NS> CB> It was a very interesting point of how just minor, reasonable changes
NS> CB> in code can totally change the benchmarking results, especially when
NS> CB> the benchmark is onlya single program.  And how compiler makers can
NS> CB> quite 'reasonably' proclaim their compiler produces the best code.
NS> Quite amazing really. Anyone could take any case for any compiler and 
claim
I thought it was interesting.  I had gone through some old mags and
happened to read it.  Very simple program with reasonable changes each
time.  And each time there was a different clear winner.  That's a
warning about simple benchmarks, and how, with small changes to your
code, you might speed up your program.
NS> something for it. I like GCC and friends because A) it's free and GNU
NS> B) it really doesnt claim anything or say how better it is and C) Its the
NS> product of 30 some years work so that there proves it must be pretty 
decent
I do like the free part....  As for the quality, .... Well, I still use
DJGPP based on 2.6.3 of the compiler.  Although the compiler itself is
okay, the DJGPP part (library etc.) is rather poor.  Not quite up to
ANSI C specs.
There is a new version, I just haven't found a place locally to grab it,
and I don't want to do long distance.
--- QScan/PCB v1.19b / 01-0162
---------------