This patch returns proper value to indicate the case when instruction throughput can't be calculated.

Differential revision https://reviews.llvm.org/D35831

llvm-svn: 309156
This commit is contained in:
Andrew V. Tischenko 2017-07-26 18:55:14 +00:00
parent 7d7f0dc08b
commit d1fefa3d7c
8 changed files with 138 additions and 134 deletions

View File

@ -339,42 +339,46 @@ computeOutputLatency(const MachineInstr *DefMI, unsigned DefOperIdx,
static Optional<double>
getRThroughputFromItineraries(unsigned schedClass,
const InstrItineraryData *IID){
double Unknown = std::numeric_limits<double>::infinity();
double Throughput = Unknown;
Optional<double> Throughput;
for (const InstrStage *IS = IID->beginStage(schedClass),
*E = IID->endStage(schedClass);
IS != E; ++IS) {
unsigned Cycles = IS->getCycles();
if (!Cycles)
continue;
Throughput =
std::min(Throughput, countPopulation(IS->getUnits()) * 1.0 / Cycles);
if (IS->getCycles()) {
double Temp = countPopulation(IS->getUnits()) * 1.0 / IS->getCycles();
Throughput = Throughput.hasValue()
? std::min(Throughput.getValue(), Temp)
: Temp;
}
}
// We need reciprocal throughput that's why we return such value.
return 1 / Throughput;
if (Throughput.hasValue())
// We need reciprocal throughput that's why we return such value.
return 1 / Throughput.getValue();
return Throughput;
}
static Optional<double>
getRThroughputFromInstrSchedModel(const MCSchedClassDesc *SCDesc,
const TargetSubtargetInfo *STI,
const MCSchedModel &SchedModel) {
double Unknown = std::numeric_limits<double>::infinity();
double Throughput = Unknown;
Optional<double> Throughput;
for (const MCWriteProcResEntry *WPR = STI->getWriteProcResBegin(SCDesc),
*WEnd = STI->getWriteProcResEnd(SCDesc);
WPR != WEnd; ++WPR) {
unsigned Cycles = WPR->Cycles;
if (!Cycles)
return Optional<double>();
unsigned NumUnits =
SchedModel.getProcResource(WPR->ProcResourceIdx)->NumUnits;
Throughput = std::min(Throughput, NumUnits * 1.0 / Cycles);
if (WPR->Cycles) {
unsigned NumUnits =
SchedModel.getProcResource(WPR->ProcResourceIdx)->NumUnits;
double Temp = NumUnits * 1.0 / WPR->Cycles;
Throughput = Throughput.hasValue()
? std::min(Throughput.getValue(), Temp)
: Temp;
}
}
// We need reciprocal throughput that's why we return such value.
return 1 / Throughput;
if (Throughput.hasValue())
// We need reciprocal throughput that's why we return such value.
return 1 / Throughput.getValue();
return Throughput;
}
Optional<double>

