[llvm] [LV] Honor -force-vector-interleave when generating diff checks (PR #225768)

[via llvm-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-transforms

Author: Benjamin Maxwell (MacDue)

<details>
<summary>Changes</summary>



---
Full diff: https://github.com/llvm/llvm-project/pull/225768.diff


2 Files Affected:

- (modified) llvm/lib/Transforms/Vectorize/LoopVectorize.cpp (+1-1) 
- (added) llvm/test/Transforms/LoopVectorize/AArch64/diff-checks-force-interleave-count.ll (+43) 


``````````diff
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index d929af8afbd1d..d494ba2765b08 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -7924,7 +7924,7 @@ bool LoopVectorizePass::processLoop(Loop *L) {
     // Select the interleave count.
     IC = LVP.selectInterleaveCount(*BestPlanPtr, VF.Width, VF.Cost);
 
-    unsigned SelectedIC = std::max(IC, UserIC);
+    unsigned SelectedIC = UserIC > 0 ? UserIC : IC;
     //  Optimistically generate runtime checks if they are needed. Drop them if
     //  they turn out to not be profitable.
     if (VF.Width.isVector() || SelectedIC > 1) {
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/diff-checks-force-interleave-count.ll b/llvm/test/Transforms/LoopVectorize/AArch64/diff-checks-force-interleave-count.ll
new file mode 100644
index 0000000000000..170b2e13a4911
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/diff-checks-force-interleave-count.ll
@@ -0,0 +1,43 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "^vector.ph:" --version 6
+; RUN: opt -passes loop-vectorize -mattr=+sve2 -force-vector-width="vscale x 2" -force-vector-interleave=1 -S < %s | FileCheck %s
+
+target triple = "aarch64-unknown-linux-gnu"
+
+; Tests that diff checks honor `-force-vector-interleave` for the check width, not the cost model.
+
+define void @diff_check_d(ptr %src, ptr %dst, i64 %n) {
+; CHECK-LABEL: define void @diff_check_d(
+; CHECK-SAME: ptr [[SRC:%.*]], ptr [[DST:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[SRC2:%.*]] = ptrtoaddr ptr [[SRC]] to i64
+; CHECK-NEXT:    [[DST1:%.*]] = ptrtoaddr ptr [[DST]] to i64
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], [[SCALAR_PH:label %.*]], label %[[VECTOR_MEMCHECK:.*]]
+; CHECK:       [[VECTOR_MEMCHECK]]:
+; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP3:%.*]] = shl i64 [[TMP2]], 4
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[TMP3]], -1
+; CHECK-NEXT:    [[TMP5:%.*]] = sub i64 [[DST1]], [[SRC2]]
+; CHECK-NEXT:    [[TMP6:%.*]] = sub i64 [[TMP5]], 1
+; CHECK-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP4]]
+; CHECK-NEXT:    br i1 [[DIFF_CHECK]], [[SCALAR_PH]], label %[[EXIT:.*]]
+; CHECK:       [[EXIT]]:
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %src.gep = getelementptr inbounds i64, ptr %src, i64 %iv
+  %value = load i64, ptr %src.gep, align 8
+  %dst.gep = getelementptr inbounds i64, ptr %dst, i64 %iv
+  store i64 %value, ptr %dst.gep, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %done = icmp eq i64 %iv.next, %n
+  br i1 %done, label %exit, label %loop
+
+exit:
+  ret void
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/225768