mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2025-11-03 09:01:54 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions
pipewire/spa/plugins/audioconvert/fmt-ops.h

487 lines
15 KiB
C
Raw Normal View History

treewide: use SPDX tags to specify copyright information SPDX tags make the licensing information easy to understand and clear, and they are machine parseable. See https://spdx.dev for more information.
2023-02-08 18:12:00 +01:00
/* Spa */
/* SPDX-FileCopyrightText: Copyright © 2019 Wim Taymans */
/* SPDX-License-Identifier: MIT */
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
#include <math.h>
#include <spa/utils/defs.h>
Add spa/utils/endian.h This provides access to GNU C library-style endian and byteswap functions. Windows doesn't provide pre-processor defines for endianness, but all current Windows architectures (X32, X64, ARM) are little-endian.
2024-06-25 16:32:13 -07:00
#include <spa/utils/endian.h>
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
#include <spa/utils/string.h>
audioconvert: round instead of truncate, to reduce distortion See #2543
2022-07-13 20:56:13 +02:00
#define f32_round(a) lrintf(a)
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define ITOF(type,v,scale,offs) \
(((type)(v)) * (1.0f / (scale)) - (offs))
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
#define FTOI(type,v,scale,offs,noise,min,max) \
audioconvert: use SPA_CLAMPF to clamp floats It generates better assembler.
2022-07-19 17:59:14 +02:00
(type)f32_round(SPA_CLAMPF((v) * (scale) + (offs) + (noise), min, max))
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
spa: use static inline for interfaces instead of macro It gives better typechecking and a path to make a library of functions.
2024-11-19 19:57:37 +01:00
#define FMT_OPS_MAX_ALIGN 32u
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define U8_MIN 0u
#define U8_MAX 255u
audioconvert: tweak more conversion constants Tweak some more constants to make lossless conversions work. Add some tests for lossless conversion. Add some more tests.
2022-07-05 15:50:05 +02:00
#define U8_SCALE 128.f
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define U8_OFFS 128.f
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define U8_TO_F32(v) ITOF(uint8_t, v, U8_SCALE, 1.0f)
#define F32_TO_U8_D(v,d) FTOI(uint8_t, v, U8_SCALE, U8_OFFS, d, U8_MIN, U8_MAX)
#define F32_TO_U8(v) F32_TO_U8_D(v, 0.0f)
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
audioconvert: tweak more conversion constants Tweak some more constants to make lossless conversions work. Add some tests for lossless conversion. Add some more tests.
2022-07-05 15:50:05 +02:00
#define S8_MIN -128
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define S8_MAX 127
audioconvert: tweak more conversion constants Tweak some more constants to make lossless conversions work. Add some tests for lossless conversion. Add some more tests.
2022-07-05 15:50:05 +02:00
#define S8_SCALE 128.0f
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define S8_TO_F32(v) ITOF(int8_t, v, S8_SCALE, 0.0f)
#define F32_TO_S8_D(v,d) FTOI(int8_t, v, S8_SCALE, 0.0f, d, S8_MIN, S8_MAX)
#define F32_TO_S8(v) F32_TO_S8_D(v, 0.0f);
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define U16_MIN 0u
#define U16_MAX 65535u
audioconvert: tweak more conversion constants Tweak some more constants to make lossless conversions work. Add some tests for lossless conversion. Add some more tests.
2022-07-05 15:50:05 +02:00
#define U16_SCALE 32768.f
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define U16_OFFS 32768.f
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define U16_TO_F32(v) ITOF(uint16_t, v, U16_SCALE, 1.0f)
#define U16S_TO_F32(v) U16_TO_F32(bswap_16(v))
#define F32_TO_U16_D(v,d) FTOI(uint16_t, v, U16_SCALE, U16_OFFS, d, U16_MIN, U16_MAX)
#define F32_TO_U16(v) F32_TO_U16_D(v, 0.0f);
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define F32_TO_U16S_D(v,d) bswap_16(F32_TO_U16_D(v,d))
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define F32_TO_U16S(v) bswap_16(F32_TO_U16(v))
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
audioconvert: tweak conversion constants Tweak the conversion constants a bit so that they handle the extreme ranges a bit better. Align the C and vector instructions. Reactivate the unit test asserts when a conversion fails.
2022-07-05 12:20:02 +02:00
#define S16_MIN -32768
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define S16_MAX 32767
audioconvert: tweak conversion constants Tweak the conversion constants a bit so that they handle the extreme ranges a bit better. Align the C and vector instructions. Reactivate the unit test asserts when a conversion fails.
2022-07-05 12:20:02 +02:00
#define S16_SCALE 32768.0f
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define S16_TO_F32(v) ITOF(int16_t, v, S16_SCALE, 0.0f)
#define S16S_TO_F32(v) S16_TO_F32(bswap_16(v))
#define F32_TO_S16_D(v,d) FTOI(int16_t, v, S16_SCALE, 0.0f, d, S16_MIN, S16_MAX)
#define F32_TO_S16(v) F32_TO_S16_D(v, 0.0f)
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define F32_TO_S16S_D(v,d) bswap_16(F32_TO_S16_D(v,d))
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define F32_TO_S16S(v) bswap_16(F32_TO_S16(v))
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define U24_MIN 0u
#define U24_MAX 16777215u
audioconvert: tweak more conversion constants Tweak some more constants to make lossless conversions work. Add some tests for lossless conversion. Add some more tests.