View File

@ -867,14 +867,14 @@ define <4 x float> @test_extractf128(<8 x float> %a0, <8 x float> %a1, <4 x floa
; SANDY: # BB#0:
; SANDY-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:1.00]
; SANDY-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [5:1.00]
; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00]
; SANDY-NEXT: vzeroupper
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_extractf128:
; HASWELL: # BB#0:
; HASWELL-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00]
; HASWELL-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [1:0.00]
; HASWELL-NEXT: vzeroupper # sched: [1:?]
; HASWELL-NEXT: retq # sched: [1:1.00]
;
; BTVER2-LABEL: test_extractf128:
@ -887,7 +887,7 @@ define <4 x float> @test_extractf128(<8 x float> %a0, <8 x float> %a1, <4 x floa
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:0.50]
; ZNVER1-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:0.50]
; ZNVER1-NEXT: vzeroupper # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vzeroupper
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = shufflevector <8 x float> %a0, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
%2 = shufflevector <8 x float> %a1, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
@ -1098,15 +1098,15 @@ define <2 x double> @test_maskmovpd(i8* %a0, <2 x i64> %a1, <2 x double> %a2) {
;
; BTVER2-LABEL: test_maskmovpd:
; BTVER2: # BB#0:
; BTVER2-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [?:0.000000e+00]
; BTVER2-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [?:0.000000e+00]
; BTVER2-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2
; BTVER2-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi)
; BTVER2-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_maskmovpd:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2
; ZNVER1-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi)
; ZNVER1-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:0.50]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call <2 x double> @llvm.x86.avx.maskload.pd(i8* %a0, <2 x i64> %a1)
@ -1120,7 +1120,7 @@ define <4 x double> @test_maskmovpd_ymm(i8* %a0, <4 x i64> %a1, <4 x double> %a2
; SANDY-LABEL: test_maskmovpd_ymm:
; SANDY: # BB#0:
; SANDY-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [5:1.00]
; SANDY-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00]
; SANDY-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi)
; SANDY-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:1.00]
; SANDY-NEXT: retq # sched: [1:1.00]
;
@ -1133,15 +1133,15 @@ define <4 x double> @test_maskmovpd_ymm(i8* %a0, <4 x i64> %a1, <4 x double> %a2
;
; BTVER2-LABEL: test_maskmovpd_ymm:
; BTVER2: # BB#0:
; BTVER2-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00]
; BTVER2-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00]
; BTVER2-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2
; BTVER2-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi)
; BTVER2-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_maskmovpd_ymm:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2
; ZNVER1-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi)
; ZNVER1-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:0.50]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call <4 x double> @llvm.x86.avx.maskload.pd.256(i8* %a0, <4 x i64> %a1)
@ -1168,15 +1168,15 @@ define <4 x float> @test_maskmovps(i8* %a0, <4 x i32> %a1, <4 x float> %a2) {
;
; BTVER2-LABEL: test_maskmovps:
; BTVER2: # BB#0:
; BTVER2-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [?:0.000000e+00]
; BTVER2-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [?:0.000000e+00]
; BTVER2-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2
; BTVER2-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi)
; BTVER2-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_maskmovps:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2
; ZNVER1-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi)
; ZNVER1-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:0.50]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call <4 x float> @llvm.x86.avx.maskload.ps(i8* %a0, <4 x i32> %a1)
@ -1190,7 +1190,7 @@ define <8 x float> @test_maskmovps_ymm(i8* %a0, <8 x i32> %a1, <8 x float> %a2)
; SANDY-LABEL: test_maskmovps_ymm:
; SANDY: # BB#0:
; SANDY-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [1:0.50]
; SANDY-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00]
; SANDY-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi)
; SANDY-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:1.00]
; SANDY-NEXT: retq # sched: [1:1.00]
;
@ -1203,15 +1203,15 @@ define <8 x float> @test_maskmovps_ymm(i8* %a0, <8 x i32> %a1, <8 x float> %a2)
;
; BTVER2-LABEL: test_maskmovps_ymm:
; BTVER2: # BB#0:
; BTVER2-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00]
; BTVER2-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00]
; BTVER2-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2
; BTVER2-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi)
; BTVER2-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_maskmovps_ymm:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2
; ZNVER1-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi)
; ZNVER1-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:0.50]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call <8 x float> @llvm.x86.avx.maskload.ps.256(i8* %a0, <8 x i32> %a1)
@ -1452,13 +1452,13 @@ define i32 @test_movmskpd(<4 x double> %a0) {
; SANDY-LABEL: test_movmskpd:
; SANDY: # BB#0:
; SANDY-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00]
; SANDY-NEXT: vzeroupper
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_movmskpd:
; HASWELL: # BB#0:
; HASWELL-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
