RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2026-07-19 23:23:38 -04:00
Code Issues Packages Projects Releases 10 Wiki Activity

[X86] Automatically generate stack folding tests. NFC

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@369876 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Amaury Sechet
2019-08-25 20:48:14 +00:00
parent ae7ce17922
commit 2e6b938e4e
22 changed files with 19916 additions and 3750 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/CodeGen/X86/stack-folding-3dnow.ll
+218 -48
View File
@@ -1,8 +1,16 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+3dnow | FileCheck %s
define x86_mmx @stack_fold_pavgusb(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pavgusb
;CHECK: pavgusb {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pavgusb:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pavgusb {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pavgusb(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -10,8 +18,15 @@ define x86_mmx @stack_fold_pavgusb(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pavgusb(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pf2id(x86_mmx %a) {
;CHECK-LABEL: stack_fold_pf2id
;CHECK: pf2id {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pf2id:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pf2id {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm1},~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pf2id(x86_mmx %a) nounwind readnone
ret x86_mmx %2
@@ -19,8 +34,15 @@ define x86_mmx @stack_fold_pf2id(x86_mmx %a) {
declare x86_mmx @llvm.x86.3dnow.pf2id(x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pf2iw(x86_mmx %a) {
;CHECK-LABEL: stack_fold_pf2iw
;CHECK: pf2iw {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pf2iw:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pf2iw {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm1},~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnowa.pf2iw(x86_mmx %a) nounwind readnone
ret x86_mmx %2
@@ -28,8 +50,15 @@ define x86_mmx @stack_fold_pf2iw(x86_mmx %a) {
declare x86_mmx @llvm.x86.3dnowa.pf2iw(x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfacc(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfacc
;CHECK: pfacc {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfacc:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfacc {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfacc(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -37,8 +66,15 @@ define x86_mmx @stack_fold_pfacc(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfacc(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfadd(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfadd
;CHECK: pfadd {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfadd:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfadd {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfadd(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -46,8 +82,15 @@ define x86_mmx @stack_fold_pfadd(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfadd(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfcmpeq(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfcmpeq
;CHECK: pfcmpeq {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfcmpeq:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfcmpeq {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfcmpeq(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -55,8 +98,15 @@ define x86_mmx @stack_fold_pfcmpeq(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfcmpeq(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfcmpge(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfcmpge
;CHECK: pfcmpge {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfcmpge:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfcmpge {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfcmpge(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -64,8 +114,15 @@ define x86_mmx @stack_fold_pfcmpge(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfcmpge(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfcmpgt(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfcmpgt
;CHECK: pfcmpgt {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfcmpgt:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfcmpgt {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfcmpgt(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -73,8 +130,15 @@ define x86_mmx @stack_fold_pfcmpgt(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfcmpgt(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfmax(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfmax
;CHECK: pfmax {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfmax:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfmax {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfmax(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -82,8 +146,15 @@ define x86_mmx @stack_fold_pfmax(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfmax(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfmin(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfmin
;CHECK: pfmin {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfmin:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfmin {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfmin(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -91,8 +162,15 @@ define x86_mmx @stack_fold_pfmin(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfmin(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfmul(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfmul
;CHECK: pfmul {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfmul:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfmul {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfmul(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -100,8 +178,15 @@ define x86_mmx @stack_fold_pfmul(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfmul(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfnacc(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfnacc
;CHECK: pfnacc {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfnacc:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfnacc {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnowa.pfnacc(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -109,8 +194,15 @@ define x86_mmx @stack_fold_pfnacc(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnowa.pfnacc(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfpnacc(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfpnacc
;CHECK: pfpnacc {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfpnacc:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfpnacc {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnowa.pfpnacc(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -118,8 +210,15 @@ define x86_mmx @stack_fold_pfpnacc(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnowa.pfpnacc(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfrcp(x86_mmx %a) {
;CHECK-LABEL: stack_fold_pfrcp
;CHECK: pfrcp {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfrcp:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfrcp {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm1},~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfrcp(x86_mmx %a) nounwind readnone
ret x86_mmx %2
@@ -127,8 +226,15 @@ define x86_mmx @stack_fold_pfrcp(x86_mmx %a) {
declare x86_mmx @llvm.x86.3dnow.pfrcp(x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfrcpit1(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfrcpit1
;CHECK: pfrcpit1 {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfrcpit1:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfrcpit1 {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfrcpit1(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -136,8 +242,15 @@ define x86_mmx @stack_fold_pfrcpit1(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfrcpit1(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfrcpit2(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfrcpit2
;CHECK: pfrcpit2 {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfrcpit2:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfrcpit2 {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfrcpit2(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -145,8 +258,15 @@ define x86_mmx @stack_fold_pfrcpit2(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfrcpit2(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfrsqit1(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfrsqit1
;CHECK: pfrsqit1 {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfrsqit1:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfrsqit1 {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfrsqit1(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -154,8 +274,15 @@ define x86_mmx @stack_fold_pfrsqit1(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfrsqit1(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfrsqrt(x86_mmx %a) {
;CHECK-LABEL: stack_fold_pfrsqrt
;CHECK: pfrsqrt {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfrsqrt:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfrsqrt {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm1},~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfrsqrt(x86_mmx %a) nounwind readnone
ret x86_mmx %2
@@ -163,8 +290,15 @@ define x86_mmx @stack_fold_pfrsqrt(x86_mmx %a) {
declare x86_mmx @llvm.x86.3dnow.pfrsqrt(x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfsub(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfsub
;CHECK: pfsub {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfsub:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfsub {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfsub(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -172,8 +306,15 @@ define x86_mmx @stack_fold_pfsub(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfsub(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pfsubr(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pfsubr
;CHECK: pfsubr {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pfsubr:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pfsubr {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pfsubr(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -181,8 +322,15 @@ define x86_mmx @stack_fold_pfsubr(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pfsubr(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pi2fd(x86_mmx %a) {
;CHECK-LABEL: stack_fold_pi2fd
;CHECK: pi2fd {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pi2fd:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pi2fd {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm1},~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pi2fd(x86_mmx %a) nounwind readnone
ret x86_mmx %2
@@ -190,8 +338,15 @@ define x86_mmx @stack_fold_pi2fd(x86_mmx %a) {
declare x86_mmx @llvm.x86.3dnow.pi2fd(x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pi2fw(x86_mmx %a) {
;CHECK-LABEL: stack_fold_pi2fw
;CHECK: pi2fw {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pi2fw:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pi2fw {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm1},~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnowa.pi2fw(x86_mmx %a) nounwind readnone
ret x86_mmx %2
@@ -199,8 +354,15 @@ define x86_mmx @stack_fold_pi2fw(x86_mmx %a) {
declare x86_mmx @llvm.x86.3dnowa.pi2fw(x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pmulhrw(x86_mmx %a, x86_mmx %b) {
;CHECK-LABEL: stack_fold_pmulhrw
;CHECK: pmulhrw {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pmulhrw:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm1, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pmulhrw {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnow.pmulhrw(x86_mmx %a, x86_mmx %b) nounwind readnone
ret x86_mmx %2
@@ -208,8 +370,16 @@ define x86_mmx @stack_fold_pmulhrw(x86_mmx %a, x86_mmx %b) {
declare x86_mmx @llvm.x86.3dnow.pmulhrw(x86_mmx, x86_mmx) nounwind readnone
define x86_mmx @stack_fold_pswapd(x86_mmx %a) {
;CHECK-LABEL: stack_fold_pswapd
;CHECK: pswapd {{-?[0-9]*}}(%rsp), {{%mm[0-7]}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pswapd:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %mm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: pswapd {{[-0-9]+}}(%r{{[sb]}}p), %mm0 # 8-byte Folded Reload
; CHECK-NEXT: # mm0 = mem[1,0]
; CHECK-NEXT: movq2dq %mm0, %xmm0
; CHECK-NEXT: retq
%1 = tail call x86_mmx asm sideeffect "nop", "=y,~{mm1},~{mm2},~{mm3},~{mm4},~{mm5},~{mm6},~{mm7}"()
%2 = call x86_mmx @llvm.x86.3dnowa.pswapd(x86_mmx %a) nounwind readnone
ret x86_mmx %2
test/CodeGen/X86/stack-folding-adx-x86_64.ll
+283 -12
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+adx < %s | FileCheck %s --check-prefix=CHECK --check-prefix=ADX
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=-adx < %s | FileCheck %s --check-prefix=CHECK --check-prefix=NOADX
@@ -10,8 +11,53 @@ target triple = "x86_64-unknown-unknown"
; relevant registers and check that the reload is correctly folded into the instruction.
