@KingKong

You still haven't made any point _why_ 48 bits is preferable to a 64 bit system. Yes it saves 16 bits per (full) pointer but raises problems in a number of areas, most of which can be included in compatibility.

Real computers use real world data and if you don't want to recreate that how would you efficiently read/write standard files?

And if we come back to the one "advantage" (that of saving 16 bits per (full) pointer) even that in itself isn't much of an advantage: 2^48 bytes is a lot of memory but in reach as real physical memory. But modern operating systems use virtual memory and limiting the virtual address space to 2^48 bytes makes a number of optimizations impossible (or at least much more expensive).
Furthermore one can create data structures with 48 bit indices (or even 32 bit if it is enough) into a 64 bit address space and then your single advantage fails completely.

My toy OS is already seriously limited by the current AMD64 limit of 2^48 bits virtual address space. But that limit can be lifted in hardware without creating any compatibility problems for software though the OS itself would need some polishing in the memory management parts.

@Kronos

I couldn't resist:
http://www.behoo-fastener.com/newsinfo-379.html

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@Trekiej

Quote:

I wish I could remember the site that talked about a 40 or 48 bit system.

I think that was here. Probably this very thread...

_________________
"Unix is supposed to fix that." -- Jay Miner

Why use less if you can have 64 bit? Would it be cheaper at a similar speed?

@megol

Quote:

megol wrote: @KingKong You still haven't made any point _why_ 48 bits is preferable to a 64 bit system. Yes it saves 16 bits per (full) pointer but raises problems in a number of areas, most of which can be included in compatibility. Real computers use real world data and if you don't want to recreate that how would you efficiently read/write standard files? And if we come back to the one "advantage" (that of saving 16 bits per (full) pointer) even that in itself isn't much of an advantage: 2^48 bytes is a lot of memory but in reach as real physical memory. But modern operating systems use virtual memory and limiting the virtual address space to 2^48 bytes makes a number of optimizations impossible (or at least much more expensive). Furthermore one can create data structures with 48 bit indices (or even 32 bit if it is enough) into a 64 bit address space and then your single advantage fails completely. My toy OS is already seriously limited by the current AMD64 limit of 2^48 bits virtual address space. But that limit can be lifted in hardware without creating any compatibility problems for software though the OS itself would need some polishing in the memory management parts.

Why do even bother to engage KingKong? By his own admission he's been an unemployed "IT pro" for years. His rantings show he's out of touch with much more than just the IT profession. It's quite obvious why he remains unemployed and out of touch with reality based on his comments. He obviously has some twisted obsession with the number 48. He's just another kook who drops in sporadically to stir things up. Stop encouraging him.

Last edited by ferrels on 16-Jun-2015 at 07:13 PM.

@olegil

Quote:

olegil wrote: @Kronos I couldn't resist: "http://www.behoo-fastener.com/newsinfo-379.html"

Yeah, KingKong lost a big bag of those a long time ago and he needs help finding them! LOL!

Last edited by ferrels on 16-Jun-2015 at 07:20 PM.

Quote:

Thorham wrote: Why use less if you can have 64 bit? Would it be cheaper at a similar speed?

Less than 64 bits "saves transistors and energy" as KingKong correctly stated. This means it's possible to be cheaper and/or faster at the same clock speed.

Advantages and disadvantages of 64 bit:

+ Practically unlimited address space is available (more directly addressable memory).
+ Work can sometimes be done twice as fast by working on data that is twice as wide.
+ Less registers are needed for data that would overflow 32 bit registers.
- More transistors are needed for wider paths, registers and DCache.
- More transistors are needed for more and larger instructions and ICache.
- Multiply, divide and shift in full 64 bits generally slows down the pipeline.
- More transistors are commonly used for better branch prediction and OoO execution to reduce the pipeline slow downs caused by 64 bit.
- Die shrinks are especially beneficial to offset the disadvantages of 64 bits but increase costs.

The 64 bit x86_64 benefited from ISA changes including 8 more registers which helped to push 64 bit CPU popularity. The speed advantages of the ISA change more than offset the speed disadvantages of a 64 bit processor. There are still a few tasks which are faster in 32 bit on the x86_64 due to better code density and cache utilization. Other processors commonly see a slow down with 64 bits. The AmigaOS is likely to be significantly slower with 64 bits due to smaller data types and less than natural alignment in structures. Some future programs may be faster though.

