>>>>> "Bennett" == Bennett Helm <[EMAIL PROTECTED]> writes:
>> Rob, could you try to run LyX under Tiger's Shark.app? The time in >> lengthcombox.C does not seem to make sense to me. Bennett> I did that, and the 65% is for Rob's version. Here's a bit Bennett> more detail: | | | | | | | | | | | | + 72.7% rowpainter.C:303 (LyX) | | | | | | | | | | | | | + 72.6% basic_string.h:909 (LyX) | | | | | | | | | | | | | | + 69.8% scoped_ptr.hpp:93 (LyX) | | | | | | | | | | | | | | | + 68.9% lengthcombo.C:63 (LyX) | | | | | | | | | | | | | | | | + 66.1% lengthcombo.C:63 (LyX) | | | | | | | | | | | | | | | | | + 64.9% lengthcombo.C:63 (LyX) | | | | | | | | | | | | | | | | | | + 0.7% lengthcombo.C:63 (LyX) rowpainter.C:303 reads: pain_.text(int(x_), yo_, str, font); This means this is the place where drawing text is actually done. This seems very reasonable. However, the code in lengthcombo.C does not look reasonable at all (look how it is recursive) since it reads for (int i = 0; i < num; i++) { Therefore I tend to think that drawing is what takes most of the time. I am not sure what kind of optimization we can do, but we'll need to do a lot of it. Bennett> I've got Shark 4.1 and Xcode 1.5; these are the latest Bennett> versions I have access to without installing OS X 10.4. (And Bennett> until VPN issues get worked out on 10.4, I'm not eager to Bennett> dive in....) Very good. JMarc