aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--Alc/ALu.c13
-rw-r--r--Alc/bformatdec.c35
-rw-r--r--Alc/bformatdec.h6
-rw-r--r--Alc/mixer.c13
-rw-r--r--Alc/mixer_c.c4
-rw-r--r--Alc/mixer_defs.h15
-rw-r--r--Alc/mixer_neon.c6
-rw-r--r--Alc/mixer_sse.c6
-rw-r--r--OpenAL32/Include/alMain.h42
-rw-r--r--OpenAL32/Include/alu.h5
10 files changed, 98 insertions, 47 deletions
diff --git a/Alc/ALu.c b/Alc/ALu.c
index 13aff5d9..5bcf039e 100644
--- a/Alc/ALu.c
+++ b/Alc/ALu.c
@@ -1489,10 +1489,9 @@ ALvoid aluMixData(ALCdevice *device, ALvoid *buffer, ALsizei size)
slot = slotroot;
while(slot)
{
- const ALeffectslot *cslot = slot;
- ALeffectState *state = cslot->Params.EffectState;
- V(state,process)(SamplesToDo, cslot->WetBuffer, state->OutBuffer,
- state->OutChannels);
+ ALeffectState *state = slot->Params.EffectState;
+ V(state,process)(SamplesToDo, SAFE_CONST(ALfloatBUFFERSIZE*,slot->WetBuffer),
+ state->OutBuffer, state->OutChannels);
slot = ATOMIC_LOAD(&slot->next, almemory_order_relaxed);
}
@@ -1540,19 +1539,19 @@ ALvoid aluMixData(ALCdevice *device, ALvoid *buffer, ALsizei size)
{
if(device->Dry.Buffer != device->FOAOut.Buffer)
bformatdec_upSample(device->AmbiDecoder,
- device->Dry.Buffer, device->FOAOut.Buffer,
+ device->Dry.Buffer, SAFE_CONST(ALfloatBUFFERSIZE*,device->FOAOut.Buffer),
device->FOAOut.NumChannels, SamplesToDo
);
bformatdec_process(device->AmbiDecoder,
device->RealOut.Buffer, device->RealOut.NumChannels,
- device->Dry.Buffer, SamplesToDo
+ SAFE_CONST(ALfloatBUFFERSIZE*,device->Dry.Buffer), SamplesToDo
);
}
else if(device->AmbiUp)
{
ambiup_process(device->AmbiUp,
device->RealOut.Buffer, device->RealOut.NumChannels,
- device->FOAOut.Buffer, SamplesToDo
+ SAFE_CONST(ALfloatBUFFERSIZE*,device->FOAOut.Buffer), SamplesToDo
);
}
else if(device->Uhj_Encoder)
diff --git a/Alc/bformatdec.c b/Alc/bformatdec.c
index 0722c061..2aab4ed8 100644
--- a/Alc/bformatdec.c
+++ b/Alc/bformatdec.c
@@ -161,19 +161,6 @@ static const ALfloat Ambi3DDecoder[8][FB_Max][MAX_AMBI_COEFFS] = {
static ALfloat Ambi3DEncoder[8][MAX_AMBI_COEFFS];
-static inline RowMixerFunc SelectRowMixer(void)
-{
-#ifdef HAVE_SSE
- if((CPUCapFlags&CPU_CAP_SSE))
- return MixRow_SSE;
-#endif
-#ifdef HAVE_NEON
- if((CPUCapFlags&CPU_CAP_NEON))
- return MixRow_Neon;
-#endif
- return MixRow_C;
-}
-
static RowMixerFunc MixMatrixRow = MixRow_C;
@@ -495,7 +482,7 @@ void bformatdec_reset(BFormatDec *dec, const AmbDecConf *conf, ALuint chancount,
}
-void bformatdec_process(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint OutChannels, ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint SamplesToDo)
+void bformatdec_process(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint OutChannels, const ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint SamplesToDo)
{
ALuint chan, i;
@@ -512,10 +499,12 @@ void bformatdec_process(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BU
memset(dec->ChannelMix, 0, SamplesToDo*sizeof(ALfloat));
MixMatrixRow(dec->ChannelMix, dec->Matrix.Dual[chan][FB_HighFreq],
- dec->SamplesHF, dec->NumChannels, SamplesToDo
+ SAFE_CONST(ALfloatBUFFERSIZE*,dec->SamplesHF), dec->NumChannels, 0,
+ SamplesToDo
);
MixMatrixRow(dec->ChannelMix, dec->Matrix.Dual[chan][FB_LowFreq],
- dec->SamplesLF, dec->NumChannels, SamplesToDo
+ SAFE_CONST(ALfloatBUFFERSIZE*,dec->SamplesLF), dec->NumChannels, 0,
+ SamplesToDo
);
if(dec->Delay[chan].Length > 0)
@@ -553,7 +542,7 @@ void bformatdec_process(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BU
memset(dec->ChannelMix, 0, SamplesToDo*sizeof(ALfloat));
MixMatrixRow(dec->ChannelMix, dec->Matrix.Single[chan], InSamples,
- dec->NumChannels, SamplesToDo);
+ dec->NumChannels, 0, SamplesToDo);
if(dec->Delay[chan].Length > 0)
{
@@ -584,7 +573,7 @@ void bformatdec_process(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BU
}
-void bformatdec_upSample(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint InChannels, ALuint SamplesToDo)
+void bformatdec_upSample(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BUFFERSIZE], const ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint InChannels, ALuint SamplesToDo)
{
ALuint i, j;
@@ -608,7 +597,9 @@ void bformatdec_upSample(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[B
/* Now write each band to the output. */
for(j = 0;j < dec->NumChannels;j++)
MixMatrixRow(OutBuffer[j], dec->UpSampler.Gains[i][j],
- dec->Samples, FB_Max, SamplesToDo);
+ SAFE_CONST(ALfloatBUFFERSIZE*,dec->Samples), FB_Max, 0,
+ SamplesToDo
+ );
}
}
@@ -659,7 +650,7 @@ void ambiup_reset(struct AmbiUpsampler *ambiup, const ALCdevice *device)
}
}
-void ambiup_process(struct AmbiUpsampler *ambiup, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint OutChannels, ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint SamplesToDo)
+void ambiup_process(struct AmbiUpsampler *ambiup, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint OutChannels, const ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint SamplesToDo)
{
ALuint i, j;
@@ -672,6 +663,8 @@ void ambiup_process(struct AmbiUpsampler *ambiup, ALfloat (*restrict OutBuffer)[
for(j = 0;j < OutChannels;j++)
MixMatrixRow(OutBuffer[j], ambiup->Gains[i][j],
- ambiup->Samples, FB_Max, SamplesToDo);
+ SAFE_CONST(ALfloatBUFFERSIZE*,ambiup->Samples), FB_Max, 0,
+ SamplesToDo
+ );
}
}
diff --git a/Alc/bformatdec.h b/Alc/bformatdec.h
index 433603dd..e78d89a7 100644
--- a/Alc/bformatdec.h
+++ b/Alc/bformatdec.h
@@ -17,10 +17,10 @@ int bformatdec_getOrder(const struct BFormatDec *dec);
void bformatdec_reset(struct BFormatDec *dec, const struct AmbDecConf *conf, ALuint chancount, ALuint srate, const ALuint chanmap[MAX_OUTPUT_CHANNELS], int flags);
/* Decodes the ambisonic input to the given output channels. */
-void bformatdec_process(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint OutChannels, ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint SamplesToDo);
+void bformatdec_process(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint OutChannels, const ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint SamplesToDo);
/* Up-samples a first-order input to the decoder's configuration. */
-void bformatdec_upSample(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint InChannels, ALuint SamplesToDo);
+void bformatdec_upSample(struct BFormatDec *dec, ALfloat (*restrict OutBuffer)[BUFFERSIZE], const ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint InChannels, ALuint SamplesToDo);
/* Stand-alone first-order upsampler. Kept here because it shares some stuff
@@ -30,7 +30,7 @@ struct AmbiUpsampler *ambiup_alloc();
void ambiup_free(struct AmbiUpsampler *ambiup);
