64-bit pipe dreams
LocalFolk
(1) that I would hard-wire my own 64-bit "integer processor"
out of discrete components and someday hook it up to
my PC.
(2) that I would someday be wealthy enough to purchase
the reclusive "Itanium" machine, that would someday be
available to consumers.
It never occurred to me that, right now, there might be a
64-bit integer co-processor somwhere on the market that
could be purchased for under $1000. US.
Anyone know of such a device? Or should I keep waiting
for "someday" to arrive?
Thank in advance,
-bear
Martin Kiewitz
> I used to have 2 fantasies:
> (1) that I would hard-wire my own 64-bit "integer processor"
> out of discrete components and someday hook it up to
> my PC.
> (2) that I would someday be wealthy enough to purchase
> the reclusive "Itanium" machine, that would someday be
> available to consumers.
> It never occurred to me that, right now, there might be a
> 64-bit integer co-processor somwhere on the market that
> could be purchased for under $1000. US.
lol
> Anyone know of such a device? Or should I keep waiting
> for "someday" to arrive?
For what do you need 64-bit ? Buy yourself a PS2. Cheaper than 1000$
and has lots of games.
cu Kiewitz
P.S.: damn, I hate troll postings <g>
--
Randall Hyde
LocalFolk wrote in message <20020121205538...@mb-mk.aol.com>...
>I used to have 2 fantasies:
>(1) that I would hard-wire my own 64-bit "integer processor"
> out of discrete components and someday hook it up to
> my PC.
The cost of communicating with the "integer processor"
would be greater than doing 64-bit extended precision
arithmetic (division *possibly* excepted). Also, keep in mind
that you can use the FPU for 64-bit signed integer arithmetic
and the MMX instruction set for 64-bit logical operations.
>(2) that I would someday be wealthy enough to purchase
> the reclusive "Itanium" machine, that would someday be
> available to consumers.
Ths isn't a fantasy, this is a fact. *Someday* you will
be able to purchase an IA-64 machine for home use
(though not necessarily the Itanium). Remember when
Pentium Pros were for "servers only" and only servers
needed that kind of power?
>It never occurred to me that, right now, there might be a
>64-bit integer co-processor somwhere on the market that
>could be purchased for under $1000. US.
Absolutely no market for such a device. 64-bit CPUs
abound and, as I noted above, 64-bit arithmetic on
32-bit CPUs is much faster than communicating with
a coprocessor.
>
>Anyone know of such a device? Or should I keep waiting
>for "someday" to arrive?
No such device exists for use on PCs (that I'm aware of).
Of course, if all you need is arithmetic, it isn't that hard
to wire up an ALU yourself. But usng it's going to be
a lot more work and be a lot slower than using the
native capabilities of your 32-bit CPU. Also, 64-bit
RISC chips are fairly cheap (hey, they're used in
video games all over the place). You could wire
up one of those. However, there'd be no performance
or "ease of programming" benefit to doing so.
Randy Hyde
LocalFolk
>> Anyone know of such a [64-bit] device? Or should I keep waiting
>> for "someday" to arrive?
Martin Kiewitz responded:
>For what do you need 64-bit ? Buy yourself a PS2. Cheaper than 1000$
>and has lots of games.
Thanks for indulging me, to date. My desire is to do ultra-speedy
set processing. I'd like to do things like setting and counting bits,
then keeping a statistical log, based on how many bits were set
after a particular process (i.e. ANDing 2 bitsets). for a number of
different real-world (although not very lucrative) and beezarro projects.
For instance, I'd like to (say, randomly) generate a bunch of (as in
a couple hundred million) 50-bit values, and then do a quick bit-count
on the generated values, and increment one of 50 variables, based on
how many bits were set. I know I can do this fairly easily using standard
methods, but then I'd like to do it, say 10,000 more times, and unless
I've got hardware counting the bits, I'm in for a long coffee break.
So, yeah. I guess I knew in the back of my mind that PS2 used
a 64-bit processor of some kind. And the games sure would be
a pleasing diversion! However, if I went the PS2 route, would I
have access to the CPU for my own beezarro purposes? My goal,
for instance, is not to merely be the proud *owner* of a 64-bit machine,
but to do my little mad-scientist thing with the CPU. I know, from
here I should probably surf the PS2 newsgroups, but I figured, what
the hey, a couple of quick follow-up posts to the originator of the
idea might be all it takes.
-bear
"Neither a borrower nor a troller be." -MacHomer
Martin Kiewitz
> >For what do you need 64-bit ? Buy yourself a PS2. Cheaper than 1000$
> >and has lots of games.
> Thanks for indulging me, to date. My desire is to do ultra-speedy
> set processing. I'd like to do things like setting and counting bits,
> then keeping a statistical log, based on how many bits were set
> after a particular process (i.e. ANDing 2 bitsets). for a number of
> different real-world (although not very lucrative) and beezarro projects.
Great. Why don't you use some 32-bit x86 cpu ?
64-bit cpus are a) extremly expensive and b) not for end-users
Even Intel does not see 64-bit architecture for SOHO use.
If your computer is too slow:
a) don't use Windoze
b) upgrade your CPU to Ghz speed
c) get more RAM
I don't see any use in getting a 64-bit CPU for your "purposes".
If you work for a company that needs this, just let them contact
Intel. You will get some 64-bit stuff for sure.
cu Kiewitz
--
LocalFolk
Looks like MMX might be the way to go.
-bear
Jerry Coffin
loca...@aol.com says...
[ ... ]
> It never occurred to me that, right now, there might be a
> 64-bit integer co-processor somwhere on the market that
> could be purchased for under $1000. US.
Sure there is -- find a used Alpha workstation (they routinely sell
for under $1000 US) and (assuming your current machine has an
Ethernet connection on it) plug the two together. A "coprocessor"
that's cheap, effective, easy to use, supports a range of operating
systems (Windows NT, OpenVMS and Linux being the most obvious) has a
good range of development tools, etc.
My main question is why anybody (including Intel) gets excited about
Itanium or any of its ilk -- the Alpha's stable, proven, and mostly a
better design than Intel's anyway. Its primary shortcoming has been
(lack of) marketing.
--
Later,
Jerry.
The Universe is a figment of its own imagination.
michelle
be out fairly soon. The P4 is getting to be old news now (you know, like
cheap enough where scrooges like me are about to buy one).
Randall Hyde
Martin Kiewitz wrote in message ...
>On Tue, 22 Jan 2002 18:41:04 UTC, loca...@aol.com (LocalFolk) wrote:
>
>Great. Why don't you use some 32-bit x86 cpu ?
>64-bit cpus are a) extremly expensive and b) not for end-users
Hmmmm....
Thousands upon thousands of Nintendo-64 users would
disagree with this statement.
Actually, a large percentage of the 64-bit RISC chips are
quite cheap.
Of course, high-end 64-bit general purpose computer
systems don't (yet) fall into this category.
Randy Hyde
Randall Hyde
LocalFolk wrote in message <20020122125445...@mb-cq.aol.com>...
>
>Thanks for indulging me, to date. My desire is to do ultra-speedy
>set processing. I'd like to do things like setting and counting bits,
>then keeping a statistical log, based on how many bits were set
>after a particular process (i.e. ANDing 2 bitsets). for a number of
>different real-world (although not very lucrative) and beezarro projects.
Counting bits is probably best done in 32-bit mode on an x86
using the high-speed DWORD bit counting algorithm that
gets posted here every now and then.
Intersection, union, set difference, and other set operations
*may* be faster using MMX, but I haven't done the analysis.
Now I won't swear by this; someone may have come up with
a fast MMX bit counting algorithm, but I've not seen one yet.
