[MachineLICM] Don't allow hoisting invariant loads across mem barrier. #116987
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
fhahn
requested review from
RKSimon,
jroelofs,
davemgreen,
igogo-x86 and
david-arm
|
@llvm/pr-subscribers-backend-aarch64 Author: Florian Hahn (fhahn) ChangesThe improvements in 63917e1 / #70796 do not check for memory barriers/unmodelled sideeffects, which means we may incorrectly hoist loads across memory barriers. Fix this by checking any machine instruction in the loop is a load-fold barrier. Full diff: https://github.com/llvm/llvm-project/pull/116987.diff 2 Files Affected:
diff --git a/llvm/lib/CodeGen/MachineLICM.cpp b/llvm/lib/CodeGen/MachineLICM.cpp
index 48c901b8d06d61..00e3b5def5d108 100644
--- a/llvm/lib/CodeGen/MachineLICM.cpp
+++ b/llvm/lib/CodeGen/MachineLICM.cpp
@@ -1501,7 +1501,7 @@ void MachineLICMImpl::InitializeLoadsHoistableLoops() {
if (!AllowedToHoistLoads[Loop])
continue;
for (auto &MI : *MBB) {
- if (!MI.mayStore() && !MI.isCall() &&
+ if (!MI.isLoadFoldBarrier() && !MI.mayStore() && !MI.isCall() &&
!(MI.mayLoad() && MI.hasOrderedMemoryRef()))
continue;
for (MachineLoop *L = Loop; L != nullptr; L = L->getParentLoop())
diff --git a/llvm/test/CodeGen/AArch64/machine-licm-hoist-load.ll b/llvm/test/CodeGen/AArch64/machine-licm-hoist-load.ll
index 932a5af264a000..17f8263560430d 100644
--- a/llvm/test/CodeGen/AArch64/machine-licm-hoist-load.ll
+++ b/llvm/test/CodeGen/AArch64/machine-licm-hoist-load.ll
@@ -499,16 +499,16 @@ for.exit: ; preds = %for.body
@a = external local_unnamed_addr global i32, align 4
-; FIXME: Load hoisted out of the loop across memory barriers.
+; Make sure the load is not hoisted out of the loop across memory barriers.
define i32 @load_between_memory_barriers() {
; CHECK-LABEL: load_between_memory_barriers:
; CHECK: // %bb.0:
; CHECK-NEXT: adrp x8, :got:a
; CHECK-NEXT: ldr x8, [x8, :got_lo12:a]
-; CHECK-NEXT: ldr w0, [x8]
; CHECK-NEXT: .LBB8_1: // %loop
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: //MEMBARRIER
+; CHECK-NEXT: ldr w0, [x8]
; CHECK-NEXT: //MEMBARRIER
; CHECK-NEXT: cbz w0, .LBB8_1
; CHECK-NEXT: // %bb.2: // %exit
|
The improvements in 63917e1 / llvm#70796 do not check for memory barriers/unmodelled sideeffects, which means we may incorrectly hoist loads across memory barriers. Fix this by checking any machine instruction in the loop is a load-fold barrier.
fhahn
force-pushed
the
machinelicm-membarrier
branch
from
d5b0f71 to
b62f928
Compare
|
/pull-request #117154 |
tru
pushed a commit
to llvmbot/llvm-project
that referenced
this pull request
Nov 25, 2024
llvm#116987) The improvements in 63917e1 / llvm#70796 do not check for memory barriers/unmodelled sideeffects, which means we may incorrectly hoist loads across memory barriers. Fix this by checking any machine instruction in the loop is a load-fold barrier. PR: llvm#116987 (cherry picked from commit ef102b4)
fhahn
added a commit
to fhahn/llvm-project
that referenced
this pull request
Jan 6, 2025
llvm#116987) The improvements in 63917e1 / llvm#70796 do not check for memory barriers/unmodelled sideeffects, which means we may incorrectly hoist loads across memory barriers. Fix this by checking any machine instruction in the loop is a load-fold barrier. PR: llvm#116987 (cherry picked from commit ef102b4)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The improvements in 63917e1 / #70796 do not check for memory barriers/unmodelled sideeffects, which means we may incorrectly hoist loads across memory barriers.
Fix this by checking any machine instruction in the loop is a load-fold barrier.