Commit graph

135 commits

Author SHA1 Message Date
Henrik Rydgard
c8e719e165 Inline ADSREnvelope::Step (thanks @unknownbrackets)
(Some compilers want it declared inline in the header too)
2014-03-22 09:31:09 +01:00
Unknown W. Brackets
da1e4a80a6 Tone down sas raw output reporting, found a game.
Crimson Gem Saga apparently uses this mode.
2014-03-01 15:32:34 -08:00
Unknown W. Brackets
be79670efb Revert "Smooth SAS resampling when pitch != 0x1000."
This reverts commit 1efcebb121.
2014-02-23 08:48:49 -08:00
Unknown W. Brackets
c806ae0982 Revert "Optimize mixing samples at common pitches."
This reverts commit dd434cd7c5.

Conflicts:
	Core/HW/SasAudio.cpp
2014-02-23 08:48:30 -08:00
Unknown W. Brackets
97e8f36a9b Revert "Actually use the resample buffer."
This reverts commit dc5d17971c.
2014-02-23 08:46:33 -08:00
Unknown W. Brackets
d6f203db99 Revert "Reset resample hist on keyon, round up read."
This reverts commit 3edebbfba9.
2014-02-23 08:45:46 -08:00
Unknown W. Brackets
3edebbfba9 Reset resample hist on keyon, round up read.
If we have an uneven pitch, round up so that we get the next sample to
resample in.
2014-02-22 22:38:03 -08:00
Unknown W. Brackets
9cba543a4b sceSasCoreWithMix() returns an error in raw mode.
And does not reschedule, specifically.
2014-02-22 08:48:55 -08:00
Unknown W. Brackets
0939065333 Report the raw output mode to find examples.
I can't find any games actually leaving the output mode at 1.
2014-02-22 08:07:51 -08:00
Unknown W. Brackets
ff920f6af6 Fix sceSas outputMode = 1 (which is not mono.)
May fix games that played audio only on the left side, or something.
2014-02-22 07:52:18 -08:00
Unknown W. Brackets
a1283644ab If never set, loop to the first vag block.
Not the second (0 = second.)
2014-02-19 01:03:11 -08:00
Unknown W. Brackets
a36f7bf267 Fill in predict_nr's above 5, fix clamping.
We weren't processing vag properly when it hit the edges.

The coefficients were found with testing about 384 samples.
2014-02-19 01:01:56 -08:00
Unknown W. Brackets
11df9dabf9 Add an extra delay to VAG keyon.
It seems to come in an extra sample late.  PCM doesn't.

