Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Audio: DRC: Change DRC to use lookup table based sine function #8491

Merged
merged 3 commits into from
Jan 11, 2024
Merged

Audio: DRC: Change DRC to use lookup table based sine function #8491

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
88a15b7
Math: Restore fast look-up table based sine function
singalsu Nov 17, 2023
4ed988b
Test: Cmocka: Add test case for lookup table sine function
singalsu Jan 9, 2024
3d1d453
Audio: DRC: Change DRC to use lookup table based sine function
singalsu Nov 17, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions src/audio/drc/Kconfig
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
config COMP_DRC
bool "Dynamic Range Compressor component"
select CORDIC_FIXED
select MATH_LUT_SINE_FIXED
select NUMBERS_NORM
select MATH_EXP
select COMP_BLOB
Expand Down
7 changes: 3 additions & 4 deletions src/audio/drc/drc_math.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,7 @@
#include <stdint.h>
#include <sof/audio/format.h>
#include <sof/math/numbers.h>
#include <sof/math/lut_trig.h>
#include <sof/math/trig.h>

#include "drc_plat_conf.h"
Expand Down Expand Up @@ -60,9 +61,8 @@ static inline int32_t drc_sin_fixed(int32_t x)
{
const int32_t lshift = drc_get_lshift(30, 30, 28);
int32_t denorm_x = drc_mult_lshift(x, PI_OVER_TWO_Q30, lshift);
int32_t sin_val = sin_fixed_16b(denorm_x);

return sin_val << 16;
return sofm_lut_sin_fixed_16b(denorm_x) << 16;
}

#ifdef DRC_USE_CORDIC_ASIN
Expand All @@ -88,9 +88,8 @@ static inline int32_t drc_asin_fixed(int32_t x)
static inline int32_t drc_sin_fixed(int32_t x)
{
const int32_t PI_OVER_TWO = Q_CONVERT_FLOAT(1.57079632679489661923, 30);
int32_t sin_val = sin_fixed_16b(Q_MULTSR_32X32((int64_t)x, PI_OVER_TWO, 30, 30, 28));

return sin_val << 16;
return sofm_lut_sin_fixed_16b(Q_MULTSR_32X32((int64_t)x, PI_OVER_TWO, 30, 30, 28)) << 16;
}

#ifdef DRC_USE_CORDIC_ASIN
Expand Down
15 changes: 15 additions & 0 deletions src/include/sof/math/lut_trig.h
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,15 @@
/* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright(c) 2016-2023 Intel Corporation. All rights reserved.
*
* Author: Seppo Ingalsuo <[email protected]>
*/

#ifndef __SOF_MATH_LUT_TRIG_H__
#define __SOF_MATH_LUT_TRIG_H__

#include <stdint.h>

int16_t sofm_lut_sin_fixed_16b(int32_t w); /* Input is Q4.28, output is Q1.15 */

#endif /* __SOF_MATH_LUT_TRIG_H__ */
2 changes: 2 additions & 0 deletions src/math/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,8 @@ if(CONFIG_CORDIC_FIXED)
add_local_sources(sof trig.c)
endif()

add_local_sources_ifdef(CONFIG_MATH_LUT_SINE_FIXED sof lut_trig.c)

add_local_sources_ifdef(CONFIG_SQRT_FIXED sof sqrt_int16.c)

add_local_sources_ifdef(CONFIG_MATH_EXP sof exp_fcn.c exp_fcn_hifi.c)
Expand Down
11 changes: 11 additions & 0 deletions src/math/Kconfig
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,17 @@ config CORDIC_FIXED
Select this to enable sin(), cos(), asin(), acos(),
and cexp() functions as 16 bit and 32 bit versions.

config MATH_LUT_SINE_FIXED
bool "Lookup table based sine function"
default n
btian1 marked this conversation as resolved.
Show resolved Hide resolved
help
Select this to enable sofm_lut_sin_fixed_16b() function. The
calculation is using 1/4 wave lookup and interpolation. Use
this for fast sine calculation in hot code parts. Sine
calculation in component initialization should use the cordic
version. This option consumes 1026 bytes .bss RAM for the
lookup table.

config POWER_FIXED
bool "Power function"
default n
Expand Down
108 changes: 108 additions & 0 deletions src/math/lut_trig.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,108 @@
// SPDX-License-Identifier: BSD-3-Clause
//
// Copyright(c) 2016-2024 Intel Corporation. All rights reserved.
//
// Author: Seppo Ingalsuo <[email protected]>

#include <sof/audio/format.h>
#include <sof/math/lut_trig.h>
#include <stdint.h>

#define SOFM_LUT_SINE_C_Q20 341782638 /* 2 * SINE_NQUART / pi in Q12.20 */
#define SOFM_LUT_SINE_NQUART 512 /* Must be 2^N */
#define SOFM_LUT_SINE_SIZE (SOFM_LUT_SINE_NQUART + 1)

