|From:||Brandon J. Van Every|
|Subject:||Re: [Chicken-users] Nursery sizing considered stupid|
|Date:||Fri, 21 Jul 2006 10:02:40 -0700|
|User-agent:||Thunderbird 18.104.22.168 (Windows/20060516)|
John Cowan wrote:
felix winkelmann scripsit:Then tell me the nursery size doesn't make a difference. It does, at least on Linux machines.My claim has never been that the nursery size as such makes no difference, but that the method for determining the best nursery size doesn't actually do so -- particularly under realistic conditions for a build. It weeds out the worst sizes but otherwise picks one more or less at random. Therefore, all it does is slow down the build.
On my box it doesn't even weed anything out. When I take 100 nsample runs, the average is the same within 1%, no matter what the stack size is. That's why I implemented NOISE_THRESHOLD of 5% for the CMake build. And really, my professional benchmarking experience is anything under 20% is not to be taken seriously. So there are clearly classes of machines for which the current nsampling ritual is a waste of time. A seriously major waste of time, actually, when 100 samples are needed for stability.
One thing I'm certain of: ./configure's generation of chicken-defaults.h in Makefile.am *is definitely* a randomizer. No question. 3 samples is way substandard.
Brandon Van Every
|[Prev in Thread]||Current Thread||[Next in Thread]|