https://www.power.org/documentation/power-org-power-architecture-silicon-roadmap-update-2013/

https://www.power.org/documentation/power-roadmap-power8/

_________________
retired

@AmigaBlitter

What was the X5000 supposed to use?

_________________
- Don't get fooled by my avatar, I'm not like that (anymore, mostly... maybe only sometimes)
> Amiga Classic and OS4 developer for OnyxSoft.

@Deniil715

P5020 and P5040.

_________________

@WolfToTheMoon

Well, it IS using P5020 and P5040. Whatever it was SUPPOSED to use, we'll probably never know

_________________
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.

@AmigaBlitter

That wasn't much of a roadmap, if I may say my opinion.

_________________
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.

@olegil

yeah, Freescale doesn't have a single new PPC core planned, according to that document. I guess they'll just keep shrinking cores as new processes become available.

_________________

@olegil

Quote:

That wasn't much of a roadmap, if I may say my opinion.

Probably cause the power architecture has been licenced to other companies, such nvidia, google ....

There in an interesting innovation in the power 8, if i read correctly, a special section of the processor usefull to attach other processor (fpga or asic) with a dedicated optimized bus.

_________________
retired

POWER 8 can only be a post-Amiga machine dream.

Quote:

WolfToTheMoon wrote: @olegil yeah, Freescale doesn't have a single new PPC core planned, according to that document. I guess they'll just keep shrinking cores as new processes become available.

Exactly. Despite of many claims of some people who don't want to accept the reality...

Why is A-EON not a member of power.org ?

_________________
Sometimes we give people a lot of credit just because they’re writing nice sentences even if it isn’t adding up to much.

@AmigaBlitter

What!? Prototype when the Motorola 68000 was new? 64-bit when Windows '95 came out? Multicore 2001? 22 years!? Then where is it! Is this an exaggeration?

If the above were true it would be easy for A-EON to locate a four core 64-bit PowerPC chip with AltiVec. Or find a company who can license it for their own boards and resell it as an AmigaOne. That's how they make it look.

Sounds too good to sound true. I don't see it. Perhaps I'm looking in the wrong places but all I see is Intel taking over the world.

Altivec is very important (and it's unbelievable that was missing for the new machines), but 64 bits are not, because they are a treat to the backward compatibility with the existing applications.

You can only use 64-bit in limited scenarios, for example to implement a cache for disk reads and/or writes, using the memory >= 2GB. Or to better support the virtual memory, where applicable (swap in >=2GB memory space instead of doing it with the much slower hard disk).

Except cases like these, you'll not see new 64-bit applications that can freely run in a 64-bit address space AND be able to "collaborate" (share resources, primarily) with the old 68K and PowerPC software, included the o.s., like it happens now.

The Amiga o.s. structures are still "hard-coded" (32/31 bits pointers, 32-bit longs, and so on).

@Birbo

Quote:

Why is A-EON not a member of power.org ?

A-eon is already member of power consortium

_________________
retired

@cdimauro

Microsoft is able to have 32bit programs running on a 64bit OS, it is possible.

Quote:

The Amiga o.s. structures are still "hard-coded" (32/31 bits pointers, 32-bit longs, and so on).

a 32bit register is just a 64bit register whit padded zero's is it not, in other words, a 64bit program has no problem communicating whit 32bit program over 32bit address pointer, the issue is the revere order, where you have a 64bit memory space you try to present to 32bit program.

In the internet you have some thing called nating where you have two or more computers shearing one wan IP, it work by maping what port belongs to what PC, in the same way 32bit program can be mapped two memory using the MMU if needed, clearly this want be backwords compatible, only programs that know how to communicate in this way will be able to.

There are many problems that has to be worked out foreshore, but many things can be virtual, the user will never know what runs inside or outside the virtual space.

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

@AmigaBlitter

What? To download the full roadmap they require people e-mail address?

How strange behaviour ask personal informations to just download a common roadmap document!

_________________
"When the Amiga came out, everyone [at Apple] was scared as hell." (J.L. Gassée, former CEO of Apple France and chief of devs of Mac II-fx, interviewed by Amazing Computing, Nov 1996).

Quote:

NutsAboutAmiga wrote: @cdimauro Microsoft is able to have 32bit programs running on a 64bit OS, it is possible.

Windows is totally different: it hasn't a single, public and fully shared, address space for 32-bit applications, 64-bit applications, and even the o.s., like happens with the Amiga o.s..

Also Windows doesn't "publish" the internal data structure which it uses, and offers a different set of APIs for the 64-bit subsystems.
Quote:

Quote: The Amiga o.s. structures are still "hard-coded" (32/31 bits pointers, 32-bit longs, and so on). a 32bit register is just a 64bit register whit padded zero's is it not, in other words, a 64bit program has no problem communicating whit 32bit program over 32bit address pointer, the issue is the revere order, where you have a 64bit memory space you try to present to 32bit program. In the internet you have some thing called nating where you have two or more computers shearing one wan IP, it work by maping what port belongs to what PC, in the same way 32bit program can be mapped two memory using the MMU if needed, clearly this want be backwords compatible, only programs that know how to communicate in this way will be able to. There are many problems that has to be worked out foreshore, but many things can be virtual, the user will never know what runs inside or outside the virtual space.

