On 10/22/2015 06:05 AM, Dominik Vogt wrote:
On Tue, Oct 13, 2015 at 05:03:36PM -0400, Vladimir Makarov wrote:
[snip]
I checked my article
ftp://ftp.uvsq.fr/pub/gcc/summit/2004/Fighting%20Register%20Pressure.pdf
and GVN gave mostly 0.2% on eon only. The current environment is
quite different (IRA, LRA) so the results might be different too.
Also as I remember I implemented GVN only for pseudos.
LRA also checks values too but again only for reload and original
pseudos.
Do you still have the branches you've tested back then? I'd
really like to try how this patch affects other targets (big
endian?). Gcc seems to do a better job optimising code for x86 in
some complicated situations, so the extra logic might pay off more
on other targets.
No, sorry. I can not find the code after 12 years since it was done. As
I remember that time we even used cvs and there was no branch for the
code. That time the the development practice was different. Even these
days results of most modern research articles in CS can not be
reproduced. If I did such research today, I'd definitely put these code
somewhere on github with detail instructions how to reproduce the results.