Planeshift and 64 bit Processor

[Grakrim]

Anyway, 64-bit is too much for any power user (let alone an average user).  64-bit processors will only be useful for people doing complex calculations.  I would imagine a 64-bit computer would have more address space, but that\'s not immediatly useful, yet (I\'m not going to pull a Gates, though.)  Of course, people will end up buying 64-bit processors anyway, thinking its more powerful.  Its really equivlent to the x87 Math Coprocessors in the pre-Pentium days, although with less impact on the average user.

64-bits is probably the future, but they\'re more difficult to produce, and, correct me if I\'m wrong, more expensive at the moment.   Although, they also say encrypted instruction sets are the future; and I doubt that will happen any time soon.

And to kbilik: It couldn\'t be an option, it would have to be rebuilt.

[kbilik]

Hmm... what about bandwidth problems with 32 bit processors? As processors get faster and faster at sending information, the 32 bit will eventually hit a bottleneck.

If you want to send a stable stream of high quality data from processor through RAM and so on, 64 bit will be better in the long run. I heard that 64 bit also supports FAR more RAM than 32 bit.

Look at this 25 gigaflop chip for example:
Here

In theory, its fast... but that can change if it is hooked up to a 32 bit processor and then it slows down.

[Fish]

I am not getting whole metaphor with the car engine however in the past 20 years I?ve seen a lot of these arguments before.

The first computer ever used was a PDP8.  A core memory computer from the Digital Corp. Then a Atari 400, Atari 800, IBM PC, 286, 386, 486, Pentium, Pentium 2, K6, and on, and on...

Grakim

Anyway, 64-bit is too much for any power user (let alone an average user).

If I had a nickel for every time I?ve read this one somewhere I would have a $1.50
The point is faster smaller cheaper is always better.  This is especially true in video games.  As an example if you can put 1000 Pentiums in one chip and stick on a wristwatch, maybe 100 years from now, there would be a lot of useful applications for it.

I think the other misconceptions is how base technology like microprocessor upgrades are integrated into finished product.  First of all if you?re not working in assembly language it has little or no impact.  This reminds me of the math coprocessor debate.  It ends up being the math coprocessor was simulated, at greatly reduce speed, in the C language thus allowing software to run in the two environments.  The same thing will probably happen in 64-bit processors.

So until the developers have a 64-bit processor and compilers to use it I seriously doubt they will dive into assembly language to make it happen.

[Grakrim]

Originally posted by kbilik
Hmm... what about bandwidth problems with 32 bit processors? As processors get faster and faster at sending information, the 32 bit will eventually hit a bottleneck.

Eventually is the keyword.  Effective processor speed is restricted by bandwidth; but it mainly has to with the size of data the processor can natively deal with.  So it takes more processing for a 32-bit processor to deal with 64-bit numbers.  Processor speed hasn\'t really hit a point where 64-bit bandwidth is nesscary, and the days of Moore\'s law are over, so we\'ll have to see how long it will take.  64-bit is faster when 64-bit or higher values are dealt with, but that\'s a rare occurance, outside of engineering and mathematics.

If you want to send a stable stream of high quality data from processor through RAM and so on, 64 bit will be better in the long run. I heard that 64 bit also supports FAR more RAM than 32 bit.

Hence I said more addressing space

[Grakrim]

Originally posted by Fish
If I had a nickel for every time I?ve read this one somewhere I would have a $1.50
The point is faster smaller cheaper is always better.  This is especially true in video games.  As an example if you can put 1000 Pentiums in one chip and stick on a wristwatch, maybe 100 years from now, there would be a lot of useful applications for it.

Yeah, why do you think the Z80 is still so popular?

I think the other misconceptions is how base technology like microprocessor upgrades are integrated into finished product.  First of all if you?re not working in assembly language it has little or no impact.  This reminds me of the math coprocessor debate.  It ends up being the math coprocessor was simulated, at greatly reduce speed, in the C language thus allowing software to run in the two environments.  The same thing will probably happen in 64-bit processors.

So until the developers have a 64-bit processor and compilers to use it I seriously doubt they will dive into assembly language to make it happen.

Well, that\'s rather flawed.  Compilers have evolved alot since the \'80s.  Compilers now produce almost as fast code as hand-written assembly, making assembly unnessicary for coding on established platforms (that is, anything that\'s ever been used by mankind, the x86 family espeically)  Although, new compilers will have to be written to take advantage of the 64-bit processors, but that\'s a given.

