[PATCH v7.1 06/13] ARC: hardware floating point support

[Vineet Gupta]

On 7/2/20 11:55 AM, Adhemerval Zanella via Libc-alpha wrote:
> 
> 
> On 30/06/2020 21:08, Vineet Gupta via Libc-alpha wrote:
>> ---
>>    Changes since v7:
>>      - Disintegrated math-use-builtins.h (deleted) and added
>>        math-use-builtins-{fma,sqrt}.h
> 
> LGTM with some clarification below.
> 
> Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>


>> +#include <features.h>
>> +
>> +#if !defined(__ARC_FPU_SP__) &&  !defined(__ARC_FPU_DP__)
> 
> My guess is __ARC_FPU_SP__ indicates single floating point and
> __ARC_FPU_DP__ a double precision support. 

Right.

> Is there a possibility
> to compiler define just one of them? If so how should we handle it
> on glibc?

Technically yes. The differences would be in gcc codegen. There's no ARC hand
written code which needs to handle that. Ofcourse we only test with
-mcpu=hs38_linux which implies both SP and DP.

-Vineet