This would be a good question to post as a separate thread
to see any any optimization/MMX mavens can figure something
out.
Randy Hyde
Martin Kiewitz
wrote:
> >Great. Why don't you use some 32-bit x86 cpu ?
> >64-bit cpus are a) extremly expensive and b) not for end-users
> Hmmmm....
> Thousands upon thousands of Nintendo-64 users would
> disagree with this statement.
Arghl. He said he wanted to use that thing for something *BUT* games.
First, I said he should by a PS2.
> Actually, a large percentage of the 64-bit RISC chips are
> quite cheap.
RISC <-> x86. You know, we are in comp.lang.asm.x86
> Of course, high-end 64-bit general purpose computer
> systems don't (yet) fall into this category.
x86 64-bit are not supposed to be for SOHO users.
cu Kiewitz
--
Phil Carmody
>
> On Tue, 22 Jan 2002 18:41:04 UTC, loca...@aol.com (LocalFolk) wrote:
>
> > >For what do you need 64-bit ? Buy yourself a PS2. Cheaper than 1000$
> > >and has lots of games.
> > Thanks for indulging me, to date. My desire is to do ultra-speedy
> > set processing. I'd like to do things like setting and counting bits,
> > then keeping a statistical log, based on how many bits were set
> > after a particular process (i.e. ANDing 2 bitsets). for a number of
> > different real-world (although not very lucrative) and beezarro projects.
You don't need a 64-bit architecture to do what you want.
Whilst playing around with message-digest functions I did a fair bit of
bit twiddling and counting, etc. and my platform of choice was 64-bit,
but everything can be done on any x86 machine. Anything with MMX almost
certainly has enough to do what you want, for example. And remember, for
some operations, MMX is 64-bit.
> Great. Why don't you use some 32-bit x86 cpu ?
> 64-bit cpus are a) extremly expensive and b) not for end-users
Where did you pull that opinion from? As an end-user (programmer) I
_almost always_ chose my 64-bit architecture. And your preconceptions
about price are ill-founded - 64-bit CPUs are _not_ expensive if you
don't mind second hand ones. Just yesterday I saw a R10000-powered
Indigo 2 on ebay (uk) with a #225 buy-now price (and bids were way less
than that). If I lived in the UK I'd have snapped that up immediately.
Phil
Phil Carmody
> LocalFolk wrote in message <20020122125445...@mb-cq.aol.com>...
> >
> >Thanks for indulging me, to date. My desire is to do ultra-speedy
> >set processing. I'd like to do things like setting and counting bits,
> >then keeping a statistical log, based on how many bits were set
> >after a particular process (i.e. ANDing 2 bitsets). for a number of
> >different real-world (although not very lucrative) and beezarro projects.
>
> Counting bits is probably best done in 32-bit mode on an x86
> using the high-speed DWORD bit counting algorithm that
> gets posted here every now and then.
The Mathisen/Harley bitslice algorithm the fastest algorithms I've seen
here, and is progressively faster the wider your word-size is.
> Intersection, union, set difference, and other set operations
> *may* be faster using MMX, but I haven't done the analysis.
>
> Now I won't swear by this; someone may have come up with
> a fast MMX bit counting algorithm, but I've not seen one yet.
> This would be a good question to post as a separate thread
> to see any any optimization/MMX mavens can figure something
> out.
Hmmm, google fails me, all I can find is:
and
which led to:
http://groups.google.com/groups?hl=en&selm=385027c1%40news.iglou.com
I'm sure there's enough there for someone to create a nice tight asm
library for MMX x86.
Phil
Ed Beroset
>In article <20020121205538...@mb-mk.aol.com>,
>loca...@aol.com says...
>
>My main question is why anybody (including Intel) gets excited about
>Itanium or any of its ilk -- the Alpha's stable, proven, and mostly a
>better design than Intel's anyway. Its primary shortcoming has been
>(lack of) marketing.
Since Compaq sold the both the Alpha technology and the chip-making
facilities involved to Intel last year, (and because Intel acquired
the fab plant and Alpha designs from DEC in 1999) I think to safe to
say that the Alpha is quite dead. Those still at Compaq who were
working on the Alpha are being transferred over to Intel where they
will be working on Itanium and derivatives.
That's not to say that it wasn't a very nice processor -- just that
it's a dead end in today's environment, at least by that name. I
anticipate Alpha features in future Itanium derivatives.
Alpha is dead! Long live Alpha!
Ed
Mark Gibson
What I really want is a CPU with 16+ general purpose registers that supports
128-bit or better arithmetic ops in hardware. The best system I've found
'til now has been the VAX line, but that's ancient history. A tight piece
of 32-bit code written for a modern x86 clone CPU will run circles around
the stuff I was running on a VAX 4000-300, but I miss that VAX!
Regards,
Mark
P.S. I'm in the market for a spiffy used VAX box. I want to do some
heavy-duty interger number crunching on it, so it has to have at least
128+ MB of RAM. I'm not so picky about the OS, but newer-is-better tends to
apply...as long as it is some flavor of VMS.
LocalFolk
I haven't been around this NG long enough
(who would've guessed! :-) to see the
aforementioned bit-counting routines.
In church last night I cranked out a
6-statement 32-bit counter, which
would probably not be too bad in a
64-bit mileau. But I was really hoping
for a hardware bit-count. Because going
from a dumass 64-loop BTST-and-add
scenario to my byte-lookup method
would convert a "years-long process"
into a "months-long process", on a
333MHz AMD. Upping to a 2GHz
Pentium would (arithmetically speaking)
turn it into a "weeks-long process."
I was kind of hoping to crank results over
a weekend or (maybe) a few hours.
Hence the need^H^H^H^Hdesire for
a 64-bit hardware bitcount.
And I don't need a completely "general"
processor, although I am looking for
something other than "just games."
And, although I am posting to x86,
I don't feel bound to the architecture.
I started life as a Mac fanatic in
68K-land.
I really appreciate the input. It's
an envelope-pushing thing, so it *is*
a lot of fun to talk about. :-)
-bear
Jerry Coffin
ber...@mindspring.com says...
[ ... ]
> Since Compaq sold the both the Alpha technology and the chip-making
> facilities involved to Intel last year, (and because Intel acquired
> the fab plant and Alpha designs from DEC in 1999) I think to safe to
> say that the Alpha is quite dead.
I quite agree that it's dead basically dead -- I just find the fact
disgusting. It looks to me like Intel's competitors noticed that the
x86 did well in spite of a lousy design, so they designed the PowerPC
and the SPARC to be even worse. Not to be outdone at their own game,
Intel responded with the IA64, the worst new design since the Data
General Nova.
I mean, I realize Intel's sick of people complaining about there
being too few registers, but now they've gone so far the other
direction that it takes all kinds of extra tricks for it to finish a
context switch in less time than some people switch spouses...
Terje Mathisen
>
> What I really want is a CPU with 16+ general purpose registers that supports
> 128-bit or better arithmetic ops in hardware. The best system I've found
> 'til now has been the VAX line, but that's ancient history. A tight piece
> of 32-bit code written for a modern x86 clone CPU will run circles around
> the stuff I was running on a VAX 4000-300, but I miss that VAX!
>
> Regards,
> Mark
>
> P.S. I'm in the market for a spiffy used VAX box. I want to do some
> heavy-duty interger number crunching on it, so it has to have at least
> 128+ MB of RAM. I'm not so picky about the OS, but newer-is-better tends to
> apply...as long as it is some flavor of VMS.
