aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorChris Robinson <[email protected]>2015-10-23 20:16:11 -0700
committerChris Robinson <[email protected]>2015-10-23 20:16:11 -0700
commit714354caeef1927c3b5ecd132d7cc090e7672106 (patch)
tree53ed957b75cfba32615c93d7ceef6d7735fa5a49
parentbca854baacf0c92038f0e2633fb92e9255b05603 (diff)
Set XYZ channel gains for source sends to 0
It's cleaner to just set the gains to 0 rather than to special-case B-Format in the mixer.
-rw-r--r--Alc/ALu.c187
-rw-r--r--Alc/mixer.c18
2 files changed, 107 insertions, 98 deletions
diff --git a/Alc/ALu.c b/Alc/ALu.c
index bcb73651..b4927ee6 100644
--- a/Alc/ALu.c
+++ b/Alc/ALu.c
@@ -531,108 +531,131 @@ ALvoid CalcNonAttnSourceParams(ALvoice *voice, const ALsource *ALSource, const A
voice->Direct.Moving = AL_TRUE;
voice->IsHrtf = AL_FALSE;
+
for(i = 0;i < NumSends;i++)
- WetGain[i] *= 1.4142f;
+ {
+ /* Only the first channel of B-Format buffers (W) goes to auxiliary
+ * sends. It also needs to be scaled by sqrt(2) to account for the
+ * signal being scaled by sqrt(1/2).
+ */
+ voice->Send[i].Gains[0].Target = WetGain[i] * 1.4142f;
+ for(c = 1;c < num_channels;c++)
+ voice->Send[i].Gains[c].Target = 0.0f;
+ UpdateWetStepping(&voice->Send[i], num_channels, (voice->Send[i].Moving ? 64 : 0));
+ voice->Send[i].Moving = AL_TRUE;
+ }
}
- else if(DirectChannels != AL_FALSE)
+ else
{
- if(Device->Hrtf)
+ if(DirectChannels)
{
- voice->Direct.OutBuffer += voice->Direct.OutChannels;
- voice->Direct.OutChannels = 2;
- for(c = 0;c < num_channels;c++)
+ if(Device->Hrtf)
+ {
+ /* DirectChannels with HRTF enabled. Skip the virtual channels
+ * and write FrontLeft and FrontRight inputs to the first and
+ * second outputs.
+ */
+ voice->Direct.OutBuffer += voice->Direct.OutChannels;
+ voice->Direct.OutChannels = 2;
+ for(c = 0;c < num_channels;c++)
+ {
+ MixGains *gains = voice->Direct.Gains[c];
+
+ for(j = 0;j < MAX_OUTPUT_CHANNELS;j++)
+ gains[j].Target = 0.0f;
+
+ if(chans[c].channel == FrontLeft)
+ gains[0].Target = DryGain;
+ else if(chans[c].channel == FrontRight)
+ gains[1].Target = DryGain;
+ }
+ }
+ else for(c = 0;c < num_channels;c++)
{
MixGains *gains = voice->Direct.Gains[c];
+ int idx;
for(j = 0;j < MAX_OUTPUT_CHANNELS;j++)
gains[j].Target = 0.0f;
-
- if(chans[c].channel == FrontLeft)
- gains[0].Target = DryGain;
- else if(chans[c].channel == FrontRight)
- gains[1].Target = DryGain;
+ if((idx=GetChannelIdxByName(Device, chans[c].channel)) != -1)
+ gains[idx].Target = DryGain;
}
- }
- else for(c = 0;c < num_channels;c++)
- {
- MixGains *gains = voice->Direct.Gains[c];
- int idx;
+ UpdateDryStepping(&voice->Direct, num_channels, (voice->Direct.Moving ? 64 : 0));
+ voice->Direct.Moving = AL_TRUE;
- for(j = 0;j < MAX_OUTPUT_CHANNELS;j++)
- gains[j].Target = 0.0f;
- if((idx=GetChannelIdxByName(Device, chans[c].channel)) != -1)
- gains[idx].Target = DryGain;
+ voice->IsHrtf = AL_FALSE;
}
- UpdateDryStepping(&voice->Direct, num_channels, (voice->Direct.Moving ? 64 : 0));
- voice->Direct.Moving = AL_TRUE;
-
- voice->IsHrtf = AL_FALSE;
- }
- else if(Device->Hrtf_Mode == FullHrtf)
- {
- voice->Direct.OutBuffer += voice->Direct.OutChannels;
- voice->Direct.OutChannels = 2;
- for(c = 0;c < num_channels;c++)
+ else if(Device->Hrtf_Mode == FullHrtf)
{
- if(chans[c].channel == LFE)
+ /* Full HRTF rendering. Skip the virtual channels and render each
+ * input channel to the real outputs.