This corrects timing for VAG samples so they match up exactly.  Really
minor, of course...
2014-02-18 23:32:37 -08:00
Unknown W. Brackets
dc5d17971c Actually use the resample buffer.
Fixes static in Final Fantasy Tactics music.
2014-02-17 23:23:44 -08:00
Unknown W. Brackets
93060e3aef Avoid some divisions, just in case. 2014-02-17 23:19:30 -08:00
Unknown W. Brackets
b5bdac211a Ignore pitch > 0x1000 for pcm.
It seems to be supported only below 0x1000.  Also, drop optimization for
double for now, simpler this way and doesn't seem common?
2014-02-17 21:09:44 -08:00
Unknown W. Brackets
dd434cd7c5 Optimize mixing samples at common pitches.
Doesn't seem to be a big help but may matter more on mobile.
2014-02-17 21:09:44 -08:00
Unknown W. Brackets
1efcebb121 Smooth SAS resampling when pitch != 0x1000. 2014-02-17 21:09:43 -08:00
Unknown W. Brackets
8ded863231 Reset sampleFrac on keyon.
Since we're starting the samples over.
2014-02-17 21:09:42 -08:00
Unknown W. Brackets
38053df4ca Don't drop the last mix of vag samples.
Wait instead until the end of mixing to end the envelope.
2014-02-17 21:09:42 -08:00
Unknown W. Brackets
3899e18422 Always key on for 32 samples regardless of pitch.
Before it would be more or less depending on pitch.  It's always 32.
2014-02-17 21:09:41 -08:00
Unknown W. Brackets
e006ab5b2d Don't read the first 32 samples after keyon.
The way I'm approximating this is ugly...
2014-02-17 01:39:59 -08:00
Unknown W. Brackets
6b984c5593 Don't loop PCM samples when loopPos < 0. 2014-02-17 01:39:59 -08:00
Unknown W. Brackets
95fc9624b4 Fix PCM looping in SAS.
The parameter specifies where the loop restarts at.
2014-02-17 01:39:58 -08:00
Unknown W. Brackets
55507d7844 Fix the default adsr sustain mode.
Probably doesn't matter, games always set it.
2014-02-17 01:39:57 -08:00
Unknown W. Brackets
d097ac5394 Don't change height when keying off a voice.
Tests show that the height does not revert to the sustain level.
2014-02-17 01:39:56 -08:00
Unknown W. Brackets
d53ab02e5b Don't change height immediately on keyon.
This replicates the PSP's behavior, which is a 32-grain lead in, plus the
height remains what it was before until processing happens.
2014-02-17 01:39:55 -08:00
Unknown W. Brackets
684caa01ec Correct SAS bent linear curve.
Tests shows that at 0x30000000 it still goes up the full amount.
2014-02-17 01:39:54 -08:00
Henrik Rydgård
caf729e01e Remove call to SetState, superfluous now that the state variables are gone. 2014-02-11 17:05:46 +01:00
Henrik Rydgård
079f653cff Sas envelopes: Simplify away some redundant state. 2014-02-11 11:56:24 +01:00
Unknown W. Brackets
0265fc757d Flip and shift instead of signed divide.
Clearer than all those zeroes and maybe faster...
2014-02-11 01:29:33 -08:00
Unknown W. Brackets
b398a84517 Attack switches to decay at max height already.
It does not wait until it goes above.
2014-02-11 00:18:36 -08:00
Unknown W. Brackets
95a8af9e60 Decay does not switch to release at height max.
This can happen with an exponent rate of 0.  It stays there.
2014-02-11 00:17:13 -08:00
Unknown W. Brackets
f11d362be3 Fix integer overflow in exponent decrease. 2014-02-11 00:11:07 -08:00
Unknown W. Brackets
f981d39f94 Sustain defaults to 0, not 0x100.
Not sure how I came to that mistaken idea.
2014-02-10 23:48:10 -08:00
Unknown W. Brackets
6392307cbd Calculate the curve on the fly per the PSP.
Values based on lots of testing from the PSP, and match exactly.
2014-02-10 08:43:01 -08:00
Unknown W. Brackets
df7d25f4bf Correct a case where release rate was set negative.
Seems to be capped at INT_MAX like the others, oops.
2014-02-09 11:30:15 -08:00
Unknown W. Brackets
a14850c16b Drop height to 0 when VAG ends.
Otherwise we will set it to !playing and never step the height.
Should fix Yu-Gi-Oh again.  Probably release before was miscalculated as
0 which is why it worked before.
2014-02-09 10:53:47 -08:00
Unknown W. Brackets
074ef84559 Correct the rates set by sceSasSetSimpleADSR().
They were wrong for a few combinations before.  Could have effected how
accurate sound effects were (they may have been too quite, or dropped off
wrong, etc.)
2014-02-07 22:29:54 -08:00
raven02
3f20b5ad77 sceSas : should be always delay thread 2014-02-04 20:03:54 +08:00
Unknown W. Brackets
a00d5a2df8 Set a voice as keyed off when its envelope ends. 2014-02-04 01:35:53 -08:00
raven02
8e8329abec Fix audio gitches 2014-01-27 19:32:35 +08:00
Henrik Rydgård
0414a3faa4 Not having voices playing is not an error. Disabling logging of it. 2014-01-26 16:32:29 +01:00
raven02
7701784608 Not reschelding if voicesPlayingCount == 0 2014-01-26 19:52:31 +08:00
Henrik Rydgård
b333e5b77f Comment the fallthroughs in SasAudio.cpp 2014-01-23 20:58:42 +01:00
Henrik Rydgård
6c5165e505 Revert "Merge pull request #5190 from raven02/patch-32"
This reverts commit aef180259a, reversing
changes made to d69f02dea0.
2014-01-23 20:57:39 +01:00
raven02
9965bbb43e Shouldn't voice.type == VOICETYPE_VAG/PCM 2014-01-24 00:28:53 +08:00
Unknown W. Brackets
4102e649d5 Key off the envelope when the VAG/etc. ends.
Fixes Yu-Gi-Oh Tag Force GX 3 hanging in some places.
2014-01-13 21:42:42 -08:00
Unknown W. Brackets
79864a5ee0 Fix some initialization order warnings. 2014-01-10 22:21:24 -08:00
Henrik Rydgard
ddfdeb2222 VAGADPCM: no need to check end_ for every sample, it's only modified in DecodeBlock (and Start). 2014-01-04 11:01:44 +01:00