[InstCombine] Lower flag check pattern to use a bitmask-shift #169557

bababuck · 2025-11-25T19:44:45Z

File Edit Options Buffers Tools Help
The following C-code:

bool or_icmp(int type) {
  return type == 0 || type == 6 || type == 15;
}

Currently lowers to:

define i1 @or_icmp(i32 signext %type) {
entry:
  %cmp = icmp eq i32 %type, 6
  %cmp1 = icmp eq i32 %type, 0
  %or.cond = or i1 %cmp, %cmp1
  %cmp3 = icmp eq i32 %type, 15
  %or.cond1 = or i1 %cmp3, %or.cond
  ret i1 %or.cond1
}

But more optimally lowers to:

define i1 @or_icmp(i32 signext %type) {
entry:
  %srl = lshr i32 32833, %type
  %srl.1 = trunc i32 %srl to i1
  %cmp = icmp ult i32 %type, 64
  %and = and i1 %srl.1, %cmp
  ret i1 %and
}

Alive Proofs for the test cases:
or_icmp_2: https://alive2.llvm.org/ce/z/9W4PZJ

or_icmp_3: https://alive2.llvm.org/ce/z/zrVHSh

or_icmp_7: https://alive2.llvm.org/ce/z/yBoMer

or_icmp_i64: https://alive2.llvm.org/ce/z/Jas5gE

or_icmp_expand: https://alive2.llvm.org/ce/z/waU32g

or_icmp_expand_trunc_type_shr: https://alive2.llvm.org/ce/z/Catc9V

or_icmp_expand_zext_cmp: https://alive2.llvm.org/ce/z/a7-d3B

or_icmp_i128: https://alive2.llvm.org/ce/z/ng6WJR

or_imcp_expand_128: https://alive2.llvm.org/ce/z/vy39kW

The following C-code: bool or_icmp(int type) { return type == 0 || type == 6 || type == 15; } Currently lowers to: define i1 @or_icmp(i32 signext %type) { entry: %cmp = icmp eq i32 %type, 6 %cmp1 = icmp eq i32 %type, 0 %or.cond = or i1 %cmp, %cmp1 %cmp3 = icmp eq i32 %type, 15 %or.cond1 = or i1 %cmp3, %or.cond ret i1 %or.cond1 } But more optimally lowers to: define i1 @or_icmp(i32 signext %type) { entry: %srl = lshr i32 32833, %type %srl.1 = trunc i32 %srl to i1 %cmp = icmp ult i32 %type, 64 %and = and i1 %srl.1, %cmp ret i1 %and }

llvmbot · 2025-11-25T19:45:17Z

@llvm/pr-subscribers-llvm-transforms

@llvm/pr-subscribers-backend-risc-v

Author: Ryan Buchner (bababuck)

Changes

File Edit Options Buffers Tools Help
The following C-code:

bool or_icmp(int type) {
  return type == 0 || type == 6 || type == 15;
}

Currently lowers to:

define i1 @<!-- -->or_icmp(i32 signext %type) {
entry:
  %cmp = icmp eq i32 %type, 6
  %cmp1 = icmp eq i32 %type, 0
  %or.cond = or i1 %cmp, %cmp1
  %cmp3 = icmp eq i32 %type, 15
  %or.cond1 = or i1 %cmp3, %or.cond
  ret i1 %or.cond1
}

But more optimally lowers to:

define i1 @<!-- -->or_icmp(i32 signext %type) {
entry:
  %srl = lshr i32 32833, %type
  %srl.1 = trunc i32 %srl to i1
  %cmp = icmp ult i32 %type, 64
  %and = and i1 %srl.1, %cmp
  ret i1 %and
}

Full diff: https://github.com/llvm/llvm-project/pull/169557.diff

2 Files Affected:

(modified) llvm/lib/Target/RISCV/RISCVISelLowering.cpp (+49)
(added) llvm/test/CodeGen/RISCV/flag_check.ll (+241)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 3b250d7d9ad1f..50076dbb4555e 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -16736,6 +16736,52 @@ static SDValue combineOrAndToBitfieldInsert(SDNode *N, SelectionDAG &DAG,
   return DAG.getNode(RISCVISD::QC_INSB, DL, MVT::i32, Ops);
 }
 
+// or (icmp eq x, imm0), (icmp eq x, imm1) -> czero.eqz (sltui x, 64), (bext x,
+// 1 << imm0 | 1 << imm1) If [imm0, imm1] < 64
+static SDValue combineOrOfImmCmpToBitExtract(SDNode *N, SelectionDAG &DAG,
+                                             const RISCVSubtarget &Subtarget) {
+  using namespace SDPatternMatch;
+
+  auto CollectSetEqImmTree = [](auto &&Self, SmallVector<APInt, 4> &FlagVals,
+                                SDNode *N, SDValue &X) -> bool {
+    APInt Imm;
+    if (X ? sd_match(N, m_OneUse(m_SetCC(m_Specific(X), m_ConstInt(Imm),
+                                         m_SpecificCondCode(ISD::SETEQ))))
+          : sd_match(N, m_OneUse(m_SetCC(m_Value(X), m_ConstInt(Imm),
+                                         m_SpecificCondCode(ISD::SETEQ))))) {
+      FlagVals.push_back(Imm);
+      return true;
+    }
+    SDValue LHS, RHS;
+    if (sd_match(N, m_OneUse(m_Or(m_Value(LHS), m_Value(RHS))))) {
+      return Self(Self, FlagVals, LHS.getNode(), X) &&
+             Self(Self, FlagVals, RHS.getNode(), X);
+    }
+    return false;
+  };
+
+  SmallVector<APInt, 4> FlagVals;
+  SDValue X;
+  if (!CollectSetEqImmTree(CollectSetEqImmTree, FlagVals, N, X))
+    return SDValue();
+
+  unsigned XLen = Subtarget.getXLen();
+  uint64_t BitMask = 0;
+  for (auto &Imm : FlagVals) {
+    if (Imm.uge(XLen))
+      return SDValue();
+    BitMask |= ((uint64_t)1 << Imm.getZExtValue());
+  }
+
+  SDLoc DL(N);
+  EVT VT = N->getValueType(0);
+  SDValue BitExtract =
+      DAG.getNode(ISD::SRL, DL, VT, DAG.getConstant(BitMask, DL, VT), X);
+  SDValue Lt64Check =
+      DAG.getSetCC(DL, VT, X, DAG.getConstant(XLen, DL, VT), ISD::SETULT);
+  return DAG.getNode(ISD::AND, DL, VT, Lt64Check, BitExtract);
+}
+
 static SDValue performORCombine(SDNode *N, TargetLowering::DAGCombinerInfo &DCI,
                                 const RISCVSubtarget &Subtarget) {
   SelectionDAG &DAG = DCI.DAG;
@@ -16748,6 +16794,9 @@ static SDValue performORCombine(SDNode *N, TargetLowering::DAGCombinerInfo &DCI,
     return V;
   if (SDValue V = combineBinOpOfExtractToReduceTree(N, DAG, Subtarget))
     return V;
+  if (DCI.isAfterLegalizeDAG())
+    if (SDValue V = combineOrOfImmCmpToBitExtract(N, DAG, Subtarget))
+      return V;
 
