[RISCV][VLOPT] Skip EMUL if it is unknown before entering EMULAndEEWAreEqual

[BeMg]

Fix #139288

[llvmbot]

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

Author: Piyou Chen (BeMg)

Changes

Fix #139288

---

Full diff: https://github.com/llvm/llvm-project/pull/139670.diff

2 Files Affected:

• (modified) llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp (+5)
• (added) llvm/test/CodeGen/RISCV/rvv/139288-VLOPT-crash.ll (+22)

diff --git a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
index 9ed2ba274bc53..88876201ae21b 100644
--- a/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
+++ b/llvm/lib/Target/RISCV/RISCVVLOptimizer.cpp
@@ -1383,6 +1383,11 @@ RISCVVLOptimizer::checkUsers(const MachineInstr &MI) const {
     // If the operand is used as a scalar operand, then the EEW must be
     // compatible. Otherwise, the EMUL *and* EEW must be compatible.
     bool IsVectorOpUsedAsScalarOp = isVectorOpUsedAsScalarOp(UserOp);
+
+    if (!IsVectorOpUsedAsScalarOp &&
+        (!ConsumerInfo->EMUL || !ProducerInfo->EMUL))
+      return std::nullopt;
+
     if ((IsVectorOpUsedAsScalarOp &&
          !OperandInfo::EEWAreEqual(*ConsumerInfo, *ProducerInfo)) ||
         (!IsVectorOpUsedAsScalarOp &&
diff --git a/llvm/test/CodeGen/RISCV/rvv/139288-VLOPT-crash.ll b/llvm/test/CodeGen/RISCV/rvv/139288-VLOPT-crash.ll
new file mode 100644
index 0000000000000..1e2f0173505c3
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/139288-VLOPT-crash.ll
@@ -0,0 +1,22 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc -mtriple=riscv64 -mattr=+v < %s | FileCheck %s
+
+define i32 @pps_is_equal(<vscale x 16 x i1> %0, <vscale x 16 x i1> %1) #0 {
+; CHECK-LABEL: pps_is_equal:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetivli zero, 1, e32, m8, ta, ma
+; CHECK-NEXT:    vmclr.m v8
+; CHECK-NEXT:    vmv.s.x v16, zero
+; CHECK-NEXT:    vmor.mm v0, v8, v8
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vredmax.vs v8, v8, v16, v0.t
+; CHECK-NEXT:    vmv.x.s a0, v8
+; CHECK-NEXT:    ret
+entry:
+    %2 = tail call <vscale x 16 x i1> @llvm.vp.or.nxv16i1(<vscale x 16 x i1> zeroinitializer, <vscale x 16 x i1> zeroinitializer, <vscale x 16 x i1> %0, i32 1)
+    %3 = tail call i32 @llvm.vp.reduce.smax.nxv16i32(i32 0, <vscale x 16 x i32> zeroinitializer, <vscale x 16 x i1> %2, i32 1)
+    ret i32 %3
+}
+
+declare <vscale x 16 x i1> @llvm.vp.or.nxv16i1(<vscale x 16 x i1>, <vscale x 16 x i1>, <vscale x 16 x i1>, i32) #1
+declare i32 @llvm.vp.reduce.smax.nxv16i32(i32, <vscale x 16 x i32>, <vscale x 16 x i1>, i32) #1

[lukel97]

Would it be better if we just handle the nullopt EMUL case in EMULAndEEWAreEqual? Does it work if we just replace it with

  static bool EMULAndEEWAreEqual(const OperandInfo &A, const OperandInfo &B) {
    return A.Log2EEW == B.Log2EEW && A.EMUL == B.EMUL;
  }

[lukel97]

Nit, can you move this into vl-opt.ll and replace the VP intrinsics with RISC-V intrinsics? Just in case the VP intrinsics lowering ever changes

[BeMg]

I update the testcase inside the vl-opt.mir.

[BeMg]

Would it be better if we just handle the nullopt EMUL case in EMULAndEEWAreEqual? Does it work if we just replace it with

  static bool EMULAndEEWAreEqual(const OperandInfo &A, const OperandInfo &B) {
    return A.Log2EEW == B.Log2EEW && A.EMUL == B.EMUL;
  }

Yes, It work.