL_Y;
}
+
+ if ( GL_FALSE == next_ins(pAsm) )
+ {
+ return GL_FALSE;
+ }
return GL_TRUE;
}
projects / mesa.git / commitdiff