search

skmp / reicast-emulator

Reicast was a multiplatform Sega Dreamcast emulator
https://reicast.emudev.org
Other
1.1k stars 346 forks source link

Unstable optimizations / Crackling sound #118

Closed skmp closed 10 years ago

skmp commented 10 years ago

It is time for the release!

I'm merging in #113, then we gotta test abit for all platforms, and rls ?

Seems like #109 will be for r7 (with some code to edit input config?)

skmp commented 10 years ago

Also, #115 if none has any objections ?

skmp commented 10 years ago

@ptitSeb - can you prepare a pull for the pandora branch? I can merge the sh4rec stuff myself in, but duno what else you want to get in for r6 ~

skmp commented 10 years ago

Also maby #115 ?

ptitSeb commented 10 years ago

@skmp, I'm not ready, there are still my frameskip nonsense that I have to remove first. Maybe it's better if you get the sh4rec alone for now, I'm not sure the rest is 100% correct (most is garded with #ifdef exept the frameskip code I have to clean).

skmp commented 10 years ago

Okay, great, will do that. Frameskip is implemented in mainline as well now (actually, entire ta-level skipping, so it should be faster)

ptitSeb commented 10 years ago

Yes, I have seen that. That's why I want to remove my code.

mrcmunir commented 10 years ago

@skmp actually is broken wrong with you recents changes in sh4/driver.cpp I think if (!settings.dynarec.unstable_opts) to if (!settings.dynarec.unstable_opt) because https://github.com/reicast/reicast-emulator/blob/master/core/nullDC.cpp settings is dynarec_unstable_opt without s right? Later I try With #113 changes.

mrcmunir commented 10 years ago

Good news progress I test with #113 it's working rendering again with Android 2.2 with touch buttons but if push any physical buttons crash by at GL2JNIActivity.java:413 Handle_Key and GL2JNIActivity.OnKeydown line 441.

skmp commented 10 years ago

Should build and work fine (fixed it). Also merged in the sh4rec changes from @ptitSeb and #113

skmp commented 10 years ago

@MrPsyMan do we save the flash properly? I believe it was fix .. perhaps it was only on the pandora version?

skmp commented 10 years ago

@LoungeKatt on startup it sure takes a while .. any idea why? I'll look into it tomorrow night if its still unknown :)

Also, added @LoungeKatt @bwRavencl and @hooby3dfx to the team. Sorry for taking so long - though I already did that last week ...

skmp commented 10 years ago

Also, while i'm spamming anyway, perhaps back should get back to the browser if you're in the options fragment?

AbandonedCart commented 10 years ago

LOL. Twistedumbrella on Google if you needed to get me without posting.

Startup is searching through the entire base directory for the game path. I can see what can be done to optimize it but the best way is setting a specific location. I'll work out a lazy load that adds as it finds so people see progress instead of delay.

Agreed on back. I'll make sure all the "sub-fragment" have their own override. On Jan 20, 2014 7:43 PM, "Stefanos Kornilios Mitsis Poiitidis" < notifications@github.com> wrote:

Also, while i'm spamming anyway, perhaps back should get back to the browser if you're in the options fragment?

— Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-32811497 .

MrPsyMan commented 10 years ago

Wasn't the Shenmue bug a blocker for r6? Lemme see about the flash thing.

hooby3dfx commented 10 years ago

tested out build from buildbot, posted results on forum: http://forums.reicast.com/index.php?topic=9.0

On Mon, Jan 20, 2014 at 11:50 PM, MrPsyMan notifications@github.com wrote:

Wasn't the Shenmue bug a blocker for r6? Lemme see about the flash thing.

— Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-32821316 .

hooby3dfx commented 10 years ago

there are some showstoppers. im going to build my own to verify the buildbot version. ill prob also try and check out some of the ui code soon to help out with that.

On Mon, Jan 20, 2014 at 11:58 PM, Bryan Barnes hooby3d@gmail.com wrote:

tested out build from buildbot, posted results on forum: http://forums.reicast.com/index.php?topic=9.0

On Mon, Jan 20, 2014 at 11:50 PM, MrPsyMan notifications@github.comwrote:

Wasn't the Shenmue bug a blocker for r6? Lemme see about the flash thing.

— Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-32821316 .

MrPsyMan commented 10 years ago

The flash is indeed saved on exit now drk.

The non-bootable shenmue gdi is probably a bad dump.

AbandonedCart commented 10 years ago

LinearLayout in controllers was supposed to still scroll, but easy to fix.

The text issue I haven't seen but I have some updates for the navigation that may have an impact as a side effect.

Eating quick then I'll get what I did earlier into commits to test.

... and that thing where I submitted a full page error log button :/

AbandonedCart commented 10 years ago

https://github.com/reicast/reicast-emulator/pull/121 <- That should cover the MAJOR layout issues for now.

helloray commented 10 years ago

My phone is Galaxy S3.Test Biohazard(Resident Evil) Code:Veronica(JP) and The Record of Lodoss War(JP) with the lastest code in Git.
The graphic related errors seems fixed. Bio CV will jump out of the game randomly during the play. The Record of Lodoss War is freezed during the first battle.

hooby3dfx commented 10 years ago

been testing on latest and found that the black triangle glitches on jet grind radio from r5 were gone!! perfect on shield!

the ui has come a long way too. great work. looks like r6 is pretty much ready to ship.

