@Radov

Not sure that assembler can radically speedup minigl. It will give "a few" fps in the best case imho. I think if problem is minigl (and not warp3d), then, its some there on logic-level, on level or organizing the whole functions and so on.

Assembler for sure can speedup a bit, but not radically. And if you talk about 440 cpu, then, its will speedup just a bit minigl for Sams (again, those 1-2 fps).

But the big interest, its that tinigl on morphos, faster in 2 times. On 100%. That for sure not because of assembler usage. And morphos tinigl also works on almos the same warp3d (which name are rave3d for morphos).

Also those slowdown problems can be not because of exactly minigl, and not because of warp3d itself , but maybe because of just system itself. Maybe some there in p96 some sucking Formit/Permit combo, or maybe some wrong sync, or dunno what... Maybe on morphos blitting fucntions of CGX are speeduped radically, maybe everything by little give in end twice performance in compare ..

Tons of maybe

Last edited by kas1e on 10-Jan-2011 at 09:58 PM.

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites

@kas1e

maybe something related to that topic?

Can be P96 SO bad?

Well...maybe some OS4 core dev can tell that to us...

_________________
Simone"Tuxedo"Monsignori, Perugia, ITALY.

@samo79

Quote:

AFAIK on MorphOS 2.x they use CyberGraphX 5 and TinyGL, the first isn't better than Picasso96

Samo

I have to stop you on that one.. Cybergfx is FAR better than p96 ANY day ANY time....

just if i may:
first of all, even if i cant speak from own experience, but the saying goes that morphos equivalent for warp3d is considerably faster (btw i think it was called goa).

on the other hand the minigl was considered to be much faster than storm mesa, which i wasnt able to confirm on 68k testing same application with both libraries. well, the application was pretty basic (gears) so maybe it gave not the whole picture.

optimizing asm might speed up the things, but as kas1e say, not by two times. matthey has proven it with w3d voodoo hardware drivers. but the biggest bottleneck still remains w3d with its limited set of functions through which (as it is) it is impossible to squeeze the effective opengl support, it lacks probably too much to count, just to name hardware t&l.

ok, noob has spoken, howgh!

@Tuxedo

from what ive been able to test p96 isnt generally much worse than cgx on 68k, after some unofficial updates have been applied, that is. but since i was testing (pretty chaotically) only a limited range of solutions, and from a quite biased perspective (sdl) i still would not want to make final statement about it. i dont think, it influences 3d to much though, am i right?

@tuxedo
Quote:

maybe something related to that topic?

Imho nope. As i know minigl already uses Altivec in some aspects.

Quote:

Can be P96 SO bad?

Very very possible. It can be not "bad" but just "too old". With permit/forbit calls, designed without supporting of 3d in mind, and so on. As well as warp3d interface and logic of working with system are too old too.

Quote:

Well...maybe some OS4 core dev can tell that to us...

I talk and Rogue (main warp3d author), and he that:

Quote:

Warp3D is neither old nor slow. The problem lies elsewhere, namely in the way that the command processor works. The issue is that the feed is slow, and the reason for that is that Picasso96 intvervenes. There is a lot of things I could do for speedup, but I lack the time, and others lack the knowledge. The lack of internal arbitrating create the problem. It requires idle times to sync whatever is coming at it. ... and so on ....

The big problem from my point of view, its that aos4-core dev not so worring about "clear performance and optimization". If something works more or less fine, then they switch to another area. Because if they (aos4 core dev) will have priority to "make good and fast 3d" they of course will do it. Maybe because of it, for a long time they think that Warp3d will be not used anymore, and because of that we in almost 10 years use 3d layers which have fixes only 1 times in years.


@wawa
As i know 3d layers for morphos called "rave3d". Goa, it was just some news from Hyperion, years ago, which mean Warp3d 5.0 (they want at these times improve warp3d more, but after a while drop that idea by some reasons)

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites

@kas1e

Quote:

The big problem from my point of view, its that aos4-core dev not so worring about "clear performance and optimization". If something works more or less fine, then they switch to another area. Because if they (aos4 core dev) will have priority to "make good and fast 3d" they of course will do it. Maybe because of it, for a long time they think that Warp3d will be not used anymore, and because of that we in almost 10 years use 3d layers which have fixes only 1 times in years.

