mirror of
https://gitlab.freedesktop.org/pipewire/pipewire.git
synced 2025-10-29 05:40:27 -04:00
audioconvert: simplify 24 bits handling
Make a new uint42_t and int24_t type and use that to handle 24 bits samples. This makes it easier because we can iterate and copy the structs like other types.
This commit is contained in:
Wim Taymans
parent
e395f62425
commit
817d5bd7a4
6 changed files with 204 additions and 247 deletions
|
|
@ -147,7 +147,7 @@ void
|
|||
conv_s24_to_f32d_1s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
|
||||
uint32_t n_channels, uint32_t n_samples)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
const int24_t *s = src;
|
||||
float *d0 = dst[0];
|
||||
uint32_t n, unrolled;
|
||||
__m128i in;
|
||||
|
|
@ -164,21 +164,21 @@ conv_s24_to_f32d_1s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
for(n = 0; n < unrolled; n += 4) {
|
||||
in = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[0 * n_channels]),
|
||||
*((uint32_t*)&s[3 * n_channels]),
|
||||
*((uint32_t*)&s[6 * n_channels]),
|
||||
*((uint32_t*)&s[9 * n_channels]));
|
||||
*((uint32_t*)&s[1 * n_channels]),
|
||||
*((uint32_t*)&s[2 * n_channels]),
|
||||
*((uint32_t*)&s[3 * n_channels]));
|
||||
in = _mm_slli_epi32(in, 8);
|
||||
in = _mm_srai_epi32(in, 8);
|
||||
out = _mm_cvtepi32_ps(in);
|
||||
out = _mm_mul_ps(out, factor);
|
||||
_mm_store_ps(&d0[n], out);
|
||||
s += 12 * n_channels;
|
||||
s += 4 * n_channels;
|
||||
}
|
||||
for(; n < n_samples; n++) {
|
||||
out = _mm_cvtsi32_ss(factor, read_s24(s));
|
||||
out = _mm_cvtsi32_ss(factor, s24_to_s32(*s));
|
||||
out = _mm_mul_ss(out, factor);
|
||||
_mm_store_ss(&d0[n], out);
|
||||
s += 3 * n_channels;
|
||||
s += n_channels;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -186,7 +186,7 @@ static void
|
|||
conv_s24_to_f32d_2s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
|
||||
uint32_t n_channels, uint32_t n_samples)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
const int24_t *s = src;
|
||||
float *d0 = dst[0], *d1 = dst[1];
|
||||
uint32_t n, unrolled;
|
||||
__m128i in[2];
|
||||
|
|
@ -205,14 +205,14 @@ conv_s24_to_f32d_2s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
for(n = 0; n < unrolled; n += 4) {
|
||||
in[0] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[0 + 0*n_channels]),
|
||||
*((uint32_t*)&s[0 + 3*n_channels]),
|
||||
*((uint32_t*)&s[0 + 6*n_channels]),
|
||||
*((uint32_t*)&s[0 + 9*n_channels]));
|
||||
*((uint32_t*)&s[0 + 1*n_channels]),
|
||||
*((uint32_t*)&s[0 + 2*n_channels]),
|
||||
*((uint32_t*)&s[0 + 3*n_channels]));
|
||||
in[1] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[3 + 0*n_channels]),
|
||||
*((uint32_t*)&s[3 + 3*n_channels]),
|
||||
*((uint32_t*)&s[3 + 6*n_channels]),
|
||||
*((uint32_t*)&s[3 + 9*n_channels]));
|
||||
*((uint32_t*)&s[1 + 0*n_channels]),
|
||||
*((uint32_t*)&s[1 + 1*n_channels]),
|
||||
*((uint32_t*)&s[1 + 2*n_channels]),
|
||||
*((uint32_t*)&s[1 + 3*n_channels]));
|
||||
|
||||
in[0] = _mm_slli_epi32(in[0], 8);
|
||||
in[1] = _mm_slli_epi32(in[1], 8);
|
||||
|
|
@ -229,23 +229,23 @@ conv_s24_to_f32d_2s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
_mm_store_ps(&d0[n], out[0]);
|
||||
_mm_store_ps(&d1[n], out[1]);
|
||||
|
||||
s += 12 * n_channels;
|
||||
s += 4 * n_channels;
|
||||
}
|
||||
for(; n < n_samples; n++) {
|
||||
out[0] = _mm_cvtsi32_ss(factor, read_s24(s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, read_s24(s+3));
|
||||
out[0] = _mm_cvtsi32_ss(factor, s24_to_s32(*s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+1)));
|
||||
out[0] = _mm_mul_ss(out[0], factor);
|
||||
out[1] = _mm_mul_ss(out[1], factor);
|
||||
_mm_store_ss(&d0[n], out[0]);
|
||||
_mm_store_ss(&d1[n], out[1]);
|
||||
s += 3 * n_channels;
|
||||
s += n_channels;
|
||||
}
|
||||
}
|
||||
static void
|
||||
conv_s24_to_f32d_4s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
|
||||
uint32_t n_channels, uint32_t n_samples)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
const int24_t *s = src;
|
||||
float *d0 = dst[0], *d1 = dst[1], *d2 = dst[2], *d3 = dst[3];
|
||||
uint32_t n, unrolled;
|
||||
__m128i in[4];
|
||||
|
|
@ -266,24 +266,24 @@ conv_s24_to_f32d_4s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
for(n = 0; n < unrolled; n += 4) {
|
||||
in[0] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[0 + 0*n_channels]),
|
||||
*((uint32_t*)&s[0 + 3*n_channels]),
|
||||
*((uint32_t*)&s[0 + 6*n_channels]),
|
||||
*((uint32_t*)&s[0 + 9*n_channels]));
|
||||
*((uint32_t*)&s[0 + 1*n_channels]),
|
||||
*((uint32_t*)&s[0 + 2*n_channels]),
|
||||
