Does MorphOS support virtual memory? No?

Virtual memory is not a cardinal criterium of technical advancement. Virtual memory working at 12 MB/s (raw disk read speed of AmigaOS 4.1 + all the patches on my Pegasos2) and crashing with numerous applications isn't a criterium at all. If I need more memory for AmigaOS 4 I would prefer to insert another 512 MB module in the second mainboard socket...

OWB 1.4 on MorphOS loads pages much faster than OWB 3.15 on AmigaOS 4.

_________________
Reggae · MorphOS Files · DigiBooster 3

@Krashan Quote:

This changed many times in the past and will change in the future. Then choosing "leading" system basing on this criterium may be misleading

It's kind of funny, because your statement applies EXACTLY to what you said earlier:
Quote:

1. Is more technically advanced. 2. Has no legal obstacles.

These have changed in the past, and will certainly change in the future too. At one point MOS was supposed to be illegally based on OS3.x source code (of course Amiga Inc never backed-up those claims with legal action, so we have to assume they were false allegations).

Last edited by ChrisH on 12-Aug-2009 at 08:47 AM.
Last edited by ChrisH on 12-Aug-2009 at 08:46 AM.

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

so we have to assume they were false allegations

Why are you using them as an argument then? Just to cover the lack of arguments?

_________________
Reggae · MorphOS Files · DigiBooster 3

Quote:

Virtual memory is not a cardinal criterium of technical advancement. Virtual memory working at 12 MB/s (raw disk read speed of AmigaOS 4.1 + all the patches on my Pegasos2)

1. virtual memory is not the same as paging
2. if raw disk speed of AmigaOS 4.1 is 12 MB/s on your Pegasos2, then you have UDMA disabled or something, because on my Pegasos2 I get over 60MB/s since the first time I installed it.
3. I agree that paging isn't that important on AmigaOS if you have 1GB+ of RAM, it will most likely never be needed in that case.

@COBRA

1. What is virtual memory by your definition then?
2. How to enable this UDMA mode on AmigaOS 4.1? I've just installed the system + all patches released.

_________________
Reggae · MorphOS Files · DigiBooster 3

@ChrisH

Quote:

MOS has only proved to be faster than OS4 on a Peg2

But OS4 on A1 is not faster than MorphOS on Peg2. When OS4 on A1 was compared against MorphOS people were crying out loud it is not fair because it is not the same hardware. There are many OS4/A1 vs MOS/Peg2 comparisons here at AWN.

SAM being without L2 cache most rest results would be the same if MorphOS/Peg2 was compared against OS4/SAM.

Quote:

of course Amiga Inc never backed-up those claims with legal action

Correction: Hyperion never backed-up those claims with legal action. But it is the past now.

Last edited by itix on 12-Aug-2009 at 09:04 AM.

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

@Krashan

Quote:

1. What is virtual memory by your definition then?

http://en.wikipedia.org/wiki/Virtual_memory

Quote:

How to enable this UDMA mode on AmigaOS 4.1? I've just installed the system + all patches released.

That's a good question, I'll have to check at home how to do that. UDMA should be enabled by default, at least I never changed any settings and the installed OS4 was running in UDMA mode from the first bootup giving me over 60MB/s transfers with scsispeed. But first of all you should try with idetool to see if UDMA is enabled or not, in a shell type:
> idetool -u peg2ide.device 'unit number of your drive'

Edit: can you show me the output of SCSIspeed?

Last edited by COBRA on 12-Aug-2009 at 09:36 AM.
Last edited by COBRA on 12-Aug-2009 at 09:12 AM.

Stephen.. instead of spend our time here.. is better to go back to develop and enhance our beloved OS.. and maybe reach MOS one day.. so.. go back at work!

