1 |
-----BEGIN PGP SIGNED MESSAGE----- |
2 |
Hash: SHA1 |
3 |
|
4 |
On Fri, 25 Aug 2006, Adam Pityszek wrote: |
5 |
|
6 |
> Dear Markus, gentoo-science guys, |
7 |
> |
8 |
> Please find below the reply from Clint to my yesterday's email related to |
9 |
> our work on ATLAS shared libraries in Gentoo. |
10 |
> |
11 |
> Markus, I think we can help with answering the questions (2) and (3). Of |
12 |
> course, volunteers from gentoo-science are welcome as well. |
13 |
> |
14 |
> BR, |
15 |
> /ediap |
16 |
> |
17 |
> (1) Is it true that the extra pointer may still be used if we restore it at |
18 |
> end of assembly routine? |
19 |
> (2) Does throwing the -fpic or other required compiler flag changes change |
20 |
> the best cases (thus necessitating doubling the arch defaults)? |
21 |
> (3) What is the overall performance affect when using .so? |
22 |
> |
23 |
> I've tried to answer (1) by looking at some docs, but never got convinced |
24 |
> either way. I've been meaning to write a resister stress-test to see if |
25 |
> I can make gcc use the reserved register in a function w/o global data. |
26 |
> Perhaps you know? |
27 |
> |
28 |
> You guys could help with (2) & (3) if you like. You could build out-of-box |
29 |
> to .a on whatever machines you can, and then build it to .so using your |
30 |
> gentoo harness, and post some head-to-head timings . . . If, as we suspect, |
31 |
> the difference is essentially zero, that makes .so a lot more attractive . . . |
32 |
> |
33 |
|
34 |
Hi Adam, |
35 |
|
36 |
Thanks for talking to upstream about this and Clint's response |
37 |
sounds encouraging. We could definitely help out with 2) and 3); |
38 |
it would be good to know anyway how well we do with our shared |
39 |
libs. In doing so we should also test the impact of using |
40 |
the 387 floating point unit versus the sse instruction set. |
41 |
According to Clint, the former can give a significant performance |
42 |
gain on some CPU's. If that is the case it might be worth a |
43 |
note in the ebuild to make our users aware of it. |
44 |
|
45 |
We should get a hold of a nice benchmark suite for this purpose; |
46 |
Clint has posted one on this gcc bug |
47 |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27827 |
48 |
which we might be able to use. I'll have a look at it. |
49 |
|
50 |
Best, |
51 |
Markus |
52 |
|
53 |
|
54 |
- -- |
55 |
Markus Dittrich (markusle) |
56 |
Gentoo Linux Developer |
57 |
Scientific applications |
58 |
-----BEGIN PGP SIGNATURE----- |
59 |
Version: GnuPG v1.4.5 (GNU/Linux) |
60 |
|
61 |
iD8DBQFE7vAixlRwCwb7k40RAu4HAJ9qZ9UZr5Nt6rTAC/XoTXezJ+yqswCdE9iF |
62 |
9766l1jX9prxDzgSjnsWPWU= |
63 |
=++uN |
64 |
-----END PGP SIGNATURE----- |
65 |
-- |
66 |
gentoo-science@g.o mailing list |