Also, just before Math Coprocessors were built-in (with the 486DX and Pentiums), wasn\'t the final solution to detect wheter the 87 was there, and emulate if not?  Of course, in the 386 and before, it was more natural to assume its not there.

One question though, what\'s the designation for the Intel\'s new IA-64? Are they going to continue with the x86 tradition, or are they starting anew?

[Fish]

?Yeah, why do you think the Z80 is still so popular??  I hate to break this to you it?s not.  It?s a dead architecture.  It died 10 years ago and for good reason, there are way better alternatives.

?Compilers now produce almost as fast code as hand-written assembly? that really depends on a lot of things.  First of all who is writing the assembly code.  Second of all what algorithm you?re shooting for.  However it kinda proves my point.  If you have a new processor class it takes awhile for people to write the compilers that fully utilized that class.  Very few people will go to assembly to make it happen.  So that means no compiler libraries that utilize 64-bit = no benefit until there is one.  Then you would think you just use the 64-bit library and it works.  That almost never happens.

As for the new 64-bit chip being in the x86 family.  I personally think they?ve beat the x86 horse into the ground.  If it weren?t for %$^&$$#^ legacy issues it would have been dead long ago.  However I?ll make a $10 bet with you that it will be.  That?s because their capitalists.  That?s one bet I?d Like to lose.

[Grakrim]

Originally posted by Fish
?Yeah, why do you think the Z80 is still so popular??  I hate to break this to you it?s not.  It?s a dead architecture.  It died 10 years ago and for good reason, there are way better alternatives.

Actually, the Z80 is still in use.  It was used up to 2-3 years ago in the Game Boy/Game Boy Color (and there\'s even a Z80 in the GBA, in addition to the ARM7TDMI that makes up its core), and it is TI\'s processor of choice for graphing calculators.  Z80s are still good, cheap, low power-consumption processors when you can get by on 2MHz.

?Compilers now produce almost as fast code as hand-written assembly? that really depends on a lot of things. First of all who is writing the assembly code. Second of all what algorithm you?re shooting for. However it kinda proves my point. If you have a new processor class it takes awhile for people to write the compilers that fully utilized that class. Very few people will go to assembly to make it happen. So that means no compiler libraries that utilize 64-bit = no benefit until there is one. Then you would think you just use the 64-bit library and it works. That almost never happens.

Actually, if the compiler is coded correctly, any C library will intristically support 64 bit.  This is because the actual size of C variables (other than chars and the floating point ones) are undefined.  Which, naturally, is a headache to portable programmers...  But assemblers and compilers are always the first applications written for a new processor, and usually by the company that designs the chip (for instance, those Intel compilers, I wouldn\'t touch them, but I\'ve heard they\'re hella fast).

[Fish]

The Game Boy was introduced in 1989.  That is significantly longer than 10 years before now in 2003.  From then on you chalk it up to legacy.  Legacy means they were stuck with the thing even though there was something better out there.

As far as the compilers being able to automatically detect new processor abilities(especially ones never anticipated) and then use them, this is simply not true.  For instance in linux C the 64-bit word is a special case that is rarely used.  It wasn?t true for the math coprocessor either, somebody actually had to write it.  This is one reason software always lags hardware by six months to a year.  Somehow graphics accelerators seem to dodge this bullet(due to a decent industry-standard) however in the case of the 64-bit microprocessor, I think not.

The point is don?t expect plane shift to be going 64 bit until the hardware and tools(that?s software tools) are in place.  However when it does it will be getting that advantage by recompiling it.  However history dictates just recompiling it doesn?t usually turn the trick.  It usually takes a fair bit of tweaking.

[Grakrim]

Originally posted by Fish
As far as the compilers being able to automatically detect new processor abilities(especially ones never anticipated) and then use them, this is simply not true.  For instance in linux C the 64-bit word is a special case that is rarely used.  It wasn?t true for the math coprocessor either, somebody actually had to write it.  This is one reason software always lags hardware by six months to a year.  Somehow graphics accelerators seem to dodge this bullet(due to a decent industry-standard) however in the case of the 64-bit microprocessor, I think not.

I never said they automatically detect it, I just said that, by standard, the number of bits is undefined, or rather, implementation defined.  So, on a 64-bit processor, you\'d imagine a 64-bit integer would be made availble to the developer; but its not automatic, that\'s why I said if the compiler was properly coded.  To restate, the compiler implementor determines the built-in type size, and its usually (and should be) implemented to take the most advantage of the hardware at hand.