Wanting to own a VMS box is something I can understand, but claiming
that it is to heavy-duty integer ops does not ring true at all:
_Any_ modern cpu, including the ARM processor in your Nokia cell phone,
will probably run rings around the last VAX ever manufactured, and will
do so using so much less mains power that it should be enough to pay for
itself many times over, relative to a free VAX.
OTOH, VMS on a 264 Alpha could be quite useful. :-)
Terje
--
- <Terje.M...@hda.hydro.com>
"almost all programming can be viewed as an exercise in caching"
Terje Mathisen
> for a hardware bit-count. Because going
> from a dumass 64-loop BTST-and-add
> scenario to my byte-lookup method
> would convert a "years-long process"
> into a "months-long process", on a
> 333MHz AMD. Upping to a 2GHz
> Pentium would (arithmetically speaking)
> turn it into a "weeks-long process."
>
> I was kind of hoping to crank results over
> a weekend or (maybe) a few hours.
> Hence the need^H^H^H^Hdesire for
> a 64-bit hardware bitcount.
Get a used 264 Alpha, with the NSA opcode additions!
This includes bitcount and bitscan operations.
Matthew Smith
memory,
you can speed up your program in other ways.
First generate a table of 256 (or 64k) values which contain the number
of bits set in each byte (or word)
then you can find the number of bits in a 64-bit word much quicker
e.g (excuse the very bad asm, its been a while)
num64: 8 bytes
align 256
table256: 256 bytes set to number of bits set in each value
mov bx,table256
mov si,num64
mov bl,[si++] ;(lodb thru al probably)
mov al,[bx]
mov bl,[si++]
add al,[bx]
mov bl,[si++]
add al,[bx]
mov bl,[si++]
add al,[bx]
mov bl,[si++]
add al,[bx]
et voila al contains number of bits set in 64-bit number
Matt
Phil Carmody
> Leaving aside the fact that Pentiums and up write 64bits at a time to
> memory,
> you can speed up your program in other ways.
>
> First generate a table of 256 (or 64k) values which contain the number
> of bits set in each byte (or word)
My gut feeling is to avoid 64K tables unless that's the only thing your
code does. Even then I'd always look for a method with a smaller
footprint.
> then you can find the number of bits in a 64-bit word much quicker
>
> e.g (excuse the very bad asm, its been a while)
>
> num64: 8 bytes
>
> align 256
> table256: 256 bytes set to number of bits set in each value
>
> mov bx,table256
> mov si,num64
>
> mov bl,[si++] ;(lodb thru al probably)
> mov al,[bx]
> mov bl,[si++]
> add al,[bx]
> mov bl,[si++]
> add al,[bx]
> mov bl,[si++]
> add al,[bx]
> mov bl,[si++]
> add al,[bx]
>
> et voila al contains number of bits set in 64-bit number
I'd try to smear the operations a bit more to separate dependent
operations
(and actually do all 8 bytes!) Do any x86-alikes permit 2 memory reads
per tick?
push ebp
mov esi, num64 ; address of 64 bit value
xor eax, eax
mov edi, table256 ; address of LUT
xor ebx, ebx
mov al, [esi]
xor ecx, ecx
mov bl, [esi+1]
xor edx, edx
mov cl, [esi+2]
add ebp, [eax]
mov dl, [esi+3]
add ebp, [ebx]
mov al, [esi+4]
add ebp, [ecx]
mov bl, [esi+5]
add ebp, [edx]
mov cl, [esi+6]
add ebp, [eax]
add dl, [esi+7]
add ebp, [ebx]
; begin next word
add ebp, [ecx]
add ebp, [edx]
If memory doesn't work that way, then possibly preloading one of the E?X
registers with the whole word, and extracting the bytes one at a time
would cope too.
Phil
Steven Graham
All I can say is
DANM INTEL AND DAMN ITANIUM
I'd personally go AMD Hammer (x86-64)
Steven Graham
--
Posted via Mailgate.ORG Server - http://www.Mailgate.ORG
Randall Hyde
Martin Kiewitz wrote in message ...
>wrote:
>
>> >Great. Why don't you use some 32-bit x86 cpu ?
>> >64-bit cpus are a) extremly expensive and b) not for end-users
>> Hmmmm....
>> Thousands upon thousands of Nintendo-64 users would
>> disagree with this statement.
>
>Arghl. He said he wanted to use that thing for something *BUT* games.
Who says you have to use a Ninetendo-64 to write or play games.
With an appropriate interface, it could be used as a general purpose
64-bit coprocessor. Of course, I still doubt the utility of doing so,
but the fact that such devices exist refutes the opinion that
64-bit processors are expensive and not for home users.
>First, I said he should by a PS2.
Question: enlighten me. What 64-bit processor did the
PS/2 use? (Are was taking about the IBM PS/2 personal
computer system here?) I've completely missed out out
this bit of computer history arcania.
>
>> Actually, a large percentage of the 64-bit RISC chips are
>> quite cheap.
>
>RISC <-> x86. You know, we are in comp.lang.asm.x86
The original poster *did* discuss wiring up his own
ALU to accomplish this. I'd assume he would be willing
to use an non-x86 device if he were willing to do that.
Besides, we are talking about connecting a 64-bit CPU
to an x86 system here (or, at least, I am), which is perfectly
appropriate for this group. Heck, I don't even think the
charter of this group would forbid suggesting an alternative
to the x86 if that's a better approach.
>
>> Of course, high-end 64-bit general purpose computer
>> systems don't (yet) fall into this category.
>
>x86 64-bit are not supposed to be for SOHO users.
I remember when they said the same thing about
32-bit x86 systems :-) ("the 80386 is not supposed
to be for SOHO users...").
Randy Hyde
Randall Hyde
LocalFolk wrote in message <20020123133706...@mb-mk.aol.com>...
>Thanks for all the creative input!
>I haven't been around this NG long enough
>(who would've guessed! :-) to see the
>aforementioned bit-counting routines.
>
>In church last night I cranked out a
>6-statement 32-bit counter, which
>would probably not be too bad in a
>64-bit mileau. But I was really hoping
>for a hardware bit-count. Because going
>from a dumass 64-loop BTST-and-add
>scenario to my byte-lookup method
>would convert a "years-long process"
>into a "months-long process", on a
>333MHz AMD. Upping to a 2GHz
>Pentium would (arithmetically speaking)
>turn it into a "weeks-long process."
>
If your assembly sequence is only six
statements long, its probably an O(n)
(n being the number of bits to check)
algorithm that isn't particularly fast.
There is an algorithm that is O(log n)
(the fabled "bit counting algorithm" that
appears around here).
I received the following email from Norbert
concerning my comments about MMX.
I'll post them here because I bet he intended
to "reply group" rather than just "reply" to
my message:
>>>>>
Check the AMD Athlon Code Optimization Guide. It gives a
popcount() implementation using MMX in the MMX section.
This can be combined with Terje Mathisen's CSA approach
to yield a O(log(n)) popcount() for very large numbers.
<<<<
Randy Hyde
Martin Kiewitz
wrote:
> >Arghl. He said he wanted to use that thing for something *BUT* games.
> Who says you have to use a Ninetendo-64 to write or play games.
> With an appropriate interface, it could be used as a general purpose
> 64-bit coprocessor. Of course, I still doubt the utility of doing so,
> but the fact that such devices exist refutes the opinion that
> 64-bit processors are expensive and not for home users.
You know that N64 was made for games, so the included CPU(s) are
speciailized for gaming purposes.
Intel x86 architecture =! N64 architecture.
> >First, I said he should by a PS2.
> Question: enlighten me. What 64-bit processor did the
> PS/2 use? (Are was taking about the IBM PS/2 personal
PS2, not PS/2. -> Playstation 2. You get it ? And that one is 64-bit.
