aboutsummaryrefslogtreecommitdiff
path: root/lib/Target/R600/R600Instructions.td
diff options
context:
space:
mode:
Diffstat (limited to 'lib/Target/R600/R600Instructions.td')
-rw-r--r--lib/Target/R600/R600Instructions.td301
1 files changed, 129 insertions, 172 deletions
diff --git a/lib/Target/R600/R600Instructions.td b/lib/Target/R600/R600Instructions.td
index 1d25da3622..1060b0a26b 100644
--- a/lib/Target/R600/R600Instructions.td
+++ b/lib/Target/R600/R600Instructions.td
@@ -983,58 +983,42 @@ def MIN : R600_2OP_Helper <0x4, "MIN", AMDGPUfmin>;
// XXX: Use the defs in TargetSelectionDAG.td instead of intrinsics.
def SETE : R600_2OP <
0x08, "SETE",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, FP_ONE, FP_ZERO,
- COND_EQ))]
+ [(set f32:$dst, (selectcc f32:$src0, f32:$src1, FP_ONE, FP_ZERO, COND_EQ))]
>;
def SGT : R600_2OP <
0x09, "SETGT",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, FP_ONE, FP_ZERO,
- COND_GT))]
+ [(set f32:$dst, (selectcc f32:$src0, f32:$src1, FP_ONE, FP_ZERO, COND_GT))]
>;
def SGE : R600_2OP <
0xA, "SETGE",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, FP_ONE, FP_ZERO,
- COND_GE))]
+ [(set f32:$dst, (selectcc f32:$src0, f32:$src1, FP_ONE, FP_ZERO, COND_GE))]
>;
def SNE : R600_2OP <
0xB, "SETNE",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, FP_ONE, FP_ZERO,
- COND_NE))]
+ [(set f32:$dst, (selectcc f32:$src0, f32:$src1, FP_ONE, FP_ZERO, COND_NE))]
>;
def SETE_DX10 : R600_2OP <
0xC, "SETE_DX10",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, (i32 -1), (i32 0),
- COND_EQ))]
+ [(set i32:$dst, (selectcc f32:$src0, f32:$src1, -1, 0, COND_EQ))]
>;
def SETGT_DX10 : R600_2OP <
0xD, "SETGT_DX10",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, (i32 -1), (i32 0),
- COND_GT))]
+ [(set i32:$dst, (selectcc f32:$src0, f32:$src1, -1, 0, COND_GT))]
>;
def SETGE_DX10 : R600_2OP <
0xE, "SETGE_DX10",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, (i32 -1), (i32 0),
- COND_GE))]
+ [(set i32:$dst, (selectcc f32:$src0, f32:$src1, -1, 0, COND_GE))]
>;
def SETNE_DX10 : R600_2OP <
0xF, "SETNE_DX10",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, (i32 -1), (i32 0),
- COND_NE))]
+ [(set i32:$dst, (selectcc f32:$src0, f32:$src1, -1, 0, COND_NE))]
>;
def FRACT : R600_1OP_Helper <0x10, "FRACT", AMDGPUfract>;
@@ -1092,38 +1076,32 @@ def MIN_UINT : R600_2OP_Helper <0x39, "MIN_UINT", AMDGPUumin>;
def SETE_INT : R600_2OP <
0x3A, "SETE_INT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETEQ))]
+ [(set i32:$dst, (selectcc i32:$src0, i32:$src1, -1, 0, SETEQ))]
>;
def SETGT_INT : R600_2OP <
0x3B, "SETGT_INT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETGT))]
+ [(set i32:$dst, (selectcc i32:$src0, i32:$src1, -1, 0, SETGT))]
>;
def SETGE_INT : R600_2OP <
0x3C, "SETGE_INT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETGE))]
+ [(set i32:$dst, (selectcc i32:$src0, i32:$src1, -1, 0, SETGE))]
>;
def SETNE_INT : R600_2OP <
0x3D, "SETNE_INT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETNE))]
+ [(set i32:$dst, (selectcc i32:$src0, i32:$src1, -1, 0, SETNE))]
>;
def SETGT_UINT : R600_2OP <
0x3E, "SETGT_UINT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETUGT))]
+ [(set i32:$dst, (selectcc i32:$src0, i32:$src1, -1, 0, SETUGT))]
>;
def SETGE_UINT : R600_2OP <
0x3F, "SETGE_UINT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETUGE))]
+ [(set i32:$dst, (selectcc i32:$src0, i32:$src1, -1, 0, SETUGE))]
>;
def PRED_SETE_INT : R600_2OP <0x42, "PRED_SETE_INT", []>;
@@ -1133,26 +1111,17 @@ def PRED_SETNE_INT : R600_2OP <0x45, "PRED_SETNE_INT", []>;
def CNDE_INT : R600_3OP <
0x1C, "CNDE_INT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), 0,
- (i32 R600_Reg32:$src1), (i32 R600_Reg32:$src2),
- COND_EQ))]