2022-07-05 15:50:05 +02:00
#define U24_SCALE 8388608.f
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define U24_OFFS 8388608.f
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define U24_TO_F32(v) ITOF(uint32_t, u24_to_u32(v), U24_SCALE, 1.0f)
#define F32_TO_U24_D(v,d) u32_to_u24(FTOI(uint32_t, v, U24_SCALE, U24_OFFS, d, U24_MIN, U24_MAX))
#define F32_TO_U24(v) F32_TO_U24_D(v, 0.0f)
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
audioconvert: tweak conversion constants Tweak the conversion constants a bit so that they handle the extreme ranges a bit better. Align the C and vector instructions. Reactivate the unit test asserts when a conversion fails.
2022-07-05 12:20:02 +02:00
#define S24_MIN -8388608
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define S24_MAX 8388607
audioconvert: tweak conversion constants Tweak the conversion constants a bit so that they handle the extreme ranges a bit better. Align the C and vector instructions. Reactivate the unit test asserts when a conversion fails.
2022-07-05 12:20:02 +02:00
#define S24_SCALE 8388608.0f
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define S24_TO_F32(v) ITOF(int32_t, s24_to_s32(v), S24_SCALE, 0.0f)
#define S24S_TO_F32(v) S24_TO_F32(bswap_s24(v))
#define F32_TO_S24_D(v,d) s32_to_s24(FTOI(int32_t, v, S24_SCALE, 0.0f, d, S24_MIN, S24_MAX))
#define F32_TO_S24(v) F32_TO_S24_D(v, 0.0f)
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.
2022-07-01 12:24:35 +02:00
#define F32_TO_S24S(v) bswap_s24(F32_TO_S24(v))
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
audioconvert: make sure shifts are defined Don't to signed left shifts, make macros to do the shifting on unsigned ints when needed. See !1505
2023-01-24 21:29:18 +01:00
#define U24_32_TO_U32(v) (((uint32_t)(v)) << 8)
#define U24_32_TO_F32(v) U32_TO_F32(U24_32_TO_U32(v))
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define U24_32S_TO_F32(v) U24_32_TO_F32(bswap_32(v))
#define F32_TO_U24_32_D(v,d) FTOI(uint32_t, v, U24_SCALE, U24_OFFS, d, U24_MIN, U24_MAX)
#define F32_TO_U24_32(v) F32_TO_U24_32_D(v, 0.0f)
fmt-ops: express 32 bits formats in terms of 32_24 bits formats
2022-07-07 20:11:08 +02:00
#define F32_TO_U24_32S(v) bswap_32(F32_TO_U24_32(v))
#define F32_TO_U24_32S_D(v,d) bswap_32(F32_TO_U24_32_D(v,d))
audioconvert: make sure shifts are defined Don't to signed left shifts, make macros to do the shifting on unsigned ints when needed. See !1505
2023-01-24 21:29:18 +01:00
#define U32_TO_U24_32(v) (((uint32_t)(v)) >> 8)
audioconvert: introduce s25_32 type, f32<->s25 cast is lossless The largest integer that 32-bit floating point can exactly represent is actually `(2^24)-1`, not`(2^23)-1` like the code assumes. This means, whenever we use s24 as an intermediate step to go between f32 and s32, we lose a bit of precision. s25_32 is really a i32 with highest byte always being a sign byte. Printing was done by adding ``` for(int e = 0; e != 13; ++e) fprintf(stderr, "%16.32e,", ((float*)m1)[e]); ``` to `compare_mem`. I don't like how these tests work. https://godbolt.org/z/abe94sedT
2024-06-13 16:18:33 +03:00
#define S25_MIN -16777216
#define S25_MAX 16777215
#define S25_SCALE 16777216.0f
#define S25_32_TO_F32(v) ITOF(int32_t, v, S25_SCALE, 0.0f)
#define S25_32S_TO_F32(v) S25_32_TO_F32(bswap_32(v))
#define F32_TO_S25_32_D(v,d) FTOI(int32_t, v, S25_SCALE, 0.0f, d, S25_MIN, S25_MAX)
#define F32_TO_S25_32(v) F32_TO_S25_32_D(v, 0.0f)
#define F32_TO_S25_32S(v) bswap_32(F32_TO_S25_32(v))
#define F32_TO_S25_32S_D(v,d) bswap_32(F32_TO_S25_32_D(v,d))
#define S25_32_TO_S32(v) ((int32_t)(((uint32_t)(v)) << 7))
#define S32_TO_S25_32(v) (((int32_t)(v)) >> 7)
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define U32_MIN 0u
audioconvert: use 24 bits for 32 formats The float only preserves 24 bits so use this. Otherwise we get overflows and errors in clang.
2022-07-07 18:38:32 +02:00
#define U32_MAX 4294967295u
audioconvert: tweak more conversion constants Tweak some more constants to make lossless conversions work. Add some tests for lossless conversion. Add some more tests.
2022-07-05 15:50:05 +02:00
#define U32_SCALE 2147483648.f
audioconvert: tweak conversion constants Tweak the conversion constants a bit so that they handle the extreme ranges a bit better. Align the C and vector instructions. Reactivate the unit test asserts when a conversion fails.