You cannot do it in the Amiga o.s. condition, where applications can easily access even the task list, the memory pool, the screens, etc. etc. ect. And you cannot provide 2 copies of the same staff, because the data structures are deeply linked. Think about a Screen structure, with all the pointers to other data structure, which themselves have pointers to data structures, and so on...

So, you cannot "tunnel" / "wrap" / "virtualize" every data structure created in a 64-bit environment to make it also accessible by 32-bit code.

That's why fully supporting 64-bit isn't possible with an Amiga o.s./like. You can take advantage of the >2GB address space in SOME cases, like the ones that I already depicted, but providing a brand new 64-bit environment (which means also new 64-bit APIs / ABIs) is out of question.

May be something like the 64-bit applications introduced by Apple with OS X Tiger can be possible, which had a very limited an restricted set of 64-bit APIs (not even the GUI ones), and can communicate with 32-bit applications through some IPC (message passing by copy). But even this requires investments, which are not worth the archievable (very) little benefits.

Last but not least, PowerPC 64-bit applications are slow on average compared with the same 32-bit ones, due to the large pointers and (usually) integers used, which create more caches and TLB pressures...

@roadmaps

Disappointing read.
I think roadmaps were updated to show that no new cores are coming for now...

btw. they are available without registering/email address via googling.

@Hypex

>What!? Prototype when the Motorola 68000 was new?

IBM did first RISC prototype in 1980.
IBM did first RISC computer in 1986.
(IBM RT PC: Successor of IBM 801, predecessor of IBM POWER architecture which eventually lead to PowerPC.)

> 64-bit when Windows '95 came out?

64bit PPC was announced 1993.

> Multicore 2001?

I think they were developing towards it already in 1993...

>22 years!? Then where is it!

Everywhere, almost. ;)
Everybody use them daily... (especially if they use cell phones)

>Is this an exaggeration?

More...
"1990: RS6000 (IBM) POWER architecture chip, predecessor of PowerPC CPU. Partitioned RISC design lends itself to superscalar processing - this is the first superscalar processor, capable of executing multiple instructions at once."

etc...

>it would be easy for A-EON to locate a four core 64-bit PowerPC chip with AltiVec.

Freescale T208x T4xxx chips have Altivec and 4...12 cores and I think they are available as samples for system designers.

And also modern POWER chips have Altivec/VMX, unless I'm mistaken.

>... all I see is Intel taking over the world.

You got it wrong. It's ARM. ;)
(so far Intel has been strongly only on desktop/laptop)

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

@cdimauro

Quote:

Windows is totally different: it hasn't a single, public and fully shared, address space for 32-bit applications, 64-bit applications, and even the o.s., like happens with the Amiga o.s.. Also Windows doesn't "publish" the internal data structure which it uses, and offers a different set of APIs for the 64-bit subsystems.

That's not entirely true. Windows still supports the win16 API that was the main API used in all DOS-based versions of Windows. Since DOS doesn't natively support an MMU, win16 applications all shared the same address space and could pass pointers to each other much like AmigaOS programs can. Private OS data was also exposed, see for example the SetWindowLong function which is a glorified poke into the internal data structures used to manage controls (in Windows everything is a window: actual windows, buttons, sliders, etc. It's like the base class of all controls). The way MS still supports that programming model in 32 bit Windows is interesting, and I expect that Hyperion's answer to running 32 bit apps in 64 bit AmigaOS will use similar techniques (the 3 second answer is that they sandbox all win16 apps into the same address space - it's obviously more complex than that since win16 apps can still interact with 32 bit apps in separate address spaces, but that's the basic technique). Win16 apps can't run in 64 bit windows because x86 CPUs can't easily mix 16 bit and 64 bit instructions like they can 16 and 32 bit or 32 and 64 bit. Finally, the API used to program 64 bit Windows programs is exactly the same that one uses for 32 bit. There are some new specialized memory management functions for dealing with the larger address space, but if you're writing a "normal" program like a text editor or an MP3 player all you have to do is pass different switches to the compiler.

@cdimauro

This highly theorical, but any way, there are ways this can be done, how Hyperion decides to do it we do not know, its not even on the road map yet.

As for now we are waiting for SMP and 3D, what is going to happen after that we do not know, If I may guess, I think the natural progression will be to make a pure 64bit version for AmigaOS for the brave.

How and if the there will be backwords compatability is another issue, but as I see it should be possible given time and resources, to support 32bit in a 64bit OS, as you point out there are many differences, but consider this many of data structures in Exec was never intended to be poked around in anyway, its of no programs business to poke in the memory pools, or in the OS structures.

All the programs should care about getting memory when they ask for it, where the memory is how the OS structures are organized and how the IPC is working is some thing that the OS will need to handel, its not program task to handel that it should only get what it expect.

Clearly there will be many differences in 64bit OS vs a 32bit OS.

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

@AmigaBlitter

Thank you very much for the information. Do you know, why A-EON is not on the list at www.power.org?

Or did I miss something?

_________________
Sometimes we give people a lot of credit just because they’re writing nice sentences even if it isn’t adding up to much.

@Birbo

Yes I can confirm I am a member of power.org and have been for a number of years.

TrevorD

_________________
No, I don't need no reason, I'm just breezin'