+ [(set i32:$dst, (selectcc i32:$src0, 0, i32:$src1, i32:$src2, COND_EQ))]
>;
def CNDGE_INT : R600_3OP <
0x1E, "CNDGE_INT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), 0,
- (i32 R600_Reg32:$src1), (i32 R600_Reg32:$src2),
- COND_GE))]
+ [(set i32:$dst, (selectcc i32:$src0, 0, i32:$src1, i32:$src2, COND_GE))]
>;
def CNDGT_INT : R600_3OP <
0x1D, "CNDGT_INT",
- [(set (i32 R600_Reg32:$dst),
- (selectcc (i32 R600_Reg32:$src0), 0,
- (i32 R600_Reg32:$src1), (i32 R600_Reg32:$src2),
- COND_GT))]
+ [(set i32:$dst, (selectcc i32:$src0, 0, i32:$src1, i32:$src2, COND_GT))]
>;
//===----------------------------------------------------------------------===//
@@ -1161,7 +1130,7 @@ def CNDGT_INT : R600_3OP <
def TEX_LD : R600_TEX <
0x03, "TEX_LD",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_txf R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_txf v4f32:$SRC_GPR,
imm:$OFFSET_X, imm:$OFFSET_Y, imm:$OFFSET_Z, imm:$RESOURCE_ID,
imm:$SAMPLER_ID, imm:$textureTarget))]
> {
@@ -1174,19 +1143,19 @@ let InOperandList = (ins R600_Reg128:$SRC_GPR, i32imm:$OFFSET_X,
def TEX_GET_TEXTURE_RESINFO : R600_TEX <
0x04, "TEX_GET_TEXTURE_RESINFO",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_txq R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_txq v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, imm:$textureTarget))]
>;
def TEX_GET_GRADIENTS_H : R600_TEX <
0x07, "TEX_GET_GRADIENTS_H",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_ddx R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_ddx v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, imm:$textureTarget))]
>;
def TEX_GET_GRADIENTS_V : R600_TEX <
0x08, "TEX_GET_GRADIENTS_V",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_ddy R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_ddy v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, imm:$textureTarget))]
>;
@@ -1202,37 +1171,37 @@ def TEX_SET_GRADIENTS_V : R600_TEX <
def TEX_SAMPLE : R600_TEX <
0x10, "TEX_SAMPLE",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_tex R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_tex v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, imm:$textureTarget))]
>;
def TEX_SAMPLE_C : R600_TEX <
0x18, "TEX_SAMPLE_C",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_tex R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_tex v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, TEX_SHADOW:$textureTarget))]
>;
def TEX_SAMPLE_L : R600_TEX <
0x11, "TEX_SAMPLE_L",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_txl R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_txl v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, imm:$textureTarget))]
>;
def TEX_SAMPLE_C_L : R600_TEX <
0x19, "TEX_SAMPLE_C_L",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_txl R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_txl v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, TEX_SHADOW:$textureTarget))]
>;
def TEX_SAMPLE_LB : R600_TEX <
0x12, "TEX_SAMPLE_LB",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_txb R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_txb v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, imm:$textureTarget))]
>;
def TEX_SAMPLE_C_LB : R600_TEX <
0x1A, "TEX_SAMPLE_C_LB",
- [(set R600_Reg128:$DST_GPR, (int_AMDGPU_txb R600_Reg128:$SRC_GPR,
+ [(set v4f32:$DST_GPR, (int_AMDGPU_txb v4f32:$SRC_GPR,
imm:$RESOURCE_ID, imm:$SAMPLER_ID, TEX_SHADOW:$textureTarget))]
>;
@@ -1262,32 +1231,22 @@ class MULADD_Common <bits<5> inst> : R600_3OP <
class MULADD_IEEE_Common <bits<5> inst> : R600_3OP <
inst, "MULADD_IEEE",
- [(set (f32 R600_Reg32:$dst),
- (fadd (fmul R600_Reg32:$src0, R600_Reg32:$src1), R600_Reg32:$src2))]
