[llvm-dev] *** GMX Spamverdacht *** Re: clang 4.0.0: Invalid code for builtin floating point function with -mfloat-abi=hard -ffast-math (ARM)

[Joerg Sonnenberger via llvm-dev]

On Thu, Jul 13, 2017 at 11:55:15AM +0100, Peter Smith via llvm-dev wrote:
> The _Complex multiply and divide helpers are always called with AAPCS,
> regardless of -mfloat-abi=hard which is not correct.

The libgcc behavior here is IMO arbitrarily insane and broken.
Essentially, there are functions in libgcc/compiler_rt that have an ABI
dictated by ARM EABI and there are functions outside that scope. Why
those should have a hard-wired ABI (AAPCS) is beyond me.

Joerg