   if (DCI.isAfterLegalizeDAG())
     if (SDValue V = combineDeMorganOfBoolean(N, DAG))
diff --git a/llvm/test/CodeGen/RISCV/flag_check.ll b/llvm/test/CodeGen/RISCV/flag_check.ll
new file mode 100644
index 0000000000000..86049bf53379c
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/flag_check.ll
@@ -0,0 +1,241 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64
+
+define i1 @or_icmp_2(i32 signext %type) {
+; RV32-LABEL: or_icmp_2:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    li a1, 65
+; RV32-NEXT:    srl a1, a1, a0
+; RV32-NEXT:    sltiu a0, a0, 32
+; RV32-NEXT:    and a0, a0, a1
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: or_icmp_2:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    li a1, 65
+; RV64-NEXT:    srl a1, a1, a0
+; RV64-NEXT:    sltiu a0, a0, 64
+; RV64-NEXT:    and a0, a0, a1
+; RV64-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp eq i32 %type, 0
+  %or.cond = or i1 %cmp, %cmp1
+  ret i1 %or.cond
+}
+
+define i1 @or_icmp_3(i32 signext %type) {
+; RV32-LABEL: or_icmp_3:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    lui a1, 8
+; RV32-NEXT:    addi a1, a1, 65
+; RV32-NEXT:    srl a1, a1, a0
+; RV32-NEXT:    sltiu a0, a0, 32
+; RV32-NEXT:    and a0, a0, a1
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: or_icmp_3:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    lui a1, 8
+; RV64-NEXT:    addi a1, a1, 65
+; RV64-NEXT:    srl a1, a1, a0
+; RV64-NEXT:    sltiu a0, a0, 64
+; RV64-NEXT:    and a0, a0, a1
+; RV64-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp eq i32 %type, 0
+  %or.cond = or i1 %cmp, %cmp1
+  %cmp3 = icmp eq i32 %type, 15
+  %or.cond1 = or i1 %cmp3, %or.cond
+  ret i1 %or.cond1
+}
+
+define i1 @or_icmp_4_tree(i32 signext %type) {
+; RV32-LABEL: or_icmp_4_tree:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    lui a1, 1032
+; RV32-NEXT:    addi a1, a1, 65
+; RV32-NEXT:    srl a1, a1, a0
+; RV32-NEXT:    sltiu a0, a0, 32
+; RV32-NEXT:    and a0, a0, a1
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: or_icmp_4_tree:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    lui a1, 1032
+; RV64-NEXT:    addi a1, a1, 65
+; RV64-NEXT:    srl a1, a1, a0
+; RV64-NEXT:    sltiu a0, a0, 64
+; RV64-NEXT:    and a0, a0, a1
+; RV64-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp eq i32 %type, 0
+  %or.cond = or i1 %cmp, %cmp1
+  %cmp2 = icmp eq i32 %type, 15
+  %cmp3 = icmp eq i32 %type, 22
+  %or.cond1 = or i1 %cmp2, %cmp3
+  %or.cond2 = or i1 %or.cond1, %or.cond
+  ret i1 %or.cond2
+}
+
+define i1 @or_icmp_7(i32 signext %type) {
+; RV32-LABEL: or_icmp_7:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    lui a1, 589860
+; RV32-NEXT:    addi a1, a1, 73
+; RV32-NEXT:    srl a1, a1, a0
+; RV32-NEXT:    sltiu a0, a0, 32
+; RV32-NEXT:    and a0, a0, a1
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: or_icmp_7:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    lui a1, 147465
+; RV64-NEXT:    slli a1, a1, 2
+; RV64-NEXT:    addi a1, a1, 73
+; RV64-NEXT:    srl a1, a1, a0
+; RV64-NEXT:    sltiu a0, a0, 64
+; RV64-NEXT:    and a0, a0, a1
+; RV64-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp eq i32 %type, 0
+  %or.cond = or i1 %cmp, %cmp1
+  %cmp2 = icmp eq i32 %type, 17
+  %or.cond1 = or i1 %cmp2, %or.cond
+  %cmp3 = icmp eq i32 %type, 3
+  %or.cond2 = or i1 %cmp3, %or.cond1
+  %cmp4 = icmp eq i32 %type, 31
+  %or.cond3 = or i1 %cmp4, %or.cond2
+  %cmp5 = icmp eq i32 %type, 14