+ [(set f32:$dst, (fadd (fmul f32:$src0, f32:$src1), f32:$src2))]
>;
class CNDE_Common <bits<5> inst> : R600_3OP <
inst, "CNDE",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), FP_ZERO,
- (f32 R600_Reg32:$src1), (f32 R600_Reg32:$src2),
- COND_EQ))]
+ [(set f32:$dst, (selectcc f32:$src0, FP_ZERO, f32:$src1, f32:$src2, COND_EQ))]
>;
class CNDGT_Common <bits<5> inst> : R600_3OP <
inst, "CNDGT",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), FP_ZERO,
- (f32 R600_Reg32:$src1), (f32 R600_Reg32:$src2),
- COND_GT))]
+ [(set f32:$dst, (selectcc f32:$src0, FP_ZERO, f32:$src1, f32:$src2, COND_GT))]
>;
class CNDGE_Common <bits<5> inst> : R600_3OP <
inst, "CNDGE",
- [(set R600_Reg32:$dst,
- (selectcc (f32 R600_Reg32:$src0), FP_ZERO,
- (f32 R600_Reg32:$src1), (f32 R600_Reg32:$src2),
- COND_GE))]
+ [(set f32:$dst, (selectcc f32:$src0, FP_ZERO, f32:$src1, f32:$src2, COND_GE))]
>;
multiclass DOT4_Common <bits<11> inst> {
@@ -1295,7 +1254,7 @@ multiclass DOT4_Common <bits<11> inst> {
def _pseudo : R600_REDUCTION <inst,
(ins R600_Reg128:$src0, R600_Reg128:$src1),
"DOT4 $dst $src0, $src1",
- [(set R600_Reg32:$dst, (int_AMDGPU_dp4 R600_Reg128:$src0, R600_Reg128:$src1))]
+ [(set f32:$dst, (int_AMDGPU_dp4 v4f32:$src0, v4f32:$src1))]
>;
def _real : R600_2OP <inst, "DOT4", []>;
@@ -1308,7 +1267,7 @@ multiclass CUBE_Common <bits<11> inst> {
(outs R600_Reg128:$dst),
(ins R600_Reg128:$src),
"CUBE $dst $src",
- [(set R600_Reg128:$dst, (int_AMDGPU_cube R600_Reg128:$src))],
+ [(set v4f32:$dst, (int_AMDGPU_cube v4f32:$src))],
VecALU
> {
let isPseudo = 1;
@@ -1398,7 +1357,7 @@ class RECIP_CLAMPED_Common <bits<11> inst> : R600_1OP <
}
class RECIP_IEEE_Common <bits<11> inst> : R600_1OP <
- inst, "RECIP_IEEE", [(set R600_Reg32:$dst, (fdiv FP_ONE, R600_Reg32:$src0))]
+ inst, "RECIP_IEEE", [(set f32:$dst, (fdiv FP_ONE, f32:$src0))]
> {
let TransOnly = 1;
let Itinerary = TransALU;
@@ -1445,19 +1404,20 @@ class COS_Common <bits<11> inst> : R600_1OP <
multiclass DIV_Common <InstR600 recip_ieee> {
def : Pat<
- (int_AMDGPU_div R600_Reg32:$src0, R600_Reg32:$src1),
- (MUL_IEEE R600_Reg32:$src0, (recip_ieee R600_Reg32:$src1))
+ (int_AMDGPU_div f32:$src0, f32:$src1),
+ (MUL_IEEE $src0, (recip_ieee $src1))
>;
def : Pat<
- (fdiv R600_Reg32:$src0, R600_Reg32:$src1),
- (MUL_IEEE R600_Reg32:$src0, (recip_ieee R600_Reg32:$src1))
+ (fdiv f32:$src0, f32:$src1),
+ (MUL_IEEE $src0, (recip_ieee $src1))
>;
}
-class TGSI_LIT_Z_Common <InstR600 mul_lit, InstR600 log_clamped, InstR600 exp_ieee> : Pat <
- (int_TGSI_lit_z R600_Reg32:$src_x, R600_Reg32:$src_y, R600_Reg32:$src_w),
- (exp_ieee (mul_lit (log_clamped (MAX R600_Reg32:$src_y, (f32 ZERO))), R600_Reg32:$src_w, R600_Reg32:$src_x))
+class TGSI_LIT_Z_Common <InstR600 mul_lit, InstR600 log_clamped, InstR600 exp_ieee>
+ : Pat <
+ (int_TGSI_lit_z f32:$src_x, f32:$src_y, f32:$src_w),
+ (exp_ieee (mul_lit (log_clamped (MAX $src_y, (f32 ZERO))), $src_w, $src_x))
>;
//===----------------------------------------------------------------------===//
@@ -1497,11 +1457,10 @@ let Predicates = [isR600] in {
def RECIP_UINT_r600 : RECIP_UINT_Common <0x78>;
defm DIV_r600 : DIV_Common<RECIP_IEEE_r600>;
- def : POW_Common <LOG_IEEE_r600, EXP_IEEE_r600, MUL, R600_Reg32>;
+ def : POW_Common <LOG_IEEE_r600, EXP_IEEE_r600, MUL>;
def TGSI_LIT_Z_r600 : TGSI_LIT_Z_Common<MUL_LIT_r600, LOG_CLAMPED_r600, EXP_IEEE_r600>;
- def : Pat<(fsqrt R600_Reg32:$src),
- (MUL R600_Reg32:$src, (RECIPSQRT_CLAMPED_r600 R600_Reg32:$src))>;
+ def : Pat<(fsqrt f32:$src), (MUL $src, (RECIPSQRT_CLAMPED_r600 $src))>;
def R600_ExportSwz : ExportSwzInst {
let Word1{20-17} = 0; // BURST_COUNT
@@ -1577,13 +1536,13 @@ let Predicates = [isR600] in {
// Helper pattern for normalizing inputs to triginomic instructions for R700+
// cards.
