AGA integration in Apollo core : work in progress

[meynaf]

Originally Posted by idrougge

Come on, be realistic. How many applications use an MMU? Can you name any?

I wrote nothing about applications, i wrote about debugging.
Enforcer and similar tools are very useful for it. But you must be a coder to know that.

Originally Posted by idrougge

Hyperthreading, on the other hand, will immediately boost all Amiga applications thanks to FPGA power.

It will boost nothing at all. Benefit is null if the app isn't written to work in parallel - and of course none is on the Amiga. Besides, as AmigaOS does not support this, it can only break applications.
In fact hyperthreading is the most useless thing ever added to that core.

[E-Penguin]

It's an outrage! The Vampire is the worst thing to happen to all of humanity! Ever! It's entirely rubbish until it gets an FP...

oh...

um...

cats.

[Overflow]

Originally Posted by meynaf

It will boost nothing at all. Benefit is null if the app isn't written to work in parallel - and of course none is on the Amiga. Besides, as AmigaOS does not support this, it can only break applications.
In fact hyperthreading is the most useless thing ever added to that core.

Well, if applications are written to support it, the useless statement doesnt apply I gather.
Thats obviously assuming developers are intrested in advancing the Apollo perspective. As far as OS goes; how about AROS?

As a non coder Im intrested in posts that goes into details, even tho I dont comprehend the deep details.
Simply saying its useless isnt helpful, just like e penguins post.

And IF you decide to accomodate my query, could you expand the discussion to encompass the usefulness IF developers decide to support hyperthreading and SoftFPU. (Aos and aros).

Thanks!

[Romanujan]

Originally Posted by meynaf

In fact hyperthreading is the most useless thing ever added to that core.

Why? According to the Apollo team it was easy to do, took almost no additional FPGA space, allows to speed up at least the Vampire drivers. Looks good to me.

Originally Posted by idrougge

Come on, be realistic. How many applications use an MMU? Can you name any?

Even if it is only WHDLoad, it is worth having it. But there are more - Enforcer, GuardianAngel, some video drivers for ShapeShifter. I agree it is not the highest priority, but still...

[meynaf]

Originally Posted by Overflow

Well, if applications are written to support it, the useless statement doesnt apply I gather.

True, but that's IF applications are written to support it, which is very iffy.

Originally Posted by Overflow

Thats obviously assuming developers are intrested in advancing the Apollo perspective.

But are they ?

Originally Posted by Overflow

As far as OS goes; how about AROS?

By how much is AROS slower than regular AOS ? Compare that to the potential of hyperthreading, which can not by construction be better than 2x.
And Amiga mostly depends on single threaded performance anyway.

Originally Posted by Overflow

As a non coder Im intrested in posts that goes into details, even tho I dont comprehend the deep details.
Simply saying its useless isnt helpful, just like e penguins post.

AmigaOS is designed in such a way that makes this kind of addition basically not usable. It's like running old DOS programs on a multicore cpu : no benefit at all.

Originally Posted by Overflow

And IF you decide to accomodate my query, could you expand the discussion to encompass the usefulness IF developers decide to support hyperthreading and SoftFPU. (Aos and aros).

Developers won't support that.
This is where the team completely misses the point - they need extra work from developers where they should instead have made their life easier.

Originally Posted by Romanujan

Why? According to the Apollo team it was easy to do, took almost no additional FPGA space, allows to speed up at least the Vampire drivers. Looks good to me.

No doubt it was easy to do, but with current Amiga architecture, it simply can't be used for anything but a handful programs.

[Overflow]

The "But are they?" argument is double-edged tho.

Its hard to support a feature thats not available. Now it is, or in case of SoftFPU work in early progress. As its developed I suspect its performance will improve.

As a non-developer my relation to MMU is non-exsistant, so Ill pass on that topic.

[TrashyMG]

@meynaf:
From what I picked up on IRC, some device drivers (IDE, SD.. etc) which will reside on the vampire kickstart are being written to take advantage of the hyperthreading.

About developers, I've already seen several people want to hop in and develop new things using the vampire features. Things are moving weather you like it or not.

Also leaving this here:
[ Show youtube player ]

[demolition]

Originally Posted by idrougge

Come on, be realistic. How many applications use an MMU? Can you name any?

Hyperthreading, on the other hand, will immediately boost all Amiga applications thanks to FPGA power.

HT won't do anything unless the OS supports multiple cores which Amiga OS does not.
MMU is required for Unix and in Amiga OS it is mainly useful for debugging.

[OlafSch]

Originally Posted by meynaf

I wrote nothing about applications, i wrote about debugging.
Enforcer and similar tools are very useful for it. But you must be a coder to know that.



It will boost nothing at all. Benefit is null if the app isn't written to work in parallel - and of course none is on the Amiga. Besides, as AmigaOS does not support this, it can only break applications.
In fact hyperthreading is the most useless thing ever added to that core.

and I do not understand why it is your problem?

Kolla is moaning that he bought two vampires because of in his view promised FPU and MMU and feels betrayed but when I advised him to sell the cards because there are lots of users interested outside who would be happy to have one he does not react

You are vampire owner? If not do not buy it, problem solved. If you own it sell it, one unhappy user less and another one happy

Gunnar and the others have their agenda they are implementing now step by step. If you like what they offer and trust them then buy one, if not leave it. The noise around only nerves average users, it will change exactly zero regarding what is developed and how priorities are

[OlafSch]

Originally Posted by Overflow

The "But are they?" argument is double-edged tho.

Its hard to support a feature thats not available. Now it is, or in case of SoftFPU work in early progress. As its developed I suspect its performance will improve.

As a non-developer my relation to MMU is non-exsistant, so Ill pass on that topic.

regarding MMU that is exactly the problem. How devs admitted themselves it would slow down the core and for users it makes zero difference. It is only useful to use certain debug tools when you are a developer. Priority for Gunnar is to implement features that benefit users. Resources on FPGA are limited, developer resources are limited, time is limited. To me it sounds rather reasonable to concentrate it on features users need.

[Overflow]

Olaf, I think we shouldnt decend into " you are only allowed to voice concerns if you own/or plan to own a Vampire". Thats not a constructive attitude, nor does it invite a friendly enviroment.

If we want Vampire to be respected, we need to accept/respect other opinions.

[robinsonb5]

Originally Posted by TrashyMG

@meynaf:
From what I picked up on IRC, some device drivers (IDE, SD.. etc) which will reside on the vampire kickstart are being written to take advantage of the hyperthreading.

At first I couldn't see the point of this at all - but the more I think about it, the more I can see the possibilities. Think of the second thread less as a second CPU for multiprocessing as we normally think of it, and more as a general purpose co-processor.

The IDE driver can use it to service IDE requests which would be the next best thing to true DMA support (assuming that's not already present?), the graphics driver could use it to provide blitter-like functions in software, WriteChunkyPixels() could be patched to use it. I'm sure there are plenty of other possible uses too. AHI driver that offloads channel mixing, perhaps?

Even if the OS never gets the requisite multiprocessing support for user applications to benefit, and even if no user-level apps are written to take advantage of it, it could still prove very worthwhile.

[OlafSch]

Originally Posted by Overflow

Olaf, I think we shouldnt decend into " you are only allowed to voice concerns if you own/or plan to own a Vampire". Thats not a constructive attitude, nor does it invite a friendly enviroment.

then why should I moan on a BMW forum about what I would like to have in a new model if I have no intention to buy this new model?

If I would do that people would think I am completely crazy, wouldn´t they?

Vampire cards and future standalone are new products, I do not need to buy if I do not like the company/developer behind, the product or both. I do not see the difference here

[OlafSch]

Originally Posted by robinsonb5

At first I couldn't see the point of this at all - but the more I think about it, the more I can see the possibilities. Think of the second thread less as a second CPU for multiprocessing as we normally think of it, and more as a general purpose co-processor.

The IDE driver can use it to service IDE requests which would be the next best thing to true DMA support (assuming that's not already present?), the graphics driver could use it to provide blitter-like functions in software, WriteChunkyPixels() could be patched to use it. I'm sure there are plenty of other possible uses too. AHI driver that offloads channel mixing, perhaps?

Even if the OS never gets the requisite multiprocessing support for user applications to benefit, and even if no user-level apps are written to take advantage of it, it could still prove very worthwhile.

Gunnar wrote that they are already working on components that will use it so I assume benefit will become obvious when new core is available

[Overflow]

Well, this is an amiga forum, and we are limited amount of users.

Ive used Dan Wood/Kookytech as an example so many times its silly, but Ill do it again; its possible to have a positive attitude to everything/everyone within the Amiga community, even if we have different opinion.

Exchange of ideas AND difference of opinions can/will usually produce new ideas or methods. AND when you produce a product, its usually to please/make it useful for a end user. That means people that are still on the fence about a product.

So discounting xyz persons opinion just creates friction and negativity. If you start a post with "your opinion doesnt matter", you know the rest of the post you make will be read with hostility, and the rest off the thread will spin out of control like pretty much every thread in Amigaworld/amiga.org.
Its that a direction you want EAB to go?

