Skip to content

[X86][APX] Prevent from emitting push2/pop2 if stack alignment<16B #143076

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jun 6, 2025
Merged

[X86][APX] Prevent from emitting push2/pop2 if stack alignment<16B #143076

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4c71a98
[X86][APX] Prevent from emitting push2/pop2 when stack alignment < 16…
fzou1 Jun 6, 2025
0f8e848
Update comments and remove the complicated sub test.
fzou1 Jun 6, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion llvm/lib/Target/X86/X86FrameLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2915,13 +2915,14 @@ bool X86FrameLowering::assignCalleeSavedSpillSlots(
// 1. Use push2 when
// a) number of CSR > 1 if no need padding
// b) number of CSR > 2 if need padding
// c) stack alignment >= 16 bytes
// 2. When the number of CSR push is odd
// a. Start to use push2 from the 1st push if stack is 16B aligned.
// b. Start to use push2 from the 2nd push if stack is not 16B aligned.
// 3. When the number of CSR push is even, start to use push2 from the 1st
// push and make the stack 16B aligned before the push
unsigned NumRegsForPush2 = 0;
if (STI.hasPush2Pop2()) {
if (STI.hasPush2Pop2() && getStackAlignment() >= 16) {
Copy link
Contributor

@KanRobert KanRobert Jun 6, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If this solution is better, you should mention the alignment requirement in above comments.

Copy link
Contributor Author

@fzou1 fzou1 Jun 6, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Updated.

unsigned NumCSGPR = llvm::count_if(CSI, [](const CalleeSavedInfo &I) {
return X86::GR64RegClass.contains(I.getReg());
});
Expand Down
2 changes: 1 addition & 1 deletion llvm/lib/Target/X86/X86MachineFunctionInfo.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -149,7 +149,7 @@ class X86MachineFunctionInfo : public MachineFunctionInfo {
/// other tools to detect the extended record.
bool HasSwiftAsyncContext = false;

/// Ajust stack for push2/pop2
/// Adjust stack for push2/pop2
bool PadForPush2Pop2 = false;

/// Candidate registers for push2/pop2
Expand Down
117 changes: 117 additions & 0 deletions llvm/test/CodeGen/X86/apx/push2-pop2-disabled-with-small-stack-alignment.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,117 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+push2pop2 | FileCheck %s --check-prefix=CHECK

; This test is used to check no push2/pop2 emitted if stack alignment is set to
; the value less than 16 bytes required by push2/pop2 instruction. Here it's set
; to 8 bytes.

define void @csr1() nounwind {
; CHECK-LABEL: csr1:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: #APP
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
entry:
tail call void asm sideeffect "", "~{rbp},~{dirflag},~{fpsr},~{flags}"()
ret void
}

define void @csr2() nounwind {
; CHECK-LABEL: csr2:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: #APP
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: popq %r15
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
entry:
tail call void asm sideeffect "", "~{rbp},~{r15},~{dirflag},~{fpsr},~{flags}"()
ret void
}

define void @csr3() nounwind {
; CHECK-LABEL: csr3:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: #APP
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
entry:
tail call void asm sideeffect "", "~{rbp},~{r15},~{r14},~{dirflag},~{fpsr},~{flags}"()
ret void
}

define void @csr4() nounwind {
; CHECK-LABEL: csr4:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: #APP
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: popq %r13
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
entry:
tail call void asm sideeffect "", "~{rbp},~{r15},~{r14},~{r13},~{dirflag},~{fpsr},~{flags}"()
ret void
}

define void @csr5() nounwind {
; CHECK-LABEL: csr5:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: #APP
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: popq %r12
; CHECK-NEXT: popq %r13
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
entry:
tail call void asm sideeffect "", "~{rbp},~{r15},~{r14},~{r13},~{r12},~{dirflag},~{fpsr},~{flags}"()
ret void
}

define void @csr6() nounwind {
; CHECK-LABEL: csr6:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: #APP
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %r12
; CHECK-NEXT: popq %r13
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
entry:
tail call void asm sideeffect "", "~{rbp},~{r15},~{r14},~{r13},~{r12},~{rbx},~{dirflag},~{fpsr},~{flags}"()
ret void
}

!llvm.module.flags = !{!0}
!0 = !{i32 1, !"override-stack-alignment", i32 8}
Loading