Best open source, mod-able RTG-friendly FPS Engine running on 68k?

[James]

Originally Posted by eXeler0

Does Blitz Quake support the extended size maps?

No, AFAIK the only Quake engine on Amiga that can handle extended maps is the experimental one included with Qbreed. It always crashed for me though.

http://aminet.net/package/game/data/Qbreed

There are lots of large and impressive maps that don't require extended limits though, speed is going to be much more of a problem. Check these out for starters:

Marcher Fortress (could really do with extended audio, but works with a few sky graphics glitches and sound drop off)
https://www.quaddicted.com/reviews/kinn_marcher.html

Bastion Of The UnderWorld
https://www.quaddicted.com/reviews/kinn_bastion.html

Castle Of The Dark Ages
https://www.quaddicted.com/reviews/cda.html

Day Of The Lords
https://www.quaddicted.com/reviews/gmsp3.html

Colony
https://www.quaddicted.com/reviews/gmsp1.html

[eXeler0]

Thanx @James good stuff
Meanwhile, Gunnar has pretty much confirmed that the FPU in Apollo will be better (and more compatible? Still not entirely clear in this matter) than the one in the 060. No timeframe, but sounds promising.

Edit: Apollo-core Should be powerful enough to run Quake 2. I know @NovaCoder ran it on AGA + 75MHz 060. 100+MIPS & any kind of RTG should maybe be seen as the minimum requirement for a decent Q2 experience.
I'm sure the Gold Core will beat that with some healthy margin.

[eXeler0]

Anyone tried to run the Hyperion QII port on Vampire?
If I remember this correctly, the 68k port required Warp3d, so I guess Wazp3d is needed att the very least? (Which is kind of ironic considering Nova Coder had it running on AGA, but ok, on overclocked 060...)

[matthey]

Originally Posted by eXeler0

Meanwhile, Gunnar has pretty much confirmed that the FPU in Apollo will be better (and more compatible? Still not entirely clear in this matter) than the one in the 060. No timeframe, but sounds promising.

I doubt the Apollo core FPU will be more compatible than the 68060 FPU. It would need to be mostly bug free even if the decision was made to add all the 68060 FPU instructions, use extended precision fp, use the same supervisor stack frame format (68040 and 68060 are different), etc. The performance with the Apollo core FPU could probably be better as larger caches, more memory bandwidth and fully pipelining should more than make up for some of the disadvantage of slower wide shifting and multiplying in FPGA.

Originally Posted by eXeler0

Edit: Apollo-core Should be powerful enough to run Quake 2. I know @NovaCoder ran it on AGA + 75MHz 060. 100+MIPS & any kind of RTG should maybe be seen as the minimum requirement for a decent Q2 experience.
I'm sure the Gold Core will beat that with some healthy margin.

The Quake 2 engine is not much slower than the Quake 1 engine. Amiga versions of Quake 1 may be relatively faster because they are better optimized. I believe NovaCoder was using a 68060@80MHz with AGA (I use a 68060@75MHz with RTG) but the c2p becomes a small percentage of the CPU time with faster processors. Most of the performance is integer dependent which includes compiler and caching issues on the 68060. The 68060 does not seem to have large enough caches for the bloat that large modern compilers produce for large modern and semi-modern programs. The Apollo core probably has fixed the cache bottlenecks (which are now much larger than PPC caches considering the code density) but the outdated and poor 68k compiler backends are more difficult to fix.

Originally Posted by eXeler0

Anyone tried to run the Hyperion QII port on Vampire?
If I remember this correctly, the 68k port required Warp3d, so I guess Wazp3d is needed att the very least? (Which is kind of ironic considering Nova Coder had it running on AGA, but ok, on overclocked 060...)

Amiga Quake 2 versions which do not require Warp3D are available including sources but you still need a compatible FPU. Most 3D engines from the PC/x86 world require a FPU and/or SIMD unit with fp support.