*((uint32_t*)&s[0 + 3*n_channels]));
|
||||
in[1] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[3 + 0*n_channels]),
|
||||
*((uint32_t*)&s[3 + 3*n_channels]),
|
||||
*((uint32_t*)&s[3 + 6*n_channels]),
|
||||
*((uint32_t*)&s[3 + 9*n_channels]));
|
||||
*((uint32_t*)&s[1 + 0*n_channels]),
|
||||
*((uint32_t*)&s[1 + 1*n_channels]),
|
||||
*((uint32_t*)&s[1 + 2*n_channels]),
|
||||
*((uint32_t*)&s[1 + 3*n_channels]));
|
||||
in[2] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[6 + 0*n_channels]),
|
||||
*((uint32_t*)&s[6 + 3*n_channels]),
|
||||
*((uint32_t*)&s[6 + 6*n_channels]),
|
||||
*((uint32_t*)&s[6 + 9*n_channels]));
|
||||
*((uint32_t*)&s[2 + 0*n_channels]),
|
||||
*((uint32_t*)&s[2 + 1*n_channels]),
|
||||
*((uint32_t*)&s[2 + 2*n_channels]),
|
||||
*((uint32_t*)&s[2 + 3*n_channels]));
|
||||
in[3] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[9 + 0*n_channels]),
|
||||
*((uint32_t*)&s[9 + 3*n_channels]),
|
||||
*((uint32_t*)&s[9 + 6*n_channels]),
|
||||
*((uint32_t*)&s[9 + 9*n_channels]));
|
||||
*((uint32_t*)&s[3 + 0*n_channels]),
|
||||
*((uint32_t*)&s[3 + 1*n_channels]),
|
||||
*((uint32_t*)&s[3 + 2*n_channels]),
|
||||
*((uint32_t*)&s[3 + 3*n_channels]));
|
||||
|
||||
in[0] = _mm_slli_epi32(in[0], 8);
|
||||
in[1] = _mm_slli_epi32(in[1], 8);
|
||||
|
|
@ -310,13 +310,13 @@ conv_s24_to_f32d_4s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
_mm_store_ps(&d2[n], out[2]);
|
||||
_mm_store_ps(&d3[n], out[3]);
|
||||
|
||||
s += 12 * n_channels;
|
||||
s += 4 * n_channels;
|
||||
}
|
||||
for(; n < n_samples; n++) {
|
||||
out[0] = _mm_cvtsi32_ss(factor, read_s24(s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, read_s24(s+3));
|
||||
out[2] = _mm_cvtsi32_ss(factor, read_s24(s+6));
|
||||
out[3] = _mm_cvtsi32_ss(factor, read_s24(s+9));
|
||||
out[0] = _mm_cvtsi32_ss(factor, s24_to_s32(*s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+1)));
|
||||
out[2] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+2)));
|
||||
out[3] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+3)));
|
||||
out[0] = _mm_mul_ss(out[0], factor);
|
||||
out[1] = _mm_mul_ss(out[1], factor);
|
||||
out[2] = _mm_mul_ss(out[2], factor);
|
||||
|
|
@ -325,7 +325,7 @@ conv_s24_to_f32d_4s_avx2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
_mm_store_ss(&d1[n], out[1]);
|
||||
_mm_store_ss(&d2[n], out[2]);
|
||||
_mm_store_ss(&d3[n], out[3]);
|
||||
s += 3 * n_channels;
|
||||
s += n_channels;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -55,7 +55,7 @@ MAKE_COPY(24);
|
|||
MAKE_COPY(32);
|
||||
MAKE_COPY(64);
|
||||
|
||||
#define MAKE_D_TO_D_F(sname,stype,dname,dtype,func) \
|
||||
#define MAKE_D_TO_D(sname,stype,dname,dtype,func) \
|
||||
void conv_ ##sname## d_to_ ##dname## d_c(struct convert *conv, \
|
||||
void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[], \
|
||||
uint32_t n_samples) \
|
||||
|
|
@ -64,15 +64,12 @@ void conv_ ##sname## d_to_ ##dname## d_c(struct convert *conv, \
|
|||
for (i = 0; i < n_channels; i++) { \
|
||||
const stype *s = src[i]; \
|
||||
dtype *d = dst[i]; \
|
||||
for (j = 0; j < n_samples; j++) { \
|
||||
func; \
|
||||
} \
|
||||
for (j = 0; j < n_samples; j++) \
|
||||
d[j] = func (s[j]); \
|
||||
} \
|
||||
}
|
||||
#define MAKE_D_TO_D(sname,stype,dname,dtype,func) \
|
||||
MAKE_D_TO_D_F(sname,stype,dname,dtype, d[j] = func (s[j])) \
|
||||
|
||||
#define MAKE_I_TO_I_F(sname,stype,dname,dtype,func) \
|
||||
#define MAKE_I_TO_I(sname,stype,dname,dtype,func) \
|
||||
void conv_ ##sname## _to_ ##dname## _c(struct convert *conv, \
|
||||
void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[], \
|
||||
uint32_t n_samples) \
|
||||
|
|
@ -81,14 +78,11 @@ void conv_ ##sname## _to_ ##dname## _c(struct convert *conv, \
|
|||
const stype *s = src[0]; \
|
||||
dtype *d = dst[0]; \
|
||||
n_samples *= conv->n_channels; \
|
||||
for (j = 0; j < n_samples; j++) { \
|
||||
func; \
|
||||
} \
|
||||
for (j = 0; j < n_samples; j++) \
|
||||
d[j] = func (s[j]); \
|
||||
}
|
||||
#define MAKE_I_TO_I(sname,stype,dname,dtype,func) \
|
||||
MAKE_I_TO_I_F(sname,stype,dname,dtype, d[j] = func (s[j])) \
|
||||
|
||||
#define MAKE_I_TO_D_F(sname,stype,dname,dtype,func) \
|
||||
#define MAKE_I_TO_D(sname,stype,dname,dtype,func) \
|
||||
void conv_ ##sname## _to_ ##dname## d_c(struct convert *conv, \
|
||||
void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[], \
|
||||
uint32_t n_samples) \
|
||||
|
|
@ -97,15 +91,12 @@ void conv_ ##sname## _to_ ##dname## d_c(struct convert *conv, \
|
|||
dtype **d = (dtype**)dst; \
|
||||
uint32_t i, j, n_channels = conv->n_channels; \
|
||||
