[PATCH] D42607: [LoopStrengthReduce, x86] don't add cost for a cmp that will be macro-fused (PR35681)

[Quentin Colombet via Phabricator via llvm-commits]

qcolombet accepted this revision.
qcolombet added a comment.
This revision is now accepted and ready to land.

LGTM

Couple of nits on the test case.



================
Comment at: test/Transforms/LoopStrengthReduce/X86/macro-fuse-cmp.ll:3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown                    | FileCheck %s --check-prefix=BASE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=macrofusion | FileCheck %s --check-prefix=FUSE
+
----------------
Could you also make a IR to IR test with opt -loop-reduce?


================
Comment at: test/Transforms/LoopStrengthReduce/X86/macro-fuse-cmp.ll:57
+  %index.next = add i64 %index, 2
+  %7 = icmp eq i64 %index.next, 65536
+  br i1 %7, label %for.cond.cleanup, label %vector.body
----------------
Could you run instnamer on the test case?


https://reviews.llvm.org/D42607