Messages in this thread |  | | From | (Dale R. Worley) | Subject | Re: MMX performance.... | Date | Thu, 6 Feb 1997 13:31:10 GMT |
| |
In article <ImxFFZK00iWY07KnJY@andrew.cmu.edu> Phillip G Ezolt <pe28+@andrew.cmu.edu> writes: From what I understand, everytime you switch between MMX mode and regular FP mode, 100 or so cycles are burned. If you are context switching alot (any multitasking enviornment), this would seem to add up.
Assuming that the "cycles" are fundamental CPU cycles (as opposed to memory accesses, or something), that could take 1 microsecond or less (depending on your clock speed), which isn't much. Even better, if you arrange that the kernel doesn't use either FP or MMX, then MMX/FP switches can only happen when switching user processes, which is fairly infrequent -- and the MMX/FP switchover overhead is probably much smaller than the scheduler overhead.
Dale -- Dale R. Worley Ariadne Internet Services Voice: +1 617-899-7949 Fax: +1 617-899-7946 E-mail: worley@ariadne.com "Internet-based electronic commerce solutions to real business problems."
|  |