[llvm] [RISCV] Mark More Relocs as Relaxable (PR #151422)

Sam Elliott via llvm-commits llvm-commits at lists.llvm.org
Fri Aug 1 23:54:06 PDT 2025


lenary wrote:

> I would prevent JAL from being assembler-relaxed to QC_E_JAL, resolving the issue. There is no interference for users who don't use the extension.
> 
> Instead, the compiler should emit qc.e.jal when the displacement might exceed JAL's range, and the linker should handles QC_E_JAL->JAL accordingly.

I don't want to side-track this patch in discussions about Xqci, which is why this patch does not reference QC.E.JAL or Xqci at all. We can come back to those issues on relevant patches/issues, there are enough of them floating around.

The psABI specifically notes relaxations involving the relocations (or their respective fixups) updated in this patch.

Yes, this will expose a layout problem in the backend (#150096). That's part of the reason I want to get this landed now, so we have time to fix it before a release. This patch is not against the 21.x branch and I do not intend to backport it.

https://github.com/llvm/llvm-project/pull/151422


More information about the llvm-commits mailing list