| 1 |
On 03/13 06:27, Stroller wrote: |
| 2 |
> |
| 3 |
> > On 13 Mar 2017, at 16:27, tuxic@××××××.de wrote: |
| 4 |
> > On 03/13 08:06, Martin Vaeth wrote: |
| 5 |
> >> You appear to have experimental CFLAGS. Try without these. |
| 6 |
> >> |
| 7 |
> > Could you specify, what parts of that settings are experimental, |
| 8 |
> > so that I will remove the correct ones? |
| 9 |
> |
| 10 |
> Your flags are: |
| 11 |
> |
| 12 |
> CFLAGS=-march=native -O2 -pipe -ffast-math -fopenmp -fprefetch-loop-arrays -ftree-parallelize-loops=12 -funroll-all-loops -fwhole-program -mtune=amdfam10 |
| 13 |
> |
| 14 |
> Google each one of them, see what they do. There's probably a gcc manual that explains them. |
| 15 |
> |
| 16 |
> "funroll-loops" has long been a joke about Gentoo ricing. https://fun.irq.dk/funroll-loops.org/ |
| 17 |
> |
| 18 |
> I would guess you'd take that out and -ftree-parallelize-loops for a start. |
| 19 |
> |
| 20 |
> If you want to use the program, compile it with just `CFLAGS=-march=native -O2 -pipe` - I'm sure that will work. Then you can go back and see what options you can add without crashing the compile. These options often do not actually make things faster. |
| 21 |
> |
| 22 |
> Stroller. |
| 23 |
> |
| 24 |
> |
| 25 |
|
| 26 |
Hi Stroller, |
| 27 |
|
| 28 |
THANKS A LOT ! :) 8) |
| 29 |
That helps me! |
| 30 |
|
| 31 |
Cheers |
| 32 |
Meino |
Archives