[PATCH 1/2] ALSA: emu10k1: fix synthesizer sample playback position and caching
Compensate for the cache delay, and actually populate the cache. Without these, the playback would start with garbage (which would be (mostly?) masqueraded by the attack phase).
Unlike for the PCM voices, this doesn't try to compensate for the interpolator read-ahead, because it's pointless to be super-exact here.
Note that this code is probably still broken for particularly short samples, because we ignore the loop-related parts of CCR. But I'm not going to reverse-engineer that now ...
Signed-off-by: Oswald Buddenhagen oswald.buddenhagen@gmx.de --- sound/pci/emu10k1/emu10k1_callback.c | 36 ++++------------------------ 1 file changed, 5 insertions(+), 31 deletions(-)
diff --git a/sound/pci/emu10k1/emu10k1_callback.c b/sound/pci/emu10k1/emu10k1_callback.c index 06440b97b5d7..aab8d64fd708 100644 --- a/sound/pci/emu10k1/emu10k1_callback.c +++ b/sound/pci/emu10k1/emu10k1_callback.c @@ -253,7 +253,7 @@ lookup_voices(struct snd_emux *emu, struct snd_emu10k1 *hw, /* check if sample is finished playing (non-looping only) */ if (bp != best + V_OFF && bp != best + V_FREE && (vp->reg.sample_mode & SNDRV_SFNT_SAMPLE_SINGLESHOT)) { - val = snd_emu10k1_ptr_read(hw, CCCA_CURRADDR, vp->ch); + val = snd_emu10k1_ptr_read(hw, CCCA_CURRADDR, vp->ch) - 64; if (val >= vp->reg.loopstart) bp = best + V_OFF; } @@ -360,7 +360,7 @@ start_voice(struct snd_emux_voice *vp)
map = (hw->silent_page.addr << hw->address_mode) | (hw->address_mode ? MAP_PTI_MASK1 : MAP_PTI_MASK0);
- addr = vp->reg.start; + addr = vp->reg.start + 64; temp = vp->reg.parm.filterQ; ccca = (temp << 28) | addr; if (vp->apitch < 0xe400) @@ -428,40 +428,14 @@ start_voice(struct snd_emux_voice *vp) /* Q & current address (Q 4bit value, MSB) */ CCCA, ccca,
+ /* cache */ + CCR, REG_VAL_PUT(CCR_CACHEINVALIDSIZE, 64), + /* reset volume */ VTFT, vtarget | vp->ftarget, CVCF, vtarget | CVCF_CURRENTFILTER_MASK,
REGLIST_END); -#if 0 - /* cache */ - { - unsigned int val, sample; - val = 32; - if (vp->reg.sample_mode & SNDRV_SFNT_SAMPLE_8BITS) - sample = 0x80808080; - else { - sample = 0; - val *= 2; - } - - /* cache */ - snd_emu10k1_ptr_write(hw, CCR, ch, 0x1c << 16); - snd_emu10k1_ptr_write(hw, CDE, ch, sample); - snd_emu10k1_ptr_write(hw, CDF, ch, sample); - - /* invalidate maps */ - temp = ((unsigned int)hw->silent_page.addr << hw_address_mode) | (hw->address_mode ? MAP_PTI_MASK1 : MAP_PTI_MASK0); - snd_emu10k1_ptr_write(hw, MAPA, ch, temp); - snd_emu10k1_ptr_write(hw, MAPB, ch, temp); - - /* fill cache */ - val -= 4; - val <<= 25; - val |= 0x1c << 16; - snd_emu10k1_ptr_write(hw, CCR, ch, val); - } -#endif
return 0; }
Make terminate_voice() actually do at all what it's supposed to do: instantly and completely shut down the note.
The bogus behavior was mostly harmless, as usually the voice is freed right afterwards, which implicitly terminates it anyway.
Signed-off-by: Oswald Buddenhagen oswald.buddenhagen@gmx.de --- sound/pci/emu10k1/emu10k1_callback.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/sound/pci/emu10k1/emu10k1_callback.c b/sound/pci/emu10k1/emu10k1_callback.c index aab8d64fd708..dcd7be2d281b 100644 --- a/sound/pci/emu10k1/emu10k1_callback.c +++ b/sound/pci/emu10k1/emu10k1_callback.c @@ -136,8 +136,13 @@ terminate_voice(struct snd_emux_voice *vp) if (snd_BUG_ON(!vp)) return; hw = vp->hw; - snd_emu10k1_ptr_write(hw, DCYSUSV, vp->ch, - DCYSUSV_PHASE1_MASK | DCYSUSV_DECAYTIME_MASK | DCYSUSV_CHANNELENABLE_MASK); + snd_emu10k1_ptr_write_multiple(hw, vp->ch, + DCYSUSV, 0, + VTFT, VTFT_FILTERTARGET_MASK, + CVCF, CVCF_CURRENTFILTER_MASK, + PTRX, 0, + CPF, 0, + REGLIST_END); if (vp->block) { struct snd_emu10k1_memblk *emem; emem = (struct snd_emu10k1_memblk *)vp->block;
On Thu, 18 May 2023 16:03:38 +0200, Oswald Buddenhagen wrote:
Compensate for the cache delay, and actually populate the cache. Without these, the playback would start with garbage (which would be (mostly?) masqueraded by the attack phase).
Unlike for the PCM voices, this doesn't try to compensate for the interpolator read-ahead, because it's pointless to be super-exact here.
Note that this code is probably still broken for particularly short samples, because we ignore the loop-related parts of CCR. But I'm not going to reverse-engineer that now ...
Signed-off-by: Oswald Buddenhagen oswald.buddenhagen@gmx.de
Applied both patches now. Thanks.
Takashi
participants (2)
-
Oswald Buddenhagen
-
Takashi Iwai