[PATCH] D31965: [SLP] Enable 64-bit wide vectorization for Cyclone

[Kristof Beyls via Phabricator via llvm-commits]

kristof.beyls added inline comments.


================
Comment at: lib/Target/AArch64/AArch64TargetTransformInfo.h:88-89
+  unsigned getMinVectorRegisterBitWidth() {
+    // FIXME: This should probably be enabled for any Neon subtarget but
+    // currently it was only tuned for Cyclone.
+    return ST->getMinVectorRegisterBitWidth();
----------------
This comment can be removed now?


================
Comment at: test/Transforms/SLPVectorizer/AArch64/64-bit-vector.ll:1-3
+; RUN: opt -S -slp-vectorizer -mtriple=aarch64-apple-ios -mcpu=cyclone < %s | FileCheck %s
+; RUN: opt -S -slp-vectorizer -mtriple=aarch64--linux-gnu -mcpu=kryo < %s | FileCheck --check-prefix=NO_SLP %s
+; RUN: opt -S -slp-vectorizer -mtriple=aarch64-apple-ios -mcpu=cyclone -slp-min-reg-size=128 < %s | FileCheck --check-prefix=NO_SLP %s
----------------
Given that we think this change is good for generic AArch64 code generation, wouldn't it be good to also have a test for -mcpu=generic. Or without an -mcpu specified at all?


https://reviews.llvm.org/D31965