Poster | Thread |
cdimauro
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 6:16:51
| | [ #301 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3650
From: Germany | | |
|
| |
Status: Offline |
|
|
cdimauro
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 6:22:52
| | [ #302 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3650
From: Germany | | |
|
| @IntuitionAmiga
Quote:
IntuitionAmiga wrote: @NutsAboutAmiga
Quote:
OS that is crippled by having to be run in a 68k emulator |
Why o you insist on repeatedly making false claims like this?
x86 and 68k binaries are freely interchangeable on Amithlon, just as PPC and 68k binaries are freely interchangeable on MorphOS and OS4.
Compile equivalents of each OS3.x component to x86-BE binaries and the whole OS will still run identically yet still the ability to run 68k binaries will be there.
Likewise, replace PPC components with 68k equivalents on MorphOS and the OS will still run identically and the ability to run PPC binaries will still be there.
I assume this is also the case for OS4 but I haven't used it on my own hardware since I sold my BPPC in 2007. I will be able to test this when my Pegasos II arrives in a week or so, but I have no reason to believe it is not the case. |
But with this implementation the o.s. is effectively crippled, because you're importing the limits of the 68K o.s..
The first and most obvious is the use of 64-bit pointers, as NutsAboutAmiga stated. AmigaOS4 and MorphOS use 32-bit pointers and the data structure are exactly the same as the good old 68K Amiga o.s.. This approach cripples a 64-bit version of the o.s..
And something similar happens with multiple cores, unless a solution is found even for the old 68K code, which I think is very hard to achieve. |
|
Status: Offline |
|
|
cdimauro
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 6:24:27
| | [ #303 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3650
From: Germany | | |
|
| @terminills
Quote:
terminills wrote: @NutsAboutAmiga
Quote:
its funny that AROS native is the faster then AROS hosted, if it was so desirable. |
Proof? |
Suppose that the native version has the same accuracy / optimization level for every component (drivers, primarily) of the hosted one, the former has one layer less to deal with... |
|
Status: Offline |
|
|
KimmoK
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 7:25:19
| | [ #304 ] |
|
|
|
Elite Member |
Joined: 14-Mar-2003 Posts: 5211
From: Ylikiiminki, Finland | | |
|
| |
Status: Offline |
|
|
terminills
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 10:52:23
| | [ #305 ] |
|
|
|
AROS Core Developer |
Joined: 8-Mar-2003 Posts: 1473
From: Unknown | | |
|
| @cdimauro
I asked for proof not speculation. :) _________________ Support AROS sponsor a developer.
"AROS is prolly illegal ~ Evert Carton" intentionally quoted out of context for dramatic effect |
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 12:58:07
| | [ #306 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12845
From: Norway | | |
|
| @terminills
Yes, but can you disprove it?
You are not getting me to waste my time on search Youtube for AROS videos, to find evidence to prove or disprove it, ether you believe what I say is true or you don't.
I strongly believe every case where you have function x calling function y and then calling function z is going to be slower than if you can just call function x and that’s it.
But it all boils down to how well the AROS drivers are written.
Test case should be whit working native Nvidia drivers and fully supported SATA controller and so on, if you want to run the experiment.
Last edited by NutsAboutAmiga on 06-Nov-2013 at 01:35 PM. Last edited by NutsAboutAmiga on 06-Nov-2013 at 12:59 PM.
_________________ http://lifeofliveforit.blogspot.no/ Facebook::LiveForIt Software for AmigaOS |
|
Status: Offline |
|
|
Hypex
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 12:58:53
| | [ #307 ] |
|
|
|
Elite Member |
Joined: 6-May-2007 Posts: 11245
From: Greensborough, Australia | | |
|
| @NutsAboutAmiga
Quote:
I don't like fork(), it make it hard to know where a tread starts and where it stops |
I can see that. Some examples I've seen simply spawn off a thread while the main process waits or doesn't do much else which seems pointless. Almost like detatching itself from CLI and that was it.
Quote:
Plus fork() its not a essential part of a modern OS, Windows does not have it. |
Then we have hope! I only bring it up as it seems to be in the mould of those feaures we miss in an ever increasing list. Multitasking with multithreading. SMP. MP. VM. 64-bit. OOP API.
The Amiga multitasking with its preemptive strike is no longer modern anymore. It lacks threads. And the rest that comes with it. With other forms of IPC.
If threading can be added to the present multitasking model with its shared memory concept and work as a thread does it should be added in. Either attached to a process itself or perhaps as a new object, NT_THREAD such as. Infact, it almost sounds easy. Of course it's not that easy. But do able. For now we have to put up with virtual threads as child tasks.
