[PATCH] D90113: [DAGCombiner] Fold BinOp into Select containing identity constant

Layton Kifer via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Oct 28 11:06:11 PDT 2020


laytonio added inline comments.


================
Comment at: llvm/test/CodeGen/X86/binop-identity.ll:2
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=x86_64-pc-linux -x86-cmov-converter=true -verify-machineinstrs < %s | FileCheck %s
+
----------------
RKSimon wrote:
> Maybe pre-commit these tests to trunk and rebase this patch to show the delta?
> 
> I'm not sure you need -verify-machineinstrs
> 
> We're going to need a binop-identity-vector.ll set of tests as well.
> We're going to need a binop-identity-vector.ll set of tests as well.

What vector widths do you think would be best for testing here? Should I do a left and right for each binop at say <4 x i32>, <8 x i32>, and <16 x i32> with SSE, AVX, and AVX512, or is that overkill?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D90113/new/

https://reviews.llvm.org/D90113



More information about the llvm-commits mailing list