This is aligned with GCC's behavior.
Also, alias -mno-fp-ret-in-387 to -mno-x87, by which we can fix pr51498.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
clang/test/Driver/x86-target-features.c | ||
---|---|---|
5 | If you split this test and the above change to clang/include/clang/Driver/Options.td into a separate patch on Phab, I'd be happy to accept that! Otherwise someone more familiar with x86 will have to take a look at the below changes. |
clang/test/Driver/x86-target-features.c | ||
---|---|---|
5 | This is a simple alias. Spliting it doesn't solve the initial problem. I think it's better to commit it together with the backend change. |
Just a comment on the ECX bit, otherwise this patch looks fine to me. Someone with more x86 knowledge may want to take a look though. Thanks for the patch!
llvm/lib/Target/X86/X86CallingConv.td | ||
---|---|---|
283 | Is there a test case that exercises the assignment to %ecx? |
llvm/lib/Target/X86/X86CallingConv.td | ||
---|---|---|
283 | Add one test case for it, though I don't believe we can really generate it from frontend. |
llvm/test/CodeGen/X86/no-ret-in-x87-reg.ll | ||
---|---|---|
147–150 ↗ | (On Diff #387855) | :( |
If you split this test and the above change to clang/include/clang/Driver/Options.td into a separate patch on Phab, I'd be happy to accept that!
Otherwise someone more familiar with x86 will have to take a look at the below changes.