for (j = 0; j < n_samples; j++) { \
|
||||
for (i = 0; i < n_channels; i++) { \
|
||||
func; \
|
||||
} \
|
||||
for (i = 0; i < n_channels; i++) \
|
||||
d[i][j] = func (*s++); \
|
||||
} \
|
||||
}
|
||||
#define MAKE_I_TO_D(sname,stype,dname,dtype,func) \
|
||||
MAKE_I_TO_D_F(sname,stype,dname,dtype, d[i][j] = func (*s++)) \
|
||||
|
||||
#define MAKE_D_TO_I_F(sname,stype,dname,dtype,func) \
|
||||
#define MAKE_D_TO_I(sname,stype,dname,dtype,func) \
|
||||
void conv_ ##sname## d_to_ ##dname## _c(struct convert *conv, \
|
||||
void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[], \
|
||||
uint32_t n_samples) \
|
||||
|
|
@ -114,13 +105,10 @@ void conv_ ##sname## d_to_ ##dname## _c(struct convert *conv, \
|
|||
dtype *d = dst[0]; \
|
||||
uint32_t i, j, n_channels = conv->n_channels; \
|
||||
for (j = 0; j < n_samples; j++) { \
|
||||
for (i = 0; i < n_channels; i++) { \
|
||||
func; \
|
||||
} \
|
||||
for (i = 0; i < n_channels; i++) \
|
||||
*d++ = func (s[i][j]); \
|
||||
} \
|
||||
}
|
||||
#define MAKE_D_TO_I(sname,stype,dname,dtype,func) \
|
||||
MAKE_D_TO_I_F(sname,stype,dname,dtype, *d++ = func (s[i][j])) \
|
||||
|
||||
/* to f32 */
|
||||
MAKE_D_TO_D(u8, uint8_t, f32, float, U8_TO_F32);
|
||||
|
|
@ -154,14 +142,14 @@ MAKE_I_TO_D(s32, int32_t, f32, float, S32_TO_F32);
|
|||
MAKE_D_TO_I(s32, int32_t, f32, float, S32_TO_F32);
|
||||
MAKE_I_TO_D(s32s, uint32_t, f32, float, S32S_TO_F32);
|
||||
|
||||
MAKE_I_TO_I_F(u24, uint8_t, f32, float, d[j] = U24_TO_F32(read_u24(s)); s += 3);
|
||||
MAKE_I_TO_D_F(u24, uint8_t, f32, float, d[i][j] = U24_TO_F32(read_u24(s)); s += 3);
|
||||
MAKE_I_TO_I(u24, uint24_t, f32, float, U24_TO_F32);
|
||||
MAKE_I_TO_D(u24, uint24_t, f32, float, U24_TO_F32);
|
||||
|
||||
MAKE_D_TO_D_F(s24, int8_t, f32, float, d[j] = S24_TO_F32(read_s24(s)); s += 3);
|
||||
MAKE_I_TO_I_F(s24, int8_t, f32, float, d[j] = S24_TO_F32(read_s24(s)); s += 3);
|
||||
MAKE_I_TO_D_F(s24, int8_t, f32, float, d[i][j] = S24_TO_F32(read_s24(s)); s += 3);
|
||||
MAKE_D_TO_I_F(s24, int8_t, f32, float, *d++ = S24_TO_F32(read_s24(&s[i][j*3])));
|
||||
MAKE_I_TO_D_F(s24s, int8_t, f32, float, d[i][j] = S24_TO_F32(read_s24s(s)); s += 3);
|
||||
MAKE_D_TO_D(s24, int24_t, f32, float, S24_TO_F32);
|
||||
MAKE_I_TO_I(s24, int24_t, f32, float, S24_TO_F32);
|
||||
MAKE_I_TO_D(s24, int24_t, f32, float, S24_TO_F32);
|
||||
MAKE_D_TO_I(s24, int24_t, f32, float, S24_TO_F32);
|
||||
MAKE_I_TO_D(s24s, int24_t, f32, float, S24S_TO_F32);
|
||||
|
||||
MAKE_I_TO_I(u24_32, uint32_t, f32, float, U24_32_TO_F32);
|
||||
MAKE_I_TO_D(u24_32, uint32_t, f32, float, U24_32_TO_F32);
|
||||
|
|
@ -211,14 +199,14 @@ MAKE_I_TO_D(f32, float, s32, int32_t, F32_TO_S32);
|
|||
MAKE_D_TO_I(f32, float, s32, int32_t, F32_TO_S32);
|
||||
MAKE_D_TO_I(f32, float, s32s, uint32_t, F32_TO_S32S);
|
||||
|
||||
MAKE_I_TO_I_F(f32, float, u24, uint8_t, write_u24(d, F32_TO_U24(s[j])); d += 3);
|
||||
MAKE_D_TO_I_F(f32, float, u24, uint8_t, write_u24(d, F32_TO_U24(s[i][j])); d += 3);
|
||||
MAKE_I_TO_I(f32, float, u24, uint24_t, F32_TO_U24);
|
||||
MAKE_D_TO_I(f32, float, u24, uint24_t, F32_TO_U24);
|
||||
|
||||
MAKE_D_TO_D_F(f32, float, s24, uint8_t, write_s24(d, F32_TO_S24(s[j])); d += 3);
|
||||
MAKE_I_TO_I_F(f32, float, s24, uint8_t, write_s24(d, F32_TO_S24(s[j])); d += 3);
|
||||
MAKE_I_TO_D_F(f32, float, s24, uint8_t, write_s24(&d[i][j*3], F32_TO_S24(*s++)));
|
||||
MAKE_D_TO_I_F(f32, float, s24, uint8_t, write_s24(d, F32_TO_S24(s[i][j])); d += 3);
|
||||
MAKE_D_TO_I_F(f32, float, s24s, uint8_t, write_s24s(d, F32_TO_S24(s[i][j])); d += 3);
|
||||
MAKE_D_TO_D(f32, float, s24, int24_t, F32_TO_S24);
|
||||
MAKE_I_TO_I(f32, float, s24, int24_t, F32_TO_S24);
|
||||
MAKE_I_TO_D(f32, float, s24, int24_t, F32_TO_S24);
|
||||
MAKE_D_TO_I(f32, float, s24, int24_t, F32_TO_S24);
|
||||
MAKE_D_TO_I(f32, float, s24s, int24_t, F32_TO_S24S);
|
||||
|
||||
MAKE_I_TO_I(f32, float, u24_32, uint32_t, F32_TO_U24_32);
|
||||
MAKE_D_TO_I(f32, float, u24_32, uint32_t, F32_TO_U24_32);
|
||||
|
|
@ -253,7 +241,7 @@ static inline void update_dither_c(struct convert *conv, uint32_t n_samples)
|
|||
dither[n] = lcnoise(state) * scale;
|
||||
}
|
||||
|
||||
#define MAKE_D_dither_F(dname,dtype,func) \
|
||||
#define MAKE_D_dither(dname,dtype,func) \
|
||||
void conv_f32d_to_ ##dname## d_dither_c(struct convert *conv, \
|
||||
void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[], \
|
||||
uint32_t n_samples) \
|
||||
|
|
@ -266,16 +254,13 @@ void conv_f32d_to_ ##dname## d_dither_c(struct convert *conv, \
|
|||
dtype *d = dst[i]; \
|
||||
for (j = 0; j < n_samples;) { \
|
||||