class COS_PAT <InstR600 trig> : Pat<
- (fcos R600_Reg32:$src),
- (trig (MUL_IEEE (MOV_IMM_I32 CONST.TWO_PI_INV), R600_Reg32:$src))
+ (fcos f32:$src),
+ (trig (MUL_IEEE (MOV_IMM_I32 CONST.TWO_PI_INV), $src))
>;
class SIN_PAT <InstR600 trig> : Pat<
- (fsin R600_Reg32:$src),
- (trig (MUL_IEEE (MOV_IMM_I32 CONST.TWO_PI_INV), R600_Reg32:$src))
+ (fsin f32:$src),
+ (trig (MUL_IEEE (MOV_IMM_I32 CONST.TWO_PI_INV), $src))
>;
//===----------------------------------------------------------------------===//
@@ -1621,11 +1580,10 @@ def RECIPSQRT_IEEE_eg : RECIPSQRT_IEEE_Common<0x89>;
def SIN_eg : SIN_Common<0x8D>;
def COS_eg : COS_Common<0x8E>;
-def : POW_Common <LOG_IEEE_eg, EXP_IEEE_eg, MUL, R600_Reg32>;
+def : POW_Common <LOG_IEEE_eg, EXP_IEEE_eg, MUL>;
def : SIN_PAT <SIN_eg>;
def : COS_PAT <COS_eg>;
-def : Pat<(fsqrt R600_Reg32:$src),
- (MUL R600_Reg32:$src, (RECIPSQRT_CLAMPED_eg R600_Reg32:$src))>;
+def : Pat<(fsqrt f32:$src), (MUL $src, (RECIPSQRT_CLAMPED_eg $src))>;
} // End Predicates = [isEG]
//===----------------------------------------------------------------------===//
@@ -1649,9 +1607,8 @@ let Predicates = [isEGorCayman] in {
// (16,8) = (Input << 8) >> 24 = (Input & 0xffffff) >> 16
// (24,8) = (Input << 0) >> 24 = (Input & 0xffffffff) >> 24
def BFE_UINT_eg : R600_3OP <0x4, "BFE_UINT",
- [(set R600_Reg32:$dst, (int_AMDIL_bit_extract_u32 R600_Reg32:$src0,
- R600_Reg32:$src1,
- R600_Reg32:$src2))],
+ [(set i32:$dst, (int_AMDIL_bit_extract_u32 i32:$src0, i32:$src1,
+ i32:$src2))],
VecALU
>;
@@ -1659,8 +1616,7 @@ let Predicates = [isEGorCayman] in {
defm : BFIPatterns <BFI_INT_eg>;
def BIT_ALIGN_INT_eg : R600_3OP <0xC, "BIT_ALIGN_INT",
- [(set R600_Reg32:$dst, (AMDGPUbitalign R600_Reg32:$src0, R600_Reg32:$src1,
- R600_Reg32:$src2))],
+ [(set i32:$dst, (AMDGPUbitalign i32:$src0, i32:$src1, i32:$src2))],
VecALU
>;
@@ -1705,11 +1661,9 @@ let hasSideEffects = 1 in {
// XXX: Lowering SELECT_CC will sometimes generate fp_to_[su]int nodes,
// which do not need to be truncated since the fp values are 0.0f or 1.0f.
// We should look into handling these cases separately.
