Dr. Garry Hare Amiwest Speech

[Waccoon]

Ok guys, I've just had to delete a tonne of posts.

Just out of curiosity, what's the ratio of decent posts VS trolling posts?  I doubt it's much different than any other platform discussion BBS.

let's get back to the roots of SUPPORTING THE AMIGA.

I'm not sure the commercial companies involved are much attuned to making a product with longetivity.

I'm waiting to see the product demonstrations of KMOS.  If all they do is keep things a secret (ie, not opening a website), they're just as doomed as all the previous Amiga owners, no matter what grand ideas they may have.

Dammit, I don't want to hear any more about effeciency, small size, innovation, technical supiriority of PPC or otherwise, or any other such bull****.  I want to see DE in action, and a full explanation of how it will solve the biggest problems with existing platforms, like security, usefulness, and maintenance.

Perhaps Bill McEwen and Fleecy made the pitch of their lives!

I wonder what happened to those two.  I find it hard to believe those guys will ever work in the IT industry ever again.

Then again, I thought the same thing about Mehdi Ali, and he is very much alive, in a management position, and making obscene amounts of money despite his complete incompetence.  Life sucks.

[Waccoon]

Moss and McEwen are employees of the new company.

Must've been part of the deal.  Personally, I'd fire their sorry butts.

[Waccoon]

And I wonder.. everyone knows DE is dead. There are no developers anymore, there is no new software anymore, most of DE developers left or went for an A1.

Heh.  I couldn't even become a developer.  Amiga Inc. must have been on vacation when I bought my SDK.

If I would employ Garda, I would become bplan?

No, but if you shuffle his debts and contractual obligations to avoid forclosure...

I haven't reached any conclustions on KMOS.  I was  happy when they announced they had bought OS4.  But, when they announced they had bought ALL of Amiga Inc., well AFTER buying OS4, that's just too fishy for me.  I'm in agreement with KennyR (at least to a point).

Yes, and what you'd need to do is create a low-level hardware interface that is basically the VP run-time engine. The OS would run on top of this.

I'm all for writing apps and libraries, and maybe even drivers on a VP engine, but the idea of writing an OS and other low-level parts in VP is pretty scary.  Tools like the shell, for example, should be in VP.  The kernel and low-level drivers should always be native -- hardware access is what they're supposed to handle.

Again, it's important to distinguish between the kernel and the OS.  The kernel is what provides the interface to the hardware and the low-level drivers, the OS consists of the tools, shell, GUI, and high-level drivers.

So what. CPU's these days are ridiculously fast.

It's not so much the speed.  Try doing to a CPU what DirectX did for graphics cards.  That's the point of VP.

Ironic.  Didn't the Amiga pioneer the concept of coprocessors?  Why do you think APIs like OpenGL exist?  When the were first introduced, they just added overhead.  Today, APIs are essential for coprocessing, hardware acceleration, and architecture independence.  Funny how slow the CPU world has been to adapt the practices of the graphics industry.  The Intel monopoly is a big factor.

The use of that extreme virtualization would also make it impossible to take full advantage of HW acceleration (like 3D).

Full advantage, yes, but that's like comparing assembly to C.  There's pros and cons for tryint to take full advantage of the hardware.

Personally, I think if you make a good architecture, performance will just fall into place.  Write your software normally and let VP and/or the compiler worry about the rest.  People obsess way too much over performance.

In the long run, optimization might actually reduce performance, too.  But, I won't get into how that works.  :-)

VP/intent does not support memory protection (MP is planned for AOS4.x) it never will.

My presumption is that each thread will have its own independent VP hook, so the underlying OS handles memory protection.  If all VP threads share a single VP thread, then that's a very bad idea.

Maybe cell phones don't need memory protection if they run 1-2 processes, but when they start to multitask and run hundreds of threads, the need for memory protection will become obvious.  Are we retrograding to the days of MS-DOS and Classic Macintosh?

The name "Amiga" has been raped by executives in our community, but Jay Miner's memory stands still unharmed and is respected by very many people, including me.

Keep in mind that the millions of the people who really made the Amiga special are long gone.  Only the hundreds of totally obsessed fanatics remain.  That's why I don't like the choice to use PPC.  I think most people who have converted to x86 would agree that a $200 PC mobo is a much better value than an old $800 PPC board, and my P4 is reliable and dead silent, too.

And, yes...  I'll stop talking about CPUs.  Nobody really cares anymore.

There is always the delete button. As I see it is very popular today.

I don't argue with the admins, but I wish whoever was deleting posts would just edit them down to a placefiller, instead.

[Waccoon]

It is not sensible to run all of the OS on top of VP.

True.  Unfortunately, people still see an "OS" as a monolithic platform, so nobody with tecnical experience will explain which parts should be native and which should be VP to ordinary people.  Most still think web browsers are part of the OS.

Again. IMO: we should outperform Wintel rather than be slower on purpose.

Does that explain why Java has exploded in popularity?  That's about as sluggish and glitchy as you can get!