chunk = SPA_MIN(n_samples - j, dither_size); \
|
||||
for (k = 0; k < chunk; k++, j++) { \
|
||||
func; \
|
||||
} \
|
||||
for (k = 0; k < chunk; k++, j++) \
|
||||
d[j] = func (s[j], dither[k]); \
|
||||
} \
|
||||
} \
|
||||
}
|
||||
#define MAKE_D_dither(dname,dtype,func) \
|
||||
MAKE_D_dither_F(dname,dtype, d[j] = func (s[j], dither[k])) \
|
||||
|
||||
#define MAKE_I_dither_F(dname,dtype,func) \
|
||||
#define MAKE_I_dither(dname,dtype,func) \
|
||||
void conv_f32d_to_ ##dname## _dither_c(struct convert *conv, \
|
||||
void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[], \
|
||||
uint32_t n_samples) \
|
||||
|
|
@ -288,14 +273,11 @@ void conv_f32d_to_ ##dname## _dither_c(struct convert *conv, \
|
|||
for (j = 0; j < n_samples;) { \
|
||||
chunk = SPA_MIN(n_samples - j, dither_size); \
|
||||
for (k = 0; k < chunk; k++, j++) { \
|
||||
for (i = 0; i < n_channels; i++) { \
|
||||
func; \
|
||||
} \
|
||||
for (i = 0; i < n_channels; i++) \
|
||||
*d++ = func (s[i][j], dither[k]); \
|
||||
} \
|
||||
} \
|
||||
}
|
||||
#define MAKE_I_dither(dname,dtype,func) \
|
||||
MAKE_I_dither_F(dname,dtype, *d++ = func (s[i][j], dither[k])) \
|
||||
|
||||
MAKE_D_dither(u8, uint8_t, F32_TO_U8_D);
|
||||
MAKE_I_dither(u8, uint8_t, F32_TO_U8_D);
|
||||
|
|
@ -307,9 +289,9 @@ MAKE_I_dither(s16s, uint16_t, F32_TO_S16S_D);
|
|||
MAKE_D_dither(s32, int32_t, F32_TO_S32_D);
|
||||
MAKE_I_dither(s32, int32_t, F32_TO_S32_D);
|
||||
MAKE_I_dither(s32s, uint32_t, F32_TO_S32S_D);
|
||||
MAKE_D_dither_F(s24, uint8_t, write_s24(d, F32_TO_S24_D(s[j], dither[k])); d += 3);
|
||||
MAKE_I_dither_F(s24, uint8_t, write_s24(d, F32_TO_S24_D(s[i][j], dither[k])); d += 3);
|
||||
MAKE_I_dither_F(s24s, uint8_t, write_s24s(d, F32_TO_S24_D(s[i][j], dither[k])); d += 3);
|
||||
MAKE_D_dither(s24, int24_t, F32_TO_S24_D);
|
||||
MAKE_I_dither(s24, int24_t, F32_TO_S24_D);
|
||||
MAKE_I_dither(s24s, int24_t, F32_TO_S24_D);
|
||||
MAKE_D_dither(s24_32, int32_t, F32_TO_S24_32_D);
|
||||
MAKE_I_dither(s24_32, int32_t, F32_TO_S24_32_D);
|
||||
MAKE_I_dither(s24_32s, int32_t, F32_TO_S24_32S_D);
|
||||
|
|
@ -335,7 +317,7 @@ MAKE_I_dither(s24_32s, int32_t, F32_TO_S24_32S_D);
|
|||
#define F32_TO_S16_SH(s,sh,d) SHAPER5(int16_t, s, S16_SCALE, 0, sh, S16_MIN, S16_MAX, d)
|
||||
#define F32_TO_S16S_SH(s,sh,d) bswap_16(F32_TO_S16_SH(s,sh,d))
|
||||
|
||||
#define MAKE_D_shaped_F(dname,dtype,func) \
|
||||
#define MAKE_D_shaped(dname,dtype,func) \
|
||||
void conv_f32d_to_ ##dname## d_shaped_c(struct convert *conv, \
|
||||
void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[], \
|
||||
uint32_t n_samples) \
|
||||
|
|
@ -350,17 +332,14 @@ void conv_f32d_to_ ##dname## d_shaped_c(struct convert *conv, \
|
|||
uint32_t idx = sh->idx; \
|
||||
for (j = 0; j < n_samples;) { \
|
||||
chunk = SPA_MIN(n_samples - j, dither_size); \
|
||||
for (k = 0; k < chunk; k++, j++) { \
|
||||
func; \
|
||||
} \
|
||||
for (k = 0; k < chunk; k++, j++) \
|
||||
d[j] = func (s[j], sh, dither[k]); \
|
||||
} \
|
||||
sh->idx = idx; \
|
||||
} \
|
||||
}
|
||||
#define MAKE_D_shaped(dname,dtype,func) \
|
||||
MAKE_D_shaped_F(dname,dtype, d[j] = func (s[j], sh, dither[k])) \
|
||||
|
||||
#define MAKE_I_shaped_F(dname,dtype,func) \
|
||||
#define MAKE_I_shaped(dname,dtype,func) \
|
||||
void conv_f32d_to_ ##dname## _shaped_c(struct convert *conv, \
|
||||
void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[], \
|
||||
uint32_t n_samples) \
|
||||
|
|
@ -376,15 +355,12 @@ void conv_f32d_to_ ##dname## _shaped_c(struct convert *conv, \
|
|||
uint32_t idx = sh->idx; \
|
||||
for (j = 0; j < n_samples;) { \
|
||||
chunk = SPA_MIN(n_samples - j, dither_size); \
|
||||
for (k = 0; k < chunk; k++, j++) { \
|
||||
func; \
|
||||
} \
|
||||
for (k = 0; k < chunk; k++, j++) \
|
||||
d[j*n_channels] = func (s[j], sh, dither[k]); \
|
||||
} \
|
||||
sh->idx = idx; \
|
||||
} \
|
||||
}
|
||||
#define MAKE_I_shaped(dname,dtype,func) \
|
||||
MAKE_I_shaped_F(dname,dtype, d[j*n_channels] = func (s[j], sh, dither[k])) \
|
||||
|
||||
MAKE_D_shaped(u8, uint8_t, F32_TO_U8_SH);
|
||||
MAKE_I_shaped(u8, uint8_t, F32_TO_U8_SH);
|
||||
|
|
@ -395,23 +371,21 @@ MAKE_I_shaped(s16, int16_t, F32_TO_S16_SH);
|
|||
MAKE_I_shaped(s16s, uint16_t, F32_TO_S16S_SH);
|
||||
|
||||
#define MAKE_DEINTERLEAVE(size,type,func) \
|
||||
MAKE_I_TO_D_F(size,type,size,type,func)
|
||||
#define DEINTERLEAVE_COPY (d[i][j] = *s++)
|
||||
MAKE_I_TO_D(size,type,size,type,func)
|
||||
|
||||
MAKE_DEINTERLEAVE(8, uint8_t, DEINTERLEAVE_COPY);
|
||||
MAKE_DEINTERLEAVE(16, uint16_t, DEINTERLEAVE_COPY);
|
||||
