[Mesa-dev] [PATCH 09/12] R600/SI: replace AllReg_* with [SV]Src_* v2
Christian König
deathsimple at vodafone.de
Thu Feb 14 09:34:22 PST 2013
From: Christian König <christian.koenig at amd.com>
Mark all the operands that can also have an immediate.
v2: SOFFSET is also an SSrc_32 operand
Signed-off-by: Christian König <christian.koenig at amd.com>
---
lib/Target/R600/SIInstrFormats.td | 36 +++++------
lib/Target/R600/SIInstructions.td | 128 ++++++++++++++++++-------------------
lib/Target/R600/SIRegisterInfo.td | 10 ++-
3 files changed, 89 insertions(+), 85 deletions(-)
diff --git a/lib/Target/R600/SIInstrFormats.td b/lib/Target/R600/SIInstrFormats.td
index bd31bc1..5c69c15 100644
--- a/lib/Target/R600/SIInstrFormats.td
+++ b/lib/Target/R600/SIInstrFormats.td
@@ -22,25 +22,25 @@
//===----------------------------------------------------------------------===//
class VOP3_32 <bits<9> op, string opName, list<dag> pattern>
- : VOP3 <op, (outs VReg_32:$dst), (ins AllReg_32:$src0, VReg_32:$src1, VReg_32:$src2, i32imm:$src3, i32imm:$src4, i32imm:$src5, i32imm:$src6), opName, pattern>;
+ : VOP3 <op, (outs VReg_32:$dst), (ins VSrc_32:$src0, VReg_32:$src1, VReg_32:$src2, i32imm:$src3, i32imm:$src4, i32imm:$src5, i32imm:$src6), opName, pattern>;
class VOP3_64 <bits<9> op, string opName, list<dag> pattern>
- : VOP3 <op, (outs VReg_64:$dst), (ins AllReg_64:$src0, VReg_64:$src1, VReg_64:$src2, i32imm:$src3, i32imm:$src4, i32imm:$src5, i32imm:$src6), opName, pattern>;
+ : VOP3 <op, (outs VReg_64:$dst), (ins VSrc_64:$src0, VReg_64:$src1, VReg_64:$src2, i32imm:$src3, i32imm:$src4, i32imm:$src5, i32imm:$src6), opName, pattern>;
class SOP1_32 <bits<8> op, string opName, list<dag> pattern>
- : SOP1 <op, (outs SReg_32:$dst), (ins SReg_32:$src0), opName, pattern>;
+ : SOP1 <op, (outs SReg_32:$dst), (ins SSrc_32:$src0), opName, pattern>;
class SOP1_64 <bits<8> op, string opName, list<dag> pattern>
- : SOP1 <op, (outs SReg_64:$dst), (ins SReg_64:$src0), opName, pattern>;
+ : SOP1 <op, (outs SReg_64:$dst), (ins SSrc_64:$src0), opName, pattern>;
class SOP2_32 <bits<7> op, string opName, list<dag> pattern>
- : SOP2 <op, (outs SReg_32:$dst), (ins SReg_32:$src0, SReg_32:$src1), opName, pattern>;
+ : SOP2 <op, (outs SReg_32:$dst), (ins SSrc_32:$src0, SSrc_32:$src1), opName, pattern>;
class SOP2_64 <bits<7> op, string opName, list<dag> pattern>
- : SOP2 <op, (outs SReg_64:$dst), (ins SReg_64:$src0, SReg_64:$src1), opName, pattern>;
+ : SOP2 <op, (outs SReg_64:$dst), (ins SSrc_64:$src0, SSrc_64:$src1), opName, pattern>;
class SOP2_VCC <bits<7> op, string opName, list<dag> pattern>
- : SOP2 <op, (outs SReg_1:$vcc), (ins SReg_64:$src0, SReg_64:$src1), opName, pattern>;
+ : SOP2 <op, (outs SReg_1:$vcc), (ins SSrc_64:$src0, SSrc_64:$src1), opName, pattern>;
class VOP1_Helper <bits<8> op, RegisterClass vrc, RegisterClass arc,
