[PATCH] D45995: [DAGCombiner] Set the right SDLoc on a newly-created zextload

Vedant Kumar via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Apr 24 11:40:34 PDT 2018


vsk planned changes to this revision.
vsk added a comment.

Thanks @niravd and Adrian for your feedback. I'll take another shot at this with an eye towards fixing similar location bugs.



================
Comment at: test/CodeGen/X86/avg.ll:296
 ; AVX1-NEXT:    vpmovzxbd {{.*#+}} xmm2 = xmm2[0],zero,zero,zero,xmm2[1],zero,zero,zero,xmm2[2],zero,zero,zero,xmm2[3],zero,zero,zero
-; AVX1-NEXT:    vmovdqa %xmm2, -{{[0-9]+}}(%rsp) # 16-byte Spill
+; AVX1-NEXT:    vmovdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-NEXT:    vpmovzxbd {{.*#+}} xmm2 = xmm5[0],zero,zero,zero,xmm5[1],zero,zero,zero,xmm5[2],zero,zero,zero,xmm5[3],zero,zero,zero
----------------
niravd wrote:
> Much of this test looks like if you ran the update llc test script before this patch they would be generalized and be equivalent. Since you're making revisions, could you land that before updating this patch?
I'll do that and rebase.


================
Comment at: test/CodeGen/X86/win-smallparams.ll:60
 ; WIN32-LABEL: _manyargs:
-; WIN32-DAG: movsbl 4(%esp),
-; WIN32-DAG: movswl 8(%esp),
-; WIN32-DAG: movzbl 12(%esp),
-; WIN32-DAG: movzwl 16(%esp),
-; WIN32-DAG: movzbl 20(%esp),
-; WIN32-DAG: movzwl 24(%esp),
+; WIN32: pushl %ebx
+; WIN32: pushl %edi
----------------
niravd wrote:
> Were these pushes here before this patch?
No, the pushes were not here previously. I was surprised by this codegen changed and meant to follow-up with @rnk about it. It wasn't clear to me whether this indicates some bug related to IROrder. Should I file a PR?


https://reviews.llvm.org/D45995





More information about the llvm-commits mailing list