blob: f20a4bbea998e08711d312436536d596a023362e [file] [log] [blame]
/*
* Copyright 2018 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* Test FlowGraph
*
* This file also tests a few different conversion techniques because
* sometimes that have caused compiler bugs.
*/
#include <iostream>
#include <gtest/gtest.h>
#include <aaudio/AAudio.h>
#include "client/AAudioFlowGraph.h"
#include "flowgraph/ClipToRange.h"
#include "flowgraph/Limiter.h"
#include "flowgraph/MonoBlend.h"
#include "flowgraph/MonoToMultiConverter.h"
#include "flowgraph/RampLinear.h"
#include "flowgraph/SinkFloat.h"
#include "flowgraph/SinkI16.h"
#include "flowgraph/SinkI24.h"
#include "flowgraph/SinkI32.h"
#include "flowgraph/SinkI8_24.h"
#include "flowgraph/SourceFloat.h"
#include "flowgraph/SourceI16.h"
#include "flowgraph/SourceI24.h"
#include "flowgraph/SourceI8_24.h"
#include "flowgraph/resampler/IntegerRatio.h"
using namespace FLOWGRAPH_OUTER_NAMESPACE::flowgraph;
using namespace RESAMPLER_OUTER_NAMESPACE::resampler;
using TestFlowgraphResamplerParams = std::tuple<int32_t, int32_t, MultiChannelResampler::Quality>;
enum {
PARAM_SOURCE_SAMPLE_RATE = 0,
PARAM_SINK_SAMPLE_RATE,
PARAM_RESAMPLER_QUALITY
};
constexpr int kInt24Min = 0xff800000;
constexpr int kInt24Max = 0x007fffff;
constexpr int kBytesPerI24Packed = 3;
constexpr int kNumSamples = 8;
constexpr std::array<float, kNumSamples> kInputFloat = {
1.0f, 0.5f, -0.25f, -1.0f,
0.0f, 53.9f, -87.2f, -1.02f};
// Corresponding PCM values as integers.
constexpr std::array<int16_t, kNumSamples> kExpectedI16 = {
INT16_MAX, 1 << 14, INT16_MIN / 4, INT16_MIN,
0, INT16_MAX, INT16_MIN, INT16_MIN};
constexpr std::array<int32_t, kNumSamples> kExpectedI32 = {
INT32_MAX, 1 << 30, INT32_MIN / 4, INT32_MIN,
0, INT32_MAX, INT32_MIN, INT32_MIN};
constexpr std::array<int32_t, kNumSamples> kExpectedI8_24 = {
kInt24Max, 1 << 22, kInt24Min / 4, kInt24Min,
0, kInt24Max, kInt24Min, kInt24Min};
// =================================== FLOAT to I16 ==============
// Simple test that tries to reproduce a Clang compiler bug.
__attribute__((noinline))
void local_convert_float_to_int16(const float *input,
int16_t *output,
int count) {
for (int i = 0; i < count; i++) {
int32_t n = (int32_t) (*input++ * 32768.0f);
*output++ = std::min(INT16_MAX, std::max(INT16_MIN, n)); // clip
}
}
TEST(test_flowgraph, local_convert_float_int16) {
std::array<int16_t, kNumSamples> output;
// Do it inline, which will probably work even with the buggy compiler.
// This validates the expected data.
const float *in = kInputFloat.data();
int16_t *out = output.data();
output.fill(777);
for (int i = 0; i < kNumSamples; i++) {
int32_t n = (int32_t) (*in++ * 32768.0f);
*out++ = std::min(INT16_MAX, std::max(INT16_MIN, n)); // clip
}
for (int i = 0; i < kNumSamples; i++) {
EXPECT_EQ(kExpectedI16.at(i), output.at(i)) << ", i = " << i;
}
// Convert audio signal using the function.
output.fill(777);
local_convert_float_to_int16(kInputFloat.data(), output.data(), kNumSamples);
for (int i = 0; i < kNumSamples; i++) {
EXPECT_EQ(kExpectedI16.at(i), output.at(i)) << ", i = " << i;
}
}
TEST(test_flowgraph, module_sinki16) {
static constexpr int kNumSamples = 8;
std::array<int16_t, kNumSamples + 10> output; // larger than input
SourceFloat sourceFloat{1};
SinkI16 sinkI16{1};
sourceFloat.setData(kInputFloat.data(), kNumSamples);
sourceFloat.output.connect(&sinkI16.input);
output.fill(777);
int32_t numRead = sinkI16.read(output.data(), output.size());
ASSERT_EQ(kNumSamples, numRead);
for (int i = 0; i < numRead; i++) {
EXPECT_EQ(kExpectedI16.at(i), output.at(i)) << ", i = " << i;
}
}
// =================================== FLOAT to I32 ==============
// Simple test that tries to reproduce a Clang compiler bug.