[James]

Originally Posted by eXeler0

Anyone tried to run the Hyperion QII port on Vampire?
If I remember this correctly, the 68k port required Warp3d, so I guess Wazp3d is needed att the very least? (Which is kind of ironic considering Nova Coder had it running on AGA, but ok, on overclocked 060...)

No, the software renderer works, the requirement for Warp3D was simply a matter of speed.

[eXeler0]

Originally Posted by James

No, the software renderer works, the requirement for Warp3D was simply a matter of speed.

Ok I see, so the minimum requirements were to ensure a certain framerate on stock 060.
Do you know how much difference it makes in fps? (Warp 3d vs AGA)?.

Btw, I played QII on my PC last night and realized that the "QII look" I remembered from back in the day was the filtered low-res texture look - whereas QI was mostly played on software renderer with unfiltered pixels with that LEGO look.
I don't suppose there's a way to get texture filtering on 68k without 3d gfx hardware?
There was talk about The Vampire (gold core or later) to get some SIMD instructions (MMX). In the future maybe a wasp3d driver could be rewritten to speed up some stuff using SIMD.

[eXeler0]

Originally Posted by matthey

Amiga Quake 2 versions which do not require Warp3D are available including sources but you still need a compatible FPU. Most 3D engines from the PC/x86 world require a FPU and/or SIMD unit with fp support.

Ye I got a little carried away there. 🍻 My brain was already seeing the gold core with pipelined FPUs etc... 😳 I know Quake needs FPU.

Can you suggest a good QII port thats running on 060 +AGA? My 060 is only at 66MHz but it's a rev 6 so I could increase that. But it would only be for testing anyway until the Vamp gets its FPU.

(Currently trying out different Editors and config them with my awkward Steam-path to QII).

[matthey]

Originally Posted by eXeler0

Do you know how much difference it makes in fps? (Warp 3d vs AGA)?.

There are many factors but a 50%-100% better frame rate on the 68060 with hardware Warp3D is likely. Higher resolutions and colors plus special effects are cheaper with Warp3D and more likely as well.

Originally Posted by eXeler0

Btw, I played QII on my PC last night and realized that the "QII look" I remembered from back in the day was the filtered low-res texture look - whereas QI was mostly played on software renderer with unfiltered pixels with that LEGO look.
I don't suppose there's a way to get texture filtering on 68k without 3d gfx hardware?

Everything done in hardware can be done in software with enough CPU grunt. The Apollo RTG could add FPGA "hardware" 3D as well.

Originally Posted by eXeler0

There was talk about The Vampire (gold core or later) to get some SIMD instructions (MMX). In the future maybe a wasp3d driver could be rewritten to speed up some stuff using SIMD.

The SIMD would need to support at least single precision floating point to be able to accelerate much (the Apollo SIMD instructions are integer only as far as I know). A Warp3D driver should be possible but would be a lot of work. SIMD units are not the easiest to program which is why it is good to make them like other SIMD units to use code and compiler support already written. The original plan for the Apollo core was to make it more like the PPC Altivec SIMD unit.

Originally Posted by eXeler0

Can you suggest a good QII port thats running on 060 +AGA? My 060 is only at 66MHz but it's a rev 6 so I could increase that. But it would only be for testing anyway until the Vamp gets its FPU.

NovaCoder's Quake 2 was written for AGA and has good speed and stability here.

http://aminet.net/game/shoot/AmiQuake2.lha

NovaCoder's port is less configurable than some in an attempt to speed it up. I believe Cowcat's port is based on the Hyperion code which has a software renderer as James states above. It has more options and reasonable performance but may have problems on some configurations (testers needed). The thread about it is here on EAB.

http://eab.abime.net/showthread.php?t=74679

[eXeler0]

Just saw that the new Vampire core (silver 8) has vector instructions enabled.
Still no FPU though..
Now, considering that no current 68k compiler will optimize for SIMD, what would it take to speed up, say Quake2 with the new vector instructions?
(Lets pretend we will have FPU soon enough to assume it's presence).
Is it a matter of manually replacing some math/FPU calls with Vector instructions in assembly? And in such cases, how likely is something like that to happen? ;-)


Skickat från min HTC One via Tapatalk

[Samurai_Crow]

A Mesa implementation using vectors will speed up 3D quite a bit but the 2D blitting will come out first.

Also, once I get my card, I'm going to try to vectorize AHI.

[Zetr0]

@eXeler0

There are many Doom type tools on the PC, some even do a total conversion - Wazp3D is *amusing* at the moment on the V2, but once the FPU/SIMD is squared (I think we have a little vectorization in Silver8) - there is going to be some FUN!

[utri007]

Best, wich means stable and fastest Quake port is ClickBoom's Quake 060. That was a separete update wich at least was freely downloadable from several web sites.

I was prety sure that it is Amigas Futuredownloads, but I could found it.

It is not Limited to 8bit screens, it supports every selectable screenmode.

Wazp3d is no go, it made possible to play some games with Amiga OS4 / PPC, but it would require too much raw CPU power and everything doesn't work.

[Samurai_Crow]

Mesa is good for running OpenGL software on AROS so I would expect the AROS 68k port would be possible on the Vampire board.

[trixster]

Originally Posted by utri007

Best, wich means stable and fastest Quake port is ClickBoom's Quake 060.

In my experience NovaCoder's AmiQuake is faster than the ClickBoom 060 version by a couple of fps. It is also stable.

[utri007]

Originally Posted by trixster

In my experience NovaCoder's AmiQuake is faster than the ClickBoom 060 version by a couple of fps. It is also stable.

It might be case with original Clickboom's Quake exe, but not with updated 68060 exe. Performance leap is huge.

Clickboom's exe (all versions) are also more configurable.

[trixster]

Do you have a link to the updated 060 exe?

the 060 version of clickboom quake im using returns 15.36 fps average for timedemo demo1, 320x200 ntsc

the amiquake version novacoder released returns 16.3 fps average for timedemo demo1, 320x200 ntsc

[matthey]

Originally Posted by eXeler0

Just saw that the new Vampire core (silver 8) has vector instructions enabled. Still no FPU though..
Now, considering that no current 68k compiler will optimize for SIMD, what would it take to speed up, say Quake2 with the new vector instructions?
(Lets pretend we will have FPU soon enough to assume it's presence).
Is it a matter of manually replacing some math/FPU calls with Vector instructions in assembly? And in such cases, how likely is something like that to happen? ;-)

Quake 1/2/3 use floating point vectors (like matrices). The Apollo core SIMD is integer only. An SIMD is a very specialized tool and without floating point it is the wrong tool for this job. The Apollo core SIMD will likely have a negative impact on Quake 1/2/3 as resources are robbed from the general purpose FPU for a specialized tool that can't be used for 3D here.

Integer OpenGL ES will *not* speed up floating point using Quake either. Don't expect the Tabor without FPU to run Quake 1/2/3 well. The Amiga has left the stone age and entered the copper age. Get ready for some Android ports but forget about a modern PC computer and games.

[eXeler0]

Originally Posted by matthey

Quake 1/2/3 use floating point vectors (like matrices). The Apollo core SIMD is integer only. It is the wrong tool for this job. The Apollo core SIMD will likely have a negative impact on Quake 1/2/3 as resources are robbed from the FPU for a tool that can't be used for 3D here.

Integer OpenGL ES will *not* speed up floating point using Quake either. Don't expect the Tabor without FPU to run Quake 1/2/3 well. The Amiga has left the stone age and entered the copper age. Get ready for some Android ports but forget about a modern PC computer and games.

