Geiss II Preview

[Illusion]

If you want to check out a beta version of Geiss II (v0.91), download it here (http://www.geisswerks.com/vis_geiss2.dll).

Save vis_geiss2.dll to your /Winamp/Plugins directory.

[dylman]

Geiss II 0.91 at 1024x768 full screen I get ~17 fps.

Milkdrop 1.03 at 1280x960x32 full screen I get ~55 fps.

XP Pro SP1, Athlon XP2100+, Geforce3 Ti200 @ 200/475, Det 30.82 drivers, Winamp2.81 + latest wa2update.exe

I'll be trying the latest detonator 40.71 drivers in a couple of days, that should improve things... :)

[Rovastar]

It is really slow for me too. I figured that is because it is only a beta as the plug-in does not appear on the face of it to do too much.

One of the best things about the orginal Geiss plug-in was the speed. Highly respected bit of code and I still recommend it for slow computers.

This should be sorted out in time. Or maybe it is not nvidia friendly. :(

[fwgx]

Looks very impressive, I can use more than 8 bit colour too which I never could with version 1. With time the speed will get better. That's what betas are for.

[Twilightseer]

Interesting...
Why was my first post saying that Geiss 2 crashed my computer was deleted ?

[Rovastar]

Wasn't me mate. Dunno. MAybe swag was still setting up the forums at the time.

[Illusion]

Geiss II has issues with my PC nas well 'seer, it seems to steal focus over everything else, even when windowed... :weird: I think it's crashed on me too :)

[geiss]

Geiss II will *definitely* be slower than MilkDrop; my goal is to make it as fast as Geiss I, but more flexible. MilkDrop does all the image 'warping' using hardware; Geiss I uses your CPU, and so does Geiss II. But that CPU-driven warping also allows very fine precisions (especially when you factor in the error-diffusion dithering) and a quality look & smoothness that is impossible to get from hardware (we'd need RGB data stored as 3 floats.. coming soon, but won't be mainstream for a long time).

Geiss II is currently just a tad slower than Geiss I (when you match fullscreen vs. fullscreen at the same resolutions). This is because,
unfortunately, DirectX 8 won't let me write pixels directly to the back buffer; it insists on making me write to a system-memory surface and then copying that over, but that is much slower (20%). I'm going to see if there isn't some way around this.

Anyway, just thought I'd clear those few things up, for now...

Cheers,
Ryan

[dylman]

Thanks for the clarification. :)

[Rovastar]

?

Nope not here mate. I was comparing it to Geiss I not MD obviously.

AMD Ath 1333 Mhz - gf2mx400 (64mb)

Geiss I I get between 28-32 fps all the time 800x600x32bit

Geiss II I get between 8-11 fps all the time 800x600x32bit

Is this what you got Ryan cause otherwise something is very wrong somewhere?

That is what I mean by slow......I am trying not to put the letters a, v and s in the post. Doh!.:)

Cheers,
John

(edit: 2x Monkey texture multipass still moves at 20fps at that res/depth on my machine.)

[Twilightseer]

I still wish I could run Milkdrop :D

[SarCaSM]

hmmmm, trying to run this using winamp 2.81 , with an ATI Mobility 4, 16 megs video ram. getting an "D3FX failed to allocate video-memory texture" upon start of the plugin.

[Twilightseer]

Originally posted by SarCaSM
hmmmm, trying to run this using winamp 2.81 , with an ATI Mobility 4, 16 megs video ram. getting an "D3FX failed to allocate video-memory texture" upon start of the plugin.


Milkdrop had this notorious attitude of hating 3dfx cards; seems like Geiss 2 will do the same.

[cerealkillah]

Geiss 2 runs pretty smoothly for me, but it only takes up the far left 15% of the screen... that slice looks tight though :)

Athlon XP 1600 1.4GHz ATI Radeon 7000

[steak]

Looks good, very good.

But what is 'fake full screen'?

[whitegold]

29.994FPS on my machine 1024x768 32bit 75hz, 2x multisampling

Dual AMD MP 2000+
GeForce4 Ti4400
Win2k SP3

[Rocker]

*tumbleweed...

heh
/me gets impatiant(sp?)

[D&B]

Just tried it out for the first time. very nice. I loved Geiss, ran it all the time, that is until Milkdrop :)

[Rocker]

i like geiss better...

[MMerc]

Originally posted by geiss
Geiss II is currently just a tad slower than Geiss I (when you match fullscreen vs. fullscreen at the same resolutions). This is because,
unfortunately, DirectX 8 won't let me write pixels directly to the back buffer; it insists on making me write to a system-memory surface and then copying that over, but that is much slower (20%). I'm going to see if there isn't some way around this.

I'm not sure what you mean. It seems like one of the most straight forward things to do.
1) Lock the BackBuffer to obtain a pointer
2) Start writing
3) Unlock

But I'm sure you know 10x more about DirectX than I do :hang:

[geiss]

You wrote:
1) Lock the BackBuffer to obtain a pointer
2) Start writing
3) Unlock

Which is correct. However, the way DX8 does it, that pointer is to a [mirrored] system memory surface. I take a long time doing my computations and writing the results to that system memory surface; then, when it's time to unlock the surface, it copies the whole thing up [to video memory] at that point. The problem with this is that it's doing 100% cpu work in the first step, then in the second step (the copy), the AGP bus is 100% saturated; so in both steps, one thing is maxed out, and the other is idle. [the two things being the cpu and the agp bus.]

However, in DX7 and earlier (and using plain old ancient DDraw as I ended up finally doing for Geiss 2 v1.0), I can write pixels directly to video memory; so they trickle over the AGP bus, virtually for free, as I work the CPU to death, with no long bus-bottlenecked copy at the end. =)

Does that explain it?

Cheers,
Ryan

[MMerc]

Those sneaky bastards!

I was under the impression that the pointer I was receiving was a pointer to the video-buffer, not a fake systemmemory-buffer blitted to a real backbuffer.

How do you tell if it's fake or not? I'm using DX9.
If that turns out it's still using a fake backbuffer, I think I might switch over to the good 'ol DDraw interfaces again.

Thanks for the explanation