Re: from elsewhere, an assembler

På Sat, 14 Apr 2007 18:34:15 +0200, skrev /\\o//\annabee <Wannabee@xxxxxxxxxxxxxxx>:

Now, in every test I do, involving some useful FP operations, I get _very_ high cyclecounts. But, in code that performs _alot_ of them, so many that adding the test
result of each should yield 100s of thousands, if not million of cycles,
actually turn out to run in a very "few" cycles, on avarage (1000-1500),
when I perform the many operations many times and divide.

this is a bit of a overestimation. But at least they addup in the 10 thousands of cycles when measure seperatly, (per particle) whereas a lot of other code also runs per particle.
What I mean is that basically, the per cycle count should be higher, much higher for the number of instruction i use, whereas they are supricingly low, for code that is pretty weird. Of course, even if "low" 1000 cycles per particle is a bit absurd, and i should maybe be happy if I could get it down to 100.
.