[AArch64][NFC] Add test as a representative of scalarizing a vector i…

[sushgokh]

…nteger division

The last resort to vectorize a bundle of integer divisions is considered scalarizing it. Currently, the cost estimates for scalarizing a vector division can be considerably overestimated as is the scenario with this motivating test case i.e. vector cost should not deviate much from the scalar cost.

Future patch will try to improve the scalarization cost.

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: Sushant Gokhale (sushgokh)

Changes

…nteger division

The last resort to vectorize a bundle of integer divisions is considered scalarizing it. Currently, the cost estimates for scalarizing a vector division can be considerably overestimated as is the scenario with this motivating test case i.e. vector cost should not deviate much from the scalar cost.

Future patch will try to improve the scalarization cost.

---

Full diff: https://github.com/llvm/llvm-project/pull/114107.diff

1 Files Affected:

• (added) llvm/test/Transforms/SLPVectorizer/AArch64/scalarizing-vector-cost.ll (+23)

diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/scalarizing-vector-cost.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/scalarizing-vector-cost.ll
new file mode 100644
index 00000000000000..12bbf6e043ca91
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/scalarizing-vector-cost.ll
@@ -0,0 +1,23 @@
+; RUN: opt -mtriple=aarch64 -passes=slp-vectorizer -debug-only=SLP -S -disable-output < %s 2>&1 | FileCheck %s
+
+define <4 x i8> @v4i8(<4 x i8> %a, <4 x i8> %b)
+{
+; CHECK: SLP: Found cost = 18 for VF=4
+  %a0 = extractelement <4 x i8> %a, i64 0
+  %a1 = extractelement <4 x i8> %a, i64 1
+  %a2 = extractelement <4 x i8> %a, i64 2
+  %a3 = extractelement <4 x i8> %a, i64 3
+  %b0 = extractelement <4 x i8> %b, i64 0
+  %b1 = extractelement <4 x i8> %b, i64 1
+  %b2 = extractelement <4 x i8> %b, i64 2
+  %b3 = extractelement <4 x i8> %b, i64 3
+  %1 = sdiv i8 %a0, undef
+  %2 = sdiv i8 %a1, 1
+  %3 = sdiv i8 %a2, 2
+  %4 = sdiv i8 %a3, 4
+  %r0 = insertelement <4 x i8> poison, i8 %1, i32 0
+  %r1 = insertelement <4 x i8> %r0, i8 %2, i32 1
+  %r2 = insertelement <4 x i8> %r1, i8 %3, i32 2
+  %r3 = insertelement <4 x i8> %r2, i8 %4, i32 3
+  ret <4 x i8> %r3
+}

[alexey-bataev]

Better to avoid checks for debug outputs, use -pass-remarks-output= instead

[sushgokh]

ping

[sushgokh]

ping

[davemgreen]

This looks OK. LGTM

Feel free to put up costmodel changes too if you have them already, even if they include the tests again. It can be good to separate the tests out but don't let the code review block other patches. If nothing else there might be other tests that are useful to add or change that becomes clearer with the cost model is changed.

[sushgokh]

Thanks @davemgreen for the approval and your opinion.