[PATCH] D60682: [AMDGPU] Fixed +DumpCode

[Scott Linder via Phabricator via llvm-commits]

scott.linder added a comment.

In D60682#1470258 <https://reviews.llvm.org/D60682#1470258>, @nhaehnle wrote:

> Mesa uses +DumpCode as well. Do we have a good solution for generation both ELF and textual output?


Yes, depending on what you mean by "good": just codegen to ASM, then assemble that to the ELF? One codegen gets you both the text and object without the need for the hack.

As far as the review goes, it looks like a hack to `setUseAssemblerInfoForParsing` here, but as long as the code behaves correctly in the cases where it is unset (seems to be just inline assembly?) then it is just a continuation of the +DumpCode hack, so it seems fine to me.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D60682/new/

https://reviews.llvm.org/D60682