If I would get a gaming console for such job that you described above,
I would get a PS2.
> computer system here?) I've completely missed out out
> this bit of computer history arcania.
Me 2.
> >> Actually, a large percentage of the 64-bit RISC chips are
> >> quite cheap.
> >RISC <-> x86. You know, we are in comp.lang.asm.x86
> The original poster *did* discuss wiring up his own
> ALU to accomplish this. I'd assume he would be willing
> to use an non-x86 device if he were willing to do that.
Then this is the wrong newsgroup.
> Besides, we are talking about connecting a 64-bit CPU
> to an x86 system here (or, at least, I am), which is perfectly
Connecting ? What do you mean with connecting ?
> >> Of course, high-end 64-bit general purpose computer
> >> systems don't (yet) fall into this category.
> >x86 64-bit are not supposed to be for SOHO users.
> I remember when they said the same thing about
> 32-bit x86 systems :-) ("the 80386 is not supposed
> to be for SOHO users...").
[ ] You know how much cooling a 64-bit x86 processor needs...
If you want to put that thing in your fridge, it will work, otherwise
not. That's why. And anyway, this CPU is way too fast for desktop and
it uses too much power as well.
cu Kiewitz
--
Jerry Coffin
rh...@cs.ucr.edu says...
[ ... ]
> >First, I said he should by a PS2.
>
> Question: enlighten me. What 64-bit processor did the
> PS/2 use? (Are was taking about the IBM PS/2 personal
> computer system here?) I've completely missed out out
> this bit of computer history arcania.
I'm reasonably certain he's talking about the Sony Playstation 2,
which uses a 64-bit MIPS.
Juzzer
> facilities involved to Intel last year, (and because Intel acquired
> the fab plant and Alpha designs from DEC in 1999) I think to safe to
> say that the Alpha is quite dead. Those still at Compaq who were
> working on the Alpha are being transferred over to Intel where they
> will be working on Itanium and derivatives.
>
>
Very sad - I was indirectly involved at Digital when Compaq took over in 98
and at a 'celebratory' barbecue heard fond reminiscences of past glories
such as Alpha. Their optimism for the future proved shortlived.
Juz
Juzzer
> Thanks, Randy!
>
> Looks like MMX might be the way to go.
>
> -bear
Perculiar thread !
Of course, nevermind 64-bit, nobody has mentioned that most new pc's come
with 128-bit coprocessors as standard - on board their graphics cards !
Must be possible to tailor the input to these to provide very quick
bit-checking, though even if the results aren't as you expect, they sure
might look pretty .......
Juz
LocalFolk
It took me awhile to interpret your code, but
it is truly sweet! You juggle your memory moves
and your register adds like a RISC machine.
Can't wait to try it out!
-JH (aka bear)
Martin Kiewitz
wrote:
> Of course, nevermind 64-bit, nobody has mentioned that most new pc's come
> with 128-bit coprocessors as standard - on board their graphics cards !
>
> Must be possible to tailor the input to these to provide very quick
> bit-checking, though even if the results aren't as you expect, they sure
> might look pretty .......
Great. That's the most portable and best solution I have ever heard
about.
It won't even run with a different graphic card. Argh.
cu, Kiewitz
--
Beth
> Martin Kiewitz wrote:
> >First, I said he should by a PS2.
>
> Question: enlighten me. What 64-bit processor did the
> PS/2 use? (Are was taking about the IBM PS/2 personal
> computer system here?) I've completely missed out out
> this bit of computer history arcania.
Randy!!! Surely you're not that much of a recluse? PS/2 is IBM, PS2 is
Sony...he means the new PlayStation...thankfully, IBM had that bizarre thing
with putting a forward slash into the name (what was that all about? OS/2 as
well...anyone?:) so it is possible to tell them apart...although, I wonder
how many complete newbies have heard the phrase "PS/2 mouse" and then tried
to stick a mouse designed for the PS2 * into the back of their PC...hehehe
:)
Geez Louise, Randy...
Although, thinking about it, one could possibly claim that the PS/2 had
better than a 64-bit processor available but actually an 80-bit (!!!) maths
co-processor (if you'd stuck a 80287 or something into it :), nicht wahr?
Modern Pentia also have impressive 128-bit internal stuff and 64-bit data
paths...the problem with these sorts of discussions is that "bitage" is so
lossely defined as to be a touch meaningless in modern machines :)...
Beth:)
---
* That is, if such a thing exists...but it does have a web browser, so I'm
guessing someone somewhere's invented such a thing...after all, the fudging
SNES had a mouse, which I've never quite understood why...hehehe :)...
Beth
> Randy typed:
> > I remember when they said the same thing about
> > 32-bit x86 systems :-) ("the 80386 is not supposed
> > to be for SOHO users...").
>
> [ ] You know how much cooling a 64-bit x86 processor needs...
Yup, just like all those Amstrad PCs that used to melt the hard drive
because they lacked a fan...or, at least, that's what IBM spread as a FUD
rumour when they realised that people were impressed with a silent PC and
reduced power supply needs (because it used the power supply on the monitor
instead:)...if you want more details - and, also, to be enlightened of the
true face of Microsoft - go check out the "Halloween documents" on
www.opensource.org ...
Basically, once you've been round half as long as Randy has, you'll realise
to take any such statement as "640k is enough for anyone", "there will only
be five computers in the world, as they'll be aircraft-hangar-sized and
ludicrously expensive", "CPUs won't pass the 100MHz mark because we're
reaching the limits that sillicon can take", etc., etc. with not just a
pinch of salt but an entire salt cellar worth of the stuff...hehehe :)...
If the cooling required is too much, they'll work on designs that use lower
voltages, add more miniture fans that are more directed and closer to the
chip, make the fans more productive in their cooling ability (i.e. make them
spin faster :), re-organise the design of the chip (say, make it bigger -
spreading the heat - and include two smaller interdependent chips in that
space, capable of running with less heat problems but, together, capable of
delivering the goods required...exploiting some parallelism to lower the
heat:)...or, if all else fails, they'll make little fridge units to house
the chips inside (sounds unlikely? Well, it really wasn't all that long ago
when people would have said the same about sticking a fan on top of the chip
itself...they haven't really been doing that for all that long on
microprocessors, you know :)...and, if necessary, a combination of all of
the above and maybe some extra ideas on top of that too...
Beth:)
ccr...@crayne.org
at 02:56 AM, "Beth" <BethS...@hotmail.NOSPICEDHAM.com> said:
:thankfully, IBM had that bizarre thing
:with putting a forward slash into the name (what was that all about? OS/2
:as well...anyone?:)
This started back in 1965, when IBM announced the "360" family of
machines. Hardware designations were of the form <hardware family>/<model
number>, and operating systems were <os name>/<hardware family name>. The
original IBM-PC did not follow this pattern, because IBM envisioned it as
competing with the Apple ][ in the home market, rather than as something
which might be used in business.
When IBM finally recognized the business potential, they decided to use
the same naming conventions as their larger systems. Thus, they
concurrently had OS/370 running on the 370/xxx mainframes, OS/400 running
on the AS/400 minicomputers, and OS/2 running on the PS/2.
-- Chuck Crayne
-----------------------------------------------------------
ccr...@crayne.org
-----------------------------------------------------------
Martin Kiewitz
<BethS...@hotmail.NOSPICEDHAM.com> wrote:
> > [ ] You know how much cooling a 64-bit x86 processor needs...
> Yup, just like all those Amstrad PCs that used to melt the hard drive
> because they lacked a fan...or, at least, that's what IBM spread as a FUD
FUD ?
