| 1 |
Ciaran McCreesh wrote: |
| 2 |
> |
| 3 |
> How much of that is memory bound? Of the things that aren't, how many |
| 4 |
> aren't written in assembly anyway? Of those things, what proportion of |
| 5 |
> the runtime is spent in those areas? |
| 6 |
> |
| 7 |
> If you double the speed of something that takes up 2% of the overall |
| 8 |
> execution time, you can't measure the improvement. |
| 9 |
> |
| 10 |
> Or looking at it the other way -- is there any reason to believe that |
| 11 |
> using icc (which can end up being a substantial pain in the arse, given |
| 12 |
> the way it tries to use gcc's c++ headers but doesn't support some of |
| 13 |
> the extensions or quirks that g++ does) will provide a genuine gain |
| 14 |
> for people who aren't already doing clever profile-directed trickery |
| 15 |
> anyway? |
| 16 |
> |
| 17 |
> |
| 18 |
> The problem with -O3 is that function inlining can lead to a |
| 19 |
> substantial cache hit. Unless you're using profile-directed |
| 20 |
> optimisations, which Gentoo doesn't support, it's extremely hit and |
| 21 |
> miss as to whether O3 helps or hurts. |
| 22 |
> |
| 23 |
|
| 24 |
I agree with all of the above. Gentoo is about choice, so if people |
| 25 |
want to make ICC work well more power to them. I agree that it would be |
| 26 |
hard to make it THE ONLY system compiler. For those who do try it I'd |
| 27 |
be really interested in their findings. |
| 28 |
-- |
| 29 |
gentoo-dev@l.g.o mailing list |
Archives