+ */
+ voice->Direct.OutBuffer += voice->Direct.OutChannels;
+ voice->Direct.OutChannels = 2;
+ for(c = 0;c < num_channels;c++)
{
- /* Skip LFE */
- voice->Direct.Hrtf[c].Params.Delay[0] = 0;
- voice->Direct.Hrtf[c].Params.Delay[1] = 0;
- for(i = 0;i < HRIR_LENGTH;i++)
+ if(chans[c].channel == LFE)
{
- voice->Direct.Hrtf[c].Params.Coeffs[i][0] = 0.0f;
- voice->Direct.Hrtf[c].Params.Coeffs[i][1] = 0.0f;
+ /* Skip LFE */
+ voice->Direct.Hrtf[c].Params.Delay[0] = 0;
+ voice->Direct.Hrtf[c].Params.Delay[1] = 0;
+ for(i = 0;i < HRIR_LENGTH;i++)
+ {
+ voice->Direct.Hrtf[c].Params.Coeffs[i][0] = 0.0f;
+ voice->Direct.Hrtf[c].Params.Coeffs[i][1] = 0.0f;
+ }
+ }
+ else
+ {
+ /* Get the static HRIR coefficients and delays for this
+ * channel. */
+ GetLerpedHrtfCoeffs(Device->Hrtf,
+ chans[c].elevation, chans[c].angle, 1.0f, DryGain,
+ voice->Direct.Hrtf[c].Params.Coeffs,
+ voice->Direct.Hrtf[c].Params.Delay
+ );
}
}
- else
- {
- /* Get the static HRIR coefficients and delays for this
- * channel. */
- GetLerpedHrtfCoeffs(Device->Hrtf,
- chans[c].elevation, chans[c].angle, 1.0f, DryGain,
- voice->Direct.Hrtf[c].Params.Coeffs,
- voice->Direct.Hrtf[c].Params.Delay);
- }
- }
- voice->Direct.Counter = 0;
- voice->Direct.Moving = AL_TRUE;
+ voice->Direct.Counter = 0;
+ voice->Direct.Moving = AL_TRUE;
- voice->IsHrtf = AL_TRUE;
- }
- else
- {
- for(c = 0;c < num_channels;c++)
+ voice->IsHrtf = AL_TRUE;
+ }
+ else
{
- MixGains *gains = voice->Direct.Gains[c];
- ALfloat Target[MAX_OUTPUT_CHANNELS];
-
- /* Special-case LFE */
- if(chans[c].channel == LFE)
+ /* Basic or no HRTF rendering. Use normal panning to the output. */
+ for(c = 0;c < num_channels;c++)
{
- int idx;
+ MixGains *gains = voice->Direct.Gains[c];
+ ALfloat Target[MAX_OUTPUT_CHANNELS];
+
+ /* Special-case LFE */
+ if(chans[c].channel == LFE)
+ {
+ int idx;
+ for(i = 0;i < MAX_OUTPUT_CHANNELS;i++)
+ gains[i].Target = 0.0f;
+ if((idx=GetChannelIdxByName(Device, chans[c].channel)) != -1)
+ gains[idx].Target = DryGain;
+ continue;
+ }
+
+ ComputeAngleGains(Device, chans[c].angle, chans[c].elevation, DryGain, Target);
for(i = 0;i < MAX_OUTPUT_CHANNELS;i++)
- gains[i].Target = 0.0f;
- if((idx=GetChannelIdxByName(Device, chans[c].channel)) != -1)
- gains[idx].Target = DryGain;
- continue;
+ gains[i].Target = Target[i];
}
+ UpdateDryStepping(&voice->Direct, num_channels, (voice->Direct.Moving ? 64 : 0));
+ voice->Direct.Moving = AL_TRUE;
- ComputeAngleGains(Device, chans[c].angle, chans[c].elevation, DryGain, Target);
- for(i = 0;i < MAX_OUTPUT_CHANNELS;i++)
- gains[i].Target = Target[i];
+ voice->IsHrtf = AL_FALSE;
+ }
+ for(i = 0;i < NumSends;i++)
+ {
+ for(c = 0;c < num_channels;c++)
+ voice->Send[i].Gains[c].Target = WetGain[i];
+ UpdateWetStepping(&voice->Send[i], num_channels, (voice->Send[i].Moving ? 64 : 0));
+ voice->Send[i].Moving = AL_TRUE;
}