void ambiup_reset(struct AmbiUpsampler *ambiup, const ALCdevice *device);
-void ambiup_process(struct AmbiUpsampler *ambiup, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint OutChannels, ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint SamplesToDo);
+void ambiup_process(struct AmbiUpsampler *ambiup, ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint OutChannels, const ALfloat (*restrict InSamples)[BUFFERSIZE], ALuint SamplesToDo);
/* Band splitter. Splits a signal into two phase-matching frequency bands. */
diff --git a/Alc/mixer.c b/Alc/mixer.c
index 2736920e..864e2395 100644
--- a/Alc/mixer.c
+++ b/Alc/mixer.c
@@ -79,6 +79,19 @@ MixerFunc SelectMixer(void)
return Mix_C;
}
+RowMixerFunc SelectRowMixer(void)
+{
+#ifdef HAVE_SSE
+ if((CPUCapFlags&CPU_CAP_SSE))
+ return MixRow_SSE;
+#endif
+#ifdef HAVE_NEON
+ if((CPUCapFlags&CPU_CAP_NEON))
+ return MixRow_Neon;
+#endif
+ return MixRow_C;
+}
+
static inline HrtfMixerFunc SelectHrtfMixer(void)
{
#ifdef HAVE_SSE
diff --git a/Alc/mixer_c.c b/Alc/mixer_c.c
index c74729a8..3e726df5 100644
--- a/Alc/mixer_c.c
+++ b/Alc/mixer_c.c
@@ -208,7 +208,7 @@ void Mix_C(const ALfloat *data, ALuint OutChans, ALfloat (*restrict OutBuffer)[B
* transform. And as the matrices are more or less static once set up, no
* stepping is necessary.
*/
-void MixRow_C(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans, ALuint BufferSize)
+void MixRow_C(ALfloat *OutBuffer, const ALfloat *Gains, const ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans, ALuint InPos, ALuint BufferSize)
{
ALuint c, i;
@@ -219,6 +219,6 @@ void MixRow_C(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict data)
continue;
for(i = 0;i < BufferSize;i++)
- OutBuffer[i] += data[c][i] * gain;
+ OutBuffer[i] += data[c][InPos+i] * gain;
}
}
diff --git a/Alc/mixer_defs.h b/Alc/mixer_defs.h
index 1572ac36..05e6f964 100644
--- a/Alc/mixer_defs.h
+++ b/Alc/mixer_defs.h
@@ -31,8 +31,9 @@ void MixDirectHrtf_C(ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint lidx, ALu
ALuint BufferSize);
void Mix_C(const ALfloat *data, ALuint OutChans, ALfloat (*restrict OutBuffer)[BUFFERSIZE],
struct MixGains *Gains, ALuint Counter, ALuint OutPos, ALuint BufferSize);
-void MixRow_C(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict data)[BUFFERSIZE],
- ALuint InChans, ALuint BufferSize);
+void MixRow_C(ALfloat *OutBuffer, const ALfloat *Gains,
+ const ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans,
+ ALuint InPos, ALuint BufferSize);
/* SSE mixers */
void MixHrtf_SSE(ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint lidx, ALuint ridx,
@@ -45,8 +46,9 @@ void MixDirectHrtf_SSE(ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint lidx, A
ALuint BufferSize);
void Mix_SSE(const ALfloat *data, ALuint OutChans, ALfloat (*restrict OutBuffer)[BUFFERSIZE],
struct MixGains *Gains, ALuint Counter, ALuint OutPos, ALuint BufferSize);
-void MixRow_SSE(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict data)[BUFFERSIZE],
- ALuint InChans, ALuint BufferSize);
+void MixRow_SSE(ALfloat *OutBuffer, const ALfloat *Gains,
+ const ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans,
+ ALuint InPos, ALuint BufferSize);
/* SSE resamplers */
inline void InitiatePositionArrays(ALuint frac, ALuint increment, ALuint *frac_arr, ALuint *pos_arr, ALuint size)
@@ -92,7 +94,8 @@ void MixDirectHrtf_Neon(ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint lidx,
ALuint BufferSize);
void Mix_Neon(const ALfloat *data, ALuint OutChans, ALfloat (*restrict OutBuffer)[BUFFERSIZE],
struct MixGains *Gains, ALuint Counter, ALuint OutPos, ALuint BufferSize);
-void MixRow_Neon(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict data)[BUFFERSIZE],
- ALuint InChans, ALuint BufferSize);
+void MixRow_Neon(ALfloat *OutBuffer, const ALfloat *Gains,
+ const ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans,
+ ALuint InPos, ALuint BufferSize);
#endif /* MIXER_DEFS_H */
diff --git a/Alc/mixer_neon.c b/Alc/mixer_neon.c
index 73395aeb..8c96aef1 100644
--- a/Alc/mixer_neon.c
+++ b/Alc/mixer_neon.c
@@ -144,7 +144,7 @@ void Mix_Neon(const ALfloat *data, ALuint OutChans, ALfloat (*restrict OutBuffer
}
}
-void MixRow_Neon(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans, ALuint BufferSize)
+void MixRow_Neon(ALfloat *OutBuffer, const ALfloat *Gains, const ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans, ALuint InPos, ALuint BufferSize)
{
float32x4_t gain4;
ALuint c;
@@ -159,12 +159,12 @@ void MixRow_Neon(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict da
gain4 = vdupq_n_f32(gain);
for(;BufferSize-pos > 3;pos += 4)
{
- const float32x4_t val4 = vld1q_f32(&data[c][pos]);
+ const float32x4_t val4 = vld1q_f32(&data[c][InPos+pos]);
float32x4_t dry4 = vld1q_f32(&OutBuffer[pos]);
dry4 = vmlaq_f32(dry4, val4, gain4);
vst1q_f32(&OutBuffer[pos], dry4);
}
for(;pos < BufferSize;pos++)
- OutBuffer[pos] += data[c][pos]*gain;
+ OutBuffer[pos] += data[c][InPos+pos]*gain;
}
}
diff --git a/Alc/mixer_sse.c b/Alc/mixer_sse.c
index 1ac78f9c..24e0e545 100644
--- a/Alc/mixer_sse.c
+++ b/Alc/mixer_sse.c
@@ -262,7 +262,7 @@ void Mix_SSE(const ALfloat *data, ALuint OutChans, ALfloat (*restrict OutBuffer)
}
}
-void MixRow_SSE(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans, ALuint BufferSize)
+void MixRow_SSE(ALfloat *OutBuffer, const ALfloat *Gains, const ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans, ALuint InPos, ALuint BufferSize)
{
__m128 gain4;
ALuint c;
@@ -277,12 +277,12 @@ void MixRow_SSE(ALfloat *OutBuffer, const ALfloat *Gains, ALfloat (*restrict dat
gain4 = _mm_set1_ps(gain);
for(;BufferSize-pos > 3;pos += 4)
{
- const __m128 val4 = _mm_load_ps(&data[c][pos]);
+ const __m128 val4 = _mm_load_ps(&data[c][InPos+pos]);
__m128 dry4 = _mm_load_ps(&OutBuffer[pos]);
dry4 = _mm_add_ps(dry4, _mm_mul_ps(val4, gain4));
_mm_store_ps(&OutBuffer[pos], dry4);
}
for(;pos < BufferSize;pos++)
- OutBuffer[pos] += data[c][pos]*gain;
+ OutBuffer[pos] += data[c][InPos+pos]*gain;
}
}
diff --git a/OpenAL32/Include/alMain.h b/OpenAL32/Include/alMain.h
index af599227..837e1082 100644
--- a/OpenAL32/Include/alMain.h
+++ b/OpenAL32/Include/alMain.h
@@ -128,6 +128,48 @@ AL_API ALboolean AL_APIENTRY alIsBufferFormatSupportedSOFT(ALenum format);
#endif
+#ifdef __GNUC__
+/* Because of a long-standing deficiency in C, you're not allowed to implicitly
+ * cast a pointer-to-type-array to a pointer-to-const-type-array. For example,
+ *
+ * int (*ptr)[10];
+ * const int (*cptr)[10] = ptr;
+ *
+ * is not allowed and most compilers will generate noisy warnings about
+ * incompatible types, even though it just makes the array elements const.
+ * Clang will allow it if you make the array type a typedef, like this:
+ *
+ * typedef int int10[10];
+ * int10 *ptr;
+ * const int10 *cptr = ptr;
+ *
+ * however GCC does not and still issues the incompatible type warning. The
+ * "proper" way to fix it is to add an explicit cast for the constified type,
+ * but that removes the vast majority of otherwise useful type-checking you'd
+ * get, and runs the risk of improper casts if types are later changed. Leaving
+ * it non-const can also be an issue if you use it as a function parameter, and
+ * happen to have a const type as input (and also reduce the capabilities of
+ * the compiler to better optimize the function).
+ *
+ * So to work around the problem, we use a macro. The macro first assigns the
+ * incoming variable to the specified non-const type to ensure it's the correct
+ * type, then casts the variable as the desired constified type. Very ugly, but
+ * I'd rather not have hundreds of lines of warnings because I want to tell the
+ * compiler that some array(s) can't be changed by the code, or have lots of
+ * error-prone casts.
+ */
+#define SAFE_CONST(T, var) __extension__({ \
+ T _tmp = (var); \
+ (const T)_tmp; \
+})
+#else
+/* Non-GNU-compatible compilers have to use a straight cast with no extra
+ * checks, due to the lack of multi-statement expressions.
+ */
+#define SAFE_CONST(T, var) ((const T)(var))
+#endif
+
+
typedef ALint64SOFT ALint64;
typedef ALuint64SOFT ALuint64;
diff --git a/OpenAL32/Include/alu.h b/OpenAL32/Include/alu.h
index 29cb00fb..3ced9628 100644
--- a/OpenAL32/Include/alu.h
+++ b/OpenAL32/Include/alu.h
@@ -162,8 +162,8 @@ typedef void (*MixerFunc)(const ALfloat *data, ALuint OutChans,
ALfloat (*restrict OutBuffer)[BUFFERSIZE], struct MixGains *Gains,
ALuint Counter, ALuint OutPos, ALuint BufferSize);
typedef void (*RowMixerFunc)(ALfloat *OutBuffer, const ALfloat *gains,
- ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans,
- ALuint BufferSize);
+ const ALfloat (*restrict data)[BUFFERSIZE], ALuint InChans,
+ ALuint InPos, ALuint BufferSize);
typedef void (*HrtfMixerFunc)(ALfloat (*restrict OutBuffer)[BUFFERSIZE], ALuint lidx, ALuint ridx,
const ALfloat *data, ALuint Counter, ALuint Offset, ALuint OutPos,
const ALuint IrSize, const MixHrtfParams *hrtfparams,
@@ -265,6 +265,7 @@ enum HrtfRequestMode {
void aluInitMixer(void);
MixerFunc SelectMixer(void);
+RowMixerFunc SelectRowMixer(void);
/* aluInitRenderer
*
generated by cgit v1.2.3 (git 2.39.1) at 2024-11-29 23:53:01 +0000