On 27/02/2014 06:13, Guillem Jover wrote: >> This change pessimizes on the common case by adding >> two extra in-loop branches, without actually saving any cost since >> the allocation needs to happen anyway. > > You mean on a loop that is performing a syscall per byte, ok.
I don't think you should speak of Baptiste's code this way. Sure, there's room for improvement. And sure, the read() could use a larger chunk size since the buffer allows this, saving a lot of syscalls. But your statement is akin to "this code performs so bad, it doesn't matter if we make it even worse". This isn't the right way to judge someone else's work, and it isn't the right way to think of problems either. We should strive to solve them (in this case, making keyboards and mice work again!), rather than to make them worse. > Anyway, I'm not the one who's going to merge such changes, so… I expected as much. I think the code is in good shape now. I already took advantage of the useful parts in your review and turned them into improvements in the actual code. I will now wait for someone with the authority to either approve or reject it before taking further action. -- Robert Millan _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
