maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

AMDGPU: Move MnemonicAlias out of instruction def hierarchy 

Unfortunately MnemonicAlias defines a "Predicates" field just like an
instruction or pattern, with a somewhat different interpretation.

This ends up overriding the intended Predicates set by
PredicateControl on the pseudoinstruction defintions with an empty
list. This allowed incorrectly selecting instructions that should have
been rejected due to the SubtargetPredicate from patterns on the
instruction definition.

This does remove the divergent predicate from the 64-bit shift
patterns, which were already not used for the 32-bit shift, so I'm not
sure what the point was. This also removes a second, redundant copy of
the 64-bit divergent patterns.

llvm-svn: 371427
This commit is contained in:
Matt Arsenault 2019-09-09 17:25:35 +00:00
parent c0728eac15
commit d2a9516a6d
7 changed files with 73 additions and 85 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
llvm/lib/Target/AMDGPU/VOP1Instructions.td
View File

@ -107,6 +107,13 @@ multiclass VOP1Inst <string opName, VOPProfile P,
def _sdwa : VOP1_SDWA_Pseudo <opName, P>;
foreach _ = BoolToList<P.HasExtDPP>.ret in
def _dpp : VOP1_DPP_Pseudo <opName, P>;
def : MnemonicAlias<opName#"_e32", opName>, LetDummies;
def : MnemonicAlias<opName#"_e64", opName>, LetDummies;
def : MnemonicAlias<opName#"_sdwa", opName>, LetDummies;
foreach _ = BoolToList<P.HasExtDPP>.ret in
def : MnemonicAlias<opName#"_dpp", opName>, LetDummies;
}
// Special profile for instructions which have clamp

6
llvm/lib/Target/AMDGPU/VOP2Instructions.td
View File

@ -539,9 +539,9 @@ defm V_MAX_LEGACY_F32 : VOP2Inst <"v_max_legacy_f32", VOP_F32_F32_F32, AMDGPUfma
let SubtargetPredicate = isGFX6GFX7GFX10 in {
let isCommutable = 1 in {
defm V_MAC_LEGACY_F32 : VOP2Inst <"v_mac_legacy_f32", VOP_F32_F32_F32>;
defm V_LSHR_B32 : VOP2Inst <"v_lshr_b32", VOP_I32_I32_I32>;
defm V_ASHR_I32 : VOP2Inst <"v_ashr_i32", VOP_I32_I32_I32>;
defm V_LSHL_B32 : VOP2Inst <"v_lshl_b32", VOP_I32_I32_I32>;
defm V_LSHR_B32 : VOP2Inst <"v_lshr_b32", VOP_I32_I32_I32, srl>;
defm V_ASHR_I32 : VOP2Inst <"v_ashr_i32", VOP_I32_I32_I32, sra>;
defm V_LSHL_B32 : VOP2Inst <"v_lshl_b32", VOP_I32_I32_I32, shl>;
} // End isCommutable = 1
} // End SubtargetPredicate = isGFX6GFX7GFX10

25
llvm/lib/Target/AMDGPU/VOP3Instructions.td
View File

@ -385,12 +385,12 @@ def V_TRIG_PREOP_F64 : VOP3Inst <"v_trig_preop_f64", VOP3_Profile<VOP_F64_F64_I3
}
let SchedRW = [Write64Bit] in {
let SubtargetPredicate = isGFX6GFX7GFX10, Predicates = [isGFX6GFX7GFX10] in {
def V_LSHL_B64 : VOP3Inst <"v_lshl_b64", VOP3_Profile<VOP_PAT_GEN<VOP_I64_I64_I32>>, shl>;
def V_LSHR_B64 : VOP3Inst <"v_lshr_b64", VOP3_Profile<VOP_PAT_GEN<VOP_I64_I64_I32>>, srl>;
def V_ASHR_I64 : VOP3Inst <"v_ashr_i64", VOP3_Profile<VOP_PAT_GEN<VOP_I64_I64_I32>>, sra>;
let SubtargetPredicate = isGFX6GFX7GFX10 in {
def V_LSHL_B64 : VOP3Inst <"v_lshl_b64", VOP3_Profile<VOP_I64_I64_I32>, shl>;
def V_LSHR_B64 : VOP3Inst <"v_lshr_b64", VOP3_Profile<VOP_I64_I64_I32>, srl>;
def V_ASHR_I64 : VOP3Inst <"v_ashr_i64", VOP3_Profile<VOP_I64_I64_I32>, sra>;
def V_MULLIT_F32 : VOP3Inst <"v_mullit_f32", VOP3_Profile<VOP_F32_F32_F32_F32>>;
} // End SubtargetPredicate = isGFX6GFX7GFX10, Predicates = [isGFX6GFX7GFX10]
} // End SubtargetPredicate = isGFX6GFX7GFX10
let SubtargetPredicate = isGFX8Plus in {
def V_LSHLREV_B64 : VOP3Inst <"v_lshlrev_b64", VOP3_Profile<VOP_I64_I32_I64>, lshl_rev>;
@ -399,21 +399,6 @@ def V_ASHRREV_I64 : VOP3Inst <"v_ashrrev_i64", VOP3_Profile<VOP_I64_I32_I64>, as
} // End SubtargetPredicate = isGFX8Plus
} // End SchedRW = [Write64Bit]
let Predicates = [isGFX8Plus] in {
def : GCNPat <
(getDivergentFrag<shl>.ret i64:$x, i32:$y),
(V_LSHLREV_B64 $y, $x)
>;
def : AMDGPUPat <
(getDivergentFrag<srl>.ret i64:$x, i32:$y),
(V_LSHRREV_B64 $y, $x)
>;
def : AMDGPUPat <
(getDivergentFrag<sra>.ret i64:$x, i32:$y),
(V_ASHRREV_I64 $y, $x)
>;
}
let SchedRW = [Write32Bit] in {
let SubtargetPredicate = isGFX8Plus in {

