rec.autos.simulators

'optimising' compilers do not optimise in the same way that writing
'pure' machine code does, Most optimise to compensate for the severe
'overhead' and redundancy that modern API's add to code, which mean that
they are optimised 'relative' to how they would be if this optimisation
didn't occur, however these routines do not IN ANY WAY compare to what a
coder can do with code written from scratch.

oh, and BTW, to the ppl deriding GC's use of Integer math, Yes FPU's
have improved, but are STILL a long way from acheiving the performance
needed,
IIRC Quake II was entirely Integer calc based too.

Alan

: 'optimising' compilers do not optimise in the same way that writing
: 'pure' machine code does, Most optimise to compensate for the severe
: 'overhead' and redundancy that modern API's add to code, which mean that
: they are optimised 'relative' to how they would be if this optimisation
: didn't occur, however these routines do not IN ANY WAY compare to what a
: coder can do with code written from scratch.

  Have you ever heard the saying "premature optimisation is the root
of all evil" - most coders I know including assembly language
programmers accept that a well optimised C program is faster than a well
optimised assembler program simply coz the C compiler will kno when to
unroll a loop and similar.  A lot of people waste a lot of time coding
assembler under the mistaken belief it is more  efficient.  It can be -
for some things.   In the general case, however, a well written C
program will almost always be faster.

  The stuff about APIs is totally irrelevant - leads me to think you're
a windows programmer.

: oh, and BTW, to the ppl deriding GC's use of Integer math, Yes FPU's
: have improved, but are STILL a long way from acheiving the performance
: needed,
: IIRC Quake II was entirely Integer calc based too.

  Integer math is faster if you can get away with it - but notby much.
It's dumb to go out of your way to limit all float ops - but reduce them
to a minimum sure.

--
Richard G. Clegg       Only the mind is waving
    Networks and Non-Linear Dynamics Group
      Dept. of Mathematics, Uni. of York
     UPDATED WWW: http://manor.york.ac.uk/

I'm sure that GP3 performance is suffering with hardware 3D because the
software engine is over-optimised. COnverting all those int to floating
point texture co-ords, polygon co-ord and depth values is taking it's
toll on GP3's realism, looks and speed. IMO :)

Sorry, have to agree to disgree here.

Hehe, nope, *nix console C++ and oracle/sql programmer these days. Used
the API term because wasn't sure what platform your/your friends
experience was in, could have easily said shared library but went for
the lowest common denominator ;)

Not by much is still faster IMHO (but ONLY when done right)

Alan.

:>   Have you ever heard the saying "premature optimisation is the root
:> of all evil" - most coders I know including assembly language
:> programmers accept that a well optimised C program is faster than a well
:> optimised assembler program simply coz the C compiler will kno when to
:> unroll a loop and similar.  

: Heh. Yeah, but i'm sure that I can find a thousand and one  posts on
: usenet saying the counter :) (this argument is as divided as the current
: gp3 vs. gpl)

  (Grin) Now c'mon the GP3 vs GPL argument isn't that divided.  The GP3
vs punch in the face argument is divided.

:> however, a well written C
:> program will almost always be faster.

: Sorry, have to agree to disgree here.

  I agree.

:>   The stuff about APIs is totally irrelevant - leads me to think you're
:> a windows programmer.

: Hehe, nope, *nix console C++ and oracle/sql programmer these days. Used
: the API term because wasn't sure what platform your/your friends
: experience was in, could have easily said shared library but went for
: the lowest common denominator ;)

  In which case I apologise for the unwarranted abuse.

  Richard.

--
Richard G. Clegg       Only the mind is waving
    Networks and Non-Linear Dynamics Group
      Dept. of Mathematics, Uni. of York
     UPDATED WWW: http://manor.york.ac.uk/