[PATCH] D46423: [WIP, RISCV] Support .option relax and .option norelax

[Alex Bradbury via Phabricator via llvm-commits]

asb added a comment.

Many thanks for the update.

With the current revision, ForceLinkerReloc is defined but seems unused?

It might be worth spending more time looking at exactly what gas currently does. Looking at the build I've got handy, it seems it will currently unconditionally emit relocations for label differences (i.e. independent of whether -mno-relax is used). I suppose this further reduces cases where you might get incorrect codegen, but it does create a bunch of extra work for the linker which you'd hope -mno-relax would avoid.

It's going to be more hassle, but I think it would be worth evaluating passing through the current STI to the MCAsmBackend, like the series of patches culmininating in https://reviews.llvm.org/D45962.


Repository:
  rL LLVM

https://reviews.llvm.org/D46423