+  %or.cond4 = or i1 %cmp5, %or.cond3
+  %cmp6 = icmp eq i32 %type, 28
+  %or.cond5 = or i1 %cmp6, %or.cond4
+  ret i1 %or.cond5
+}
+
+define i1 @or_icmp_gte_64(i32 signext %type) {
+; CHECK-LABEL: or_icmp_gte_64:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi a1, a0, -6
+; CHECK-NEXT:    addi a0, a0, -64
+; CHECK-NEXT:    seqz a1, a1
+; CHECK-NEXT:    seqz a0, a0
+; CHECK-NEXT:    or a0, a1, a0
+; CHECK-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp eq i32 %type, 64
+  %or.cond = or i1 %cmp, %cmp1
+  ret i1 %or.cond
+}
+
+define i1 @or_icmp_multiple_uses(i32 signext %type) {
+; CHECK-LABEL: or_icmp_multiple_uses:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi a1, a0, -6
+; CHECK-NEXT:    seqz a1, a1
+; CHECK-NEXT:    seqz a0, a0
+; CHECK-NEXT:    or a0, a1, a0
+; CHECK-NEXT:    xor a0, a1, a0
+; CHECK-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp eq i32 %type, 0
+  %or.cond = or i1 %cmp, %cmp1
+  %or.cond1 = xor i1 %cmp, %or.cond
+  ret i1 %or.cond1
+}
+
+
+define i1 @or_icmp_not_eq(i32 signext %type) {
+; CHECK-LABEL: or_icmp_not_eq:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi a1, a0, -6
+; CHECK-NEXT:    seqz a1, a1
+; CHECK-NEXT:    snez a0, a0
+; CHECK-NEXT:    or a0, a1, a0
+; CHECK-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp ugt i32 %type, 0
+  %or.cond = or i1 %cmp, %cmp1
+  ret i1 %or.cond
+}
+
+define i1 @or_icmp_xlen(i32 signext %type) {
+; RV32-LABEL: or_icmp_xlen:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    addi a1, a0, -6
+; RV32-NEXT:    addi a0, a0, -32
+; RV32-NEXT:    seqz a1, a1
+; RV32-NEXT:    seqz a0, a0
+; RV32-NEXT:    or a0, a1, a0
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: or_icmp_xlen:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    li a1, 1
+; RV64-NEXT:    slli a1, a1, 32
+; RV64-NEXT:    addi a1, a1, 64
+; RV64-NEXT:    srl a1, a1, a0
+; RV64-NEXT:    sltiu a0, a0, 64
+; RV64-NEXT:    and a0, a0, a1
+; RV64-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp eq i32 %type, 32
+  %or.cond = or i1 %cmp, %cmp1
+  ret i1 %or.cond
+}
+
+define i1 @or_icmp_i64(i64 signext %type) {
+; RV32-LABEL: or_icmp_i64:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    xori a2, a0, 6
+; RV32-NEXT:    or a3, a0, a1
+; RV32-NEXT:    xori a0, a0, 15
+; RV32-NEXT:    or a2, a2, a1
+; RV32-NEXT:    seqz a3, a3
+; RV32-NEXT:    or a0, a0, a1
+; RV32-NEXT:    seqz a1, a2
+; RV32-NEXT:    or a1, a1, a3
+; RV32-NEXT:    seqz a0, a0
+; RV32-NEXT:    or a0, a0, a1
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: or_icmp_i64:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    lui a1, 8
+; RV64-NEXT:    addi a1, a1, 65
+; RV64-NEXT:    srl a1, a1, a0
+; RV64-NEXT:    sltiu a0, a0, 64
+; RV64-NEXT:    and a0, a0, a1
+; RV64-NEXT:    ret
+entry:
+  %cmp = icmp eq i64 %type, 6
+  %cmp1 = icmp eq i64 %type, 0
+  %or.cond = or i1 %cmp, %cmp1
+  %cmp3 = icmp eq i64 %type, 15
+  %or.cond1 = or i1 %cmp3, %or.cond
+  ret i1 %or.cond1
+}
+
+define i1 @or_icmp_specific(i32 signext %type, i32 signext %type1) {
+; CHECK-LABEL: or_icmp_specific:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi a0, a0, -6
+; CHECK-NEXT:    addi a1, a1, -32
+; CHECK-NEXT:    seqz a0, a0
+; CHECK-NEXT:    seqz a1, a1
+; CHECK-NEXT:    or a0, a0, a1
+; CHECK-NEXT:    ret
+entry:
+  %cmp = icmp eq i32 %type, 6
+  %cmp1 = icmp eq i32 %type1, 32
+  %or.cond = or i1 %cmp, %cmp1
+  ret i1 %or.cond
+}