2022-07-05 12:20:02 +02:00
#define U32_OFFS 2147483648.f
audioconvert: make sure shifts are defined Don't to signed left shifts, make macros to do the shifting on unsigned ints when needed. See !1505
2023-01-24 21:29:18 +01:00
#define U32_TO_F32(v) ITOF(uint32_t, U32_TO_U24_32(v), U24_SCALE, 1.0f)
#define F32_TO_U32(v) U24_32_TO_U32(F32_TO_U24_32(v))
#define F32_TO_U32_D(v,d) U24_32_TO_U32(F32_TO_U24_32_D(v,d))
fmt-ops: express 32 bits formats in terms of 32_24 bits formats
2022-07-07 20:11:08 +02:00
audioconvert: make sure shifts are defined Don't to signed left shifts, make macros to do the shifting on unsigned ints when needed. See !1505
2023-01-24 21:29:18 +01:00
#define S24_32_TO_S32(v) ((int32_t)(((uint32_t)(v)) << 8))
#define S24_32_TO_F32(v) S32_TO_F32(S24_32_TO_S32(v))
audiconvert: make macros for conversions Make a common macro for float to int and int to float so that we can change the algorithms easily.
2022-07-13 18:10:25 +02:00
#define S24_32S_TO_F32(v) S24_32_TO_F32(bswap_32(v))
#define F32_TO_S24_32_D(v,d) FTOI(int32_t, v, S24_SCALE, 0.0f, d, S24_MIN, S24_MAX)
#define F32_TO_S24_32(v) F32_TO_S24_32_D(v, 0.0f)
fmt-ops: express 32 bits formats in terms of 32_24 bits formats
2022-07-07 20:11:08 +02:00
#define F32_TO_S24_32S(v) bswap_32(F32_TO_S24_32(v))
#define F32_TO_S24_32S_D(v,d) bswap_32(F32_TO_S24_32_D(v,d))
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
audioconvert: make sure shifts are defined Don't to signed left shifts, make macros to do the shifting on unsigned ints when needed. See !1505
2023-01-24 21:29:18 +01:00
#define S32_TO_S24_32(v) (((int32_t)(v)) >> 8)
audioconvert: avoid even more precision loss in S32 to F32 conversion There's really no point in doing that s25_32 intermediate step, to be honest i don't have a clue why the original implementation did that \_(ツ)_/¯. Both `S25_SCALE` and `S32_SCALE` are powers of two, and thus are both exactly representable as floats, and reprocial of power-of-two is also exactly representable, so it's not like that rescaling results in precision loss. This additionally avoids right-shift, and thus is even faster. As `test_lossless_s32_lossless_subset` shows, if the integer is in the form of s25+shift, the maximal absolute error became even lower, but not zero, because F32->S32 still goes through S25 intermediate. I think we could theoretically do better, but then the clamping becomes pretty finicky, so i don't feel like touching that here.
2024-06-14 06:05:18 +03:00
#define S32_MIN -2147483648
#define S32_MAX 2147483647
audioconvert: avoid even more precision loss in F32 to S32 conversion This is somewhat similar to the S32->F32 conversion improvements, but here things a bit more tricky... The main consideration is that the limits to which we clamp must be valid 32-bit signed integers, but not all such integers are exactly losslessly representable in `float32_t`. For example it we'd clamp to `2147483647`, that is actually a `2147483648.0f`, and `2147483648` is not a valid 32-bit signed integer, so the post-clamp conversion would basically be UB. We don't have this problem for negative bound, though. But as we know, any 25-bit signed integer is losslessly round-trippable through float32_t, and since multiplying by 2 only changes the float's exponent, we can clamp to `2147483520`! The algorithm of selection of the pre-clamping scale is unaffected. This additionally avoids right-shift, and thus is even faster. As `test_lossless_s32_lossless_subset` shows, if the integer is in the form of s25+shift, the maximal absolute error is finally zero. Without going through `float`->`double`->`int`, i'm not sure if the `float`->`int` conversion can be improved further.
2024-06-25 19:20:42 +03:00
#define S32_SCALE_I2F 2147483648.0f
#define S32_TO_F32(v) ITOF(int32_t, v, S32_SCALE_I2F, 0.0f)
fmt-ops: express 32 bits formats in terms of 32_24 bits formats
2022-07-07 20:11:08 +02:00
#define S32S_TO_F32(v) S32_TO_F32(bswap_32(v))
audioconvert: avoid even more precision loss in F32 to S32 conversion This is somewhat similar to the S32->F32 conversion improvements, but here things a bit more tricky... The main consideration is that the limits to which we clamp must be valid 32-bit signed integers, but not all such integers are exactly losslessly representable in `float32_t`. For example it we'd clamp to `2147483647`, that is actually a `2147483648.0f`, and `2147483648` is not a valid 32-bit signed integer, so the post-clamp conversion would basically be UB. We don't have this problem for negative bound, though. But as we know, any 25-bit signed integer is losslessly round-trippable through float32_t, and since multiplying by 2 only changes the float's exponent, we can clamp to `2147483520`! The algorithm of selection of the pre-clamping scale is unaffected. This additionally avoids right-shift, and thus is even faster. As `test_lossless_s32_lossless_subset` shows, if the integer is in the form of s25+shift, the maximal absolute error is finally zero. Without going through `float`->`double`->`int`, i'm not sure if the `float`->`int` conversion can be improved further.
2024-06-25 19:20:42 +03:00
#define S32_MIN_F2I ((int32_t)(((uint32_t)(S25_MIN)) << 7))
#define S32_MAX_F2I ((S25_MAX) << 7)
#define S32_SCALE_F2I (-((float)(S32_MIN_F2I)))
#define F32_TO_S32_D(v,d) FTOI(int32_t, v, S32_SCALE_F2I, 0.0f, d, S32_MIN_F2I, S32_MAX_F2I)
audioconvert: somewhat avoid precision loss in F32 to S32 conversion At the very least, we should go through s25_32 intermediate instead of s24_32, to avoid needlessly loosing 1 LSB precision bit. FIXME: the noise codepath is not covered with tests.
2024-06-14 04:48:03 +03:00
#define F32_TO_S32(v) F32_TO_S32_D(v, 0.0f)
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
#define F32_TO_S32S(v) bswap_32(F32_TO_S32(v))
#define F32_TO_S32S_D(v,d) bswap_32(F32_TO_S32_D(v,d))
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
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.
2022-07-01 12:24:35 +02:00
typedef struct {
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
#if __BYTE_ORDER == __LITTLE_ENDIAN
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.
2022-07-01 12:24:35 +02:00
uint8_t v3;
uint8_t v2;
uint8_t v1;
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
#else
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.
2022-07-01 12:24:35 +02:00
uint8_t v1;
uint8_t v2;
uint8_t v3;
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
#endif
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.
2022-07-01 12:24:35 +02:00
} __attribute__ ((packed)) uint24_t;
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -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.
2022-07-01 12:24:35 +02:00
typedef struct {
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
#if __BYTE_ORDER == __LITTLE_ENDIAN
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.
2022-07-01 12:24:35 +02:00
uint8_t v3;
uint8_t v2;
int8_t v1;
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
#else
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.
2022-07-01 12:24:35 +02:00
int8_t v1;
uint8_t v2;
uint8_t v3;
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
#endif
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.
2022-07-01 12:24:35 +02:00
} __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;
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
}
audioconvert: fix pack/unpack of s24/u24 Copy the values into the right place in the struct.
2022-07-05 12:18:57 +02:00
#define U32_TO_U24(s) (uint24_t) { .v1 = (uint8_t)(((uint32_t)s) >> 16), \
.v2 = (uint8_t)(((uint32_t)s) >> 8), .v3 = (uint8_t)((uint32_t)s) }
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.
2022-07-01 12:24:35 +02:00
static inline uint24_t u32_to_u24(uint32_t src)
fmt-ops: add support for s24 with endian conversion
2020-01-27 15:46:25 +01:00
{
audioconvert: fix pack/unpack of s24/u24 Copy the values into the right place in the struct.
2022-07-05 12:18:57 +02:00
return U32_TO_U24(src);
fmt-ops: add support for s24 with endian conversion
2020-01-27 15:46:25 +01:00
}
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.
2022-07-01 12:24:35 +02:00
static inline int32_t s24_to_s32(int24_t src)
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
{
audioconvert: don't left shift negative values See #3572
2023-10-15 21:00:01 +02:00
return ((uint32_t)((int32_t)src.v1 & 0xFFFF) << 16) | ((uint32_t)src.v2 << 8) | (uint32_t)src.v3;
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
}
audioconvert: fix pack/unpack of s24/u24 Copy the values into the right place in the struct.
2022-07-05 12:18:57 +02:00
#define S32_TO_S24(s) (int24_t) { .v1 = (int8_t)(((int32_t)s) >> 16), \
.v2 = (uint8_t)(((uint32_t)s) >> 8), .v3 = (uint8_t)((uint32_t)s) }
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.
2022-07-01 12:24:35 +02:00
static inline int24_t s32_to_s24(int32_t src)
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
{
audioconvert: fix pack/unpack of s24/u24 Copy the values into the right place in the struct.
2022-07-05 12:18:57 +02:00
return S32_TO_S24(src);
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
}
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.
2022-07-01 12:24:35 +02:00
static inline uint24_t bswap_u24(uint24_t src)
fmt-ops: add support for s24 with endian conversion
2020-01-27 15:46:25 +01:00
{
audioconvert: Fix s24/u24 byte swapping
2022-07-05 14:56:54 +02:00
return (uint24_t) { .v1 = src.v3, .v2 = src.v2, .v3 = src.v1 };
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.
2022-07-01 12:24:35 +02:00
}
static inline int24_t bswap_s24(int24_t src)
{
audioconvert: Fix s24/u24 byte swapping
2022-07-05 14:56:54 +02:00
return (int24_t) { .v1 = src.v3, .v2 = src.v2, .v3 = src.v1 };
fmt-ops: add support for s24 with endian conversion
2020-01-27 15:46:25 +01:00
}
audioconvert: implement f64s Add swapping functions for f64s. Fix the awkward interleave/deinterleave names for 32s.
2022-07-11 10:58:51 +02:00
#define F32_TO_F32S(v) \
bswap_32((union { uint32_t i; float f; }){ .f = (v) }.i)
#define F32S_TO_F32(v) \
((union { uint32_t i; float f; }){ .i = bswap_32(v) }.f)
#define F64_TO_F64S(v) \
bswap_32((union { uint64_t i; double d; }){ .d = (v) }.i)
#define F64S_TO_F64(v) \
((union { uint64_t i; double d; }){ .i = bswap_32(v) }.d)
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
#define NS_MAX 8
#define NS_MASK (NS_MAX-1)
struct shaper {
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
float e[NS_MAX * 2];
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
uint32_t idx;
float r;
};
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
struct convert {
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
uint32_t noise_bits;
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
#define DITHER_METHOD_NONE 0
#define DITHER_METHOD_RECTANGULAR 1
#define DITHER_METHOD_TRIANGULAR 2
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
#define DITHER_METHOD_TRIANGULAR_HF 3
audioconvert: Fix Wannamaker name
2022-07-18 10:21:56 +02:00
#define DITHER_METHOD_WANNAMAKER_3 4
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
#define DITHER_METHOD_LIPSHITZ 5
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
uint32_t method;
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
uint32_t src_fmt;
uint32_t dst_fmt;
uint32_t n_channels;
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
uint32_t rate;
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
uint32_t cpu_flags;
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
const char *func_name;
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
make bitfields unsigned Remove driver property from client-node of client-stream.