- UpdateDryStepping(&voice->Direct, num_channels, (voice->Direct.Moving ? 64 : 0));
- voice->Direct.Moving = AL_TRUE;
-
- voice->IsHrtf = AL_FALSE;
- }
- for(i = 0;i < NumSends;i++)
- {
- for(c = 0;c < num_channels;c++)
- voice->Send[i].Gains[c].Target = WetGain[i];
- UpdateWetStepping(&voice->Send[i], num_channels, (voice->Send[i].Moving ? 64 : 0));
- voice->Send[i].Moving = AL_TRUE;
}
{
@@ -998,7 +1021,9 @@ ALvoid CalcSourceParams(ALvoice *voice, const ALsource *ALSource, const ALCconte
if(Device->Hrtf_Mode == FullHrtf)
{
- /* Use a binaural HRTF algorithm for stereo headphone playback */
+ /* Full HRTF rendering. Skip the virtual channels and render to the
+ * real outputs.
+ */
aluVector dir = {{ 0.0f, 0.0f, -1.0f, 0.0f }};
ALfloat ev = 0.0f, az = 0.0f;
ALfloat radius = ALSource->Radius;
@@ -1035,7 +1060,8 @@ ALvoid CalcSourceParams(ALvoice *voice, const ALsource *ALSource, const ALCconte
delta = CalcFadeTime(voice->Direct.LastGain, DryGain,
&voice->Direct.LastDir, &dir);
/* If the delta is large enough, get the moving HRIR target
- * coefficients, target delays, steppping values, and counter. */
+ * coefficients, target delays, steppping values, and counter.
+ */
if(delta > 0.000015f)
{
ALuint counter = GetMovingHrtfCoeffs(Device->Hrtf,
@@ -1064,6 +1090,7 @@ ALvoid CalcSourceParams(ALvoice *voice, const ALsource *ALSource, const ALCconte
}
else
{
+ /* Basic or no HRTF rendering. Use normal panning to the output. */
MixGains *gains = voice->Direct.Gains[0];
ALfloat dir[3] = { 0.0f, 0.0f, -1.0f };
ALfloat radius = ALSource->Radius;
diff --git a/Alc/mixer.c b/Alc/mixer.c
index 35c38431..e8b94b17 100644
--- a/Alc/mixer.c
+++ b/Alc/mixer.c
@@ -290,7 +290,6 @@ ALvoid MixSource(ALvoice *voice, ALsource *Source, ALCdevice *Device, ALuint Sam
ResamplerFunc Resample;
ALbufferlistitem *BufferListItem;
ALuint DataPosInt, DataPosFrac;
- ALboolean isbformat = AL_FALSE;
ALboolean Looping;
ALuint increment;
ALenum State;
@@ -311,19 +310,6 @@ ALvoid MixSource(ALvoice *voice, ALsource *Source, ALCdevice *Device, ALuint Sam
SampleSize = Source->SampleSize;
increment = voice->Step;
- while(BufferListItem)
- {
- ALbuffer *buffer;
- if((buffer=BufferListItem->buffer) != NULL)
- {
- isbformat = (buffer->FmtChannels == FmtBFormat2D ||
- buffer->FmtChannels == FmtBFormat3D);
- break;
- }
- BufferListItem = BufferListItem->next;
- }
- assert(BufferListItem != NULL);
-
IrSize = (Device->Hrtf ? GetHrtfIrSize(Device->Hrtf) : 0);
Resample = ((increment == FRACTIONONE && DataPosFrac == 0) ?
@@ -488,10 +474,6 @@ ALvoid MixSource(ALvoice *voice, ALsource *Source, ALCdevice *Device, ALuint Sam
&parms->Hrtf[chan].State, DstBufferSize);
}
- /* Only the first channel for B-Format buffers (W channel) goes to
- * the send paths. */
- if(chan > 0 && isbformat)
- continue;
for(j = 0;j < Device->NumAuxSends;j++)
{
SendParams *parms = &voice->Send[j];