; HASWELL-NEXT: vzeroupper # sched: [1:0.00]
; HASWELL-NEXT: vzeroupper # sched: [1:?]
; HASWELL-NEXT: retq # sched: [1:1.00]
;
; BTVER2-LABEL: test_movmskpd:
@ -1469,7 +1469,7 @@ define i32 @test_movmskpd(<4 x double> %a0) {
; ZNVER1-LABEL: test_movmskpd:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vmovmskpd %ymm0, %eax # sched: [1:0.25]
; ZNVER1-NEXT: vzeroupper # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vzeroupper
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call i32 @llvm.x86.avx.movmsk.pd.256(<4 x double> %a0)
ret i32 %1
@ -1480,13 +1480,13 @@ define i32 @test_movmskps(<8 x float> %a0) {
; SANDY-LABEL: test_movmskps:
; SANDY: # BB#0:
; SANDY-NEXT: vmovmskps %ymm0, %eax # sched: [3:1.00]
; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00]
; SANDY-NEXT: vzeroupper
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_movmskps:
; HASWELL: # BB#0:
; HASWELL-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
; HASWELL-NEXT: vzeroupper # sched: [1:0.00]
; HASWELL-NEXT: vzeroupper # sched: [1:?]
; HASWELL-NEXT: retq # sched: [1:1.00]
;
; BTVER2-LABEL: test_movmskps:
@ -1497,7 +1497,7 @@ define i32 @test_movmskps(<8 x float> %a0) {
; ZNVER1-LABEL: test_movmskps:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vmovmskps %ymm0, %eax # sched: [1:0.25]
; ZNVER1-NEXT: vzeroupper # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vzeroupper
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call i32 @llvm.x86.avx.movmsk.ps.256(<8 x float> %a0)
ret i32 %1
@ -2499,7 +2499,7 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a
; SANDY-NEXT: setb %al # sched: [1:1.00]
; SANDY-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:1.00]
; SANDY-NEXT: adcl $0, %eax # sched: [1:0.33]
; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00]
; SANDY-NEXT: vzeroupper
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_testpd_ymm:
@ -2509,7 +2509,7 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a
; HASWELL-NEXT: setb %al # sched: [1:0.50]
; HASWELL-NEXT: vtestpd (%rdi), %ymm0 # sched: [5:0.50]
; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50]
; HASWELL-NEXT: vzeroupper # sched: [1:0.00]
; HASWELL-NEXT: vzeroupper # sched: [1:?]
; HASWELL-NEXT: retq # sched: [1:1.00]
;
; BTVER2-LABEL: test_testpd_ymm:
@ -2528,7 +2528,7 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a
; ZNVER1-NEXT: setb %al # sched: [1:0.25]
; ZNVER1-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:0.50]
; ZNVER1-NEXT: adcl $0, %eax # sched: [1:0.25]
; ZNVER1-NEXT: vzeroupper # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vzeroupper
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call i32 @llvm.x86.avx.vtestc.pd.256(<4 x double> %a0, <4 x double> %a1)
%2 = load <4 x double>, <4 x double> *%a2, align 32
@ -2590,7 +2590,7 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2)
; SANDY-NEXT: setb %al # sched: [1:1.00]
; SANDY-NEXT: vtestps (%rdi), %ymm0 # sched: [8:1.00]
; SANDY-NEXT: adcl $0, %eax # sched: [1:0.33]
; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00]
; SANDY-NEXT: vzeroupper
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_testps_ymm:
@ -2600,7 +2600,7 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2)
; HASWELL-NEXT: setb %al # sched: [1:0.50]
; HASWELL-NEXT: vtestps (%rdi), %ymm0 # sched: [5:0.50]
; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50]
; HASWELL-NEXT: vzeroupper # sched: [1:0.00]
; HASWELL-NEXT: vzeroupper # sched: [1:?]
; HASWELL-NEXT: retq # sched: [1:1.00]
;
; BTVER2-LABEL: test_testps_ymm:
@ -2619,7 +2619,7 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2)
; ZNVER1-NEXT: setb %al # sched: [1:0.25]
; ZNVER1-NEXT: vtestps (%rdi), %ymm0 # sched: [8:0.50]
; ZNVER1-NEXT: adcl $0, %eax # sched: [1:0.25]
; ZNVER1-NEXT: vzeroupper # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vzeroupper
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call i32 @llvm.x86.avx.vtestc.ps.256(<8 x float> %a0, <8 x float> %a1)
%2 = load <8 x float>, <8 x float> *%a2, align 32
@ -2840,22 +2840,22 @@ define <8 x float> @test_xorps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a
define void @test_zeroall() {
; SANDY-LABEL: test_zeroall:
; SANDY: # BB#0:
; SANDY-NEXT: vzeroall # sched: [?:0.000000e+00]
; SANDY-NEXT: vzeroall
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_zeroall:
; HASWELL: # BB#0:
; HASWELL-NEXT: vzeroall # sched: [1:0.00]
; HASWELL-NEXT: vzeroall # sched: [1:?]
; HASWELL-NEXT: retq # sched: [1:1.00]
;
; BTVER2-LABEL: test_zeroall:
; BTVER2: # BB#0:
; BTVER2-NEXT: vzeroall # sched: [?:0.000000e+00]
; BTVER2-NEXT: vzeroall
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_zeroall:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vzeroall # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vzeroall
; ZNVER1-NEXT: retq # sched: [5:0.50]
call void @llvm.x86.avx.vzeroall()
ret void
@ -2865,22 +2865,22 @@ declare void @llvm.x86.avx.vzeroall() nounwind
define void @test_zeroupper() {
; SANDY-LABEL: test_zeroupper:
; SANDY: # BB#0:
; SANDY-NEXT: vzeroupper # sched: [?:0.000000e+00]
; SANDY-NEXT: vzeroupper
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_zeroupper:
; HASWELL: # BB#0:
; HASWELL-NEXT: vzeroupper # sched: [1:0.00]
; HASWELL-NEXT: vzeroupper # sched: [1:?]
; HASWELL-NEXT: retq # sched: [1:1.00]
;
; BTVER2-LABEL: test_zeroupper:
; BTVER2: # BB#0:
; BTVER2-NEXT: vzeroupper # sched: [?:0.000000e+00]
; BTVER2-NEXT: vzeroupper
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_zeroupper:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vzeroupper # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vzeroupper
; ZNVER1-NEXT: retq # sched: [5:0.50]
call void @llvm.x86.avx.vzeroupper()
ret void