That's what Intel will tell you theirselves. -> Ask them.
> Basically, once you've been round half as long as Randy has, you'll realise
> to take any such statement as "640k is enough for anyone", "there will only
I don't see any use for 64-bits for Office or something. Really.
I see use in using that much power for games. But then, I wouldn't
choose something built on x86 architecture.
I would buy myself a PS2. And that's what I did.
> If the cooling required is too much, they'll work on designs that use lower
> voltages, add more miniture fans that are more directed and closer to the
That's for sure. The current fact is, that no one is able to use a
64-bit x86 processor at his normal home. Things will change. Anyway I
don't see any reason for it, but games. And for games there are better
and cheaper possibilites.
> heat:)...or, if all else fails, they'll make little fridge units to house
> the chips inside (sounds unlikely? Well, it really wasn't all that long ago
lol
But actually you are right :-))
> when people would have said the same about sticking a fan on top of the chip
> itself...they haven't really been doing that for all that long on
> microprocessors, you know :)...and, if necessary, a combination of all of
> the above and maybe some extra ideas on top of that too...
Anyway, a normal "user" wouldn't need it. Really.
It's like having a Porsche to drive to McDonalds. Imho a waste.
And...like I said...Intel said "64-bit are not meant for SOHO users".
That's what *they* are saying. So don't call it FUD, because they
would FUD themselves.
cu, Kiewitz
--
Phil Carmody
I am a RISC programmer 99% of the time. Please _don't_ trust my x86 code
just because it looks neat.
Only trust it when either
a) one of the many superb x86 programmers on this list tweaks it and
makes it better (in which case trust their version instead!)
or
b) you try it yourself and RDTSC tells you it runs as fast as you'd
hope.
But drop it like a stone if someone comes up with a cogent reason why
it's rubbish.
Phil
Jerry Coffin
BethS...@hotmail.NOSPICEDHAM.com says...
[ ... ]
> Randy!!! Surely you're not that much of a recluse? PS/2 is IBM, PS2 is
> Sony...he means the new PlayStation...thankfully, IBM had that bizarre thing
> with putting a forward slash into the name (what was that all about? OS/2 as
> well...anyone?:) so it is possible to tell them apart...although, I wonder
> how many complete newbies have heard the phrase "PS/2 mouse" and then tried
> to stick a mouse designed for the PS2 * into the back of their PC...hehehe
>:)
The might easily work -- the PS2 has a couple of USB ports, so with
the right driver it would have no real difficulty using a perfectly
normal mouse from a PC. It's also got a FireWire port, so using hard
drives (and such) is really only a driver away as well. OTOH, the
normal game controller ports on a PS2 are completely different.
That's a decided contrast to the Xbox -- though they have different
plugs on them, its game controller ports ARE normal USB. It might
almost be interesting to graft a normal USB plug onto an Xbox game
controller and plug it into a PC...
Randall Hyde
Martin Kiewitz wrote in message ...
><BethS...@hotmail.NOSPICEDHAM.com> wrote:
>
>
>That's for sure. The current fact is, that no one is able to use a
>64-bit x86 processor at his normal home. Things will change. Anyway I
>don't see any reason for it, but games. And for games there are better
>and cheaper possibilites.
I do a lot of video editing and I see a big need for 64 bits at home
(and yes, my video editing is a hobby, not a professional endeavor).
I'm talking both the ability to work with large blocks of integer
data *and* the ability to address more than 4GB of main memory.
It would be totally cool to put 32 GB of RAM in my machine so I
could edit an hour's worth of Digital Video (DV format) all in RAM.
And as long as the general purpose integer registers are 64 bits,
I'd like 256-bit MMX registers while we're at it. I'd also like a CPU
that supports 16x4->16 32-bit matrix multiplies too. Yes, this stuff
would be great for games, but it's great for other multimedia apps
too. When I can render photo-realistic 3D objects in real time and
I can apply video filters in real time, along with audio filters (in a
multi-track environment), and still have processing power left over
to run several software-based MIDI synthesizers concurrently,
I'll be happy.
Now you might think that the average person doesn't need this
kind of power. But keep in mind that the stuff found only on
high-end multimedia development machines of yesterday is
built into today's browsers and many web pages require this
capability to display properly.
Particularly as the public begins to accept and require technologies
like voice recognition, you're going to see the need for much more
powerful (i.e., 64-bit and greater) CPUs.
>
>Anyway, a normal "user" wouldn't need it. Really.
>It's like having a Porsche to drive to McDonalds. Imho a waste.
>
>And...like I said...Intel said "64-bit are not meant for SOHO users".
>That's what *they* are saying. So don't call it FUD, because they
>would FUD themselves.
*They* always say this when a new technology comes out
in order to justify high prices. However, nobody believes *them*
anymore, so one only looks foolish when parroting such a line.
Randy Hyde
Randall Hyde
Beth wrote in message ...
arcania.
>
>Randy!!! Surely you're not that much of a recluse? PS/2 is IBM, PS2 is
>Sony...he means the new PlayStation...thankfully, IBM had that bizarre
thing
>with putting a forward slash into the name (what was that all about? OS/2
as
>well...anyone?:) so it is possible to tell them apart...although, I wonder
>how many complete newbies have heard the phrase "PS/2 mouse" and then tried
>to stick a mouse designed for the PS2 * into the back of their PC...hehehe
Actually, I *am* that much of a recluse.
I got burned out on computer games way back in
1975 when I found myself playing a "Star Trek" game
until 3:00am the day before an 8:00am final exam
in freshman Physics.
Since I don't watch too much commercial television
(Sci-Fi channel is about it, and not much of that), I don't
get bombarded with enough Sony ads to automatically
associate PS2 with Playstation. We actually have a
Playstation around the house (for the nieces and nephews
to play with), but we've always called it the "Playstation"
and I'm not into the latest and greatest gaming gear,
so I never got used to the PS2 moniker. You'll have
to forgive me. It's obvious once it was mentioned.
Randy Hyde
Jerry Coffin
[ ... ]
> When IBM finally recognized the business potential, they decided to use
> the same naming conventions as their larger systems. Thus, they
> concurrently had OS/370 running on the 370/xxx mainframes, OS/400 running
> on the AS/400 minicomputers, and OS/2 running on the PS/2.
These were not concurrent -- the IBM 370 series was obsolete well
before PS/2 or OS/2 had been dreamt up at all. When OS/2 was new, I
believe the 308x series mainframes were still in production, and
these did NOT use the "slash" style designation -- e.g. the type of
IBM's line in the early '80s was the 3084 Model K.
Fortunately, since then IBM has come back to their senses -- the
current mainframes are the 390 series and unless I'm badly mistaken,
they do use the "slash" between the series and the model again.
ccr...@crayne.org
at 02:55 AM, Jerry Coffin <jco...@taeus.com> said:
:> When IBM finally recognized the business potential, they decided to use
:> the same naming conventions as their larger systems. Thus, they
:> concurrently had OS/370 running on the 370/xxx mainframes, OS/400 running
:> on the AS/400 minicomputers, and OS/2 running on the PS/2.
:These were not concurrent -- the IBM 370 series was obsolete well before
:PS/2 or OS/2 had been dreamt up at all.
I didn't want to get into this level of detail, but since you are a good
guy in my books, I will explain why I am right, and you are . . . well,
mostly right. :-).
To begin with, while you are correct that the 308x series was already in
production when the PS/2 was introduced, but the life cycle of mainframes
is pretty long, and there were a lot of IBM customers who were still
running 370s. In fact, I once called on a guy who was still making a good
living by selling time on his IBM 7094s, almost twenty years after the 360
family had been announced.
