@Hans

Quote:

Symmetrical Multi-Processing (SMP), yet.

I don't recall very well, but what would be better SMP or ASMP?

_________________
retired

@Leo

Quote:

I have nothing against it. But I don't see the use of taking the time to write a sandbox right now... A sandbox approach should have been taken back in 2002 (yeah, that's the MorphOS way), not now... And, btw, breaking bad written applications would help eliminating bad written apps and give better habbits to users&developers... No one ever did that at Commodore. That's a problem.

This change will break maybe 70-80% of every thing that runs on AmigaOS4.x, thats the problem (Classic AmigaOS3.x software too), and no the applications them self are not badly written infarct the old NDK3.9 and older SDK's does advice the use of permits and forbids in combination of other exec and dos commands, forbids and permits are pretty common in AmigaOS applications.

Its simply not possible to implement SMP whit out writing a sandbox, debating when it the right time to do it is useless, ExecSG1 is good foundation for writing ExecSG2, if like AmigaOS to continue to be AmigaOS, and not a Linux or BSD clone, I think this right thing to do even if we are talking about single processing, because forbids are often used for to long time intervals, inflicting pain on every thing else.

Last edited by NutsAboutAmiga on 26-Apr-2009 at 01:58 PM.

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

@Hans

"Where as a symmetric multiprocessor or SMP treats all of the processing elements in the system identically, an ASMP system assigns certain tasks only to certain processors. In particular, only one processor may be responsible for fielding all of the interrupts in the system or perhaps even performing all of the I/O in the system. This makes the design of the I/O system much simpler, although it tends to limit the ultimate performance of the system. Graphics cards, physics cards and cryptographic accelerators which are subordinate to a CPU in modern computers can be considered a form of asymmetric multiprocessing.[citation needed] SMP is extremely common in the modern computing world, when people refer to "multi core" or "multi processing" they are most commonly referring to SMP."

http://wapedia.mobi/en/Asymmetric_multiprocessing

I think i have somewhere OS/2 warp ASMP

_________________
retired

@AmigaBlitter

actually, implementing the second core as a coprocessor might actually work. Yes, the software would need to support it, but come on. This is Amigans we're talking about here

_________________
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.

@Leo

Fair enough. I see things taking a more gradual route. Constructing a sandbox back in 2002 would have only delayed things even further. They had their hands full with doing a proper migration to PowerPC, a task for which no sandbox was required. IMHO, now's a good time to take the next step.

@AmigaBlitter

I prefer SMP because it gives greater flexibility. Having said that, I'm not all that knowledgeable in multi-processing. In fact, I have no multicore hardware at all.

I don't think that ASMP would require treating the other cores as a coprocessor that requires special programming; you'd still be able to run multiple threads on different processors. It's just that certain tasks would have to be run on certain processors.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@AmigaBlitter

SMP implies equal (symmetric) access to all resources--memory bus, peripheral bus, etc.--by all processors.

ASMP implies unequal or disparate access to resources, e.g. one processor accesses the memory bus directly while the other relies on the main processor to page data to its local buffer (that's how the Cell's SPEs work). (How the "paging" works is specific to the hardware.)

EDIT: Depending on how the state of the m68k emulation in OS4 is managed, there may already be an SMP-like relationship between the emulator's core and the PowerPC CPU.

EDIT: Except for the Forbid()/Permit() snafu (and Disable()/Enable() as well), most AmigaOS software should alread be mostly SMP ready. Since AmigaOS is fully preemptive, all tasks have the appearance of running simultaneously. For example, if I have two tasks A and B:

A
global_var = 1;
global_var++;

B
global_var = 2;
global_var--;

and neither task takes steps to synchronize access to global_var, then there are absolutely no guarentees regarding order of execution. Any of the following could happen:

global_var = 1;
global_var++;
global_var = 2;
global_var--;

global_var = 1;
global_var = 2;
global_var--;
global_var++;

global_var = 2;
global_var = 1;
global_var--;
global_var++;

etc.

And depending on how the optimizer sees global_var, you'll probably end up with these two possibilities:

global_var = 2;
global_var = 1;

global_var = 1;
global_var = 2;

This is true whether tasks A and B are running on the same CPU via task switching or on separate CPUs simultaneously and is why synchronization primitives exist in the first place.

EDIT: And finally, even Forbid()/Permit() and Disable()/Enable() could be made to work properly in an SMP environment--the scheduler owns these, after all, and it can suspend tasks on all CPUs just as well as it can suspend them on one. Obviously, there's a performance penalty involved, but it's not a barrier to implementing SMP.

Last edited by Trev on 26-Apr-2009 at 06:12 AM.
Last edited by Trev on 26-Apr-2009 at 06:08 AM.
Last edited by Trev on 26-Apr-2009 at 05:56 AM.

_________________
Sam440ep-flex 733 MHz/1 GB RAM/Radeon 9250/AmigaOS4.1 Update 2
borked A1200/Blizzard1260+SCSI-IV/Z4+MediatorZIV/Deneb/Voodoo3/CatweaselMk3
more borked A1200/MBX1200z/Indivision
A500/clockport/RRNet
A600/A603

@olegil

Quote:

actually, implementing the second core as a coprocessor might actually work. Yes, the software would need to support it, but come on. This is Amigans we're talking about here

Heh, something like PowerUp boards going Hack & Patch way? There was only handful of software written for PPC and most of them were very poor. No sound, crashing, slowness etc...