There are (semi-)modern processors with 48 (and 40 bit) hardware data types. They are processors for specific (usually embedded) uses like DSPs, micro-controllers and media processors. The purpose is usually to have extra precision for multiplication products (24x24=48 for example in 56k DSP). The extra precision is helpful with cumulative rounding errors and saturation math while having the speed possible with narrower data multiplications. Much of the high end market for DSPs has been replaced by more general purpose CPUs with SIMD processors. FPGA DSPs with limited data width multiply support have helped the popularity of low end DSPs with odd data type sizes though. Specialty processors are usually difficult to program and odd data type sizes make them more so. Compilers will have fewer problems and better support for 32 bit and 64 bit sizes which are better choices for general purpose computers. It's too bad good code density 32 bit CPUs like the 68k don't get more development love as they are easy to program, address enough memory for most general purpose computing needs and would use a lot less transistors than a modern x86_64. RIP 68k.

Last edited by matthey on 16-Jun-2015 at 08:50 PM.
Last edited by matthey on 16-Jun-2015 at 08:47 PM.

@matthey

You're failing to point out that all the major and minor chip designers waded through the same criteria that you've just listed here. And the pros of 64-bit computing far outweighed the cons. They chose to go with 64-bit designs for good reasons, one of which was to lessen the effects of obsolescence by skipping 48-bit architectures altogether.

Your argument that a narrower bus of 48-bits making for a cheaper and faster architecture at the same clock speed is also erroneous. You need to re-check your math. At the same clock speed on a 64-bit bus, I'm moving 16 more bits at a time than on a 48-bit bus at the same clock speed. It might be cheaper to design and operate, but certainly not faster. And as the saying goes, you get what you pay for.

You also point out that more transistors are somehow a con. It is not a con. The transistor count plays a huge, if not primary role, in just how powerful a CPU really is. And CPU dies continue to shrink roughly in accordance with Moore's Law so if anything, transistor count should be a neutral.
I for one, want more transistors on my CPU die. It's a hedge against obsolescence and I'll have a more powerful, albeit more power-hungry CPU.

And in your own words, 48-bit processors are "semi" modern. KinKong is arguing that a 48-bit architecture would be superior to what's already on the market which is patently false and/or delusional. The thread title is about making AmigaOS a leading OS. That will never happen using current technologies let alone on semi-modern architectures using embedded processors or by going with a 48-bit design of any kind.

If you're seeking credibility, you won't get it by arguing in support of KingKong.

Last edited by ferrels on 16-Jun-2015 at 10:16 PM.
Last edited by ferrels on 16-Jun-2015 at 09:56 PM.
Last edited by ferrels on 16-Jun-2015 at 09:55 PM.
Last edited by ferrels on 16-Jun-2015 at 09:50 PM.
Last edited by ferrels on 16-Jun-2015 at 09:49 PM.
Last edited by ferrels on 16-Jun-2015 at 09:43 PM.

@ferrels

AmigaOS should run on DNA computers.

http://computer.howstuffworks.com/dna-computer.htm

And we have decade to do it, by article estimate, if get product to market first, then AmigaOS has a chance to come back,



anyway I'm sure people run Linux on it instead

Last edited by NutsAboutAmiga on 16-Jun-2015 at 10:06 PM.
Last edited by NutsAboutAmiga on 16-Jun-2015 at 10:05 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Quote:

NutsAboutAmiga wrote: @ferrels AmigaOS should run on DNA computers. http://computer.howstuffworks.com/dna-computer.htm And we have decade to do it, by article estimate, if get product to market first, then AmigaOS has a chance to come back, anyway I'm sure people run Linux on it instead

Is that according to KingKong because I'd never make such a ridiculous statement?

And what happens when it gets a REAL virus or infection, instead of a cyber virus?
Then I have to have a health insurance plan for it and take it to a doctor. Not to mention the already nasty cyber viruses floating around! LOL!

Last edited by ferrels on 16-Jun-2015 at 10:20 PM.
Last edited by ferrels on 16-Jun-2015 at 10:19 PM.

@ferrels

Quote:

Is that according to KingKong because I'd never make such a ridiculous statement?

Ok, how about this what if we invented a time machine, we go back in time tell everyone what they did wrong

https://www.youtube.com/watch?v=i0cVdPHOIxw

Stephen Hawking thinks it might be possible

Last edited by NutsAboutAmiga on 16-Jun-2015 at 10:23 PM.
Last edited by NutsAboutAmiga on 16-Jun-2015 at 10:23 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Quote:

NutsAboutAmiga wrote: @ferrels Quote: Is that according to KingKong because I'd never make such a ridiculous statement? Ok, how about this what if we invented a time machine, we go back in time tell everyone what they did wrong

You've been hanging out with KingKong and his time machine again! I told you to wear your tin foil hat if you went visiting him again but you obviously ignored me! LOL!

