RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-31 12:41:49 +01:00
Code Issues Projects Releases Wiki Activity

[X86] Fix VZEROUPPER scheduling info on SNB,HSW,BDW,SXL,SKX.

Browse Source 
Summary:
Starting from SNB, VZEROUPPER is handled by the renamer and uses no proc resources.
After HSW, it also has zero latency.

This fixes PR35606.

To reproduce:
Uops:
  llvm-exegesis -mode=uops -opcode-name=VZEROUPPER
Latency:
  echo -e '#LLVM-EXEGESIS-DEFREG XMM0 1\n#LLVM-EXEGESIS-DEFREG XMM1 1\nvzeroupper' | /tmp/llvm-exegesis -mode=latency -snippets-file=-
  echo -e '#LLVM-EXEGESIS-DEFREG XMM0 1\n#LLVM-EXEGESIS-DEFREG XMM1 1\nvzeroupper\naddps %xmm0, %xmm1' | /tmp/llvm-exegesis -mode=latency -snippets-file=-

Reviewers: RKSimon, craig.topper, andreadb

Subscribers: gbedwell, llvm-commits

Differential Revision: https://reviews.llvm.org/D54107

llvm-svn: 346482
This commit is contained in:
Clement Courbet 2018-11-09 09:49:06 +00:00
parent 2629fa33cc
commit ee1472f937
19 changed files with 255 additions and 248 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
lib/Target/X86/X86SchedBroadwell.td
View File

@ -878,10 +878,10 @@ def BWWriteResGroup45 : SchedWriteRes<[BWPort0156]> {
}
def: InstRW<[BWWriteResGroup45], (instrs FNCLEX)>;
def BWWriteResGroup46 : SchedWriteRes<[BWPort015,BWPort0156]> {
let Latency = 4;
def BWWriteResGroup46 : SchedWriteRes<[]> {
let Latency = 0;
let NumMicroOps = 4;
let ResourceCycles = [1,3];
let ResourceCycles = [];
}
def: InstRW<[BWWriteResGroup46], (instrs VZEROUPPER)>;

6
lib/Target/X86/X86SchedHaswell.td
View File

@ -1408,10 +1408,10 @@ def HWWriteResGroup81 : SchedWriteRes<[HWPort0156]> {
}
def: InstRW<[HWWriteResGroup81], (instrs FNCLEX)>;
def HWWriteResGroup82 : SchedWriteRes<[HWPort015,HWPort0156]> {
let Latency = 4;
def HWWriteResGroup82 : SchedWriteRes<[]> {
let Latency = 0;
let NumMicroOps = 4;
let ResourceCycles = [1,3];
let ResourceCycles = [];
}
def: InstRW<[HWWriteResGroup82], (instrs VZEROUPPER)>;

7
lib/Target/X86/X86SchedSandyBridge.td
View File

@ -1112,6 +1112,13 @@ def SBWriteResGroupVzeroall : SchedWriteRes<[SBPort5]> {
}
def: InstRW<[SBWriteResGroupVzeroall], (instrs VZEROALL)>;
def SBWriteResGroupVzeroupper : SchedWriteRes<[]> {
let Latency = 1;
let NumMicroOps = 4;
let ResourceCycles = [];
}
def: InstRW<[SBWriteResGroupVzeroupper], (instrs VZEROUPPER)>;
def: InstRW<[WriteZero], (instrs CLC)>;
// Intruction variants handled by the renamer. These might not need execution

6
lib/Target/X86/X86SchedSkylakeClient.td
View File

@ -897,10 +897,10 @@ def SKLWriteResGroup55 : SchedWriteRes<[SKLPort6,SKLPort0156]> {
}
def: InstRW<[SKLWriteResGroup55], (instrs PAUSE)>;
def SKLWriteResGroup56 : SchedWriteRes<[SKLPort015,SKLPort0156]> {
let Latency = 4;
def SKLWriteResGroup56 : SchedWriteRes<[]> {
let Latency = 0;
let NumMicroOps = 4;
let ResourceCycles = [1,3];
let ResourceCycles = [];
}
def: InstRW<[SKLWriteResGroup56], (instrs VZEROUPPER)>;

6
lib/Target/X86/X86SchedSkylakeServer.td
View File

@ -1009,10 +1009,10 @@ def SKXWriteResGroup55 : SchedWriteRes<[SKXPort0156]> {
}
def: InstRW<[SKXWriteResGroup55], (instrs FNCLEX)>;
def SKXWriteResGroup56 : SchedWriteRes<[SKXPort015,SKXPort0156]> {
let Latency = 4;
def SKXWriteResGroup56 : SchedWriteRes<[]> {
let Latency = 0;
let NumMicroOps = 4;
let ResourceCycles = [1,3];
let ResourceCycles = [];
}
def: InstRW<[SKXWriteResGroup56], (instrs VZEROUPPER)>;