string opName, list<dag> pattern> :
@@ -49,7 +49,7 @@ class VOP1_Helper <bits<8> op, RegisterClass vrc, RegisterClass arc,
>;
multiclass VOP1_32 <bits<8> op, string opName, list<dag> pattern> {
- def _e32: VOP1_Helper <op, VReg_32, AllReg_32, opName, pattern>;
+ def _e32: VOP1_Helper <op, VReg_32, VSrc_32, opName, pattern>;
def _e64 : VOP3_32 <{1, 1, op{6}, op{5}, op{4}, op{3}, op{2}, op{1}, op{0}},
opName, []
>;
@@ -57,7 +57,7 @@ multiclass VOP1_32 <bits<8> op, string opName, list<dag> pattern> {
multiclass VOP1_64 <bits<8> op, string opName, list<dag> pattern> {
- def _e32 : VOP1_Helper <op, VReg_64, AllReg_64, opName, pattern>;
+ def _e32 : VOP1_Helper <op, VReg_64, VSrc_64, opName, pattern>;
def _e64 : VOP3_64 <
{1, 1, op{6}, op{5}, op{4}, op{3}, op{2}, op{1}, op{0}},
@@ -73,7 +73,7 @@ class VOP2_Helper <bits<6> op, RegisterClass vrc, RegisterClass arc,
multiclass VOP2_32 <bits<6> op, string opName, list<dag> pattern> {
- def _e32 : VOP2_Helper <op, VReg_32, AllReg_32, opName, pattern>;
+ def _e32 : VOP2_Helper <op, VReg_32, VSrc_32, opName, pattern>;
def _e64 : VOP3_32 <{1, 0, 0, op{5}, op{4}, op{3}, op{2}, op{1}, op{0}},
opName, []
@@ -81,7 +81,7 @@ multiclass VOP2_32 <bits<6> op, string opName, list<dag> pattern> {
}
multiclass VOP2_64 <bits<6> op, string opName, list<dag> pattern> {
- def _e32: VOP2_Helper <op, VReg_64, AllReg_64, opName, pattern>;
+ def _e32: VOP2_Helper <op, VReg_64, VSrc_64, opName, pattern>;
def _e64 : VOP3_64 <
{1, 0, 0, op{5}, op{4}, op{3}, op{2}, op{1}, op{0}},
@@ -112,16 +112,16 @@ multiclass VOPC_Helper <bits<8> op, RegisterClass vrc, RegisterClass arc,
}
multiclass VOPC_32 <bits<8> op, string opName, list<dag> pattern>
- : VOPC_Helper <op, VReg_32, AllReg_32, opName, pattern>;
+ : VOPC_Helper <op, VReg_32, VSrc_32, opName, pattern>;
multiclass VOPC_64 <bits<8> op, string opName, list<dag> pattern>
- : VOPC_Helper <op, VReg_64, AllReg_64, opName, pattern>;
+ : VOPC_Helper <op, VReg_64, VSrc_64, opName, pattern>;
class SOPC_32 <bits<7> op, string opName, list<dag> pattern>
- : SOPC <op, (outs SCCReg:$dst), (ins SReg_32:$src0, SReg_32:$src1), opName, pattern>;
+ : SOPC <op, (outs SCCReg:$dst), (ins SSrc_32:$src0, SSrc_32:$src1), opName, pattern>;
class SOPC_64 <bits<7> op, string opName, list<dag> pattern>
- : SOPC <op, (outs SCCReg:$dst), (ins SReg_64:$src0, SReg_64:$src1), opName, pattern>;
+ : SOPC <op, (outs SCCReg:$dst), (ins SSrc_64:$src0, SSrc_64:$src1), opName, pattern>;
class MIMG_Load_Helper <bits<7> op, string asm> : MIMG <
op,
@@ -140,7 +140,7 @@ class MTBUF_Store_Helper <bits<3> op, string asm, RegisterClass regClass> : MTBU
(outs),
(ins regClass:$vdata, i16imm:$offset, i1imm:$offen, i1imm:$idxen, i1imm:$glc,
i1imm:$addr64, i8imm:$dfmt, i8imm:$nfmt, VReg_32:$vaddr,
