>I've ported ReCache to Linux, and tested it out. Oddly enough, it didn't help.
>I'm not sure if I've ported the spawnl() call in a wrong way (I'm doing a
>fork() and then an exec()), but it certainly doesn't help (the iteration
>time goes up from 0.201 to 0.203 secs). Perhaps the Linux MM is better than
>Windows after all.
Ah, well, worth a try I suppose.
Actually the virtual to physical memory mapping mechanism is a part of
the processor architecture, not the operating system - which is why I
thought it was worth a try. What may well be different is the way in
which the OS actually allocates physical memory and/or its strategy in
balancing disk buffers against process memory allocation, especially
when there is heavy demand on swap/page file I/O.
Your interpretation of spawnl() is probably correct. The child process
should start to execute immediately and the parent should not hang
around waiting for the child's exit code.
With Windoze I found that starting the child process halfway through
the second pass seemed to work best, maybe a different strategy would
work better with linux. It might be worth trying at the end of the
first pass then exiting ReCache immediately, or at the end of the
first pass & starting the second pass - you could then abort the
second pass at the halfway mark, the idea being that ReCache should
continue thrashing long enough for Prime95/mprime to get itself
initialized as far as allocating its work vectors.
>
>What _does_ help, however, is killing everything (by going to a different
>runlevel) and starting again (by going back). This is a bit surprising; I
>thought this kind of thing was reserved Windows users. (The iteration time
>is still at 0.201 secs, but it is in fact needed to get maximum speed back
>after running ReCache.)
>
>Result: I don't have any clue about what's going on.
Me neither, so far as linux is concerned.
I did find that ReCache "worked" in the sense of at least making things
no worse on a wide selection of systems running Win 9x & NT. It had least
effect on systems which had minimal physical memory, and most effect on
systems with lots of memory & high clock rate multipliers. e.g. on my
PII-333 system (96 MB, NT WS 4.0) on a 256K FFT a "random" start of
Prime95 gets an iteration time somewhere between 0.190 & 0.195 - towards
the high end if Prime95 is started automatically by means of a shortcut
in the "Startup" folder - whereas using ReCache I get 0.188 _consistently_.
If you find ReCache doesn't work for you - even on a Windoze machine -
then I'm sorry, but you do have the option not to use it!
Regards
Brian Beesley
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