View File

@ -139,15 +139,15 @@ define i32 @test_bextr_i32(i32 %a0, i32 %a1, i32 *%a2) {
;
; BTVER2-LABEL: test_bextr_i32:
; BTVER2: # BB#0:
; BTVER2-NEXT: bextrl %edi, (%rdx), %ecx # sched: [?:0.000000e+00]
; BTVER2-NEXT: bextrl %edi, %esi, %eax # sched: [?:0.000000e+00]
; BTVER2-NEXT: bextrl %edi, (%rdx), %ecx
; BTVER2-NEXT: bextrl %edi, %esi, %eax
; BTVER2-NEXT: addl %ecx, %eax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_bextr_i32:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: bextrl %edi, (%rdx), %ecx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: bextrl %edi, %esi, %eax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: bextrl %edi, (%rdx), %ecx
; ZNVER1-NEXT: bextrl %edi, %esi, %eax
; ZNVER1-NEXT: addl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i32, i32 *%a2
@ -175,15 +175,15 @@ define i64 @test_bextr_i64(i64 %a0, i64 %a1, i64 *%a2) {
;
; BTVER2-LABEL: test_bextr_i64:
; BTVER2: # BB#0:
; BTVER2-NEXT: bextrq %rdi, (%rdx), %rcx # sched: [?:0.000000e+00]
; BTVER2-NEXT: bextrq %rdi, %rsi, %rax # sched: [?:0.000000e+00]
; BTVER2-NEXT: bextrq %rdi, (%rdx), %rcx
; BTVER2-NEXT: bextrq %rdi, %rsi, %rax
; BTVER2-NEXT: addq %rcx, %rax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_bextr_i64:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: bextrq %rdi, (%rdx), %rcx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: bextrq %rdi, %rsi, %rax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: bextrq %rdi, (%rdx), %rcx
; ZNVER1-NEXT: bextrq %rdi, %rsi, %rax
; ZNVER1-NEXT: addq %rcx, %rax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i64, i64 *%a2
@ -211,15 +211,15 @@ define i32 @test_blsi_i32(i32 %a0, i32 *%a1) {
;
; BTVER2-LABEL: test_blsi_i32:
; BTVER2: # BB#0:
; BTVER2-NEXT: blsil (%rsi), %ecx # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsil %edi, %eax # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsil (%rsi), %ecx
; BTVER2-NEXT: blsil %edi, %eax
; BTVER2-NEXT: addl %ecx, %eax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_blsi_i32:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: blsil (%rsi), %ecx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsil %edi, %eax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsil (%rsi), %ecx
; ZNVER1-NEXT: blsil %edi, %eax
; ZNVER1-NEXT: addl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i32, i32 *%a1
@ -248,15 +248,15 @@ define i64 @test_blsi_i64(i64 %a0, i64 *%a1) {
;
; BTVER2-LABEL: test_blsi_i64:
; BTVER2: # BB#0:
; BTVER2-NEXT: blsiq (%rsi), %rcx # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsiq %rdi, %rax # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsiq (%rsi), %rcx
; BTVER2-NEXT: blsiq %rdi, %rax
; BTVER2-NEXT: addq %rcx, %rax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_blsi_i64:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: blsiq (%rsi), %rcx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsiq %rdi, %rax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsiq (%rsi), %rcx
; ZNVER1-NEXT: blsiq %rdi, %rax
; ZNVER1-NEXT: addq %rcx, %rax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i64, i64 *%a1
@ -285,15 +285,15 @@ define i32 @test_blsmsk_i32(i32 %a0, i32 *%a1) {
;
; BTVER2-LABEL: test_blsmsk_i32:
; BTVER2: # BB#0:
; BTVER2-NEXT: blsmskl (%rsi), %ecx # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsmskl %edi, %eax # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsmskl (%rsi), %ecx
; BTVER2-NEXT: blsmskl %edi, %eax
; BTVER2-NEXT: addl %ecx, %eax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_blsmsk_i32:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: blsmskl (%rsi), %ecx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsmskl %edi, %eax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsmskl (%rsi), %ecx
; ZNVER1-NEXT: blsmskl %edi, %eax
; ZNVER1-NEXT: addl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i32, i32 *%a1
@ -322,15 +322,15 @@ define i64 @test_blsmsk_i64(i64 %a0, i64 *%a1) {
;
; BTVER2-LABEL: test_blsmsk_i64:
; BTVER2: # BB#0:
; BTVER2-NEXT: blsmskq (%rsi), %rcx # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsmskq %rdi, %rax # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsmskq (%rsi), %rcx
; BTVER2-NEXT: blsmskq %rdi, %rax