On Tue, Jan 21, 2014 at 4:37 AM, helloray notifications@github.com wrote:

My phone is Galaxy S3.Test Biohazard(Resident Evil) Code:Veronica(JP) and The Record of Lodoss War(JP) with the lastest code in Git.

The graphic related errors seems fixed. Bio CV will jump out of the game randomly during the play. The Record of Lodoss War is freezed during the first battle.

— Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-32833161 .

AbandonedCart commented 10 years ago

I still get them on shield. Haven't tried that one on n5. It's sad but the issue was that the app just needed the "have at you gpu" flag LOL.

I wonder if we should start configuring egl for special scenarios...

skmp commented 10 years ago

Testing with psy so far:

2.3 ui is broken (after some time doesnt handle input) Moga is broken in mode a Folder changing requires respawn Sometimes kets get messed up

MrPsyMan commented 10 years ago

Unfortunately the release will be put back for a while. Some crap came up and we could not get all the stoppers fixed today (we were supposed to release today).

It's a pity too cause it's like just a couple of hours work to fix them.

skmp commented 10 years ago

Okay, let's go for a release thursday. Considering how many commits make it into the master, I think we should branch off.

Also, gotta prepare up the forums and site stuff on my end.

Ya all registed on the forums (forums.reicast.com)? Send me a pm if I haven't added ya to the developers group yet (I keep getting confused with github/irc/etc nicknames >.>)

AbandonedCart commented 10 years ago

androidui is up to date with master

hooby3dfx commented 10 years ago

what would you all think of potentially jumping straight to r7? are all the r6 blockers done?

On Mon, Jan 27, 2014 at 10:28 AM, TwistedUmbrella notifications@github.comwrote:

androidui is up to date with master

Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-33376941 .

AbandonedCart commented 10 years ago

At this rate, it will be r0.6 final

skmp commented 10 years ago

There's a threading issue on the input stuff that I gotta fix, everything else is fixed!

AbandonedCart commented 10 years ago

The new logging needs a prompt but it isnt vital. Right now it just makes you submit a log because it assumes fatal error.

hooby3dfx commented 10 years ago

@loungekatt - about the sliding menu: looks way better now! 2 issues: -title text gets out of sync when you hit the back button. ex go to input then hit back, main fragment will be the browser but the title still says 'input'. -love how you can pull open the menu via dpad left/right but can we make the rows there focusable too?

On Mon, Feb 3, 2014 at 4:51 PM, TwistedUmbrella notifications@github.comwrote:

The new logging needs a prompt but it isnt vital. Right now it just makes you submit a log because it assumes fatal error.

Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-34004615 .

AbandonedCart commented 10 years ago

NO! Wait, yeah... https://github.com/reicast/reicast-emulator/pull/225

hooby3dfx commented 10 years ago

i added a couple other potential r6 blockers that just came in to the milestone, feel free to take them off. one was crazy taxi not working (regression), the other was one report of lost video output (another regression). also marked the input threading issue as r6.

are we going to cut the build from current master?

On Mon, Feb 3, 2014 at 10:58 PM, TwistedUmbrella notifications@github.comwrote:

NO! Wait, yeah... #225https://github.com/reicast/reicast-emulator/pull/225

Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-34028685 .

hooby3dfx commented 10 years ago

i made a branch for the release so we can approach something of a code freeze there with critical changes, and keep working on other stuff in master. https://github.com/reicast/reicast-emulator/tree/r6-rc sounds like the crazy taxi issue might just be with the pal version. @loungekatt, did you want to revert the changes to the rtc? also, is the input threading issue a regression? i havent seen it happen - just wondering that should go into r7 instead.

On Mon, Feb 3, 2014 at 11:26 PM, TwistedUmbrella notifications@github.comwrote:

That video issue seems like something is up. I can't replicate it on anything.

Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-34029680 .

AbandonedCart commented 10 years ago

https://github.com/reicast/reicast-emulator/commit/bf4e05391c2e2f77f883507896f519a0a34aeb99 -> https://github.com/reicast/reicast-emulator/commit/3db43e691e5368d00bdb6f976485f90aba61be64

That should take care of that.

AbandonedCart commented 10 years ago

Just for the sake of saying it, the purpose of getting androidui caught up was so master could be releases and that could be other stuff. We can probably delete androidui now and do release branches. That works much better with the beta changelog tracker anyway.

hooby3dfx commented 10 years ago

@skmp + @MrPsyMan let us know if you need help with any other aspects of managing the project like the site/forums/store/handling emails or whatever. i think the current r6 candidate plays awesome and we should get it into the hands of the masses. as stated its an alpha so its not going to be perfect but an update would give people hope and keep the interest up.

