Federico Mena Quintero <[EMAIL PROTECTED]> writes: > - The X server uses 30% of the time. 5.5% is fbSolidFillmmx(). 7.9% is > "in kernel" - no idea what that means. Some DMA or something?
"In kernel" just means the kernel is running. It could be a system call or a page fault or an interrupt handler. For system calls and page faults the blame gets assigned to the application that caused them. For interrupt the blame gets assigned to whatever application happened to be running. So basically, in this case it means that 7.9% of the time spent in the X server is system calls or page faults. I do have a longterm plan to have sysprof report (1) _what_ the kernel was doing and (2) what parts of userspace apps were responsible for making it do those things. In sysprof HEAD "In kernel" does actually get assigned to the stacktrace that caused it to happen, but it still doesn't tell you what the "In kernel" code was doing. Sysprof HEAD is a little rough at the moment though. > > - Is there any way to get to know how many times a specific function was > > called and/or how long it took to proceed? > > Not with sysprof, unfortunately. You'd need something like gprof for > that. A valgrind skin to do the call counting would probably be a better bet. Writing valgrind skins is actually not that hard. Soren _______________________________________________ Performance-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/performance-list