__attribute__((noinline))
static int32_t clamp32FromFloat(float f)
{
static const float scale = (float)(1UL << 31);
static const float limpos = 1.;
static const float limneg = -1.;
if (f <= limneg) {
return INT32_MIN;
} else if (f >= limpos) {
return INT32_MAX;
}
f *= scale;
/* integer conversion is through truncation (though int to float is not).
* ensure that we round to nearest, ties away from 0.
*/
return f > 0 ? f + 0.5 : f - 0.5;
}
void local_convert_float_to_int32(const float *input,
int32_t *output,
int count) {
for (int i = 0; i < count; i++) {
*output++ = clamp32FromFloat(*input++);
}
}
TEST(test_flowgraph, simple_convert_float_int32) {
std::array<int32_t, kNumSamples> output;
// Do it inline, which will probably work even with a buggy compiler.
// This validates the expected data.
const float *in = kInputFloat.data();
output.fill(777);
int32_t *out = output.data();
for (int i = 0; i < kNumSamples; i++) {
int64_t n = (int64_t) (*in++ * 2147483648.0f);
*out++ = (int32_t)std::min((int64_t)INT32_MAX,
std::max((int64_t)INT32_MIN, n)); // clip
}
for (int i = 0; i < kNumSamples; i++) {
EXPECT_EQ(kExpectedI32.at(i), output.at(i)) << ", i = " << i;
}
}
TEST(test_flowgraph, local_convert_float_int32) {
std::array<int32_t, kNumSamples> output;
// Convert audio signal using the function.
output.fill(777);
local_convert_float_to_int32(kInputFloat.data(), output.data(), kNumSamples);
for (int i = 0; i < kNumSamples; i++) {
EXPECT_EQ(kExpectedI32.at(i), output.at(i)) << ", i = " << i;
}
}
TEST(test_flowgraph, module_sinki32) {
std::array<int32_t, kNumSamples + 10> output; // larger than input
SourceFloat sourceFloat{1};
SinkI32 sinkI32{1};
sourceFloat.setData(kInputFloat.data(), kNumSamples);
sourceFloat.output.connect(&sinkI32.input);
output.fill(777);
int32_t numRead = sinkI32.read(output.data(), output.size());
ASSERT_EQ(kNumSamples, numRead);
for (int i = 0; i < numRead; i++) {
EXPECT_EQ(kExpectedI32.at(i), output.at(i)) << ", i = " << i;
}
}
TEST(test_flowgraph, module_mono_to_stereo) {
static const float input[] = {1.0f, 2.0f, 3.0f};
float output[100] = {};
SourceFloat sourceFloat{1};
MonoToMultiConverter monoToStereo{2};
SinkFloat sinkFloat{2};
sourceFloat.setData(input, 3);
sourceFloat.output.connect(&monoToStereo.input);
monoToStereo.output.connect(&sinkFloat.input);
int32_t numRead = sinkFloat.read(output, 8);
ASSERT_EQ(3, numRead);
EXPECT_EQ(input[0], output[0]);
EXPECT_EQ(input[0], output[1]);
EXPECT_EQ(input[1], output[2]);
EXPECT_EQ(input[1], output[3]);
EXPECT_EQ(input[2], output[4]);
EXPECT_EQ(input[2], output[5]);
}
TEST(test_flowgraph, module_ramp_linear) {
constexpr int singleNumOutput = 1;
constexpr int rampSize = 5;
constexpr int numOutput = 100;
constexpr float value = 1.0f;
constexpr float initialTarget = 10.0f;
constexpr float finalTarget = 100.0f;
constexpr float tolerance = 0.0001f; // arbitrary
float output[numOutput] = {};
RampLinear rampLinear{1};
SinkFloat sinkFloat{1};
rampLinear.input.setValue(value);
rampLinear.setLengthInFrames(rampSize);
rampLinear.output.connect(&sinkFloat.input);
// Check that the values go to the initial target instantly.