MAKE_DEINTERLEAVE(24, uint8_t, write_s24(&d[i][j*3], read_s24(s)); s+=3);
|
||||
MAKE_DEINTERLEAVE(32, uint32_t, DEINTERLEAVE_COPY);
|
||||
MAKE_DEINTERLEAVE(32s, uint32_t, d[i][j] = bswap_32(*s++));
|
||||
MAKE_DEINTERLEAVE(64, uint64_t, DEINTERLEAVE_COPY);
|
||||
MAKE_DEINTERLEAVE(8, uint8_t, (uint8_t));
|
||||
MAKE_DEINTERLEAVE(16, uint16_t, (uint16_t));
|
||||
MAKE_DEINTERLEAVE(24, uint24_t, (uint24_t));
|
||||
MAKE_DEINTERLEAVE(32, uint32_t, (uint32_t));
|
||||
MAKE_DEINTERLEAVE(32s, uint32_t, bswap_32);
|
||||
MAKE_DEINTERLEAVE(64, uint64_t, (uint64_t));
|
||||
|
||||
#define MAKE_INTERLEAVE(size,type,func) \
|
||||
MAKE_D_TO_I_F(size,type,size,type,func)
|
||||
#define INTERLEAVE_COPY (*d++ = s[i][j])
|
||||
MAKE_D_TO_I(size,type,size,type,func)
|
||||
|
||||
MAKE_INTERLEAVE(8, uint8_t, INTERLEAVE_COPY);
|
||||
MAKE_INTERLEAVE(16, uint16_t, INTERLEAVE_COPY);
|
||||
MAKE_INTERLEAVE(24, uint8_t, write_s24(d, read_s24(&s[i][j*3])); d+=3);
|
||||
MAKE_INTERLEAVE(32, uint32_t, INTERLEAVE_COPY);
|
||||
MAKE_INTERLEAVE(32s, uint32_t, *d++ = bswap_32(s[i][j]));
|
||||
MAKE_INTERLEAVE(64, uint64_t, INTERLEAVE_COPY);
|
||||
MAKE_INTERLEAVE(8, uint8_t, (uint8_t));
|
||||
MAKE_INTERLEAVE(16, uint16_t, (uint16_t));
|
||||
MAKE_INTERLEAVE(24, uint24_t, (uint24_t));
|
||||
MAKE_INTERLEAVE(32, uint32_t, (uint32_t));
|
||||
MAKE_INTERLEAVE(32s, uint32_t, bswap_32);
|
||||
MAKE_INTERLEAVE(64, uint64_t, (uint64_t));
|
||||
|
|
|
|||
|
|
@ -132,7 +132,7 @@ void
|
|||
conv_s24_to_f32d_1s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
|
||||
uint32_t n_channels, uint32_t n_samples)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
const int24_t *s = src;
|
||||
float *d0 = dst[0];
|
||||
uint32_t n, unrolled;
|
||||
__m128i in;
|
||||
|
|
@ -149,21 +149,21 @@ conv_s24_to_f32d_1s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
for(n = 0; n < unrolled; n += 4) {
|
||||
in = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[0 * n_channels]),
|
||||
*((uint32_t*)&s[3 * n_channels]),
|
||||
*((uint32_t*)&s[6 * n_channels]),
|
||||
*((uint32_t*)&s[9 * n_channels]));
|
||||
*((uint32_t*)&s[1 * n_channels]),
|
||||
*((uint32_t*)&s[2 * n_channels]),
|
||||
*((uint32_t*)&s[3 * n_channels]));
|
||||
in = _mm_slli_epi32(in, 8);
|
||||
in = _mm_srai_epi32(in, 8);
|
||||
out = _mm_cvtepi32_ps(in);
|
||||
out = _mm_mul_ps(out, factor);
|
||||
_mm_store_ps(&d0[n], out);
|
||||
s += 12 * n_channels;
|
||||
s += 4 * n_channels;
|
||||
}
|
||||
for(; n < n_samples; n++) {
|
||||
out = _mm_cvtsi32_ss(factor, read_s24(s));
|
||||
out = _mm_cvtsi32_ss(factor, s24_to_s32(*s));
|
||||
out = _mm_mul_ss(out, factor);
|
||||
_mm_store_ss(&d0[n], out);
|
||||
s += 3 * n_channels;
|
||||
s += n_channels;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -171,7 +171,7 @@ static void
|
|||
conv_s24_to_f32d_2s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
|
||||
uint32_t n_channels, uint32_t n_samples)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
const int24_t *s = src;
|
||||
float *d0 = dst[0], *d1 = dst[1];
|
||||
uint32_t n, unrolled;
|
||||
__m128i in[2];
|
||||
|
|
@ -190,14 +190,14 @@ conv_s24_to_f32d_2s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
for(n = 0; n < unrolled; n += 4) {
|
||||
in[0] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[0 + 0*n_channels]),
|
||||
*((uint32_t*)&s[0 + 3*n_channels]),
|
||||
*((uint32_t*)&s[0 + 6*n_channels]),
|
||||
*((uint32_t*)&s[0 + 9*n_channels]));
|
||||
*((uint32_t*)&s[0 + 1*n_channels]),
|
||||
*((uint32_t*)&s[0 + 2*n_channels]),
|
||||
*((uint32_t*)&s[0 + 3*n_channels]));
|
||||
in[1] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[3 + 0*n_channels]),
|
||||
*((uint32_t*)&s[3 + 3*n_channels]),
|
||||
*((uint32_t*)&s[3 + 6*n_channels]),
|
||||
*((uint32_t*)&s[3 + 9*n_channels]));
|
||||
*((uint32_t*)&s[1 + 0*n_channels]),
|
||||
*((uint32_t*)&s[1 + 1*n_channels]),
|
||||
*((uint32_t*)&s[1 + 2*n_channels]),
|
||||
*((uint32_t*)&s[1 + 3*n_channels]));
|
||||
|
||||
in[0] = _mm_slli_epi32(in[0], 8);
|
||||
in[1] = _mm_slli_epi32(in[1], 8);
|
||||
|
|
@ -214,23 +214,23 @@ conv_s24_to_f32d_2s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
_mm_store_ps(&d0[n], out[0]);
|
||||
_mm_store_ps(&d1[n], out[1]);
|
||||
|
||||
s += 12 * n_channels;
|
||||
s += 4 * n_channels;
|
||||
}
|
||||
for(; n < n_samples; n++) {
|
||||
out[0] = _mm_cvtsi32_ss(factor, read_s24(s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, read_s24(s+3));
|
||||
out[0] = _mm_cvtsi32_ss(factor, s24_to_s32(*s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+1)));
|
||||
out[0] = _mm_mul_ss(out[0], factor);
|
||||
out[1] = _mm_mul_ss(out[1], factor);
|
||||
_mm_store_ss(&d0[n], out[0]);
|
||||
_mm_store_ss(&d1[n], out[1]);
|
||||
s += 3 * n_channels;
|
||||
s += n_channels;
|
||||
}
|
||||
}
|
||||
static void
|
||||
conv_s24_to_f32d_4s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
|
||||
uint32_t n_channels, uint32_t n_samples)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
const int24_t *s = src;
|
||||
float *d0 = dst[0], *d1 = dst[1], *d2 = dst[2], *d3 = dst[3];
|
||||
uint32_t n, unrolled;
|
||||
__m128i in[4];
|
||||
|
|
@ -251,24 +251,24 @@ conv_s24_to_f32d_4s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
for(n = 0; n < unrolled; n += 4) {
|
||||
in[0] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[0 + 0*n_channels]),
|
||||
*((uint32_t*)&s[0 + 3*n_channels]),
|
||||
*((uint32_t*)&s[0 + 6*n_channels]),
|
||||
*((uint32_t*)&s[0 + 9*n_channels]));
|
||||
*((uint32_t*)&s[0 + 1*n_channels]),
|
||||
*((uint32_t*)&s[0 + 2*n_channels]),
|
||||
*((uint32_t*)&s[0 + 3*n_channels]));
|
||||
in[1] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[3 + 0*n_channels]),
|
||||
*((uint32_t*)&s[3 + 3*n_channels]),
|
||||
*((uint32_t*)&s[3 + 6*n_channels]),
|
||||
*((uint32_t*)&s[3 + 9*n_channels]));
|
||||
*((uint32_t*)&s[1 + 0*n_channels]),
|
||||
*((uint32_t*)&s[1 + 1*n_channels]),
|
||||
*((uint32_t*)&s[1 + 2*n_channels]),
|
||||
*((uint32_t*)&s[1 + 3*n_channels]));
|
||||
in[2] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[6 + 0*n_channels]),
|
||||
*((uint32_t*)&s[6 + 3*n_channels]),
|
||||
*((uint32_t*)&s[6 + 6*n_channels]),
|
||||
*((uint32_t*)&s[6 + 9*n_channels]));
|
||||
*((uint32_t*)&s[2 + 0*n_channels]),
|
||||
*((uint32_t*)&s[2 + 1*n_channels]),
|
||||
*((uint32_t*)&s[2 + 2*n_channels]),
|
||||
*((uint32_t*)&s[2 + 3*n_channels]));
|
||||
in[3] = _mm_setr_epi32(
|
||||
*((uint32_t*)&s[9 + 0*n_channels]),
|
||||
*((uint32_t*)&s[9 + 3*n_channels]),
|
||||
*((uint32_t*)&s[9 + 6*n_channels]),
|
||||
*((uint32_t*)&s[9 + 9*n_channels]));
|
||||
*((uint32_t*)&s[3 + 0*n_channels]),
|
||||
*((uint32_t*)&s[3 + 1*n_channels]),
|
||||
*((uint32_t*)&s[3 + 2*n_channels]),
|
||||
*((uint32_t*)&s[3 + 3*n_channels]));
|
||||
|
||||
in[0] = _mm_slli_epi32(in[0], 8);
|
||||
in[1] = _mm_slli_epi32(in[1], 8);
|
||||
|
|
@ -295,13 +295,13 @@ conv_s24_to_f32d_4s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
_mm_store_ps(&d2[n], out[2]);
|
||||
_mm_store_ps(&d3[n], out[3]);
|
||||
|
||||
s += 12 * n_channels;
|
||||
s += 4 * n_channels;
|
||||
}
|
||||
for(; n < n_samples; n++) {
|
||||
out[0] = _mm_cvtsi32_ss(factor, read_s24(s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, read_s24(s+3));
|
||||
out[2] = _mm_cvtsi32_ss(factor, read_s24(s+6));
|
||||
out[3] = _mm_cvtsi32_ss(factor, read_s24(s+9));
|
||||
out[0] = _mm_cvtsi32_ss(factor, s24_to_s32(*s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+1)));
|
||||
out[2] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+2)));
|
||||
out[3] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+3)));
|
||||
out[0] = _mm_mul_ss(out[0], factor);
|
||||
out[1] = _mm_mul_ss(out[1], factor);
|
||||
out[2] = _mm_mul_ss(out[2], factor);
|
||||
|
|
@ -310,7 +310,7 @@ conv_s24_to_f32d_4s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
|
|||
_mm_store_ss(&d1[n], out[1]);
|
||||
_mm_store_ss(&d2[n], out[2]);
|
||||
_mm_store_ss(&d3[n], out[3]);
|
||||
s += 3 * n_channels;
|
||||
s += n_channels;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -30,7 +30,7 @@ static void
|
|||
conv_s24_to_f32d_1s_sse41(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
|
||||
uint32_t n_channels, uint32_t n_samples)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
const int24_t *s = src;
|
||||
float *d0 = dst[0];
|
||||
uint32_t n, unrolled;
|
||||
__m128i in = _mm_setzero_si128();
|
||||
|
|
@ -43,21 +43,21 @@ conv_s24_to_f32d_1s_sse41(void *data, void * SPA_RESTRICT dst[], const void * SP
|
|||
|
||||
for(n = 0; n < unrolled; n += 4) {
|
||||
in = _mm_insert_epi32(in, *((uint32_t*)&s[0 * n_channels]), 0);
|
||||
in = _mm_insert_epi32(in, *((uint32_t*)&s[3 * n_channels]), 1);
|
||||
in = _mm_insert_epi32(in, *((uint32_t*)&s[6 * n_channels]), 2);
|
||||
in = _mm_insert_epi32(in, *((uint32_t*)&s[9 * n_channels]), 3);
|
||||
in = _mm_insert_epi32(in, *((uint32_t*)&s[1 * n_channels]), 1);
|
||||
in = _mm_insert_epi32(in, *((uint32_t*)&s[2 * n_channels]), 2);
|
||||
in = _mm_insert_epi32(in, *((uint32_t*)&s[3 * n_channels]), 3);
|
||||
in = _mm_slli_epi32(in, 8);
|
||||
in = _mm_srai_epi32(in, 8);
|
||||
out = _mm_cvtepi32_ps(in);
|
||||
out = _mm_mul_ps(out, factor);
|
||||
_mm_store_ps(&d0[n], out);
|
||||
s += 12 * n_channels;
|
||||
s += 4 * n_channels;
|
||||
}
|
||||
for(; n < n_samples; n++) {
|
||||
out = _mm_cvtsi32_ss(factor, read_s24(s));
|
||||
out = _mm_cvtsi32_ss(factor, s24_to_s32(*s));
|
||||
out = _mm_mul_ss(out, factor);
|
||||
_mm_store_ss(&d0[n], out);
|
||||
s += 3 * n_channels;
|
||||
s += n_channels;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -30,7 +30,7 @@ static void
|
|||
conv_s24_to_f32d_4s_ssse3(void *data, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src,
|
||||
uint32_t n_channels, uint32_t n_samples)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
const int24_t *s = src;
|
||||
float *d0 = dst[0], *d1 = dst[1], *d2 = dst[2], *d3 = dst[3];
|
||||
uint32_t n, unrolled;
|
||||
__m128i in[4];
|
||||
|
|
@ -48,9 +48,9 @@ conv_s24_to_f32d_4s_ssse3(void *data, void * SPA_RESTRICT dst[], const void * SP
|
|||
|
||||
for(n = 0; n < unrolled; n += 4) {
|
||||
in[0] = _mm_loadu_si128((__m128i*)(s + 0*n_channels));
|
||||
in[1] = _mm_loadu_si128((__m128i*)(s + 3*n_channels));
|
||||
in[2] = _mm_loadu_si128((__m128i*)(s + 6*n_channels));
|
||||
in[3] = _mm_loadu_si128((__m128i*)(s + 9*n_channels));
|
||||
in[1] = _mm_loadu_si128((__m128i*)(s + 1*n_channels));
|
||||
in[2] = _mm_loadu_si128((__m128i*)(s + 2*n_channels));
|
||||
in[3] = _mm_loadu_si128((__m128i*)(s + 3*n_channels));
|
||||
in[0] = _mm_shuffle_epi8(in[0], mask);
|
||||
in[1] = _mm_shuffle_epi8(in[1], mask);
|
||||
in[2] = _mm_shuffle_epi8(in[2], mask);
|
||||
|
|
@ -74,13 +74,13 @@ conv_s24_to_f32d_4s_ssse3(void *data, void * SPA_RESTRICT dst[], const void * SP
|
|||
_mm_store_ps(&d1[n], out[1]);
|
||||
_mm_store_ps(&d2[n], out[2]);
|
||||
_mm_store_ps(&d3[n], out[3]);
|
||||
s += 12 * n_channels;
|
||||
s += 4 * n_channels;
|
||||
}
|
||||
for(; n < n_samples; n++) {
|
||||
out[0] = _mm_cvtsi32_ss(factor, read_s24(s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, read_s24(s+3));
|
||||
out[2] = _mm_cvtsi32_ss(factor, read_s24(s+6));
|
||||
out[3] = _mm_cvtsi32_ss(factor, read_s24(s+9));
|
||||
out[0] = _mm_cvtsi32_ss(factor, s24_to_s32(*s));
|
||||
out[1] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+1)));
|
||||
out[2] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+2)));
|
||||
out[3] = _mm_cvtsi32_ss(factor, s24_to_s32(*(s+3)));
|
||||
out[0] = _mm_mul_ss(out[0], factor);
|
||||
out[1] = _mm_mul_ss(out[1], factor);
|
||||
out[2] = _mm_mul_ss(out[2], factor);
|
||||
|
|
@ -89,7 +89,7 @@ conv_s24_to_f32d_4s_ssse3(void *data, void * SPA_RESTRICT dst[], const void * SP
|
|||
_mm_store_ss(&d1[n], out[1]);
|
||||
_mm_store_ss(&d2[n], out[2]);
|
||||
_mm_store_ss(&d3[n], out[3]);
|
||||
s += 3 * n_channels;
|
||||
s += n_channels;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -79,17 +79,19 @@
|
|||
#define U24_MAX 16777215u
|
||||
#define U24_SCALE 8388607.5f
|
||||
#define U24_OFFS 8388608.f
|
||||
#define U24_TO_F32(v) ((((uint32_t)(v)) * (1.0f / U24_OFFS)) - 1.0)
|
||||
#define F32_TO_U24(v) (uint32_t)SPA_CLAMP((v) * U24_SCALE + U24_OFFS, U24_MIN, U24_MAX)
|
||||
#define F32_TO_U24_D(v,d) (uint32_t)SPA_CLAMP((v) * U24_SCALE + U24_OFFS + (d), U24_MIN, U24_MAX)
|
||||
#define U24_TO_F32(v) ((u24_to_u32(v) * (1.0f / U24_OFFS)) - 1.0)
|
||||
#define F32_TO_U24(v) u32_to_u24(SPA_CLAMP((v) * U24_SCALE + U24_OFFS, U24_MIN, U24_MAX))
|
||||
#define F32_TO_U24_D(v,d) u32_to_u24(SPA_CLAMP((v) * U24_SCALE + U24_OFFS + (d), U24_MIN, U24_MAX))
|
||||
|
||||
#define S24_MIN -8388607
|
||||
#define S24_MAX 8388607
|
||||
#define S24_MAX_F 8388607.0f
|
||||
#define S24_SCALE 8388607.0f
|
||||
#define S24_TO_F32(v) (((int32_t)(v)) * (1.0f / S24_SCALE))
|
||||
#define F32_TO_S24(v) (int32_t)SPA_CLAMP((v) * S24_SCALE, S24_MIN, S24_MAX)
|
||||
#define F32_TO_S24_D(v,d) (int32_t)SPA_CLAMP((v) * S24_SCALE + (d), S24_MIN, S24_MAX)
|
||||
#define S24_TO_F32(v) (s24_to_s32(v) * (1.0f / S24_SCALE))
|
||||
#define S24S_TO_F32(v) (s24_to_s32(bswap_s24(v)) * (1.0f / S24_SCALE))
|
||||
#define F32_TO_S24(v) s32_to_s24(SPA_CLAMP((v) * S24_SCALE, S24_MIN, S24_MAX))
|
||||
#define F32_TO_S24S(v) bswap_s24(F32_TO_S24(v))
|
||||
#define F32_TO_S24_D(v,d) s32_to_s24(SPA_CLAMP((v) * S24_SCALE + (d), S24_MIN, S24_MAX))
|
||||
|
||||
#define U32_MIN 0u
|
||||
#define U32_MAX 4294967040u
|
||||
|
|
@ -112,88 +114,69 @@
|
|||
|
||||
#define U24_32_TO_F32(v) U32_TO_F32((v)<<8)
|
||||
#define U24_32S_TO_F32(v) U32_TO_F32(((int32_t)bswap_32(v))<<8)
|
||||
#define F32_TO_U24_32(v) F32_TO_U24(v)
|
||||
#define F32_TO_U24_32S(v) bswap_32(F32_TO_U24(v))
|
||||
#define F32_TO_U24_32_D(v,d) F32_TO_U24_D(v,d)