The point is don?t expect plane shift to be going 64 bit until the hardware and tools(that?s software tools) are in place.  However when it does it will be getting that advantage by recompiling it.  However history dictates just recompiling it doesn?t usually turn the trick.  It usually takes a fair bit of tweaking.

Naturally.

[Fish]

Changing the size based on processor type and size is something C specifically does not do.  Why?

Supposing you write a program based on a integer being 16-bit word.  All of a sudden in compiler version its 32-bit.  This would wreak all kinds of havoc.  Word size is a tightly controlled parameter.  So by running the 64-bit processor you don?t automatically get more precision.  It might be faster for other reasons, say it takes two 32-bit words rather than one, but then the program itself would have to be optimized along with the compiler to take advantage of this.  This a reason why going to a 64-bit processor is a big deal.  It?s not cut and dry there is no? if the compiler was properly coded? they?re not.

[Grakrim]

Originally posted by Fish
It?s not cut and dry there is no? if the compiler was properly coded? they?re not.

I think you misunderstand me... I mean, if the compiler is properly coded for that particular processor.  Its not the sort of adjustment that\'s made at runtime, its made at designtime.  I\'d imagine the Intel compilers for the IA-64 are properly coded.

[Zeonire]

I personally don\'t think they should change a thing about it. Look at the game we\'re talking about here:

1. It has ASTOUNDING graphics, especially for an mmorpg.
2. For a game like this that is in prealpha, its certainly fast enough. Lag here and there is only inexorable.

I think it\'s fine.

[Cha0s]

Just to make a few corrections and confirm a few things:

IBM and Apple co-released the first 64-bit computer several weeks ago. You can see it here: http://www.apple.com/powermac/

It supports up to 8 gigs of main RAM and, believe it or not with 1 GB of RAM (no display or speakers) you can get 1.6 gHz model for $2,820 (US dollars). The specs:
? 1.6GHz PowerPC G5
? 1GB DDR333 SDRAM (PC2700) - 2x512
? 160GB Serial ATA - 7200rpm
? ATI Radeon 9800 Pro
? SuperDrive (DVD-R/CD-RW)
? Apple Keyboard & Apple Mouse - U.S. English
? Mac OS X - U.S. English

This is ultimate graphics without all the RAM (that gets expensive... up to an additional $2,100 to get 4 GB; I don\'t want to get into the dual 2 gHz processor machine...memory for that...ouch).

Anyway, 64 bit processing is here and whether or not the PlaneShift devs choose to use it now, if PS hangs around for more than five years they will probably be forced to, as new software creates new demands for processors and eventually PlaneShift will become obsolete without more advanced software and graphics. It\'s the trend of computers and technology.

Example: I have some old Mac G3s in my basement. I can\'t play Warcraft 3 on them and they\'re barely five years old. I can play Warcraft 1 and 2, yea sure... But do I want to? No way, WC3 is a better game. If PlaneShift lives for an extended time and gets to the level of the Blizzard games (WC3 sold 1 million copies it\'s first weekend) then it will need to switch to 64 bit processing. Not now necessarily, maybe not in five years, but eventually.

[Xandria]

Cha0s, did you read the rest of this thread?  64-bit processing gives your processor an internal register size of 64 bits, rather than the 32 bits most today have.  All this means is that 1) you computer can do faster calculations on very long numbers, and 2) you have a larger maximum limit on RAM.

What does this mean for Planeshift?  NOTHING!!

I\'ve said this in every post, and I\'ll say it again here: stop bugging the PS devs about 64-bit and go bug the CrystalSpace devs about 64-bit, because if you want PS to run fastser on the new processors, CS is where it\'s going to happen.

OKAY?!?!

[Cha0s]

As I said in my post, PS will not need 64-bit support for quite a while, and I\'m not necessarily saying the PS devs should implement it themselves (I stated that badly, i was thinking that they should bug the CrystalSpace people to   ). As for what advantages a 64-bit processor has, I\'d like to draw attention to your second point. As games become more and more complex, there may end up being a lot to process and 64-bit power (and more RAM) would be useful then. Anyway, no need for now, but no one knows what the future holds. In five years virtual reality with smell, ambient sounds, touch, and 3d-surround vision may be commonplace creating a need for a better processor. No one knows for sure, but it could happen...