On 10/25/2017 10:28 AM, Joseph Myers wrote:
> On Wed, 25 Oct 2017, vladimir.mezent...@oracle.com wrote:
>
>> +# Disable FMA (floating-point multiply-add) instructions for complex
>> division.
>> +# These instructions can produce different result if two operations
>> executed separately.
>> +LIBGCC2_FFP_CONTRAST_CFLAGS = -ffp-contract=off
>> +LIB2_DIV3_FUNCS = _divdc3 _divhc3 _divsc3
> Without regard to whether the change is appropriate in the first place,
> I'm doubtful of the logic for selecting floating-point modes. I'd expect
> something that applies to all floating-point modes. Even if nothing has
> fused XFmode operations, some architectures (e.g. s390, powerpc) have
> fused operations on TFmode / KFmode.
Hello Joseph and GCC-team,
Different people are giving me different directions.
I am glad to change my fix but please provide clear directions on how
you want to see it.
michael.hud...@linaro.org created PR 59714 and suggested to set
-ffp-contract=off:
> int main(int argc, char** argv)
> {
> __complex double c = 1.0 + 3.0i;
> printf("%g\n", __imag__ (c/c));
> }
> ubuntu@arm64:~$ gcc cplx.c -o cplx
> ubuntu@arm64:~$ ./cplx
> -1.66533e-17
>
> This is because libgcc2.c is compiled in a way that lets the compiler used
> fused multiply add instructions.
> It shouldn't be! > ...
> In general, fma makes it hard to reason about expressions
> like "a*b-c*d" -- which of the multiplications is being done at the
higher precision?
> But I guess arguing to fp-contract to default to off is a war I don't
want to get into fighting.
Thank you,
-Vladimir