define i8 @stack_fold_addcarry_u32(i8 %a0, i32 %a1, i32 %a2, i8* %a3) {
;CHECK-LABEL: stack_fold_addcarry_u32
;CHECK: adcl {{-?[0-9]*}}(%rsp), %{{.*}} {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_addcarry_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movl %edx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: addb $-1, %al
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %edx # 4-byte Reload
; CHECK-NEXT: adcl {{[-0-9]+}}(%r{{[sb]}}p), %edx # 4-byte Folded Reload
; CHECK-NEXT: setb %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; CHECK-NEXT: movl %edx, (%rcx)
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call { i8, i32 } @llvm.x86.addcarry.32(i8 %a0, i32 %a1, i32 %a2)
%3 = extractvalue { i8, i32 } %2, 1
@@ -22,8 +68,53 @@ define i8 @stack_fold_addcarry_u32(i8 %a0, i32 %a1, i32 %a2, i8* %a3) {
}
define i8 @stack_fold_addcarry_u64(i8 %a0, i64 %a1, i64 %a2, i8* %a3) {
;CHECK-LABEL: stack_fold_addcarry_u64
;CHECK: adcq {{-?[0-9]*}}(%rsp), %{{.*}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_addcarry_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: addb $-1, %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; CHECK-NEXT: adcq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Folded Reload
; CHECK-NEXT: setb %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; CHECK-NEXT: movq %rdx, (%rcx)
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call { i8, i64 } @llvm.x86.addcarry.64(i8 %a0, i64 %a1, i64 %a2)
%3 = extractvalue { i8, i64 } %2, 1
@@ -34,8 +125,53 @@ define i8 @stack_fold_addcarry_u64(i8 %a0, i64 %a1, i64 %a2, i8* %a3) {
}
define i8 @stack_fold_addcarryx_u32(i8 %a0, i32 %a1, i32 %a2, i8* %a3) {
;CHECK-LABEL: stack_fold_addcarryx_u32
;CHECK: adcl {{-?[0-9]*}}(%rsp), %{{.*}} {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_addcarryx_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movl %edx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: addb $-1, %al
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %edx # 4-byte Reload
; CHECK-NEXT: adcl {{[-0-9]+}}(%r{{[sb]}}p), %edx # 4-byte Folded Reload
; CHECK-NEXT: setb %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; CHECK-NEXT: movl %edx, (%rcx)
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call { i8, i32 } @llvm.x86.addcarry.32(i8 %a0, i32 %a1, i32 %a2)
%3 = extractvalue { i8, i32 } %2, 1
@@ -46,8 +182,53 @@ define i8 @stack_fold_addcarryx_u32(i8 %a0, i32 %a1, i32 %a2, i8* %a3) {
}
define i8 @stack_fold_addcarryx_u64(i8 %a0, i64 %a1, i64 %a2, i8* %a3) {
;CHECK-LABEL: stack_fold_addcarryx_u64
;CHECK: adcq {{-?[0-9]*}}(%rsp), %{{.*}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_addcarryx_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: addb $-1, %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; CHECK-NEXT: adcq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Folded Reload
; CHECK-NEXT: setb %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; CHECK-NEXT: movq %rdx, (%rcx)
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call { i8, i64 } @llvm.x86.addcarry.64(i8 %a0, i64 %a1, i64 %a2)
%3 = extractvalue { i8, i64 } %2, 1
@@ -58,8 +239,53 @@ define i8 @stack_fold_addcarryx_u64(i8 %a0, i64 %a1, i64 %a2, i8* %a3) {
}
define i8 @stack_fold_subborrow_u32(i8 %a0, i32 %a1, i32 %a2, i8* %a3) {
;CHECK-LABEL: stack_fold_subborrow_u32
;CHECK: sbbl {{-?[0-9]*}}(%rsp), %{{.*}} {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_subborrow_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movl %edx, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: addb $-1, %al
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %edx # 4-byte Reload
; CHECK-NEXT: sbbl {{[-0-9]+}}(%r{{[sb]}}p), %edx # 4-byte Folded Reload
; CHECK-NEXT: setb %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; CHECK-NEXT: movl %edx, (%rcx)
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call { i8, i32 } @llvm.x86.subborrow.32(i8 %a0, i32 %a1, i32 %a2)
%3 = extractvalue { i8, i32 } %2, 1
@@ -70,8 +296,53 @@ define i8 @stack_fold_subborrow_u32(i8 %a0, i32 %a1, i32 %a2, i8* %a3) {
}
define i8 @stack_fold_subborrow_u64(i8 %a0, i64 %a1, i64 %a2, i8* %a3) {
;CHECK-LABEL: stack_fold_subborrow_u64
;CHECK: sbbq {{-?[0-9]*}}(%rsp), %{{.*}} {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_subborrow_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: addb $-1, %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; CHECK-NEXT: sbbq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Folded Reload
; CHECK-NEXT: setb %al
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
; CHECK-NEXT: movq %rdx, (%rcx)
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call { i8, i64 } @llvm.x86.subborrow.64(i8 %a0, i64 %a1, i64 %a2)
%3 = extractvalue { i8, i64 } %2, 1
test/CodeGen/X86/stack-folding-avx512bf16.ll
+234 -30
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512bf16,+avx512vl < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -9,8 +10,14 @@ target triple = "x86_64-unknown-unknown"
; relevant registers and check that the reload is correctly folded into the instruction.
define <32 x i16> @stack_fold_cvtne2ps2bf16(<16 x float> %a0, <16 x float> %a1) {
;CHECK-LABEL: stack_fold_cvtne2ps2bf16:
;CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_cvtne2ps2bf16:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <32 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.512(<16 x float> %a0, <16 x float> %a1)
ret <32 x i16> %2
@@ -19,7 +26,16 @@ declare <32 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.512(<16 x float>, <16 x fl
define <32 x i16> @stack_fold_cvtne2ps2bf16_mask(<16 x float> %a0, <16 x float> %a1, <32 x i16>* %passthru, i32 %U) {
; CHECK-LABEL: stack_fold_cvtne2ps2bf16_mask:
; CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vmovdqa64 (%rdi), %zmm2
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
; CHECK-NEXT: vmovdqa64 %zmm2, %zmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <32 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.512(<16 x float> %a0, <16 x float> %a1)
%3 = bitcast i32 %U to <32 x i1>
@@ -31,7 +47,14 @@ define <32 x i16> @stack_fold_cvtne2ps2bf16_mask(<16 x float> %a0, <16 x float>
define <32 x i16> @stack_fold_cvtne2ps2bf16_maskz(<16 x float> %a0, <16 x float> %a1, i32 %U) {
; CHECK-LABEL: stack_fold_cvtne2ps2bf16_maskz:
; CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %edi, %k1
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <32 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.512(<16 x float> %a0, <16 x float> %a1)
%3 = bitcast i32 %U to <32 x i1>
@@ -41,7 +64,13 @@ define <32 x i16> @stack_fold_cvtne2ps2bf16_maskz(<16 x float> %a0, <16 x float>
define <16 x i16> @stack_fold_cvtneps2bf16(<16 x float> %a0) {
; CHECK-LABEL: stack_fold_cvtneps2bf16:
; CHECK: vcvtneps2bf16 {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vcvtneps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <16 x i16> @llvm.x86.avx512bf16.cvtneps2bf16.512(<16 x float> %a0)
ret <16 x i16> %2
@@ -50,7 +79,16 @@ declare <16 x i16> @llvm.x86.avx512bf16.cvtneps2bf16.512(<16 x float>)
define <16 x i16> @stack_fold_cvtneps2bf16_mask(<16 x float> %a0, <16 x i16>* %passthru, i16 %U) {
; CHECK-LABEL: stack_fold_cvtneps2bf16_mask:
; CHECK: vcvtneps2bf16 {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vmovaps (%rdi), %ymm1
; CHECK-NEXT: vcvtneps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 {%k1} # 64-byte Folded Reload
; CHECK-NEXT: vmovaps %ymm1, %ymm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <16 x i16> @llvm.x86.avx512bf16.cvtneps2bf16.512(<16 x float> %a0)
%3 = bitcast i16 %U to <16 x i1>
@@ -62,7 +100,14 @@ define <16 x i16> @stack_fold_cvtneps2bf16_mask(<16 x float> %a0, <16 x i16>* %p
define <16 x i16> @stack_fold_cvtneps2bf16_maskz(<16 x float> %a0, i16 %U) {
; CHECK-LABEL: stack_fold_cvtneps2bf16_maskz:
; CHECK: vcvtneps2bf16 {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %edi, %k1
; CHECK-NEXT: vcvtneps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 {%k1} {z} # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <16 x i16> @llvm.x86.avx512bf16.cvtneps2bf16.512(<16 x float> %a0)
%3 = bitcast i16 %U to <16 x i1>
@@ -72,7 +117,13 @@ define <16 x i16> @stack_fold_cvtneps2bf16_maskz(<16 x float> %a0, i16 %U) {
define <16 x float> @stack_fold_vdpbf16ps(<16 x float> %a0, <16 x i32> %a1, <16 x i32> %a2) {