Absolutely correct imho!
And...
Also if Gallium was lie Catalyst...probably we dont have so much benfit from it until the 2D gfx subsytem wasnt updated...

The old Amiga-way-of-programming wasnt respected anymore unfortunately...

I think was better to have less things that works (really)good that many that works but not satisfy noone...

_________________
Simone"Tuxedo"Monsignori, Perugia, ITALY.

@kas1e

http://www.biclodon.com/software/goa/

_________________
x86/MorphOS 4.0

"Delving into the past can be a dangerous exercise." -hyperionmp

"I've been a supporter of "REACTION" GUI because is an Amiga OS thing." -Snuffy

"I personally prefer a vision of do'ers and makers rather than

@Cheese
Yep, right.

@All
Btw, from the link which post cheese, you can see what is wrong with warp3d itself. That all about authors;s wish, to make fast 3d.

Last edited by kas1e on 11-Jan-2011 at 01:49 PM.

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites

@Tuxedo

Quote:

Also if Gallium was lie Catalyst...probably we dont have so much benfit from it until the 2D gfx subsytem wasnt updated... I think was better to have less things that works (really)good that many that works but not satisfy noone...

We will see in end anyway. But strange why just no going easy way before, and just not do some actual benchmarks of CGX vs P96 on peg2 with aos4.x / mos2.x, then actual benchmarks of the same tests beetwen warp3d-aos4 and goa-mos, then actual benchmarks of minigl over tinigl. Compare all of this, see where difference start to be big, and invisigate deep those problems.

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites

@kas1e

Quote:

kas1e wrote: The big problem from my point of view, its that aos4-core dev not so worring about "clear performance and optimization". If something works more or less fine, then they switch to another area. Because if they (aos4 core dev) will have priority to "make good and fast 3d" they of course will do it. Maybe because of it, for a long time they think that Warp3d will be not used anymore, and because of that we in almost 10 years use 3d layers which have fixes only 1 times in years.

In their defence, AmigaOS 4.x is a lot more than just the Warp3D drivers, so the aos4-core devs have a lot more other things to work on than just 3D graphics. For you 3D is incredibly important, whereas for someone else, having app X work a little better, or USB 2.0, or hardware Y supported is more important. There's always going to be someone telling them that they have their priorities wrong. On top of this there is fixing bugs, porting to new motherboards, etc.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@kas1e

Quote:

kas1e wrote: We will see in end anyway. But strange why just no going easy way before, and just not do some actual benchmarks of CGX vs P96 on peg2 with aos4.x / mos2.x, then actual benchmarks of the same tests beetwen warp3d-aos4 and goa-mos, then actual benchmarks of minigl over tinigl. Compare all of this, see where difference start to be big, and invisigate deep those problems.

People have compared Warp3D vs Goa often enough. I have never seen any P96 vs CGX benchmarks. However, all that it will tell you is that X is n ops/s faster than Y under condition Z. It says nothing about why, and what can be improved.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans
Quote:

People have compared Warp3D vs Goa often enough. I have never seen any P96 vs CGX benchmarks. However, all that it will tell you is that X is n ops/s faster than Y under condition Z. It says nothing about why, and what can be improved.

Imho , those benchmarks will show us in which area difference. If we already compare w3d vs goa (by the same for example quake3, which if i remember right, on boch oses use warp3d, and not opengl), we still not have normal tests (well, maybe have, but i not see). For example: upload xxxxx textures how fast, draw xxxx textures - how fast, draw, scretch, move, cut off, i.e. compare not by games (which indeed will say nothing), but by little benchmark-test programs.

Then the same for CGX and P96. Draw something on screen by anologic functions. See if difference have place or not, check all the functions which we have there and there. And if no difference : then its no gfx subsystem problem. If difference is here, then P96 , and then, os4-dev-team coders can see in that way.

Step by step, test one over another, and imho it possible to found in which area problems. For now we even not have clear answer on question "why on the same HW minigl slower than tinigl in 2 times". Only maybes, but maybes for sure will not helps.

Rogue say back in times, that he can speedup w3d there and there, just have no time. So he know already some speedup tricks as initial w3d author ..

Quote:

For you 3D is incredibly important, whereas for someone else, having app X work a little better, or USB 2.0, or hardware Y supported is more important.

