e66470df68
initial commit for examples/dsptest - unit test for Nuttx DSP library Approved-by: GregoryN <gnutt@nuttx.org>
634 lines
14 KiB
C
634 lines
14 KiB
C
/****************************************************************************
|
|
* examples/dsptest/test_misc.c
|
|
*
|
|
* Copyright (C) 2018 Gregory Nutt. All rights reserved.
|
|
* Author: Mateusz Szafoni <raiden00@railab.me>
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
*
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in
|
|
* the documentation and/or other materials provided with the
|
|
* distribution.
|
|
* 3. Neither the name NuttX nor the names of its contributors may be
|
|
* used to endorse or promote products derived from this software
|
|
* without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
|
|
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
|
|
* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
|
|
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
|
|
* OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
|
|
* AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
|
|
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
*
|
|
****************************************************************************/
|
|
|
|
/****************************************************************************
|
|
* Included Files
|
|
****************************************************************************/
|
|
|
|
#include "dsptest.h"
|
|
|
|
/****************************************************************************
|
|
* Pre-processor Definitions
|
|
****************************************************************************/
|
|
|
|
/* Set float precision for this module */
|
|
|
|
#undef UNITY_FLOAT_PRECISION
|
|
#define UNITY_FLOAT_PRECISION (0.0001f)
|
|
|
|
#define TEST_SINCOS_ANGLE_STEP 0.01
|
|
#define TEST_SINCOS2_ANGLE_STEP 0.01
|
|
#define TEST_ATAN2_XY_STEP 0.01
|
|
|
|
/* Angle sin/cos depends on libdsp precision */
|
|
|
|
#if CONFIG_LIBDSP_PRECISION == 1
|
|
# define ANGLE_SIN(val) fast_sin2(val)
|
|
# define ANGLE_COS(val) fast_cos2(val)
|
|
#elif CONFIG_LIBDSP_PRECISION == 2
|
|
# define ANGLE_SIN(val) sin(val)
|
|
# define ANGLE_COS(val) cos(val)
|
|
#else
|
|
# define ANGLE_SIN(val) fast_sin(val)
|
|
# define ANGLE_COS(val) fast_cos(val)
|
|
#endif
|
|
|
|
/****************************************************************************
|
|
* Private Types
|
|
****************************************************************************/
|
|
|
|
/****************************************************************************
|
|
* Private Function Protototypes
|
|
****************************************************************************/
|
|
|
|
/****************************************************************************
|
|
* Private Data
|
|
****************************************************************************/
|
|
|
|
/****************************************************************************
|
|
* Private Functions
|
|
****************************************************************************/
|
|
|
|
/* Test float value saturation */
|
|
|
|
static void test_f_saturate(void)
|
|
{
|
|
float val = 0.0;
|
|
|
|
/* Value in the range */
|
|
|
|
val = 0.9;
|
|
f_saturate(&val, 0.0, 1.0);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.9, val);
|
|
|
|
/* Upper limit */
|
|
|
|
val = 1.2;
|
|
f_saturate(&val, 0.0, 1.0);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(1.0, val);
|
|
|
|
/* Lower limit */
|
|
|
|
val = 0.0;
|
|
f_saturate(&val, 0.1, 1.0);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.1, val);
|
|
}
|
|
|
|
/* Test 2D vector magnitude */
|
|
|
|
static void test_vector2d_mag(void)
|
|
{
|
|
float mag = 0.0;
|
|
|
|
/* mag([0.0, 0.0]) = 0.0 */
|
|
|
|
mag = vector2d_mag(0.0, 0.0);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, mag);
|
|
|
|
/* mag([1.0, 1.0]) = 1.4142 */
|
|
|
|
mag = vector2d_mag(1.0, 1.0);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(1.4142, mag);
|
|
|
|
/* mag([-1.0, 1.0]) = 1.4142 */
|
|
|
|
mag = vector2d_mag(-1.0, 1.0);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(1.4142, mag);
|
|
}
|
|
|
|
/* Test 2D vector saturation */
|
|
|
|
static void test_vector2d_saturate(void)
|
|
{
|
|
float x = 0.0;
|
|
float y = 0.0;
|
|
float mag = 0.0;
|
|
float mag_ref = 0.0;
|
|
|
|
/* Vector magnitude 0.0 */
|
|
|
|
x = 0.0;
|
|
y = 0.0;
|
|
mag_ref = 1.0;
|
|
vector2d_saturate(&x, &y, mag_ref);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, x);
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, y);
|
|
|
|
/* Vector magnitude 0.0, saturation 0.0 */
|
|
|
|
x = 0.0;
|
|
y = 0.0;
|
|
mag_ref = 0.0;
|
|
vector2d_saturate(&x, &y, mag_ref);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(mag_ref, x);
|
|
TEST_ASSERT_EQUAL_FLOAT(mag_ref, y);
|
|
|
|
/* Vector magnitude 1.4142, saturation 0.0 */
|
|
|
|
x = 1.0;
|
|
y = 1.0;
|
|
mag_ref = 0.0;
|
|
vector2d_saturate(&x, &y, mag_ref);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, x);
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, y);
|
|
|
|
/* Vector magnitude 1.4142, saturation 3.0 */
|
|
|
|
x = 1.0;
|
|
y = 1.0;
|
|
mag_ref = 3.0;
|
|
vector2d_saturate(&x, &y, mag_ref);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(1.0, x);
|
|
TEST_ASSERT_EQUAL_FLOAT(1.0, y);
|
|
|
|
/* Vector magnitude 1.4142, saturation 1.0 - truncate */
|
|
|
|
x = 1.0;
|
|
y = 1.0;
|
|
mag_ref = 1.0;
|
|
vector2d_saturate(&x, &y, mag_ref);
|
|
mag = vector2d_mag(x, y);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(mag_ref, mag);
|
|
}
|
|
|
|
/* Test dq vector magnitude */
|
|
|
|
static void test_dq_mag(void)
|
|
{
|
|
TEST_IGNORE_MESSAGE("test_dq_mag not implemented");
|
|
}
|
|
|
|
/* Test dq vector saturation */
|
|
|
|
static void test_dq_saturate(void)
|
|
{
|
|
dq_frame_t dq;
|
|
float mag_ref = 0.0;
|
|
float mag = 0.0;
|
|
|
|
/* Vector magnitude 0.0 */
|
|
|
|
dq.d = 0.0;
|
|
dq.q = 0.0;
|
|
mag_ref = 1.0;
|
|
dq_saturate(&dq, mag_ref);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, dq.d);
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, dq.q);
|
|
|
|
/* Vector magnitude 0.0, saturation 0.0 */
|
|
|
|
dq.d = 0.0;
|
|
dq.q = 0.0;
|
|
mag_ref = 0.0;
|
|
dq_saturate(&dq, mag_ref);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(mag_ref, dq.d);
|
|
TEST_ASSERT_EQUAL_FLOAT(mag_ref, dq.q);
|
|
|
|
/* Vector magnitude 1.4142, saturation 0.0 */
|
|
|
|
dq.d = 1.0;
|
|
dq.q = 1.0;
|
|
mag_ref = 0.0;
|
|
dq_saturate(&dq, mag_ref);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(mag_ref, dq.d);
|
|
TEST_ASSERT_EQUAL_FLOAT(mag_ref, dq.q);
|
|
|
|
/* Vector magnitude 1.4142, saturation 3.0 */
|
|
|
|
dq.d = 1.0;
|
|
dq.q = 1.0;
|
|
mag_ref = 3.0;
|
|
dq_saturate(&dq, mag_ref);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(1.0, dq.d);
|
|
TEST_ASSERT_EQUAL_FLOAT(1.0, dq.q);
|
|
|
|
/* Vector magnitude 1.4142, saturation 1.0 - truncate */
|
|
|
|
dq.d = 1.0;
|
|
dq.q = 1.0;
|
|
mag_ref = 1.0;
|
|
dq_saturate(&dq, mag_ref);
|
|
mag = dq_mag(&dq);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(mag_ref, mag);
|
|
}
|
|
|
|
/* Test fast sine */
|
|
|
|
static void test_fast_sin(void)
|
|
{
|
|
float s_ref = 0.0;
|
|
float angle = 0.0;
|
|
float s = 0.0;
|
|
|
|
/* Compare with LIBC sine */
|
|
|
|
for(angle = 0.0; angle < 2*M_PI_F; angle += TEST_SINCOS_ANGLE_STEP)
|
|
{
|
|
s_ref = sinf(angle);
|
|
s = fast_sin(angle);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_SINCOS_DELTA, s_ref, s);
|
|
}
|
|
}
|
|
|
|
/* Test fast cosine */
|
|
|
|
static void test_fast_cos(void)
|
|
{
|
|
float c_ref = 0.0;
|
|
float angle = 0.0;
|
|
float c = 0.0;
|
|
|
|
/* Compare with LIBC cosine */
|
|
|
|
for(angle = 0.0; angle < 2*M_PI_F; angle += TEST_SINCOS_ANGLE_STEP)
|
|
{
|
|
c_ref = cosf(angle);
|
|
c = fast_cos(angle);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_SINCOS_DELTA, c_ref, c);
|
|
}
|
|
}
|
|
|
|
/* Test fast sine (better accuracy) */
|
|
|
|
static void test_fast_sin2(void)
|
|
{
|
|
float s_ref = 0.0;
|
|
float angle = 0.0;
|
|
float s = 0.0;
|
|
|
|
/* Compare with LIBC sine */
|
|
|
|
for(angle = 0.0; angle < 2*M_PI_F; angle += TEST_SINCOS2_ANGLE_STEP)
|
|
{
|
|
s_ref = sinf(angle);
|
|
s = fast_sin2(angle);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_SINCOS2_DELTA, s_ref, s);
|
|
}
|
|
}
|
|
|
|
/* Test fast cosine (better accuracy) */
|
|
|
|
static void test_fast_cos2(void)
|
|
{
|
|
float c_ref = 0.0;
|
|
float angle = 0.0;
|
|
float c = 0.0;
|
|
|
|
/* Compare with LIBC cosine */
|
|
|
|
for(angle = 0.0; angle < 2*M_PI_F; angle += TEST_SINCOS2_ANGLE_STEP)
|
|
{
|
|
c_ref = cosf(angle);
|
|
c = fast_cos2(angle);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_SINCOS2_DELTA, c_ref, c);
|
|
}
|
|
}
|
|
|
|
/* Test fast atan2 */
|
|
|
|
static void test_fast_atan2(void)
|
|
{
|
|
float angle_ref = 0.0;
|
|
float angle = 0.0;
|
|
float x = 0.0;
|
|
float y = 0.0;
|
|
|
|
/* atan2(0, 0) - special case when atan2 is not defined */
|
|
|
|
angle = fast_atan2(y, x);
|
|
|
|
/* Expect non inf and non nan */
|
|
|
|
TEST_ASSERT_FLOAT_IS_DETERMINATE(angle);
|
|
|
|
/* Compare with LIBC atan2 */
|
|
|
|
for (x = TEST_ATAN2_XY_STEP; x < M_PI_F; x += TEST_ATAN2_XY_STEP)
|
|
{
|
|
for (y = TEST_ATAN2_XY_STEP; y < M_PI_F; y += TEST_ATAN2_XY_STEP)
|
|
{
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
}
|
|
}
|
|
|
|
for (x = -TEST_ATAN2_XY_STEP; x > -M_PI_F; x -= TEST_ATAN2_XY_STEP)
|
|
{
|
|
for (y = -TEST_ATAN2_XY_STEP; y > -M_PI_F; y -= TEST_ATAN2_XY_STEP)
|
|
{
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
}
|
|
}
|
|
|
|
for (x = TEST_ATAN2_XY_STEP; x < M_PI_F; x += TEST_ATAN2_XY_STEP)
|
|
{
|
|
for (y = -TEST_ATAN2_XY_STEP; y > -M_PI_F; y -= TEST_ATAN2_XY_STEP)
|
|
{
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
}
|
|
}
|
|
|
|
for (x = -TEST_ATAN2_XY_STEP; x > -M_PI_F; x -= TEST_ATAN2_XY_STEP)
|
|
{
|
|
for (y = TEST_ATAN2_XY_STEP; y < M_PI_F; y += TEST_ATAN2_XY_STEP)
|
|
{
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
}
|
|
}
|
|
|
|
/* Test some big numbers */
|
|
|
|
x = 1000000.0;
|
|
y = 2.0;
|
|
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
|
|
x = 2.0;
|
|
y = 1000000.0;
|
|
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
|
|
x = 1000000.0;
|
|
y = 1000000.0;
|
|
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
|
|
x = -1000000.0;
|
|
y = 1000000.0;
|
|
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
|
|
x = 1000000.0;
|
|
y = -1000000.0;
|
|
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
|
|
x = -1000000.0;
|
|
y = -1000000.0;
|
|
|
|
angle_ref = atan2f(y, x);
|
|
angle = fast_atan2(y, x);
|
|
|
|
TEST_ASSERT_FLOAT_WITHIN(TEST_ATAN2_DELTA, angle_ref, angle);
|
|
}
|
|
|
|
/* Test angle normalization */
|
|
|
|
static void test_angle_norm(void)
|
|
{
|
|
float angle = 0.0;
|
|
float per = 0.0;
|
|
float bottom = 0.0;
|
|
float top = 0.0;
|
|
|
|
/* range = (0.0, 2PI) */
|
|
|
|
per = 2 * M_PI_F;
|
|
bottom = 0.0;
|
|
top = 2 * M_PI_F;
|
|
|
|
/* in range */
|
|
|
|
angle = 0.0;
|
|
angle_norm(&angle, per, bottom, top);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, angle);
|
|
|
|
/* in range */
|
|
|
|
angle = 1.0;
|
|
angle_norm(&angle, per, bottom, top);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(1.0, angle);