@ferrels

Quote:

Told you to wear your tin foil hat

I have been trying to catch signals with large antennas on my head.





Just been looking at TV the Disc. Sci channel, with TV show "into the wormhole" with Morgan Freeman. Because everyone knows actors can do everything, and knows anything

Governor of California Arnold, is perfect example of this.

Last edited by NutsAboutAmiga on 16-Jun-2015 at 11:04 PM.
Last edited by NutsAboutAmiga on 16-Jun-2015 at 10:42 PM.
Last edited by NutsAboutAmiga on 16-Jun-2015 at 10:34 PM.
Last edited by NutsAboutAmiga on 16-Jun-2015 at 10:31 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@ferrels

So what if we installed AmigaOS in the Cosmos computer.
https://www.youtube.com/watch?v=atMuFCpxnUQ

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@KingKong well nothing, what we can do though is make it affordable by making it available ( native Aros or AOS4) for the Raspberry Pi 2, a 35$ computer.



Raspberry Pi 2 into A600 case running AEROS

_________________
So we must exercise ourselves in the things which bring happiness, since, if that be present, we have everything, and, if that be absent, all our actions are directed towards attaining it

Quote:

ferrels wrote: @matthey You're failing to point out that all the major and minor chip designers waded through the same criteria that you've just listed here. And the pros of 64-bit computing far outweighed the cons. They chose to go with 64-bit designs for good reasons, one of which was to lessen the effects of obsolescence by skipping 48-bit architectures altogether.

Oligopolies tend to produce a few quantity competitors and many specialty (niche) competitors. The "quantity" processor chip designers for general purpose computing have produced 32 bit and 64 bit designs. The x86_64 choice to go 64 bit was made easier by the advantages of ISA changes as I mentioned. The other "quantity" chip developer is ARM and they are mostly still 32 bit (they have a new and different 64 bit offering but it is not obvious that it will be successful). One could argue that 64 bit PPC and MIPS designs made it to "quantity" production but they have largely failed to maintain market share. Economies of scale with chip production are huge and have more to do with success than any ISA design. I see only one successful 64 bit ISA for "quantity" production which is x86_64. I would rather not judge based on one success in an oligopoly with many variables. However, your reasoning for skipping 48 bits is a good reason for general purpose computers. Memory and compilers being more efficient with powers of two are other good reasons. I'm not advocating making a modern 48 bit CPU. My point was that there is a significant cost to 64 bit processors for the extra bits.

Quote:

Your argument that a narrower bus of 48-bits making for a cheaper and faster architecture at the same clock speed is also erroneous. You need to re-check your math. At the same clock speed on a 64-bit bus, I'm moving 16 more bits at a time than on a 48-bit bus at the same clock speed. It might be cheaper to design and operate, but certainly not faster. And as the saying goes, you get what you pay for.

A 64 bit CPU is not always able to take advantage of a 64 bit bus or use a 64 bit operation. Actually, the majority of the time it doesn't. In many cases it ends up extending and aligning with padded zeros or ones which just plug up the DCache. The instructions are also bigger which plug up the ICache. Wide multiplication, division and shifting can slow down the pipeline. There is no free lunch. A 32 bit processor can be faster than a 64 bit processor, especially with limited number of transistors. Again, I'm not advocating for a 48 bit CPU because any gains in hardware savings would likely be lost due to more inefficiencies in software.

Quote:

You also point out that more transistors are somehow a con. It is not a con. The transistor count plays a huge, if not primary role, in just how powerful a CPU really is. And CPU dies continue to shrink roughly in accordance with Moore's Law so if anything, transistor count should be a neutral. I for one, want more transistors on my CPU die. It's a hedge against obsolescence and I'll have a more powerful, albeit more power-hungry CPU.

Unused transistors are more expense and potentially more heat. You must be the guy who has bought a dozen AmigaOS 4 computers and thinks the high price is "a hedge against obsolescence". Chip die shrinkage is becoming more difficult. There is no guarantee that Moore's Law will continue forever. I doubt those DNA processors will be available in time to take up the gap. Also, God may have trademarks on the design :P.

Quote:

And in your own words, 48-bit processors are "semi" modern. KingKong is arguing that a 48-bit architecture would be superior to what's already on the market which is patently false and/or delusional. The thread title is about making AmigaOS a leading OS. That will never happen using current technologies let alone on semi-modern architectures using embedded processors or by going with a 48-bit design of any kind.