- def : Pat<(fp_to_sint R600_Reg32:$src0),
- (FLT_TO_INT_eg (TRUNC R600_Reg32:$src0))>;
+ def : Pat<(fp_to_sint f32:$src0), (FLT_TO_INT_eg (TRUNC $src0))>;
- def : Pat<(fp_to_uint R600_Reg32:$src0),
- (FLT_TO_UINT_eg (TRUNC R600_Reg32:$src0))>;
+ def : Pat<(fp_to_uint f32:$src0), (FLT_TO_UINT_eg (TRUNC $src0))>;
def EG_ExportSwz : ExportSwzInst {
let Word1{19-16} = 0; // BURST_COUNT
@@ -1811,14 +1765,14 @@ class RAT_WRITE_CACHELESS_eg <dag ins, bits<4> comp_mask, string name,
def RAT_WRITE_CACHELESS_32_eg : RAT_WRITE_CACHELESS_eg <
(ins R600_TReg32_X:$rw_gpr, R600_TReg32_X:$index_gpr, InstFlag:$eop),
0x1, "RAT_WRITE_CACHELESS_32_eg",
- [(global_store (i32 R600_TReg32_X:$rw_gpr), R600_TReg32_X:$index_gpr)]
+ [(global_store i32:$rw_gpr, i32:$index_gpr)]
>;
//128-bit store
def RAT_WRITE_CACHELESS_128_eg : RAT_WRITE_CACHELESS_eg <
(ins R600_Reg128:$rw_gpr, R600_TReg32_X:$index_gpr, InstFlag:$eop),
0xf, "RAT_WRITE_CACHELESS_128",
- [(global_store (v4i32 R600_Reg128:$rw_gpr), R600_TReg32_X:$index_gpr)]
+ [(global_store v4i32:$rw_gpr, i32:$index_gpr)]
>;
class VTX_READ_eg <string name, bits<8> buffer_id, dag outs, list<dag> pattern>
@@ -1943,19 +1897,19 @@ class VTX_READ_128_eg <bits<8> buffer_id, list<dag> pattern>
//===----------------------------------------------------------------------===//
def VTX_READ_PARAM_8_eg : VTX_READ_8_eg <0,
- [(set (i32 R600_TReg32_X:$dst), (load_param_zexti8 ADDRVTX_READ:$ptr))]
+ [(set i32:$dst, (load_param_zexti8 ADDRVTX_READ:$ptr))]
>;
def VTX_READ_PARAM_16_eg : VTX_READ_16_eg <0,
- [(set (i32 R600_TReg32_X:$dst), (load_param_zexti16 ADDRVTX_READ:$ptr))]
+ [(set i32:$dst, (load_param_zexti16 ADDRVTX_READ:$ptr))]
>;
def VTX_READ_PARAM_32_eg : VTX_READ_32_eg <0,
- [(set (i32 R600_TReg32_X:$dst), (load_param ADDRVTX_READ:$ptr))]
+ [(set i32:$dst, (load_param ADDRVTX_READ:$ptr))]
>;
def VTX_READ_PARAM_128_eg : VTX_READ_128_eg <0,
- [(set (v4i32 R600_Reg128:$dst), (load_param ADDRVTX_READ:$ptr))]
+ [(set v4i32:$dst, (load_param ADDRVTX_READ:$ptr))]
>;
//===----------------------------------------------------------------------===//
@@ -1964,17 +1918,17 @@ def VTX_READ_PARAM_128_eg : VTX_READ_128_eg <0,
// 8-bit reads
def VTX_READ_GLOBAL_8_eg : VTX_READ_8_eg <1,
- [(set (i32 R600_TReg32_X:$dst), (zextloadi8_global ADDRVTX_READ:$ptr))]
+ [(set i32:$dst, (zextloadi8_global ADDRVTX_READ:$ptr))]
>;
// 32-bit reads
def VTX_READ_GLOBAL_32_eg : VTX_READ_32_eg <1,
- [(set (i32 R600_TReg32_X:$dst), (global_load ADDRVTX_READ:$ptr))]
+ [(set i32:$dst, (global_load ADDRVTX_READ:$ptr))]
>;
// 128-bit reads
def VTX_READ_GLOBAL_128_eg : VTX_READ_128_eg <1,
- [(set (v4i32 R600_Reg128:$dst), (global_load ADDRVTX_READ:$ptr))]
+ [(set v4i32:$dst, (global_load ADDRVTX_READ:$ptr))]
>;
//===----------------------------------------------------------------------===//
@@ -1983,7 +1937,7 @@ def VTX_READ_GLOBAL_128_eg : VTX_READ_128_eg <1,
//===----------------------------------------------------------------------===//
def CONSTANT_LOAD_eg : VTX_READ_32_eg <1,
- [(set (i32 R600_TReg32_X:$dst), (constant_load ADDRVTX_READ:$ptr))]
+ [(set i32:$dst, (constant_load ADDRVTX_READ:$ptr))]
