🚀 go-pugleaf

Chad Irby wrote:
> 
> Robert Salasidis (salasidis@medcor.mcgill.ca) wrote:
> :
> : Nobody argues that the PPC is faster in floating point. However for most
> : users, 99% of the time is spent in integer functions. Integer benchmarks
> : show that there is very little diffeence between the 2 processors (and
> : at least the x86 camp has a Pentium Pro to upgrade to today).
> 
> That might have been the case several years ago, but with the
> availability of machines with truly fast floating-point operations, a lot
> of the modern use of computers has moved further into types of software
> that *needs* a fast FPU.
> 
> Besides the obvious things like spreadsheets and CAD, a fast floating-point
> machine is useful for lots of graphics operations, not to mention basic
> (and common) things like games and sound processing.  The list gets
> bigger every day.

All games that are worth anything use absolutely zero floating point in critical 
areas. Floating point is too inefficient, and integers are used exclusively 
(this is for all games that require real time action).

I doubt most users spend more than a few seconds doing a spreadsheet recalc, but 
I will admit there are some that do - in my circles I can safely say that most 
computers are used for common business duties, WP, database (entry, display, 
reports), games, etc. All use very little floating point. Given a choice I would 
take faster integer processing over floating point any day for my applications 
(unless I could have both).

> 
> Even some of the basic display operations (like font manipulation) are
> starting to take advantage of the faster math operations involved.
> 
> Not to mention- why should people upgrade to the Pentium Pro, unless
> they're part of the 3% or so of the market using Windows NT?  If they run
> W95 or 3.1, they'll get a *slower* machine for their money.

At 200 MHz it does run 16 bit apps faster than the 133 MHz Pentium (and probably 
the 166 as well - though I haven't seen a benchmark yet (however for 16bit apps 
I will agree power/$ is lower). Although most use Windows 95, code in a 32 bit 
app (or any 32 bit game - and most new ones are) will still run at the 32 bit 
speed (as long as a system 16 bit system call is not being used). (therefore how 
much improvement over a Pentium there will be will depend on the application).

I use UNIX for most of my development work, so I would definitely see an 
improvement (add the UNIX folks to the 3% as well)

In article <30F35518.4209@medcor.mcgill.ca>, Robert Salasidis
<salasidis@medcor.mcgill.ca> wrote:

>All games that are worth anything use absolutely zero floating point in
critical 
>areas. Floating point is too inefficient, and integers are used exclusively 
>(this is for all games that require real time action).

Au contraire.  There is at least one 3-D gaming engine for 68040 and
PowerPC Macs that makes extensive use of floating point.  Integer math
being truly more efficient is a thing of the past, since modern CPUs
provide very fast FP.

The calculations made in 3-D games are much more natural in a FP number
system.  The only reason integer was ever used, really, is that the
programmer had no guarantee that large percentages of customers had FPUs. 
Now that a fast FPU is always present with either the PowerPC or the
Pentium, usage of FP in games should rise.

+-----------------------------------------------------------+
|Tim Seufert, bwanga@cats.ucsc.edu | UselessWastedSpace(tm) |
| "I never give them hell. I just tell the truth, and they  |
|  think it is hell."  -Harry S Truman                      |
+-----------------------------------------------------------+

Robert Salasidis <salasidis@medcor.mcgill.ca> wrote:
[snipt]
: All games that are worth anything use absolutely zero floating point in critical 
: areas. Floating point is too inefficient, and integers are used exclusively 
: (this is for all games that require real time action).

Really?

You're sure that no Mac-only 3D VR games make use of the FP on the PowerPCs?

--
-------------------------------------------------------------------------------
Lawson English                            __  __     ____  ___       ___ ____
english@primenet.com                     /__)/__) / / / / /_  /\  / /_    /
                                        /   / \  / / / / /__ /  \/ /___  /
-------------------------------------------------------------------------------

> I will admit there are some that do - in my circles I can safely say
that most 
> computers are used for common business duties, WP, database (entry, display, 
> reports), games, etc. All use very little floating point. Given a choice
I would 
> take faster integer processing over floating point any day for my
applications 
> (unless I could have both).
> 
> > 

Maybe on the PC side...most Mac users are working in graphics apps most of
time (the Mac has the low-to-mid-end graphics and high end DTP markets
practically locked--so much that more than a few prepress service bureaus
charge extra for working from PC files, and can get away with it). Most of
these programs, especially those compiled for the PPC, were written to use
the FPU as much as they can. Given the fact that the Power Mac 8100/80 ran
integer and boolean operations 3x as fast as the Quadra 840, but FPU
operations 15-20x as fast, it's easy to understand why.