; CHECK-LABEL: stack_fold_vdpbf16ps:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <16 x float> @llvm.x86.avx512bf16.dpbf16ps.512(<16 x float> %a0, <16 x i32> %a1, <16 x i32> %a2)
ret <16 x float> %2
@@ -81,7 +132,16 @@ declare <16 x float> @llvm.x86.avx512bf16.dpbf16ps.512(<16 x float>, <16 x i32>,
define <16 x float> @stack_fold_vdpbf16ps_mask(<16 x float>* %a0, <16 x i32> %a1, <16 x i32> %a2, <16 x float>* %passthru, i16 %U) {
; CHECK-LABEL: stack_fold_vdpbf16ps_mask:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %zmm2
; CHECK-NEXT: kmovd %edx, %k1
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
; CHECK-NEXT: vmovaps %zmm2, %zmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
; load needed to keep the operation from being scheduled above the asm block
%2 = load <16 x float>, <16 x float>* %a0
@@ -93,7 +153,14 @@ define <16 x float> @stack_fold_vdpbf16ps_mask(<16 x float>* %a0, <16 x i32> %a1
define <16 x float> @stack_fold_vdpbf16ps_maskz(<16 x float> %a0, <16 x i32> %a1, <16 x i32> %a2, i16* %U) {
; CHECK-LABEL: stack_fold_vdpbf16ps_maskz:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovw (%rdi), %k1
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} {z} # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <16 x float> @llvm.x86.avx512bf16.dpbf16ps.512(<16 x float> %a0, <16 x i32> %a1, <16 x i32> %a2)
%3 = load i16, i16* %U
@@ -105,8 +172,14 @@ define <16 x float> @stack_fold_vdpbf16ps_maskz(<16 x float> %a0, <16 x i32> %a1
define <16 x i16> @stack_fold_cvtne2ps2bf16_ymm(<8 x float> %a0, <8 x float> %a1) {
;CHECK-LABEL: stack_fold_cvtne2ps2bf16_ymm:
;CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
; CHECK-LABEL: stack_fold_cvtne2ps2bf16_ymm:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.256(<8 x float> %a0, <8 x float> %a1)
ret <16 x i16> %2
@@ -115,7 +188,16 @@ declare <16 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.256(<8 x float>, <8 x floa
define <16 x i16> @stack_fold_cvtne2ps2bf16_mask_ymm(<8 x float> %a0, <8 x float> %a1, <16 x i16>* %passthru, i16 %U) {
; CHECK-LABEL: stack_fold_cvtne2ps2bf16_mask_ymm:
; CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vmovdqa (%rdi), %ymm2
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 {%k1} # 32-byte Folded Reload
; CHECK-NEXT: vmovdqa %ymm2, %ymm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.256(<8 x float> %a0, <8 x float> %a1)
%3 = bitcast i16 %U to <16 x i1>
@@ -127,7 +209,14 @@ define <16 x i16> @stack_fold_cvtne2ps2bf16_mask_ymm(<8 x float> %a0, <8 x float
define <16 x i16> @stack_fold_cvtne2ps2bf16_maskz_ymm(<8 x float> %a0, <8 x float> %a1, i16 %U) {
; CHECK-LABEL: stack_fold_cvtne2ps2bf16_maskz_ymm:
; CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %edi, %k1
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 {%k1} {z} # 32-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.256(<8 x float> %a0, <8 x float> %a1)
%3 = bitcast i16 %U to <16 x i1>
@@ -137,7 +226,14 @@ define <16 x i16> @stack_fold_cvtne2ps2bf16_maskz_ymm(<8 x float> %a0, <8 x floa
define <8 x i16> @stack_fold_cvtneps2bf16_ymm(<8 x float> %a0) {
; CHECK-LABEL: stack_fold_cvtneps2bf16_ymm:
; CHECK: vcvtneps2bf16y {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vcvtneps2bf16y {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 32-byte Folded Reload
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <8 x i16> @llvm.x86.avx512bf16.cvtneps2bf16.256(<8 x float> %a0)
ret <8 x i16> %2
@@ -146,7 +242,17 @@ declare <8 x i16> @llvm.x86.avx512bf16.cvtneps2bf16.256(<8 x float>)
define <8 x i16> @stack_fold_cvtneps2bf16_mask_ymm(<8 x float> %a0, <8 x i16>* %passthru, i8 %U) {
; CHECK-LABEL: stack_fold_cvtneps2bf16_mask_ymm:
; CHECK: vcvtneps2bf16y {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vmovaps (%rdi), %xmm1
; CHECK-NEXT: vcvtneps2bf16y {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 {%k1} # 32-byte Folded Reload
; CHECK-NEXT: vmovaps %xmm1, %xmm0
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <8 x i16> @llvm.x86.avx512bf16.cvtneps2bf16.256(<8 x float> %a0)
%3 = bitcast i8 %U to <8 x i1>
@@ -158,7 +264,15 @@ define <8 x i16> @stack_fold_cvtneps2bf16_mask_ymm(<8 x float> %a0, <8 x i16>* %
define <8 x i16> @stack_fold_cvtneps2bf16_maskz_ymm(<8 x float> %a0, i8 %U) {
; CHECK-LABEL: stack_fold_cvtneps2bf16_maskz_ymm:
; CHECK: vcvtneps2bf16y {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %edi, %k1
; CHECK-NEXT: vcvtneps2bf16y {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 {%k1} {z} # 32-byte Folded Reload
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <8 x i16> @llvm.x86.avx512bf16.cvtneps2bf16.256(<8 x float> %a0)
%3 = bitcast i8 %U to <8 x i1>
@@ -168,7 +282,13 @@ define <8 x i16> @stack_fold_cvtneps2bf16_maskz_ymm(<8 x float> %a0, i8 %U) {
define <8 x float> @stack_fold_vdpbf16ps_ymm(<8 x float> %a0, <8 x i32> %a1, <8 x i32> %a2) {
; CHECK-LABEL: stack_fold_vdpbf16ps_ymm:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <8 x float> @llvm.x86.avx512bf16.dpbf16ps.256(<8 x float> %a0, <8 x i32> %a1, <8 x i32> %a2)
ret <8 x float> %2
@@ -177,7 +297,16 @@ declare <8 x float> @llvm.x86.avx512bf16.dpbf16ps.256(<8 x float>, <8 x i32>, <8
define <8 x float> @stack_fold_vdpbf16ps_mask_ymm(<8 x float>* %a0, <8 x i32> %a1, <8 x i32> %a2, <8 x float>* %passthru, i8 %U) {
; CHECK-LABEL: stack_fold_vdpbf16ps_mask_ymm:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %ymm2
; CHECK-NEXT: kmovd %edx, %k1
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 {%k1} # 32-byte Folded Reload
; CHECK-NEXT: vmovaps %ymm2, %ymm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
; load needed to keep the operation from being scheduled above the asm block
%2 = load <8 x float>, <8 x float>* %a0
@@ -189,7 +318,15 @@ define <8 x float> @stack_fold_vdpbf16ps_mask_ymm(<8 x float>* %a0, <8 x i32> %a
define <8 x float> @stack_fold_vdpbf16ps_maskz_ymm(<8 x float> %a0, <8 x i32> %a1, <8 x i32> %a2, i8* %U) {
; CHECK-LABEL: stack_fold_vdpbf16ps_maskz_ymm:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%ymm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movzbl (%rdi), %eax
; CHECK-NEXT: kmovd %eax, %k1
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 {%k1} {z} # 32-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <8 x float> @llvm.x86.avx512bf16.dpbf16ps.256(<8 x float> %a0, <8 x i32> %a1, <8 x i32> %a2)
%3 = load i8, i8* %U
@@ -202,8 +339,14 @@ define <8 x float> @stack_fold_vdpbf16ps_maskz_ymm(<8 x float> %a0, <8 x i32> %a
define <8 x i16> @stack_fold_cvtne2ps2bf16_xmm(<4 x float> %a0, <4 x float> %a1) {
;CHECK-LABEL: stack_fold_cvtne2ps2bf16_xmm:
;CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK-LABEL: stack_fold_cvtne2ps2bf16_xmm:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <8 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.128(<4 x float> %a0, <4 x float> %a1)
ret <8 x i16> %2
@@ -212,7 +355,16 @@ declare <8 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.128(<4 x float>, <4 x float
define <8 x i16> @stack_fold_cvtne2ps2bf16_mask_xmm(<4 x float> %a0, <4 x float> %a1, <8 x i16>* %passthru, i8 %U) {
; CHECK-LABEL: stack_fold_cvtne2ps2bf16_mask_xmm:
; CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vmovdqa (%rdi), %xmm2
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 {%k1} # 16-byte Folded Reload
; CHECK-NEXT: vmovdqa %xmm2, %xmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <8 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.128(<4 x float> %a0, <4 x float> %a1)
%3 = bitcast i8 %U to <8 x i1>
@@ -224,7 +376,14 @@ define <8 x i16> @stack_fold_cvtne2ps2bf16_mask_xmm(<4 x float> %a0, <4 x float>
define <8 x i16> @stack_fold_cvtne2ps2bf16_maskz_xmm(<4 x float> %a0, <4 x float> %a1, i8 %U) {
; CHECK-LABEL: stack_fold_cvtne2ps2bf16_maskz_xmm:
; CHECK: vcvtne2ps2bf16 {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %edi, %k1
; CHECK-NEXT: vcvtne2ps2bf16 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 {%k1} {z} # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <8 x i16> @llvm.x86.avx512bf16.cvtne2ps2bf16.128(<4 x float> %a0, <4 x float> %a1)
%3 = bitcast i8 %U to <8 x i1>
@@ -234,7 +393,13 @@ define <8 x i16> @stack_fold_cvtne2ps2bf16_maskz_xmm(<4 x float> %a0, <4 x float
define <8 x i16> @stack_fold_cvtneps2bf16_xmm(<4 x float> %a0) {
; CHECK-LABEL: stack_fold_cvtneps2bf16_xmm:
; CHECK: vcvtneps2bf16x {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vcvtneps2bf16x {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <8 x i16> @llvm.x86.avx512bf16.mask.cvtneps2bf16.128(<4 x float> %a0, <8 x i16> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
ret <8 x i16> %2
@@ -243,7 +408,16 @@ declare <8 x i16> @llvm.x86.avx512bf16.mask.cvtneps2bf16.128(<4 x float>, <8 x i
define <8 x i16> @stack_fold_cvtneps2bf16_mask_xmm(<4 x float> %a0, <8 x i16>* %passthru, i8 %U) {
; CHECK-LABEL: stack_fold_cvtneps2bf16_mask_xmm:
; CHECK: vcvtneps2bf16x {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %xmm1
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vcvtneps2bf16x {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 {%k1} # 16-byte Folded Reload
; CHECK-NEXT: vmovaps %xmm1, %xmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <8 x i16>, <8 x i16>* %passthru
%3 = bitcast i8 %U to <8 x i1>
@@ -254,7 +428,14 @@ define <8 x i16> @stack_fold_cvtneps2bf16_mask_xmm(<4 x float> %a0, <8 x i16>* %
define <8 x i16> @stack_fold_cvtneps2bf16_maskz_xmm(<4 x float> %a0, i8 %U) {
; CHECK-LABEL: stack_fold_cvtneps2bf16_maskz_xmm:
; CHECK: vcvtneps2bf16x {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovd %edi, %k1
; CHECK-NEXT: vcvtneps2bf16x {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 {%k1} {z} # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = bitcast i8 %U to <8 x i1>
%3 = shufflevector <8 x i1> %2, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
@@ -264,7 +445,13 @@ define <8 x i16> @stack_fold_cvtneps2bf16_maskz_xmm(<4 x float> %a0, i8 %U) {
define <4 x float> @stack_fold_vdpbf16ps_xmm(<4 x float> %a0, <4 x i32> %a1, <4 x i32> %a2) {
; CHECK-LABEL: stack_fold_vdpbf16ps_xmm:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <4 x float> @llvm.x86.avx512bf16.dpbf16ps.128(<4 x float> %a0, <4 x i32> %a1, <4 x i32> %a2)
ret <4 x float> %2
@@ -273,7 +460,16 @@ declare <4 x float> @llvm.x86.avx512bf16.dpbf16ps.128(<4 x float>, <4 x i32>, <4
define <4 x float> @stack_fold_vdpbf16ps_mask_xmm(<4 x float>* %a0, <4 x i32> %a1, <4 x i32> %a2, <4 x float>* %passthru, i8 %U) {
; CHECK-LABEL: stack_fold_vdpbf16ps_mask_xmm:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %xmm2
; CHECK-NEXT: kmovd %edx, %k1
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 {%k1} # 16-byte Folded Reload
; CHECK-NEXT: vmovaps %xmm2, %xmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
; load needed to keep the operation from being scheduled above the asm block
%2 = load <4 x float>, <4 x float>* %a0
@@ -286,7 +482,15 @@ define <4 x float> @stack_fold_vdpbf16ps_mask_xmm(<4 x float>* %a0, <4 x i32> %a
define <4 x float> @stack_fold_vdpbf16ps_maskz_xmm(<4 x float> %a0, <4 x i32> %a1, <4 x i32> %a2, i8* %U) {
; CHECK-LABEL: stack_fold_vdpbf16ps_maskz_xmm:
; CHECK: vdpbf16ps {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%xmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movzbl (%rdi), %eax
; CHECK-NEXT: kmovd %eax, %k1
; CHECK-NEXT: vdpbf16ps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm0 {%k1} {z} # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = tail call <4 x float> @llvm.x86.avx512bf16.dpbf16ps.128(<4 x float> %a0, <4 x i32> %a1, <4 x i32> %a2)
%3 = load i8, i8* %U
test/CodeGen/X86/stack-folding-avx512vp2intersect.ll
+75 -6
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512vp2intersect,+avx512vl < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -5,7 +6,17 @@ target triple = "x86_64-unknown-unknown"
define void @stack_fold_vp2intersectd(<16 x i32>* %a, <16 x i32> %b, <16 x i1>* nocapture %m0, <16 x i1>* nocapture %m1) {
; CHECK-LABEL: stack_fold_vp2intersectd:
; CHECK: vp2intersectd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %zmm0
; CHECK-NEXT: vp2intersectd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k0 # 64-byte Folded Reload
; CHECK-NEXT: kmovw %k0, (%rsi)
; CHECK-NEXT: kmovw %k1, (%rdx)
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <16 x i32>, <16 x i32>* %a
%3 = tail call { <16 x i1>, <16 x i1> } @llvm.x86.avx512.vp2intersect.d.512(<16 x i32> %2, <16 x i32> %b)
@@ -19,7 +30,19 @@ declare { <16 x i1>, <16 x i1> } @llvm.x86.avx512.vp2intersect.d.512(<16 x i32>,
define void @stack_fold_vp2intersectq(<8 x i64>* %a, <8 x i64> %b, <8 x i1>* nocapture %m0, <8 x i1>* nocapture %m1) {
; CHECK-LABEL: stack_fold_vp2intersectq:
; CHECK: vp2intersectq {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 64-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %zmm0
; CHECK-NEXT: vp2intersectq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k0 # 64-byte Folded Reload
; CHECK-NEXT: kmovw %k1, %eax
; CHECK-NEXT: kmovw %k0, %ecx
; CHECK-NEXT: movb %cl, (%rsi)
; CHECK-NEXT: movb %al, (%rdx)
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <8 x i64>, <8 x i64>* %a
%3 = tail call { <8 x i1>, <8 x i1> } @llvm.x86.avx512.vp2intersect.q.512(<8 x i64> %2, <8 x i64> %b)
@@ -33,7 +56,19 @@ declare { <8 x i1>, <8 x i1> } @llvm.x86.avx512.vp2intersect.q.512(<8 x i64>, <8
define void @stack_fold_vp2intersectd_256(<8 x i32>* %a, <8 x i32> %b, <8 x i1>* nocapture %m0, <8 x i1>* nocapture %m1) {
; CHECK-LABEL: stack_fold_vp2intersectd_256:
; CHECK: vp2intersectd {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %ymm0
; CHECK-NEXT: vp2intersectd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %k0 # 32-byte Folded Reload
; CHECK-NEXT: kmovw %k1, %eax
; CHECK-NEXT: kmovw %k0, %ecx
; CHECK-NEXT: movb %cl, (%rsi)
; CHECK-NEXT: movb %al, (%rdx)
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <8 x i32>, <8 x i32>* %a
%3 = tail call { <8 x i1>, <8 x i1> } @llvm.x86.avx512.vp2intersect.d.256(<8 x i32> %2, <8 x i32> %b)
@@ -47,7 +82,19 @@ declare { <8 x i1>, <8 x i1> } @llvm.x86.avx512.vp2intersect.d.256(<8 x i32>, <8
define void @stack_fold_vp2intersectq_256(<4 x i64>* %a, <4 x i64> %b, <4 x i1>* nocapture %m0, <4 x i1>* nocapture %m1) {
; CHECK-LABEL: stack_fold_vp2intersectq_256:
; CHECK: vp2intersectq {{-?[0-9]*}}(%rsp), {{%ymm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 32-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %ymm0
; CHECK-NEXT: vp2intersectq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %k0 # 32-byte Folded Reload
; CHECK-NEXT: kmovw %k1, %eax
; CHECK-NEXT: kmovw %k0, %ecx
; CHECK-NEXT: movb %cl, (%rsi)
; CHECK-NEXT: movb %al, (%rdx)
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <4 x i64>, <4 x i64>* %a
%3 = tail call { <4 x i1>, <4 x i1> } @llvm.x86.avx512.vp2intersect.q.256(<4 x i64> %2, <4 x i64> %b)
@@ -61,7 +108,18 @@ declare { <4 x i1>, <4 x i1> } @llvm.x86.avx512.vp2intersect.q.256(<4 x i64>, <4
define void @stack_fold_vp2intersectd_128(<4 x i32>* %a, <4 x i32> %b, <4 x i1>* nocapture %m0, <4 x i1>* nocapture %m1) {
; CHECK-LABEL: stack_fold_vp2intersectd_128:
; CHECK: vp2intersectd {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %xmm0
; CHECK-NEXT: vp2intersectd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %k0 # 16-byte Folded Reload
; CHECK-NEXT: kmovw %k1, %eax
; CHECK-NEXT: kmovw %k0, %ecx
; CHECK-NEXT: movb %cl, (%rsi)
; CHECK-NEXT: movb %al, (%rdx)
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <4 x i32>, <4 x i32>* %a
%3 = tail call { <4 x i1>, <4 x i1> } @llvm.x86.avx512.vp2intersect.d.128(<4 x i32> %2, <4 x i32> %b)
@@ -75,7 +133,18 @@ declare { <4 x i1>, <4 x i1> } @llvm.x86.avx512.vp2intersect.d.128(<4 x i32>, <4
define void @stack_fold_vp2intersectq_128(<2 x i64>* %a, <2 x i64> %b, <2 x i1>* nocapture %m0, <2 x i1>* nocapture %m1) {
; CHECK-LABEL: stack_fold_vp2intersectq_128:
; CHECK: vp2intersectq {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}}, {{%k[0-7]}} {{.*#+}} 16-byte Folded Reload
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %xmm0
; CHECK-NEXT: vp2intersectq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %k0 # 16-byte Folded Reload
; CHECK-NEXT: kmovw %k1, %eax
; CHECK-NEXT: kmovw %k0, %ecx
; CHECK-NEXT: movb %cl, (%rsi)
; CHECK-NEXT: movb %al, (%rdx)
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <2 x i64>, <2 x i64>* %a
%3 = tail call { <2 x i1>, <2 x i1> } @llvm.x86.avx512.vp2intersect.q.128(<2 x i64> %2, <2 x i64> %b)
test/CodeGen/X86/stack-folding-bmi.ll
+465 -26
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+bmi < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -9,8 +10,46 @@ target triple = "x86_64-unknown-unknown"
; relevant registers and check that the reload is correctly folded into the instruction.
