[PATCH] D70068: [SLP] Enhance SLPVectorizer to vectorize vector aggregate

[Alexey Bataev via Phabricator via llvm-commits]

ABataev added inline comments.


================
Comment at: llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:2879-2880
+  if (auto *VT = dyn_cast<VectorType>(EltTy)) {
+    if (VT->isScalable())
+      return 0;
+    EltTy = VT->getElementType();
----------------
Why do we have this check?


================
Comment at: llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:6762
 static bool findBuildAggregate(InsertValueInst *IV,
-                               SmallVectorImpl<Value *> &BuildVectorOpds) {
+                             TargetTransformInfo *TTI,
+                             SmallVectorImpl<Value *> &BuildVectorOpds,
----------------
Not formatted


================
Comment at: llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:6772-6774
+      for (auto OpdIt = TmpBuildVectorOpds.rbegin();
+           OpdIt != TmpBuildVectorOpds.rend(); OpdIt++)
+        BuildVectorOpds.push_back(*OpdIt);
----------------
Just `BuildVectorOpds.append(TmpBuildVectorOpds.rbegin(), TmpBuildVectorOpds.rend());`


================
Comment at: llvm/test/Transforms/SLPVectorizer/X86/pr42022.ll:4
+
+%struct.Vector4 = type { float, float, float, float }
+
----------------
anton-afanasyev wrote:
> RKSimon wrote:
> > Regenerate with update_test_checks.py ?
> Should it really be autogenerated? I don't like test autogeneration for excessiveness and false positives while testing. Here we just checking vector `<4 x float>` is generated. Is autogeneration really standard for tests now?
It is common practice to use auto checks and demonstrate the difference then.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D70068/new/

https://reviews.llvm.org/D70068