I still remember how surprised we all were, the day that the 308x machines
were announced, that they were not 380s, although all three digits were
there -- in a slightly different order. One consequence of this was that
OS/370 did not undergo a name change, nor did the architecture name.
:Fortunately, since then IBM has come back to their senses -- the
:current mainframes are the 390 series and unless I'm badly mistaken,
:they do use the "slash" between the series and the model again.
You are correct about the 390s, except that they are no longer in new
production, an thus -- by your definition -- obsolete.
The current mainframe machines are the zSeries machines, which -- among
other enhancements -- have 64 bit registers.
Jerry Coffin
[ ... ]
> To begin with, while you are correct that the 308x series was already in
> production when the PS/2 was introduced, but the life cycle of mainframes
> is pretty long, and there were a lot of IBM customers who were still
> running 370s. In fact, I once called on a guy who was still making a good
> living by selling time on his IBM 7094s, almost twenty years after the 360
> family had been announced.
Interesting logic. Let's see: there are still WW2 Willies Jeeps in
use, and Daimler now owns the Jeep trademark, so Germany lost the war
because Daimler built Jeeps for the Allies concurrently with World
War 2.
Not to belabor the point, but I think this is stretching
"concurrent" to the point that it's rendered meaningless at best.
As to your other points: you're more or less correct -- I hadn't had
any reason to keep track of new developments at IBM in the last
couple of months, so I hadn't noticed the introduction of the
zSeries. OTOH, IBM still seems to consider at least some 390 series
machines current as well. E.g:
http://www-1.ibm.com/servers/s390/multiprise/
In fact, given the timeframes involved in mainframe acquisitions, it
may be open to question whether a single zSeries machine has been
delivered to a customer yet.
That's quite a bit different from the situation with OS/2 vs. the 370
series -- customers had been using 308x machines for over 5 years by
the time OS/2 1.0 came on the market.
Beth
> Beth wrote:
> [ ... ]
>
> > Randy!!! Surely you're not that much of a recluse? PS/2 is IBM, PS2 is
> > Sony...he means the new PlayStation...thankfully, IBM had that bizarre
thing
> > with putting a forward slash into the name (what was that all about?
OS/2 as
> > well...anyone?:) so it is possible to tell them apart...although, I
wonder
> > how many complete newbies have heard the phrase "PS/2 mouse" and then
tried
> > to stick a mouse designed for the PS2 * into the back of their
PC...hehehe
> >:)
>
> The might easily work -- the PS2 has a couple of USB ports, so with
> the right driver it would have no real difficulty using a perfectly
> normal mouse from a PC. It's also got a FireWire port, so using hard
> drives (and such) is really only a driver away as well.
Well, of course, I was only saying it for a touch of humour...but, hey, if
it can actually be made to work, that's cool...after all, even jokes have
their serious side, right? :)
> OTOH, the normal game controller ports on a PS2 are completely different.
On the other hand, I was talking about a PS2 mouse (i.e. PS2 connection :)
slotting into the back of a PC, so whether it's possible in reverse is a
touch irrelevent (though, very interesting...I ain't knocking what you said
there at all...didn't know that stuff and now I do...just wondering why you
said it, is all :)...on the other hand, I was talking about newbies not
spotting the slash and realising that there was a difference between PS2 and
PS/2, so is it reasonable to expect them to know they have to get some
drivers?
On the other hand, the point/joke was that people would get confused between
the two names, due to the near identical names (as Randy seems to have done
and if he can do it as an ASM guru, then anyone can...I was only really
pulling Randy's leg in the other post, as if you're not interested in games
consoles, you might not even know that the PS2 exists...similarly, an
American wouldn't be "stupid" if they didn't know who Richard and Judy are
because Americans - I'm assuming - don't get to see this very famous British
married couple who presented some crap daytime TV show for years...everyone
in the UK knows who they are because they've been on our screens for years
and years...but, I'm guessing, they've probably done a big fat _zero_ on
American TV, so no-one over there knows who the hell they are...which, in
Richard and Judy's case, is very much a blessing...hehehe :) and I have
known newbies who've in the past got confused and actually gone out and
bought things mistakenly, which don't fit...though I make a joke, the
occasion I faintly remember where this happened to a friend was actually a
touch depressing because he'd actually saved up his pocket-money to get
himself a joystick or whatever (for a sort of birthday present to himself :)
and was really excited, except it didn't fit...felt really sorry for him
because he'd put so much into it all but all because of one simple perfectly
understandable mistake...luckily, he took it back - pretending it was
broken, rather than admit to the mistake - and swapped it for something
else, which he could use...anyway, if there is a moral to this story, I
suppose it's "don't ever be afraid to ask, be afraid of not knowing", as, if
the shop hadn't given refunds and replacements, he would have been stuck
with something useless and lost all that money he'd got with lots of hard
work...and the fact that he pretended it was broken when he took it back -
where I'm sure the shop would've swapped it anyway, if he'd explained the
mix-up - shows the root of the problem, I think...finding it too difficult
to swallow his pride and just ask someone who knows...he could've asked me,
he could've asked in the shop, he could've checked it out first in more than
one place but didn't...
Anyway...
> That's a decided contrast to the Xbox -- though they have different
> plugs on them, its game controller ports ARE normal USB. It might
> almost be interesting to graft a normal USB plug onto an Xbox game
> controller and plug it into a PC...
Now, there's interesting...would these game controllers also be primarily
manufactured by Microsoft, perchance? But, of course, the difference in
connection is all to do with trying to make it as simple as the Nintendo
connections (those Nintendo connections look like big lego bricks...talk
about simple...although, at the same time, they work perfectly so why not?
And, of course, the reason, no doubt, is the Nintendo people going that
extra mile in the design to make sure it causes absolutely no problems at
all...even for, say, eight year olds...and talking about design, Sony's
joypad - complete with the very good thinking and cool design of
colour-coding and shape-coding the buttons - is a classic...in fact, when
Sony entered the console market, I think they shook a few of the traditions
there because, of course, Sony came at it with experience gained from years
of creating "dumbo" products for the mass-public and they're approach was
totally different and quite, quite refreshing...they, unlike soooo many in
the computing industry, realised that it's consumer first _then_ technology
second...it's so easy to get caught up in the finer points of the technical
aspects that it is sometimes hard to remember to just take a step back and
think over whether it actually makes sense...yes, it may be fast, it may be
clever, it may be a technical wonder...but if it isn't usable by most
average people, you probably won't be able to sell many of them and it'll
head the way of the dodo :)...it would have nothing to do with Microsoft
making sure, despite being the same thing as manufactured by innumerable
companies, that their manufacture is pushed to the front and, of course,
even if other companies do make XBox controllers, it'll be a market with
much less competition as it's specific to one console rather than millions
of PCs and Macs and stuff...
