Skip to content

[MemProf] Improve metadata cleanup in LTO backend #113039

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 3 commits into from
Oct 21, 2024
Merged

[MemProf] Improve metadata cleanup in LTO backend #113039

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a5cbba5
[MemProf] Improve metadata cleanup in LTO backend
teresajohnson Oct 19, 2024
f1c49e7
Remove stale comment
teresajohnson Oct 19, 2024
d41836b
Address comment
teresajohnson Oct 21, 2024
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
23 changes: 17 additions & 6 deletions llvm/lib/Transforms/IPO/MemProfContextDisambiguation.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4264,9 +4264,6 @@ bool MemProfContextDisambiguation::applyImport(Module &M) {
AllocVersionsThinBackend++;
if (!MaxAllocVersionsThinBackend)
MaxAllocVersionsThinBackend = 1;
// Remove any remaining callsite metadata and we can skip the rest of
// the handling for this instruction, since no cloning needed.
I.setMetadata(LLVMContext::MD_callsite, nullptr);
continue;
}

Expand Down Expand Up @@ -4419,16 +4416,30 @@ bool MemProfContextDisambiguation::applyImport(Module &M) {
CloneCallsite(Callsite->second, CB, CalledFunction);
}
}
// Memprof and callsite metadata on memory allocations no longer needed.
I.setMetadata(LLVMContext::MD_memprof, nullptr);
I.setMetadata(LLVMContext::MD_callsite, nullptr);
}
}

// Now do any promotion required for cloning.
performICP(M, FS->callsites(), VMaps, ICallAnalysisInfo, ORE);
}

// We skip some of the functions and instructions above, so remove all the
// metadata in a single sweep here.
for (auto &F : M) {
// We can skip memprof clones because createFunctionClones already strips
// the metadata from the newly created clones.
if (F.isDeclaration() || isMemProfClone(F))
continue;
for (auto &BB : F) {
for (auto &I : BB) {
Copy link

@snehasish snehasish Oct 20, 2024

Choose a reason for hiding this comment

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

Should we also filter by CallInst to so that we don't call setMetadata unnecessarily?

Copy link
Contributor Author

@teresajohnson teresajohnson Oct 21, 2024

Choose a reason for hiding this comment

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

done

if (!isa<CallBase>(I))
continue;
I.setMetadata(LLVMContext::MD_memprof, nullptr);
I.setMetadata(LLVMContext::MD_callsite, nullptr);
}
}
}

return Changed;
}

Expand Down
5 changes: 4 additions & 1 deletion llvm/test/ThinLTO/X86/memprof-icp.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -176,7 +176,10 @@
; RUN: -pass-remarks=. -save-temps \
; RUN: -o %t.noicp.out 2>&1 | FileCheck %s --implicit-check-not "created clone"

; RUN: llvm-dis %t.noicp.out.2.4.opt.bc -o - | FileCheck %s --implicit-check-not "_Z3fooR2B0j.memprof"
;; Verify that we did not do any cloning of the function with the indirect call
;; when memprof ICP is off. However, we should still have removed the callsite
;; metadata.
; RUN: llvm-dis %t.noicp.out.2.4.opt.bc -o - | FileCheck %s --implicit-check-not "_Z3fooR2B0j.memprof" --implicit-check-not "!callsite"

; REMARKS-MAIN: call in clone main assigned to call function clone _Z3fooR2B0j.memprof.1
; REMARKS-MAIN: call in clone main assigned to call function clone _Z3fooR2B0j.memprof.1
Expand Down
Loading