---------------
Chris Woodfield - rekol@charm.net - http://www.charm.net/~rekoil

-- 
Christopher A. Woodfield
rekoil@charm.net - http://www.charm.net/~rekoil

Robert Salasidis <salasidis@medcor.mcgill.ca> wrote:
: All games that are worth anything use absolutely zero floating point in critical 
: areas. Floating point is too inefficient, and integers are used exclusively 
: (this is for all games that require real time action).

Sure? You know that some compiler optimize code by using the FPU for integer
calculations? How about the following code?

----------------
#include <stdio.h>

#define COUNT 101


main()
{
   double a=0;
   unsigned int i,j;
   double b[COUNT];

   for (i=0;i<COUNT;i++)
      b[i]=i;

   for (j=0;j<4000000;j++) {
      for (i=0;i<COUNT-1;i++)
	 a+=b[i]*b[i+1];
   }
   printf("%d: %d\n",j,a);
}
--------------------------------

On a Pentium, changing a and b to be "int" rather than "double" slows it
down. And that is despite "int" being 32 bits while "double" is 64 bits.
Change the "*" to "/", and using the FPU is _still_ faster.

Bernie

-- 
==============================================================================
I'm using an experimental version of "nn" to read news. So if anything seems
strange about the headers of this mail or post, please inform me. Thanks!

In article <4d2id2$8po@nnrp1.news.primenet.com>
Lawson English <english@primenet.com> writes:

> Really?
> 
> You're sure that no Mac-only 3D VR games make use of the FP on the PowerPCs?
> 
> --
> -------------------------------------------------------------------------------
> Lawson English                            __  __     ____  ___       ___ ____

Correct me if I'm wrong, but I've been led to believe that fixed-point
"integer" math is almost always quicker than fpu floating-point math.


Matt Peterson
University of Kansas, Experimental Psychology

Chad Irby (cirby@gate.net) wrote:

: Not to mention- why should people upgrade to the Pentium Pro, unless 
: they're part of the 3% or so of the market using Windows NT?  If they run 
: W95 or 3.1, they'll get a *slower* machine for their money.

	I've got one word for ya:  UNIX.

	The 80x86 UNIX (ala FreeBSD, Linux, etc.) workstation market 
might actualy have a chance in the mid-ranged systems market now that the 
PP is around.  Before, if you wanted power, you got a SPARC, Alpha, or 
MIPS.  Now we've got another option.
	Of course, there are the UNIXes for the 68k chips.

	And, of course, if you really wanted the power, you'd forget 
Intel and Motorola, and buy yourself an R10k, or an UltraSPARC.  ;)

--
 __                       __                                           __
/_/\  Peter Blackthorne  /_/\  "Reality is just a collective hunch."  /_/\ 
\_\/    pb@europa.com    \_\/       http://www.europa.com/~pb/        \_\/

In article <1996Jan11.185649.111671@kuhub.cc.ukans.edu>,
mattsp@stat1.cc.ukans.edu (Matt Peterson) wrote:

> Correct me if I'm wrong, but I've been led to believe that fixed-point
> "integer" math is almost always quicker than fpu floating-point math.
> 
> 
> Matt Peterson
> University of Kansas, Experimental Psychology

Then why did so many people buy math coprocessor upgrades for their
computers in the 386 era? Obviously because FPU math _can_ be much faster
then integer, at least in some circumstances.

-- 
Regards,       Joe Ragosta

Copyright Joseph M. Ragosta, 1995. Non-exclusive, royalty free license to distribute this post granted to any service provider except Microsoft. Fee for Microsoft is $1,000 per posting.

In article <4d4tuo$o4s@atheria.europa.com>, pb@europa.com (Peat Blackthorne) writes:

|> 	The 80x86 UNIX (ala FreeBSD, Linux, etc.) workstation market might
|> actualy have a chance in the mid-ranged systems market now that the
|> PP is around.  Before, if you wanted power, you got a SPARC, Alpha,
|> or MIPS.  Now we've got another option.