; BTVER2-NEXT: addq %rcx, %rax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_blsmsk_i64:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: blsmskq (%rsi), %rcx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsmskq %rdi, %rax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsmskq (%rsi), %rcx
; ZNVER1-NEXT: blsmskq %rdi, %rax
; ZNVER1-NEXT: addq %rcx, %rax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i64, i64 *%a1
@ -359,15 +359,15 @@ define i32 @test_blsr_i32(i32 %a0, i32 *%a1) {
;
; BTVER2-LABEL: test_blsr_i32:
; BTVER2: # BB#0:
; BTVER2-NEXT: blsrl (%rsi), %ecx # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsrl %edi, %eax # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsrl (%rsi), %ecx
; BTVER2-NEXT: blsrl %edi, %eax
; BTVER2-NEXT: addl %ecx, %eax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_blsr_i32:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: blsrl (%rsi), %ecx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsrl %edi, %eax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsrl (%rsi), %ecx
; ZNVER1-NEXT: blsrl %edi, %eax
; ZNVER1-NEXT: addl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i32, i32 *%a1
@ -396,15 +396,15 @@ define i64 @test_blsr_i64(i64 %a0, i64 *%a1) {
;
; BTVER2-LABEL: test_blsr_i64:
; BTVER2: # BB#0:
; BTVER2-NEXT: blsrq (%rsi), %rcx # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsrq %rdi, %rax # sched: [?:0.000000e+00]
; BTVER2-NEXT: blsrq (%rsi), %rcx
; BTVER2-NEXT: blsrq %rdi, %rax
; BTVER2-NEXT: addq %rcx, %rax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_blsr_i64:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: blsrq (%rsi), %rcx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsrq %rdi, %rax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: blsrq (%rsi), %rcx
; ZNVER1-NEXT: blsrq %rdi, %rax
; ZNVER1-NEXT: addq %rcx, %rax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i64, i64 *%a1
@ -435,16 +435,16 @@ define i16 @test_cttz_i16(i16 zeroext %a0, i16 *%a1) {
;
; BTVER2-LABEL: test_cttz_i16:
; BTVER2: # BB#0:
; BTVER2-NEXT: tzcntw (%rsi), %cx # sched: [?:0.000000e+00]
; BTVER2-NEXT: tzcntw %di, %ax # sched: [?:0.000000e+00]
; BTVER2-NEXT: tzcntw (%rsi), %cx
; BTVER2-NEXT: tzcntw %di, %ax
; BTVER2-NEXT: orl %ecx, %eax # sched: [1:0.50]
; BTVER2-NEXT: # kill: %AX<def> %AX<kill> %EAX<kill>
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_cttz_i16:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: tzcntw (%rsi), %cx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: tzcntw %di, %ax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: tzcntw (%rsi), %cx
; ZNVER1-NEXT: tzcntw %di, %ax
; ZNVER1-NEXT: orl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: # kill: %AX<def> %AX<kill> %EAX<kill>
; ZNVER1-NEXT: retq # sched: [5:0.50]
@ -473,15 +473,15 @@ define i32 @test_cttz_i32(i32 %a0, i32 *%a1) {
;
; BTVER2-LABEL: test_cttz_i32:
; BTVER2: # BB#0:
; BTVER2-NEXT: tzcntl (%rsi), %ecx # sched: [?:0.000000e+00]
; BTVER2-NEXT: tzcntl %edi, %eax # sched: [?:0.000000e+00]
; BTVER2-NEXT: tzcntl (%rsi), %ecx
; BTVER2-NEXT: tzcntl %edi, %eax
; BTVER2-NEXT: orl %ecx, %eax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_cttz_i32:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: tzcntl (%rsi), %ecx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: tzcntl %edi, %eax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: tzcntl (%rsi), %ecx
; ZNVER1-NEXT: tzcntl %edi, %eax
; ZNVER1-NEXT: orl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i32, i32 *%a1
@ -509,15 +509,15 @@ define i64 @test_cttz_i64(i64 %a0, i64 *%a1) {
;
; BTVER2-LABEL: test_cttz_i64:
; BTVER2: # BB#0:
; BTVER2-NEXT: tzcntq (%rsi), %rcx # sched: [?:0.000000e+00]
; BTVER2-NEXT: tzcntq %rdi, %rax # sched: [?:0.000000e+00]
; BTVER2-NEXT: tzcntq (%rsi), %rcx
; BTVER2-NEXT: tzcntq %rdi, %rax
; BTVER2-NEXT: orq %rcx, %rax # sched: [1:0.50]
; BTVER2-NEXT: retq # sched: [4:1.00]
;
; ZNVER1-LABEL: test_cttz_i64:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: tzcntq (%rsi), %rcx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: tzcntq %rdi, %rax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: tzcntq (%rsi), %rcx
; ZNVER1-NEXT: tzcntq %rdi, %rax
; ZNVER1-NEXT: orq %rcx, %rax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i64, i64 *%a1