84
test/CodeGen/X86/avx-schedule.ll
View File

@ -1843,42 +1843,42 @@ define <4 x float> @test_extractf128(<8 x float> %a0, <8 x float> %a1, <4 x floa
; GENERIC: # %bb.0:
; GENERIC-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:1.00]
; GENERIC-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SANDY-LABEL: test_extractf128:
; SANDY: # %bb.0:
; SANDY-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:1.00]
; SANDY-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; SANDY-NEXT: vzeroupper # sched: [100:0.33]
; SANDY-NEXT: vzeroupper # sched: [1:1.00]
; 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: [1:1.00]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_extractf128:
; BROADWELL: # %bb.0:
; BROADWELL-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00]
; BROADWELL-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_extractf128:
; SKYLAKE: # %bb.0:
; SKYLAKE-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00]
; SKYLAKE-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_extractf128:
; SKX: # %bb.0:
; SKX-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00]
; SKX-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; BDVER2-LABEL: test_extractf128:
@ -3008,37 +3008,37 @@ define i32 @test_movmskpd(<4 x double> %a0) {
; GENERIC-LABEL: test_movmskpd:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SANDY-LABEL: test_movmskpd:
; SANDY: # %bb.0:
; SANDY-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
; SANDY-NEXT: vzeroupper # sched: [100:0.33]
; SANDY-NEXT: vzeroupper # sched: [1:1.00]
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_movmskpd:
; HASWELL: # %bb.0:
; HASWELL-NEXT: vmovmskpd %ymm0, %eax # sched: [3:1.00]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_movmskpd:
; BROADWELL: # %bb.0:
; BROADWELL-NEXT: vmovmskpd %ymm0, %eax # sched: [3:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_movmskpd:
; SKYLAKE: # %bb.0:
; SKYLAKE-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_movmskpd:
; SKX: # %bb.0:
; SKX-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; BDVER2-LABEL: test_movmskpd:
@ -3066,37 +3066,37 @@ define i32 @test_movmskps(<8 x float> %a0) {
; GENERIC-LABEL: test_movmskps:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SANDY-LABEL: test_movmskps:
; SANDY: # %bb.0:
; SANDY-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
; SANDY-NEXT: vzeroupper # sched: [100:0.33]
; SANDY-NEXT: vzeroupper # sched: [1:1.00]
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_movmskps:
; HASWELL: # %bb.0:
; HASWELL-NEXT: vmovmskps %ymm0, %eax # sched: [3:1.00]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_movmskps:
; BROADWELL: # %bb.0:
; BROADWELL-NEXT: vmovmskps %ymm0, %eax # sched: [3:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_movmskps:
; SKYLAKE: # %bb.0:
; SKYLAKE-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_movmskps:
; SKX: # %bb.0:
; SKX-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; BDVER2-LABEL: test_movmskps:
@ -3126,7 +3126,7 @@ define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
; GENERIC-NEXT: #APP
; GENERIC-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SANDY-LABEL: test_movntdq:
@ -3134,7 +3134,7 @@ define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
; SANDY-NEXT: #APP
; SANDY-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
; SANDY-NEXT: #NO_APP
; SANDY-NEXT: vzeroupper # sched: [100:0.33]
; SANDY-NEXT: vzeroupper # sched: [1:1.00]
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_movntdq:
@ -3142,7 +3142,7 @@ define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
; HASWELL-NEXT: #APP
; HASWELL-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_movntdq:
@ -3150,7 +3150,7 @@ define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
; BROADWELL-NEXT: #APP
; BROADWELL-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_movntdq:
@ -3158,7 +3158,7 @@ define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
; SKYLAKE-NEXT: #APP
; SKYLAKE-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_movntdq:
@ -3166,7 +3166,7 @@ define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
; SKX-NEXT: #APP
; SKX-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; BDVER2-LABEL: test_movntdq:
@ -5240,7 +5240,7 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a
; GENERIC-NEXT: setb %al # sched: [1:0.50]
; GENERIC-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:1.00]
; GENERIC-NEXT: adcl $0, %eax # sched: [2:0.67]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SANDY-LABEL: test_testpd_ymm:
@ -5250,7 +5250,7 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a
; SANDY-NEXT: setb %al # sched: [1:0.50]
; SANDY-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:1.00]
; SANDY-NEXT: adcl $0, %eax # sched: [2:0.67]
; SANDY-NEXT: vzeroupper # sched: [100:0.33]
; SANDY-NEXT: vzeroupper # sched: [1:1.00]
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_testpd_ymm:
@ -5260,7 +5260,7 @@ 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: [8:1.00]
; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_testpd_ymm:
@ -5270,7 +5270,7 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a
; BROADWELL-NEXT: setb %al # sched: [1:0.50]
; BROADWELL-NEXT: vtestpd (%rdi), %ymm0 # sched: [7:1.00]
; BROADWELL-NEXT: adcl $0, %eax # sched: [1:0.50]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_testpd_ymm:
@ -5280,7 +5280,7 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a
; SKYLAKE-NEXT: setb %al # sched: [1:0.50]
; SKYLAKE-NEXT: vtestpd (%rdi), %ymm0 # sched: [9:1.00]
; SKYLAKE-NEXT: adcl $0, %eax # sched: [1:0.50]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_testpd_ymm:
@ -5290,7 +5290,7 @@ define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a
; SKX-NEXT: setb %al # sched: [1:0.50]
; SKX-NEXT: vtestpd (%rdi), %ymm0 # sched: [9:1.00]
; SKX-NEXT: adcl $0, %eax # sched: [1:0.50]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; BDVER2-LABEL: test_testpd_ymm:
@ -5426,7 +5426,7 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2)
; GENERIC-NEXT: setb %al # sched: [1:0.50]
; GENERIC-NEXT: vtestps (%rdi), %ymm0 # sched: [8:1.00]
; GENERIC-NEXT: adcl $0, %eax # sched: [2:0.67]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SANDY-LABEL: test_testps_ymm:
@ -5436,7 +5436,7 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2)
; SANDY-NEXT: setb %al # sched: [1:0.50]
; SANDY-NEXT: vtestps (%rdi), %ymm0 # sched: [8:1.00]
; SANDY-NEXT: adcl $0, %eax # sched: [2:0.67]
; SANDY-NEXT: vzeroupper # sched: [100:0.33]
; SANDY-NEXT: vzeroupper # sched: [1:1.00]
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_testps_ymm:
@ -5446,7 +5446,7 @@ 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: [8:1.00]
; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_testps_ymm:
@ -5456,7 +5456,7 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2)
; BROADWELL-NEXT: setb %al # sched: [1:0.50]
; BROADWELL-NEXT: vtestps (%rdi), %ymm0 # sched: [7:1.00]
; BROADWELL-NEXT: adcl $0, %eax # sched: [1:0.50]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_testps_ymm:
@ -5466,7 +5466,7 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2)
; SKYLAKE-NEXT: setb %al # sched: [1:0.50]
; SKYLAKE-NEXT: vtestps (%rdi), %ymm0 # sched: [9:1.00]
; SKYLAKE-NEXT: adcl $0, %eax # sched: [1:0.50]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_testps_ymm:
@ -5476,7 +5476,7 @@ define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2)
; SKX-NEXT: setb %al # sched: [1:0.50]
; SKX-NEXT: vtestps (%rdi), %ymm0 # sched: [9:1.00]
; SKX-NEXT: adcl $0, %eax # sched: [1:0.50]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; BDVER2-LABEL: test_testps_ymm:
@ -5976,32 +5976,32 @@ declare void @llvm.x86.avx.vzeroall() nounwind
define void @test_zeroupper() {
; GENERIC-LABEL: test_zeroupper:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SANDY-LABEL: test_zeroupper:
; SANDY: # %bb.0:
; SANDY-NEXT: vzeroupper # sched: [100:0.33]
; SANDY-NEXT: vzeroupper # sched: [1:1.00]
; SANDY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_zeroupper:
; HASWELL: # %bb.0:
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_zeroupper:
; BROADWELL: # %bb.0:
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_zeroupper:
; SKYLAKE: # %bb.0:
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_zeroupper:
; SKX: # %bb.0:
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; BDVER2-LABEL: test_zeroupper:

40
test/CodeGen/X86/avx2-schedule.ll
View File

@ -177,7 +177,7 @@ define <4 x i32> @test_extracti128(<8 x i16> %a0, <4 x i32> *%a1) {
; GENERIC-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:1.00]
; GENERIC-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [1:1.00]
; GENERIC-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_extracti128:
@ -185,7 +185,7 @@ define <4 x i32> @test_extracti128(<8 x i16> %a0, <4 x i32> *%a1) {
; HASWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
; HASWELL-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00]
; HASWELL-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_extracti128:
@ -193,7 +193,7 @@ define <4 x i32> @test_extracti128(<8 x i16> %a0, <4 x i32> *%a1) {
; BROADWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
; BROADWELL-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00]
; BROADWELL-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_extracti128:
@ -201,7 +201,7 @@ define <4 x i32> @test_extracti128(<8 x i16> %a0, <4 x i32> *%a1) {
; SKYLAKE-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
; SKYLAKE-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00]
; SKYLAKE-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_extracti128:
@ -209,7 +209,7 @@ define <4 x i32> @test_extracti128(<8 x i16> %a0, <4 x i32> *%a1) {
; SKX-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
; SKX-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00]
; SKX-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_extracti128:
@ -474,31 +474,31 @@ define <4 x float> @test_gatherqps_ymm(<4 x float> %a0, i8* %a1, <4 x i64> %a2,
; GENERIC-LABEL: test_gatherqps_ymm:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [5:0.50]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_gatherqps_ymm:
; HASWELL: # %bb.0:
; HASWELL-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [28:3.67]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_gatherqps_ymm:
; BROADWELL: # %bb.0:
; BROADWELL-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [24:5.00]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_gatherqps_ymm:
; SKYLAKE: # %bb.0:
; SKYLAKE-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [25:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_gatherqps_ymm:
; SKX: # %bb.0:
; SKX-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [25:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_gatherqps_ymm:
@ -2920,31 +2920,31 @@ define <4 x i32> @test_pgatherqd_ymm(<4 x i32> %a0, i8* %a1, <4 x i64> %a2, <4 x
; GENERIC-LABEL: test_pgatherqd_ymm:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [5:0.50]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_pgatherqd_ymm:
; HASWELL: # %bb.0:
; HASWELL-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [28:5.00]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_pgatherqd_ymm:
; BROADWELL: # %bb.0:
; BROADWELL-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [5:0.50]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_pgatherqd_ymm:
; SKYLAKE: # %bb.0:
; SKYLAKE-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [25:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_pgatherqd_ymm:
; SKX: # %bb.0:
; SKX-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [25:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_pgatherqd_ymm:
@ -4089,31 +4089,31 @@ define i32 @test_pmovmskb(<32 x i8> %a0) {
; GENERIC-LABEL: test_pmovmskb:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_pmovmskb:
; HASWELL: # %bb.0:
; HASWELL-NEXT: vpmovmskb %ymm0, %eax # sched: [3:1.00]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_pmovmskb:
; BROADWELL: # %bb.0:
; BROADWELL-NEXT: vpmovmskb %ymm0, %eax # sched: [3:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_pmovmskb:
; SKYLAKE: # %bb.0:
; SKYLAKE-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; SKX-LABEL: test_pmovmskb:
; SKX: # %bb.0:
; SKX-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_pmovmskb:

140
test/CodeGen/X86/avx512-schedule.ll
View File

@ -1447,13 +1447,13 @@ define <4 x float> @slto4f32(<4 x i64> %a) {
; GENERIC-LABEL: slto4f32:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvtqq2ps %ymm0, %xmm0 # sched: [3:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: slto4f32:
; SKX: # %bb.0:
; SKX-NEXT: vcvtqq2ps %ymm0, %xmm0 # sched: [7:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%b = sitofp <4 x i64> %a to <4 x float>
ret <4 x float> %b
@ -1463,13 +1463,13 @@ define <4 x float> @ulto4f32(<4 x i64> %a) {
; GENERIC-LABEL: ulto4f32:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvtuqq2ps %ymm0, %xmm0 # sched: [3:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: ulto4f32:
; SKX: # %bb.0:
; SKX-NEXT: vcvtuqq2ps %ymm0, %xmm0 # sched: [7:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%b = uitofp <4 x i64> %a to <4 x float>
ret <4 x float> %b
@ -1538,14 +1538,14 @@ define <16 x i8> @f32to16uc(<16 x float> %f) {
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvttps2dq %zmm0, %zmm0 # sched: [3:1.00]
; GENERIC-NEXT: vpmovdb %zmm0, %xmm0 # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: f32to16uc:
; SKX: # %bb.0:
; SKX-NEXT: vcvttps2dq %zmm0, %zmm0 # sched: [4:0.50]
; SKX-NEXT: vpmovdb %zmm0, %xmm0 # sched: [4:2.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%res = fptoui <16 x float> %f to <16 x i8>
ret <16 x i8> %res
@ -1614,14 +1614,14 @@ define <8 x i16> @f64to8us(<8 x double> %f) {
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvttpd2dq %zmm0, %ymm0 # sched: [4:1.00]
; GENERIC-NEXT: vpmovdw %ymm0, %xmm0 # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: f64to8us:
; SKX: # %bb.0:
; SKX-NEXT: vcvttpd2dq %zmm0, %ymm0 # sched: [7:1.00]
; SKX-NEXT: vpmovdw %ymm0, %xmm0 # sched: [4:2.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%res = fptoui <8 x double> %f to <8 x i16>
ret <8 x i16> %res
@ -1632,14 +1632,14 @@ define <8 x i8> @f64to8uc(<8 x double> %f) {
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvttpd2dq %zmm0, %ymm0 # sched: [4:1.00]
; GENERIC-NEXT: vpmovdw %ymm0, %xmm0 # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: f64to8uc:
; SKX: # %bb.0:
; SKX-NEXT: vcvttpd2dq %zmm0, %ymm0 # sched: [7:1.00]
; SKX-NEXT: vpmovdw %ymm0, %xmm0 # sched: [4:2.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%res = fptoui <8 x double> %f to <8 x i8>
ret <8 x i8> %res
@ -1649,13 +1649,13 @@ define <4 x i32> @f64to4ui(<4 x double> %a) nounwind {
; GENERIC-LABEL: f64to4ui:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvttpd2udq %ymm0, %xmm0 # sched: [4:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: f64to4ui:
; SKX: # %bb.0:
; SKX-NEXT: vcvttpd2udq %ymm0, %xmm0 # sched: [7:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%b = fptoui <4 x double> %a to <4 x i32>
ret <4 x i32> %b
@ -1737,13 +1737,13 @@ define <4 x i32> @f64to4si(<4 x double> %a) {
; GENERIC-LABEL: f64to4si:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [4:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: f64to4si:
; SKX: # %bb.0:
; SKX-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [7:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%b = fptosi <4 x double> %a to <4 x i32>
ret <4 x i32> %b
@ -1771,13 +1771,13 @@ define <4 x float> @f64to4f32(<4 x double> %b) {
; GENERIC-LABEL: f64to4f32:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [4:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: f64to4f32:
; SKX: # %bb.0:
; SKX-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [7:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%a = fptrunc <4 x double> %b to <4 x float>
ret <4 x float> %a
@ -1789,7 +1789,7 @@ define <4 x float> @f64to4f32_mask(<4 x double> %b, <4 x i1> %mask) {
; GENERIC-NEXT: vpslld $31, %xmm1, %xmm1 # sched: [1:1.00]
; GENERIC-NEXT: vpmovd2m %xmm1, %k1 # sched: [1:0.33]
; GENERIC-NEXT: vcvtpd2ps %ymm0, %xmm0 {%k1} {z} # sched: [4:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: f64to4f32_mask:
@ -1797,7 +1797,7 @@ define <4 x float> @f64to4f32_mask(<4 x double> %b, <4 x i1> %mask) {
; SKX-NEXT: vpslld $31, %xmm1, %xmm1 # sched: [1:0.50]
; SKX-NEXT: vpmovd2m %xmm1, %k1 # sched: [1:1.00]
; SKX-NEXT: vcvtpd2ps %ymm0, %xmm0 {%k1} {z} # sched: [7:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%a = fptrunc <4 x double> %b to <4 x float>
%c = select <4 x i1>%mask, <4 x float>%a, <4 x float> zeroinitializer
@ -4307,7 +4307,7 @@ define i16 @trunc_16i32_to_16i1(<16 x i32> %a) {
; GENERIC-NEXT: vpmovd2m %zmm0, %k0 # sched: [1:0.33]
; GENERIC-NEXT: kmovd %k0, %eax # sched: [1:0.33]
; GENERIC-NEXT: # kill: def $ax killed $ax killed $eax
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: trunc_16i32_to_16i1:
@ -4316,7 +4316,7 @@ define i16 @trunc_16i32_to_16i1(<16 x i32> %a) {
; SKX-NEXT: vpmovd2m %zmm0, %k0 # sched: [1:1.00]
; SKX-NEXT: kmovd %k0, %eax # sched: [3:1.00]
; SKX-NEXT: # kill: def $ax killed $ax killed $eax
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%mask_b = trunc <16 x i32>%a to <16 x i1>
%mask = bitcast <16 x i1> %mask_b to i16
@ -4422,14 +4422,14 @@ define <8 x i16> @sext_8i1_8i16(<8 x i32> %a1, <8 x i32> %a2) nounwind {
; GENERIC: # %bb.0:
; GENERIC-NEXT: vpcmpgtd %ymm0, %ymm1, %k0 # sched: [1:0.50]
; GENERIC-NEXT: vpmovm2w %k0, %xmm0 # sched: [1:0.33]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: sext_8i1_8i16:
; SKX: # %bb.0:
; SKX-NEXT: vpcmpgtd %ymm0, %ymm1, %k0 # sched: [3:1.00]
; SKX-NEXT: vpmovm2w %k0, %xmm0 # sched: [1:0.25]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%x = icmp slt <8 x i32> %a1, %a2
%y = sext <8 x i1> %x to <8 x i16>
@ -4475,14 +4475,14 @@ define void @extload_v8i64(<8 x i8>* %a, <8 x i64>* %res) {
; GENERIC: # %bb.0:
; GENERIC-NEXT: vpmovsxbq (%rdi), %zmm0 # sched: [8:1.00]
; GENERIC-NEXT: vmovdqa64 %zmm0, (%rsi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: extload_v8i64:
; SKX: # %bb.0:
; SKX-NEXT: vpmovsxbq (%rdi), %zmm0 # sched: [10:1.00]
; SKX-NEXT: vmovdqa64 %zmm0, (%rsi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%sign_load = load <8 x i8>, <8 x i8>* %a
%c = sext <8 x i8> %sign_load to <8 x i64>
@ -6124,13 +6124,13 @@ define void @mov_test18(i8 * %addr, <8 x i64> %data) {
; GENERIC-LABEL: mov_test18:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: mov_test18:
; SKX: # %bb.0:
; SKX-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%vaddr = bitcast i8* %addr to <8 x i64>*
store <8 x i64>%data, <8 x i64>* %vaddr, align 64
@ -6141,13 +6141,13 @@ define void @mov_test19(i8 * %addr, <16 x i32> %data) {
; GENERIC-LABEL: mov_test19:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovups %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: mov_test19:
; SKX: # %bb.0:
; SKX-NEXT: vmovups %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%vaddr = bitcast i8* %addr to <16 x i32>*
store <16 x i32>%data, <16 x i32>* %vaddr, align 1
@ -6158,13 +6158,13 @@ define void @mov_test20(i8 * %addr, <16 x i32> %data) {
; GENERIC-LABEL: mov_test20:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: mov_test20:
; SKX: # %bb.0:
; SKX-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%vaddr = bitcast i8* %addr to <16 x i32>*
store <16 x i32>%data, <16 x i32>* %vaddr, align 64
@ -6190,13 +6190,13 @@ define void @mov_test22(i8 * %addr, <8 x i64> %data) {
; GENERIC-LABEL: mov_test22:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovups %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: mov_test22:
; SKX: # %bb.0:
; SKX-NEXT: vmovups %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%vaddr = bitcast i8* %addr to <8 x i64>*
store <8 x i64>%data, <8 x i64>* %vaddr, align 1
@ -6222,13 +6222,13 @@ define void @mov_test24(i8 * %addr, <8 x double> %data) {
; GENERIC-LABEL: mov_test24:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: mov_test24:
; SKX: # %bb.0:
; SKX-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%vaddr = bitcast i8* %addr to <8 x double>*
store <8 x double>%data, <8 x double>* %vaddr, align 64
@ -6254,13 +6254,13 @@ define void @mov_test26(i8 * %addr, <16 x float> %data) {
; GENERIC-LABEL: mov_test26:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: mov_test26:
; SKX: # %bb.0:
; SKX-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%vaddr = bitcast i8* %addr to <16 x float>*
store <16 x float>%data, <16 x float>* %vaddr, align 64
@ -6286,13 +6286,13 @@ define void @mov_test28(i8 * %addr, <8 x double> %data) {
; GENERIC-LABEL: mov_test28:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovups %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: mov_test28:
; SKX: # %bb.0:
; SKX-NEXT: vmovups %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%vaddr = bitcast i8* %addr to <8 x double>*
store <8 x double>%data, <8 x double>* %vaddr, align 1
@ -6318,13 +6318,13 @@ define void @mov_test30(i8 * %addr, <16 x float> %data) {
; GENERIC-LABEL: mov_test30:
; GENERIC: # %bb.0:
; GENERIC-NEXT: vmovups %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: mov_test30:
; SKX: # %bb.0:
; SKX-NEXT: vmovups %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%vaddr = bitcast i8* %addr to <16 x float>*
store <16 x float>%data, <16 x float>* %vaddr, align 1
@ -6877,7 +6877,7 @@ define i32 @zext_test1(<16 x i32> %a, <16 x i32> %b) {
; GENERIC-NEXT: kshiftrw $5, %k0, %k0 # sched: [1:1.00]
; GENERIC-NEXT: kmovd %k0, %eax # sched: [1:0.33]
; GENERIC-NEXT: andl $1, %eax # sched: [1:0.33]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: zext_test1:
@ -6886,7 +6886,7 @@ define i32 @zext_test1(<16 x i32> %a, <16 x i32> %b) {
; SKX-NEXT: kshiftrw $5, %k0, %k0 # sched: [3:1.00]
; SKX-NEXT: kmovd %k0, %eax # sched: [3:1.00]
; SKX-NEXT: andl $1, %eax # sched: [1:0.25]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%cmp_res = icmp ugt <16 x i32> %a, %b
%cmp_res.i1 = extractelement <16 x i1> %cmp_res, i32 5
@ -6902,7 +6902,7 @@ define i16 @zext_test2(<16 x i32> %a, <16 x i32> %b) {
; GENERIC-NEXT: kmovd %k0, %eax # sched: [1:0.33]
; GENERIC-NEXT: andl $1, %eax # sched: [1:0.33]
; GENERIC-NEXT: # kill: def $ax killed $ax killed $eax
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: zext_test2:
@ -6912,7 +6912,7 @@ define i16 @zext_test2(<16 x i32> %a, <16 x i32> %b) {
; SKX-NEXT: kmovd %k0, %eax # sched: [3:1.00]
; SKX-NEXT: andl $1, %eax # sched: [1:0.25]
; SKX-NEXT: # kill: def $ax killed $ax killed $eax
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%cmp_res = icmp ugt <16 x i32> %a, %b
%cmp_res.i1 = extractelement <16 x i1> %cmp_res, i32 5
@ -6928,7 +6928,7 @@ define i8 @zext_test3(<16 x i32> %a, <16 x i32> %b) {
; GENERIC-NEXT: kmovd %k0, %eax # sched: [1:0.33]
; GENERIC-NEXT: andb $1, %al # sched: [1:0.33]
; GENERIC-NEXT: # kill: def $al killed $al killed $eax
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: zext_test3:
@ -6938,7 +6938,7 @@ define i8 @zext_test3(<16 x i32> %a, <16 x i32> %b) {
; SKX-NEXT: kmovd %k0, %eax # sched: [3:1.00]
; SKX-NEXT: andb $1, %al # sched: [1:0.25]
; SKX-NEXT: # kill: def $al killed $al killed $eax
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%cmp_res = icmp ugt <16 x i32> %a, %b
%cmp_res.i1 = extractelement <16 x i1> %cmp_res, i32 5
@ -6976,7 +6976,7 @@ define <4 x i32> @test4(<4 x i64> %x, <4 x i64> %y, <4 x i64> %x1, <4 x i64> %y1
; GENERIC-NEXT: vpcmpleq %ymm1, %ymm0, %k1 # sched: [1:0.50]
; GENERIC-NEXT: vpcmpgtq %ymm3, %ymm2, %k0 {%k1} # sched: [1:0.50]
; GENERIC-NEXT: vpmovm2d %k0, %xmm0 # sched: [1:0.33]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: test4:
@ -6984,7 +6984,7 @@ define <4 x i32> @test4(<4 x i64> %x, <4 x i64> %y, <4 x i64> %x1, <4 x i64> %y1
; SKX-NEXT: vpcmpleq %ymm1, %ymm0, %k1 # sched: [3:1.00]
; SKX-NEXT: vpcmpgtq %ymm3, %ymm2, %k0 {%k1} # sched: [3:1.00]
; SKX-NEXT: vpmovm2d %k0, %xmm0 # sched: [1:0.25]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%x_gt_y = icmp sgt <4 x i64> %x, %y
%x1_gt_y1 = icmp sgt <4 x i64> %x1, %y1
@ -7061,13 +7061,13 @@ define <16 x i8> @vcmp_test8(<16 x i32>%a, <16 x i32>%b, i32 %a1, i32 %b1) {
; GENERIC-NEXT: # %bb.2:
; GENERIC-NEXT: kxorw %k0, %k0, %k0 # sched: [1:0.33]
; GENERIC-NEXT: vpmovm2b %k0, %xmm0 # sched: [1:0.33]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
; GENERIC-NEXT: .LBB386_1:
; GENERIC-NEXT: vpxor %xmm1, %xmm1, %xmm1 # sched: [0:0.25]
; GENERIC-NEXT: vpcmpgtd %zmm1, %zmm0, %k0 # sched: [1:0.50]
; GENERIC-NEXT: vpmovm2b %k0, %xmm0 # sched: [1:0.33]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: vcmp_test8:
@ -7077,13 +7077,13 @@ define <16 x i8> @vcmp_test8(<16 x i32>%a, <16 x i32>%b, i32 %a1, i32 %b1) {
; SKX-NEXT: # %bb.2:
; SKX-NEXT: kxorw %k0, %k0, %k0 # sched: [1:1.00]
; SKX-NEXT: vpmovm2b %k0, %xmm0 # sched: [1:0.25]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
; SKX-NEXT: .LBB386_1:
; SKX-NEXT: vpxor %xmm1, %xmm1, %xmm1 # sched: [1:0.33]
; SKX-NEXT: vpcmpgtd %zmm1, %zmm0, %k0 # sched: [3:1.00]
; SKX-NEXT: vpmovm2b %k0, %xmm0 # sched: [1:0.25]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%cond = icmp sgt i32 %a1, %b1
%cmp1 = icmp sgt <16 x i32> %a, zeroinitializer
@ -7617,11 +7617,11 @@ define void @ktest_1(<8 x double> %in, double * %base) {
; GENERIC-NEXT: je .LBB410_2 # sched: [1:1.00]
; GENERIC-NEXT: # %bb.1: # %L1
; GENERIC-NEXT: vmovapd %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
; GENERIC-NEXT: .LBB410_2: # %L2
; GENERIC-NEXT: vmovapd %zmm0, 8(%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: ktest_1:
@ -7634,11 +7634,11 @@ define void @ktest_1(<8 x double> %in, double * %base) {
; SKX-NEXT: je .LBB410_2 # sched: [1:0.50]
; SKX-NEXT: # %bb.1: # %L1
; SKX-NEXT: vmovapd %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
; SKX-NEXT: .LBB410_2: # %L2
; SKX-NEXT: vmovapd %zmm0, 8(%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%addr1 = getelementptr double, double * %base, i64 0
%addr2 = getelementptr double, double * %base, i64 1
@ -7686,12 +7686,12 @@ define void @ktest_2(<32 x float> %in, float * %base) {
; GENERIC-NEXT: # %bb.1: # %L1
; GENERIC-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vmovaps %zmm1, 64(%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
; GENERIC-NEXT: .LBB411_2: # %L2
; GENERIC-NEXT: vmovaps %zmm0, 4(%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vmovaps %zmm1, 68(%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: ktest_2:
@ -7711,12 +7711,12 @@ define void @ktest_2(<32 x float> %in, float * %base) {
; SKX-NEXT: # %bb.1: # %L1
; SKX-NEXT: vmovaps %zmm0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vmovaps %zmm1, 64(%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
; SKX-NEXT: .LBB411_2: # %L2
; SKX-NEXT: vmovaps %zmm0, 4(%rdi) # sched: [1:1.00]
; SKX-NEXT: vmovaps %zmm1, 68(%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%addr1 = getelementptr float, float * %base, i64 0
%addr2 = getelementptr float, float * %base, i64 1
@ -7908,7 +7908,7 @@ define void @store_32i1(<32 x i1>* %a, <32 x i1> %v) {
; GENERIC-NEXT: vpsllw $7, %ymm0, %ymm0 # sched: [1:1.00]
; GENERIC-NEXT: vpmovb2m %ymm0, %k0 # sched: [1:0.33]
; GENERIC-NEXT: kmovd %k0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: store_32i1:
@ -7916,7 +7916,7 @@ define void @store_32i1(<32 x i1>* %a, <32 x i1> %v) {
; SKX-NEXT: vpsllw $7, %ymm0, %ymm0 # sched: [1:0.50]
; SKX-NEXT: vpmovb2m %ymm0, %k0 # sched: [1:1.00]
; SKX-NEXT: kmovd %k0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
store <32 x i1> %v, <32 x i1>* %a
ret void
@ -7928,7 +7928,7 @@ define void @store_32i1_1(<32 x i1>* %a, <32 x i16> %v) {
; GENERIC-NEXT: vpsllw $15, %zmm0, %zmm0 # sched: [1:1.00]
; GENERIC-NEXT: vpmovw2m %zmm0, %k0 # sched: [1:0.33]
; GENERIC-NEXT: kmovd %k0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: store_32i1_1:
@ -7936,7 +7936,7 @@ define void @store_32i1_1(<32 x i1>* %a, <32 x i16> %v) {
; SKX-NEXT: vpsllw $15, %zmm0, %zmm0 # sched: [1:1.00]
; SKX-NEXT: vpmovw2m %zmm0, %k0 # sched: [1:1.00]
; SKX-NEXT: kmovd %k0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%v1 = trunc <32 x i16> %v to <32 x i1>
store <32 x i1> %v1, <32 x i1>* %a
@ -7951,7 +7951,7 @@ define void @store_64i1(<64 x i1>* %a, <64 x i1> %v) {
; GENERIC-NEXT: vpsllw $7, %zmm0, %zmm0 # sched: [1:1.00]
; GENERIC-NEXT: vpmovb2m %zmm0, %k0 # sched: [1:0.33]
; GENERIC-NEXT: kmovq %k0, (%rdi) # sched: [1:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: store_64i1:
@ -7959,7 +7959,7 @@ define void @store_64i1(<64 x i1>* %a, <64 x i1> %v) {
; SKX-NEXT: vpsllw $7, %zmm0, %zmm0 # sched: [1:1.00]
; SKX-NEXT: vpmovb2m %zmm0, %k0 # sched: [1:1.00]
; SKX-NEXT: kmovq %k0, (%rdi) # sched: [1:1.00]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
store <64 x i1> %v, <64 x i1>* %a
ret void
@ -7971,7 +7971,7 @@ define i32 @test_bitcast_v8i1_zext(<16 x i32> %a) {
; GENERIC-NEXT: vptestnmd %zmm0, %zmm0, %k0 # sched: [1:0.33]
; GENERIC-NEXT: kmovb %k0, %eax # sched: [1:0.33]
; GENERIC-NEXT: addl %eax, %eax # sched: [1:0.33]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: test_bitcast_v8i1_zext:
@ -7979,7 +7979,7 @@ define i32 @test_bitcast_v8i1_zext(<16 x i32> %a) {
; SKX-NEXT: vptestnmd %zmm0, %zmm0, %k0 # sched: [3:1.00]
; SKX-NEXT: kmovb %k0, %eax # sched: [3:1.00]
; SKX-NEXT: addl %eax, %eax # sched: [1:0.25]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%v1 = icmp eq <16 x i32> %a, zeroinitializer
%mask = shufflevector <16 x i1> %v1, <16 x i1> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@ -7995,7 +7995,7 @@ define i32 @test_bitcast_v16i1_zext(<16 x i32> %a) {
; GENERIC-NEXT: vptestnmd %zmm0, %zmm0, %k0 # sched: [1:0.33]
; GENERIC-NEXT: kmovw %k0, %eax # sched: [1:0.33]
; GENERIC-NEXT: addl %eax, %eax # sched: [1:0.33]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; SKX-LABEL: test_bitcast_v16i1_zext:
@ -8003,7 +8003,7 @@ define i32 @test_bitcast_v16i1_zext(<16 x i32> %a) {
; SKX-NEXT: vptestnmd %zmm0, %zmm0, %k0 # sched: [3:1.00]
; SKX-NEXT: kmovw %k0, %eax # sched: [3:1.00]
; SKX-NEXT: addl %eax, %eax # sched: [1:0.25]
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
%v1 = icmp eq <16 x i32> %a, zeroinitializer
%mask1 = bitcast <16 x i1> %v1 to i16

8
test/CodeGen/X86/avx512vpopcntdq-schedule.ll
View File

@ -18,7 +18,7 @@ define void @test_vpopcntd(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> *%a2, i16
; GENERIC-NEXT: vpopcntd (%rdi){1to16}, %zmm0 {%k1} # sched: [8:0.50]
; GENERIC-NEXT: vpopcntd (%rdi){1to16}, %zmm0 {%k1} {z} # sched: [8:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; ICELAKE-LABEL: test_vpopcntd:
@ -35,7 +35,7 @@ define void @test_vpopcntd(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> *%a2, i16
; ICELAKE-NEXT: vpopcntd (%rdi){1to16}, %zmm0 {%k1} # sched: [8:1.00]
; ICELAKE-NEXT: vpopcntd (%rdi){1to16}, %zmm0 {%k1} {z} # sched: [8:1.00]
; ICELAKE-NEXT: #NO_APP
; ICELAKE-NEXT: vzeroupper # sched: [4:1.00]
; ICELAKE-NEXT: vzeroupper # sched: [0:0.67]
; ICELAKE-NEXT: retq # sched: [7:1.00]
tail call void asm "vpopcntd $1, $0 \0A\09 vpopcntd $1, $0 {$3} \0A\09 vpopcntd $1, $0 {$3} {z} \0A\09 vpopcntd $2, $0 \0A\09 vpopcntd $2, $0 {$3} \0A\09 vpopcntd $2, $0 {$3} {z} \0A\09 vpopcntd $2{1to16}, $0 \0A\09 vpopcntd $2{1to16}, $0 {$3} \0A\09 vpopcntd $2{1to16}, $0 {$3} {z}", "v,v,*m,^Yk"(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> *%a2, i16 %a3) nounwind
ret void
@ -56,7 +56,7 @@ define void @test_vpopcntq(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> *%a2, i8 %a3)
; GENERIC-NEXT: vpopcntq (%rdi){1to8}, %zmm0 {%k1} # sched: [8:0.50]
; GENERIC-NEXT: vpopcntq (%rdi){1to8}, %zmm0 {%k1} {z} # sched: [8:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; ICELAKE-LABEL: test_vpopcntq:
@ -73,7 +73,7 @@ define void @test_vpopcntq(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> *%a2, i8 %a3)
; ICELAKE-NEXT: vpopcntq (%rdi){1to8}, %zmm0 {%k1} # sched: [8:1.00]
; ICELAKE-NEXT: vpopcntq (%rdi){1to8}, %zmm0 {%k1} {z} # sched: [8:1.00]
; ICELAKE-NEXT: #NO_APP
; ICELAKE-NEXT: vzeroupper # sched: [4:1.00]
; ICELAKE-NEXT: vzeroupper # sched: [0:0.67]
; ICELAKE-NEXT: retq # sched: [7:1.00]
tail call void asm "vpopcntq $1, $0 \0A\09 vpopcntq $1, $0 {$3} \0A\09 vpopcntq $1, $0 {$3} {z} \0A\09 vpopcntq $2, $0 \0A\09 vpopcntq $2, $0 {$3} \0A\09 vpopcntq $2, $0 {$3} {z} \0A\09 vpopcntq $2{1to8}, $0 \0A\09 vpopcntq $2{1to8}, $0 {$3} \0A\09 vpopcntq $2{1to8}, $0 {$3} {z}", "v,v,*m,^Yk"(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> *%a2, i8 %a3) nounwind
ret void

10
test/CodeGen/X86/f16c-schedule.ll
View File

@ -197,35 +197,35 @@ define <8 x i16> @test_vcvtps2ph_256(<8 x float> %a0, <8 x float> %a1, <8 x i16>
; GENERIC: # %bb.0:
; GENERIC-NEXT: vcvtps2ph $0, %ymm0, %xmm0 # sched: [3:1.00]
; GENERIC-NEXT: vcvtps2ph $0, %ymm1, (%rdi) # sched: [4:1.00]
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; IVY-LABEL: test_vcvtps2ph_256:
; IVY: # %bb.0:
; IVY-NEXT: vcvtps2ph $0, %ymm0, %xmm0 # sched: [3:1.00]
; IVY-NEXT: vcvtps2ph $0, %ymm1, (%rdi) # sched: [4:1.00]
; IVY-NEXT: vzeroupper # sched: [100:0.33]
; IVY-NEXT: vzeroupper # sched: [1:1.00]
; IVY-NEXT: retq # sched: [1:1.00]
;
; HASWELL-LABEL: test_vcvtps2ph_256:
; HASWELL: # %bb.0:
; HASWELL-NEXT: vcvtps2ph $0, %ymm0, %xmm0 # sched: [6:1.00]
; HASWELL-NEXT: vcvtps2ph $0, %ymm1, (%rdi) # sched: [7:1.00]
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vcvtps2ph_256:
; BROADWELL: # %bb.0:
; BROADWELL-NEXT: vcvtps2ph $0, %ymm0, %xmm0 # sched: [6:1.00]
; BROADWELL-NEXT: vcvtps2ph $0, %ymm1, (%rdi) # sched: [7:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vcvtps2ph_256:
; SKYLAKE: # %bb.0:
; SKYLAKE-NEXT: vcvtps2ph $0, %ymm0, %xmm0 # sched: [7:1.00]
; SKYLAKE-NEXT: vcvtps2ph $0, %ymm1, (%rdi) # sched: [8:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; BDVER2-LABEL: test_vcvtps2ph_256:

120
test/CodeGen/X86/fma-schedule.ll
View File

@ -123,7 +123,7 @@ define void @test_vfmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; GENERIC-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [10:0.50]
; GENERIC-NEXT: vfmadd231pd {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfmaddpd_256:
@ -149,7 +149,7 @@ define void @test_vfmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; HASWELL-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [12:0.50]
; HASWELL-NEXT: vfmadd231pd {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfmaddpd_256:
@ -162,7 +162,7 @@ define void @test_vfmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; BROADWELL-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [11:0.50]
; BROADWELL-NEXT: vfmadd231pd {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfmaddpd_256:
@ -175,7 +175,7 @@ define void @test_vfmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; SKYLAKE-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [11:0.50]
; SKYLAKE-NEXT: vfmadd231pd {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfmaddpd_256:
@ -200,7 +200,7 @@ define void @test_vfmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; SKX-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [11:0.50]
; SKX-NEXT: vfmadd231pd {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfmaddpd_256:
@ -330,7 +330,7 @@ define void @test_vfmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; GENERIC-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [10:0.50]
; GENERIC-NEXT: vfmadd231ps {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfmaddps_256:
@ -356,7 +356,7 @@ define void @test_vfmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; HASWELL-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [12:0.50]
; HASWELL-NEXT: vfmadd231ps {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfmaddps_256:
@ -369,7 +369,7 @@ define void @test_vfmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; BROADWELL-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [11:0.50]
; BROADWELL-NEXT: vfmadd231ps {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfmaddps_256:
@ -382,7 +382,7 @@ define void @test_vfmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; SKYLAKE-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [11:0.50]
; SKYLAKE-NEXT: vfmadd231ps {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfmaddps_256:
@ -407,7 +407,7 @@ define void @test_vfmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; SKX-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + mem sched: [11:0.50]
; SKX-NEXT: vfmadd231ps {{.*#+}} ymm0 = (ymm1 * mem) + ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfmaddps_256:
@ -741,7 +741,7 @@ define void @test_vfmaddsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; GENERIC-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [10:0.50]
; GENERIC-NEXT: vfmaddsub231pd {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfmaddsubpd_256:
@ -767,7 +767,7 @@ define void @test_vfmaddsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; HASWELL-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [12:0.50]
; HASWELL-NEXT: vfmaddsub231pd {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfmaddsubpd_256:
@ -780,7 +780,7 @@ define void @test_vfmaddsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; BROADWELL-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [11:0.50]
; BROADWELL-NEXT: vfmaddsub231pd {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfmaddsubpd_256:
@ -793,7 +793,7 @@ define void @test_vfmaddsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; SKYLAKE-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [11:0.50]
; SKYLAKE-NEXT: vfmaddsub231pd {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfmaddsubpd_256:
@ -818,7 +818,7 @@ define void @test_vfmaddsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; SKX-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [11:0.50]
; SKX-NEXT: vfmaddsub231pd {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfmaddsubpd_256:
@ -948,7 +948,7 @@ define void @test_vfmaddsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; GENERIC-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [10:0.50]
; GENERIC-NEXT: vfmaddsub231ps {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfmaddsubps_256:
@ -974,7 +974,7 @@ define void @test_vfmaddsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; HASWELL-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [12:0.50]
; HASWELL-NEXT: vfmaddsub231ps {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfmaddsubps_256:
@ -987,7 +987,7 @@ define void @test_vfmaddsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; BROADWELL-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [11:0.50]
; BROADWELL-NEXT: vfmaddsub231ps {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfmaddsubps_256:
@ -1000,7 +1000,7 @@ define void @test_vfmaddsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; SKYLAKE-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [11:0.50]
; SKYLAKE-NEXT: vfmaddsub231ps {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfmaddsubps_256:
@ -1025,7 +1025,7 @@ define void @test_vfmaddsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; SKX-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) +/- mem sched: [11:0.50]
; SKX-NEXT: vfmaddsub231ps {{.*#+}} ymm0 = (ymm1 * mem) +/- ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfmaddsubps_256:
@ -1159,7 +1159,7 @@ define void @test_vfmsubaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; GENERIC-NEXT: vfmsubadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [10:0.50]
; GENERIC-NEXT: vfmsubadd231pd {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfmsubaddpd_256:
@ -1185,7 +1185,7 @@ define void @test_vfmsubaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; HASWELL-NEXT: vfmsubadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [12:0.50]
; HASWELL-NEXT: vfmsubadd231pd {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfmsubaddpd_256:
@ -1198,7 +1198,7 @@ define void @test_vfmsubaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; BROADWELL-NEXT: vfmsubadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [11:0.50]
; BROADWELL-NEXT: vfmsubadd231pd {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfmsubaddpd_256:
@ -1211,7 +1211,7 @@ define void @test_vfmsubaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; SKYLAKE-NEXT: vfmsubadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [11:0.50]
; SKYLAKE-NEXT: vfmsubadd231pd {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfmsubaddpd_256:
@ -1236,7 +1236,7 @@ define void @test_vfmsubaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; SKX-NEXT: vfmsubadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [11:0.50]
; SKX-NEXT: vfmsubadd231pd {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfmsubaddpd_256:
@ -1366,7 +1366,7 @@ define void @test_vfmsubaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; GENERIC-NEXT: vfmsubadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [10:0.50]
; GENERIC-NEXT: vfmsubadd231ps {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfmsubaddps_256:
@ -1392,7 +1392,7 @@ define void @test_vfmsubaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; HASWELL-NEXT: vfmsubadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [12:0.50]
; HASWELL-NEXT: vfmsubadd231ps {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfmsubaddps_256:
@ -1405,7 +1405,7 @@ define void @test_vfmsubaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; BROADWELL-NEXT: vfmsubadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [11:0.50]
; BROADWELL-NEXT: vfmsubadd231ps {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfmsubaddps_256:
@ -1418,7 +1418,7 @@ define void @test_vfmsubaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; SKYLAKE-NEXT: vfmsubadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [11:0.50]
; SKYLAKE-NEXT: vfmsubadd231ps {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfmsubaddps_256:
@ -1443,7 +1443,7 @@ define void @test_vfmsubaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; SKX-NEXT: vfmsubadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) -/+ mem sched: [11:0.50]
; SKX-NEXT: vfmsubadd231ps {{.*#+}} ymm0 = (ymm1 * mem) -/+ ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfmsubaddps_256:
@ -1577,7 +1577,7 @@ define void @test_vfmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; GENERIC-NEXT: vfmsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [10:0.50]
; GENERIC-NEXT: vfmsub231pd {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfmsubpd_256:
@ -1603,7 +1603,7 @@ define void @test_vfmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; HASWELL-NEXT: vfmsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [12:0.50]
; HASWELL-NEXT: vfmsub231pd {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfmsubpd_256:
@ -1616,7 +1616,7 @@ define void @test_vfmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; BROADWELL-NEXT: vfmsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [11:0.50]
; BROADWELL-NEXT: vfmsub231pd {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfmsubpd_256:
@ -1629,7 +1629,7 @@ define void @test_vfmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; SKYLAKE-NEXT: vfmsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [11:0.50]
; SKYLAKE-NEXT: vfmsub231pd {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfmsubpd_256:
@ -1654,7 +1654,7 @@ define void @test_vfmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; SKX-NEXT: vfmsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [11:0.50]
; SKX-NEXT: vfmsub231pd {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfmsubpd_256:
@ -1784,7 +1784,7 @@ define void @test_vfmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; GENERIC-NEXT: vfmsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [10:0.50]
; GENERIC-NEXT: vfmsub231ps {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfmsubps_256:
@ -1810,7 +1810,7 @@ define void @test_vfmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; HASWELL-NEXT: vfmsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [12:0.50]
; HASWELL-NEXT: vfmsub231ps {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfmsubps_256:
@ -1823,7 +1823,7 @@ define void @test_vfmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; BROADWELL-NEXT: vfmsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [11:0.50]
; BROADWELL-NEXT: vfmsub231ps {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfmsubps_256:
@ -1836,7 +1836,7 @@ define void @test_vfmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; SKYLAKE-NEXT: vfmsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [11:0.50]
; SKYLAKE-NEXT: vfmsub231ps {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfmsubps_256:
@ -1861,7 +1861,7 @@ define void @test_vfmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; SKX-NEXT: vfmsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) - mem sched: [11:0.50]
; SKX-NEXT: vfmsub231ps {{.*#+}} ymm0 = (ymm1 * mem) - ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfmsubps_256:
@ -2195,7 +2195,7 @@ define void @test_vfnmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; GENERIC-NEXT: vfnmadd213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [10:0.50]
; GENERIC-NEXT: vfnmadd231pd {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfnmaddpd_256:
@ -2221,7 +2221,7 @@ define void @test_vfnmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; HASWELL-NEXT: vfnmadd213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [12:0.50]
; HASWELL-NEXT: vfnmadd231pd {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfnmaddpd_256:
@ -2234,7 +2234,7 @@ define void @test_vfnmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; BROADWELL-NEXT: vfnmadd213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [11:0.50]
; BROADWELL-NEXT: vfnmadd231pd {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfnmaddpd_256:
@ -2247,7 +2247,7 @@ define void @test_vfnmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; SKYLAKE-NEXT: vfnmadd213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [11:0.50]
; SKYLAKE-NEXT: vfnmadd231pd {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfnmaddpd_256:
@ -2272,7 +2272,7 @@ define void @test_vfnmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; SKX-NEXT: vfnmadd213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [11:0.50]
; SKX-NEXT: vfnmadd231pd {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfnmaddpd_256:
@ -2402,7 +2402,7 @@ define void @test_vfnmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; GENERIC-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [10:0.50]
; GENERIC-NEXT: vfnmadd231ps {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfnmaddps_256:
@ -2428,7 +2428,7 @@ define void @test_vfnmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; HASWELL-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [12:0.50]
; HASWELL-NEXT: vfnmadd231ps {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfnmaddps_256:
@ -2441,7 +2441,7 @@ define void @test_vfnmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; BROADWELL-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [11:0.50]
; BROADWELL-NEXT: vfnmadd231ps {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfnmaddps_256:
@ -2454,7 +2454,7 @@ define void @test_vfnmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; SKYLAKE-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [11:0.50]
; SKYLAKE-NEXT: vfnmadd231ps {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfnmaddps_256:
@ -2479,7 +2479,7 @@ define void @test_vfnmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; SKX-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + mem sched: [11:0.50]
; SKX-NEXT: vfnmadd231ps {{.*#+}} ymm0 = -(ymm1 * mem) + ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfnmaddps_256:
@ -2813,7 +2813,7 @@ define void @test_vfnmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; GENERIC-NEXT: vfnmsub213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [10:0.50]
; GENERIC-NEXT: vfnmsub231pd {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfnmsubpd_256:
@ -2839,7 +2839,7 @@ define void @test_vfnmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; HASWELL-NEXT: vfnmsub213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [12:0.50]
; HASWELL-NEXT: vfnmsub231pd {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfnmsubpd_256:
@ -2852,7 +2852,7 @@ define void @test_vfnmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; BROADWELL-NEXT: vfnmsub213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [11:0.50]
; BROADWELL-NEXT: vfnmsub231pd {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfnmsubpd_256:
@ -2865,7 +2865,7 @@ define void @test_vfnmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; SKYLAKE-NEXT: vfnmsub213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [11:0.50]
; SKYLAKE-NEXT: vfnmsub231pd {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfnmsubpd_256:
@ -2890,7 +2890,7 @@ define void @test_vfnmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; SKX-NEXT: vfnmsub213pd {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [11:0.50]
; SKX-NEXT: vfnmsub231pd {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfnmsubpd_256:
@ -3020,7 +3020,7 @@ define void @test_vfnmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; GENERIC-NEXT: vfnmsub213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [10:0.50]
; GENERIC-NEXT: vfnmsub231ps {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER2-LABEL: test_vfnmsubps_256:
@ -3046,7 +3046,7 @@ define void @test_vfnmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; HASWELL-NEXT: vfnmsub213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [12:0.50]
; HASWELL-NEXT: vfnmsub231ps {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [12:0.50]
; HASWELL-NEXT: #NO_APP
; HASWELL-NEXT: vzeroupper # sched: [4:1.00]
; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
; HASWELL-NEXT: retq # sched: [7:1.00]
;
; BROADWELL-LABEL: test_vfnmsubps_256:
@ -3059,7 +3059,7 @@ define void @test_vfnmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; BROADWELL-NEXT: vfnmsub213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [11:0.50]
; BROADWELL-NEXT: vfnmsub231ps {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [11:0.50]
; BROADWELL-NEXT: #NO_APP
; BROADWELL-NEXT: vzeroupper # sched: [4:1.00]
; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
; BROADWELL-NEXT: retq # sched: [7:1.00]
;
; SKYLAKE-LABEL: test_vfnmsubps_256:
@ -3072,7 +3072,7 @@ define void @test_vfnmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; SKYLAKE-NEXT: vfnmsub213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [11:0.50]
; SKYLAKE-NEXT: vfnmsub231ps {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [11:0.50]
; SKYLAKE-NEXT: #NO_APP
; SKYLAKE-NEXT: vzeroupper # sched: [4:1.00]
; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
; SKYLAKE-NEXT: retq # sched: [7:1.00]
;
; KNL-LABEL: test_vfnmsubps_256:
@ -3097,7 +3097,7 @@ define void @test_vfnmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; SKX-NEXT: vfnmsub213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) - mem sched: [11:0.50]
; SKX-NEXT: vfnmsub231ps {{.*#+}} ymm0 = -(ymm1 * mem) - ymm0 sched: [11:0.50]
; SKX-NEXT: #NO_APP
; SKX-NEXT: vzeroupper # sched: [4:1.00]
; SKX-NEXT: vzeroupper # sched: [0:0.67]
; SKX-NEXT: retq # sched: [7:1.00]
;
; ZNVER1-LABEL: test_vfnmsubps_256:

24
test/CodeGen/X86/fma4-schedule.ll
View File

@ -48,7 +48,7 @@ define void @test_vfmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; GENERIC-NEXT: vfmaddpd (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfmaddpd %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfmaddpd_256:
@ -113,7 +113,7 @@ define void @test_vfmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; GENERIC-NEXT: vfmaddps (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfmaddps %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfmaddps_256:
@ -244,7 +244,7 @@ define void @test_vfmaddsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; GENERIC-NEXT: vfmaddsubpd (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfmaddsubpd %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfmaddsubpd_256:
@ -309,7 +309,7 @@ define void @test_vfmaddsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; GENERIC-NEXT: vfmaddsubps (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfmaddsubps %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfmaddsubps_256:
@ -378,7 +378,7 @@ define void @test_vfmsubaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x doubl
; GENERIC-NEXT: vfmsubaddpd (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfmsubaddpd %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfmsubaddpd_256:
@ -443,7 +443,7 @@ define void @test_vfmsubaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float>
; GENERIC-NEXT: vfmsubaddps (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfmsubaddps %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfmsubaddps_256:
@ -512,7 +512,7 @@ define void @test_vfmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; GENERIC-NEXT: vfmsubpd (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfmsubpd %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfmsubpd_256:
@ -577,7 +577,7 @@ define void @test_vfmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2
; GENERIC-NEXT: vfmsubps (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfmsubps %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfmsubps_256:
@ -708,7 +708,7 @@ define void @test_vfnmaddpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; GENERIC-NEXT: vfnmaddpd (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfnmaddpd %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfnmaddpd_256:
@ -773,7 +773,7 @@ define void @test_vfnmaddps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; GENERIC-NEXT: vfnmaddps (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfnmaddps %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfnmaddps_256:
@ -904,7 +904,7 @@ define void @test_vfnmsubpd_256(<4 x double> %a0, <4 x double> %a1, <4 x double>
; GENERIC-NEXT: vfnmsubpd (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfnmsubpd %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfnmsubpd_256:
@ -969,7 +969,7 @@ define void @test_vfnmsubps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a
; GENERIC-NEXT: vfnmsubps (%rdi), %ymm1, %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: vfnmsubps %ymm1, (%rdi), %ymm0, %ymm0 # sched: [10:0.50]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfnmsubps_256:

10
test/CodeGen/X86/xop-schedule.ll
View File

@ -14,7 +14,7 @@ define void @test_vfrczpd(<2 x double> %a0, <4 x double> %a1, <2 x double> *%a2,
; GENERIC-NEXT: vfrczpd (%rdi), %xmm0 # sched: [9:1.00]
; GENERIC-NEXT: vfrczpd (%rsi), %ymm1 # sched: [10:1.00]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfrczpd:
@ -62,7 +62,7 @@ define void @test_vfrczps(<4 x float> %a0, <4 x double> %a1, <4 x float> *%a2, <
; GENERIC-NEXT: vfrczps (%rdi), %xmm0 # sched: [9:1.00]
; GENERIC-NEXT: vfrczps (%rsi), %ymm1 # sched: [10:1.00]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vfrczps:
@ -221,7 +221,7 @@ define void @test_vpcmov_256(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> %a2, <4 x i
; GENERIC-NEXT: vpcmov (%rdi), %ymm1, %ymm0, %ymm0 # sched: [8:1.00]
; GENERIC-NEXT: vpcmov %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vpcmov_256:
@ -425,7 +425,7 @@ define void @test_vpermil2pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double
; GENERIC-NEXT: vpermil2pd $3, %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
; GENERIC-NEXT: vpermil2pd $3, (%rdi), %ymm1, %ymm0, %ymm0 # sched: [8:1.00]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vpermil2pd_256:
@ -509,7 +509,7 @@ define void @test_vpermil2ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %
; GENERIC-NEXT: vpermil2ps $3, %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
; GENERIC-NEXT: vpermil2ps $3, (%rdi), %ymm1, %ymm0, %ymm0 # sched: [8:1.00]
; GENERIC-NEXT: #NO_APP
; GENERIC-NEXT: vzeroupper # sched: [100:0.33]
; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
; GENERIC-NEXT: retq # sched: [1:1.00]
;
; BDVER12-LABEL: test_vpermil2ps_256:

