Re: Results from SPEC2006 FP analysis done at Richard`s request {late July / early August}

[Richard Biener <richard dot guenther at gmail dot com>]

On Fri, Aug 14, 2015 at 3:35 PM, Abe Skolnik <abe_skolnik@yahoo.com> wrote:
> [Alan wrote:]
>
>> Interesting, thanks.  For what kind of architecture are these -
>
> You are welcome.
>
> You raised 2 or 3 good points, I think.
>
> First: the numbers are all from builds on and for the AMD64 ISA, AKA "x86_64".  My apologies for forgetting to mention that vital fact.
>
> Second:  I did not tell the SPEC build system to use any particular "-march" or "-mtune" flags, and AFAIK it [SPEC] did not add any that I didn`t specify.  In other words, those compiler-tuning values were almost certainly at their GCC defaults.
>
> [A question about the preceding: is "-march=native" the default nowadays, at least on GNU/Linux?      AFAIK the default GCC behavior is (still?) to generate code for the most generic form of the target ISA unless an explicit flag overrides this.]
>
> Third: the server in question has Intel silicon for its CPUs.  If an implicit "-march=native" or similar is suspected of having been a factor, then please let me know and I`ll report back on the specifics.  [I am at home right now, so I have no easy way of getting that data right now.]

Ah, so for a meaningful comparison -march=native should be used.
Otherwise we don't get much store if-conversion anyway.

Richard.

>
>> specifically: with/out masked/gathering loads/stores ??
>
>
> TTBOMK, generic AMD64/x86_64 does _not_ have the gathering stuff and the very latest from Intel _does_.
>
> Sorry, but I don`t know about the masked form[s].  If that`s important to know, then please tell me and I will investigate.
>
> Regards,
>
> Abe