define i32 @stack_fold_andn_u32(i32 %a0, i32 %a1) {
;CHECK-LABEL: stack_fold_andn_u32
;CHECK: andnl {{-?[0-9]*}}(%rsp), %eax, %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_andn_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: andnl {{[-0-9]+}}(%r{{[sb]}}p), %eax, %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = xor i32 %a0, -1
%3 = and i32 %a1, %2
@@ -18,8 +57,46 @@ define i32 @stack_fold_andn_u32(i32 %a0, i32 %a1) {
}
define i64 @stack_fold_andn_u64(i64 %a0, i64 %a1) {
;CHECK-LABEL: stack_fold_andn_u64
;CHECK: andnq {{-?[0-9]*}}(%rsp), %rax, %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_andn_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: andnq {{[-0-9]+}}(%r{{[sb]}}p), %rax, %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = xor i64 %a0, -1
%3 = and i64 %a1, %2
@@ -27,9 +104,46 @@ define i64 @stack_fold_andn_u64(i64 %a0, i64 %a1) {
}
define i32 @stack_fold_bextr_u32(i32 %a0, i32 %a1) {
;CHECK-LABEL: stack_fold_bextr_u32
;CHECK: # %bb.0:
;CHECK: bextrl %eax, {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_bextr_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: bextrl %eax, {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.x86.bmi.bextr.32(i32 %a0, i32 %a1)
ret i32 %2
@@ -37,9 +151,46 @@ define i32 @stack_fold_bextr_u32(i32 %a0, i32 %a1) {
declare i32 @llvm.x86.bmi.bextr.32(i32, i32)
define i64 @stack_fold_bextr_u64(i64 %a0, i64 %a1) {
;CHECK-LABEL: stack_fold_bextr_u64
;CHECK: # %bb.0:
;CHECK: bextrq %rax, {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_bextr_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: bextrq %rax, {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.x86.bmi.bextr.64(i64 %a0, i64 %a1)
ret i64 %2
@@ -47,8 +198,44 @@ define i64 @stack_fold_bextr_u64(i64 %a0, i64 %a1) {
declare i64 @llvm.x86.bmi.bextr.64(i64, i64)
define i32 @stack_fold_blsi_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blsi_u32
;CHECK: blsil {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blsi_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsil {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i32 0, %a0
%3 = and i32 %2, %a0
@@ -56,8 +243,44 @@ define i32 @stack_fold_blsi_u32(i32 %a0) {
}
define i64 @stack_fold_blsi_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blsi_u64
;CHECK: blsiq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blsi_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsiq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i64 0, %a0
%3 = and i64 %2, %a0
@@ -65,8 +288,44 @@ define i64 @stack_fold_blsi_u64(i64 %a0) {
}
define i32 @stack_fold_blsmsk_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blsmsk_u32
;CHECK: blsmskl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blsmsk_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsmskl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i32 %a0, 1
%3 = xor i32 %2, %a0
@@ -74,8 +333,44 @@ define i32 @stack_fold_blsmsk_u32(i32 %a0) {
}
define i64 @stack_fold_blsmsk_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blsmsk_u64
;CHECK: blsmskq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blsmsk_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsmskq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i64 %a0, 1
%3 = xor i64 %2, %a0
@@ -83,8 +378,44 @@ define i64 @stack_fold_blsmsk_u64(i64 %a0) {
}
define i32 @stack_fold_blsr_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blsr_u32
;CHECK: blsrl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blsr_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsrl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i32 %a0, 1
%3 = and i32 %2, %a0
@@ -92,8 +423,44 @@ define i32 @stack_fold_blsr_u32(i32 %a0) {
}
define i64 @stack_fold_blsr_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blsr_u64
;CHECK: blsrq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blsr_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsrq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i64 %a0, 1
%3 = and i64 %2, %a0
@@ -103,8 +470,44 @@ define i64 @stack_fold_blsr_u64(i64 %a0) {
;TODO stack_fold_tzcnt_u16
define i32 @stack_fold_tzcnt_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_tzcnt_u32
;CHECK: tzcntl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_tzcnt_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: tzcntl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.cttz.i32(i32 %a0, i1 0)
ret i32 %2
@@ -112,8 +515,44 @@ define i32 @stack_fold_tzcnt_u32(i32 %a0) {
declare i32 @llvm.cttz.i32(i32, i1)
define i64 @stack_fold_tzcnt_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_tzcnt_u64
;CHECK: tzcntq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_tzcnt_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: tzcntq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.cttz.i64(i64 %a0, i1 0)
ret i64 %2
test/CodeGen/X86/stack-folding-bmi2.ll
+241 -12
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+bmi,+bmi2 < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -9,8 +10,46 @@ target triple = "x86_64-unknown-unknown"
; relevant registers and check that the reload is correctly folded into the instruction.