_________________
Amiga Developer
Amiga 500, Efika, Mac Mini and PowerBook

Quote:

25th February they said "Hyperion Entertainment ... embarks on its most ambitious project to date".

pushing May anything further on this?

_________________
"The system no longer works " -- Young Anakin Skywalker

@Hans Quote:

I see things taking a more gradual route. Constructing a sandbox back in 2002 would have only delayed things even further. They had their hands full with doing a proper migration to PowerPC, a task for which no sandbox was required. IMHO, now's a good time to take the next step.

While I have previously criticised Hyperion for not going the sandbox route straight away, what you say does make some sense. If writing a sandbox would have been much more expensive/complicated that our current "mixed 68k + PPC" system, and Hyperion were lacking in resources even then (which I guess is true), then while the sandbox might have been the best way to go, it was not the quickest (and time was of the essence).

Last edited by ChrisH on 27-Apr-2009 at 06:53 PM.

_________________
Author of the PortablE programming language.
It is pitch black. You are likely to be eaten by a grue...

About the topic...

I hope AOS4.1 FINAL is the highest priority.

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

@KimmoK

It depends on what we consider "ambitious".
This could be only dreams and users speculations

_________________
retired

@AmigaBlitter

Quote:

It depends on what we consider "ambitious". This could be only dreams and users speculations

I think that is all it is.

_________________
Peter Swallow.
A1XEG3-800 [IBM 750FX PowerPC], running OS4.1FE, using ac97 onboard sound.

"There are 10 types of people in the world: those who understand binary, and those who don't."

@Hans

>My understanding is that the problem with SMP is more that forbit()/permit() would >be a real performance killer since all cores would have to be stopped.

yes all cores must stop, but in real world thats no problem, not many time is spend in forbid permit, you can log how many time is between forbid and permit call in real world if you dont believe it is near nothing time.

you can add a patch to forbid and permit.here is your OS4 HW usefull.the PPC cpu have a hardware cpu clock counter.

every forbid read the PPC counter to a var forbid_store.then on permit read the ppc counter again and substract the time from forbid_store and save value to clocksinforbid_permit.

then you can print out every sec this value.My experience is amiga stay on programs max 1-2% in forbid.

So no performance killer

Last edited by bernd_afa on 03-May-2009 at 07:56 PM.

@KimmoK

Quote:

KimmoK wrote: About the topic... I hope AOS4.1 FINAL is the highest priority.

Dunno about the SAM, but on Micro A1 and Peg 2, I've already got it. thanks.

_________________
Rage quited 29th May 2011

I have a feeling its working on AmigaOS5.0 and the ambitious part is working with Amiga Inc.
:)

Whats happened with the court case, anyone know, was'nt the last thing it said was actually coming to some kind of agreement by the 27th jan, I heard nothing since.

_________________
AmigaNG, YouTube, LeaveReality Studio

@AmigaBlitter

Quote:

AmigaBlitter wrote: @Trev I contacted this company, but after a few replies i never heard other info from them. http://www.cellsupercomputer.com/ One of the link and other search lead me to an ex amigans, an engineer, owner of the walker and other amiga prototype: He is one of the engineer that have the necessary knowledge to create Cell Boards, imho. http://www.blachford.info/computer/Cell/Cell0_v2.html Would be great to have him again in the Amiga team. http://www.blachford.info/computer/walker/walker.html I swapped some e-mail to him. I think he still an amigans.

I have debated with Mr Blachford in the past i.e. to debunk some his assertions.

In relation to http://www.blachford.info/computer/Cell/archive/Cell4.html

"The "Cray factor" is something to which Intel, AMD, Nvidia and ATI may have no answer to."

AMD has answered the specfic
1. Sony's 1080p PS3 hype with AMD GAME! platform.
http://www.youtube.com/watch?v=jN_aGKt7uhw
"AMD Takes You (198%) Beyond HD".

2. Sony's And Intel's raytracing FUD against GPUs.
http://www.tgdaily.com/content/view/38145/135/
"Radeon 4800 supports a 100% ray-traced pipeline".

Note that ATI != NVIDIA in regards to GPU raytracing.

Toshiba's SpursEngine SE1000 HD doesn't include PPE i.e. it uses the host's X86 CPU. X86 PC world effectively assimilated SPEs.

Quote:

The second reason is cost, other platforms have often used expensive custom components and have been made in smaller numbers. This has put their cost above that of PCs, putting them at immediate disadvantage. Cell may be expensive initially but once Sony and Toshiba's fabs ramp up it will be manufactured in massive volumes forcing the prices down, the fact it's going into the PS3 and TVs is an obvious help for getting the massive volumes that will be required. IBM will also be making Cells and many companies use IBM's silicon process technologies, if truly vast numbers of Cells were required Samsung, Chartered, Infineon and even AMD could manufacture them (provided they had a license of course).

