@Lou

Quote:

Gunnar's solution to everthing is to add more cpu instructions creating more incompatibilities...then wondering why there is no GCC support. This is a CISC-to-the-extreme when long a go everyone else realized that a cpu is good a cpu things and other chips are better at specific things and can work independently speeding up the system as a whole.

I hear you, but you know very well why there is 'no' GCC support, and it's not because someone wants to extend the architecture.

Have you seen the Intel compiler intrinsics list? I thought of it in this context because "AMMX" is (more than) somewhat modeled after MMX, of course - point being Intel didn't have any hesitation to utterly transform not only the fetch, decode, and execution of the existing instruction set (in a strictly compatible sense), they kept piling on - and on and on:

Intel Intrinsics Guide

If you want to use the intrinsics, you call them in your code as if they were regular functions (but for the register and data-type constraints). The gnu compiler maintainers did not rework the instruction rendering and optimization algorithms to (implicitly) employ the MMX, SSE, AVX, etc functions - you explicitly invoke the intrinsics you want and the compiler emits the appropriate binary encoding for the corresponding machine instructions. It is a perfectly reasonable approach, though far from ideal.

My point is, a similar approach could be taken in the case of the 68080.

Is there a better way? Doubtless, but I don't see anyone rising to the challenge - and honestly, I don't see the point without a top-to-bottom re-creation of the hardware and software to properly address a number of shortcomings. But then, that wouldn't be an Amiga either, by the standards I'm inferring in this thread.

Quote:

The Amiga was about custom chips working in unison with the cpu not about the cpu doing everything.

I guess the operative word here, I hate to say, is "was." Maybe you or some other will win the argument(s), but I suspect it will not have proven or accomplished anything.

Respectfully,

Snorg

@ErikBauer

Quote:

Define partially compatible: more and more 68K demos, games and programs are running upon SAGA. Thee harware registers are there, the blitter blits, the copper cops and audio chip plays sounds executing commands directed to Paula. Where's the problem?

vs

"OCS/AGA: Many games and demos work perfectly, but not 100%."

http://apollo-core.com/knowledge.php?b=1¬e=23461

Some gvb puppets has as usuallly problem with reality.
Of course there are some natami/apollo/vampire users/developers who learn that lies about performance and compatybility will be soon revealed after people buy vampire.

saga contain nothing from original Commodore chips.
saga does not share history with original Commodore chips.
saga is not evolution od original Commodore chips.
It is just made from scratch partially compatible implementaion of Commodore chips.
It is NG and will be treated as NG.

Last edited by ppcamiga1 on 06-Oct-2019 at 01:48 PM.

@Snorg

Quote:

Maybe you or some other will win the argument(s), but I suspect it will not have proven or accomplished anything.

I'm ok with this. I use and will use vampire as only fast cpu and RTG card.
saga is crap at 3D, so someone made worth of use graphics card for vampire users.
We will relive year 1997 yet another time. For me it is ok.

@ppcamiga1

Quote:

ppcamiga1 wrote: @ErikBauer Quote: Define partially compatible: more and more 68K demos, games and programs are running upon SAGA. Thee harware registers are there, the blitter blits, the copper cops and audio chip plays sounds executing commands directed to Paula. Where's the problem? vs "OCS/AGA: Many games and demos work perfectly, but not 100%." http://apollo-core.com/knowledge.php?b=1¬e=23461 Some gvb puppets has as usuallly problem with reality. Of course there are some natami/apollo/vampire users/developers who learn that lies about performance and compatybility will be soon revealed after people buy vampire.

I can't see your point. My sentence and theirs are mostly the same, just worded differently
Care to elaborate?

Quote:

saga contain nothing from original Commodore chips. saga does not share history with original Commodore chips. saga is not evolution od original Commodore chips. It is just made from scratch partially compatible implementaion of Commodore chips. It is NG and will be treated as NG.

Hummm... yes? Seems pretty obvious

_________________
God created Paula so that Allister Brimble and Dave Whittaker could do music

Check my Amiga gameplays (ITA)!

As I wrote some gvb puppets has as usually problem with full and partial compatybility.
Their 68k wunderwaffe is not as good as they think.
The some thing was with FPU. It was also declared as fully compatible, but soon it was clear that works only in selected apps.
And Amiga Os calculator was not one of them.
future customers should be aware of this.

Last edited by ppcamiga1 on 06-Oct-2019 at 04:35 PM.

@ppcamiga1

Well, all your claims should not be new to someone that carefully reads Apollo Core threads
And should not be a surprise to someone that understand just a little how computers work

For someone who lived in first person the passage from an A500 to an A1200, some compatibility issues are granted every time you make a step forward.
Windows grants more compatibility through evolution than Amiga because of it's abstraction layer, and even then, it is not 100%, Amiga does not have that layer and so badly written "Heavy Hardware Banging" programs fail to run even if you set the Chipset to NTSC instead of PAL, figure out in front of a new CPU that did not exist when the program was written.

On top of that, everything is backward engineered and re-created from scratch into an FPGA.
So, while on paper 68080 can run every instruction of the whole 68K family, timings and behaviours can surely change at very low level, thus creating incompatibility.
Add FPGA programming bugs on top of that and for sure you don't have a machine that is 100% compatible with any Native Amiga.
But then, even between native Amiga models, 100% compatibility was not ever granted (Try running "Kult" on a 1.3+ A500 or "Dungeon Master" on an A3000 for instance)

Now, my wonder is: how all this can be such a big issue to somebody when everything seems so clear?

Last edited by ErikBauer on 06-Oct-2019 at 05:50 PM.
Last edited by ErikBauer on 06-Oct-2019 at 05:49 PM.

_________________
God created Paula so that Allister Brimble and Dave Whittaker could do music

Check my Amiga gameplays (ITA)!

@ErikBauer

There were big problems with running amiga games almost exclusively made for Amiga 500 on anything better than Amiga 500.
Future customers should be aware that v4 is not classic, saga is not evolution of original chips made by Commodore, timigs are different and they may face problems running software (games) for classic.

It’s clear vampire v4 isn’t best choice for guys who want play a500 games.
I hope instead it will be much more, a NG Amiga based on 68k instruction set with some simd and gpu like calc extension inside.
I hope there will be a new os like aros and possibility to install classic Amiga os, morphos and even better os4, to maintain api compatibility with all recent developed apps who needs only a recompile to run on vampire.

This is all I want about vampire project.
Maybe, at least some of these things, are also the same vampire team pursued in these years..

_________________
Pegasos II G4@1GHz 2GB Radeon 9250 256MB
AmigaOS4.1 fe - MorphOS - Debian 9 Jessie

@Fl@sh

in an outlook aros might potentially offer better compatibility across the range of the genuine kickstarts then each of those against beach other. already now you can run 1.x-3.x applications next to each other.

@wawa

Yes, I know.
I hope aros and vampire will not be another product to make retrocomputing.
For this we already have a lot of good choices for free and personally I have zero interest on it.
I’d like amiga os, from 3.x to 4.x, api compatibility and new features to make easy porting from old Amiga code and produce new software with faster cpu and other coprocessors or expansions.
Nothing more nothing less

_________________
Pegasos II G4@1GHz 2GB Radeon 9250 256MB
AmigaOS4.1 fe - MorphOS - Debian 9 Jessie

@Lou

Quote:

But the SAGA "core" is not an enhanced chipset or even a chipset at all. It's another thread on the cpu doing functions that the PAULA and the blitter would normally do. Marketing vs reality. SAGA is already RTG+AHI.

In the chipset sense it's integrated in the one FPGA chip. But is it really a thread as the CPU is running in an FPGA. In any case SuperAGA is enhanced over AGA offering greater resolutions, better sprites, better audio, etc. Actually, I just did a a search, and the information on SAGA is hard to come by. On official Apollo site they don't give any info at all. Except for the extra SAGA registers. Nice site but they don't even list what what SAGA actually does. Now I want to know what's so super about SuperAGA?

Quote:

Why not just let a real gpu chipset do it's job rather than create another mis-guided 68K version of the CELL processor but with only 1 or 2 SPUs?

Because that wouldn't be an Amiga chipset. They don't do planar, don't do sprites, and most of all don't have Amiga compatible hardware registers at $DFF000. Besides, it's all been done before. Picasso, CyberVision3d, Mediator.

The same thing was done for OS4 and Amiga people hated it. What I mean is "only real" Amiga people hated it. I've spoken to a few Amiga people, a few into the Vampire, and when talking about the AmigaOne look down on it because it has no Amiga compatible graphics included. Now, they aren't going to program it themselves. But, in reality, they don't consider it to be Amiga because there's no Amiga graphics chipset mapped to $DFF000!

@NutsAboutAmiga

Quote:

"Even if it had a PCIe slot for video expansion" who says it has to be for video expansion, there is lots of interesting stuff to plug into PCIe, USB3, sound cards, SATA controllers, PCIe active raiser cards.

I suppose it doesn't have to be for video but exclusively basing the graphics on AGA with extras is being questioned here. Which I thought was the whole point of it. For other things you mention, great, if drivers can be written.

@Lou

Quote:

As I said before, an 020 with FASTRAM can blit faster than Agnus...so guess what's doing the blitting...

What kind of blitting? I mean, an aligned block of graphics could easily be blitted faster by CPU. But what about CPU copy of a bitmap into an arbitrary bitmap position?

Given this would require shifting, likely 020 optimised bit field or rotate instructions would make this easier. With masking of first and last word. There are also multiple source and destination to consider as well as different logic operations.

There is also line drawing with patterns the CPU would have to emulate and here I think using the CPU to blit a rotated line pattern would be slightly taxing. Even if a C16 had to line draw with only CPU by comparison.

In lower resolutions 020 is little faster in line drawing in 256 colors (about 2%) than blitter, and three times faster i clearing bitmaps.
In resolutions like 640x480 or 800x600 256 colors AGA take all bandwith, blitter has not access to chip RAM during display and 020 is always faster.

Last edited by ppcamiga1 on 18-Oct-2019 at 08:36 AM.

@all

Are there any recent Sysspeed Vampire V4 benchmarks ?

I am curious about the AGA output performance over HDMI with higher resolutions 8bit.
As we know 640x512x8 bit is slow on A1200.

Last edited by bennymee on 19-Oct-2019 at 08:23 AM.

@bennymee

All that has been shown of AGA (“core3”) so far (as far as I have seen), has been default PAL Hires 640x256, as if it cannot do interlaced or dbl modes.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@bennymee

Is there any reason for 640x512-8bit AGA screen? Most games use 320x200/256 and RTG screens are much better solution for anything else.

@pavlor

There are games that use interlaced modes, multiball levels in certain popular pinball games for example.

The point is - SAGA is supposed to be Super AGA, a superset of AGA, so it should at least be capable of using all modes that AGA is capable of. This is also what has been described since the start. And open sourcing. Thumbs are twiddling.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@pavlor

640 x 512 8-bit 256 colour screens are quite nice to animate on using Deluxe Paint V and a Picasso IV. I guess RTG would have been better but since DP never supported it, it is important to have and is nicely deinterlaced on the PIV!

I'm surprised the mode has been omitted from SAGA thus far!

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios