[PATCH] D86784: [ARM] Skip combining base updates for vld1x NEON intrinsics

[Sjoerd Meijer via Phabricator via llvm-commits]

SjoerdMeijer added inline comments.


================
Comment at: llvm/test/CodeGen/ARM/pr45824.ll:29
+  %6 = tail call { <8 x i8>, <8 x i8> } @llvm.arm.neon.vld1x2.v8i8.p0i8(i8* %4)
+  %7 = getelementptr inbounds i8, i8* %4, i32 %2
+  %8 = tail call { <8 x i8>, <8 x i8> } @llvm.arm.neon.vld1x2.v8i8.p0i8(i8* %7)
----------------
SjoerdMeijer wrote:
> dmgreen wrote:
> > You can possibly simplify these tests a bit. Something like the tests in mve-vld2-post.ll might work, which return the postinc value.
> Hi Dave, thanks for looking, and you're right that the test can be simplified. I had a quick look, but need more time for this. That's why I will commit this first to unblock other work, and follow up on this to simplify the tests when I am back in the office.
FYI: I've committed a slightly reduced test case in rG288c582fc939. I couldn't really get it down to something smaller than this and still trigger the bug.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D86784/new/

https://reviews.llvm.org/D86784