[Aldor-l] GCC / ICC differences

[Gabriel Dos Reis]

2008/8/17  <pip88nl at gmail.com>:
> On Sat, Aug 16, 2008 at 10:53:23PM -0500, Gabriel Dos Reis wrote:
>> What optimization flags did you use for both compilers?
>>
>> -- Gaby
>
> For this test: none, but looking at the ICC asm output shows me that ICC
> does actually optimise (it uses the 2-byte xorl %eax, %eax instead of the
> 5-byte mov $0, %eax in return 0, for example). I passed -fwrapv to GCC
> just to be sure, but it didn't blow up there, anyway. I don't know how to
> force ICC not to optimise, yet.

A couple of notes
  (1) If the test is about floating points, then -fwrapv will alost
have no effect,
       or, if it has then it is an orthogonal issue

  (2) If ICC optimizes the code, then it would seem OK to ask GCC to do
       similar optimization (-O1 ?)  This is tricky, because on x86,
you may have
       other side effect kicking i as the fact that fp registers are wider than
       main memory datatype storage; and the results seems to difffer only
       on the sign.


>
> Pippijn
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
>
> iD8DBQFIp/y2Jc+zqGNdDgoRAhf0AKCw9sGTcxdSnQ/CT5NXmz/H1QzIWgCeLFlU
> AG1wDOUgsc5B88L4ZxXCbZ0=
> =YZkh
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> Aldor-l mailing list
> Aldor-l at aldor.org
> http://aldor.org/mailman/listinfo/aldor-l_aldor.org
>
>