Yes, you are right, time and resources and so on, but in last year, the most complaining from users its all about 3d (and even about usb less).

Last edited by kas1e on 11-Jan-2011 at 08:14 PM.

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites

@Tuxedo

Quote:

Tuxedo wrote: Also if Gallium was lie Catalyst...probably we dont have so much benfit from it until the 2D gfx subsytem wasnt updated...

I'm not sure what you're trying to say here, but I think that people need to take a wait-and-see approach. Picasso96's ageing design might pose some technical challenges, but that doesn't make it impossible to bully it into playing nice with 3D drivers.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@kas1e

Try this for benchmarking 2D performance of OS4 and MorphOS. It's 68K and system friendly.

http://aminet.net/util/moni/AmiGOD2.lha

@kas1e
Quote:

The big problem from my point of view, its that aos4-core dev not so worring about "clear performance and optimization". If something works more or less fine, then they switch to another area. Because if they (aos4 core dev) will have priority to "make good and fast 3d" they of course will do it. Maybe because of it, for a long time they think that Warp3d will be not used anymore, and because of that we in almost 10 years use 3d layers which have fixes only 1 times in years.

The plan (for many years) was always to replace Warp3D with something modern. Hyperion has now decided to invest in Gallium3D to that end. Simple as that.

_________________
ExecSG Team Lead

@kas1e

Quote:

kas1e wrote: Imho , those benchmarks will show us in which area difference. If we already compare w3d vs goa (by the same for example quake3, which if i remember right, on boch oses use warp3d, and not opengl), we still not have normal tests (well, maybe have, but i not see). For example: upload xxxxx textures how fast, draw xxxx textures - how fast, draw, scretch, move, cut off, i.e. compare not by games (which indeed will say nothing), but by little benchmark-test programs.

You're assuming that there are simple areas where improvements are needed, but graphics drivers are never so simple. IMHO, time would be better spent working on implementing and testing existing ideas to improve the drivers rather than writing benchmarks to compare with someone else's product. Even with smaller tests, you still don't know the why or the "what to do about it."

Actually, IMHO, time would be better spent working on bringing Gallium 3D to the platform so that we can move on from 10+ year old graphics cards.

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Hans

Quote:

so that we can move on from 10+ year old graphics cards.

Your tright here, but remember that our hw (except Sam460 and the future X1000) cant handle Much that Radeon9800 AGP card...
And getting PCI RadeonHD cards was really useless imho...

_________________
Simone"Tuxedo"Monsignori, Perugia, ITALY.

@Tuxedo

Quote:

Tuxedo wrote: Your tright here, but remember that our hw (except Sam460 and the future X1000) cant handle Much that Radeon9800 AGP card... And getting PCI RadeonHD cards was really useless imho...

Why not wait and see before deciding that PCI Radeon HD cards are useless?

Hans

_________________
http://hdrlab.org.nz/ - Amiga OS 4 projects, programming articles and more. Home of the RadeonHD driver for Amiga OS 4.x project.
https://keasigmadelta.com/ - More of my work.

@Tuxedo
Quote:

Your tright here, but remember that our hw (except Sam460 and the future X1000) cant handle Much that Radeon9800 AGP card...

AmigaOnes and Sam440s can handle all those RadeonHD cards as well. In other words, all HW which we have can handle these PCI radeonHD cards. Pegasos2 can handle it too, just genesi no in interest to update their firmware to detect cards normally. Dunno why you think that only sam460 and x1000 can handle it ..

Quote:

And getting PCI RadeonHD cards was really useless imho...

Maybe i not so good understand what you mean, but radeonsHD cards are really very cool to have. It will speedup everything. How you can say that new cards are useless .. dunno. Its the same as say that new HW and new CPU are useless. Or 1gb of memory useless, yep ?:)

@hans
Quote:

Actually, IMHO, time would be better spent working on bringing Gallium 3D to the platform so that we can move on from 10+ year old graphics cards.

Of course. I just hope that when you involved in all of this, it will be done in right way for now, and in terms of speed, and in terms of stability :)

Last edited by kas1e on 11-Jan-2011 at 09:35 PM.

_________________
Join us to improve dopus5!
zerohero's mirror of os4/os3 crosscompiler suites