Quote:
Its only natural that CreateNewProc() is extended whit NP_SMP or some other tag that tells the system that process can run on any core, |
I've been told the system runs in an interrupt; that is Exec, for most of the time. But I think in the current model where each process is its own this should be more possible. That is, because of the shared memory across each process and how any child task is another seperate process, should me it easier to split the processes across cores. As long as there is no memory collision since multi-core hardware really does multitask.
We are already seeing the proof of this concept in the X1000 X-Kernel.
Quote:
Anyway if you where to duplicate memory when you do not need it the system is going to be memory hungry. |
I'd advise against this also. Perhaps this is where the Amiga memory model can be made use of. At least there a process can simply share memory with threads. Which is the intended purpose AFAIK. I don't know if this concept would cause problems when porting foreign code relying on standard MP, SMP and thread features.Last edited by Hypex on 06-Nov-2013 at 01:04 PM.
|
|
Status: Offline |
|
|
damocles
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 13:49:23
| | [ #308 ] |
|
|
|
Super Member |
Joined: 22-Dec-2007 Posts: 1719
From: Unknown | | |
|
| @NutsAboutAmiga
Quote:
es, but can you disprove it? You are not getting me to waste my time on search Youtube for AROS videos, to find evidence to prove or disprove it, ether you believe what I say is true or you don't. I strongly believe every case where you have function x calling function y and then calling function z is going to slower than if you can just call function x and that’s it. But it all boils down to how well the AROS drivers are written. Test case should be whit working native Nvidia drivers and fully supported SATA controller and so on, if you want to run the experiment. |
So you are going to make the charge and demand the other person prove you wrong solely because you think it's a logical assessment that you refuse to confirm by doing testing yourself? What makes this even more preposterous of a stance is that Terminills is one of a dozen or so people who has in depth knowledge of AROS native and hosted.
_________________ Dammy |
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 14:03:56
| | [ #309 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12845
From: Norway | | |
|
| |
Status: Offline |
|
|
bison
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 17:19:43
| | [ #310 ] |
|
|
|
Elite Member |
Joined: 18-Dec-2007 Posts: 2112
From: N-Space | | |
|
| @cdimauro
Quote:
Yes, that's right, but as you stated there are also open source drivers for graphics cards. The quality is not that good compared to the close ones and they let you use your card, albeit it's very difficult for games or some applications. |
I hope that's changing. Both Nvidia and ATI recently released more information, prodded along by Valve, I think. Or Nvidia was induced by Valve, and ATI by Nvidia, in order to keep something close to parity.
There's this theory that both Nvidia and ATI will never release all their specs, since it would reveal that they've been stealing each others IP. Of course there's no way to know if this is true or just a conspiracy theory.
_________________ "Unix is supposed to fix that." -- Jay Miner |
|
Status: Offline |
|
|
Arko
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 17:40:24
| | [ #311 ] |
|
|
|
Super Member |
Joined: 17-Jan-2007 Posts: 1989
From: Unknown | | |
|
| @cdimauro
Quote:
Suppose that the native version has the same accuracy / optimization level for every component (drivers, primarily) of the hosted one, the former has one layer less to deal with... |
Linux has some really good drivers vor Disk I/O, GFX and Ethernrt. I would not be suprised if Aros hosted on most x86 would beat Aros native when it comes to speed tests._________________ AmigaONE. Haha. Just because you can put label on it does not make it Amiga.
I borrowed this comments from here (#27 & #28): http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=38873&forum=2&start=20&order=0 |
|
Status: Offline |
|
|
cdimauro
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 19:20:28
| | [ #312 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3650
From: Germany | | |
|
| @terminills
Quote:
terminills wrote: @cdimauro
I asked for proof not speculation. :) |
So if the hosted AROS version performs better than the native one, why the developers are wasting their (little) time with the latter? They can just focus on the former, right?
The truth is that IF the hosted version(s) has some advantage over the native one, it's solely because of the efforts that the developers have put on the hosting o.s.. and/or the drivers. And you should know it, if you're a developer and have some knowledge about how a platform (o.s., drivers, application) works. |
|
Status: Offline |
|
|
cdimauro
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 19:23:25
| | [ #313 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3650
From: Germany | | |
|
| Quote:
Hypex wrote: [quote]The Amiga multitasking with its preemptive strike is no longer modern anymore. It lacks threads. And the rest that comes with it. With other forms of IPC.
If threading can be added to the present multitasking model with its shared memory concept and work as a thread does it should be added in. Either attached to a process itself or perhaps as a new object, NT_THREAD such as. Infact, it almost sounds easy. Of course it's not that easy. But do able. For now we have to put up with virtual threads as child tasks. |
In reality it's the exact opposite: an Amiga o.s. has only threads, and lacks processes. The Amiga o.s., drivers, and application platform runs as if there's only one process, and all the tasks are threads (which share all the memory in a common address space). |
|
Status: Offline |
|
|
cdimauro
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 19:27:50
| | [ #314 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3650
From: Germany | | |
|
| @bison
Quote:
bison wrote: @cdimauro
Quote:
Yes, that's right, but as you stated there are also open source drivers for graphics cards. The quality is not that good compared to the close ones and they let you use your card, albeit it's very difficult for games or some applications. |
I hope that's changing. Both Nvidia and ATI recently released more information, prodded along by Valve, I think. Or Nvidia was induced by Valve, and ATI by Nvidia, in order to keep something close to parity. |
Not so much. AMD unveiled its old GPU, whereas nVidia discovered only some very limited information about a part of its GPUs.