View File

@ -22,8 +22,8 @@ define i32 @test_bzhi_i32(i32 %a0, i32 %a1, i32 *%a2) {
;
; ZNVER1-LABEL: test_bzhi_i32:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: bzhil %edi, (%rdx), %ecx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: bzhil %edi, %esi, %eax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: bzhil %edi, (%rdx), %ecx
; ZNVER1-NEXT: bzhil %edi, %esi, %eax
; ZNVER1-NEXT: addl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i32, i32 *%a2
@ -51,8 +51,8 @@ define i64 @test_bzhi_i64(i64 %a0, i64 %a1, i64 *%a2) {
;
; ZNVER1-LABEL: test_bzhi_i64:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: bzhiq %rdi, (%rdx), %rcx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: bzhiq %rdi, %rsi, %rax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: bzhiq %rdi, (%rdx), %rcx
; ZNVER1-NEXT: bzhiq %rdi, %rsi, %rax
; ZNVER1-NEXT: addq %rcx, %rax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i64, i64 *%a2
@ -80,8 +80,8 @@ define i32 @test_pdep_i32(i32 %a0, i32 %a1, i32 *%a2) {
;
; ZNVER1-LABEL: test_pdep_i32:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: pdepl (%rdx), %edi, %ecx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: pdepl %esi, %edi, %eax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: pdepl (%rdx), %edi, %ecx
; ZNVER1-NEXT: pdepl %esi, %edi, %eax
; ZNVER1-NEXT: addl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i32, i32 *%a2
@ -109,8 +109,8 @@ define i64 @test_pdep_i64(i64 %a0, i64 %a1, i64 *%a2) {
;
; ZNVER1-LABEL: test_pdep_i64:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: pdepq (%rdx), %rdi, %rcx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: pdepq %rsi, %rdi, %rax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: pdepq (%rdx), %rdi, %rcx
; ZNVER1-NEXT: pdepq %rsi, %rdi, %rax
; ZNVER1-NEXT: addq %rcx, %rax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i64, i64 *%a2
@ -138,8 +138,8 @@ define i32 @test_pext_i32(i32 %a0, i32 %a1, i32 *%a2) {
;
; ZNVER1-LABEL: test_pext_i32:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: pextl (%rdx), %edi, %ecx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: pextl %esi, %edi, %eax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: pextl (%rdx), %edi, %ecx
; ZNVER1-NEXT: pextl %esi, %edi, %eax
; ZNVER1-NEXT: addl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i32, i32 *%a2
@ -167,8 +167,8 @@ define i64 @test_pext_i64(i64 %a0, i64 %a1, i64 *%a2) {
;
; ZNVER1-LABEL: test_pext_i64:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: pextq (%rdx), %rdi, %rcx # sched: [?:0.000000e+00]
; ZNVER1-NEXT: pextq %rsi, %rdi, %rax # sched: [?:0.000000e+00]
; ZNVER1-NEXT: pextq (%rdx), %rdi, %rcx
; ZNVER1-NEXT: pextq %rsi, %rdi, %rax
; ZNVER1-NEXT: addq %rcx, %rax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = load i64, i64 *%a2