- GPR4Align<SReg_128>:$srsrc, i1imm:$slc, i1imm:$tfe, SReg_32:$soffset),
+ GPR4Align<SReg_128>:$srsrc, i1imm:$slc, i1imm:$tfe, SSrc_32:$soffset),
asm,
[]> {
let mayStore = 1;
@@ -152,7 +152,7 @@ class MUBUF_Load_Helper <bits<7> op, string asm, RegisterClass regClass> : MUBUF
(outs regClass:$dst),
(ins i16imm:$offset, i1imm:$offen, i1imm:$idxen, i1imm:$glc, i1imm:$addr64,
i1imm:$lds, VReg_32:$vaddr, GPR4Align<SReg_128>:$srsrc, i1imm:$slc,
- i1imm:$tfe, SReg_32:$soffset),
+ i1imm:$tfe, SSrc_32:$soffset),
asm,
[]> {
let mayLoad = 1;
@@ -164,7 +164,7 @@ class MTBUF_Load_Helper <bits<3> op, string asm, RegisterClass regClass> : MTBUF
(outs regClass:$dst),
(ins i16imm:$offset, i1imm:$offen, i1imm:$idxen, i1imm:$glc, i1imm:$addr64,
i8imm:$dfmt, i8imm:$nfmt, VReg_32:$vaddr, GPR4Align<SReg_128>:$srsrc,
- i1imm:$slc, i1imm:$tfe, SReg_32:$soffset),
+ i1imm:$slc, i1imm:$tfe, SSrc_32:$soffset),
asm,
[]> {
let mayLoad = 1;
diff --git a/lib/Target/R600/SIInstructions.td b/lib/Target/R600/SIInstructions.td
index a09f243..63ef9ef 100644
--- a/lib/Target/R600/SIInstructions.td
+++ b/lib/Target/R600/SIInstructions.td
@@ -119,33 +119,33 @@ def S_GETREG_REGRD_B32 : SOPK_32 <0x00000014, "S_GETREG_REGRD_B32", []>;
defm V_CMP_F_F32 : VOPC_32 <0x00000000, "V_CMP_F_F32", []>;
defm V_CMP_LT_F32 : VOPC_32 <0x00000001, "V_CMP_LT_F32", []>;
def : Pat <
- (i1 (setcc (f32 AllReg_32:$src0), VReg_32:$src1, COND_LT)),
- (V_CMP_LT_F32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (f32 VSrc_32:$src0), VReg_32:$src1, COND_LT)),
+ (V_CMP_LT_F32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_EQ_F32 : VOPC_32 <0x00000002, "V_CMP_EQ_F32", []>;
def : Pat <
- (i1 (setcc (f32 AllReg_32:$src0), VReg_32:$src1, COND_EQ)),
- (V_CMP_EQ_F32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (f32 VSrc_32:$src0), VReg_32:$src1, COND_EQ)),
+ (V_CMP_EQ_F32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_LE_F32 : VOPC_32 <0x00000003, "V_CMP_LE_F32", []>;
def : Pat <
- (i1 (setcc (f32 AllReg_32:$src0), VReg_32:$src1, COND_LE)),
- (V_CMP_LE_F32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (f32 VSrc_32:$src0), VReg_32:$src1, COND_LE)),
+ (V_CMP_LE_F32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_GT_F32 : VOPC_32 <0x00000004, "V_CMP_GT_F32", []>;
def : Pat <
- (i1 (setcc (f32 AllReg_32:$src0), VReg_32:$src1, COND_GT)),
- (V_CMP_GT_F32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (f32 VSrc_32:$src0), VReg_32:$src1, COND_GT)),
+ (V_CMP_GT_F32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_LG_F32 : VOPC_32 <0x00000005, "V_CMP_LG_F32", []>;
def : Pat <
- (i1 (setcc (f32 AllReg_32:$src0), VReg_32:$src1, COND_NE)),
- (V_CMP_LG_F32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (f32 VSrc_32:$src0), VReg_32:$src1, COND_NE)),
+ (V_CMP_LG_F32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_GE_F32 : VOPC_32 <0x00000006, "V_CMP_GE_F32", []>;
