Trying to fix sceAtracGetBufferInfoForResetting #4303
Conversation
| @@ -730,7 +736,7 @@ u32 sceAtracGetBufferInfoForResetting(int atracID, int sample, u32 bufferInfoAdd | |||
| bufferInfo->first.filePos = Sampleoffset; | |||
|
|
|||
| // TODO: It seems like this is always the same as the first buffer's pos? | |||
| bufferInfo->second.writePosPtr = atrac->first.addr; | |||
| bufferInfo->second.writePosPtr = atrac->second.addr; | |||
There was a problem hiding this comment.
Read the comment. Test results show it uses the first value. Did this change affect games? It didn't for me. Didn't even seem like games used it.
-[Unknown]
|
GTA VIce City Stories undergoes a further audio regression as compared to the current master with this commit. All ingame BGM playback (excluding the FMV cutscenes which still sound normal, just like the current master) is just reduced to a repeated beeping sound that loops ad infinitum Here's a sample of what it sounds like, taken ingame: GTA Chinatown Wars is unchanged with this commit. The game does not crash after the white jet flies through the screen in the opening sequence, but the audio is still cutoff after that point, just like in the current master. |
|
Does this change improve the game? |
|
GTA Chinatown Wars is definitely improved with this last update you made to the commit. The BGM doesn't cutoff anymore during the entire jet opening sequence where you're introduced to the main character and his background; no crashes either. Good work here. GTA VIce City Stories however, is a bit of a mixed bag. On one hand, the BGM is back to sounding normal (no more beeping noises), but on the other hand, the BGM it self is a lot more stuttery and performance definitely takes a hit as compared to before. This is easily experienced by riding on a bike/any other fast moving vehicle and driving around town with the radio switched on. Still, its an overall improvement to the current situation, but I'm not sure if it's correct, off to work I go. |
|
@solarmystic , stutteryness seems like an unlikely consequence of this change. Are you sure? |
|
I was rather taken aback by that finding myself in that particular GTA game (tested with GTA VCS), and so I proceeded to boot up the latest master for comparison to see whether or not it was already present there. The current master (v0.9.1-2122-g36d433c) does not have that particular issue, but on the other hand, the BGM just cuts off and on randomly while playing audio tracks in a vehicle. The commit by @papel resolves that to a larger extent (no more BGM cutoffs like the current master has), but introduces increased stutteryness when switching between radio stations and/or heading to/loading newer areas while in a fast vehicle, with regards to the performance (greater occurances of momentary speed hitches) -Posted from my mobile device- |
|
Does this introduce problems in any other games? I'm trying to decide whether or not to merge this before 0.9.5. |
|
A last attempt. It reverts #3772, but does not break Ys Seven, because it handles what is different in Ys7 variables (https://github.com/papel/ppsspp/compare/test). |
Fix stencil testing in the softgpu, and also some small things along the way
Windows: Remove MemUsage (from Dolphin) as it was unused. Fix some tabbing.
Was breaking non-fastmem lwl/lwr/etc.
I analysed the input values of sceAtracGetBufferInfoForResetting to choose the output.
It works with both Trails in the Sky (#4003) and Ys Seven (#2391). I do not have enough games to test. Henrik asked to open a pull request to discuss about it.
@unknownbrackets, does this change make sense?
@solarmystic, could you test the GTA games?