You also have to figure that Windows programmers aren't exactly concerned about reliablity, and most low-level drivers are so hell-bent on performance, there's little we can do to compete.  Do you think any Amiga graphics drivers can go up against the likes of nVidia or ATI?  They pour millions of dollars into making those drivers as fast as possible, and everything runs as close to the hardware as possible in kernel mode.  Poke fun at Windows all you want, but most drivers are designed to slice through the bloat and get as close to the hardware as possible.  I doubt you can exceed the performance of Windows in many ways.

For audio & pleasant sound experience, near "real time OS" performance is required.

Realtime has nothing to do with speed.  It has to do with latency.  Doom3 is doing all its sound mixing in software with the CPU, instead of using the hardware features of, for example, Audigy cards.  Why waste CPU cycles to do that?  For one thing, you can mix it just as you like, instead of limiting yourself to the capabilities of the hardware.

Unless it comes to games, most people don't care about performance except on paper.  Responsiveness counts.

I'm starting to hate using that word, though.  Responsiveness has become a major buzz word as of late, but nobody really seems to know what it means.  Fast screen refreshes in OS4 is only a small part of the whole story.

IIRC, not even the FPU "problem" is solved yet. Intent FPU is slow.

I thought they were allowing you to use native "tools" to optimize your code, and only force VP if no optimized tools were available?  I know little about FP, though, other than the fact that each CPU has different levels of accuracy and will return unpredictable numbers.

I think making a SIMD VP would be totally insane attempt. (intent gives the possibility to use native executables amongs VP ones, perhaps that could be used...)

Yeah, that's the tools I was talking about.  Any programmer who wants to release code on multiple systems MUST break down the critical stuff into native code, and use generic alternatives as a last resort.

The problem is, many do no bother.  Even Perl scripts are not always portable because people keep using home-made routines using symlinks and UNIX shell calls, instead of using proper libraries.  It shocks me how many people still try  to parse CGI data themselves instead of using CGI.pm!

IMO: AOS would not perform like AOS should if it is run on top of virtual machine.

That depends what parts of it are running on a virtual machine.  A VP CLI and GUI would make little difference in terms of performance if the graphics drawing routines are still native.

As we all know, if you write AmigaDE software it will be compatable with every platform

Potential to be compatible?  IMO there is no such thing as a fully portable application.  There's always some nut that wants to screw everything up.  Hell, we can't even get webpages to look the same in every browser!

I think DE should offer a set of generic and native tools and let the developer decide what to use.  If they want to enjoy the profit margin of many different markets, they'll have to program accordingly.  If they need ultra performance on a particular CPU, there should be nothing to prevent them from using native code.  That's what I don't like about Java.  Write Once Run Everywhere is a dream that doesn't exist.  Java is notorious for cross-platform bugs, which makes me wonder why everyone likes it so damn much.

On top of this, the Amiga set-top boxes will act as wireless routers and automatically save and secure(encrypt?) data that doesn't fit on your desktop hard disk

Funny how they don't call them file servers anymore.  They're all set-top boxes.

HE ALSO SAID THAT ALL DEVICES THAT RUN AMIGAOS4 CAN BE BADGED AS AN AMIGA!!!!

Well, at least someone realized that the Amiga is not just about slick hardware...

I am soooooo going to puchase the new SDK

Don't you want to see the product, first?  The first SDK was simply horrible - very raw and terribly organized.

Those Crusoe devices are getting incredibly small, powerful and SLICK

Do they have native instructions, now, or are they still only emulating x86?  Code morphing is still a major performance hog, and I haven't heard of how good Transmeta processors are working native.

OS4 being a closed source platform is actually more appealing to companies.

True.  If you have no legal right to make changes to the code, there's really no point in having it open source, as it'll probably take just as long as a closed source product for fixes to be implemented.

Oh yeah, and open source does not mean fewer bugs.  Fewer bugs are because of the dedication of the developers and their debugging techniques.  I can give you a long, long list of open source projects that are swamped with bugs -- seeing how the developers are all on vacation and I always end up having to fork the code into my own, new project just to fix all the damn problems!

Security through Obscurity as Ben put it

I won't raise my opinion on that comment again (as my last post got deleted).  I will say he'll have to do much, much better than that.  Security is not magic, as any script writer will tell you, and as we move to more virtualized processors running on dynamicly compiled or translated code, security will start becoming a very, very big problem.

As an example, I still see people doing lots of evals in their Perl code, such as:

$val = <<...something which includes CGI input...
EOF

eval( $var, {...} );

These people are insane!  Never, ever use eval except to force run-time compilation to improve performance.

I felt like it was soooo clear, but maybe I heard something you guys did not.

I want to see a working prototype.  If they need investor money, they're not getting it from me buying an SDK.

[Waccoon]

ARM

Well, I'm not sure any processor is going to hold a monopoly seeing how practically everything written for cell phones these days is done with Java.  My brother-in-law works for Nokia, and Java is pretty much the only language they use.

It'd be much easier to overtake Symbian on PDAs than to displace Windows on the PC.  I don't follow the PDA scene, but what about TabletPC and PocketPC?  Those don't run on ARM, do they?