[llvm] [DAGCombine] Fix an incorrect folding of extract_subvector (PR #153709)

Min-Yih Hsu via llvm-commits llvm-commits at lists.llvm.org
Thu Aug 14 15:55:13 PDT 2025


https://github.com/mshockwave created https://github.com/llvm/llvm-project/pull/153709

Reported from https://github.com/llvm/llvm-project/pull/153393#issuecomment-3189898813

During DAGCombine, an intermediate extract_subvector sequence was generated:
```
  t8: v9i16 = extract_subvector t3, Constant:i64<9>
t24: v8i16 = extract_subvector t8, Constant:i64<0>
```
And one of the DAGCombine rule which turns `(extract_subvector (extract_subvector X, C), 0)` into `(extract_subvector X, C)` kicked in and turn that into `v8i16 = extract_subvector t3, Constant:i64<9>`. But it forgot to check if the extracted index is a multiple of the minimum vector length of the result type, hence the crash.

This patch fixes this by adding an additional check.

>From 2d8d901679b45fec39d0ac2c814e9fc3b49c4a72 Mon Sep 17 00:00:00 2001
From: Min-Yih Hsu <min.hsu at sifive.com>
Date: Thu, 14 Aug 2025 15:45:01 -0700
Subject: [PATCH] [DAGCombine] Fix an incorrect folding of extract_subvector

---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp |  5 ++-
 .../incorrect-extract-subvector-combine.ll    | 36 +++++++++++++++++++
 2 files changed, 40 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/CodeGen/RISCV/rvv/incorrect-extract-subvector-combine.ll

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 17703f58f2824..feb8f8556067b 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -26020,7 +26020,10 @@ SDValue DAGCombiner::visitEXTRACT_SUBVECTOR(SDNode *N) {
   if (ExtIdx == 0 && V.getOpcode() == ISD::EXTRACT_SUBVECTOR && V.hasOneUse()) {
     if (TLI.isExtractSubvectorCheap(NVT, V.getOperand(0).getValueType(),
                                     V.getConstantOperandVal(1)) &&
-        TLI.isOperationLegalOrCustom(ISD::EXTRACT_SUBVECTOR, NVT)) {
+        TLI.isOperationLegalOrCustom(ISD::EXTRACT_SUBVECTOR, NVT) &&
+        // The index has to be a multiple of the new result type's known minimum
+        // vector length.
+        V.getConstantOperandVal(1) % NVT.getVectorMinNumElements() == 0) {
       return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, NVT, V.getOperand(0),
                          V.getOperand(1));
     }
diff --git a/llvm/test/CodeGen/RISCV/rvv/incorrect-extract-subvector-combine.ll b/llvm/test/CodeGen/RISCV/rvv/incorrect-extract-subvector-combine.ll
new file mode 100644
index 0000000000000..6a0c03f339717
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/incorrect-extract-subvector-combine.ll
@@ -0,0 +1,36 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=riscv64 -mattr='+zve64f,+zvl512b' < %s | FileCheck %s
+
+; Previously, an incorrect (extract_subvector (extract_subvector X, C), 0) DAG combine crashed
+; this snippet.
+
+define <8 x i16> @gsm_encode() {
+; CHECK-LABEL: gsm_encode:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetivli zero, 19, e16, m1, ta, ma
+; CHECK-NEXT:    vle16.v v8, (zero)
+; CHECK-NEXT:    vslidedown.vi v9, v8, 12
+; CHECK-NEXT:    vmv.x.s a0, v9
+; CHECK-NEXT:    vsetivli zero, 8, e16, mf4, ta, ma
+; CHECK-NEXT:    vmv.v.i v9, -1
+; CHECK-NEXT:    vsetivli zero, 1, e16, m1, ta, ma
+; CHECK-NEXT:    vslidedown.vi v8, v8, 9
+; CHECK-NEXT:    vmv.x.s a1, v8
+; CHECK-NEXT:    vsetivli zero, 8, e16, mf4, ta, ma
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vslide1down.vx v9, v9, zero
+; CHECK-NEXT:    vslide1down.vx v8, v8, zero
+; CHECK-NEXT:    vslide1down.vx v8, v8, zero
+; CHECK-NEXT:    vslide1down.vx v8, v8, zero
+; CHECK-NEXT:    vslide1down.vx v8, v8, zero
+; CHECK-NEXT:    vslide1down.vx v8, v8, a1
+; CHECK-NEXT:    vslide1down.vx v8, v8, a0
+; CHECK-NEXT:    vslidedown.vi v8, v8, 1
+; CHECK-NEXT:    vand.vv v8, v8, v9
+; CHECK-NEXT:    ret
+entry:
+  %0 = load <19 x i16>, ptr null, align 2
+  %1 = shufflevector <19 x i16> zeroinitializer, <19 x i16> %0, <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 28, i32 31, i32 poison, i32 poison>
+  %2 = shufflevector <9 x i16> %1, <9 x i16> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 15>
+  ret <8 x i16> %2
+}



More information about the llvm-commits mailing list