<RANT POINTLESS="YES" VERBOSE="YES">
I have to give it to Bill, he's a genius at making money...the one thing
I've yet to work out, though, is why? I can understand why us poor people
would try to better ourselves and make some more money...but why _him_ in
_his_ situation? He's obviously well beyond that needed to get by in
life...he's even well beyond enough to live the "high life"...he could even
plough millions into charities and still not even notice anything's
gone...he could blow millions on something completely silly and frivolous
and recoop it all in interest...I'm not saying he shouldn't do it
necessarily (though I do question that :), what I'm confused is the reason
behind it...as daft as it sounds when we're talking about making billions
without breaking a sweat, but what's actually in it for him to keep up this
bullcrap? Or, in other words, hasn't he seen yet that money is great and
powerful but that there are other things besides it? All it's buying him now
is more and more resentment from everyone and court cases and more
responsibility from all the extra work and greater personal pressures on
himself...that in focusing on money making, he's probably using up time he
should be spending on his personal life (I mean, he must have a personal
life, surely? You know, friends and family, dreams of learning to play the
guitar or how to paint or whatever interests him :)...what I mean is, what
exactly is the eventual goal here? I mean, if he's really clever, he can
keep on going almost indefinitely _but_ why bother? I mean, money's
certainly no object anymore for him so what additional things could he be
getting from this? Pride, perhaps? Or maybe he still feels he hasn't proved
himself (but, in which case, he must have serious emotional issues because
he's obviously shown he's more than capable of achieving success :)...now,
again, maybe people who've been brought up on the infallibility of
capitalism will think me entirely mad, but what is the final goal here? I've
been thinking it through and there doesn't seem to be one...so, ironically,
even Big Bill with dollar bills bursting out of his ears is, essentially,
just a slave to money...no matter how much he gets, there's always more to
be got...it reminds me of two kids trying to better each other: "My dad's
got loads", "So? My dad's got thousands", "So? Mine's got millions", "So?
Mine's got billions", "Trillions", "Zillions", "Millions of Zillions",
"Billions of Zillions", "He's got infinity"...and then we get the answer
that says it all: "So? Mine's got infinity plus one"...infinity plus
one...just think it over :)...
</RANT>
Beth:)
Jerry Coffin
BethS...@hotmail.NOSPICEDHAM.com says...
[ ... ]
> Well, of course, I was only saying it for a touch of humour...but, hey, if
> it can actually be made to work, that's cool...after all, even jokes have
> their serious side, right? :)
I'm reminded of one of an American who wrote politically oriented
humor. He talked about a time back in the 1960's that a friend of
his noted that (at that time) white people didn't mind standing in
lines with black people -- it was only when they had to sit down that
the whites wanted to be segregated. Obviously the way to integrate
people was to simply make them stand up all the time.
Unfortunately, he never got to write about this even though he
thought it was hilarious: shortly before anything could be published,
one of the local libraries had decided that in an effort to
integrate, they'd (of course) pull out all the chairs and make
everybody stand up all the time... <G>
[ ... ]
> I have to give it to Bill, he's a genius at making money...the one thing
> I've yet to work out, though, is why? I can understand why us poor people
> would try to better ourselves and make some more money...but why _him_ in
> _his_ situation?
I don't think for him it's ever really been about making money. It's
about doing cool stuff, using (of course) his own definition of
what's cool. He's also always pushed toward the idea of things not
just being about technology, but about technology being accessible
and useful to everyday people. You can argue about the degree to
which he's succeeded at that (and people certainly have argued
endlessly about it) but I don't see any real way to question that
it's his intent.
I'd bet almost anything that the motivation behind the Xbox was not
somebody saying "this is a large and lucrative market" -- it was Bill
playing some existing video game, and saying something like "We can
do something a LOT better than this!"
Chrissie
anyone writing
a DirectNT couterpart library, that could interface with all graphics cards
and utilize
their different abilities for other stuff then graphics.
When you dont play games, most of thoose chips highspeed memory and GPU
arent beeing used.
Maybe, by programming it directly it would be possible to use it for some
highspeed calculations
or something (whatever). I have no idee if this would be useful, just a wild
though, but since most
asm programmers seem to break the rules all the time with great results to
show for it,
maybe such a library, if well written could boost performance on any future
PC.
I realize that many of you allready thought of it. Maybe some of you could
sheed a light upon the
usefulness of such a library ??? Or the lack of it ?
--Chrissie.
A try except finally end programmer.
Beth
> This "insane" thread got me thinking (about a lot of stuff), but how about
> anyone writing a DirectNT couterpart library, that could interface with
> all graphics cards and utilize their different abilities for other stuff
then
> graphics.
>
> When you dont play games, most of thoose chips highspeed memory and GPU
> arent beeing used.
> Maybe, by programming it directly it would be possible to use it for some
> highspeed calculations or something (whatever). I have no idee if this
would be
> the rules all the time with great results to show for it, maybe such a
library, if
> well written could boost performance on any future PC.
>
> I realize that many of you allready thought of it. Maybe some of you could
> sheed a light upon the usefulness of such a library ??? Or the lack of it
?
Hi, Chrissie...
The essential problem is that those "polygon accelerator" * cards are too
specific...there may be one or two tricks you could think up exploiting the
card to do a job different to its design but, essentially, the problem is
that those cards accelerate very specific algorithms, calculate specific
geometries and stuff...this was why as part of my earlier "they are not 3D
accelerators, they are polygon accelerators" I was kicking up such a fuss
about the things being so bloody specific...if they sat as a large array of
programmable maths co-processors (which they are essentially but can't be
accessed as such in anything but polygon accelerating) then it would open up
soooo much scope for amazing tricks...especially when, sitting as most do on
an AGP slot, they can go about their business in parallel with the CPU at a
very reasonable speed...that's part of why I'm so annoyed with the
things...it's just too much power sitting idle the majority of the time and,
thanks to an unhealthy obsession with polygons, it's next to impossible to
use them for anything else beyond that...if they just loosened it up a bit
so that all those large arrays of maths processors could be exploited for
other tasks, it would be sooo brilliant...plus, they might then be able to
aid in accelerating intersection calculations for ray-tracing and, perhaps,
actually _earn_ their "3D accelerator" (implying all types of 3D :) title
rather than accelerating Quake clones and thinking that's all there is
available out there...
Yup, probably many of us have thought of it but thanks to all this "black
boxes that do specific jobs are great!!" thinking, there's precious little
that could really be done...well, one or two things could be exploited but
it's not general enough to be generally exploitable in most situations...an
array of programmable FPUs, though, would be an entirely different
story...anything maths (or that could be twisted to be maths :) would then
be accelerated into the stratosphere (because there is a lot of potentiaal
power in those cards :)...unfortunately, though, from my experience in R&D,
there's an unhealthy obsession with "black boxes" for everything...so,
essentially, we pay the price because the engineers don't want to talk to
each other, as the chief benefit of why they obsess over "black boxes" is
that the modular design it brings means they can proceed in their own little
worlds and just present nice black box interfaces to the outside world and
never have to communicate with anyone about anything...that's the trouble
with humans, anything for the easy life...
Beth:)
---
* I will continue to refuse to call them "3D accelerators" until they
actually all cover a reasonable amount of all possible methods of doing
3D...primarily, these things accelerate producing texture-mapped
polygons...yup, some add bump mapping and complex lighting to make it seem
less "polygonised" but they do so on polygons, in a polygonised manner...I
mean, by this convention of naming, maybe keyboards should be called
"thought accelerators" or something instead...after all, it sounds clever
and has a tiny, tiny element of twisted truth behind it, which is all the
criteria they seemed to use in naming them "3D accelerators"...
Chrissie
bitRAKE
> specific...there may be one or two tricks you could think up exploiting the
> card to do a job different to its design but, essentially, the problem is
> that those cards accelerate very specific algorithms, calculate specific
> geometries and stuff...this was why as part of my earlier "they are not 3D
> accelerators, they are polygon accelerators"
Hello Beth,
My GeForce3 card is quite programable! It does the polygon stuff, too. ;)
bitRAKE
Beth
> > The essential problem is that those "polygon accelerator" * cards are
too
> > specific...there may be one or two tricks you could think up exploiting
the
> > card to do a job different to its design but, essentially, the problem
is
> > that those cards accelerate very specific algorithms, calculate specific
> > geometries and stuff...this was why as part of my earlier "they are not
3D
> > accelerators, they are polygon accelerators"
>
> Hello Beth,
Hi:)
> My GeForce3 card is quite programmable! It does the polygon stuff, too.
