| 1 |
ive just finished some updates to flag-o-matic functions so that they should |
| 2 |
be more robust now ... however, a few flag usages in cvs are broken ... note |
| 3 |
that they were broken before this update, i'm just giving people heads up for |
| 4 |
future usage |
| 5 |
|
| 6 |
before, people used to do this: |
| 7 |
replace-flags k6 586 |
| 8 |
replace-flags k6-1 586 |
| 9 |
replace-flags k6-2 586 |
| 10 |
this was because k6 code generation was pretty bad with gcc-3.2.3 and earlier |
| 11 |
(and i think 3.3.x has resolved most of them) |
| 12 |
|
| 13 |
the way replace-flags (and well, all flag-o-matic functions) is to work off of |
| 14 |
'flag atoms' ... that is, they dont match/replace partial flags (if they did, |
| 15 |
this could lead to unforeseen behavior with users who use some really custom |
| 16 |
flags) |
| 17 |
|
| 18 |
so, to help out in general with replacing march/mcpu flags, ive added another |
| 19 |
function, replace-cpu-flags |
| 20 |
replace-cpu-flags 586 k6 k6-1 k6-2 |
| 21 |
this will replace all instances of -march=k6, -march=k6-1, and -march=k6-2 |
| 22 |
with -march=i586 and -mcpu=k6, -mcpu=k6-1, and -march=k6-2 with -mcpu=i586 |
| 23 |
-mike |
Archives