2019-04-10 17:52:42 +02:00
unsigned int is_passthrough:1;
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
audioconvert: improve format conversion Make dither noise as a value between -0.5 and 0.5 and add this to the scaled samples. For this, we first need to do the scaling and then the CLAMP to the target depth. This optimizes to the same code but allows us to avoid under and overflows when we add the dither noise. Add more dithering methods. Expose a dither.method property on audioconvert. Disable dither when the target depth > 16.
2022-06-29 14:10:15 +02:00
float scale;
audioconvert: make separate noise functions So that we can reuse optimized versions in unoptimized noise functions. Do allocation a little different so that we can align everything from the start.
2022-09-29 20:34:15 +02:00
uint32_t *random;
int32_t *prev;
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
#define NOISE_METHOD_NONE 0
#define NOISE_METHOD_RECTANGULAR 1
#define NOISE_METHOD_TRIANGULAR 2
#define NOISE_METHOD_TRIANGULAR_HF 3
audioconvert: improve noise bits Make a new noise method called PATTERN and use it to add a slow (every 1024 samples) repeating pattern of -1, 0. Only use this method when we don't already use triangular dither. See #2540
2022-07-18 11:39:24 +02:00
#define NOISE_METHOD_PATTERN 4
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
uint32_t noise_method;
float *noise;
uint32_t noise_size;
const float *ns;
uint32_t n_ns;
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
struct shaper shaper[64];
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
audioconvert: make separate noise functions So that we can reuse optimized versions in unoptimized noise functions. Do allocation a little different so that we can align everything from the start.
2022-09-29 20:34:15 +02:00
void (*update_noise) (struct convert *conv, float *noise, uint32_t n_samples);
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
void (*process) (struct convert *conv, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[],
uint32_t n_samples);
void (*free) (struct convert *conv);
audioconvert: make separate noise functions So that we can reuse optimized versions in unoptimized noise functions. Do allocation a little different so that we can align everything from the start.
2022-09-29 20:34:15 +02:00
void *data;
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
};
int convert_init(struct convert *conv);
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
static const struct dither_method_info {
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
uint32_t method;
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
const char *label;
const char *description;
} dither_method_info[] = {
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
[DITHER_METHOD_NONE] = { DITHER_METHOD_NONE,
"none", "Disabled", },
[DITHER_METHOD_RECTANGULAR] = { DITHER_METHOD_RECTANGULAR,
"rectangular", "Rectangular dithering", },
[DITHER_METHOD_TRIANGULAR] = { DITHER_METHOD_TRIANGULAR,
"triangular", "Triangular dithering", },
[DITHER_METHOD_TRIANGULAR_HF] = { DITHER_METHOD_TRIANGULAR_HF,
"triangular-hf", "Sloped Triangular dithering", },
audioconvert: Fix Wannamaker name
2022-07-18 10:21:56 +02:00
[DITHER_METHOD_WANNAMAKER_3] = { DITHER_METHOD_WANNAMAKER_3,
"wannamaker3", "Wannamaker 3 dithering", },
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
[DITHER_METHOD_LIPSHITZ] = { DITHER_METHOD_LIPSHITZ,
"shaped5", "Lipshitz 5 dithering", },
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
};
static inline uint32_t dither_method_from_label(const char *label)
{
spa: add macro to simplify array iterations some more uint32_t i; for (i = 0; i < SPA_N_ELEMENTS(some_array); i++) .. stuff with some_array[i].foo ... becomes: SPA_FOR_EACH_ELEMENT_VAR(some_array, p) .. stuff with p->foo ..
2022-09-30 16:21:28 +02:00
SPA_FOR_EACH_ELEMENT_VAR(dither_method_info, i) {
if (spa_streq(i->label, label))
return i->method;
audioconvert: move dither and noise to fmt-ops We need to do dithering and noise when converting f32 to the target format. This is more natural because we can work in 32 bits integers instead of floats. This will also make it possible to actually calculate the error between source and target values and implement some sort of feedback and noise shaping later.
2022-06-28 16:55:50 +02:00
}
return DITHER_METHOD_NONE;
}
audioconvert: make separate noise functions So that we can reuse optimized versions in unoptimized noise functions. Do allocation a little different so that we can align everything from the start.
2022-09-29 20:34:15 +02:00
#define convert_update_noise(conv,...) (conv)->update_noise(conv, __VA_ARGS__)
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
#define convert_process(conv,...) (conv)->process(conv, __VA_ARGS__)
#define convert_free(conv) (conv)->free(conv)
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
audioconvert: make separate noise functions So that we can reuse optimized versions in unoptimized noise functions. Do allocation a little different so that we can align everything from the start.
