@OlafS25

Quote:

OlafS25 wrote: @Leo one correction... AROS did not break "compatibility" but is aims source code compatibility instead of binary compatibility, in short every program must be compiled for every platform to run and you cannot mix different binaries like X86 with 68k.

For the 68K platform, AROS goal is to have also binary compatibility.

@ferrels

-8



a OS is not a self-purpose or a "proof-of-concept", people use it to run software on it

@cdimauro

if you want a updated polished OS supporting more RAM, SMP and so on you have to break compatibility. Alone having 64bit means new compilation of software.


I am always wondering that people expect a "modern" OS and then all stays the same. And because AROS already offers some of the requested features and "feels" different because of that they do not want to use it.

@cdimauro

Yes but most people think of X86 if they say AROS

@OlafS25

Quote:

OlafS25 wrote: @cdimauro if you want a updated polished OS supporting more RAM, SMP and so on you have to break compatibility. Alone having 64bit means new compilation of software.

Absolutely.
Quote:

I am always wondering that people expect a "modern" OS and then all stays the same. And because AROS already offers some of the requested features and "feels" different because of that they do not want to use it.

Who cares, Olaf? That's not our problem.

Quote:

OlafS25 wrote: @cdimauro Yes but most people think of X86 if they say AROS

Again, we don't have to care about such mistakes. We know that AROS works even on 68K machines, where other post-Amiga/like o.ses don't.

And we know that "Intel outside" is just a childish slogan which also cannot be true anymore (because even non x86 machines use technologies developed by Intel).

@cdimauro

Quote:

cdimauro wrote: @OlafS25 And we know that "Intel outside" is just a childish slogan which also cannot be true anymore (because even non x86 machines use technologies developed by Intel).

In this day and age, absolutely - ten years ago it was a nice tagline for platforms on alternative architectures. Haven't seen anyone using one of those stickers on their iPad (or other generic ARM device) in the modern day though.

_________________
Troll - n., A disenfranchised former potential customer who remains interested enough to stay informed and express critical opinions.
opp., the vast majority who voted silently with their feet.

@Boot_WB: even more than 10 years ago, I think nobody can claim that his computer was "Intel outside".

Consider that Intel developed by herself the following technologies: PCI, AGP, AC'97, HD Audio. She also developed, with other companies, these: USB, PCI-Express, SATA, ACPI, UEFI.
Many peripherals use x86 code on/for their firmwares. Almost all graphic cards firmwares still are x86, and in particular they continue to integrate the very old VGA BIOS which is 8086 code (yes, the first Intel processor of the IA-32/x86 family!).
And, if I remember correctly, many chipsets (not only for x86) implement very old Intel chips, like the 8259 to manage interrupts.

So, it's very very hard that someone can claim that his computer is "Intel Outside". Basically with more modern hardware (e.g.: post-Amiga era) only ignorant or fanatic people can continue to sustain such false statement.

@cdimauro

You are right ofcourse, but the thing is that instead of going
always for the main stream there are also people who try as
much as possible to go for the underdog in some way or another.

_________________
A1G3-SE + OS4.1 u1 iso (x2)

@cdimauro

Still I wonder if you'd not complain if someone was calling Intel based systems "AMD inside"? ;)

@megol: LOL Should we take a look at the CPUID manufacturer string? ;)

@cdimauro
Just wanted to point out that even Intel systems use things from other parts - including Microsoft that have had a hand in many standards. :P

[ The 64 bit extension comes from AMD and is used by Intel via the patent-sharing agreements, other things also comes from AMD. The latest example is the PrefetchW instruction added in Broadwell which AMD have had defined since 1998. ]

@OlafS25

Surely you jest....I actually use hardware to run my software. Including my OS.

@megol: absolutely! Nothing to say here, and this enforces the fact that the "xyz outside" slogals don't make sense at all.

How abaout a 40 bit CPU/computer?

100% 40 bit, thus data and address bus with full 40 bits, with 40 bit data types also. 80 bit data types should be supported in hardware.

With 40 bit one can address 1 TB RAM ... but 40 bit are 37,5% less than 64 bit (or 64 bit are 60% more than 40 bit) and that's a huge advantage in chip production. Think of less energy consumption at possibly higher performance. More cores, possibly even faster.

Thus 40 bit can be a considerable advantage over 64 bit.

Some more text in German: Wie wäre es mit einem 40 Bit Computer? - sorry for not translating all.

@KingKong

Quote:

How abaout a 40 bit CPU/computer?

Totally pointless.

Quote:

100% 40 bit, thus data and address bus with full 40 bits

That would make horrible performance drop when accessing your imaginary 40 bit types or in general any types which are larger than these 40 bits.

Quote:

With 40 bit one can address 1 TB RAM ...

I hope you do realise that current 64bit cpus *DO NOT* implement full 64-bit physical address space?

Quote:

but 40 bit are 37,5% less than 64 bit (or 64 bit are 60% more than 40 bit) and that's a huge advantage in chip production. Think of less energy consumption at possibly higher performance. More cores, possibly even faster.

