From Bugzilla Helper: User-Agent: Mozilla/4.75 [en] (X11; U; Linux 2.2.16-22 i686) Using shared memory for moving image data between the X server and the graphical application is very slow when server acceleration is activated. This was not the case in previous XFree86 versions. On my PIII 933MHz/RIVA TNT2/i815, moving a 800x600 8-bit image with XShmGetImage() takes around 72 milliseconds with accerelation turned on, as is the default setup. When turning off acceleration with 'Option "NoAccel" "1"' in /etc/X11/XF86Config-4, this drops to the more reasonable 3 milliseconds. Similar slowdowns are noticed with the same software on a Celeron 677/i815/Matrox G200. Reproducible: Always Steps to Reproduce: 1. Install Red Hat 7.1; X server acceleration is default set to active 2. Run an application which moves image data around with XShmGetImage() 3. Notice low framerate 4. Turn off acceleration with Option "NoAccel" "1" in XF86Config-4 5. Framerate increases dramatically The tested application is our Olex seabed mapping and visualisation system, previously installed on several hundred Red Hat 5.1 to 7.0 systems with XFree86 3.3.x. A demonstration copy can be downloaded from www.olex.no.
Attach your config file and X server logs using the "Create a new attachment" link below.
Please feel free to reopen this report after attaching the requested information, if this is still a problem.