|
||||
#define F32_TO_U24_32S_D(v,d) bswap_32(F32_TO_U24_D(v,d))
|
||||
#define F32_TO_U24_32(v) (uint32_t)SPA_CLAMP((v) * U24_SCALE + U24_OFFS, U24_MIN, U24_MAX)
|
||||
#define F32_TO_U24_32S(v) bswap_32(F32_TO_U24_32(v))
|
||||
#define F32_TO_U24_32_D(v,d) (uint32_t)SPA_CLAMP((v) * U24_SCALE + U24_OFFS + (d), U24_MIN, U24_MAX)
|
||||
#define F32_TO_U24_32S_D(v,d) bswap_32(F32_TO_U24_32_D(v,d))
|
||||
|
||||
#define S24_32_TO_F32(v) S32_TO_F32((v)<<8)
|
||||
#define S24_32S_TO_F32(v) S32_TO_F32(((int32_t)bswap_32(v))<<8)
|
||||
#define F32_TO_S24_32(v) F32_TO_S24(v)
|
||||
#define F32_TO_S24_32S(v) bswap_32(F32_TO_S24(v))
|
||||
#define F32_TO_S24_32_D(v,d) F32_TO_S24_D(v,d)
|
||||
#define F32_TO_S24_32S_D(v,d) bswap_32(F32_TO_S24_D(v,d))
|
||||
#define F32_TO_S24_32(v) (int32_t)SPA_CLAMP((v) * S24_SCALE, S24_MIN, S24_MAX)
|
||||
#define F32_TO_S24_32S(v) bswap_32(F32_TO_S24_32(v))
|
||||
#define F32_TO_S24_32_D(v,d) (int32_t)SPA_CLAMP((v) * S24_SCALE + (d), S24_MIN, S24_MAX)
|
||||
#define F32_TO_S24_32S_D(v,d) bswap_32(F32_TO_S24_32_D(v,d))
|
||||
|
||||
static inline uint32_t read_u24(const void *src)
|
||||
{
|
||||
const uint8_t *s = src;
|
||||
typedef struct {
|
||||
#if __BYTE_ORDER == __LITTLE_ENDIAN
|
||||
return (((uint32_t)s[2] << 16) | ((uint32_t)(uint8_t)s[1] << 8) | (uint32_t)(uint8_t)s[0]);
|
||||
uint8_t v3;
|
||||
uint8_t v2;
|
||||
uint8_t v1;
|
||||
#else
|
||||
return (((uint32_t)s[0] << 16) | ((uint32_t)(uint8_t)s[1] << 8) | (uint32_t)(uint8_t)s[2]);
|
||||
uint8_t v1;
|
||||
uint8_t v2;
|
||||
uint8_t v3;
|
||||
#endif
|
||||
} __attribute__ ((packed)) uint24_t;
|
||||
|
||||
typedef struct {
|
||||
#if __BYTE_ORDER == __LITTLE_ENDIAN
|
||||
uint8_t v3;
|
||||
uint8_t v2;
|
||||
int8_t v1;
|
||||
#else
|
||||
int8_t v1;
|
||||
uint8_t v2;
|
||||
uint8_t v3;
|
||||
#endif
|
||||
} __attribute__ ((packed)) int24_t;
|
||||
|
||||
static inline uint32_t u24_to_u32(uint24_t src)
|
||||
{
|
||||
return ((uint32_t)src.v1 << 16) | ((uint32_t)src.v2 << 8) | (uint32_t)src.v3;
|
||||
}
|
||||
|
||||
static inline int32_t read_s24(const void *src)
|
||||
static inline uint24_t u32_to_u24(uint32_t src)
|
||||
{
|
||||
const int8_t *s = src;
|
||||
#if __BYTE_ORDER == __LITTLE_ENDIAN
|
||||
return (((int32_t)s[2] << 16) | ((uint32_t)(uint8_t)s[1] << 8) | (uint32_t)(uint8_t)s[0]);
|
||||
#else
|
||||
return (((int32_t)s[0] << 16) | ((uint32_t)(uint8_t)s[1] << 8) | (uint32_t)(uint8_t)s[2]);
|
||||
#endif
|
||||
return (uint24_t) { src >> 16, src >> 8, src };
|
||||
}
|
||||
|
||||
static inline int32_t read_s24s(const void *src)
|
||||
static inline int32_t s24_to_s32(int24_t src)
|
||||
{
|
||||
const int8_t *s = src;
|
||||
#if __BYTE_ORDER == __LITTLE_ENDIAN
|
||||
return (((int32_t)s[0] << 16) | ((uint32_t)(uint8_t)s[1] << 8) | (uint32_t)(uint8_t)s[2]);
|
||||
#else
|
||||
return (((int32_t)s[2] << 16) | ((uint32_t)(uint8_t)s[1] << 8) | (uint32_t)(uint8_t)s[0]);
|
||||
#endif
|
||||
return ((int32_t)src.v1 << 16) | ((uint32_t)src.v2 << 8) | (uint32_t)src.v3;
|
||||
}
|
||||
|
||||
static inline void write_u24(void *dst, uint32_t val)
|
||||
static inline int24_t s32_to_s24(int32_t src)
|
||||
{
|
||||
uint8_t *d = dst;
|
||||
#if __BYTE_ORDER == __LITTLE_ENDIAN
|
||||
d[0] = (uint8_t) (val);
|
||||
d[1] = (uint8_t) (val >> 8);
|
||||
d[2] = (uint8_t) (val >> 16);
|
||||
#else
|
||||
d[0] = (uint8_t) (val >> 16);
|
||||
d[1] = (uint8_t) (val >> 8);
|
||||
d[2] = (uint8_t) (val);
|
||||
#endif
|
||||
return (int24_t) { src >> 16, src >> 8, src };
|
||||
}
|
||||
|
||||
static inline void write_s24(void *dst, int32_t val)
|
||||
static inline uint24_t bswap_u24(uint24_t src)
|
||||
{
|
||||
uint8_t *d = dst;
|
||||
#if __BYTE_ORDER == __LITTLE_ENDIAN
|
||||
d[0] = (uint8_t) (val);
|
||||
d[1] = (uint8_t) (val >> 8);
|
||||
d[2] = (uint8_t) (val >> 16);
|
||||
#else
|
||||
d[0] = (uint8_t) (val >> 16);
|
||||
d[1] = (uint8_t) (val >> 8);
|
||||
d[2] = (uint8_t) (val);
|
||||
#endif
|
||||
return (uint24_t) { src.v3, src.v2, src.v1 };
|
||||
}
|
||||
|
||||
static inline void write_s24s(void *dst, int32_t val)
|
||||
static inline int24_t bswap_s24(int24_t src)
|
||||
{
|
||||
uint8_t *d = dst;
|
||||
#if __BYTE_ORDER == __LITTLE_ENDIAN
|
||||
d[0] = (uint8_t) (val >> 16);
|
||||
d[1] = (uint8_t) (val >> 8);
|
||||
d[2] = (uint8_t) (val);
|
||||
#else
|
||||
d[0] = (uint8_t) (val);
|
||||
d[1] = (uint8_t) (val >> 8);
|
||||
d[2] = (uint8_t) (val >> 16);
|
||||
#endif
|
||||
return (int24_t) { src.v3, src.v2, src.v1 };
|
||||
}
|
||||
|
||||
#define NS_MAX 8
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue