@Boot_WB

Quote:

Posted by Steven Solie on October 24, 2013 X-Kernel Update: Task scheduler rewritten in C Removed reliance on data structures (e.g. ExecBase task lists and ThisTask pointer) Moving scheduler to run on auxiliary cores All cores schedule tasks independently Load balancing between cores

So the kernel has been rewritten again ?

It was first rewritten years back :



AmigaOS 4.0 Prerelease Update #3

page 2
Quote:

The following is a non-exhaustive list of the features of update #3 New Kernel. The new kernel includes numerous optimizations, bug fixes and new features. It can now run a special variety of ixemul library and thus run all of the ADE. ELF library has a memory leak fixed. New IDE devices. IDE drivers are included for the VIA 686B onboard and Silicon Images 0680 Paralell ATA controllers, and the Silicon Images SiI3112, SiI3512 and SiI3114 Serial ATA controllers. All of these support UDMA on hard disks and CD/DVD-Drives. New Picasso96/Graphics and monitor drivers. The new monitor drivers support usage of vertical blanking interrupts for display synchronization. Limited Gamma support is available in the Voodoo driver. Warp3D V51. This update contains Warp3D V51, the all new, OS4-native version of the classic 3D hardware driver system. This release includes completely re-written drivers for the Voodoo 3 (Avenger), Voodoo 4/5 (Napalm) and the Radeon 7x00 series of graphics cards. New features include single-cycle multitexture support (two simultaneous units on the Voodoo's, three on the Radeon) with simplified and extended combiner modes, interleaved vertex arrays, up to 8 bit stencil buffer support (where supported by hardware), and more. WarpUp support. The new powerpc.library emulates loading and execution of WarpUp binaries. There are a few remaining compatibility issues, especially in conjunction with Warp3D, that are still being worked on. New Shell and updated shell commands Update reaction classes and new intuition New Roadshow TCP/IP stack and updated eth3com.device All new version of the Sirion USB stack and usbprinter.device Updated datatypes Updated Preferences programs Updated Locale

with all these versions no one is keeping kernel timings/ test speed logs, no regression testing for efficiency or compatibility.

it is good to see progress updates, i hope we continue to see updates from all departments :). Good work.

_________________
Download 499.26 Mbps, 659.94 Mbps Upload :)

@retro

Quote:

way is smp soo importent ??? i dont say multiprocessing is not needet becuse it is.. but way is AMP not cool anuf ?

With AMP each program needs to be written to take advantage of it, and dependng on the design a limited number may be able to benefit at any one time, with SMP all programs gain for "free", (but those designed for multithreading gain much more ofcourse).

_________________
BroadBlues On Blues BroadBlues On Amiga Walker Broad

@asymetrix

Quote:

So the kernel has been rewritten again ? It was first rewritten years back : AmigaOS 4.0 Prerelease Update #3

Well it's been in contiunous development since then, does each step count as a rewrite?
The fact of the first rewrite no doubt made the sunsequent updates easier, that was the point of it (or one of the points).

Quote:

with all these versions no one is keeping kernel timings/ test speed logs, no regression testing for efficiency or compatibility.

What makes you think that? If anything legacy compatabilty has improved over time, with the occasional quirk here an there, some considerable effort was made at one stage to keep certain 68k programs working, recently there have been few issues, but I do know for certain that gfx library bugs were fixed to assist in getting PPaint running correctly.

100% compatabilty has never been the goal (progress would then be impossible) but that doesn't not mean compatabilty is discarded completely.

Last edited by broadblues on 27-Feb-2015 at 04:25 PM.

_________________
BroadBlues On Blues BroadBlues On Amiga Walker Broad

@asymetrix

The latest info presented on the blog was NOT an update. It was a short tutorial on multi-core SMP along with the same info that was announced back in 2013 regarding forbid(). Nothing new at all. It's too bad that people are writing in between the lines of this blog so to speak.

Last edited by ferrels on 27-Feb-2015 at 04:21 PM.

@broadblues

Quote:

broadblues wrote: @retro Quote: way is smp soo importent ??? i dont say multiprocessing is not needet becuse it is.. but way is AMP not cool anuf ? With AMP each program needs to be written to take advantage of it, and dependng on the design a limited number may be able to benefit at any one time, with SMP all programs gain for "free", (but those designed for multithreading gain much more ofcourse).

No. SMP emphasizes that the system is symmetric - each processor an equal peer*. What you are referring to can be called a coprocessor system but that is more specialized than an AMP or simply MP system. Gains for a MP system are in throughput and not every task will run faster**.

