Hyrule Field Slowdown Observation

[deathjijn]

(05-05-2010, 05:31 AM)Xtreme2damax Wrote: What other games, I've tested quite a few games and have had absolutely no problems and I have a lot of games. Also these patches shouldn't affect much anyways, Kiesel-Steins patch has no affect on any other games except for Zelda Twilight Princess from what I tested so far. The only other alteration I made also shouldn't cause many issues if it causes any at all.

Reverting the JIT change might of affected something, but it shouldn't of broken anything and any issues relatively minor.

Actualy even without the jit change the patch screws up textures in certain games ill name 3 all the metroids, windwaker, and tales of symphonia gets some issues it messes with textures in certain games.
I think more testing needs to be done on your compile before you commit it to the public but its your choice not mine. And its not the JIT you can try it with the older patched version compared to the official Svns and you'll notice the texture corruptions. Not that it matters i barely play on dolphin Between my son, College, a wii, 360, and ps3 i don't really mess with it i just give feedback. Just though thought id let you know what I've seen. If its not welcome i'll just go elsewhere.

[Xtreme2damax]

Well it might be an another issue that is causing it but I don't think it's related to the patch, I can't say that I have noticed anything as everything seems the same as unpatched except for it being faster now.

Can you provide some screens so I can check to see if I have the same problem in the same areas or if the problem also exists on unpatched/older builds?

[deathjijn]

(05-05-2010, 05:56 AM)Xtreme2damax Wrote: Well it might be an another issue that is causing it but I don't think it's related to the patch, I can't say that I have noticed anything as everything seems the same as unpatched except for it being faster now.

Can you provide some screens so I can check to see if I have the same problem in the same areas or if the problem also exists on unpatched/older builds?

Ill be back in a few days with the screens if it hasn't been noticed by then got lots of studying i need ro do over the course of the next few days but im going to bed thought id make a quick post.

[Xtreme2damax]

Well it would be nice if you can post them today, Kiesel-Stein may know what the issues are and be able to fix them.

I do see what you mean on Metroid Prime at least, but I couldn't find anything in Zelda Windwaker and I don't own the other title you mentioned.

[Xtreme2damax]

Comment 10 in this issue mentions something I think may be related to the Hyrule Field slowdown issue as well:

Quote:Apparently fifo breakpoints don't always get deleted in dual core mode. this causes
an infinite loop in the gpu thread.

debugging and manually resetting _fifo.bFF_BPEnable and _fifo.bFF_Breakpoint causes
the values to be set again immediately. in all cases i tested the breakpoint is
successfully removed thereafter and the game continues execution.

i guess this might be some timing/syncing issue.

[Xtreme2damax]

Yeah kiesel, your latest patch is still causing some texture errors/corruption/oddities. Check the games deathjijn mentioned, so far I've only noticed an issue on Metroid Prime 1 when you use the weapons the blast effect is messed up.

deathjijn may wish to post more screen shots so this can be looked into.

[kiesel-stein]

(05-05-2010, 06:26 AM)Xtreme2damax Wrote: Well it would be nice if you can post them today, Kiesel-Stein may know what the issues are and be able to fix them.

Please don't overestimate what can be done with a change as small as this. My change essentially introduces a bug into dolphin, which has the side-effect of speeding up Hyrule Field. It is pure coincidence that this change has nearly invisible effects on Zelda TP. I expected a lot more broken rendering.

The change was intended to pinpoint an important(if not the most important) bottleneck when rendering Hyrule Field(at least on my systems). Improving gameplay of this single game was a nice sideeffect, i admit.

For correct rendering, the pipeline flush really needs to be done, since different texture environments currently need different fragment shader programs, and the flush updates the shaders.

The real fix, imho, needs to be done elsewhere, either by using a different path for the texture environment information or by reducing the cost of the pipeline flush.

It looks like the driver revalidates all its state on the machines with slowdowns, but for some reason does not do that on others. This is the reason for my suspicion that a newer directx may be faster in that regard, because, although dolphin talks to the dx9 interfaces, they are only (at worst) a thin wrapper around dx11, it is not talking to the old code in dx9 releases that only get bugfixes. Or, to put it another way: there is not much of dx9 code left in dx11 releases.

[Xtreme2damax]

Kiesel, this may also be part of the issue as well and is contributing to the issue you mentioned:

Quote:Apparently fifo breakpoints don't always get deleted in dual core mode. this causes
an infinite loop in the gpu thread.

debugging and manually resetting _fifo.bFF_BPEnable and _fifo.bFF_Breakpoint causes
the values to be set again immediately. in all cases i tested the breakpoint is
successfully removed thereafter and the game continues execution.

i guess this might be some timing/syncing issue.

I have a feeling that Skids fifo fixes might actually fix this issue as well, if not I'm not sure what the problem could be. Your theory is partially true, but there is other factor(s) also contributing to the performance issues.

[skid]

Xtreme2damax, yes, my current FIFO code has fixed the FIFO breakpoint behaviour. I have written the code according to the docs we have on the FIFO.

The code in SVN is so wrong that it probably should not have been submitted. It is a hack built on top of another hack (very bad!). This is known by the devs including ayuanx who submitted the change.

The FIFO breakpoint code will not affect performance. It is code that affects compatibility (games that hang).

[Xtreme2damax]

Could the hangs in the fifo code theoretically be what is causing these performance issues and slowdowns in Hyrule Field?