Rurouni Kenshin : Meiji Kenkaku Romantan Saisen

[srdjan1995]

(02-19-2013 09:13 PM)KingPepper Wrote:  

(02-19-2013 08:37 PM)srdjan1995 Wrote:  

(02-19-2013 08:01 PM)KingPepper Wrote:  

(02-19-2013 03:21 PM)srdjan1995 Wrote:  

(02-19-2013 02:50 PM)ckgbr Wrote:  with 6.1 build at my desktop there is no graphical glitch, but at a motorola razr (powerVR) there is the same graphical glitch, and at galaxy note 2, there isn´t.

Do u have Nvidia or ATI graphic on ur PC (desktop)? Don't know about Motorola Razr, but i have that graphical glitch on windows even with latest drivers (314.07) and latest git build.

Thanks for making me realise there was an update for my Nvidia drivers

Hahahaha, you're welcome man.. So, u have nvidia graphic on ur PC, but no graphical glitch as i do? Hmm, that's weird.

No my friend, I don't even have this game, I just noticed you said a newer version of my driver (310.90) I think it was

Oh, sorry.. I didn't saw ur nickname. I tought ur same guy (the one that don't have graphical glitch on windows, but have on motorola razr). Yeah, those are new drivers, especialy for Crysis 3 full version.. Too bad i have DX10.1 graphic so i can't play it.. They should have add DX10 support too, and not only DX11.. Graphic would be awesome even od DX10 in Crysis 3.

[ckgbr]

It stop working at pre 0.7 at windows and android...
Please check that, thanks a a lot.

[[Unknown]]

Which version did it stop in and which one was it working last in?

-[Unknown]

[rasyid]

(02-26-2013 01:24 AM)[Unknown] Wrote:  Which version did it stop in and which one was it working last in?

-[Unknown]

me too
last 754 did work
last working looks like on 0.61 at android market

[[Unknown]]

Oh, so it's working again in 754?

If not, that's a big range. The 754 is a count of changes - so if it was working in 0.6.1 and not working in 0.6.1-754, there are 754 things that could've broken it.

If possible, could you try a couple other versions? For example, if it works in 400, that reduces the number of things that could've broken it from 754 to 354. If you try a couple like that, say even 5, it'll cut it down to only ~23 possibilities - 3% of the original, and a lot easier to tell what might've caused the problem.

You could find the exact change that broke it in ~9-10 tries. Knowing the change that broke it would make it like 1000 times (or that is, 754 times) easier to fix.

-[Unknown]

[rasyid]

(02-26-2013 02:43 AM)[Unknown] Wrote:  Oh, so it's working again in 754?

If not, that's a big range. The 754 is a count of changes - so if it was working in 0.6.1 and not working in 0.6.1-754, there are 754 things that could've broken it.

If possible, could you try a couple other versions? For example, if it works in 400, that reduces the number of things that could've broken it from 754 to 354. If you try a couple like that, say even 5, it'll cut it down to only ~23 possibilities - 3% of the original, and a lot easier to tell what might've caused the problem.

You could find the exact change that broke it in ~9-10 tries. Knowing the change that broke it would make it like 1000 times (or that is, 754 times) easier to fix.

-[Unknown]

sorry sir its my fault,, 754 did not work

but on build that still caould work if im remember that version still have a slighty fast interpreter

---

last working is v0.6-470-gde626e3

[[Unknown]]

That commit:

https://github.com/hrydgard/ppsspp/commi...d2cd9a96dd

Broke some other games and was fixed later. Did you try 500 for example just in case?

-[Unknown]

[Carter07]

@[Unknown], it's a problem with vertex cache because if disabled game works, tested in windows.

[[Unknown]]

Wow, that's interesting.

If you replace this line in GPU/GLES/TransformPipeline:

Code:

vai->drawsUntilNextFullHash = std::min(24, vai->numFrames);

With:

Code:

vai->drawsUntilNextFullHash = 0;

Does it work (probably slower than before), or does it still not work?

-[Unknown]

[Carter07]

Unfortunately it still crashes when pressing start at the main screen, if vertex cache is enabled after the main screen it works well.

If it's useful I can find the specific commit that broke it.

[ckgbr]

Just tried with last v0.6-754-gd2159d4 for android at motorola razr, if I turn off vertex cache, it works without grafics gliches found at v0.61 but very slow.
I think something at vertex cache crashes after pressing start at the main screen.

[[Unknown]]

That's very interesting. So it's not the caching of the vertices that is the problem (because my above change makes it rehash every time), but some of the mechanics in the caching.

Does the game crash or the emulator itself? If the emulator, a stack trace would be really helpful if possible.

-[Unknown]

[ckgbr]

I think is the emulator that crashed because it closes back to the android home screen.
How do I get a stack trace?

[[Unknown]]

Okay, sounds like a crash. I don't know how to get a trace on Android, unfortunately..

A couple things to try in GPU/GLES/TransformPipeline.cpp:

Change:

Code:

switch (vai->status) {

To:

Code:

vai->status = VertexArrayInfo::VAI_NEW;
switch (vai->status) {

If that doesn't help, it's probably in ComputeFastDCID() or something. If it does help, revert and change:

Code:

if (newHash != vai->hash) {

To:

Code:

if (newHash != vai->hash || true) {

If that helps, there's something probably wrong with vertice reuse. If it does, revert and change:

Code:

if (vai->vbo == 0) {

To:

Code:

if (vai->vbo) {
    glDeleteBuffers(1, &vai->vbo);
    vai->vbo = 0;
}
if (vai->ebo) {
    glDeleteBuffers(1, &vai->ebo);
    vai->ebo = 0;
}
if (vai->vbo == 0) {

This will test the vbo creation code path only. I think the above would really narrow it down.

Sorry, I don't have this game...

-[Unknown]

[Carter07]

It's not only an android problem.
I attached the stack trace of the crash.

However this change made the game not crash anymore:

Change:

Code:

switch (vai->status) {

To:

Code:

vai->status = VertexArrayInfo::VAI_NEW;
switch (vai->status) {