define i32 @stack_fold_bzhi_u32(i32 %a0, i32 %a1) {
;CHECK-LABEL: stack_fold_bzhi_u32
;CHECK: bzhil %eax, {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_bzhi_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: bzhil %eax, {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i32 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.x86.bmi.bzhi.32(i32 %a0, i32 %a1)
ret i32 %2
@@ -18,8 +57,46 @@ define i32 @stack_fold_bzhi_u32(i32 %a0, i32 %a1) {
declare i32 @llvm.x86.bmi.bzhi.32(i32, i32)
define i64 @stack_fold_bzhi_u64(i64 %a0, i64 %a1) {
;CHECK-LABEL: stack_fold_bzhi_u64
;CHECK: bzhiq %rax, {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_bzhi_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: bzhiq %rax, {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.x86.bmi.bzhi.64(i64 %a0, i64 %a1)
ret i64 %2
@@ -27,8 +104,46 @@ define i64 @stack_fold_bzhi_u64(i64 %a0, i64 %a1) {
declare i64 @llvm.x86.bmi.bzhi.64(i64, i64)
define i32 @stack_fold_pdep_u32(i32 %a0, i32 %a1) {
;CHECK-LABEL: stack_fold_pdep_u32
;CHECK: pdepl {{-?[0-9]*}}(%rsp), %eax, %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_pdep_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: pdepl {{[-0-9]+}}(%r{{[sb]}}p), %eax, %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i32 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.x86.bmi.pdep.32(i32 %a0, i32 %a1)
ret i32 %2
@@ -36,8 +151,46 @@ define i32 @stack_fold_pdep_u32(i32 %a0, i32 %a1) {
declare i32 @llvm.x86.bmi.pdep.32(i32, i32)
define i64 @stack_fold_pdep_u64(i64 %a0, i64 %a1) {
;CHECK-LABEL: stack_fold_pdep_u64
;CHECK: pdepq {{-?[0-9]*}}(%rsp), %rax, %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pdep_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: pdepq {{[-0-9]+}}(%r{{[sb]}}p), %rax, %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.x86.bmi.pdep.64(i64 %a0, i64 %a1)
ret i64 %2
@@ -45,8 +198,46 @@ define i64 @stack_fold_pdep_u64(i64 %a0, i64 %a1) {
declare i64 @llvm.x86.bmi.pdep.64(i64, i64)
define i32 @stack_fold_pext_u32(i32 %a0, i32 %a1) {
;CHECK-LABEL: stack_fold_pext_u32
;CHECK: pextl {{-?[0-9]*}}(%rsp), %eax, %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_pext_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: pextl {{[-0-9]+}}(%r{{[sb]}}p), %eax, %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i32 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.x86.bmi.pext.32(i32 %a0, i32 %a1)
ret i32 %2
@@ -54,8 +245,46 @@ define i32 @stack_fold_pext_u32(i32 %a0, i32 %a1) {
declare i32 @llvm.x86.bmi.pext.32(i32, i32)
define i64 @stack_fold_pext_u64(i64 %a0, i64 %a1) {
;CHECK-LABEL: stack_fold_pext_u64
;CHECK: pextq {{-?[0-9]*}}(%rsp), %rax, %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_pext_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: pextq {{[-0-9]+}}(%r{{[sb]}}p), %rax, %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.x86.bmi.pext.64(i64 %a0, i64 %a1)
ret i64 %2
test/CodeGen/X86/stack-folding-fp-avx1.ll
+2078 -420
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-fp-avx512.ll
+1140 -224
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-fp-avx512vl.ll
+748 -168
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-fp-sse42.ll
+1467 -276
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-int-avx1.ll
+1250 -234
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-int-avx2.ll
+1128 -268
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-int-avx512.ll
+4036 -806
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-int-avx512vl.ll
+2348 -534
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-int-avx512vnni.ll
+113 -24
View File
@@ -1,27 +1,49 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512dq,+avx512vbmi,+avx512cd,+avx512vpopcntdq,+avx512vnni < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-unknown"
define <16 x i32> @stack_fold_vpdpwssd(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2) {
;CHECK-LABEL: stack_fold_vpdpwssd:
;CHECK: vpdpwssd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssd:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vpdpwssd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i32> @llvm.x86.avx512.vpdpwssd.512(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2)
ret <16 x i32> %2
}
define <16 x i32> @stack_fold_vpdpwssd_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2) {
;CHECK-LABEL: stack_fold_vpdpwssd_commuted:
;CHECK: vpdpwssd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssd_commuted:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vpdpwssd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i32> @llvm.x86.avx512.vpdpwssd.512(<16 x i32> %a0, <16 x i32> %a2, <16 x i32> %a1)
ret <16 x i32> %2
}
define <16 x i32> @stack_fold_vpdpwssd_mask(<16 x i32>* %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
;CHECK-LABEL: stack_fold_vpdpwssd_mask:
;CHECK: vpdpwssd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssd_mask:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %zmm2
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vpdpwssd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
; CHECK-NEXT: vmovaps %zmm2, %zmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <16 x i32>, <16 x i32>* %a0
%3 = call <16 x i32> @llvm.x86.avx512.vpdpwssd.512(<16 x i32> %2, <16 x i32> %a1, <16 x i32> %a2)
@@ -31,8 +53,17 @@ define <16 x i32> @stack_fold_vpdpwssd_mask(<16 x i32>* %a0, <16 x i32> %a1, <16
}
define <16 x i32> @stack_fold_vpdpwssd_mask_commuted(<16 x i32>* %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
;CHECK-LABEL: stack_fold_vpdpwssd_mask_commuted:
;CHECK: vpdpwssd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssd_mask_commuted:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %zmm2
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vpdpwssd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
; CHECK-NEXT: vmovaps %zmm2, %zmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <16 x i32>, <16 x i32>* %a0
%3 = call <16 x i32> @llvm.x86.avx512.vpdpwssd.512(<16 x i32> %2, <16 x i32> %a2, <16 x i32> %a1)
@@ -42,8 +73,15 @@ define <16 x i32> @stack_fold_vpdpwssd_mask_commuted(<16 x i32>* %a0, <16 x i32>
}
define <16 x i32> @stack_fold_vpdpwssd_maskz(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16* %mask) {
;CHECK-LABEL: stack_fold_vpdpwssd_maskz:
;CHECK: vpdpwssd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssd_maskz:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovw (%rdi), %k1
; CHECK-NEXT: vpdpwssd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} {z} # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i32> @llvm.x86.avx512.vpdpwssd.512(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2)
%3 = load i16, i16* %mask
@@ -53,8 +91,15 @@ define <16 x i32> @stack_fold_vpdpwssd_maskz(<16 x i32> %a0, <16 x i32> %a1, <16
}
define <16 x i32> @stack_fold_vpdpwssd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16* %mask) {
;CHECK-LABEL: stack_fold_vpdpwssd_maskz_commuted:
;CHECK: vpdpwssd {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssd_maskz_commuted:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovw (%rdi), %k1
; CHECK-NEXT: vpdpwssd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} {z} # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i32> @llvm.x86.avx512.vpdpwssd.512(<16 x i32> %a0, <16 x i32> %a2, <16 x i32> %a1)
%3 = load i16, i16* %mask
@@ -64,24 +109,45 @@ define <16 x i32> @stack_fold_vpdpwssd_maskz_commuted(<16 x i32> %a0, <16 x i32>
}
define <16 x i32> @stack_fold_vpdpwssds(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2) {
;CHECK-LABEL: stack_fold_vpdpwssds:
;CHECK: vpdpwssds {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssds:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vpdpwssds {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i32> @llvm.x86.avx512.vpdpwssds.512(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2)
ret <16 x i32> %2
}
define <16 x i32> @stack_fold_vpdpwssds_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2) {
;CHECK-LABEL: stack_fold_vpdpwssds_commuted:
;CHECK: vpdpwssds {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssds_commuted:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vpdpwssds {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i32> @llvm.x86.avx512.vpdpwssds.512(<16 x i32> %a0, <16 x i32> %a2, <16 x i32> %a1)
ret <16 x i32> %2
}
define <16 x i32> @stack_fold_vpdpwssds_mask(<16 x i32>* %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
;CHECK-LABEL: stack_fold_vpdpwssds_mask:
;CHECK: vpdpwssds {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssds_mask:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %zmm2
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vpdpwssds {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
; CHECK-NEXT: vmovaps %zmm2, %zmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <16 x i32>, <16 x i32>* %a0
%3 = call <16 x i32> @llvm.x86.avx512.vpdpwssds.512(<16 x i32> %2, <16 x i32> %a1, <16 x i32> %a2)
@@ -91,8 +157,17 @@ define <16 x i32> @stack_fold_vpdpwssds_mask(<16 x i32>* %a0, <16 x i32> %a1, <1
}
define <16 x i32> @stack_fold_vpdpwssds_mask_commuted(<16 x i32>* %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
;CHECK-LABEL: stack_fold_vpdpwssds_mask_commuted:
;CHECK: vpdpwssds {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssds_mask_commuted:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmovaps (%rdi), %zmm2
; CHECK-NEXT: kmovd %esi, %k1
; CHECK-NEXT: vpdpwssds {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
; CHECK-NEXT: vmovaps %zmm2, %zmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = load <16 x i32>, <16 x i32>* %a0
%3 = call <16 x i32> @llvm.x86.avx512.vpdpwssds.512(<16 x i32> %2, <16 x i32> %a2, <16 x i32> %a1)
@@ -102,8 +177,15 @@ define <16 x i32> @stack_fold_vpdpwssds_mask_commuted(<16 x i32>* %a0, <16 x i32
}
define <16 x i32> @stack_fold_vpdpwssds_maskz(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16* %mask) {
;CHECK-LABEL: stack_fold_vpdpwssds_maskz:
;CHECK: vpdpwssds {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssds_maskz:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovw (%rdi), %k1
; CHECK-NEXT: vpdpwssds {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} {z} # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i32> @llvm.x86.avx512.vpdpwssds.512(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2)
%3 = load i16, i16* %mask
@@ -113,8 +195,15 @@ define <16 x i32> @stack_fold_vpdpwssds_maskz(<16 x i32> %a0, <16 x i32> %a1, <1
}
define <16 x i32> @stack_fold_vpdpwssds_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16* %mask) {
;CHECK-LABEL: stack_fold_vpdpwssds_maskz_commuted:
;CHECK: vpdpwssds {{-?[0-9]*}}(%rsp), {{%zmm[0-9][0-9]*}}, {{%zmm[0-9][0-9]*}} {{{%k[0-7]}}} {z} {{.*#+}} 64-byte Folded Reload
; CHECK-LABEL: stack_fold_vpdpwssds_maskz_commuted:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: kmovw (%rdi), %k1
; CHECK-NEXT: vpdpwssds {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} {z} # 64-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
%2 = call <16 x i32> @llvm.x86.avx512.vpdpwssds.512(<16 x i32> %a0, <16 x i32> %a2, <16 x i32> %a1)
%3 = load i16, i16* %mask
test/CodeGen/X86/stack-folding-int-sse42.ll
+1538 -268
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-lwp.ll
+163 -8
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+lwp < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -9,9 +10,48 @@ target triple = "x86_64-unknown-unknown"
; relevant registers and check that the reload is correctly folded into the instruction.