Well I Think we've already agreed that Quake III is beyond the capabilities of Apollo, however Quake 2 is already sort of playable on overclocked 060 and AGA so it's a good and realistic target to make *really* playable on the Vampire. Probably lots of mods that can be run on it too...
So the question is; is there anything in Silver core 8 can be of use when trying to improve the Q2 experience. (Granted, we need to wait for the FPU to be enabled. And hopefully the FPU will be faster than the one in 060.)
@matthey I know you don't agree with everything the Apollo team does, in your opinion, what would be the most useful feature to implement within the current level of ambition and a reasonable timeframe (and FPGA "real estate") with respect to having a useful platform for solid 3d gaming (quake 2 generation).



Skickat från min HTC One via Tapatalk

[matthey]

Originally Posted by eXeler0

Well I Think we've already agreed that Quake III is beyond the capabilities of Apollo, however Quake 2 is already sort of playable on overclocked 060 and AGA so it's a good and realistic target to make *really* playable on the Vampire. Probably lots of mods that can be run on it too...

Quake III would likely be playable if single precision floating point was added to the SIMD. This should be possible in an FPGA but would require a larger more expensive FPGA.

Originally Posted by eXeler0

So the question is; is there anything in Silver core 8 can be of use when trying to improve the Q2 experience. (Granted, we need to wait for the FPU to be enabled. And hopefully the FPU will be faster than the one in 060.)

If the FPU was as fast as the 68060 FPU, I would expect Quake 1/2 to do 20-30FPS without any SIMD. It is possible to improve the performance of the FPU substantially although an FPGA doesn't like wide operations and more pipelining is expensive. Memory bandwidth in the FPGA helps the FPU performance though. I don't know what has been done to enhance the FPU or what is missing so I can't comment further.

Originally Posted by eXeler0

@matthey I know you don't agree with everything the Apollo team does, in your opinion, what would be the most useful feature to implement within the current level of ambition and a reasonable timeframe (and FPGA "real estate") with respect to having a useful platform for solid 3d gaming (quake 2 generation).

The easiest ways to enhance 3D on the Amiga are improving the integer and FPU performance and improving compilers to take advantage of enhancements. An SIMD with floating point is too expensive in an FPGA and compiler support is too complex for short term goals.

[eXeler0]

Originally Posted by matthey

Quake III would likely be playable if single precision floating point was added to the SIMD. This should be possible in an FPGA but would require a larger more expensive FPGA.

You sure? Don't think I've ever seen the original Q3 run on say something like a Pentium MMX without a 3d graphics card. I remember playing it on my PII 400MHz + nVidia TNT at the time. And I remember noticing improvements when I moved to TNT2. (Mostly when running in 24/32-bit color mode).

Originally Posted by matthey

If the FPU was as fast as the 68060 FPU, I would expect Quake 1/2 to do 20-30FPS without any SIMD. It is possible to improve the performance of the FPU substantially although an FPGA doesn't like wide operations and more pipelining is expensive. Memory bandwidth in the FPGA helps the FPU performance though. I don't know what has been done to enhance the FPU or what is missing so I can't comment further.

Just speculating, but I expect the Apollo FPU beat the 060 by some margin. ;-)

Originally Posted by matthey

The easiest ways to enhance 3D on the Amiga are improving the integer and FPU performance and improving compilers to take advantage of enhancements. An SIMD with floating point is too expensive in an FPGA and compiler support is too complex for short term goals.

Well, tbh that has already been done in the Apollo core, hasn't it? The IPC is improved over 060 even if compilers aren't up to speed yet.. Clock is higher, memory bandwidth by far exceeds anything previously seen on Amiga.
FPU is still shrouded in mystery though.

Compiler optimizations should come one day, but then I suppose the easiest solution is the brute force alternative. The cost of performance is passed on to the buyer. (Meaning, were paying top dollar for Apollo core in an Arria 10 FPGA. But then we magically have a 400MHz 060 and we can screw all optimizations ;-)
.

Skickat från min HTC One via Tapatalk