>;
}
@@ -2013,7 +1967,7 @@ def SIN_cm : SIN_Common<0x8D>;
def COS_cm : COS_Common<0x8E>;
} // End isVector = 1
-def : POW_Common <LOG_IEEE_cm, EXP_IEEE_cm, MUL, R600_Reg32>;
+def : POW_Common <LOG_IEEE_cm, EXP_IEEE_cm, MUL>;
def : SIN_PAT <SIN_cm>;
def : COS_PAT <COS_cm>;
@@ -2022,8 +1976,8 @@ defm DIV_cm : DIV_Common<RECIP_IEEE_cm>;
// RECIP_UINT emulation for Cayman
// The multiplication scales from [0,1] to the unsigned integer range
def : Pat <
- (AMDGPUurecip R600_Reg32:$src0),
- (FLT_TO_UINT_eg (MUL_IEEE (RECIP_IEEE_cm (UINT_TO_FLT_eg R600_Reg32:$src0)),
+ (AMDGPUurecip i32:$src0),
+ (FLT_TO_UINT_eg (MUL_IEEE (RECIP_IEEE_cm (UINT_TO_FLT_eg $src0)),
(MOV_IMM_I32 CONST.FP_UINT_MAX_PLUS_1)))
>;
@@ -2033,8 +1987,7 @@ def : Pat <
let COUNT = 0;
}
-def : Pat<(fsqrt R600_Reg32:$src),
- (MUL R600_Reg32:$src, (RECIPSQRT_CLAMPED_cm R600_Reg32:$src))>;
+def : Pat<(fsqrt f32:$src), (MUL R600_Reg32:$src, (RECIPSQRT_CLAMPED_cm $src))>;
} // End isCayman
@@ -2099,17 +2052,23 @@ def MASK_WRITE : AMDGPUShaderInst <
def TXD: InstR600 <
(outs R600_Reg128:$dst),
- (ins R600_Reg128:$src0, R600_Reg128:$src1, R600_Reg128:$src2, i32imm:$resourceId, i32imm:$samplerId, i32imm:$textureTarget),
+ (ins R600_Reg128:$src0, R600_Reg128:$src1, R600_Reg128:$src2,
+ i32imm:$resourceId, i32imm:$samplerId, i32imm:$textureTarget),
"TXD $dst, $src0, $src1, $src2, $resourceId, $samplerId, $textureTarget",
- [(set R600_Reg128:$dst, (int_AMDGPU_txd R600_Reg128:$src0, R600_Reg128:$src1, R600_Reg128:$src2, imm:$resourceId, imm:$samplerId, imm:$textureTarget))], NullALU> {
+ [(set v4f32:$dst, (int_AMDGPU_txd v4f32:$src0, v4f32:$src1, v4f32:$src2,
+ imm:$resourceId, imm:$samplerId, imm:$textureTarget))],
+ NullALU > {
let TEXInst = 1;
}
def TXD_SHADOW: InstR600 <
(outs R600_Reg128:$dst),
- (ins R600_Reg128:$src0, R600_Reg128:$src1, R600_Reg128:$src2, i32imm:$resourceId, i32imm:$samplerId, i32imm:$textureTarget),
+ (ins R600_Reg128:$src0, R600_Reg128:$src1, R600_Reg128:$src2,
+ i32imm:$resourceId, i32imm:$samplerId, i32imm:$textureTarget),
"TXD_SHADOW $dst, $src0, $src1, $src2, $resourceId, $samplerId, $textureTarget",
- [(set R600_Reg128:$dst, (int_AMDGPU_txd R600_Reg128:$src0, R600_Reg128:$src1, R600_Reg128:$src2, imm:$resourceId, imm:$samplerId, TEX_SHADOW:$textureTarget))], NullALU
+ [(set v4f32:$dst, (int_AMDGPU_txd v4f32:$src0, v4f32:$src1, v4f32:$src2,
+ imm:$resourceId, imm:$samplerId, TEX_SHADOW:$textureTarget))],
+ NullALU
> {
let TEXInst = 1;
}
@@ -2149,7 +2108,7 @@ def CONST_COPY : Instruction {
def TEX_VTX_CONSTBUF :
InstR600ISA <(outs R600_Reg128:$dst), (ins MEMxi:$ptr, i32imm:$BUFFER_ID), "VTX_READ_eg $dst, $ptr",
- [(set R600_Reg128:$dst, (CONST_ADDRESS ADDRGA_VAR_OFFSET:$ptr, (i32 imm:$BUFFER_ID)))]>,
+ [(set v4i32:$dst, (CONST_ADDRESS ADDRGA_VAR_OFFSET:$ptr, (i32 imm:$BUFFER_ID)))]>,
VTX_WORD1_GPR, VTX_WORD0 {
let VC_INST = 0;
@@ -2203,7 +2162,7 @@ def TEX_VTX_CONSTBUF :
def TEX_VTX_TEXBUF:
InstR600ISA <(outs R600_Reg128:$dst), (ins MEMxi:$ptr, i32imm:$BUFFER_ID), "TEX_VTX_EXPLICIT_READ $dst, $ptr",
- [(set R600_Reg128:$dst, (int_R600_load_texbuf ADDRGA_VAR_OFFSET:$ptr, imm:$BUFFER_ID))]>,