2022-09-29 20:34:15 +02:00
#define DEFINE_NOISE_FUNCTION(name,arch) \
void conv_noise_##name##_##arch(struct convert *conv, float *noise, \
uint32_t n_samples)
DEFINE_NOISE_FUNCTION(none, c);
DEFINE_NOISE_FUNCTION(rect, c);
DEFINE_NOISE_FUNCTION(tri, c);
DEFINE_NOISE_FUNCTION(tri_hf, c);
DEFINE_NOISE_FUNCTION(pattern, c);
#if defined(HAVE_SSE2)
DEFINE_NOISE_FUNCTION(rect, sse2);
DEFINE_NOISE_FUNCTION(tri, sse2);
DEFINE_NOISE_FUNCTION(tri_hf, sse2);
#endif
#undef DEFINE_NOISE_FUNCTION
#define DEFINE_FUNCTION(name,arch) \
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
void conv_##name##_##arch(struct convert *conv, void * SPA_RESTRICT dst[], \
audioconvert: make separate noise functions So that we can reuse optimized versions in unoptimized noise functions. Do allocation a little different so that we can align everything from the start.
2022-09-29 20:34:15 +02:00
const void * SPA_RESTRICT src[], uint32_t n_samples)
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(copy8d, c);
DEFINE_FUNCTION(copy8, c);
DEFINE_FUNCTION(copy16d, c);
DEFINE_FUNCTION(copy16, c);
DEFINE_FUNCTION(copy24d, c);
DEFINE_FUNCTION(copy24, c);
DEFINE_FUNCTION(copy32d, c);
DEFINE_FUNCTION(copy32, c);
audioconvert: add f64 conversion See #1990
2022-01-10 13:04:31 +01:00
DEFINE_FUNCTION(copy64d, c);
DEFINE_FUNCTION(copy64, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(u8d_to_f32d, c);
DEFINE_FUNCTION(u8_to_f32, c);
DEFINE_FUNCTION(u8_to_f32d, c);
DEFINE_FUNCTION(u8d_to_f32, c);
audioconvert: handle S8/S8P raw audio formats
2021-05-05 08:56:59 +03:00
DEFINE_FUNCTION(s8d_to_f32d, c);
DEFINE_FUNCTION(s8_to_f32, c);
DEFINE_FUNCTION(s8_to_f32d, c);
DEFINE_FUNCTION(s8d_to_f32, c);
audioconvert: add converstion to and from alaw/ulaw
2021-08-17 17:32:25 +02:00
DEFINE_FUNCTION(ulaw_to_f32d, c);
DEFINE_FUNCTION(alaw_to_f32d, c);
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
DEFINE_FUNCTION(u16_to_f32, c);
DEFINE_FUNCTION(u16_to_f32d, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s16d_to_f32d, c);
DEFINE_FUNCTION(s16_to_f32, c);
DEFINE_FUNCTION(s16_to_f32d, c);
audioconvert: add more conversions Add the other endian conversions.
2021-08-17 14:45:00 +02:00
DEFINE_FUNCTION(s16s_to_f32d, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s16d_to_f32, c);
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
DEFINE_FUNCTION(u32_to_f32, c);
DEFINE_FUNCTION(u32_to_f32d, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s32d_to_f32d, c);
DEFINE_FUNCTION(s32_to_f32, c);
DEFINE_FUNCTION(s32_to_f32d, c);
audioconvert: add more conversions Add the other endian conversions.
2021-08-17 14:45:00 +02:00
DEFINE_FUNCTION(s32s_to_f32d, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s32d_to_f32, c);
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
DEFINE_FUNCTION(u24_to_f32, c);
DEFINE_FUNCTION(u24_to_f32d, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s24d_to_f32d, c);
DEFINE_FUNCTION(s24_to_f32, c);
DEFINE_FUNCTION(s24_to_f32d, c);
fmt-ops: add support for s24 with endian conversion
2020-01-27 15:46:25 +01:00
DEFINE_FUNCTION(s24s_to_f32d, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s24d_to_f32, c);
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
DEFINE_FUNCTION(u24_32_to_f32, c);
DEFINE_FUNCTION(u24_32_to_f32d, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s24_32d_to_f32d, c);
DEFINE_FUNCTION(s24_32_to_f32, c);
DEFINE_FUNCTION(s24_32_to_f32d, c);
audioconvert: add more conversions Add the other endian conversions.
2021-08-17 14:45:00 +02:00
DEFINE_FUNCTION(s24_32s_to_f32d, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s24_32d_to_f32, c);
audioconvert: add f64 conversion See #1990
2022-01-10 13:04:31 +01:00
DEFINE_FUNCTION(f64d_to_f32d, c);
DEFINE_FUNCTION(f64_to_f32, c);
DEFINE_FUNCTION(f64_to_f32d, c);
DEFINE_FUNCTION(f64s_to_f32d, c);
DEFINE_FUNCTION(f64d_to_f32, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32d_to_u8d, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_u8d_noise, c);
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
DEFINE_FUNCTION(f32d_to_u8d_shaped, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32_to_u8, c);
DEFINE_FUNCTION(f32_to_u8d, c);
DEFINE_FUNCTION(f32d_to_u8, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_u8_noise, c);
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
DEFINE_FUNCTION(f32d_to_u8_shaped, c);
audioconvert: handle S8/S8P raw audio formats
2021-05-05 08:56:59 +03:00
DEFINE_FUNCTION(f32d_to_s8d, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s8d_noise, c);
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
DEFINE_FUNCTION(f32d_to_s8d_shaped, c);
audioconvert: handle S8/S8P raw audio formats
2021-05-05 08:56:59 +03:00
DEFINE_FUNCTION(f32_to_s8, c);
DEFINE_FUNCTION(f32_to_s8d, c);
DEFINE_FUNCTION(f32d_to_s8, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s8_noise, c);
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
DEFINE_FUNCTION(f32d_to_s8_shaped, c);
audioconvert: add converstion to and from alaw/ulaw
2021-08-17 17:32:25 +02:00
DEFINE_FUNCTION(f32d_to_alaw, c);
DEFINE_FUNCTION(f32d_to_ulaw, c);
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
DEFINE_FUNCTION(f32_to_u16, c);
DEFINE_FUNCTION(f32d_to_u16, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32d_to_s16d, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s16d_noise, c);
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
DEFINE_FUNCTION(f32d_to_s16d_shaped, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32_to_s16, c);
DEFINE_FUNCTION(f32_to_s16d, c);
DEFINE_FUNCTION(f32d_to_s16, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s16_noise, c);
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
DEFINE_FUNCTION(f32d_to_s16_shaped, c);
audioconvert: add more conversions Add the other endian conversions.
