On Mon, 22 Oct 2012, Glauber Costa wrote: > It is still unclear to me if the above is really better than > ilog2(size -1) + 1 Hmmm... We could change that if ilog2 is now supported on all platforms and works right. That was not the case a couple of years ago (I believe 2008) when I tried to use ilog. > For that case, gcc seems to generate dec + brs + inc which at some point > will be faster than walking a jump table. At least for dynamically-sized > allocations. The code size is definitely smaller, and this is always > inline... Anyway, this is totally separate. Indeed I would favor that approach but it did not work out for all platforms the last time around. Compiler was getting into issues to do the constant folding too. > The patch also seem to have some churn for the slob for no reason: you > have a patch just to move the kmalloc definitions, would maybe be better > to do it in there to decrease the # of changes in this one, which is > more complicated. Ok. Will look at that. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>