On Sun, Feb 15, 2009 at 5:42 PM, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote: > On Sun, Feb 15, 2009 at 04:33:16PM -0800, Christopher Li wrote: >> All applied and pushed. I update some of the validations error because >> of the tab width change. > > BTW, I'm not sure that this messing with tabstops is a good idea. > Note that tokenizer is _still_ the hottest part of the entire > thing, so we need to be damn careful around it. You are getting the > slow path of nextchar() on \t now, which can add up to a lot of > extra overhead... How about the attached patch? I move white space handling out side of the nextchar_slow(). Let nextchar_slow() only handle two case: possible EOF and '\\'. Chris
Attachment:
hot-nextchar
Description: Binary data