_________________
http://www.amigasoft.net

5.AmigaOS4:> idetool -u peg2ide.device 0
-------------------Unit 0 information------------------------
Flags : $000001A5 - present, supports DMA, ATA, supports acoustic mgmt, supports cache flush, interrupts used,
Xfer mode : best pio 12 (PIO 4, 16 MB/s) / best dma 69 (UDMA 5, 100 MB/s) / current 69 (UDMA 5, 100 MB/s)
Total blocks : 117231408
Blocksize : 512
Power mode : 2 / idle (ready for operation)
IO1 / IO2 / BMCR @ : $1000 / $100E / $1020


Unfortunately I haven't been able to run SCSIspeed (downloaded from OS4Depot). I've tried "scsispeed drive HDB4:" and "scsispeed drive AmigaOS4:", where both names refer to the same partition. Then scsispeed just printed "cannot open device".

_________________
Reggae · MorphOS Files · DigiBooster 3

@Krashan

You definitely have UDMA enabled.

Try scsispeed with these options:

scsispeed DRIVE peg2ide.device:0 FAST BUF1=8192 BUF2=16384 BUF3=32768 BUF4=65536

@afxgroup

Quote:

Stephen.. instead of spend our time here.. is better to go back to develop and enhance our beloved OS.. and maybe reach MOS one day.. so.. go back at work!

I'm still waiting for you to add that port of xvidcore to ffmpeg which I sent you a few weeks ago

Holidays!!! you have to wait september..

_________________
http://www.amigasoft.net

@afxgroup

Quote:

Holidays!!! you have to wait september..

Ohh, I can't wait that long, PLEASE PLEASE PLEASE PLEASE!!!

@amije:
Yes, OS4.1 is BETA on Peg-II.
Even the site, that published these results is explicitly writing it in its review:
> Il faut mettre tout de même un bémol à ce comparatif. Il a eu lieu sur un matériel sur lequel l'AmigaOS 4.1 est soi-disant en version "bêta"
= The [comparision] took place on a hardware, where AmigaOS 4.1 is in "beta" version.

Last edited by _PAB_ on 12-Aug-2009 at 11:43 AM.

_________________

Quote:

This can only mean one thing: Under MorphOS the memory setup/configuration is different than under OS4

However, that does not mean that the memory CONTROLLER setup is different. There are other components involved in getting to memory.

In particular, the way the MMU is set up can make a big difference. In order of obviousness:

a) An OS might satisfy multi-page memory allocations by handing out multiple copy-on-write mapping to a single, zero-filled page. A benchmark that does not take this into account (i.e. does not fill the memory it uses for bechmarking before running the test) will hit the same page over and over again, and thus hit cache.

b) OSs might use different page sizes to mape memory, or (on the PPC), use the BATs. Using the rather common page size of 4kB, the 8MB test involves 2048 pages, and thus 2048 page MMU table entries. As the TLBs on the G4 have no hope of holding that many PTEs, there are tens of thousands of TLB reloads every second. The PPC uses hashed MMU tables. Given that a TLB miss, and the resulting MMU entry search, stops the processor pretty much dead in its tracks until the search has succeeded, the choices the OS makes with respect to the size of the MMU tables may make quite a difference (and even more so if the OS doesn't even use paging, but instead uses always-on-chip BATs to do the address translation).
(Some PPCs, although IIRC not the G3/G4 ones, do not even do the MMU search in hardware, but instead cause an exception on each TLB miss and leave it to the OS. Major potential for performance differences there --- and the G4 can be switched into a software-only update mode)

c) As discussed in the Natami thread, DDR memory is anything but simple. Different virtual-to-physical mappings by different OSs may very well result in significantly more or less favourable row/page changing overhead for the exact same accesses into contiguous virtual address space.


These are just a few MMU-related possibilities from the top of my head. There are almost certainly quite a few more.
Also, I seem to recall something about OS4 not using L3 cache; Does that G4 module have the L3 cache installed?

@umisef

You are absolutely correct that MMU setup can also make a difference, but that doesn't explain why it is so much faster on AmigaOne hardware, unless for some reason the OS uses a radically different MMU setup on the AmigaOne with the same processors, which is very unlikely, and I also don't think that the MMU setup even if it was different, can make such a big (2.5x) speed difference.

