Richard Henderson <richard.hender...@linaro.org> writes:
> On 12/21/18 11:30 AM, Emilio G. Cota wrote:
>> + ua.s = 0x0020000000000001;
>> + ub.s = 0x3ca0000000000000;
>> + uc.s = 0x0020000000000000;
>> + ur.h = fma(ua.h, ub.h, uc.h);
>> + if (ur.s != 0x0020000000000001) {
>
> Forgot your ull's, but otherwise ok.
>
> In email to Alex, I did wonder if we should check for fma hardware (at least
> on
> x86). Without a hardware insn, the libm routine is probably no faster than
> softmmu.
My only worry is we could end up doing a bunch of these processor
id/capability type checks. Is a correct but slow libm FMA going to be
much slower than our FMA? Will users even notice?
--
Alex Bennée
