| 1 |
On Sat, 3 May 2008 14:17:39 +0100 |
| 2 |
Uwe Thiem wrote: |
| 3 |
|
| 4 |
> On Saturday 03 May 2008, David Relson wrote: |
| 5 |
> > On Fri, 2 May 2008 07:10:02 -0400 |
| 6 |
> > David Relson wrote: |
| 7 |
> > |
| 8 |
> > ...[snip]... |
| 9 |
> > |
| 10 |
> > > As part of identifying the capabilities and files of your |
| 11 |
> > > operating system (distro) ./configure creates a lot of small |
| 12 |
> > > programs and compiles them. I can see how caching compilation |
| 13 |
> > > info would help with this. |
| 14 |
> > |
| 15 |
> > I ran a quick test with ccache and ./configure. Using bogofilter's |
| 16 |
> > configure script (because it was available): |
| 17 |
> > |
| 18 |
> > before installing ccache: |
| 19 |
> > |
| 20 |
> > 3 runs of ./configure averaged 6.07s |
| 21 |
> > |
| 22 |
> > ... installed ccache ... |
| 23 |
> > |
| 24 |
> > next run took 6.5 sec (slower than before) |
| 25 |
> > next 3 runs averaged 4.96s (18.4% better than before) |
| 26 |
> |
| 27 |
> All your configure runs were for the same package, right? |
| 28 |
|
| 29 |
Correct. |
| 30 |
|
| 31 |
> > conclusion: ccache _does_ help ./configure |
| 32 |
> |
| 33 |
> It does not for different packages, AFAIK. And that was the question |
| 34 |
> of the OP. |
| 35 |
|
| 36 |
There _should_ be some effect as there are many operations common |
| 37 |
amongst configure scripts, for example determining the name of |
| 38 |
the header file defining malloc or determining whether to use bzero or |
| 39 |
memset, etc, etc. As with any form of caching the rate of cache hits |
| 40 |
will vary depending on load and cache size. |
| 41 |
k |
| 42 |
-- |
| 43 |
gentoo-user@l.g.o mailing list |
Archives