2021-08-17 14:45:00 +02:00
DEFINE_FUNCTION(f32d_to_s16s, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s16s_noise, c);
audioconvert: implement noise shaping To compare the results: pw-cat -r test.wav --format=u8 -P '{ dither.method=none }' vs: pw-cat -r test.wav --format=u8 -P '{ dither.method=shaped5 }'
2022-06-29 17:31:32 +02:00
DEFINE_FUNCTION(f32d_to_s16s_shaped, c);
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
DEFINE_FUNCTION(f32_to_u32, c);
DEFINE_FUNCTION(f32d_to_u32, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32d_to_s32d, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s32d_noise, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32_to_s32, c);
DEFINE_FUNCTION(f32_to_s32d, c);
DEFINE_FUNCTION(f32d_to_s32, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s32_noise, c);
audioconvert: add more conversions Add the other endian conversions.
2021-08-17 14:45:00 +02:00
DEFINE_FUNCTION(f32d_to_s32s, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s32s_noise, c);
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
DEFINE_FUNCTION(f32_to_u24, c);
DEFINE_FUNCTION(f32d_to_u24, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32d_to_s24d, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s24d_noise, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32_to_s24, c);
DEFINE_FUNCTION(f32_to_s24d, c);
DEFINE_FUNCTION(f32d_to_s24, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s24_noise, c);
fmt: support f32p -> s24s conversion As used by MAudio Fasttrack Pro See rhbz#916017
2021-01-14 17:58:54 +01:00
DEFINE_FUNCTION(f32d_to_s24s, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s24s_noise, c);
fmtconvert: add support for U16, U24, U32 and U24_32 formats
2021-09-08 09:07:49 -04:00
DEFINE_FUNCTION(f32_to_u24_32, c);
DEFINE_FUNCTION(f32d_to_u24_32, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32d_to_s24_32d, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s24_32d_noise, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32_to_s24_32, c);
DEFINE_FUNCTION(f32_to_s24_32d, c);
DEFINE_FUNCTION(f32d_to_s24_32, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s24_32_noise, c);
audioconvert: add more conversions Add the other endian conversions.
2021-08-17 14:45:00 +02:00
DEFINE_FUNCTION(f32d_to_s24_32s, c);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s24_32s_noise, c);
audioconvert: add f64 conversion See #1990
2022-01-10 13:04:31 +01:00
DEFINE_FUNCTION(f32d_to_f64d, c);
DEFINE_FUNCTION(f32_to_f64, c);
DEFINE_FUNCTION(f32_to_f64d, c);
DEFINE_FUNCTION(f32d_to_f64, c);
DEFINE_FUNCTION(f32d_to_f64s, c);
audioconvert: implement (de)interleave with existing functions
2022-06-30 18:24:05 +02:00
DEFINE_FUNCTION(8_to_8d, c);
DEFINE_FUNCTION(16_to_16d, c);
DEFINE_FUNCTION(24_to_24d, c);
DEFINE_FUNCTION(32_to_32d, c);
audioconvert: implement f64s Add swapping functions for f64s. Fix the awkward interleave/deinterleave names for 32s.
2022-07-11 10:58:51 +02:00
DEFINE_FUNCTION(32s_to_32d, c);
audioconvert: implement (de)interleave with existing functions
2022-06-30 18:24:05 +02:00
DEFINE_FUNCTION(64_to_64d, c);
DEFINE_FUNCTION(64s_to_64sd, c);
DEFINE_FUNCTION(8d_to_8, c);
DEFINE_FUNCTION(16d_to_16, c);
DEFINE_FUNCTION(24d_to_24, c);
DEFINE_FUNCTION(32d_to_32, c);
audioconvert: implement f64s Add swapping functions for f64s. Fix the awkward interleave/deinterleave names for 32s.
2022-07-11 10:58:51 +02:00
DEFINE_FUNCTION(32d_to_32s, c);
audioconvert: implement (de)interleave with existing functions
2022-06-30 18:24:05 +02:00
DEFINE_FUNCTION(64d_to_64, c);
DEFINE_FUNCTION(64sd_to_64s, c);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
fmt-ops: add neon optimizations for format conversion
2020-03-30 10:59:21 -04:00
#if defined(HAVE_NEON)
audioconvert: add stereo deinterleave neon asm This can take some shortcuts and convert twice as many samples in one iteration as the strided stereo deinterleave one.
2021-10-28 11:30:04 +02:00
DEFINE_FUNCTION(s16_to_f32d_2, neon);
fmt-ops: add neon optimizations for format conversion
2020-03-30 10:59:21 -04:00
DEFINE_FUNCTION(s16_to_f32d, neon);
DEFINE_FUNCTION(f32d_to_s16, neon);
#endif
fmt-ops: add R-V V optimizations for f32_s16
2024-09-17 08:58:14 +08:00
#if defined(HAVE_RVV)
fmt-ops: add RVV optimizations for f32d_to_s32
2024-09-25 13:11:08 +08:00
DEFINE_FUNCTION(f32d_to_s32, rvv);
fmt-ops: add R-V V optimizations for f32_s16
2024-09-17 08:58:14 +08:00
DEFINE_FUNCTION(f32_to_s16, rvv);
fmt-ops: add RVV optimizations for f32d_s16d
2024-09-23 10:22:44 +08:00
DEFINE_FUNCTION(f32d_to_s16d, rvv);
fmt-ops: add RVV optimizations for f32d_s16
2024-09-23 10:53:01 +08:00
DEFINE_FUNCTION(f32d_to_s16, rvv);
fmt-ops: add RVV optimizations for s16_to_f32d
2024-09-23 21:03:28 +08:00
DEFINE_FUNCTION(s16_to_f32d, rvv);
fmt-ops: add RVV optimizations for s32_to_f32d
2024-09-29 11:17:18 +08:00
DEFINE_FUNCTION(s32_to_f32d, rvv);
fmt-ops: add R-V V optimizations for f32_s16
2024-09-17 08:58:14 +08:00
#endif
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
#if defined(HAVE_SSE2)
audioconvert: pass state to functions Pass some state to convert and channelmix functions. This makes it possible to select per channel optimized convert functions but also makes it possible to implement noise shaping later. Pass the channelmix matrix and volume in the state. Handle specialized 2 channel s16 -> f32 conversion
2019-03-29 17:39:59 +01:00
DEFINE_FUNCTION(s16_to_f32d_2, sse2);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s16_to_f32d, sse2);
DEFINE_FUNCTION(s24_to_f32d, sse2);
fmt: add sse2 optimized s32->f32d
2019-10-24 11:06:04 +02:00
DEFINE_FUNCTION(s32_to_f32d, sse2);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32d_to_s32, sse2);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s32_noise, sse2);
fmt-ops: add sse2 version of f32 to s16 planar/interleaved
2020-07-03 11:04:39 +02:00
DEFINE_FUNCTION(f32_to_s16, sse2);
fmt-ops: add optimized f32 to s16 stereo conversion
2020-03-16 13:04:21 +01:00
DEFINE_FUNCTION(f32d_to_s16_2, sse2);
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(f32d_to_s16, sse2);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s16_noise, sse2);
fmt-ops: add sse2 version of f32 to s16 planar/interleaved
2020-07-03 11:04:39 +02:00
DEFINE_FUNCTION(f32d_to_s16d, sse2);
audioconvert: improve noise shaping Reorganize things a little so we can add more noise shapers. Add sloped triangular noise. Add wanamaker3 noise shaping.
2022-07-15 12:36:15 +02:00
DEFINE_FUNCTION(f32d_to_s16d_noise, sse2);
audioconvert: implement (de)interleave with existing functions
2022-06-30 18:24:05 +02:00
DEFINE_FUNCTION(32_to_32d, sse2);
audioconvert: implement f64s Add swapping functions for f64s. Fix the awkward interleave/deinterleave names for 32s.
2022-07-11 10:58:51 +02:00
DEFINE_FUNCTION(32s_to_32d, sse2);
audioconvert: implement (de)interleave with existing functions
2022-06-30 18:24:05 +02:00
DEFINE_FUNCTION(32d_to_32, sse2);
audioconvert: implement f64s Add swapping functions for f64s. Fix the awkward interleave/deinterleave names for 32s.
2022-07-11 10:58:51 +02:00
DEFINE_FUNCTION(32d_to_32s, sse2);
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
#endif
#if defined(HAVE_SSSE3)
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s24_to_f32d, ssse3);
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
#endif
#if defined(HAVE_SSE41)
audioconvert: compile c version separately Also compile the c versions in a separate module with their own flags.
2019-03-28 21:07:53 +01:00
DEFINE_FUNCTION(s24_to_f32d, sse41);
fmt-ops: add avx2 optimized version Only one optimized version but the sse2 version are compiled with the avx2 flags so that they get optimized better.
2020-03-16 16:11:29 +01:00
#endif
#if defined(HAVE_AVX2)
DEFINE_FUNCTION(s16_to_f32d_2, avx2);
DEFINE_FUNCTION(s16_to_f32d, avx2);
DEFINE_FUNCTION(s24_to_f32d, avx2);
DEFINE_FUNCTION(s32_to_f32d, avx2);
DEFINE_FUNCTION(f32d_to_s32, avx2);
fmt-ops: flesh out avx optimizations
2020-03-17 17:27:47 +01:00
DEFINE_FUNCTION(f32d_to_s16_4, avx2);
fmt-ops: add avx2 optimized version Only one optimized version but the sse2 version are compiled with the avx2 flags so that they get optimized better.
2020-03-16 16:11:29 +01:00
DEFINE_FUNCTION(f32d_to_s16_2, avx2);
DEFINE_FUNCTION(f32d_to_s16, avx2);
audioconvert: handle more optimizations Compile an optimized library for the given CPU with the right flags, then link it with the main library.
2019-03-27 17:58:48 +01:00
#endif
audioconvert: add volume ops Add volume helpers and use this for the monitor output ports so that we can add volumes to the monitor output signal as well.
2021-02-05 15:08:26 +01:00
#undef DEFINE_FUNCTION
Reference in a new issue Copy permalink