I was not even aware that Pegasos2 G4's had such a thing as L3 cache, how can I check this?

It worked this way. It turned out, that raw read is much faster than 12 MB/s. This 12 MB/s was a copy speed when copying files between disk partitions. Raw read is at 40 MB/s:

5.AmigaOS4:Work/Scsispeed4> scsispeed drive=peg2ide.device:0 fast buf1=8192 buf2=16384 buf3=32768 buf4=65536
MKSoft ScsiSpeed 4.2 Copyright © 1989-92 MKSoft Development
AmigaOS4 port by Stéphane Guillard
------------------------------------------------------------
CPU: 68020 AmigaOS Version: 53.5
Device: peg2ide.device:0

Testing with a 8192 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 33622835 bytes/sec

Testing with a 16384 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 40403763 bytes/sec

Testing with a 32768 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 40244019 bytes/sec

Testing with a 65536 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 40068710 bytes/sec

I couldn't miss an opportunity to perform the same test with MorphOS:

Ram Disk:DiskSpeed> ScsiSpeed drive=ide.device:0 fast buf1 8192 buf2 16384 buf3 32768 buf4 65536
Aros ScsiSpeed 4.10 Copyright © 2004-2006 The AROS Development Team
------------------------------------------------------------------
Processor: PPC MorphOS Build: 51.34 Normal Video DMA
Device: ide.device:0

CPU Calibration shows that CPU availability tests
would be inaccurate in the current system state.
No CPU Speed Rating -- CPU % not available.

Testing with a 8192 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 30730649 bytes/sec

Testing with a 16384 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 40265318 bytes/sec

Testing with a 32768 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 40165376 bytes/sec

Testing with a 65536 byte, MEMF_FAST, LONG-aligned buffer.
Read from SCSI: 39924531 bytes/sec

Except of 8192 bytes buffer, where AmigaOS4 is faster by 10%, the rest of results differs only within 1% of measure error margin. The conclusion is performance of IDE drivers is comparable, limited by my disk I guess (which is Seagate Barracuda 60 GB one). On the other hand file copying with DOS command 'copy' (large, single 200 MB file, HDD -> the same HDD) is 16 MB/s on MorphOS and 12 MB/s on AmigaOS 4.

_________________
Reggae · MorphOS Files · DigiBooster 3

@Krashan:

Quote:

Why are you using them as an argument then? Just to cover the lack of arguments?

Maybe to illustrate the fact that things can change in the future just like they changed in the past using one of the point you used ?

Oh and to add to the discussion I'll say like Jahc : who cares ? Why can't we just enjoy our choice respecting other's without trying to convert them to our own choice ? I really don't understand that spirit.

_________________
AOS 4.1 : I dream it, Hyperion did it !
Now dreaming AOS 4.2...
Thank you to all devs involved for this great job !

Why can't we just enjoy our choice respecting other's without trying to convert them to our own choice?

There is no try to convert. Most of us have choosen our hobby operating system because of other things than speedtest results. My point of view is explained in post #19 in this thread for example. I doubt any dedicated AmigaOS 4 user would switch to MorphOS after reading the article or our nice flamefest here.

_________________
Reggae · MorphOS Files · DigiBooster 3

Is it better to wait for MOS for Minimacs?

It seems it is a lot better than OS4.

I was waiting for SAM prices to become cheaper to finally buy a new AMIGA hardware.
Now it looks like MOS is better choice.
And Minimacs are more powerful than SAMs too.
They have 1.4 GHZ and SAM only has MHZ.

I'm really confused and uncertain about this.
Situation never gets clear. Waiting for many years to resolve now.

Excuse for bad english.

_________________
AMIGA 500/EXT.FLOPPY
AMIGA 1200/030/50MHz/FPU/SCSI
AMIGA 4000/060/50MHz/SCSI/CYBERVISION
AMIGA CD32
AMIGA CDTV
AMIGA T-Shirt
AMIGA Mousepad
Commodore Underwear