[OlafSch]

Originally Posted by Romanujan

Why? According to the Apollo team it was easy to do, took almost no additional FPGA space, allows to speed up at least the Vampire drivers. Looks good to me.



Even if it is only WHDLoad, it is worth having it. But there are more - Enforcer, GuardianAngel, some video drivers for ShapeShifter. I agree it is not the highest priority, but still...

WHDLoad not needs MMU

[grond]

Originally Posted by robinsonb5

The IDE driver can use it to service IDE requests which would be the next best thing to true DMA support

Actually a CPU servicing the request is better than DMA as a CPU is more versatile and could do some extra data processing while copying data around. A CPU is also more versatile than a blitter. It can do everything a blitter can do plus more. The reason why we all have a heart for blitters, DMA controllers and more is because they work in parallel to the CPU. But once we have more than one CPU, the CPU will always beat dedicated hardware. I know this is difficult to accept because it contradicts decades of single-core user experience.

[meynaf]

Originally Posted by TrashyMG

@meynaf:
From what I picked up on IRC, some device drivers (IDE, SD.. etc) which will reside on the vampire kickstart are being written to take advantage of the hyperthreading.

Hmm must i understand that these drivers currently use too much cpu ?

Originally Posted by TrashyMG

About developers, I've already seen several people want to hop in and develop new things using the vampire features. Things are moving weather you like it or not.

So what will it be ? There will be ONE program making use of this, just like ammx. Not impressive.

Originally Posted by OlafSch

and I do not understand why it is your problem?

It doesn't have to be my problem for me to point out a few bad details...

Originally Posted by OlafSch

Kolla is moaning that he bought two vampires because of in his view promised FPU and MMU and feels betrayed but when I advised him to sell the cards because there are lots of users interested outside who would be happy to have one he does not react

So the team is losing customers because of promises not fulfilled, and doesn't care. Replying "then sell it back" isn't the best of marketing methods...

Originally Posted by OlafSch

You are vampire owner? If not do not buy it, problem solved. If you own it sell it, one unhappy user less and another one happy

I wouldn't buy one even if i wanted to - no trapdoor V1200 available.
But as former team member I have a few things to say, sorry if you don't like it.

Originally Posted by OlafSch

Gunnar and the others have their agenda they are implementing now step by step. If you like what they offer and trust them then buy one, if not leave it. The noise around only nerves average users, it will change exactly zero regarding what is developed and how priorities are

So you've just admitted they don't listen to anyone...

Originally Posted by OlafSch

then why should I moan on a BMW forum about what I would like to have in a new model if I have no intention to buy this new model?

If I would do that people would think I am completely crazy, wouldn´t they?

Vampire cards and future standalone are new products, I do not need to buy if I do not like the company/developer behind, the product or both. I do not see the difference here

But, if you would buy a BMW if they listened to your suggestions, would you still close your mouth ?

Originally Posted by Overflow

Well, this is an amiga forum, and we are limited amount of users.

Ive used Dan Wood/Kookytech as an example so many times its silly, but Ill do it again; its possible to have a positive attitude to everything/everyone within the Amiga community, even if we have different opinion.

Exchange of ideas AND difference of opinions can/will usually produce new ideas or methods. AND when you produce a product, its usually to please/make it useful for a end user. That means people that are still on the fence about a product.

So discounting xyz persons opinion just creates friction and negativity. If you start a post with "your opinion doesnt matter", you know the rest of the post you make will be read with hostility, and the rest off the thread will spin out of control like pretty much every thread in Amigaworld/amiga.org.
Its that a direction you want EAB to go?

+1

Originally Posted by OlafSch

WHDLoad not needs MMU

No software "needs" hyperthreading, so your argument can be returned to you...
And Whdload does not "need" MMU but it can use it and is safer that way.

[meynaf]

Originally Posted by grond

Actually a CPU servicing the request is better than DMA as a CPU is more versatile and could do some extra data processing while copying data around. A CPU is also more versatile than a blitter. It can do everything a blitter can do plus more. The reason why we all have a heart for blitters, DMA controllers and more is because they work in parallel to the CPU. But once we have more than one CPU, the CPU will always beat dedicated hardware. I know this is difficult to accept because it contradicts decades of single-core user experience.

There is no extra data processing to be done for IDE and the like, it's just data copying, and you forget that the CPU can't use all available memory bandwidth, while DMA can use a lot more.

[TrashyMG]

@Meynaf, why do you care? just don't bother with it.