Skip to content
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

[ARM] Check all terms in emitPopInst when clearing Restored for LR. #75527

Merged
merged 4 commits into from
Dec 20, 2023
Merged

[ARM] Check all terms in emitPopInst when clearing Restored for LR. #75527

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f6896d1
[ARM] Check all terms in emitPopInst when clearing Restored for LR.
fhahn Dec 14, 2023
cd4021f
!fixup move setting IsRestored to processFunctionBeforeFrameFinalized
fhahn Dec 19, 2023
53d8fc4
!fixup remove FIXME.
fhahn Dec 19, 2023
0f799ea
!fixup add break after setRestored.
fhahn Dec 20, 2023
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
18 changes: 15 additions & 3 deletions llvm/lib/Target/ARM/ARMFrameLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1645,9 +1645,21 @@ void ARMFrameLowering::emitPopInst(MachineBasicBlock &MBB,
// Fold the return instruction into the LDM.
DeleteRet = true;
LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_RET : ARM::LDMIA_RET;
// We 'restore' LR into PC so it is not live out of the return block:
// Clear Restored bit.
Info.setRestored(false);
// Check if all terminators do not implicitly use LR. Then we can
// 'restore' LR into PC so it is not live out of the return block: Clear
// Restored bit.
if (all_of(MF, [MI](const MachineBasicBlock &MBB) {
return all_of(MBB.terminators(), [MI](const MachineInstr &Term) {
// MI's terminator is to be re-written, don't check the old
// opcode.
if (&*MI == &Term)
return true;
return Term.getOpcode() == ARM::LDMIA_RET ||
Term.getOpcode() == ARM::t2LDMIA_RET ||
Term.getOpcode() == ARM::tPOP_RET;
Copy link
Collaborator

Choose a reason for hiding this comment

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

Should we checking for isReturn()? This is going to reject any block containing a branch.

Checking all blocks here is O(N^2) in the number of epilogues; can we move this later somehow?

There's corresponding code in Thumb1FrameLowering.cpp which needs a similar fix.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Thanks, I added an isReturn check in the latest version and moved setting IsRestored to processFunctionBeforeFrameFinalized, which runs just after spilling the callee-saved registers

});
}))
Info.setRestored(false);
}

// If NoGap is true, pop consecutive registers and then leave the rest
Expand Down
1 change: 1 addition & 0 deletions llvm/test/CodeGen/Thumb2/mve-float16regloops.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -831,6 +831,7 @@ define void @arm_fir_f32_1_4_mve(ptr nocapture readonly %S, ptr nocapture readon
; CHECK-NEXT: mov r0, r1
; CHECK-NEXT: .LBB15_10: @ %while.end55
; CHECK-NEXT: ands r1, r9, #3
; CHECK-NEXT: @ implicit-def: $lr
; CHECK-NEXT: beq .LBB15_12
; CHECK-NEXT: @ %bb.11: @ %if.then59
; CHECK-NEXT: vldrw.u32 q0, [r0]
Expand Down
1 change: 1 addition & 0 deletions llvm/test/CodeGen/Thumb2/mve-float32regloops.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -822,6 +822,7 @@ define void @arm_fir_f32_1_4_mve(ptr nocapture readonly %S, ptr nocapture readon
; CHECK-NEXT: mov r0, r1
; CHECK-NEXT: .LBB15_10: @ %while.end55
; CHECK-NEXT: ands r1, r10, #3
; CHECK-NEXT: @ implicit-def: $lr
; CHECK-NEXT: beq .LBB15_12
; CHECK-NEXT: @ %bb.11: @ %if.then59
; CHECK-NEXT: vldrw.u32 q0, [r0]
Expand Down
12 changes: 10 additions & 2 deletions llvm/test/CodeGen/Thumb2/outlined-fn-may-clobber-lr-in-caller.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,11 +22,19 @@ define void @test(ptr nocapture noundef writeonly %arg, i32 noundef %arg1, i8 no
; CHECK-NEXT: cmp r1, #1
; CHECK-NEXT: bne .LBB0_5
; CHECK-NEXT: @ %bb.2: @ %bb4
; CHECK-NEXT: bl OUTLINED_FUNCTION_0
; CHECK-NEXT: movs r1, #1
; CHECK-NEXT: strb.w r1, [r0, #36]
; CHECK-NEXT: movs r1, #30
; CHECK-NEXT: strb.w r1, [r0, #34]
; CHECK-NEXT: add.w r1, r2, r2, lsl #3
; CHECK-NEXT: ldr r2, .LCPI0_1
; CHECK-NEXT: b .LBB0_4
; CHECK-NEXT: .LBB0_3: @ %bb14
; CHECK-NEXT: bl OUTLINED_FUNCTION_0
; CHECK-NEXT: movs r1, #1
; CHECK-NEXT: strb.w r1, [r0, #36]
; CHECK-NEXT: movs r1, #30
; CHECK-NEXT: strb.w r1, [r0, #34]
; CHECK-NEXT: add.w r1, r2, r2, lsl #3
; CHECK-NEXT: ldr r2, .LCPI0_0
; CHECK-NEXT: .LBB0_4: @ %bb4
; CHECK-NEXT: add.w r1, r2, r1, lsl #2
Expand Down