[PATCH] Set "thumbv6k" CPU subtype to ARM_V6

Vedant Kumar via llvm-commits llvm-commits at lists.llvm.org
Wed Aug 12 09:06:17 PDT 2015


>> I think we default to CPU_SUBTYPE_ARM_V7K when we don't know the specific ARM architecture name passed in, which seems sane.
> 
> This looks really odd. Then again, this is Darwin specific, so the
> rules are different.

Sorry, Tim's right, I meant to type CPU_SUBTYPE_ARM_V7.


> I'd be surprised if this is the only case that gets it wrong, and as
> fix goes, this is pretty vague.

It looks like MachO is the only file format that cares about the ARM CPU Subtype.

The llvm-readobj output for ELF and COFF files don't report this information (at least, I couldn't find a way to make it do that).

By vague, do you mean that the patch should not be MachO-specific? There don't appear to be any issues with compiling with -mtriple=thumbv6k-windows-gnu or -mtriple=thumbv6k-linux-gnu..


> O Linux we default to ARMv4T for soft-float and ARMv6Z for hard-float.
> We also don't have "thumbv*" as we tend to normalise to "armv*" so not
> to get that wrong.

Hm. This would require a clang driver change. Would it be better if I sent a patch for this to cfe-commits?


> I'd expect a more thorough fix, and re-using a test
> that already exists for that.

Could you clarify what's missing in the patch?

best
vedant


More information about the llvm-commits mailing list