| 1 |
Boyd Stephen Smith Jr. wrote: |
| 2 |
> On Wednesday 27 September 2006 11:11, "Hemmann, Volker Armin" |
| 3 |
> <volker.armin.hemmann@××××××××××××.de> wrote about 'Re: [gentoo-amd64] |
| 4 |
> First Impressions': |
| 5 |
> |
| 6 |
>> -O3 don't do it. O2 is much, much safer and not really slower. It will |
| 7 |
>> prevent a lot of breakage. |
| 8 |
>> |
| 9 |
> |
| 10 |
> Bah! |
| 11 |
> |
| 12 |
> -O3 breaking just doesn't happen anymore. I'm been running it for years. |
| 13 |
> In fact -O3 has never generated invalid code [1] on x86 or amd64 |
| 14 |
> platforms. It did for a period of time on ppc, but I think that was like |
| 15 |
> 10 years ago. |
| 16 |
> |
| 17 |
> Also, some applications break with just -O2. I count 8 ebuilds that |
| 18 |
> replace this cflag specifically, there could be others that filter it or |
| 19 |
> filter/replace optimization options entirely. My methodology is available |
| 20 |
> on request. |
| 21 |
> |
| 22 |
> Now, it is an open question whether -O3 is significantly faster than -O2 |
| 23 |
> and depending on a number of factors it may actually be slower. -O3 will |
| 24 |
> enlarge code to make execution paths shorter, but a cache miss is probably |
| 25 |
> goes to blow away any advantage gained. Conversely, -Os will make |
| 26 |
> execution paths longer to shrink code, and a cache hits may make up for |
| 27 |
> the longer path. [2] |
| 28 |
> |
| 29 |
> So, which is best depends on the program and on the CPU used (not only |
| 30 |
> instruction set, but also cache sizes). Therefore, making a sweeping |
| 31 |
> recommendations is not recommended, but that doesn't stop anyone. My |
| 32 |
> recommendation is -O3. :) |
| 33 |
> |
| 34 |
> Naturally, Gentoo wins over binary distros on this point, since you can use |
| 35 |
> CFLAGS that are appropriate for your specific processor (including cflags |
| 36 |
> which control the optimization tradeoff parameters) as well as tuning for |
| 37 |
> the specific application. |
| 38 |
> |
| 39 |
> |
| 40 |
I've always used "-O3". The only exception is busybox, when I compile it |
| 41 |
for initrd/initramfs. In those cases I prefer "-Os". I've never seen a |
| 42 |
package which explodes with "-O3" and compiles with "-O2". That's the |
| 43 |
reason I hate to find that "replace-flags -O? -O2" stuff in the ebuilds. |
| 44 |
I think I've read somewhere in *.gentoo.org that It should be reported |
| 45 |
if a "testing" version works fine so it could get in the stable tree |
| 46 |
faster. Should this "replace-flags" stuff be reported when the package |
| 47 |
compiles successfully with "-O3"? |
| 48 |
|
| 49 |
|
| 50 |
-- |
| 51 |
Best regards, |
| 52 |
Daniel |
| 53 |
|
| 54 |
|
| 55 |
-- |
| 56 |
gentoo-amd64@g.o mailing list |
Archives