|
|
|
|
/* wrap to 0.2 */
|
|
|
|
angle = 2 * M_PI_F + 0.2;
|
|
angle_norm(&angle, per, bottom, top);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.2, angle);
|
|
|
|
/* wrap to 0.2 */
|
|
|
|
angle = -2 * M_PI_F + 0.2;
|
|
angle_norm(&angle, per, bottom, top);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.2, angle);
|
|
}
|
|
|
|
/* Test angle normalization with 2*PI period length */
|
|
|
|
static void test_angle_norm_2pi(void)
|
|
{
|
|
float angle = 0.0;
|
|
float bottom = 0.0;
|
|
float top = 0.0;
|
|
|
|
/* range = (0.0, 2PI) */
|
|
|
|
bottom = 0.0;
|
|
top = 2 * M_PI_F;
|
|
|
|
/* in range */
|
|
|
|
angle = 0.0;
|
|
angle_norm_2pi(&angle, bottom, top);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.0, angle);
|
|
|
|
/* in range */
|
|
|
|
angle = 1.0;
|
|
angle_norm_2pi(&angle, bottom, top);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(1.0, angle);
|
|
|
|
/* wrap to 0.2 */
|
|
|
|
angle = 2 * M_PI_F + 0.2;
|
|
angle_norm_2pi(&angle, bottom, top);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.2, angle);
|
|
|
|
/* wrap to 0.2 */
|
|
|
|
angle = -2 * M_PI_F + 0.2;
|
|
angle_norm_2pi(&angle, bottom, top);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(0.2, angle);
|
|
}
|
|
|
|
/* Test phase angle update */
|
|
|
|
static void test_phase_angle_update(void)
|
|
{
|
|
struct phase_angle_s angle;
|
|
float val = 0.0;
|
|
float s = 0.0;
|
|
float c = 0.0;
|
|
|
|
/* angle = 0.0 */
|
|
|
|
val = 0.0;
|
|
s = ANGLE_SIN(val);
|
|
c = ANGLE_COS(val);
|
|
phase_angle_update(&angle, val);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(val, angle.angle);
|
|
TEST_ASSERT_EQUAL_FLOAT(s, angle.sin);
|
|
TEST_ASSERT_EQUAL_FLOAT(c, angle.cos);
|
|
|
|
/* angle = 1.5 */
|
|
|
|
val = 1.5;
|
|
s = ANGLE_SIN(val);
|
|
c = ANGLE_COS(val);
|
|
phase_angle_update(&angle, val);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(val, angle.angle);
|
|
TEST_ASSERT_EQUAL_FLOAT(s, angle.sin);
|
|
TEST_ASSERT_EQUAL_FLOAT(c, angle.cos);
|
|
|
|
/* angle = 8, should be normalize to (0.0, 2PI) range */
|
|
|
|
val = 8;
|
|
s = ANGLE_SIN(val);
|
|
c = ANGLE_COS(val);
|
|
phase_angle_update(&angle, val);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(val-2*M_PI_F, angle.angle);
|
|
TEST_ASSERT_EQUAL_FLOAT(s, angle.sin);
|
|
TEST_ASSERT_EQUAL_FLOAT(c, angle.cos);
|
|
|
|
/* angle = -1.5, should be normalize to (0.0, 2PI) range */
|
|
|
|
val = -1.5;
|
|
s = ANGLE_SIN(val);
|
|
c = ANGLE_COS(val);
|
|
phase_angle_update(&angle, val);
|
|
|
|
TEST_ASSERT_EQUAL_FLOAT(val+2*M_PI_F, angle.angle);
|
|
TEST_ASSERT_EQUAL_FLOAT(s, angle.sin);
|
|
TEST_ASSERT_EQUAL_FLOAT(c, angle.cos);
|
|
}
|
|
|
|
/****************************************************************************
|
|
* Public Functions
|
|
****************************************************************************/
|
|
|
|
/****************************************************************************
|
|
* Name: test_misc
|
|
****************************************************************************/
|
|
|
|
void test_misc(void)
|
|
{
|
|
UNITY_BEGIN();
|
|
|
|
TEST_SEPARATOR();
|
|
|
|
/* Test helper funcions */
|
|
|
|
RUN_TEST(test_f_saturate);
|
|
|
|
/* Test vector functions */
|
|
|
|
RUN_TEST(test_vector2d_mag);
|
|
RUN_TEST(test_vector2d_saturate);
|
|
RUN_TEST(test_dq_mag);
|
|
RUN_TEST(test_dq_saturate);
|
|
|
|
/* Test fast trigonometric functions */
|
|
|
|
RUN_TEST(test_fast_sin);
|
|
RUN_TEST(test_fast_cos);
|
|
RUN_TEST(test_fast_sin2);
|
|
RUN_TEST(test_fast_cos2);
|
|
RUN_TEST(test_fast_atan2);
|
|
|
|
/* Test angle functions */
|
|
|
|
RUN_TEST(test_angle_norm);
|
|
RUN_TEST(test_angle_norm_2pi);
|
|
RUN_TEST(test_phase_angle_update);
|
|
|
|
UNITY_END();
|
|
}
|