MrPsyMan commented 10 years ago

We still have a couple of stoppers that have to be fixed. At least until the input problem with physical buttons that stop working correctly is fixed, we don't release.

We should find a commit with the rest stoppers fixed (ui, moga, droid 2.x, etcetera fixes) and make an r6 release branch from it though. Someone has to check in which recent commit the UI behaves correctly on droid 2.x devices and devices without touch inputs (like ouya).

skmp commented 10 years ago

Sorry, guys, i've been busy hacking with the oculus this weekend ...

The input is on the wrong thread right now, thats what kills input on Shield after a while.

(in related news, i'm totally overloaded for this month at 70+ hours/week worktime, thats why i'm so MIA)

On Sun, Feb 9, 2014 at 12:21 PM, MrPsyMan notifications@github.com wrote:

We still have a couple of stoppers that have to be fixed. At least until the input problem with physical buttons that stop working correctly is fixed, we don't release.

We should find a commit with the rest stoppers fixed (ui, moga, droid 2.x, etcetera fixes) and make an r6 release branch from it though. Someone has to check in which recent commit the UI behaves correctly on droid 2.x devices and devices without touch inputs (like ouya).

— Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-34570298 .

~skmp

AbandonedCart commented 10 years ago

@skmp If there is anything else you need help with, my school and job are remote so I get to manage my time on a somewhat flexible schedule. I was actually given an assignment of joining a project like this, so it gets filtered in as "school work" in a way haha.

AbandonedCart commented 10 years ago

@MrPsyMan the r6-rc branch @hooby3dfx made is the "if we had to release right now" branch. Everything I added since it was made was verified against the blockers. Running some GB tests now as a precaution.

MrPsyMan commented 10 years ago

Ah that's great. So drk will just need to push the input fix, we check if it works, then release. Problem is finding the time to do the fix.

hooby3dfx commented 10 years ago

r6 testing (actually tested on 853db51) on shield -in paths menu browse buttons not focusable (not new, just noticed) -hitting back when browsing paths closes app -hitting back after going into the in-game popup's more settings is weird; doesnt close popup or go back -how is the widescreen setting suppose to work (some games get some graphics on the sides - is that expected?) -limitfps takes a few taps of the button to activate -noticing new general slowdown in some games (virtua tennis) -couldnt get software renderer to enable. is it obvious once enabled?

On Sun, Feb 9, 2014 at 11:43 AM, MrPsyMan notifications@github.com wrote:

Ah that's great. So drk will just need to push the input fix, we check if it works, then release. Problem is finding the time to do the fix.

Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-34578561 .

helloray commented 10 years ago

@skmp , can you help to check the memory leak issue #252 and #254. It can be easily duplicated. I think it's also a showstopper.

AbandonedCart commented 10 years ago

@hooby3dfx I know you're anxious to get r6 out, but give me today to finish a massive overhaul. The entire project will be clean and organized, settings will be categorized and make sense, and there are a bunch of optimizations. Sadly I can only type so fast and Dragon dictation doesn't work well with Eclipse.

psennermann commented 10 years ago

I downloaded reicast-emulatror-036be40.apk (036be40) from build archive of http://loungekatt.com/ReicastBot because I thought it was the latest version, then I noticed that new versions were added downward so I downloaded the much newer f59da37.apk (althoug in the info always says r6(6) for both)...well I was very surprised to discover that the first one (older one) is much better: it's faster and (most important) the audio is smooth and doesn't crackle all the time like in f59da37.apk ...so I'm here to ask what has happened? (I use a Quad core 1.2 ghz tabled with poverVR gpu and Android 4.2.2)

AbandonedCart commented 10 years ago

@psennermann The comment above yours could also be applied to your question.

http://loungekatt.com/ReicastBot/experimental/compiled/reicast-experimental-b2e3a2a.apk You can see if that one has the same results, but it is just a test for the input push.

AbandonedCart commented 10 years ago

@hooby3dfx I am beginning to think that your search for the lag you experienced should start with your input handling. My phone is fast enough to not notice a difference, but I didn't heavily test the before / after with something slower.

It is also possible the audio was a result of giving thread priority to the emu. I took care of needing that by adding the option to disable audio.

hooby3dfx commented 10 years ago

testing from my fork (5 days old) that includes the input changes doesnt show the lag.

On Wed, Feb 12, 2014 at 4:26 PM, TwistedUmbrella notifications@github.comwrote:

@hooby3dfx https://github.com/hooby3dfx I am beginning to think that your search for the lag you experienced should start with your input handling. My phone is fast enough to not notice a difference, but I didn't heavily test the before / after with something slower.

Reply to this email directly or view it on GitHubhttps://github.com/reicast/reicast-emulator/issues/118#issuecomment-34919794 .

AbandonedCart commented 10 years ago

@hooby3dfx I tried with and without that. I tried with and without the thread priority. The only other thing I could think of it some of these devices were using software rendering and now that it is either on or off without just being whatever, the lacking hardware support (drivers, etc) is showing. Maybe I should set software as default.