Abdelrazak Younes wrote: > Jean-Marc Lasgouttes wrote: >>>>>>> "Abdelrazak" == Abdelrazak Younes <[EMAIL PROTECTED]> writes: >> >> Abdelrazak> The cursor and dociterator pist is a good one I think. I >> Abdelrazak> attach again the test file. At the end of file, appending >> Abdelrazak> characters at the end of the last line is _slow_ but >> Abdelrazak> hitting return and typing in a new line is very quick (as >> Abdelrazak> quick as usual I mean). So the relationship between two >> Abdelrazak> consecutive paragraph is the key to our problem I think. >> >> And this slowness is visible in csrss.exe? > > I am not sure of what you mean... By slowness I mean that characters are > slowly inserted. While inserting the characters one by one, csrss.exe > occupies 50% of the CPU and lyx the rest. > This behaviour is only visible for the paragraph immediately following > the big formula. Insertion of characters in the following paragraph is > rapid again. > >> This is without gcc debug STL, right? > > No but this is with MSVC STL debug which is normally quite slow. You're > right, I should rebuild in release mode and see if things have improved > after all the different micro-optimisations that have already taken place. > There's a good chance that we have solved already most of the slowness > as LyX-svn in debug mode is way faster than the (gcc-optimized) 1.4.3 > for this test document. > > Abdel. > >
Here some numbers, msvc/Release: when crss eats 100 seconds lyx only needs 32 seconds (cpu-time in the tsak manager). If these 100 sec are really wasted by lyxerr, the we waste 75% of the cpu time for nothing. -- Peter Kümmel
