I get the feeling the answer to my ultimate question is "no," but... I'm curious as to how much 2D acceleration features (hw blit, hw cursor, etc.) actually save in terms of data transferred to the graphics chip relative to not having them turned on. In theory, hw blit alone is orders of magnitude lower for things like moving a window across screen (the work is offloaded to the graphics chip, so the transfer over PCI/AGP/whatever is minimal--assuming discrete graphics componentry, yada yada). I know this seems kinda old-skool, but I'm wondering if there's any way (short of a hardware PCI bus analyzer), to do a test where one can get an approximate byte count of data transferred to a discrete graphics chip (say a cheap old S3, Trident TGUI or similar) from the host. Basically, I want to know about what quantity of data gets shoved over the bus during, say, a 5-minute Gnome session both with and without accel. I was thinking of taking the XF source and sprinkling one of the drivers with log output (tanking performance during the test, but that's OK) and compiling the results on a per-call basis, but my cursory inspection of the code indicates that various DMAisms would get in the way (invalidating any results I may get). I also get the creeping feeling that either this sort of thing has been done before or can't be done without major mods to the code (like logging damn near every function call everywhere and asking questions later). Thoughts? -- I've got a lovely bunch of coconuts. Lodged in my thoat. Help! -- B.J. Black _______________________________________________ XFree86 mailing list XFree86@xxxxxxxxxxx http://XFree86.Org/mailman/listinfo/xfree86