11
llvm/lib/Target/AMDGPU/VOPInstructions.td
View File

@ -14,6 +14,7 @@ class LetDummies {
bit isReMaterializable;
bit isAsCheapAsAMove;
bit VOPAsmPrefer32Bit;
bit FPDPRounding;
Predicate SubtargetPredicate;
string Constraints;
string DisableEncoding;
@ -41,9 +42,7 @@ class VOP_Pseudo <string opName, string suffix, VOPProfile P, dag outs, dag ins,
string asm, list<dag> pattern> :
InstSI <outs, ins, asm, pattern>,
VOP <opName>,
SIMCInstr <opName#suffix, SIEncodingFamily.NONE>,
MnemonicAlias<opName#suffix, opName> {
SIMCInstr <opName#suffix, SIEncodingFamily.NONE> {
let isPseudo = 1;
let isCodeGenOnly = 1;
let UseNamedOperandTable = 1;
@ -473,8 +472,7 @@ class VOP_SDWA9Be<VOPProfile P> : VOP_SDWA9e<P> {
class VOP_SDWA_Pseudo <string opName, VOPProfile P, list<dag> pattern=[]> :
InstSI <P.OutsSDWA, P.InsSDWA, "", pattern>,
VOP <opName>,
SIMCInstr <opName#"_sdwa", SIEncodingFamily.NONE>,
MnemonicAlias <opName#"_sdwa", opName> {
SIMCInstr <opName#"_sdwa", SIEncodingFamily.NONE> {
let isPseudo = 1;
let isCodeGenOnly = 1;
@ -595,8 +593,7 @@ class VOP_DPPe<VOPProfile P, bit IsDPP16=0> : Enc64 {
class VOP_DPP_Pseudo <string OpName, VOPProfile P, list<dag> pattern=[]> :
InstSI <P.OutsDPP, P.InsDPP, OpName#P.AsmDPP, pattern>,
VOP <OpName>,
SIMCInstr <OpName#"_dpp", SIEncodingFamily.NONE>,
MnemonicAlias <OpName#"_dpp", OpName> {
SIMCInstr <OpName#"_dpp", SIEncodingFamily.NONE> {
let isPseudo = 1;
let isCodeGenOnly = 1;

37
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-ashr.mir
View File

@ -216,13 +216,13 @@ body: |
; GFX6-LABEL: name: ashr_s64_sv
; GFX6: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX6: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX6: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX6: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
; GFX7-LABEL: name: ashr_s64_sv
; GFX7: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX7: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX7: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
; GFX8-LABEL: name: ashr_s64_sv
; GFX8: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
@ -237,8 +237,8 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX10: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX10: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
%0:sgpr(s64) = COPY $sgpr0_sgpr1
%1:vgpr(s32) = COPY $vgpr0
%2:vgpr(s64) = G_ASHR %0, %1
@ -256,13 +256,13 @@ body: |
; GFX6-LABEL: name: ashr_s64_vs
; GFX6: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX6: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX6: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX6: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
; GFX7-LABEL: name: ashr_s64_vs
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX7: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX7: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
; GFX8-LABEL: name: ashr_s64_vs
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
@ -277,8 +277,8 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX10: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX10: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
%0:vgpr(s64) = COPY $vgpr0_vgpr1
%1:sgpr(s32) = COPY $sgpr0
%2:vgpr(s64) = G_ASHR %0, %1
@ -296,13 +296,13 @@ body: |
; GFX6-LABEL: name: ashr_s64_vv
; GFX6: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX6: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX6: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX6: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
; GFX7-LABEL: name: ashr_s64_vv
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX7: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX7: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
; GFX8-LABEL: name: ashr_s64_vv
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
@ -317,11 +317,10 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: [[V_ASHRREV_I64_:%[0-9]+]]:vreg_64 = V_ASHRREV_I64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_ASHRREV_I64_]]
; GFX10: [[V_ASHR_I64_:%[0-9]+]]:vreg_64 = V_ASHR_I64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_ASHR_I64_]]
%0:vgpr(s64) = COPY $vgpr0_vgpr1
%1:vgpr(s32) = COPY $vgpr2
%2:vgpr(s64) = G_ASHR %0, %1
S_ENDPGM 0, implicit %2
...

