Skip to content

[WIP][InstCombine] Add assume-based optimizations for equality and AMDGPU ballot patterns #160670

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

Draft
wants to merge 6 commits into
base: main
Choose a base branch
from
+47 −0
Draft

[WIP][InstCombine] Add assume-based optimizations for equality and AMDGPU ballot patterns #160670

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
ff9e194
[InstCombine] Optimize AMDGPU ballot + assume uniformity patterns
TejaX-Alaghari Sep 25, 2025
d86e492
[InstCombine] Add constant folding for AMDGPU ballot intrinsics
TejaX-Alaghari Sep 29, 2025
a711d52
[InstCombine] Implement generic assume-based uniformity optimization
TejaX-Alaghari Sep 29, 2025
b16fe02
[InstCombine] Add focused assume-based optimizations
TejaX-Alaghari Oct 2, 2025
40eeaff
Address @ssahasra's review feedback
TejaX-Alaghari Oct 5, 2025
08355f9
Address feedback on the location of the opt
TejaX-Alaghari Oct 7, 2025
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: 18 additions & 0 deletions llvm/lib/Transforms/Scalar/GVN.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,7 @@
#include "llvm/IR/Instruction.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/IntrinsicsAMDGPU.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Metadata.h"
#include "llvm/IR/Module.h"
Expand Down Expand Up @@ -2206,6 +2207,23 @@ bool GVNPass::processAssumeIntrinsic(AssumeInst *IntrinsicI) {
std::swap(CmpLHS, CmpRHS);
}

// Optimize AMDGPU ballot pattern: assume(ballot(cmp) == -1) or
// assume(ballot(cmp) == exec_mask). This implies cmp is true on all
// active lanes and hence can be replaced with true.
if (isa<IntrinsicInst>(CmpLHS) && isa<Constant>(CmpRHS)) {
auto *IntrCall = cast<IntrinsicInst>(CmpLHS);
// Check if CmpLHS is a ballot intrinsic
if (IntrCall->getIntrinsicID() ==
Intrinsic::AMDGCNIntrinsics::amdgcn_ballot) {
Value *BallotArg = IntrCall->getArgOperand(0);
if (BallotArg->getType()->isIntegerTy(1) &&
(match(CmpRHS, m_AllOnes()) || !isa<Constant>(CmpRHS))) {
Copy link
Collaborator

Choose a reason for hiding this comment

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

The second part here, !isa<Constant>(CmpRHS) ... is this meant to be execmask? What if it is simply a value that the programmer knows will match the current state of cmp across active threads? For example:

unsigned threadmask = foo(); // where a bit is 1 for whatever reason
if (divergent_condition) {
  cmp = bar();
  assume(ballot(cmp) == threadmask);
}

Copy link
Author

@TejaX-Alaghari TejaX-Alaghari Oct 9, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

Per the discussion so far, I can infer that the %cmp value is not to be directly replaced with true after assume instruction.

Instead, the decision should be made depending the RHS of assume instruction, which could be -1 or some %mask (const/var).

  • -1: %cmp is true across all threads.
    %cmp = true
  • %mask: %cmp is only true for the masked threads
    %cmp = (0b1) & (%mask >> thread_id)

Please correct my understanding if I'm missing something.

CmpLHS = BallotArg;
CmpRHS = ConstantInt::getTrue(BallotArg->getType());
}
}
}

// Handle degenerate case where we either haven't pruned a dead path or a
// removed a trivial assume yet.
if (isa<Constant>(CmpLHS) && isa<Constant>(CmpRHS))
Expand Down
29 changes: 29 additions & 0 deletions llvm/test/Transforms/GVN/assume-equal.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -387,6 +387,35 @@ define i8 @assume_ptr_eq_same_prov(ptr %p, i64 %x) {
ret i8 %v
}

; Test AMDGPU ballot pattern optimization
; assume(ballot(cmp) == -1) means cmp is true on all active lanes
; so uses of cmp can be replaced with true
define void @assume_ballot_uniform(i32 %x) {
; CHECK-LABEL: @assume_ballot_uniform(
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[X:%.*]], 0
; CHECK-NEXT: [[BALLOT:%.*]] = call i64 @llvm.amdgcn.ballot.i64(i1 [[CMP]])
; CHECK-NEXT: [[ALL:%.*]] = icmp eq i64 [[BALLOT]], -1
; CHECK-NEXT: call void @llvm.assume(i1 [[ALL]])
; CHECK-NEXT: br i1 true, label [[FOO:%.*]], label [[BAR:%.*]]
; CHECK: foo:
; CHECK-NEXT: ret void
; CHECK: bar:
; CHECK-NEXT: ret void
;
%cmp = icmp eq i32 %x, 0
%ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp)
%all = icmp eq i64 %ballot, -1
call void @llvm.assume(i1 %all)
br i1 %cmp, label %foo, label %bar

foo:
ret void

bar:
ret void
}

declare i64 @llvm.amdgcn.ballot.i64(i1)
declare noalias ptr @_Znwm(i64)
declare void @_ZN1AC1Ev(ptr)
declare void @llvm.assume(i1)
Expand Down