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

Will move to InstCombine

The following C-code: bool or_icmp(int type) { return type == 0 || type == 6 || type == 15; } Currently lowers to: define i1 @or_icmp(i32 signext %type) { entry: %cmp = icmp eq i32 %type, 6 %cmp1 = icmp eq i32 %type, 0 %or.cond = or i1 %cmp, %cmp1 %cmp3 = icmp eq i32 %type, 15 %or.cond1 = or i1 %cmp3, %or.cond ret i1 %or.cond1 } But more optimally lowers to: define i1 @or_icmp(i32 signext %type) { entry: %srl = lshr i32 32833, %type %srl.1 = trunc i32 %srl to i1 %cmp = icmp ult i32 %type, 64 %and = select i1 %cmd, i1 %srl.1, i1 false ret i1 %and }

github-actions · 2025-11-26T22:51:55Z

✅ With the latest revision this PR passed the undef deprecator.

bababuck · 2025-11-26T22:55:02Z

llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp

+    if (!validImm(LHSAP) || !validImm(RHSAP))
+      return nullptr;
+    LHSAP = LHSAP.zextOrTrunc(XLen);
+    RHSAP = RHSAP.zextOrTrunc(XLen);


I don't know if this is the best way to wrangle various APInt's that may have differing BitWidth's- effectively I ensure they can be represented within XLen and then force them all to have a BitWidth or XLen.

bababuck · 2025-11-26T22:58:36Z

llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp

+    return Imm && (*Imm < XLen);
+  };
+
+  // Match (or (icmp eq X, Imm0), (icmp eq X, Imm1))


The transform with just two options is about a net neutral (at least going to the RISCV backend). However, the benefit comes once we can expand to combining 3 or more comparisons. From a code standpoint, the implementation seemed cleaner to operate on a single or at a time though.

lukel97 · 2025-11-27T04:20:41Z

Does this transform already exist somewhere in LLVM to some degree? Just checking in godbolt Clang trunk already seems to do a very similar transform on the C snippet in the PR description: https://compiler-explorer.com/z/qTbzc9e33

dtcxzyw · 2025-11-27T04:23:10Z

Does this transform already exist somewhere in LLVM to some degree? Just checking in godbolt Clang trunk already seems to do a very similar transform on the C snippet in the PR description: https://compiler-explorer.com/z/qTbzc9e33

See SimplifyCFGOpt::simplifyBranchOnICmpChain and simplifySwitchLookup

bababuck added 2 commits November 25, 2025 11:33

[RISCV] Add new test for RISCV flag check optimization

5d4f0f7

llvmbot added the backend:RISC-V label Nov 25, 2025

bababuck commented Nov 25, 2025

View reviewed changes

llvm/lib/Target/RISCV/RISCVISelLowering.cpp Outdated Show resolved Hide resolved

bababuck commented Nov 25, 2025

View reviewed changes

llvm/lib/Target/RISCV/RISCVISelLowering.cpp Outdated Show resolved Hide resolved

bababuck added 4 commits November 26, 2025 14:42

[RISCV] Revert implementation in RISCV

ccfc7a8

Will move to InstCombine

[InstCombine] Add new tests for flag check optimization

3f57238

bababuck requested a review from nikic as a code owner November 26, 2025 22:50

llvmbot added llvm:instcombine Covers the InstCombine, InstSimplify and AggressiveInstCombine passes llvm:transforms labels Nov 26, 2025

bababuck removed the backend:RISC-V label Nov 26, 2025

bababuck changed the title ~~[RISCV] Lower flag check pattern to use a bitmask-shift~~ [InstCombine] Lower flag check pattern to use a bitmask-shift Nov 26, 2025

bababuck commented Nov 26, 2025

View reviewed changes

Remove undef in comment to pass formatter

2090e8f

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[InstCombine] Lower flag check pattern to use a bitmask-shift #169557

[InstCombine] Lower flag check pattern to use a bitmask-shift #169557

bababuck commented Nov 25, 2025 •

edited

Loading

Uh oh!

llvmbot commented Nov 25, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

github-actions bot commented Nov 26, 2025 •

edited

Loading

Uh oh!

bababuck Nov 26, 2025

Uh oh!

bababuck Nov 26, 2025

Uh oh!

lukel97 commented Nov 27, 2025 •

edited

Loading

Uh oh!

dtcxzyw commented Nov 27, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

[InstCombine] Lower flag check pattern to use a bitmask-shift #169557

Are you sure you want to change the base?

[InstCombine] Lower flag check pattern to use a bitmask-shift #169557

Conversation

bababuck commented Nov 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmbot commented Nov 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

github-actions bot commented Nov 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bababuck Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

bababuck Nov 26, 2025

Choose a reason for hiding this comment

Uh oh!

lukel97 commented Nov 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dtcxzyw commented Nov 27, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

bababuck commented Nov 25, 2025 •

edited

Loading

llvmbot commented Nov 25, 2025 •

edited

Loading

github-actions bot commented Nov 26, 2025 •

edited

Loading

lukel97 commented Nov 27, 2025 •

edited

Loading