36
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-lshr.mir
View File

@ -216,13 +216,13 @@ body: |
; GFX6-LABEL: name: lshr_s64_sv
; GFX6: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX6: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX6: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX6: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
; GFX7-LABEL: name: lshr_s64_sv
; GFX7: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX7: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX7: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
; GFX8-LABEL: name: lshr_s64_sv
; GFX8: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
@ -237,8 +237,8 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX10: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX10: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
%0:sgpr(s64) = COPY $sgpr0_sgpr1
%1:vgpr(s32) = COPY $vgpr0
%2:vgpr(s64) = G_LSHR %0, %1
@ -256,13 +256,13 @@ body: |
; GFX6-LABEL: name: lshr_s64_vs
; GFX6: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX6: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX6: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX6: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
; GFX7-LABEL: name: lshr_s64_vs
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX7: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX7: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
; GFX8-LABEL: name: lshr_s64_vs
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
@ -277,8 +277,8 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX10: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX10: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
%0:vgpr(s64) = COPY $vgpr0_vgpr1
%1:sgpr(s32) = COPY $sgpr0
%2:vgpr(s64) = G_LSHR %0, %1
@ -296,13 +296,13 @@ body: |
; GFX6-LABEL: name: lshr_s64_vv
; GFX6: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX6: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX6: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX6: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
; GFX7-LABEL: name: lshr_s64_vv
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX7: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX7: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
; GFX8-LABEL: name: lshr_s64_vv
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
@ -317,8 +317,8 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: [[V_LSHRREV_B64_:%[0-9]+]]:vreg_64 = V_LSHRREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHRREV_B64_]]
; GFX10: [[V_LSHR_B64_:%[0-9]+]]:vreg_64 = V_LSHR_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHR_B64_]]
%0:vgpr(s64) = COPY $vgpr0_vgpr1
%1:vgpr(s32) = COPY $vgpr2
%2:vgpr(s64) = G_LSHR %0, %1

36
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-shl.mir
View File

@ -216,13 +216,13 @@ body: |
; GFX6-LABEL: name: shl_s64_sv
; GFX6: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX6: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX6: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX6: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
; GFX7-LABEL: name: shl_s64_sv
; GFX7: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX7: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX7: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
; GFX8-LABEL: name: shl_s64_sv
; GFX8: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
@ -237,8 +237,8 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:sreg_64_xexec = COPY $sgpr0_sgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX10: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX10: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
%0:sgpr(s64) = COPY $sgpr0_sgpr1
%1:vgpr(s32) = COPY $vgpr0
%2:vgpr(s64) = G_SHL %0, %1
@ -256,13 +256,13 @@ body: |
; GFX6-LABEL: name: shl_s64_vs
; GFX6: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX6: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX6: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX6: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
; GFX7-LABEL: name: shl_s64_vs
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX7: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX7: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
; GFX8-LABEL: name: shl_s64_vs
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
@ -277,8 +277,8 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0
; GFX10: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX10: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
%0:vgpr(s64) = COPY $vgpr0_vgpr1
%1:sgpr(s32) = COPY $sgpr0
%2:vgpr(s64) = G_SHL %0, %1
@ -296,13 +296,13 @@ body: |
; GFX6-LABEL: name: shl_s64_vv
; GFX6: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX6: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX6: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX6: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
; GFX7-LABEL: name: shl_s64_vv
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX7: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX7: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX7: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
; GFX8-LABEL: name: shl_s64_vv
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
@ -317,8 +317,8 @@ body: |
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: [[V_LSHLREV_B64_:%[0-9]+]]:vreg_64 = V_LSHLREV_B64 [[COPY1]], [[COPY]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHLREV_B64_]]
; GFX10: [[V_LSHL_B64_:%[0-9]+]]:vreg_64 = V_LSHL_B64 [[COPY]], [[COPY1]], implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_LSHL_B64_]]
%0:vgpr(s64) = COPY $vgpr0_vgpr1
%1:vgpr(s32) = COPY $vgpr2
%2:vgpr(s64) = G_SHL %0, %1