View File

@ -114,14 +114,14 @@ define <8 x i16> @test_vcvtps2ph_256(<8 x float> %a0, <8 x float> %a1, <8 x i16>
; IVY: # BB#0:
; IVY-NEXT: vcvtps2ph $0, %ymm0, %xmm0 # sched: [3:1.00]
; IVY-NEXT: vcvtps2ph $0, %ymm1, (%rdi) # sched: [7:1.00]
; IVY-NEXT: vzeroupper # sched: [?:0.000000e+00]
; IVY-NEXT: vzeroupper
; IVY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_vcvtps2ph_256:
; HASWELL: # BB#0:
; HASWELL-NEXT: vcvtps2ph $0, %ymm0, %xmm0 # sched: [4:1.00]
; HASWELL-NEXT: vcvtps2ph $0, %ymm1, (%rdi) # sched: [8:1.00]
; HASWELL-NEXT: vzeroupper # sched: [1:0.00]
; HASWELL-NEXT: vzeroupper # sched: [1:?]
; HASWELL-NEXT: retq # sched: [1:1.00]
;
; BTVER2-LABEL: test_vcvtps2ph_256:
@ -134,7 +134,7 @@ define <8 x i16> @test_vcvtps2ph_256(<8 x float> %a0, <8 x float> %a1, <8 x i16>
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vcvtps2ph $0, %ymm0, %xmm0 # sched: [5:1.00]
; ZNVER1-NEXT: vcvtps2ph $0, %ymm1, (%rdi) # sched: [12:1.00]
; ZNVER1-NEXT: vzeroupper # sched: [?:0.000000e+00]
; ZNVER1-NEXT: vzeroupper
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %a0, i32 0)
%2 = call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %a1, i32 0)

View File

@ -394,8 +394,8 @@ define <8 x i16> @test_mpsadbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
;
; ZNVER1-LABEL: test_mpsadbw:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [100:?]
; ZNVER1-NEXT: vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [100:?]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i8 7)
%2 = bitcast <8 x i16> %1 to <16 x i8>

