[wellylug] distro performance

[Cliff Pratt]

Jim Cheetham wrote:
> On Fri, Aug 25, 2006 at 05:32:48PM -0700, Brent Wood wrote:
>> I've mentioned before how running this under Linux is much faster
>> than Windows. Possibly a fuction of mingw/gcc efficiency, but as
>> pragmatists, we'll go with Linux, why is not really an issue. We
>> just want it to run fast.
> 
> An interesting result, basically observing that AMD64 and SLED10 may
> be a very good combination for speed ;-)
> 
> But it strikes me you've described a lot of effort gone into testing 
> different machines/kernels/compilers, without putting an equivalent 
> amount of effort into profiling the code itself to eliminate 
> bottlenecks.
> 
> I can anecdotally say that my first real-customer perl program took
> over 4 minutes to execute on my 486 with (errm ... red hat? this was
> years ago; I've forgotten) and when it was moved onto a 586 with a
> similar Linux, it took around 1 minute ... but the difference was
> primarily memory and disk, not cpu. Optimising to get rid of the disk
> access (in other words, learning to use perl properly!) took the
> whole thing down to 1:30 on the 486 and ~20 secs on the 586.
> 
I agree. I still remember an old mainframe tuning course I went on. All
week we tweaked and tuned, seeing barely measurable differences in
speed. Then we did a small tweak to the program, recompiled and it ran
like it was turbocharged. OK, the change was a pretty obvious one, but
even so....

Cheers,

Cliff