If you were right (but you are not) then it would mean the 32 bit CPUs should be even more energy efficient and even more faster than such nonsense 40 bit cpu. Are they?

Quote:

Some more text in German: Wie wäre es mit einem 40 Bit Computer? - sorry for not translating all.

Just like in case of all your posts, this text in German is just a text written without necessary technical background. I just wasted time reading it.

@KingKong: please, computer scientist, can you elaborate such 40-bit computer theory and give PRACTICAL examples on how to use it, and how an ordinary coder or compiler should deal with it?

Thanks

EDIT. @michalsc: TOTALLY agree.

Last edited by cdimauro on 05-Mar-2015 at 06:18 AM.

I can not use any Amiga OS for my work, none of the amigoid systems have a modern network stack that I need, they are all stuck in 1992 or thereabouts in terms of networking functionality. It's not like IETF have been doing nothing for 20+ years.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

Quote:

KingKong wrote: How about a 40 bit CPU/computer?

From a hardware perspective, I see several advantages to a 40 bit CPU over a 64 bit CPU but one big disadvantage. Memory can't be accessed in 40 bit chunks. This means wasted padding, alignment issues and more complexity (perhaps exceeding the waste of 64 bit processors). It's interesting that such a CPU could still be byte addressable as there are 5 bytes in 40 bits. Data types could be byte (8 bits), word (20 bits) and longword (40 bits). 20 bits would be enough for most foreign languages (a smaller version of 32 bit unicode), a great word size for variable length CISC encodings, and a fast while high quality 20 bit chunky gfx (although custom 20 bit gfx hardware would be needed also). A 40 bit address size allows plenty of memory without clogging up caches and memory with a bunch of zeroes like 64 bit. 40 bit chunky gfx would be good for professional quality still images and photography.

From a software perspective, there would be extra issues from compilers having some unusual data type sizes. Programs would be difficult to port. It would be difficult to gain software support considering the extra effort required for relatively small advantages.

Quote:

With 40 bit one can address 1 TB RAM ... but 40 bit are 37,5% less than 64 bit (or 64 bit are 60% more than 40 bit) and that's a huge advantage in chip production. Think of less energy consumption at possibly higher performance. More cores, possibly even faster. Thus 40 bit can be a considerable advantage over 64 bit.

32 bits is a considerable advantage over 64 bits in logic savings and energy efficiency. Choose a 32 bit CPU with good code density and an efficient OS and even 1 GB of memory is a lot. A PPC commonly needs 40-50% more memory for code and alignment than a 68k Amiga. An enhanced 68k could improve code density another 5%-15%. It would be less trouble to optimize a little more for 32 bits than try to make 40 bits work.

Data types could be 10, 20, 40 and 80 bits with support for 8, 16, 32 and 64 bits. Data types with 8, 16, 32, 64 bits should be reasonably supported by hardware, think of rotate and shift operations.

Memory access is only an issue of compatibility, there could be special 40 bit memory chips in the future. If 40 bit (or perhaps 48 bit) computing is a good idea - kind of combining the best of 32 bit and 64 bit - then it could become mainstream. Memory is quite cheap these days.

A very interesting option is that some bits could be used for ECC. One could have 10, 20, 40 bit without ECC or 8, 16, 32 bit with extraordinary ECC and easy 32 bit compatibility, which is combined with 40 bit adressing-capability and hardware support for 64 bit computing. Memory chips could be 40 oder 80 bits wide, to be used with 8 bit data + 2 bit ECC for 32 bit and 64 bit computing.

How much extra money would you pay for a super secure computer with bottom up checksums and extraordinary ECC? ... and perhaps it even must not be more expensive than 64 bit computing. The extra use of memory (ECC, padding) is compensated by more efficient data usage compared to 64 bit.

Because ECC is so desirable, this could lead to 48 bit = 40 bit + 8 bit ECC computing, to have 40 bit address bus and 40 bit data with 8 bit ECC.
40 bit or 48 bit? Well, it's not 64 bit for sure.

3 * 64 bit = 4 * 48 bit and such conventional 64 bit data bus memory chips could be used for 40 bit + 8 bit ECC computing (4 * 10 bit data + 4 * 2 bit ECC). There could also be a single read/write of 48 bit data with padding to 64 bit. The CPU-cache management could do this trick and automatically read/write with 64 bit alignment to conventional DDR SDRAM. Another variant: 2 * 64 bit = 3 * 40 bit + 8 bit ECC.

At least, there are some interesting possibilities - it should be worth to do some thinking about.

@cdimauro

Quote:

cdimauro wrote: @KingKong: please, computer scientist, can you elaborate such 40-bit computer theory and give PRACTICAL examples on how to use it, and how an ordinary coder or compiler should deal with it? Thanks

How about : compile your portable C code.
That's it.

Edit: But there is almost no advantages going to 40 bit processing - the power savings would be in the noise compared to a fully 64 bit processor.

Last edited by megol on 05-Mar-2015 at 10:39 AM.