+ [(set v4f32:$dst, (int_R600_load_texbuf ADDRGA_VAR_OFFSET:$ptr, imm:$BUFFER_ID))]>,
VTX_WORD1_GPR, VTX_WORD0 {
let VC_INST = 0;
@@ -2329,9 +2288,8 @@ let isTerminator=1 in {
// CND*_INT Pattterns for f32 True / False values
class CND_INT_f32 <InstR600 cnd, CondCode cc> : Pat <
- (selectcc (i32 R600_Reg32:$src0), 0, (f32 R600_Reg32:$src1),
- R600_Reg32:$src2, cc),
- (cnd R600_Reg32:$src0, R600_Reg32:$src1, R600_Reg32:$src2)
+ (selectcc i32:$src0, 0, f32:$src1, f32:$src2, cc),
+ (cnd $src0, $src1, $src2)
>;
def : CND_INT_f32 <CNDE_INT, SETEQ>;
@@ -2340,9 +2298,8 @@ def : CND_INT_f32 <CNDGE_INT, SETGE>;
//CNDGE_INT extra pattern
def : Pat <
- (selectcc (i32 R600_Reg32:$src0), -1, (i32 R600_Reg32:$src1),
- (i32 R600_Reg32:$src2), COND_GT),
- (CNDGE_INT R600_Reg32:$src0, R600_Reg32:$src1, R600_Reg32:$src2)
+ (selectcc i32:$src0, -1, i32:$src1, i32:$src2, COND_GT),
+ (CNDGE_INT $src0, $src1, $src2)
>;
// KIL Patterns
@@ -2352,56 +2309,56 @@ def KILP : Pat <
>;
def KIL : Pat <
- (int_AMDGPU_kill R600_Reg32:$src0),
- (MASK_WRITE (KILLGT (f32 ZERO), (f32 R600_Reg32:$src0)))
+ (int_AMDGPU_kill f32:$src0),
+ (MASK_WRITE (KILLGT (f32 ZERO), $src0))
>;
// SGT Reverse args
def : Pat <
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, FP_ONE, FP_ZERO, COND_LT),
- (SGT R600_Reg32:$src1, R600_Reg32:$src0)
+ (selectcc f32:$src0, f32:$src1, FP_ONE, FP_ZERO, COND_LT),
+ (SGT $src1, $src0)
>;
// SGE Reverse args
def : Pat <
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, FP_ONE, FP_ZERO, COND_LE),
- (SGE R600_Reg32:$src1, R600_Reg32:$src0)
+ (selectcc f32:$src0, f32:$src1, FP_ONE, FP_ZERO, COND_LE),
+ (SGE $src1, $src0)
>;
// SETGT_DX10 reverse args
def : Pat <
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, COND_LT),
- (SETGT_DX10 R600_Reg32:$src1, R600_Reg32:$src0)
+ (selectcc f32:$src0, f32:$src1, -1, 0, COND_LT),
+ (SETGT_DX10 $src1, $src0)
>;
// SETGE_DX10 reverse args
def : Pat <
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, COND_LE),
- (SETGE_DX10 R600_Reg32:$src1, R600_Reg32:$src0)
+ (selectcc f32:$src0, f32:$src1, -1, 0, COND_LE),
+ (SETGE_DX10 $src1, $src0)
>;
// SETGT_INT reverse args
def : Pat <
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETLT),
- (SETGT_INT R600_Reg32:$src1, R600_Reg32:$src0)
+ (selectcc i32:$src0, i32:$src1, -1, 0, SETLT),
+ (SETGT_INT $src1, $src0)
>;
// SETGE_INT reverse args
def : Pat <
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETLE),
- (SETGE_INT R600_Reg32:$src1, R600_Reg32:$src0)
+ (selectcc i32:$src0, i32:$src1, -1, 0, SETLE),
+ (SETGE_INT $src1, $src0)
>;
// SETGT_UINT reverse args
def : Pat <
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETULT),
- (SETGT_UINT R600_Reg32:$src1, R600_Reg32:$src0)
+ (selectcc i32:$src0, i32:$src1, -1, 0, SETULT),
+ (SETGT_UINT $src1, $src0)
>;
// SETGE_UINT reverse args
def : Pat <
- (selectcc (i32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETULE),
- (SETGE_UINT R600_Reg32:$src1, R600_Reg32:$src0)
+ (selectcc i32:$src0, i32:$src1, -1, 0, SETULE),
+ (SETGE_UINT $src1, $src0)
>;