/* Sine values 0 to pi/2, calculated with Octave
* w = linspace(0, pi/2, 513);
* s = 2^16;
* x = min(round(s * sin(w)), s - 1);
*/

static const uint16_t sofm_lut_sine_table_s16[SOFM_LUT_SINE_SIZE] = {
0, 201, 402, 603, 804, 1005, 1206, 1407, 1608, 1809, 2010,
2211, 2412, 2613, 2814, 3015, 3216, 3417, 3617, 3818, 4019, 4219,
4420, 4621, 4821, 5022, 5222, 5422, 5623, 5823, 6023, 6224, 6424,
6624, 6824, 7024, 7224, 7423, 7623, 7823, 8022, 8222, 8421, 8621,
8820, 9019, 9218, 9417, 9616, 9815, 10014, 10212, 10411, 10609, 10808,
11006, 11204, 11402, 11600, 11798, 11996, 12193, 12391, 12588, 12785, 12983,
13180, 13376, 13573, 13770, 13966, 14163, 14359, 14555, 14751, 14947, 15143,
15338, 15534, 15729, 15924, 16119, 16314, 16508, 16703, 16897, 17091, 17285,
17479, 17673, 17867, 18060, 18253, 18446, 18639, 18832, 19024, 19216, 19409,
19600, 19792, 19984, 20175, 20366, 20557, 20748, 20939, 21129, 21320, 21510,
21699, 21889, 22078, 22268, 22457, 22645, 22834, 23022, 23210, 23398, 23586,
23774, 23961, 24148, 24335, 24521, 24708, 24894, 25080, 25265, 25451, 25636,
25821, 26005, 26190, 26374, 26558, 26742, 26925, 27108, 27291, 27474, 27656,
27838, 28020, 28202, 28383, 28564, 28745, 28926, 29106, 29286, 29466, 29645,
29824, 30003, 30182, 30360, 30538, 30716, 30893, 31071, 31248, 31424, 31600,
31776, 31952, 32127, 32303, 32477, 32652, 32826, 33000, 33173, 33347, 33520,
33692, 33865, 34037, 34208, 34380, 34551, 34721, 34892, 35062, 35231, 35401,
35570, 35738, 35907, 36075, 36243, 36410, 36577, 36744, 36910, 37076, 37241,
37407, 37572, 37736, 37900, 38064, 38228, 38391, 38554, 38716, 38878, 39040,
39201, 39362, 39523, 39683, 39843, 40002, 40161, 40320, 40478, 40636, 40794,
40951, 41108, 41264, 41420, 41576, 41731, 41886, 42040, 42194, 42348, 42501,
42654, 42806, 42958, 43110, 43261, 43412, 43562, 43713, 43862, 44011, 44160,
44308, 44456, 44604, 44751, 44898, 45044, 45190, 45335, 45480, 45625, 45769,
45912, 46056, 46199, 46341, 46483, 46624, 46765, 46906, 47046, 47186, 47325,
47464, 47603, 47741, 47878, 48015, 48152, 48288, 48424, 48559, 48694, 48828,
48962, 49095, 49228, 49361, 49493, 49624, 49756, 49886, 50016, 50146, 50275,
50404, 50532, 50660, 50787, 50914, 51041, 51166, 51292, 51417, 51541, 51665,
51789, 51911, 52034, 52156, 52277, 52398, 52519, 52639, 52759, 52878, 52996,
53114, 53232, 53349, 53465, 53581, 53697, 53812, 53926, 54040, 54154, 54267,
54379, 54491, 54603, 54714, 54824, 54934, 55043, 55152, 55260, 55368, 55476,
55582, 55689, 55794, 55900, 56004, 56108, 56212, 56315, 56418, 56520, 56621,
56722, 56823, 56923, 57022, 57121, 57219, 57317, 57414, 57511, 57607, 57703,
57798, 57892, 57986, 58079, 58172, 58265, 58356, 58448, 58538, 58628, 58718,
58807, 58896, 58983, 59071, 59158, 59244, 59330, 59415, 59499, 59583, 59667,
59750, 59832, 59914, 59995, 60075, 60156, 60235, 60314, 60392, 60470, 60547,
60624, 60700, 60776, 60851, 60925, 60999, 61072, 61145, 61217, 61288, 61359,
61429, 61499, 61568, 61637, 61705, 61772, 61839, 61906, 61971, 62036, 62101,
62165, 62228, 62291, 62353, 62415, 62476, 62536, 62596, 62655, 62714, 62772,
62830, 62886, 62943, 62998, 63054, 63108, 63162, 63215, 63268, 63320, 63372,
63423, 63473, 63523, 63572, 63621, 63668, 63716, 63763, 63809, 63854, 63899,
63944, 63987, 64031, 64073, 64115, 64156, 64197, 64237, 64277, 64316, 64354,
64392, 64429, 64465, 64501, 64536, 64571, 64605, 64639, 64672, 64704, 64735,
64766, 64797, 64827, 64856, 64884, 64912, 64940, 64967, 64993, 65018, 65043,
65067, 65091, 65114, 65137, 65159, 65180, 65200, 65220, 65240, 65259, 65277,
65294, 65311, 65328, 65343, 65358, 65373, 65387, 65400, 65413, 65425, 65436,
65447, 65457, 65467, 65476, 65484, 65492, 65499, 65505, 65511, 65516, 65521,
65525, 65528, 65531, 65533, 65535, 65535, 65535
};