def : Pat <
- (i1 (setcc (f32 AllReg_32:$src0), VReg_32:$src1, COND_GE)),
- (V_CMP_GE_F32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (f32 VSrc_32:$src0), VReg_32:$src1, COND_GE)),
+ (V_CMP_GE_F32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_O_F32 : VOPC_32 <0x00000007, "V_CMP_O_F32", []>;
defm V_CMP_U_F32 : VOPC_32 <0x00000008, "V_CMP_U_F32", []>;
@@ -155,8 +155,8 @@ defm V_CMP_NGT_F32 : VOPC_32 <0x0000000b, "V_CMP_NGT_F32", []>;
defm V_CMP_NLE_F32 : VOPC_32 <0x0000000c, "V_CMP_NLE_F32", []>;
defm V_CMP_NEQ_F32 : VOPC_32 <0x0000000d, "V_CMP_NEQ_F32", []>;
def : Pat <
- (i1 (setcc (f32 AllReg_32:$src0), VReg_32:$src1, COND_NE)),
- (V_CMP_NEQ_F32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (f32 VSrc_32:$src0), VReg_32:$src1, COND_NE)),
+ (V_CMP_NEQ_F32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_NLT_F32 : VOPC_32 <0x0000000e, "V_CMP_NLT_F32", []>;
defm V_CMP_TRU_F32 : VOPC_32 <0x0000000f, "V_CMP_TRU_F32", []>;
@@ -289,33 +289,33 @@ defm V_CMPSX_TRU_F64 : VOPC_64 <0x0000007f, "V_CMPSX_TRU_F64", []>;
defm V_CMP_F_I32 : VOPC_32 <0x00000080, "V_CMP_F_I32", []>;
defm V_CMP_LT_I32 : VOPC_32 <0x00000081, "V_CMP_LT_I32", []>;
def : Pat <
- (i1 (setcc (i32 AllReg_32:$src0), VReg_32:$src1, COND_LT)),
- (V_CMP_LT_I32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (i32 VSrc_32:$src0), VReg_32:$src1, COND_LT)),
+ (V_CMP_LT_I32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_EQ_I32 : VOPC_32 <0x00000082, "V_CMP_EQ_I32", []>;
def : Pat <
- (i1 (setcc (i32 AllReg_32:$src0), VReg_32:$src1, COND_EQ)),
- (V_CMP_EQ_I32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (i32 VSrc_32:$src0), VReg_32:$src1, COND_EQ)),
+ (V_CMP_EQ_I32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_LE_I32 : VOPC_32 <0x00000083, "V_CMP_LE_I32", []>;
def : Pat <
- (i1 (setcc (i32 AllReg_32:$src0), VReg_32:$src1, COND_LE)),
- (V_CMP_LE_I32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (i32 VSrc_32:$src0), VReg_32:$src1, COND_LE)),
+ (V_CMP_LE_I32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_GT_I32 : VOPC_32 <0x00000084, "V_CMP_GT_I32", []>;
def : Pat <
- (i1 (setcc (i32 AllReg_32:$src0), VReg_32:$src1, COND_GT)),
- (V_CMP_GT_I32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (i32 VSrc_32:$src0), VReg_32:$src1, COND_GT)),
+ (V_CMP_GT_I32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_NE_I32 : VOPC_32 <0x00000085, "V_CMP_NE_I32", []>;
def : Pat <
- (i1 (setcc (i32 AllReg_32:$src0), VReg_32:$src1, COND_NE)),
- (V_CMP_NE_I32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (i32 VSrc_32:$src0), VReg_32:$src1, COND_NE)),
+ (V_CMP_NE_I32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_GE_I32 : VOPC_32 <0x00000086, "V_CMP_GE_I32", []>;
def : Pat <
- (i1 (setcc (i32 AllReg_32:$src0), VReg_32:$src1, COND_GE)),