// The next two patterns are special cases for handling 'true if ordered' and
@@ -2414,50 +2371,50 @@ def : Pat <
//SETE - 'true if ordered'
def : Pat <
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, FP_ONE, FP_ZERO, SETO),
- (SETE R600_Reg32:$src0, R600_Reg32:$src1)
+ (selectcc f32:$src0, f32:$src1, FP_ONE, FP_ZERO, SETO),
+ (SETE $src0, $src1)
>;
//SETE_DX10 - 'true if ordered'
def : Pat <
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETO),
- (SETE_DX10 R600_Reg32:$src0, R600_Reg32:$src1)
+ (selectcc f32:$src0, f32:$src1, -1, 0, SETO),
+ (SETE_DX10 $src0, $src1)
>;
//SNE - 'true if unordered'
def : Pat <
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, FP_ONE, FP_ZERO, SETUO),
- (SNE R600_Reg32:$src0, R600_Reg32:$src1)
+ (selectcc f32:$src0, f32:$src1, FP_ONE, FP_ZERO, SETUO),
+ (SNE $src0, $src1)
>;
//SETNE_DX10 - 'true if ordered'
def : Pat <
- (selectcc (f32 R600_Reg32:$src0), R600_Reg32:$src1, -1, 0, SETUO),
- (SETNE_DX10 R600_Reg32:$src0, R600_Reg32:$src1)
+ (selectcc f32:$src0, f32:$src1, -1, 0, SETUO),
+ (SETNE_DX10 $src0, $src1)
>;
-def : Extract_Element <f32, v4f32, R600_Reg128, 0, sub0>;
-def : Extract_Element <f32, v4f32, R600_Reg128, 1, sub1>;
-def : Extract_Element <f32, v4f32, R600_Reg128, 2, sub2>;
-def : Extract_Element <f32, v4f32, R600_Reg128, 3, sub3>;
+def : Extract_Element <f32, v4f32, 0, sub0>;
+def : Extract_Element <f32, v4f32, 1, sub1>;
+def : Extract_Element <f32, v4f32, 2, sub2>;
+def : Extract_Element <f32, v4f32, 3, sub3>;
-def : Insert_Element <f32, v4f32, R600_Reg32, R600_Reg128, 0, sub0>;
-def : Insert_Element <f32, v4f32, R600_Reg32, R600_Reg128, 1, sub1>;
-def : Insert_Element <f32, v4f32, R600_Reg32, R600_Reg128, 2, sub2>;
-def : Insert_Element <f32, v4f32, R600_Reg32, R600_Reg128, 3, sub3>;
+def : Insert_Element <f32, v4f32, 0, sub0>;
+def : Insert_Element <f32, v4f32, 1, sub1>;
+def : Insert_Element <f32, v4f32, 2, sub2>;
+def : Insert_Element <f32, v4f32, 3, sub3>;
-def : Extract_Element <i32, v4i32, R600_Reg128, 0, sub0>;
-def : Extract_Element <i32, v4i32, R600_Reg128, 1, sub1>;
-def : Extract_Element <i32, v4i32, R600_Reg128, 2, sub2>;
-def : Extract_Element <i32, v4i32, R600_Reg128, 3, sub3>;
+def : Extract_Element <i32, v4i32, 0, sub0>;
+def : Extract_Element <i32, v4i32, 1, sub1>;
+def : Extract_Element <i32, v4i32, 2, sub2>;
+def : Extract_Element <i32, v4i32, 3, sub3>;
-def : Insert_Element <i32, v4i32, R600_Reg32, R600_Reg128, 0, sub0>;
-def : Insert_Element <i32, v4i32, R600_Reg32, R600_Reg128, 1, sub1>;
-def : Insert_Element <i32, v4i32, R600_Reg32, R600_Reg128, 2, sub2>;
-def : Insert_Element <i32, v4i32, R600_Reg32, R600_Reg128, 3, sub3>;
+def : Insert_Element <i32, v4i32, 0, sub0>;
+def : Insert_Element <i32, v4i32, 1, sub1>;
+def : Insert_Element <i32, v4i32, 2, sub2>;
+def : Insert_Element <i32, v4i32, 3, sub3>;
-def : Vector4_Build <v4f32, R600_Reg128, f32, R600_Reg32>;
-def : Vector4_Build <v4i32, R600_Reg128, i32, R600_Reg32>;
+def : Vector4_Build <v4f32, f32>;
+def : Vector4_Build <v4i32, i32>;
// bitconvert patterns
generated by cgit v1.2.3-18-g5258 (git 2.32.0) at 2024-11-22 11:24:57 +0000