| 1 |
To explain, keep in mind that optimisation and chost are two different |
| 2 |
things. |
| 3 |
|
| 4 |
i386 is a "lowest common denominator" instruction set that will run on |
| 5 |
most 386 and above x86 processors. i4/5/686 adds few specialised |
| 6 |
instructions and I believe the compiler is able to use them to produce |
| 7 |
faster code in some cases. The downside is the loss of compatibility - |
| 8 |
apparent if you switch processors. Is the system faster - my tests |
| 9 |
(done ages ago now) say yes, but not by much and its highly dependent on |
| 10 |
the actual code/data in use at the time. |
| 11 |
|
| 12 |
Generally, you will get more gain by smarter configuration, better |
| 13 |
software etc. Thats not to say optimised CFLAGS and compiler choices |
| 14 |
wont give a useful speedup, especially when crunching data. It just |
| 15 |
wont turn a 667Mhz P3 into the equivalent 1G P3 - I know I recently |
| 16 |
tried to "get a little more" out of one :) |
| 17 |
|
| 18 |
BillK |
| 19 |
|
| 20 |
On Wed, 2005-11-23 at 21:24 -0400, Robin wrote: |
| 21 |
> > > Wouldn't leaving the CHOST at |
| 22 |
> > > "i386-pc-linux-gnu" build unoptimized binaries? |
| 23 |
> > |
| 24 |
> > No. |
| 25 |
> > |
| 26 |
> > Alexander Skwar |
| 27 |
> |
| 28 |
> Thanks for that. My CHOST flag is set to i386-pc-linux-gnu even though |
| 29 |
> it is not. Just a piece of mind I guess not building unoptimized |
| 30 |
> binaries. |
| 31 |
> |
| 32 |
-- |
| 33 |
gentoo-user@g.o mailing list |
Archives