Quote:
There's this theory that both Nvidia and ATI will never release all their specs, since it would reveal that they've been stealing each others IP. Of course there's no way to know if this is true or just a conspiracy theory. |
I think that it's true. Hardware companies have A LOT of patents and trade secrets to protect their IP against the competitors. In the same conditions, I'll do the same: it's the business the demands it. |
|
Status: Offline |
|
|
cdimauro
| |
Re: Should Linux kernel power future AOS solution ? Posted on 6-Nov-2013 19:37:10
| | [ #315 ] |
|
|
|
Elite Member |
Joined: 29-Oct-2012 Posts: 3650
From: Germany | | |
|
| Quote:
Arko wrote: @cdimauro
Quote:
Suppose that the native version has the same accuracy / optimization level for every component (drivers, primarily) of the hosted one, the former has one layer less to deal with... |
Linux has some really good drivers vor Disk I/O, GFX and Ethernrt. I would not be suprised if Aros hosted on most x86 would beat Aros native when it comes to speed tests. |
Me too. But it's an unfair comparison, because Linux can count on THOUSANDS of developers (most of them coming from very big companies) whereas AROS has 2-3 order of magnitude below coders. Even if AROS is a very lightweight o.s., it can easily lose the battle from this PoV.
But the "fault" isn't related to its "native" nature. It's a question of the forces playing in the field: too much unbalanced in favor of the other o.ses.
Besides that, there's no technical reason for an hosted version to have better performance than the native one. I wonder of the contrary, but in this case I'm pretty curious to know the (again, technical) explanations. |
|
Status: Offline |
|
|
terminills
| |
Re: Should Linux kernel power future AOS solution ? Posted on 8-Nov-2013 11:05:57
| | [ #316 ] |
|
|
|
AROS Core Developer |
Joined: 8-Mar-2003 Posts: 1473
From: Unknown | | |
|
| @cdimauro
Quote:
Me too. But it's an unfair comparison, because Linux can count on THOUSANDS of developers (most of them coming from very big companies) whereas AROS has 2-3 order of magnitude below coders. Even if AROS is a very lightweight o.s., it can easily lose the battle from this PoV. |
It's not unfair to compare the reality of the situation.
_________________ Support AROS sponsor a developer.
"AROS is prolly illegal ~ Evert Carton" intentionally quoted out of context for dramatic effect |
|
Status: Offline |
|
|
OlafS25
| |
Re: Should Linux kernel power future AOS solution ? Posted on 8-Nov-2013 11:28:29
| | [ #317 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6369
From: Unknown | | |
|
| @cdimauro
People should not speculate too much and better wait |
|
Status: Offline |
|
|
OlafS25
| |
Re: Should Linux kernel power future AOS solution ? Posted on 8-Nov-2013 12:44:27
| | [ #318 ] |
|
|
|
Elite Member |
Joined: 12-May-2010 Posts: 6369
From: Unknown | | |
|
| @cdimauro
the problem in the amiga-community is not only the lack of resources, it is the desire to reinvent every wheel (even if the result will be worse and it is wasting resources) |
|
Status: Offline |
|
|
Yo
| |
Re: Should Linux kernel power future AOS solution ? Posted on 8-Nov-2013 13:17:54
| | [ #319 ] |
|
|
|
Team Member |
Joined: 8-Oct-2004 Posts: 2043
From: France, on an ADSL line | | |
|
| @OlafS25
Quote:
OlafS25 wrote: @cdimauro
the problem in the amiga-community is not only the lack of resources, it is the desire to reinvent every wheel (even if the result will be worse and it is wasting resources) |
+1 (No kidding....)_________________ ¤¤ Official Hyperion Zealot ¤¤
(No, I didn't type that with a straight face.) |
|
Status: Offline |
|
|
Hypex
| |
Re: Should Linux kernel power future AOS solution ? Posted on 8-Nov-2013 13:22:47
| | [ #320 ] |
|
|
|
Elite Member |
Joined: 6-May-2007 Posts: 11245
From: Greensborough, Australia | | |
|
| @cdimauro
Quote:
In reality it's the exact opposite: an Amiga o.s. has only threads, and lacks processes. |
I understand your point? So where is the main process? |
|
Status: Offline |
|
|