- (V_CMP_GE_I32_e64 AllReg_32:$src0, VReg_32:$src1)
+ (i1 (setcc (i32 VSrc_32:$src0), VReg_32:$src1, COND_GE)),
+ (V_CMP_GE_I32_e64 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_CMP_T_I32 : VOPC_32 <0x00000087, "V_CMP_T_I32", []>;
@@ -583,12 +583,12 @@ defm V_READFIRSTLANE_B32 : VOP1_32 <0x00000002, "V_READFIRSTLANE_B32", []>;
//defm V_CVT_I32_F64 : VOP1_32 <0x00000003, "V_CVT_I32_F64", []>;
//defm V_CVT_F64_I32 : VOP1_64 <0x00000004, "V_CVT_F64_I32", []>;
defm V_CVT_F32_I32 : VOP1_32 <0x00000005, "V_CVT_F32_I32",
- [(set VReg_32:$dst, (sint_to_fp AllReg_32:$src0))]
+ [(set VReg_32:$dst, (sint_to_fp VSrc_32:$src0))]
>;
//defm V_CVT_F32_U32 : VOP1_32 <0x00000006, "V_CVT_F32_U32", []>;
//defm V_CVT_U32_F32 : VOP1_32 <0x00000007, "V_CVT_U32_F32", []>;
defm V_CVT_I32_F32 : VOP1_32 <0x00000008, "V_CVT_I32_F32",
- [(set (i32 VReg_32:$dst), (fp_to_sint AllReg_32:$src0))]
+ [(set (i32 VReg_32:$dst), (fp_to_sint VSrc_32:$src0))]
>;
defm V_MOV_FED_B32 : VOP1_32 <0x00000009, "V_MOV_FED_B32", []>;
////def V_CVT_F16_F32 : VOP1_F16 <0x0000000a, "V_CVT_F16_F32", []>;
@@ -605,33 +605,33 @@ defm V_MOV_FED_B32 : VOP1_32 <0x00000009, "V_MOV_FED_B32", []>;
//defm V_CVT_U32_F64 : VOP1_32 <0x00000015, "V_CVT_U32_F64", []>;
//defm V_CVT_F64_U32 : VOP1_64 <0x00000016, "V_CVT_F64_U32", []>;
defm V_FRACT_F32 : VOP1_32 <0x00000020, "V_FRACT_F32",
- [(set VReg_32:$dst, (AMDGPUfract AllReg_32:$src0))]
+ [(set VReg_32:$dst, (AMDGPUfract VSrc_32:$src0))]
>;
defm V_TRUNC_F32 : VOP1_32 <0x00000021, "V_TRUNC_F32", []>;
defm V_CEIL_F32 : VOP1_32 <0x00000022, "V_CEIL_F32", []>;
defm V_RNDNE_F32 : VOP1_32 <0x00000023, "V_RNDNE_F32",
- [(set VReg_32:$dst, (frint AllReg_32:$src0))]
+ [(set VReg_32:$dst, (frint VSrc_32:$src0))]
>;
defm V_FLOOR_F32 : VOP1_32 <0x00000024, "V_FLOOR_F32",
- [(set VReg_32:$dst, (ffloor AllReg_32:$src0))]
+ [(set VReg_32:$dst, (ffloor VSrc_32:$src0))]
>;
defm V_EXP_F32 : VOP1_32 <0x00000025, "V_EXP_F32",
- [(set VReg_32:$dst, (fexp2 AllReg_32:$src0))]
+ [(set VReg_32:$dst, (fexp2 VSrc_32:$src0))]
>;
defm V_LOG_CLAMP_F32 : VOP1_32 <0x00000026, "V_LOG_CLAMP_F32", []>;
defm V_LOG_F32 : VOP1_32 <0x00000027, "V_LOG_F32",
- [(set VReg_32:$dst, (flog2 AllReg_32:$src0))]
+ [(set VReg_32:$dst, (flog2 VSrc_32:$src0))]
>;
defm V_RCP_CLAMP_F32 : VOP1_32 <0x00000028, "V_RCP_CLAMP_F32", []>;
defm V_RCP_LEGACY_F32 : VOP1_32 <0x00000029, "V_RCP_LEGACY_F32", []>;
defm V_RCP_F32 : VOP1_32 <0x0000002a, "V_RCP_F32",
- [(set VReg_32:$dst, (fdiv FP_ONE, AllReg_32:$src0))]
+ [(set VReg_32:$dst, (fdiv FP_ONE, VSrc_32:$src0))]
>;
defm V_RCP_IFLAG_F32 : VOP1_32 <0x0000002b, "V_RCP_IFLAG_F32", []>;
defm V_RSQ_CLAMP_F32 : VOP1_32 <0x0000002c, "V_RSQ_CLAMP_F32", []>;
defm V_RSQ_LEGACY_F32 : VOP1_32 <
0x0000002d, "V_RSQ_LEGACY_F32",
- [(set VReg_32:$dst, (int_AMDGPU_rsq AllReg_32:$src0))]