;)
Yes, of course; But there's many cards that aren't...it's only a more
current trend to add more flexibility to them (probably because they
realised how it was all a touch needlessly limiting...and there's also not a
proper standard for what's available from these cards (and using DirectX or
whatever almost completely locks out the possibility of exploiting the cards
differently to intended) and so a "requires a GeForce3 v1.235 3D accelerator
card" stuck on the side of your spreadsheet program would hardly be
acceptable...it's bad enough that they keep pushing minimum requirements
through the roof (and, often, just lie about them..the amount of programs
I've seen that say one thing - like "requires 64Mb" - but run quite
comfortably, without problems, on a 32Mb machine is both stupid and
worrying), without making it worse by requiring specific models and makes of
cards to run something...so, because it's not universal, as brilliant as the
particular features of that card might be, except for stuff you want to
write for yourself to work only on your own machine, it's nice those things
are there (and, maybe, in time, others will follow suit and the ones that
are less programmable will go the way of the dodo :) but it can't be used as
a solution in general until a majority of cards can offer at least a base
set of similar features...
That is, we've got a case of "technology drag"; Because not everyone has a
card capable of this stuff, because the de facto standards for such cards
have not yet matured to be reliably known to be on most cards...then we have
"technology drag", where it may indeed be possible for the very latest card
to calculate pi to the millionth billionth decimal place in a blink of an
eye but this can't be universally used until they all get that good (or
close to it) and, to be useful and applicable to enough people to be worth
the effort, we have to work with the lowest spec of card that is
common...thanks to this "technology drag", technology can race away into the
future but, except for strict "black box" use as intended, this doesn't help
the overall situation much...I suppose it would be like having an athlete
that's so far ahead of the game, no-one comes anywhere close to them and, in
fact, this ends up boring everyone because there's no more tension and
excitement about who will win...the race is already won because our
"SuperAthlete" is so good she can run around the track in half the time
anyone else can and no-one comes anywhere near her anywhere in the race...it
might seem clever to race ahead like that but, in fact, it just makes things
boring and a touch pointless for everyone else but our "SuperAthlete" to
enter the race...or, similarly, being "ahead of your time" in art or science
or something, tends to be detrimental, as others can't understand or use
what you do...in other words, it _is_ possible to be too big for your own
boots...
And, anyway, I wasn't disputing that such cards are programmable...I was
saying that, with all these things considered, they aren't all that useful
for wider use in helping to accelerate spreadsheet calculations and such, as
the original post was asking about...put it another way, most video
recorders are entirely "programmable" too but _only_ in the context of
recording TV programmes, it would just be impractical (or even impossible)
to write a single piece of software to run on all video recorders that could
let you play a simple Mario game on them (despite all of them having buttons
to send control messages and the capability of controlling audio and video
on the screen and having a CPU stuck inside them...they are "programmable",
sure, but not in a useful way :)...
Beth:)
Chewy509
"Beth" wrote in message...
> ...it's bad enough that they keep pushing minimum requirements
> through the roof (and, often, just lie about them..the amount of programs
> I've seen that say one thing - like "requires 64Mb" - but run quite
> comfortably, without problems, on a 32Mb machine is both stupid and
> worrying),
I remember a while back, a game called Seal Team, which according to the box
required a i386DX/25, which I found ran quite well with a i286/16 + i287
with a 2MB EMS Card. And more recently Windows 2K Pro, which I had running
on a P60 w/32MB RAM, or WinXPPro on a P233MMX w/64MB Ram. So it seems even
Software companies have a hand in the upgrade market.
Chewy509...
PS. I guess with W2K+, MS understood not to quote the minimum requirements
too low, (if anyone remembers running Win95 on a i386 w/4MB Ram)
Beth
> Beth wrote in message...
> > ...it's bad enough that they keep pushing minimum requirements
> > through the roof (and, often, just lie about them..the amount of
programs
> > I've seen that say one thing - like "requires 64Mb" - but run quite
> > comfortably, without problems, on a 32Mb machine is both stupid and
> > worrying),
>
> I remember a while back, a game called Seal Team, which according to the
box
> required a i386DX/25, which I found ran quite well with a i286/16 + i287
> with a 2MB EMS Card. And more recently Windows 2K Pro, which I had running
> on a P60 w/32MB RAM, or WinXPPro on a P233MMX w/64MB Ram.
Yup; It tends to be quite miraclous to be able to find a minimum spec. that
actually tells the truth about the requirements of programs, rather than the
other way round, as you might expect it to be...
> So it seems even Software companies have a hand in the upgrade market.
Possibly; But I think the reason is a touch simpler than that...for
instance, if I told you that I'd just written a really cool 3D game that has
minimum specs of a P200 w/16Mb RAM, that creates expectations in your mind
about what sort of quality of 3D game to expect, even though you have no
idea what the game is about...it's all about the user's expectations...it's
entirely possible someone might discover some radical new way to do 3D
graphics or something that halves the minimum specs. required for a
game...but if users see the minimum specs being lower, they'll tend to
assume the game isn't as good (even though, in fact, the clever programmers
have managed to squeeze twice as good a game into half the space and speed
:)...thus, they tend to want to orientate the specs to be ever-increasing to
give the impression the games are getting better and better (whether they
actually are or not :)...
An even more mundane reason, as your P.S. points to, is that by delibrately
knocking the minimum spec. higher than the true minimum, it ensures that
there won't be any nasty problems with certain hardware configurations and
that, if it gets demonstrated in a shop window or someone sees the
game/program running on a friend's PC, then it'll be running without any
hitches or slowdowns, which might help to shift a few extra copies...
Even more mundane again, they might, indeed, have an interest in the upgrade
market, in that if more people have better PCs, this makes developing their
next games and stuff a whole lot simpler (I mean, if you can force everyone
to upgrade to 1Gb RAM then why bother being clever and trying to compress
the graphics and such? Just draw them as BMPs and stick them on the
DVD...why bother making an effort when you can just convince everyone to
upgrade and potentially save a packet on development costs :)...
Plus, there's stupid reasons like not wanting to spend cash on a bunch of
PCs of different specs to see how they run...rather, have one or two
"standard" PCs to measure by and make a guess...much cheaper than having
every piece of hardware ever constructed (although, in MS's case, as they
make OSes, this isn't an acceptable thing, really...but, then, when has
anything like trying to do things right ever stopped MS from arsing
everything up :)...
Randall Hyde
Beth wrote in message ...
Most likely, the software developers are too lazy
to figure out the minimum machine it will run on
and they just state the specifications of their
development or test machines.
Randy Hyde
p.s. It could also be that they simply don't have
access to lower performance machines in order
to run their tests, so they specify the lowest
performing machine they own.
LocalFolk
that shrink-wrapped a product or two, I can relate.
Yes. Not having 75 different PCs in your test
bed is a major reason, as is the laziness factor.
Or, more approproately, prioritization of effort.
But the big reason for not finding an absolute
minimum configuration and printing that as the
requirement, is the all-too-common occurance
of someone with those exact specs who happens
to be running a web server or a company database
on the same machine as the one you spec'ed.
So to avoid requirements like:
"PII 400MHz, 64M, 2G (75M free), unless you are
running a webserver concurrently, in which case
it's PIII 600MHz,128M, 4G (100M free) or if you
are running a company database, in which case
it's PIII 450MHz, 64M, 20G (2G free), etc, etc,
yadda yadda yadda..."
To avoid that, the requirements guys usually leave
a little slack in the requirement. The reason is
not, I assure you, to push PCs for Uncle Bill.
-uncle bear