/* Sine lookup table read */
static inline int32_t sofm_sine_lookup_16b(int idx)
{
uint16_t s;
lyakh marked this conversation as resolved.
Show resolved Hide resolved
int i1;

i1 = idx & (2 * SOFM_LUT_SINE_NQUART - 1);
if (i1 > SOFM_LUT_SINE_NQUART)
i1 = 2 * SOFM_LUT_SINE_NQUART - i1;

s = sofm_lut_sine_table_s16[i1];
lyakh marked this conversation as resolved.
Show resolved Hide resolved
if (idx > 2 * SOFM_LUT_SINE_NQUART)
return -((int32_t)s);

return (int32_t)s;
}

/* Compute fixed point sine with table lookup and interpolation */
int16_t sofm_lut_sin_fixed_16b(int32_t w)
{
int64_t idx;
int32_t sine;
int32_t frac;
int32_t delta;
int32_t s0;
int32_t s1;
int64_t idx_tmp;

/* Q4.28 x Q12.20 -> Q16.48 --> Q16.31*/
idx_tmp = ((int64_t)w * SOFM_LUT_SINE_C_Q20) >> 17;
idx = (idx_tmp >> 31); /* Shift to Q0 */
frac = (int32_t)(idx_tmp - (idx << 31)); /* Get fraction Q1.31*/
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

that seems to boil down to

idx_tmp - (idx_tmp & 0xffffffff80000000ULL) ==
idx_tmp & 0x7fffffff

would the compiler optimise that out by itself?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I was thinking that but it looks awkward in arithmetic that is not bit-banging to HW registers etc. The shifts have association to Qx.y format. But if that gives MCPS advantage I can change, and comment what happens, I'll try.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

With this modification 69.976 to 69.925 MCPS, not worth it I think, because of bit-and awkwardness here. I think our perf measurement works down to 0.1 MCPS level, below that it's probably noise.

s0 = sofm_sine_lookup_16b(idx); /* Q1.16 */
s1 = sofm_sine_lookup_16b(idx + 1); /* Q1.16 */
delta = s1 - s0; /* Q1.16 */
sine = s0 + q_mults_32x32(frac, delta, Q_SHIFT_BITS_64(31, 16, 16)); /* Q1.16 */
return sat_int16((sine + 1) >> 1); /* Round to Q1.15 */
}
4 changes: 4 additions & 0 deletions test/cmocka/src/math/trig/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,3 +40,7 @@ cmocka_test(acos_16b_fixed
${PROJECT_SOURCE_DIR}/src/math/trig.c
)

cmocka_test(lut_sin_16b_fixed
lut_sin_16b_fixed.c
${PROJECT_SOURCE_DIR}/src/math/lut_trig.c
)
54 changes: 54 additions & 0 deletions test/cmocka/src/math/trig/lut_sin_16b_fixed.c
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,54 @@
// SPDX-License-Identifier: BSD-3-Clause
//
// Copyright(c) 2024 Intel Corporation. All rights reserved.
//
// Author: Slawomir Blauciak <[email protected]>
// Author: Seppo Ingalsuo <[email protected]>

#include <stdio.h>
#include <stdint.h>
#include <stdarg.h>
#include <stddef.h>
#include <setjmp.h>
#include <math.h>
#include <cmocka.h>

#include <sof/audio/format.h>
#include <sof/math/lut_trig.h>

#include "trig_tables.h"

#define CMP_TOLERANCE 3.1e-5
#define _M_PI 3.14159265358979323846 /* pi */

static void test_math_trig_lut_sin_fixed(void **state)
{
(void)state;

int theta;

for (theta = 0; theta < 360; ++theta) {
double rad = _M_PI / 180.0 * theta;
int32_t rad_q28 = Q_CONVERT_FLOAT(rad, 28);
float r = Q_CONVERT_QTOF(sofm_lut_sin_fixed_16b(rad_q28), 15);
float diff = fabsf(sin_ref_table[theta] - r);

if (diff > CMP_TOLERANCE) {
printf("%s: diff for %d deg = %g\n", __func__,
theta, diff);
}

assert_true(diff <= CMP_TOLERANCE);
}
}

int main(void)
{
const struct CMUnitTest tests[] = {
cmocka_unit_test(test_math_trig_lut_sin_fixed)
};

cmocka_set_message_output(CM_OUTPUT_TAP);

return cmocka_run_group_tests(tests, NULL, NULL);
}
5 changes: 5 additions & 0 deletions zephyr/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -416,6 +416,11 @@ zephyr_library_sources(
lib.c
)

# Optional math utility
zephyr_library_sources_ifdef(CONFIG_MATH_LUT_SINE_FIXED
${SOF_MATH_PATH}/lut_trig.c
)

# SOF module interface functions
add_subdirectory(../src/module module_unused_install/)

Expand Down
Loading