+ [(set VReg_32:$dst, (int_AMDGPU_rsq VSrc_32:$src0))]
>;
defm V_RSQ_F32 : VOP1_32 <0x0000002e, "V_RSQ_F32", []>;
defm V_RCP_F64 : VOP1_64 <0x0000002f, "V_RCP_F64", []>;
@@ -764,7 +764,7 @@ def S_WAITCNT : SOPP <0x0000000c, (ins i32imm:$simm16), "S_WAITCNT $simm16",
//def S_TTRACEDATA : SOPP_ <0x00000016, "S_TTRACEDATA", []>;
def V_CNDMASK_B32_e32 : VOP2 <0x00000000, (outs VReg_32:$dst),
- (ins AllReg_32:$src0, VReg_32:$src1, VCCReg:$vcc), "V_CNDMASK_B32_e32",
+ (ins VSrc_32:$src0, VReg_32:$src1, VCCReg:$vcc), "V_CNDMASK_B32_e32",
[]
>{
let DisableEncoding = "$vcc";
@@ -787,35 +787,35 @@ defm V_WRITELANE_B32 : VOP2_32 <0x00000002, "V_WRITELANE_B32", []>;
defm V_ADD_F32 : VOP2_32 <0x00000003, "V_ADD_F32", []>;
def : Pat <
- (f32 (fadd AllReg_32:$src0, VReg_32:$src1)),
- (V_ADD_F32_e32 AllReg_32:$src0, VReg_32:$src1)
+ (f32 (fadd VSrc_32:$src0, VReg_32:$src1)),
+ (V_ADD_F32_e32 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_SUB_F32 : VOP2_32 <0x00000004, "V_SUB_F32", []>;
def : Pat <
- (f32 (fsub AllReg_32:$src0, VReg_32:$src1)),
- (V_SUB_F32_e32 AllReg_32:$src0, VReg_32:$src1)
+ (f32 (fsub VSrc_32:$src0, VReg_32:$src1)),
+ (V_SUB_F32_e32 VSrc_32:$src0, VReg_32:$src1)
>;
defm V_SUBREV_F32 : VOP2_32 <0x00000005, "V_SUBREV_F32", []>;
defm V_MAC_LEGACY_F32 : VOP2_32 <0x00000006, "V_MAC_LEGACY_F32", []>;
defm V_MUL_LEGACY_F32 : VOP2_32 <
0x00000007, "V_MUL_LEGACY_F32",
- [(set VReg_32:$dst, (int_AMDGPU_mul AllReg_32:$src0, VReg_32:$src1))]
+ [(set VReg_32:$dst, (int_AMDGPU_mul VSrc_32:$src0, VReg_32:$src1))]
>;
defm V_MUL_F32 : VOP2_32 <0x00000008, "V_MUL_F32",
- [(set VReg_32:$dst, (fmul AllReg_32:$src0, VReg_32:$src1))]
+ [(set VReg_32:$dst, (fmul VSrc_32:$src0, VReg_32:$src1))]
>;
//defm V_MUL_I32_I24 : VOP2_32 <0x00000009, "V_MUL_I32_I24", []>;
//defm V_MUL_HI_I32_I24 : VOP2_32 <0x0000000a, "V_MUL_HI_I32_I24", []>;
//defm V_MUL_U32_U24 : VOP2_32 <0x0000000b, "V_MUL_U32_U24", []>;
//defm V_MUL_HI_U32_U24 : VOP2_32 <0x0000000c, "V_MUL_HI_U32_U24", []>;
defm V_MIN_LEGACY_F32 : VOP2_32 <0x0000000d, "V_MIN_LEGACY_F32",
- [(set VReg_32:$dst, (AMDGPUfmin AllReg_32:$src0, VReg_32:$src1))]
+ [(set VReg_32:$dst, (AMDGPUfmin VSrc_32:$src0, VReg_32:$src1))]
>;
defm V_MAX_LEGACY_F32 : VOP2_32 <0x0000000e, "V_MAX_LEGACY_F32",
- [(set VReg_32:$dst, (AMDGPUfmax AllReg_32:$src0, VReg_32:$src1))]
+ [(set VReg_32:$dst, (AMDGPUfmax VSrc_32:$src0, VReg_32:$src1))]
>;
defm V_MIN_F32 : VOP2_32 <0x0000000f, "V_MIN_F32", []>;
defm V_MAX_F32 : VOP2_32 <0x00000010, "V_MAX_F32", []>;
@@ -830,13 +830,13 @@ defm V_ASHRREV_I32 : VOP2_32 <0x00000018, "V_ASHRREV_I32", []>;
defm V_LSHL_B32 : VOP2_32 <0x00000019, "V_LSHL_B32", []>;
defm V_LSHLREV_B32 : VOP2_32 <0x0000001a, "V_LSHLREV_B32", []>;
defm V_AND_B32 : VOP2_32 <0x0000001b, "V_AND_B32",
- [(set VReg_32:$dst, (and AllReg_32:$src0, VReg_32:$src1))]
