| 1 |
Am 23.11.2011 18:05, schrieb Michael Mol: |
| 2 |
> On Wed, Nov 23, 2011 at 11:55 AM, Stefan G. Weichinger <lists@×××××.at> wrote: |
| 3 |
>> |
| 4 |
>> Ad CFLAGS for i7-2600: |
| 5 |
>> |
| 6 |
>> Is that too much ricer-style? --> |
| 7 |
>> |
| 8 |
>> ### gcc -march=native -E -v - </dev/null 2>&1 | sed -n 's/.* -v - //p' |
| 9 |
>> |
| 10 |
>> CFLAGS="-O2 -D_FORTIFY_SOURCE=2 -march=core2 -mcx16 -msahf -maes |
| 11 |
>> -mpclmul -mpopcnt -mavx --param l1-cache-size=32 --param |
| 12 |
>> l1-cache-line-size=64 --param l2-cache-size=8192 -mtune=generic" |
| 13 |
>> |
| 14 |
>> gcc-4.5.3-r1 ... |
| 15 |
> |
| 16 |
> That's equivalent to, what, "-O2 -D_FORTIFY_SOURCE=2 -march=native" ? |
| 17 |
> I don't see anything to complain about. |
| 18 |
|
| 19 |
Ok, thanks. |
| 20 |
|
| 21 |
compiled gcc-4.6.2 (hey, it only takes ~16min now!): |
| 22 |
|
| 23 |
# /usr/bin/gcc-4.6.2 -march=native -E -v - </dev/null 2>&1 | sed -n |
| 24 |
's/.* -v - //p' |
| 25 |
-D_FORTIFY_SOURCE=2 -march=corei7-avx -mcx16 -msahf -mno-movbe -maes |
| 26 |
-mpclmul -mpopcnt -mno-abm -mno-lwp -mno-fma -mno-fma4 -mno-xop -mno-bmi |
| 27 |
-mno-tbm -mavx -msse4.2 -msse4.1 --param l1-cache-size=32 --param |
| 28 |
l1-cache-line-size=64 --param l2-cache-size=8192 -mtune=corei7-avx |
| 29 |
|
| 30 |
If I am "feeling adventurous" (Neil) I could now try my luck by choosing |
| 31 |
this nice new compiler w/ options. Maybe *after* a nice backup. |
| 32 |
|
| 33 |
> Actually, it's pretty interesting seeing what that processor comes |
| 34 |
> down to for -march=native. I wish there were a database of processors |
| 35 |
> and their decomposed compiler tuning flags for comparison. That would |
| 36 |
> be *very* interesting, from the standpoint of proc shopping and |
| 37 |
> looking at the evolution of CPUs. |
| 38 |
|
| 39 |
Wouldn't a wiki-page do the trick? gentoo-wiki ... |
| 40 |
|
| 41 |
S |
Archives