View File

@ -323,11 +323,11 @@ define i32 @test_pcmpestri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
; ZNVER1: # BB#0:
; ZNVER1-NEXT: movl $7, %eax # sched: [1:0.25]
; ZNVER1-NEXT: movl $7, %edx # sched: [1:0.25]
; ZNVER1-NEXT: vpcmpestri $7, %xmm1, %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vpcmpestri $7, %xmm1, %xmm0 # sched: [100:?]
; ZNVER1-NEXT: movl $7, %eax # sched: [1:0.25]
; ZNVER1-NEXT: movl $7, %edx # sched: [1:0.25]
; ZNVER1-NEXT: movl %ecx, %esi # sched: [1:0.25]
; ZNVER1-NEXT: vpcmpestri $7, (%rdi), %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vpcmpestri $7, (%rdi), %xmm0 # sched: [100:?]
; ZNVER1-NEXT: # kill: %ECX<def> %ECX<kill> %RCX<def>
; ZNVER1-NEXT: leal (%rcx,%rsi), %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
@ -394,10 +394,10 @@ define <16 x i8> @test_pcmpestrm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
; ZNVER1: # BB#0:
; ZNVER1-NEXT: movl $7, %eax # sched: [1:0.25]
; ZNVER1-NEXT: movl $7, %edx # sched: [1:0.25]
; ZNVER1-NEXT: vpcmpestrm $7, %xmm1, %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vpcmpestrm $7, %xmm1, %xmm0 # sched: [100:?]
; ZNVER1-NEXT: movl $7, %eax # sched: [1:0.25]
; ZNVER1-NEXT: movl $7, %edx # sched: [1:0.25]
; ZNVER1-NEXT: vpcmpestrm $7, (%rdi), %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vpcmpestrm $7, (%rdi), %xmm0 # sched: [100:?]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call <16 x i8> @llvm.x86.sse42.pcmpestrm128(<16 x i8> %a0, i32 7, <16 x i8> %a1, i32 7, i8 7)
%2 = load <16 x i8>, <16 x i8> *%a2, align 16
@ -454,9 +454,9 @@ define i32 @test_pcmpistri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
;
; ZNVER1-LABEL: test_pcmpistri:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vpcmpistri $7, %xmm1, %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vpcmpistri $7, %xmm1, %xmm0 # sched: [100:?]
; ZNVER1-NEXT: movl %ecx, %eax # sched: [1:0.25]
; ZNVER1-NEXT: vpcmpistri $7, (%rdi), %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vpcmpistri $7, (%rdi), %xmm0 # sched: [100:?]
; ZNVER1-NEXT: # kill: %ECX<def> %ECX<kill> %RCX<def>
; ZNVER1-NEXT: leal (%rcx,%rax), %eax # sched: [1:0.25]
; ZNVER1-NEXT: retq # sched: [5:0.50]
@ -501,8 +501,8 @@ define <16 x i8> @test_pcmpistrm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
;
; ZNVER1-LABEL: test_pcmpistrm:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: vpcmpistrm $7, %xmm1, %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vpcmpistrm $7, (%rdi), %xmm0 # sched: [100:0.00]
; ZNVER1-NEXT: vpcmpistrm $7, %xmm1, %xmm0 # sched: [100:?]
; ZNVER1-NEXT: vpcmpistrm $7, (%rdi), %xmm0 # sched: [100:?]
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = call <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8> %a0, <16 x i8> %a1, i8 7)
%2 = load <16 x i8>, <16 x i8> *%a2, align 16

View File

@ -16,7 +16,7 @@ define <2 x i64> @test_extrq(<2 x i64> %a0, <16 x i8> %a1) {
;
; ZNVER1-LABEL: test_extrq:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: extrq %xmm1, %xmm0 # sched: [?:0.000000e+00]
; ZNVER1-NEXT: extrq %xmm1, %xmm0
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = tail call <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64> %a0, <16 x i8> %a1)
ret <2 x i64> %1
@ -36,7 +36,7 @@ define <2 x i64> @test_extrqi(<2 x i64> %a0) {
;
; ZNVER1-LABEL: test_extrqi:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: extrq $2, $3, %xmm0 # sched: [?:0.000000e+00]
; ZNVER1-NEXT: extrq $2, $3, %xmm0
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = tail call <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64> %a0, i8 3, i8 2)
ret <2 x i64> %1
@ -56,7 +56,7 @@ define <2 x i64> @test_insertq(<2 x i64> %a0, <2 x i64> %a1) {
;
; ZNVER1-LABEL: test_insertq:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: insertq %xmm1, %xmm0 # sched: [?:0.000000e+00]
; ZNVER1-NEXT: insertq %xmm1, %xmm0
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = tail call <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64> %a0, <2 x i64> %a1)
ret <2 x i64> %1
@ -76,7 +76,7 @@ define <2 x i64> @test_insertqi(<2 x i64> %a0, <2 x i64> %a1) {
;
; ZNVER1-LABEL: test_insertqi:
; ZNVER1: # BB#0:
; ZNVER1-NEXT: insertq $6, $5, %xmm1, %xmm0 # sched: [?:0.000000e+00]
; ZNVER1-NEXT: insertq $6, $5, %xmm1, %xmm0
; ZNVER1-NEXT: retq # sched: [5:0.50]
%1 = tail call <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64> %a0, <2 x i64> %a1, i8 5, i8 6)
ret <2 x i64> %1