Do somone know how to get such a whole list?

It would be fine, when funktions like WSASTARTUP also are testet in the report.

Well done...
Posted on 2003-05-18 04:47:09 by Bubu-Boy

There's really not much use in having clock cycle tables for API functions, if possible at all. Most of them take *a lot* of clock cycles, often depending on many factors so it's nearly impossible to create some kind of clock cycle table for it.

Posted on 2003-05-18 04:59:16 by Thomas
The idea of a list of clock cycle lengths is becoming more and more out dated.
The processors can execute instructions out of order if the dependancies allow, the operating system will time slice between various tasks, and you've got to consider memory latency for anything that isn't purely register based.

There are some basic rules to follow, but no real way of getting the speed of a procedure other than testing it, in the situation it was designed to be used in.

#1 APIs are slow, but may be the only way to solve your problem
#2 Memory is slow, if you've got to use it, try to be "predictable" - it'll help any prefetcher
#3 The algorithm is key. No use trying to optimise a bubble sort!

Posted on 2003-05-18 05:31:22 by Mirno