2008/11/4 Tony Mechelynck <[EMAIL PROTECTED]>: > > On 04/11/08 22:25, Matt Wozniski wrote: >> On Tue, Nov 4, 2008 at 3:48 PM, Tony Mechelynck wrote: >>> On 04/11/08 21:25, Kazuo Teramoto wrote: >>>> On Tue, Nov 4, 2008 at 5:07 PM, Dominique Pelle >>>> <[EMAIL PROTECTED]> wrote: >>>>> Those are strange results. You have a faster processor (2.16 Ghz >>>>> core2 vs 1.73Ghz core2) and yet redraw in gvim is about 52 times >>>>> slower than on my laptop (6min 41.49s vs 7.715s)!? >>>> strangeness++; >>>> >>>> *urxvt with screen*: >>>> vim -u NONE -U NONE -S test-redraw-speed.vim 1.80s user 0.10s system >>>> 13% cpu *13.995* total >>>> >>>> urxvt without screen: >>>> vim -u NONE -U NONE -S test-redraw-speed.vim 1.80s user 0.11s system >>>> 3% cpu 1:03.07 total >>>> >>>> xterm with screen: >>>> vim -u NONE -U NONE -S test-redraw-speed.vim 2.04s user 0.15s system >>>> 1% cpu 1:53.05 total >>>> >>>> xterm without screen: >>>> vim -u NONE -U NONE -S test-redraw-speed.vim 2.12s user 0.12s system >>>> 2% cpu 1:36.51 total >>>> >>>> linux console(uvesafb) with screen: >>>> vim -u NONE -U NONE -S test-redraw-speed.vim 2.21s user 0.135s >>>> system 3% cpu 59.707 total >>>> >>>> *linux console(uvesafb) without screen*: >>>> vim -u NONE -U NONE -S test-redraw-speed.vim 2.15s user *51.62s* >>>> system *94%* cpu 56.711 total >>>> >>>> My vim binary is compiled with +X11. Same fonts and geometry on urxvt >>>> and xterm. Processor is a AMD Turion 1.8GHZ (single core). I have a >>>> binary compiled without X11 and the results are the same. On every >>>> test inside X the X cpu usage go to 90%. My linux distro is Arch Linux >>>> with vim 7.2.25. >>>> >>>> I repeated the test many many times to the same results. >>>> >>>> Looks like urxvt+screen make a pact to work together... >>> Since the Linux console has no connection to the X server, in that >>> terminal you can speed up Vim startup considerably by using the -X >>> command-line switch. >>> >>> Even in Console mode, Vim compiled with +x11 will try to connect to the >>> X server at startup because it still uses X for the +clipboard and/or >>> +clientserver features, to save and restore the console window title, >>> its contents (+xterm_save), and to "save itself" at X-windows closedown >>> (+xsmp or +xsmp_interact). >> >> I'm not sure I follow - what does this have to do with his results? X >> connection or no X connection shouldn't affect the time spent >> redrawing, and certainly not by 50 seconds. >> >> ~Matt > > no X connection affects the time starting up, because Vim waits for its > connection with the X server to time out. This may or may not influence > the results shown above, depending on how they are collected (they will > if it is from invoking to exiting Vim) but it's always good to know. > > > Best regards, > Tony.
Adding -X to the test does not affect performances for me (but I'm not affected by this "slow redraw" bug) $ time gvim -X -u NONE -U NONE -f -S test-redraw-speed.vim vim7/src/README.txt real 0m12.262s user 0m10.233s sys 0m0.236s $ time gvim -u NONE -U NONE -f -S test-redraw-speed.vim vim7/src/README.txt real 0m12.411s user 0m10.465s sys 0m0.196s The difference is just noise. What I see though, is that the locale significantly affects speed. My tests so far were with a UTF-8 locale (eo_XX.UTF-8). en_US.UTF-8 gives the same results, but if I set locale to "C", then redraw is significantly slower (though not as bad as what reports François. $ export LANG="C" $ export LC_ALL="C" $ time gvim -u NONE -U NONE -f -S test-redraw-speed.vim vim7/src/README.txt real 0m38.200s user 0m36.206s sys 0m0.484s I tried several times, results are consistent. Setting locale to C instead of eo_XX.UTF-8 (or en_US.UTF-8) makes this benchmark more than 3 times slower. But it still does not explain the fact the redraw is 52 times slower on François's computer than on mine, even though he seems to have better hardware than mine. -- Dominique --~--~---------~--~----~------------~-------~--~----~ You received this message from the "vim_use" maillist. For more information, visit http://www.vim.org/maillist.php -~----------~----~----~----~------~----~------~--~---
