http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49263
--- Comment #9 from Oleg Endo <oleg.e...@t-online.de> 2011-10-10 23:48:17 UTC --- Created attachment 25461 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=25461 CSiBE comparisons (In reply to comment #8) > > Another combine pass to reduce size less than 0.3% on one target > would be not acceptable, I guess. I'm sorry, I forgot to mention that it was just a proof of concept hack of mine, just to see whether it has any chance to work at all. I think it would be better to change/fix the behavior of the combine pass in this regard, so that it tries matching combined patterns without sophisticated transformations. I will try asking on the gcc list about that. > ~10 new patterns would be > overkill for that result, though I'm still expecting that a few > patterns of them were dominant. Yep, even if it turned out to be actually only 8 patterns in total, but still.. I haven't looked at the issue with SH4A but most likely it would add another one or two patterns... so basically ~10 :) > Could you get numbers which pattern > was used in the former option? I think it would be a bit too much checking out each individual pattern. Instead I grouped them into what they are effectively doing. While I was at it, I also added your peephole idea, and a top 10 listing of the individual files.