The BeBox comes to mind.

|> Of course, there are the UNIXes for the 68k chips.

SunOS 1, SunOS 2, and SunOS3 come to mind.

|> 	And, of course, if you really wanted the power, you'd forget 
|> Intel and Motorola, and buy yourself an R10k, or an UltraSPARC.  ;)

Or a PowerPC 604.  Check out the Jan '96 Byte.  BYTEmark Tests (higher
is better; Model, processor, and then two scores listed, integer, then
floating point.)  I've sorted in order of integer performance

DeskStation Raptor 3
266 MHz Alpha 21164
   3.75, 4.49

IBM Personal Computer Power Series 850
133 MHz PowerPC 604
   2.90, 2.92

Sun Ultra I Creator3D Model 170E 
167 MHz UltraSparc
   2.16, 1.54

Intel Reference System
150 MHz P6 (Pentium Pro)
   2.10, 2.60

Netpower Fastseries MP
200 MHz Mips R4400
   2.03, 2.08

Chris
-- 
Speaking only for myself, of course.
Chris Wood    christw@lexis-nexis.com   ChrisCWood@eworld.com

mattsp@stat1.cc.ukans.edu (Matt Peterson) writes:

>Correct me if I'm wrong, but I've been led to believe that fixed-point
>"integer" math is almost always quicker than fpu floating-point math.

You are wrong. See the source code I posted here recently.

Bernie
-- 
==============================================================================
I'm using an experimental version of "nn" to read news. So if anything seems
strange about the headers of this mail or post, please inform me. Thanks!

Bernd Bernie Meyer wrote:
> 
> Robert Salasidis <salasidis@medcor.mcgill.ca> wrote:
> : All games that are worth anything use absolutely zero floating point in critical
> : areas. Floating point is too inefficient, and integers are used exclusively
> : (this is for all games that require real time action).
> 
> Sure? You know that some compiler optimize code by using the FPU for integer
> calculations? How about the following code?
> 
> ----------------
> #include <stdio.h>
> 
> #define COUNT 101
> 
> main()
> {
>    double a=0;
>    unsigned int i,j;
>    double b[COUNT];
> 
>    for (i=0;i<COUNT;i++)
>       b[i]=i;
> 
>    for (j=0;j<4000000;j++) {
>       for (i=0;i<COUNT-1;i++)
>          a+=b[i]*b[i+1];
>    }
>    printf("%d: %d\n",j,a);
> }
> --------------------------------
> 
> On a Pentium, changing a and b to be "int" rather than "double" slows it
> down. And that is despite "int" being 32 bits while "double" is 64 bits.
> Change the "*" to "/", and using the FPU is _still_ faster.
> 
> Bernie

I tried out your code on a Pentium 120 with a QNX version of Watcom's
9.5 C compiler. This is what I get

using double 40.98 sec
      int    43.88

however changing the * to a +  (b[i]+b[i+i])
      double 40.98 sec
      int    17.05

I tried it out and surprisingly you are right that using double in the 
above example is more efficient (the compiler does not optimize by using 
double instead of 32 bit int - double is simply faster). Using doubles 
instead of ints would result in different results, as one allows numbers 
up to 10^307 while the other overflows at 2^32.

Eliminating a multiplication altogether however results in much faster 
integer performance. The moral of this story is in a critical loop, 
avoid multiplications (int or float) at all costs.

On Fri, 12 Jan 1996 08:06:31 -0500, jragosta@dca.net (Joe Ragosta)
wrote:

>In article <1996Jan11.185649.111671@kuhub.cc.ukans.edu>,
>mattsp@stat1.cc.ukans.edu (Matt Peterson) wrote:
>
>> Correct me if I'm wrong, but I've been led to believe that fixed-point
>> "integer" math is almost always quicker than fpu floating-point math.
>> 
>> 
>> Matt Peterson
>> University of Kansas, Experimental Psychology
>
>Then why did so many people buy math coprocessor upgrades for their
>computers in the 386 era? 

Likely, to run Autocad and 3D Studio.  And a lot of people bought
Windows- sometimes people don't always know better.