+ [(set VReg_32:$dst, (and VSrc_32:$src0, VReg_32:$src1))]
>;
defm V_OR_B32 : VOP2_32 <0x0000001c, "V_OR_B32",
- [(set VReg_32:$dst, (or AllReg_32:$src0, VReg_32:$src1))]
+ [(set VReg_32:$dst, (or VSrc_32:$src0, VReg_32:$src1))]
>;
defm V_XOR_B32 : VOP2_32 <0x0000001d, "V_XOR_B32",
- [(set VReg_32:$dst, (xor AllReg_32:$src0, VReg_32:$src1))]
+ [(set VReg_32:$dst, (xor VSrc_32:$src0, VReg_32:$src1))]
>;
defm V_BFM_B32 : VOP2_32 <0x0000001e, "V_BFM_B32", []>;
defm V_MAC_F32 : VOP2_32 <0x0000001f, "V_MAC_F32", []>;
@@ -847,10 +847,10 @@ defm V_MADAK_F32 : VOP2_32 <0x00000021, "V_MADAK_F32", []>;
//defm V_MBCNT_HI_U32_B32 : VOP2_32 <0x00000024, "V_MBCNT_HI_U32_B32", []>;
let Defs = [VCC] in { // Carry-out goes to VCC
defm V_ADD_I32 : VOP2_32 <0x00000025, "V_ADD_I32",
- [(set VReg_32:$dst, (add (i32 AllReg_32:$src0), (i32 VReg_32:$src1)))]
+ [(set VReg_32:$dst, (add (i32 VSrc_32:$src0), (i32 VReg_32:$src1)))]
>;
defm V_SUB_I32 : VOP2_32 <0x00000026, "V_SUB_I32",
- [(set VReg_32:$dst, (sub (i32 AllReg_32:$src0), (i32 VReg_32:$src1)))]
+ [(set VReg_32:$dst, (sub (i32 VSrc_32:$src0), (i32 VReg_32:$src1)))]
>;
} // End Defs = [VCC]
defm V_SUBREV_I32 : VOP2_32 <0x00000027, "V_SUBREV_I32", []>;
@@ -862,7 +862,7 @@ defm V_LDEXP_F32 : VOP2_32 <0x0000002b, "V_LDEXP_F32", []>;
////def V_CVT_PKNORM_I16_F32 : VOP2_I16 <0x0000002d, "V_CVT_PKNORM_I16_F32", []>;
////def V_CVT_PKNORM_U16_F32 : VOP2_U16 <0x0000002e, "V_CVT_PKNORM_U16_F32", []>;
defm V_CVT_PKRTZ_F16_F32 : VOP2_32 <0x0000002f, "V_CVT_PKRTZ_F16_F32",
- [(set VReg_32:$dst, (int_SI_packf16 AllReg_32:$src0, VReg_32:$src1))]
+ [(set VReg_32:$dst, (int_SI_packf16 VSrc_32:$src0, VReg_32:$src1))]
>;
////def V_CVT_PK_U16_U32 : VOP2_U16 <0x00000030, "V_CVT_PK_U16_U32", []>;
////def V_CVT_PK_I16_I32 : VOP2_I16 <0x00000031, "V_CVT_PK_I16_I32", []>;
@@ -933,8 +933,8 @@ def V_MUL_LO_U32 : VOP3_32 <0x00000169, "V_MUL_LO_U32", []>;
def V_MUL_HI_U32 : VOP3_32 <0x0000016a, "V_MUL_HI_U32", []>;
def V_MUL_LO_I32 : VOP3_32 <0x0000016b, "V_MUL_LO_I32", []>;
def : Pat <
- (mul AllReg_32:$src0, VReg_32:$src1),
- (V_MUL_LO_I32 AllReg_32:$src0, VReg_32:$src1, (IMPLICIT_DEF), 0, 0, 0, 0)
+ (mul VSrc_32:$src0, VReg_32:$src1),
+ (V_MUL_LO_I32 VSrc_32:$src0, VReg_32:$src1, (IMPLICIT_DEF), 0, 0, 0, 0)
>;
def V_MUL_HI_I32 : VOP3_32 <0x0000016c, "V_MUL_HI_I32", []>;
def V_DIV_SCALE_F32 : VOP3_32 <0x0000016d, "V_DIV_SCALE_F32", []>;
@@ -973,10 +973,10 @@ def : Pat <
def S_AND_B32 : SOP2_32 <0x0000000e, "S_AND_B32", []>;
def S_AND_B64 : SOP2_64 <0x0000000f, "S_AND_B64",
- [(set SReg_64:$dst, (and SReg_64:$src0, SReg_64:$src1))]
+ [(set SReg_64:$dst, (and SSrc_64:$src0, SSrc_64:$src1))]
>;
def S_AND_VCC : SOP2_VCC <0x0000000f, "S_AND_B64",
- [(set SReg_1:$vcc, (SIvcc_and SReg_64:$src0, SReg_64:$src1))]
+ [(set SReg_1:$vcc, (SIvcc_and SSrc_64:$src0, SSrc_64:$src1))]
