Skip to content

[bolt] Support arm64 FP register spills #73021

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 1 commit into from
Dec 5, 2023
Merged

[bolt] Support arm64 FP register spills #73021

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 6 additions & 0 deletions bolt/lib/Target/AArch64/AArch64MCPlusBuilder.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -314,6 +314,12 @@ class AArch64MCPlusBuilder : public MCPlusBuilder {

bool isRegToRegMove(const MCInst &Inst, MCPhysReg &From,
MCPhysReg &To) const override {
if (Inst.getOpcode() == AArch64::FMOVDXr) {
From = Inst.getOperand(1).getReg();
To = Inst.getOperand(0).getReg();
return true;
}
Copy link
Member

Choose a reason for hiding this comment

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

new line please after }


if (Inst.getOpcode() != AArch64::ORRXrs)
return false;
if (Inst.getOperand(1).getReg() != AArch64::XZR)
Expand Down
19 changes: 19 additions & 0 deletions bolt/test/AArch64/fp-reg-spill.s
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,19 @@
# Check that we correctly handle arm64 fp register spills in
# bolt when we are processing jump tables.
Comment on lines +1 to +2
Copy link
Contributor

Choose a reason for hiding this comment

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

Sorry for jumping in late after the change landed. This test checks that BOLT doesn't crash, but it doesn't really check for the correct handling of the jump table. You'll need to add FileCheck checks for --print-cfg dumps or other checks for the output.

Copy link
Member

Choose a reason for hiding this comment

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

The idea behind the test was not to check the jump table (although it would be nice), but the previous bolt assertion during analyzation. So the test is good-enough I think

# REQUIRES: system-linux
Copy link
Member

Choose a reason for hiding this comment

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

Please add test scription at the beginning of the file

# RUN: llvm-mc -filetype=obj -triple=aarch64 %s -o %t.o
# RUN: ld.lld --emit-relocs %t.o -o %t.elf
# RUN: llvm-bolt --jump-tables=move %t.elf -o %t.bolt

.globl _foo, _start

_foo:
ret

_start:
adr x6, _foo
fmov d18,x6
fmov x5,d18
ldrb w5, [x5, w1, uxtw]
add x5, x6, w5, sxtb #2
br x5