AMD's cross licensing deals with Motorola(now Freescale) bars AMD from building a PowerPC compatiable processor.

Quote:

Cell is designed from the ground up to be more general purpose than GPUs, the APUs are not graphics specific so adapting non 3D algorithms will likely mean less work for developers.

NVIDIA CUDA 1.0/1.1/1.3 processors presents a SIMT (single instruction multi-threading) software model i.e. think of SIMD with "MD" part as multi-threaded.

Quote:

GPUs will provide the only viable competition to the Cell but even then for a number of reasons I don't think they will be able to catch the Cell.

To begin with, PS3's CELL+RSX combo didn't catch a PC powered by Geforce 8800 GTX on running multi-platform games beyond PS3's average 720p rendering resolution. NVIDIA's Geforce 8800 GTX was release a few days ahead of PS3.

Quote:

There is another reason I don't think Nvidia or ATI will be able to match the Cell's performance anytime soon. Last time around the PC rapidly caught up with and surpassed the PS2, I think it is one of Sony's aims this time to make that very difficult so, as such Cell has been designed in a highly aggressive manner.

On 8th of November 2006, NVIDIA's Geforce 8800 GTX (G80, CUDA 1.0 device) was release a few days ahead of PS3 and exceeded PS3's CELL at non-3D Fold@Home performance.

Quote:

CPUs and GPUs use a cache memory to hide access to main memory

Even since Radeon X1800/X1900, the GPU employs a large hyper-threading architecture to hide memory access latency. Besides the L1/L2 cache, GPUs also employs large register sets that numbers into thousands.

Each of G80's processing elements has the following specs
[MT IU]
[SP][SP]
[SP][SP]
[SP][SP]
[SP][SP]
[SSP][SSP]
[Shared memory]

SSP = Special SP not factored in.

8 scalar threads processors(SP) . Each SP issues two instructions per cycle i.e. 1 FMAD, 1 FMUL.
32 GFLOPS peak at 1.35Ghz. (fastest being clocked at +1.5Ghz)
8192 32bit registers (32KByte).
- 512KByte total register file space.
768 thread active threads in total
16KByte on-chip memory
- shared amongst threads of a block
- supports thread communication

Source: http://s08.idav.ucdavis.edu/luebke-nvidia-gpu-architecture.pdf

For example;
Geforce 8600 GT/Geforce 9650M GT has four of these processing elements i.e. 32 SPs with 32768 32bit registers.
Geforce 9600 GSO has eight of these processing elements i.e. 96 SPs with 98304 32bit registers i.e. these values are stored next to actual execution cores.
...

PS3's 7 SPEs only has 896 registers or effectively 3584 32bit data registers.
For processing math arrays, SPE's design pretty tame by comparison.

Quote:

The aggressiveness in the design of the Cell architecture means that it is going to be very, very difficult to produce a comparably performing part. x86 has no hope of getting there, they ultimately need to duplicate the Cell design in order to match it. GPUs will also have a hard time, they are currently at a 10 fold clock speed disadvantage, generate large amounts of heat and the highest performance parts are made in tiny numbers compared to what cell will be made. It will require a complete rethink of the GPUs design in order to get even close to the Cell's clock rate.

Geforce 8800 GTX's stream processors are clocked at 1.35 GHz. Laptop variants consumes less than PS3 CELL's 80 watts. A complete rethink of the GPUs design is not required.

http://www.notebookcheck.net/NVIDIA-GeForce-GTS-160M.14561.0.html
NVIDIA Geforce GTS 160M/9800M GS consumes 60 watts.

AMD/ATI Radeon HD 38x0 and NVIDIA Geforce GT300 series continues the GpGPU way.

Last edited by Hammer on 05-May-2009 at 01:54 PM.
Last edited by Hammer on 05-May-2009 at 01:39 PM.
Last edited by Hammer on 05-May-2009 at 01:37 PM.
Last edited by Hammer on 05-May-2009 at 01:35 PM.
Last edited by Hammer on 05-May-2009 at 12:55 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@ChrisH

Maybe they're going to Open Source OS4, or, write new RKRM docs?



tiffers

A new OS...for x86 or ARM.

@Al4

If I were them, I would go ARM.

ARM is the future. ARM is almost everywhere already. And now they want the desktop / laptop / netbook markets it seems.

_________________
Courtesy of SAM440Flex & Amiga OS4.1 only
Flex is 800mhz
A1000 with Classic 520 Amiga OS3.2.1
AmiKit 12
MorphOS PowerBook G4 (which can play youtube vids)

https://blitterwolf.blogspot.com

one of the biggest advantages with using x86 and compability with some brand motherboards is that if its readily availble
i belive Alot of people can go the credit route, and get a computer that they wouldnt be possible to buy at an regular amiga dealer.. atleast around here in sweden one of the main credit sources is via the store and not via an credit card, or use an special credit card that came with an previous store credit... of course some buy with their visa or american express cards on credit but anyway, not that credit is the most optimal way to buy a computer but still. i belive alot pc user buy their computers that way...
or heck i wouldnt mind finding an sam motherboard at one of the bigger swedish dealers, like komplett.se or something..