My notation was meant to say semi-modern to modern. There are some modern specialty processors with 48 bit data types (most used for FPGAs which have limited width hardware multipliers). It would not be possible for the Amiga to make a comeback on that specialized or low end of processor so I agree with your point that they are not suitable (or desirable) for the CPU of an AmigaOS as a leading OS. The only cheap "quantity" designed CPUs that make sense for a future Amiga are x86_64 and maybe ARMv8 (although the AmigaOS would have to be competitive on an alien processor and may lose its identity). Niche processors are unlikely to be able to compete in price/performance, features vary widely by application and availability can stop at any time. IMO, the best chance of Amiga survival is for the Amiga owners to become vertically integrated as processor and chip designers. It is now possible (cheap enough) to own and develop specialized CPU and custom chip designs which allows a company to control it's own destiny. Many SoC designs have gone back to gfx near the CPU for performance and the Amiga only needs 3D. AmigaOS and CPU efficiencies could allow the Amiga to do more with less kind of like how a console is more efficient than a desktop. It would still be niche but could be cheap enough and high enough production to be successful. This was also the idea Dave Haynie commented as the best chance for the Amiga when he heard about the FPGA Amiga development.

Quote:

If you're seeking credibility, you won't get it by arguing in support of KingKong.

I used KingKong's point which is correct. It would be a fallacy to throw out all of his points because some are wrong. You make some good points also but the more transistors are not a con argument was pretty ridiculous, IMO.

Last edited by matthey on 17-Jun-2015 at 02:33 AM.

@matthey

I stand by my earlier comments and you're only making yourself look even less credible by continuing to defend anything that KingKong says. Even a blind squirrel gets lucky once in a while and finds a nut.....including KingKong. No one really cares the he was correct regarding the TPW of CPUs and no one was disputing it either.

Classic AmigaOS nor OS4 having anything to offer the world that the world doesn't already have, and arguing that they would take the world by storm by adapting them to a retro-grade CPU architecture is ludicrous. The last systems that even used a 48-bit architecture was in 1957 thru the mid-1960s, not including oddball embedded systems and FPGAs. In the 21st century, OS4 and AmigaOS are curiosities for hobbyists and that isn't going to change. Stop encouraging the delusional.

And I stand by my statement regarding transistor count as a gauge of CPU horsepower. It is an accepted industry standard for determining the raw computing ability of modern processors. Transistor count doubles every 18 months to two years, as does capability, or are you now saying that Moore's Law is ridiculous? It has stood the test of time and continues to hold true even if you feel it's "ridiculous. Adding transistors is a good thing It adds capability. It is not a con, although it has a cost as does just about any added capability.

Go ahead, design systems that keep stepping back the number of transistors and you'll eventually arrive back at the Commodore 64. How's that for progress!?

Last edited by ferrels on 17-Jun-2015 at 05:18 AM.
Last edited by ferrels on 17-Jun-2015 at 05:04 AM.
Last edited by ferrels on 17-Jun-2015 at 04:48 AM.
Last edited by ferrels on 17-Jun-2015 at 04:47 AM.
Last edited by ferrels on 17-Jun-2015 at 04:45 AM.

@ferrels

Glass half empty much?

_________________
All the way, with 68k

@agami


My, how original of you. That probably took you all of two hours to come up with and you probably strained yourself in the process.

My glass is quite full, thank you. I just have a low tolerance for the stupid and the delusional and for those who indulge the stupid and delusional.

Last edited by ferrels on 17-Jun-2015 at 07:19 AM.
Last edited by ferrels on 17-Jun-2015 at 06:46 AM.
Last edited by ferrels on 17-Jun-2015 at 06:26 AM.
Last edited by ferrels on 17-Jun-2015 at 06:23 AM.

@How to make AmigaOS a leading operating system?

1) Do not keep these kind of silly topics on the front page of the main community web site. (move to "free for all")
2) Be optimistic & realistic.
3) Focus on improving things, being constructive. Step by step, for decades to come.
4) ...
5) profit (olegil (tm) ?)


To become stronger:

HW1 - Affordable & simple & easy & warranty. (not available yet as "ReadyToRun", no info of "When")
HW2 - Expandable power user system (available, recycledMacHW, SAM and A1-X)

OS
- Focus, have a vision and a roadmap
- Implement R&D ways to speed up development of the most wanted OS features

SW
- enable building blocks for faster & high quality SW production (nice progress ongoing, gimme more)

Co-operation
- implement ways to make all flavours to work together (I see improvements at least by Application and HW developers, gimme more)

Open communication:
- let the community to see & understand what is being done, show the progress
- listen to community

I'm like an echo of my self I guess...

Last edited by KimmoK on 17-Jun-2015 at 09:28 AM.
Last edited by KimmoK on 17-Jun-2015 at 09:09 AM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?