define i8 @stack_fold_lwpins_u32(i32 %a0, i32 %a1) {
; CHECK-LABEL: stack_fold_lwpins_u32
; CHECK-LABEL: stack_fold_lwpins_u32:
; CHECK: # %bb.0:
; CHECK: lwpins $2814, {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: lwpins $2814, {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: # imm = 0xAFE
; CHECK-NEXT: setb %al
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i8 @llvm.x86.lwpins32(i32 %a0, i32 %a1, i32 2814)
ret i8 %2
@@ -19,9 +59,48 @@ define i8 @stack_fold_lwpins_u32(i32 %a0, i32 %a1) {
declare i8 @llvm.x86.lwpins32(i32, i32, i32)
define i8 @stack_fold_lwpins_u64(i64 %a0, i32 %a1) {
; CHECK-LABEL: stack_fold_lwpins_u64
; CHECK-LABEL: stack_fold_lwpins_u64:
; CHECK: # %bb.0:
; CHECK: lwpins $2814, {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 4-byte Folded Reload
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: lwpins $2814, {{[-0-9]+}}(%r{{[sb]}}p), %rax # 4-byte Folded Reload
; CHECK-NEXT: # imm = 0xAFE
; CHECK-NEXT: setb %al
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i8 @llvm.x86.lwpins64(i64 %a0, i32 %a1, i32 2814)
ret i8 %2
@@ -29,9 +108,47 @@ define i8 @stack_fold_lwpins_u64(i64 %a0, i32 %a1) {
declare i8 @llvm.x86.lwpins64(i64, i32, i32)
define void @stack_fold_lwpval_u32(i32 %a0, i32 %a1) {
; CHECK-LABEL: stack_fold_lwpval_u32
; CHECK-LABEL: stack_fold_lwpval_u32:
; CHECK: # %bb.0:
; CHECK: lwpval $2814, {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
; CHECK-NEXT: lwpval $2814, {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: # imm = 0xAFE
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
tail call void @llvm.x86.lwpval32(i32 %a0, i32 %a1, i32 2814)
ret void
@@ -39,9 +156,47 @@ define void @stack_fold_lwpval_u32(i32 %a0, i32 %a1) {
declare void @llvm.x86.lwpval32(i32, i32, i32)
define void @stack_fold_lwpval_u64(i64 %a0, i32 %a1) {
; CHECK-LABEL: stack_fold_lwpval_u64
; CHECK-LABEL: stack_fold_lwpval_u64:
; CHECK: # %bb.0:
; CHECK: lwpval $2814, {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 4-byte Folded Reload
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %esi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: lwpval $2814, {{[-0-9]+}}(%r{{[sb]}}p), %rax # 4-byte Folded Reload
; CHECK-NEXT: # imm = 0xAFE
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
tail call void @llvm.x86.lwpval64(i64 %a0, i32 %a1, i32 2814)
ret void
test/CodeGen/X86/stack-folding-mmx.ll
+710 -156
View File
File diff suppressed because it is too large Load Diff
test/CodeGen/X86/stack-folding-sha.ll
+74 -14
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+sha < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -9,8 +10,16 @@ target triple = "x86_64-unknown-unknown"
; relevant registers and check that the reload is correctly folded into the instruction.
define <4 x i32> @stack_fold_sha1msg1(<4 x i32> %a0, <4 x i32> %a1) {
;CHECK-LABEL: stack_fold_sha1msg1
;CHECK: sha1msg1 {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK-LABEL: stack_fold_sha1msg1:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: sha1msg1 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
%2 = tail call <4 x i32> @llvm.x86.sha1msg1(<4 x i32> %a0, <4 x i32> %a1)
ret <4 x i32> %2
@@ -18,8 +27,16 @@ define <4 x i32> @stack_fold_sha1msg1(<4 x i32> %a0, <4 x i32> %a1) {
declare <4 x i32> @llvm.x86.sha1msg1(<4 x i32>, <4 x i32>) nounwind readnone
define <4 x i32> @stack_fold_sha1msg2(<4 x i32> %a0, <4 x i32> %a1) {
;CHECK-LABEL: stack_fold_sha1msg2
;CHECK: sha1msg2 {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK-LABEL: stack_fold_sha1msg2:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: sha1msg2 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
%2 = tail call <4 x i32> @llvm.x86.sha1msg2(<4 x i32> %a0, <4 x i32> %a1)
ret <4 x i32> %2
@@ -27,8 +44,16 @@ define <4 x i32> @stack_fold_sha1msg2(<4 x i32> %a0, <4 x i32> %a1) {
declare <4 x i32> @llvm.x86.sha1msg2(<4 x i32>, <4 x i32>) nounwind readnone
define <4 x i32> @stack_fold_sha1nexte(<4 x i32> %a0, <4 x i32> %a1) {
;CHECK-LABEL: stack_fold_sha1nexte
;CHECK: sha1nexte {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK-LABEL: stack_fold_sha1nexte:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: sha1nexte {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
%2 = tail call <4 x i32> @llvm.x86.sha1nexte(<4 x i32> %a0, <4 x i32> %a1)
ret <4 x i32> %2
@@ -36,8 +61,16 @@ define <4 x i32> @stack_fold_sha1nexte(<4 x i32> %a0, <4 x i32> %a1) {
declare <4 x i32> @llvm.x86.sha1nexte(<4 x i32>, <4 x i32>) nounwind readnone
define <4 x i32> @stack_fold_sha1rnds4(<4 x i32> %a0, <4 x i32> %a1) {
;CHECK-LABEL: stack_fold_sha1rnds4
;CHECK: sha1rnds4 $3, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK-LABEL: stack_fold_sha1rnds4:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: sha1rnds4 $3, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
%2 = tail call <4 x i32> @llvm.x86.sha1rnds4(<4 x i32> %a0, <4 x i32> %a1, i8 3)
ret <4 x i32> %2
@@ -45,8 +78,16 @@ define <4 x i32> @stack_fold_sha1rnds4(<4 x i32> %a0, <4 x i32> %a1) {
declare <4 x i32> @llvm.x86.sha1rnds4(<4 x i32>, <4 x i32>, i8) nounwind readnone
define <4 x i32> @stack_fold_sha256msg1(<4 x i32> %a0, <4 x i32> %a1) {
;CHECK-LABEL: stack_fold_sha256msg1
;CHECK: sha256msg1 {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK-LABEL: stack_fold_sha256msg1:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: sha256msg1 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
%2 = tail call <4 x i32> @llvm.x86.sha256msg1(<4 x i32> %a0, <4 x i32> %a1)
ret <4 x i32> %2
@@ -54,8 +95,16 @@ define <4 x i32> @stack_fold_sha256msg1(<4 x i32> %a0, <4 x i32> %a1) {
declare <4 x i32> @llvm.x86.sha256msg1(<4 x i32>, <4 x i32>) nounwind readnone
define <4 x i32> @stack_fold_sha256msg2(<4 x i32> %a0, <4 x i32> %a1) {
;CHECK-LABEL: stack_fold_sha256msg2
;CHECK: sha256msg2 {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK-LABEL: stack_fold_sha256msg2:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: sha256msg2 {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
%2 = tail call <4 x i32> @llvm.x86.sha256msg2(<4 x i32> %a0, <4 x i32> %a1)
ret <4 x i32> %2
@@ -63,8 +112,19 @@ define <4 x i32> @stack_fold_sha256msg2(<4 x i32> %a0, <4 x i32> %a1) {
declare <4 x i32> @llvm.x86.sha256msg2(<4 x i32>, <4 x i32>) nounwind readnone
define <4 x i32> @stack_fold_sha256rnds2(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> %a2) {
;CHECK-LABEL: stack_fold_sha256rnds2
;CHECK: sha256rnds2 {{%xmm[0-9][0-9]*}}, {{-?[0-9]*}}(%rsp), {{%xmm[0-9][0-9]*}} {{.*#+}} 16-byte Folded Reload
; CHECK-LABEL: stack_fold_sha256rnds2:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; CHECK-NEXT: sha256rnds2 %xmm0, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; CHECK-NEXT: movaps %xmm1, %xmm0
; CHECK-NEXT: retq
%1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{flags}"()
%2 = tail call <4 x i32> @llvm.x86.sha256rnds2(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> %a2)
ret <4 x i32> %2
test/CodeGen/X86/stack-folding-tbm.ll
+763 -42
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+bmi,+tbm < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -9,9 +10,45 @@ target triple = "x86_64-unknown-unknown"
; relevant registers and check that the reload is correctly folded into the instruction.
define i32 @stack_fold_bextri_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_bextri_u32
;CHECK: # %bb.0:
;CHECK: bextrl $3841, {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_bextri_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: bextrl $3841, {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: # imm = 0xF01
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.x86.tbm.bextri.u32(i32 %a0, i32 3841)
ret i32 %2
@@ -19,9 +56,45 @@ define i32 @stack_fold_bextri_u32(i32 %a0) {
declare i32 @llvm.x86.tbm.bextri.u32(i32, i32)
define i64 @stack_fold_bextri_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_bextri_u64
;CHECK: # %bb.0:
;CHECK: bextrq $3841, {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_bextri_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: bextrq $3841, {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: # imm = 0xF01
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.x86.tbm.bextri.u64(i64 %a0, i64 3841)
ret i64 %2
@@ -29,8 +102,44 @@ define i64 @stack_fold_bextri_u64(i64 %a0) {
declare i64 @llvm.x86.tbm.bextri.u64(i64, i64)
define i32 @stack_fold_blcfill_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blcfill_u32
;CHECK: blcfilll {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blcfill_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcfilll {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i32 %a0, 1
%3 = and i32 %a0, %2
@@ -38,8 +147,44 @@ define i32 @stack_fold_blcfill_u32(i32 %a0) {
}
define i64 @stack_fold_blcfill_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blcfill_u64
;CHECK: blcfillq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blcfill_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcfillq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i64 %a0, 1
%3 = and i64 %a0, %2
@@ -47,8 +192,44 @@ define i64 @stack_fold_blcfill_u64(i64 %a0) {
}
define i32 @stack_fold_blci_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blci_u32
;CHECK: blcil {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blci_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcil {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i32 %a0, 1
%3 = xor i32 %2, -1
@@ -57,8 +238,44 @@ define i32 @stack_fold_blci_u32(i32 %a0) {
}
define i64 @stack_fold_blci_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blci_u64
;CHECK: blciq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blci_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blciq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i64 %a0, 1
%3 = xor i64 %2, -1
@@ -67,8 +284,44 @@ define i64 @stack_fold_blci_u64(i64 %a0) {
}
define i32 @stack_fold_blcic_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blcic_u32
;CHECK: blcicl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blcic_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcicl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i32 %a0, 1
%3 = xor i32 %a0, -1
@@ -77,8 +330,44 @@ define i32 @stack_fold_blcic_u32(i32 %a0) {
}
define i64 @stack_fold_blcic_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blcic_u64
;CHECK: blcicq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blcic_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcicq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i64 %a0, 1
%3 = xor i64 %a0, -1
@@ -87,8 +376,44 @@ define i64 @stack_fold_blcic_u64(i64 %a0) {
}
define i32 @stack_fold_blcmsk_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blcmsk_u32
;CHECK: blcmskl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blcmsk_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcmskl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i32 %a0, 1
%3 = xor i32 %a0, %2
@@ -96,8 +421,44 @@ define i32 @stack_fold_blcmsk_u32(i32 %a0) {
}
define i64 @stack_fold_blcmsk_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blcmsk_u64
;CHECK: blcmskq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blcmsk_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcmskq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i64 %a0, 1
%3 = xor i64 %a0, %2
@@ -105,8 +466,44 @@ define i64 @stack_fold_blcmsk_u64(i64 %a0) {
}
define i32 @stack_fold_blcs_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blcs_u32
;CHECK: blcsl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blcs_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcsl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i32 %a0, 1
%3 = or i32 %a0, %2
@@ -114,8 +511,44 @@ define i32 @stack_fold_blcs_u32(i32 %a0) {
}
define i64 @stack_fold_blcs_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blcs_u64
;CHECK: blcsq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blcs_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blcsq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i64 %a0, 1
%3 = or i64 %a0, %2
@@ -123,8 +556,44 @@ define i64 @stack_fold_blcs_u64(i64 %a0) {
}
define i32 @stack_fold_blsfill_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blsfill_u32
;CHECK: blsfilll {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blsfill_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsfilll {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i32 %a0, 1
%3 = or i32 %a0, %2
@@ -132,8 +601,44 @@ define i32 @stack_fold_blsfill_u32(i32 %a0) {
}
define i64 @stack_fold_blsfill_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blsfill_u64
;CHECK: blsfillq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blsfill_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsfillq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i64 %a0, 1
%3 = or i64 %a0, %2
@@ -141,8 +646,44 @@ define i64 @stack_fold_blsfill_u64(i64 %a0) {
}
define i32 @stack_fold_blsic_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_blsic_u32
;CHECK: blsicl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_blsic_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsicl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i32 %a0, 1
%3 = xor i32 %a0, -1
@@ -151,8 +692,44 @@ define i32 @stack_fold_blsic_u32(i32 %a0) {
}
define i64 @stack_fold_blsic_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_blsic_u64
;CHECK: blsicq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_blsic_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: blsicq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i64 %a0, 1
%3 = xor i64 %a0, -1
@@ -161,8 +738,44 @@ define i64 @stack_fold_blsic_u64(i64 %a0) {
}
define i32 @stack_fold_t1mskc_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_t1mskc_u32
;CHECK: t1mskcl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_t1mskc_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: t1mskcl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i32 %a0, 1
%3 = xor i32 %a0, -1
@@ -171,8 +784,44 @@ define i32 @stack_fold_t1mskc_u32(i32 %a0) {
}
define i64 @stack_fold_t1mskc_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_t1mskc_u64
;CHECK: t1mskcq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_t1mskc_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: t1mskcq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = add i64 %a0, 1
%3 = xor i64 %a0, -1
@@ -181,8 +830,44 @@ define i64 @stack_fold_t1mskc_u64(i64 %a0) {
}
define i32 @stack_fold_tzmsk_u32(i32 %a0) {
;CHECK-LABEL: stack_fold_tzmsk_u32
;CHECK: tzmskl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_tzmsk_u32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: tzmskl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i32 %a0, 1
%3 = xor i32 %a0, -1
@@ -191,8 +876,44 @@ define i32 @stack_fold_tzmsk_u32(i32 %a0) {
}
define i64 @stack_fold_tzmsk_u64(i64 %a0) {
;CHECK-LABEL: stack_fold_tzmsk_u64
;CHECK: tzmskq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_tzmsk_u64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: tzmskq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = sub i64 %a0, 1
%3 = xor i64 %a0, -1
test/CodeGen/X86/stack-folding-x86_64.ll
+155 -8
View File
@@ -1,3 +1,4 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
@@ -12,8 +13,44 @@ target triple = "x86_64-unknown-unknown"
declare i16 @llvm.cttz.i16(i16, i1)
define i32 @stack_fold_bsf_i32(i32 %a0) {
;CHECK-LABEL: stack_fold_bsf_i32
;CHECK: bsfl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_bsf_i32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: bsfl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call i32 @llvm.cttz.i32(i32 %a0, i1 -1)
ret i32 %2
@@ -21,8 +58,44 @@ define i32 @stack_fold_bsf_i32(i32 %a0) {
declare i32 @llvm.cttz.i32(i32, i1)
define i64 @stack_fold_bsf_i64(i64 %a0) {
;CHECK-LABEL: stack_fold_bsf_i64
;CHECK: bsfq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_bsf_i64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: bsfq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call i64 @llvm.cttz.i64(i64 %a0, i1 -1)
ret i64 %2
@@ -33,8 +106,45 @@ declare i64 @llvm.cttz.i64(i64, i1)
declare i16 @llvm.ctlz.i16(i16, i1)
define i32 @stack_fold_bsr_i32(i32 %a0) {
;CHECK-LABEL: stack_fold_bsr_i32
;CHECK: bsrl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
; CHECK-LABEL: stack_fold_bsr_i32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: bsrl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Folded Reload
; CHECK-NEXT: xorl $31, %eax
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call i32 @llvm.ctlz.i32(i32 %a0, i1 -1)
ret i32 %2
@@ -42,8 +152,45 @@ define i32 @stack_fold_bsr_i32(i32 %a0) {
declare i32 @llvm.ctlz.i32(i32, i1)
define i64 @stack_fold_bsr_i64(i64 %a0) {
;CHECK-LABEL: stack_fold_bsr_i64
;CHECK: bsrq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
; CHECK-LABEL: stack_fold_bsr_i64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 56
; CHECK-NEXT: .cfi_offset %rbx, -56
; CHECK-NEXT: .cfi_offset %r12, -48
; CHECK-NEXT: .cfi_offset %r13, -40
; CHECK-NEXT: .cfi_offset %r14, -32
; CHECK-NEXT: .cfi_offset %r15, -24
; CHECK-NEXT: .cfi_offset %rbp, -16
; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: nop
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: bsrq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
; CHECK-NEXT: xorq $63, %rax
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: popq %r12
; CHECK-NEXT: .cfi_def_cfa_offset 40
; CHECK-NEXT: popq %r13
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: popq %r14
; CHECK-NEXT: .cfi_def_cfa_offset 24
; CHECK-NEXT: popq %r15
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: .cfi_def_cfa_offset 8
; CHECK-NEXT: retq
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = call i64 @llvm.ctlz.i64(i64 %a0, i1 -1)
ret i64 %2
test/CodeGen/X86/stack-folding-xop.ll
+689 -166
View File
File diff suppressed because it is too large Load Diff