Skip to content

[InstCombine] Remove scalable vector extracts to and from the same type #69702

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.

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 2 commits into from
Oct 23, 2023

Conversation

kmclaughlin-arm
Copy link
Contributor

visitCallInst already looks for fixed width vector extracts where number of elements in
the source and destination types are equal. This patch modifies the function to also
identify scalable extracts which can be removed.

visitCallInst already looks for fixed width vector extracts where the
source and destination types are equal. This patch modifies the function
to also identify scalable extracts which can be removed.
@kmclaughlin-arm kmclaughlin-arm added llvm:instcombine Covers the InstCombine, InstSimplify and AggressiveInstCombine passes llvm Umbrella label for LLVM issues labels Oct 20, 2023
@llvmbot
Copy link
Member

llvmbot commented Oct 20, 2023

@llvm/pr-subscribers-llvm-transforms

Author: Kerry McLaughlin (kmclaughlin-arm)

Changes

visitCallInst already looks for fixed width vector extracts where number of elements in
the source and destination types are equal. This patch modifies the function to also
identify scalable extracts which can be removed.


Full diff: https://github.com/llvm/llvm-project/pull/69702.diff

2 Files Affected:

  • (modified) llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp (+10-8)
  • (modified) llvm/test/Transforms/InstCombine/canonicalize-vector-extract.ll (+10)
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index e29fb869686ca0b..00505b5edb58dca 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -2997,24 +2997,26 @@ Instruction *InstCombinerImpl::visitCallInst(CallInst &CI) {
         return replaceOperand(CI, 0, InsertTuple);
     }
 
-    auto *DstTy = dyn_cast<FixedVectorType>(ReturnType);
-    auto *VecTy = dyn_cast<FixedVectorType>(Vec->getType());
+    auto *DstTy = dyn_cast<VectorType>(ReturnType);
+    auto *VecTy = dyn_cast<VectorType>(Vec->getType());
 
-    // Only canonicalize if the destination vector and Vec are fixed
-    // vectors.
     if (DstTy && VecTy) {
-      unsigned DstNumElts = DstTy->getNumElements();
-      unsigned VecNumElts = VecTy->getNumElements();
+      auto DstEltCnt = DstTy->getElementCount();
       unsigned IdxN = cast<ConstantInt>(Idx)->getZExtValue();
 
       // Extracting the entirety of Vec is a nop.
-      if (VecNumElts == DstNumElts) {
+      if (DstEltCnt == VecTy->getElementCount()) {
         replaceInstUsesWith(CI, Vec);
         return eraseInstFromFunction(CI);
       }
 
+      // Only canonicalize to shufflevector if the destination vector and
+      // Vec are fixed vectors.
+      if (dyn_cast<ScalableVectorType>(VecTy) || DstEltCnt.isScalable())
+        break;
+
       SmallVector<int, 8> Mask;
-      for (unsigned i = 0; i != DstNumElts; ++i)
+      for (unsigned i = 0; i != DstEltCnt.getKnownMinValue(); ++i)
         Mask.push_back(IdxN + i);
 
       Value *Shuffle = Builder.CreateShuffleVector(Vec, Mask);
diff --git a/llvm/test/Transforms/InstCombine/canonicalize-vector-extract.ll b/llvm/test/Transforms/InstCombine/canonicalize-vector-extract.ll
index 26af633698f3552..0b7f1ebc040c2a1 100644
--- a/llvm/test/Transforms/InstCombine/canonicalize-vector-extract.ll
+++ b/llvm/test/Transforms/InstCombine/canonicalize-vector-extract.ll
@@ -10,6 +10,7 @@ declare <3 x i32> @llvm.vector.extract.v3i32.v8i32(<8 x i32> %vec, i64 %idx)
 declare <4 x i32> @llvm.vector.extract.v4i32.nxv4i32(<vscale x 4 x i32> %vec, i64 %idx)
 declare <4 x i32> @llvm.vector.extract.v4i32.v8i32(<8 x i32> %vec, i64 %idx)
 declare <8 x i32> @llvm.vector.extract.v8i32.v8i32(<8 x i32> %vec, i64 %idx)
+declare <vscale x 8 x i32> @llvm.vector.extract.nxv8i32.nxv8i32(<vscale x 8 x i32> %vec, i64 %idx)
 
 ; ============================================================================ ;
 ; Trivial cases
@@ -24,6 +25,15 @@ define <8 x i32> @trivial_nop(<8 x i32> %vec) {
   ret <8 x i32> %1
 }
 
+define <vscale x 8 x i32> @trivial_nop_scalable(<vscale x 8 x i32> %vec) {
+; CHECK-LABEL: define <vscale x 8 x i32> @trivial_nop_scalable(
+; CHECK-SAME: <vscale x 8 x i32> [[VEC:%.*]]) {
+; CHECK-NEXT:    ret <vscale x 8 x i32> [[VEC]]
+;
+  %ext = call <vscale x 8 x i32> @llvm.vector.extract.nxv8i32.nxv8i32(<vscale x 8 x i32> %vec, i64 0)
+  ret <vscale x 8 x i32> %ext
+}
+
 ; ============================================================================ ;
 ; Valid canonicalizations
 ; ============================================================================ ;

Copy link
Contributor

@david-arm david-arm left a comment

Choose a reason for hiding this comment

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

LGTM! I suggested a possible tweak you could make before landing the patch, but only if you think it's worth it. :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
llvm:instcombine Covers the InstCombine, InstSimplify and AggressiveInstCombine passes llvm:transforms llvm Umbrella label for LLVM issues
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants