| 1 |
On Wednesday 08 September 2004 12:15, Klavs Klavsen wrote: |
| 2 |
> Hi guys, |
| 3 |
> |
| 4 |
> Just read an interesting article about Xeon vs. Opteron from anandtech |
| 5 |
> - where they really show how much difference compile optimizations (or |
| 6 |
> not) does - and how it differs for different programs for different |
| 7 |
> processors. |
| 8 |
> |
| 9 |
> http://www.anandtech.com/linux/showdoc.aspx?i=2163&p=1 |
| 10 |
> |
| 11 |
> To me this clearly shows, that if Gentoo wants the best performance - |
| 12 |
> we can't use a "one cflags fits them all" approach. I do know that if a |
| 13 |
> program breaks, those CFLAGS are pulled out in the individual ebuild, |
| 14 |
> but this is not due to poor performance. |
| 15 |
> |
| 16 |
> IMHO the only way for Gentoo to prove its true potential - is to |
| 17 |
> somehow build an array of compile options, with CPU's on X, programs on |
| 18 |
> Y and GCC-version on Z. Getting the numbers for each CPU, will ofcourse |
| 19 |
> require writing tests, for each program - but IMHO this can be done, if |
| 20 |
> we do it one at a time. |
| 21 |
> |
| 22 |
> I would suggest these tests be included like the gentoo-stats program, |
| 23 |
> as something the individual Gentooist can choose to run after each |
| 24 |
> compile - which would give him the optimal performance (and recompile X |
| 25 |
> number of times to test different flags out) on his |
| 26 |
> CPU/program/GCCversion combination, and at the same time, send the |
| 27 |
> result to a Gentoo database. |
| 28 |
> |
| 29 |
> I know I would definetely have the patience to let it test and test |
| 30 |
> again, if it meant more performance for me Smile |
| 31 |
> |
| 32 |
> The end result should be, that Gentoo automagically selects the optimal |
| 33 |
> CFLAGS (in performance and stability - perhaps with some optimizations |
| 34 |
> flagged as "unstable" so people can select "optimize for performance" |
| 35 |
> vs. "optimize for stability") depending on the X, Y and Z from above. |
| 36 |
> |
| 37 |
> I would very much like to be one of the guys that gets the ball |
| 38 |
> rolling, but as I'm not a Gentoo Dev - We (or just I) need to agree |
| 39 |
> with the Gentoo Dev's on how this could best be done. |
| 40 |
> |
| 41 |
> What do you think? am I crazy? It seems to me that the anandtech tests |
| 42 |
> shows that it is more than just a 1% or 2% difference, with the right |
| 43 |
> CFLAGS - and that the right CFLAGS for one program, can be the worst |
| 44 |
> for another on same CPU/GCC combination. |
| 45 |
|
| 46 |
To do this for programs, one would need to have a realistic suite of |
| 47 |
"tests" that simulate the real world use of the application. Of course |
| 48 |
that also allows -fprofile_arcs to be used. |
| 49 |
|
| 50 |
Paul |
| 51 |
|
| 52 |
-- |
| 53 |
Paul de Vrieze |
| 54 |
Gentoo Developer |
| 55 |
Mail: pauldv@g.o |
| 56 |
Homepage: http://www.devrieze.net |
Archives