A system where the scheduler is run only on one processor isn't SMP.
A system where interrupts are handled on a subset of processors isn't SMP.
A system where device drivers are handled on a subset also isn't SMP.

(* http://en.wikipedia.org/wiki/Symmetric_multiprocessing )
(** http://en.wikipedia.org/wiki/Amdahl%27s_law )

@megol

Quote:

No. SMP emphasizes that the system is symmetric - each processor an equal peer*. What you are referring to can be called a coprocessor system but that is more specialized than an AMP or simply MP system.

There are so many varaints, I suppose my AMP defineition does fit the coprocessor version best.

Quote:

Gains for a MP system are in throughput and not every task will run faster**.

True, the overall system sgould run faster (or what's the point?) but a single task can only run as fast as a single core (simplisticly speaking) , and so a task (job of work not necessarily CPU task) that was taking 95% CPU won't go much faster, unless designed to be multithreaded.

Quote:

A system where the scheduler is run only on one processor isn't SMP. A system where interrupts are handled on a subset of processors isn't SMP. A system where device drivers are handled on a subset also isn't SMP.

Perhaps not but the important aspect is application software doesn't need to care about which cores it's running on. If we end up with a "nearly symetric" system becuase of faling foul of one of the above split hairs, that's good enough for me. An if it allows those that say it can't be done to save face when it is, then I'm sure that'll be cool too


(* http://en.wikipedia.org/wiki/Symmetric_multiprocessing )
(** http://en.wikipedia.org/wiki/Amdahl%27s_law )

_________________
BroadBlues On Blues BroadBlues On Amiga Walker Broad

@All

SMP for OS4 will run an instance of the scheduler for each cpu core?

Or it is rather 1 instance of the scheduler being executed on a dedicated cpu core?

_________________
SAM440EP-FLEX @ 733 Mhz, AmigaOS 4.1 Update 1

@Massi

Based on the article, a scheduler for each core.

_________________
BroadBlues On Blues BroadBlues On Amiga Walker Broad

@broadblues

OK, thanks.

_________________
SAM440EP-FLEX @ 733 Mhz, AmigaOS 4.1 Update 1

@Hondo

Okay OS4 catched up to AROS in regard of multiprocessing. Even the approach is like with "silly SMP". Nice study case though.

_________________
My programs: via.bckrs.de
MorphOS user since V0.4 (2001)

@Jose,Sundown

yup love u guys!.

@all
or maybe this is freidens way of saying covertly without saying it, reguardless of bankrupcy, work continues ?

_________________
Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway.

I serve King Jesus.
What/who do you serve?

@pavlor

Quote:

This is AmigaOS, not MorphOS.

You apparently dont realize that Morphos team are the fathers of the best software for AmigaOS right? Chris Hodges, Frank Mariak, Stuntzi, etc etc

as a matter of fact morphos was going to be "OS4" also "OS4" wouldnt even exsist if it wasnt for Phase 5.

You would think that you guys would finally wake up, but I guess its better to live in a fantasy than reality.

@realize,

Of course you realize that everyone here will have a good laugh at your expense when it's done.

Chris

as a matter of fact morphos was going to be "OS4" also "OS4" wouldnt even exsist if it wasnt for Phase 5.

You would think that you guys would finally wake up, but I guess its better to live in a fantasy than reality.


am i a racist if i say my general perception of french pepole is that there are arrogant ???




bon appetite
https://encrypted-tbn2.gstatic.com/images?q=tbn:ANd9GcT0y_SzWYC07nOiv1hfxbWC14Vw7-bQH1CmOACjdW_tcXvNaXqX

@Zylesea
Unfortunately it didn't, and that was my point. The progress status is the same is it was in 2013, a new scheduler runs on a single core.

_________________
This is just like television, only you can see much further.

@BSzili

Quote:

The progress status is the same is it was in 2013, a new scheduler runs on a single core.

The next step is to have each core in the development system (currently, the X1000) to run the scheduler. Test code will then start tasks on the different cores and see how they behave. We have already experimented with this and the results look promising. The tests basically showed that the lockout mechanism for Forbid works as planned.

Last edited by pavlor on 28-Feb-2015 at 08:31 AM.

@pavlor

How is the text you quoted contradicts what I've said?

_________________
This is just like television, only you can see much further.

@BSzili

You wrote:

a new scheduler runs on a single core.


In the blog post:

... each core ... to run the scheduler.

AND

We have already experimented with this and the results look promising.

@pavlor

Please don't cherry pick when quoting others. They experimented with starting tasks on the additional cores, which is not the same as running the scheduler on each core, their stated goal.

_________________
This is just like television, only you can see much further.

@BSzili

Well, then our understanding differs there.