[llvm] r331755 - [AArch64] Disallow vector operand if FPR128 Q register is required.

[Sander De Smalen via llvm-commits]

Hi Amara,

Thanks for pointing out. The issue with 'invalid predicate register'  when using -mattr=+sve should be fixed by https://reviews.llvm.org/D46688.

Cheers,

Sander


On 08/05/2018, 18:03, "aemerson at apple.com on behalf of Amara Emerson" <aemerson at apple.com> wrote:    
    
    Hi Sander,
    
    One thing I noticed while testing this manually was that if you specify -mattr=+sve with llvm-mc, the error message you get is "error: invalid predicate register” rather than "error: invalid operand for instruction” when using just neon. This isn’t a big issue as SVE isn’t enabled by default but is a diagnostics issue for SVE users.
    
    Cheers,
    Amara