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.

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

[StackSlotColoring] Ignore non-spill objects in RemoveDeadStores. #80242

Merged
merged 3 commits into from
Feb 1, 2024
Merged
Show file tree
Hide file tree
Changes from 2 commits
Commits
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
2 changes: 1 addition & 1 deletion llvm/lib/CodeGen/StackSlotColoring.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -480,7 +480,7 @@ bool StackSlotColoring::RemoveDeadStores(MachineBasicBlock* MBB) {
if (!(StoreReg = TII->isStoreToStackSlot(*NextMI, SecondSS, StoreSize)))
continue;
if (FirstSS != SecondSS || LoadReg != StoreReg || FirstSS == -1 ||
LoadSize != StoreSize)
LoadSize != StoreSize || !MFI->isSpillSlotObjectIndex(FirstSS))
continue;

++NumDead;
Expand Down
125 changes: 125 additions & 0 deletions llvm/test/CodeGen/RISCV/pr80052.mir
Original file line number Diff line number Diff line change
@@ -0,0 +1,125 @@
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 4
# RUN: llc %s -mtriple=riscv64 -run-pass=greedy,virtregrewriter,stack-slot-coloring -o - | FileCheck %s

--- |
define void @foo() {
entry:
; Dummy test just to hold the alloca
%alpha = alloca i32, i32 0, align 4
ret void
}

...
---
name: foo
alignment: 4
tracksRegLiveness: true
frameInfo:
maxAlignment: 4
localFrameSize: 4
stack:
- { id: 0, name: alpha, size: 1, alignment: 4, local-offset: -4 }
machineFunctionInfo:
varArgsFrameIndex: 0
varArgsSaveSize: 0
body: |
bb.0.entry:
; To trick stack-slot-colouring to run its dead-store-elimination phase,
; which is at fault, we need the register allocator to run, and spill in two
; places that can have their slots merged. Achieve this by volatile-loading
; data into all allocatable GPRs except $x31. Then, volatile storing them
; later, leaving regalloc only $x31 to play with in the middle.
$x1 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
Copy link
Contributor

Choose a reason for hiding this comment

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

you can replace all the MMO IR references with %stack.0 and drop the IR section

$x5 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x6 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x7 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x8 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x9 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x10 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x11 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x12 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x13 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x14 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x15 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x16 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x17 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x18 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x19 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x20 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x21 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x22 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x23 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x24 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x25 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x26 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x27 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x28 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x29 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
$x30 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)

; Force the first spill.
%0:gpr = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
%1:gpr = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
SW %1, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW %0, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)

; CHECK: renamable $x31 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
; CHECK-NEXT: SD killed renamable $x31, %stack.1, 0 :: (store (s64) into %stack.1)
; CHECK-NEXT: renamable $x31 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
; CHECK-NEXT: SW killed renamable $x31, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
; CHECK-NEXT: renamable $x31 = LD %stack.1, 0 :: (load (s64) from %stack.1)
; CHECK-NEXT: SW killed renamable $x31, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)

; stack-slot-coloring doesn't know that a write to $x0 is discarded.
dead $x0 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
; This stores 0 rather than the result of the preceding load since $x0
; is special.
; We don't want this store to be deleted.
SW $x0, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)

; CHECK-NEXT: dead $x0 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
; CHECK-NEXT: SW $x0, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)

; Force a second spill
%2:gpr = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
%3:gpr = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
SW %3, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW %2, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)

; CHECK-NEXT: renamable $x31 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
; CHECK-NEXT: SD killed renamable $x31, %stack.1, 0 :: (store (s64) into %stack.1)
; CHECK-NEXT: renamable $x31 = LW %stack.0.alpha, 0 :: (volatile load (s32) from %ir.alpha)
; CHECK-NEXT: SW killed renamable $x31, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
; CHECK-NEXT: renamable $x31 = LD %stack.1, 0 :: (load (s64) from %stack.1)
; CHECK-NEXT: SW killed renamable $x31, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)

SW $x1, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x5, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x6, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x7, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x8, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x9, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x10, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x11, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x12, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x13, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x14, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x15, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x16, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x17, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x18, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x19, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x20, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x21, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x22, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x23, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x24, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x25, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x26, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x27, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x28, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x29, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
SW $x30, %stack.0.alpha, 0 :: (volatile store (s32) into %ir.alpha)
PseudoRET

...
2 changes: 1 addition & 1 deletion llvm/test/CodeGen/X86/pr30821.mir
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,7 @@ stack:
- { id: 1, name: foxtrot, type: default, offset: 0, size: 16, alignment: 16,
stack-id: default, callee-saved-register: '', callee-saved-restored: true,
debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
- { id: 2, name: india, type: default, offset: 0, size: 16, alignment: 16,
- { id: 2, name: india, type: spill-slot, offset: 0, size: 16, alignment: 16,
stack-id: default, callee-saved-register: '', callee-saved-restored: true,
debug-info-variable: '', debug-info-expression: '',
debug-info-location: '' }
Expand Down
Loading