← Back to team overview

yade-dev team mailing list archive

Re: LLVM and Yade

 

On 09/05/2010 11:13 AM, Vaclav Smilauer wrote:
There is a small table, showing the relative calculation speed.

GCC, optimized........................... 1.0
LLVM, optimized........................ 0.91

GCC, optimized........................... 1.32
LLVM, optimized........................ 1.21

Anyway, GCC wins.
Hi Anton,

thanks for this, it is a very useful comparison.

c++ support in LLVM is quite a new thing, so I didn't expect any
miracles. It is a "young" and very dynamic project so I would expect
LLVM to become much much better in near future (gcc as well, but it is
quite conservative in its development, perhaps because it does not have
such a well-defined internal structure); once they add (which might take
a long, since that is definitely not easy) openmp, the results might be
even more interesting.

Vaclav, could you not have a look at "optimize" option in SConstruct?
It seems, there is a bug, it does not optimize a code, although "scons
-h" shows optimize=True.
Is this solved now for? I understand from the other e-mail that it is,
though confirm that to be sure.


Hi, Vaclav!

Difficult to say, what it was. It seems, I have got this problem because of Intel's TurboBoost technology, when the frequency increases for certain conditions.
Will try to check it again and report.



Cheers, Vaclav



_______________________________________________
Mailing list: https://launchpad.net/~yade-dev
Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp





References