>;
def S_OR_B32 : SOP2_32 <0x00000010, "S_OR_B32", []>;
def S_OR_B64 : SOP2_64 <0x00000011, "S_OR_B64", []>;
@@ -1379,23 +1379,23 @@ def : Pat <
def : POW_Common <V_LOG_F32_e32, V_EXP_F32_e32, V_MUL_F32_e32, VReg_32>;
def : Pat <
- (int_AMDGPU_div AllReg_32:$src0, AllReg_32:$src1),
- (V_MUL_LEGACY_F32_e32 AllReg_32:$src0, (V_RCP_LEGACY_F32_e32 AllReg_32:$src1))
+ (int_AMDGPU_div VSrc_32:$src0, VSrc_32:$src1),
+ (V_MUL_LEGACY_F32_e32 VSrc_32:$src0, (V_RCP_LEGACY_F32_e32 VSrc_32:$src1))
>;
def : Pat<
- (fdiv AllReg_32:$src0, AllReg_32:$src1),
- (V_MUL_F32_e32 AllReg_32:$src0, (V_RCP_F32_e32 AllReg_32:$src1))
+ (fdiv VSrc_32:$src0, VSrc_32:$src1),
+ (V_MUL_F32_e32 VSrc_32:$src0, (V_RCP_F32_e32 VSrc_32:$src1))
>;
def : Pat <
- (fcos AllReg_32:$src0),
- (V_COS_F32_e32 (V_MUL_F32_e32 AllReg_32:$src0, (V_MOV_IMM_I32 CONST.TWO_PI_INV)))
+ (fcos VSrc_32:$src0),
+ (V_COS_F32_e32 (V_MUL_F32_e32 VSrc_32:$src0, (V_MOV_IMM_I32 CONST.TWO_PI_INV)))
>;
def : Pat <
- (fsin AllReg_32:$src0),
- (V_SIN_F32_e32 (V_MUL_F32_e32 AllReg_32:$src0, (V_MOV_IMM_I32 CONST.TWO_PI_INV)))
+ (fsin VSrc_32:$src0),
+ (V_SIN_F32_e32 (V_MUL_F32_e32 VSrc_32:$src0, (V_MOV_IMM_I32 CONST.TWO_PI_INV)))
>;
def : Pat <
@@ -1423,8 +1423,8 @@ def : Pat <
/********** VOP3 Patterns **********/
/********** ================== **********/
-def : Pat <(f32 (IL_mad AllReg_32:$src0, VReg_32:$src1, VReg_32:$src2)),
- (V_MAD_LEGACY_F32 AllReg_32:$src0, VReg_32:$src1, VReg_32:$src2,
+def : Pat <(f32 (IL_mad VSrc_32:$src0, VReg_32:$src1, VReg_32:$src2)),
+ (V_MAD_LEGACY_F32 VSrc_32:$src0, VReg_32:$src1, VReg_32:$src2,
0, 0, 0, 0)>;
/********** ================== **********/
diff --git a/lib/Target/R600/SIRegisterInfo.td b/lib/Target/R600/SIRegisterInfo.td
index 809d503..150c92e 100644
--- a/lib/Target/R600/SIRegisterInfo.td
+++ b/lib/Target/R600/SIRegisterInfo.td
@@ -177,10 +177,14 @@ def VReg_256 : RegisterClass<"AMDGPU", [v8i32], 256, (add VGPR_256)>;
def VReg_512 : RegisterClass<"AMDGPU", [v16i32], 512, (add VGPR_512)>;
-// AllReg_* - A set of all scalar and vector registers of a given width.
-def AllReg_32 : RegisterClass<"AMDGPU", [f32, i32], 32, (add VReg_32, SReg_32)>;
+// [SV]Src_* operands can have either an immediate or an register
+def SSrc_32 : RegisterClass<"AMDGPU", [i32, f32], 32, (add SReg_32)>;
-def AllReg_64 : RegisterClass<"AMDGPU", [f64, i64], 64, (add SReg_64, VReg_64)>;
+def SSrc_64 : RegisterClass<"AMDGPU", [i64], 64, (add SReg_64)>;
+
+def VSrc_32 : RegisterClass<"AMDGPU", [i32, f32], 32, (add VReg_32, SReg_32)>;
+
+def VSrc_64 : RegisterClass<"AMDGPU", [i64], 64, (add SReg_64, VReg_64)>;
// Special register classes for predicates and the M0 register
def SCCReg : RegisterClass<"AMDGPU", [i1], 1, (add SCC)>;
--
1.7.10.4
More information about the mesa-dev
mailing list