6
test/tools/llvm-mca/X86/Broadwell/resources-avx1.s
View File

@ -1720,7 +1720,7 @@ vzeroupper
# CHECK-NEXT: 1 1 1.00 vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: 2 7 1.00 * vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: 20 8 1.00 * * U vzeroall
# CHECK-NEXT: 4 4 1.00 * * U vzeroupper
# CHECK-NEXT: 4 0 1.00 * * U vzeroupper
# CHECK: Resources:
# CHECK-NEXT: [0] - BWDivider
@ -1736,7 +1736,7 @@ vzeroupper
# CHECK: Resource pressure per iteration:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9]
# CHECK-NEXT: - 257.00 216.33 236.33 176.17 176.17 38.00 427.33 3.00 12.67
# CHECK-NEXT: - 257.00 215.25 235.25 176.17 176.17 38.00 426.25 2.25 12.67
# CHECK: Resource pressure by instruction:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] Instructions:
@ -2430,4 +2430,4 @@ vzeroupper
# CHECK-NEXT: - - - - - - - 1.00 - - vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: - - - - - - - 1.00 1.00 - vzeroall
# CHECK-NEXT: - - 1.08 1.08 - - - 1.08 0.75 - vzeroupper
# CHECK-NEXT: - - - - - - - - - - vzeroupper

6
test/tools/llvm-mca/X86/Generic/resources-avx1.s
View File

@ -1720,7 +1720,7 @@ vzeroupper
# CHECK-NEXT: 1 1 1.00 vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: 2 8 1.00 * vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: 20 9 2.00 * * U vzeroall
# CHECK-NEXT: 1 100 0.33 * * U vzeroupper
# CHECK-NEXT: 4 1 1.00 * * U vzeroupper
# CHECK: Resources:
# CHECK-NEXT: [0] - SBDivider
@ -1734,7 +1734,7 @@ vzeroupper
# CHECK: Resource pressure per iteration:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6.0] [6.1]
# CHECK-NEXT: - 572.00 246.83 317.33 39.00 365.83 179.50 179.50
# CHECK-NEXT: - 572.00 246.50 317.00 39.00 365.50 179.50 179.50
# CHECK: Resource pressure by instruction:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6.0] [6.1] Instructions:
@ -2428,4 +2428,4 @@ vzeroupper
# CHECK-NEXT: - - - - - 1.00 - - vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: - - - - - 2.00 - - vzeroall
# CHECK-NEXT: - - 0.33 0.33 - 0.33 - - vzeroupper
# CHECK-NEXT: - - - - - - - - vzeroupper

6
test/tools/llvm-mca/X86/Haswell/resources-avx1.s
View File

@ -1720,7 +1720,7 @@ vzeroupper
# CHECK-NEXT: 1 1 1.00 vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: 2 8 1.00 * vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: 20 8 1.00 * * U vzeroall
# CHECK-NEXT: 4 4 1.00 * * U vzeroupper
# CHECK-NEXT: 4 0 1.00 * * U vzeroupper
# CHECK: Resources:
# CHECK-NEXT: [0] - HWDivider
@ -1736,7 +1736,7 @@ vzeroupper
# CHECK: Resource pressure per iteration:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9]
# CHECK-NEXT: - 336.00 215.67 237.67 176.17 176.17 38.00 430.67 3.00 12.67
# CHECK-NEXT: - 336.00 214.58 236.58 176.17 176.17 38.00 429.58 2.25 12.67
# CHECK: Resource pressure by instruction:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] Instructions:
@ -2430,4 +2430,4 @@ vzeroupper
# CHECK-NEXT: - - - - - - - 1.00 - - vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: - - - - 0.50 0.50 - 1.00 - - vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: - - - - - - - 1.00 1.00 - vzeroall
# CHECK-NEXT: - - 1.08 1.08 - - - 1.08 0.75 - vzeroupper
# CHECK-NEXT: - - - - - - - - - - vzeroupper

6
test/tools/llvm-mca/X86/SandyBridge/resources-avx1.s
View File

@ -1720,7 +1720,7 @@ vzeroupper
# CHECK-NEXT: 1 1 1.00 vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: 2 8 1.00 * vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: 20 9 2.00 * * U vzeroall
# CHECK-NEXT: 1 100 0.33 * * U vzeroupper
# CHECK-NEXT: 4 1 1.00 * * U vzeroupper
# CHECK: Resources:
# CHECK-NEXT: [0] - SBDivider
@ -1734,7 +1734,7 @@ vzeroupper
# CHECK: Resource pressure per iteration:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6.0] [6.1]
# CHECK-NEXT: - 572.00 246.83 317.33 39.00 365.83 179.50 179.50
# CHECK-NEXT: - 572.00 246.50 317.00 39.00 365.50 179.50 179.50
# CHECK: Resource pressure by instruction:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6.0] [6.1] Instructions:
@ -2428,4 +2428,4 @@ vzeroupper
# CHECK-NEXT: - - - - - 1.00 - - vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: - - - - - 1.00 0.50 0.50 vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: - - - - - 2.00 - - vzeroall
# CHECK-NEXT: - - 0.33 0.33 - 0.33 - - vzeroupper
# CHECK-NEXT: - - - - - - - - vzeroupper

6
test/tools/llvm-mca/X86/SkylakeClient/resources-avx1.s
View File

@ -1720,7 +1720,7 @@ vzeroupper
# CHECK-NEXT: 1 1 0.33 vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: 2 8 0.50 * vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: 16 16 4.00 * * U vzeroall
# CHECK-NEXT: 4 4 1.00 * * U vzeroupper
# CHECK-NEXT: 4 0 0.67 * * U vzeroupper
# CHECK: Resources:
# CHECK-NEXT: [0] - SKLDivider
@ -1736,7 +1736,7 @@ vzeroupper
# CHECK: Resource pressure per iteration:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9]
# CHECK-NEXT: - 123.00 333.17 204.17 173.17 173.17 34.00 324.67 6.00 12.67
# CHECK-NEXT: - 123.00 332.08 203.08 173.17 173.17 34.00 323.58 5.25 12.67
# CHECK: Resource pressure by instruction:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] Instructions:
@ -2430,4 +2430,4 @@ vzeroupper
# CHECK-NEXT: - - 0.33 0.33 - - - 0.33 - - vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: - - 0.33 0.33 0.50 0.50 - 0.33 - - vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: - - 4.00 4.00 - - - 4.00 4.00 - vzeroall
# CHECK-NEXT: - - 1.08 1.08 - - - 1.08 0.75 - vzeroupper
# CHECK-NEXT: - - - - - - - - - - vzeroupper

6
test/tools/llvm-mca/X86/SkylakeServer/resources-avx1.s
View File

@ -1720,7 +1720,7 @@ vzeroupper
# CHECK-NEXT: 1 1 0.33 vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: 2 8 0.50 * vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: 34 12 5.00 * * U vzeroall
# CHECK-NEXT: 4 4 1.00 * * U vzeroupper
# CHECK-NEXT: 4 0 0.67 * * U vzeroupper
# CHECK: Resources:
# CHECK-NEXT: [0] - SKXDivider
@ -1736,7 +1736,7 @@ vzeroupper
# CHECK: Resource pressure per iteration:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9]
# CHECK-NEXT: - 123.00 318.50 197.50 173.17 173.17 34.00 339.00 7.00 12.67
# CHECK-NEXT: - 123.00 317.42 196.42 173.17 173.17 34.00 337.92 6.25 12.67
# CHECK: Resource pressure by instruction:
# CHECK-NEXT: [0] [1] [2] [3] [4] [5] [6] [7] [8] [9] Instructions:
@ -2430,4 +2430,4 @@ vzeroupper
# CHECK-NEXT: - - 0.33 0.33 - - - 0.33 - - vxorps %ymm0, %ymm1, %ymm2
# CHECK-NEXT: - - 0.33 0.33 0.50 0.50 - 0.33 - - vxorps (%rax), %ymm1, %ymm2
# CHECK-NEXT: - - 2.00 1.00 - - - 2.00 5.00 - vzeroall
# CHECK-NEXT: - - 1.08 1.08 - - - 1.08 0.75 - vzeroupper
# CHECK-NEXT: - - - - - - - - - - vzeroupper