rampLinear.setTarget(initialTarget);
int32_t singleNumRead = sinkFloat.read(output, singleNumOutput);
ASSERT_EQ(singleNumRead, singleNumOutput);
EXPECT_NEAR(value * initialTarget, output[0], tolerance);
// Now set target and check that the linear ramp works as expected.
rampLinear.setTarget(finalTarget);
int32_t numRead = sinkFloat.read(output, numOutput);
const float incrementSize = (finalTarget - initialTarget) / rampSize;
ASSERT_EQ(numOutput, numRead);
int i = 0;
for (; i < rampSize; i++) {
float expected = value * (initialTarget + i * incrementSize);
EXPECT_NEAR(expected, output[i], tolerance);
}
for (; i < numOutput; i++) {
float expected = value * finalTarget;
EXPECT_NEAR(expected, output[i], tolerance);
}
}
// It is easiest to represent packed 24-bit data as a byte array.
// This test will read from input, convert to float, then write
// back to output as bytes.
TEST(test_flowgraph, module_packed_24) {
static const uint8_t input[] = {0x01, 0x23, 0x45,
0x67, 0x89, 0xAB,
0xCD, 0xEF, 0x5A};
uint8_t output[99] = {};
SourceI24 sourceI24{1};
SinkI24 sinkI24{1};
int numInputFrames = sizeof(input) / kBytesPerI24Packed;
sourceI24.setData(input, numInputFrames);
sourceI24.output.connect(&sinkI24.input);
int32_t numRead = sinkI24.read(output, sizeof(output) / kBytesPerI24Packed);
ASSERT_EQ(numInputFrames, numRead);
for (size_t i = 0; i < sizeof(input); i++) {
EXPECT_EQ(input[i], output[i]);
}
}
TEST(test_flowgraph, module_clip_to_range) {
constexpr float myMin = -2.0f;
constexpr float myMax = 1.5f;
static const float input[] = {-9.7, 0.5f, -0.25, 1.0f, 12.3};
static const float expected[] = {myMin, 0.5f, -0.25, 1.0f, myMax};
float output[100];
SourceFloat sourceFloat{1};
ClipToRange clipper{1};
SinkFloat sinkFloat{1};
int numInputFrames = sizeof(input) / sizeof(input[0]);
sourceFloat.setData(input, numInputFrames);
clipper.setMinimum(myMin);
clipper.setMaximum(myMax);
sourceFloat.output.connect(&clipper.input);
clipper.output.connect(&sinkFloat.input);
int numOutputFrames = sizeof(output) / sizeof(output[0]);
int32_t numRead = sinkFloat.read(output, numOutputFrames);
ASSERT_EQ(numInputFrames, numRead);
constexpr float tolerance = 0.000001f; // arbitrary
for (int i = 0; i < numRead; i++) {
EXPECT_NEAR(expected[i], output[i], tolerance);
}
}
TEST(test_flowgraph, module_mono_blend) {
// Two channel to two channel with 3 inputs and outputs.
constexpr int numChannels = 2;
constexpr int numFrames = 3;
static const float input[] = {-0.7, 0.5, -0.25, 1.25, 1000, 2000};
static const float expected[] = {-0.1, -0.1, 0.5, 0.5, 1500, 1500};
float output[100];
SourceFloat sourceFloat{numChannels};
MonoBlend monoBlend{numChannels};
SinkFloat sinkFloat{numChannels};
sourceFloat.setData(input, numFrames);
sourceFloat.output.connect(&monoBlend.input);
monoBlend.output.connect(&sinkFloat.input);
int32_t numRead = sinkFloat.read(output, numFrames);
ASSERT_EQ(numRead, numFrames);
constexpr float tolerance = 0.000001f; // arbitrary
for (int i = 0; i < numRead; i++) {
EXPECT_NEAR(expected[i], output[i], tolerance);
}
}
TEST(test_flowgraph, module_limiter) {
constexpr int kNumSamples = 101;
constexpr float kLastSample = 3.0f;
constexpr float kFirstSample = -kLastSample;
constexpr float kDeltaBetweenSamples = (kLastSample - kFirstSample) / (kNumSamples - 1);
constexpr float kTolerance = 0.00001f;
float input[kNumSamples];
float output[kNumSamples];
SourceFloat sourceFloat{1};
Limiter limiter{1};
SinkFloat sinkFloat{1};
for (int i = 0; i < kNumSamples; i++) {
input[i] = kFirstSample + i * kDeltaBetweenSamples;
}
const int numInputFrames = std::size(input);
sourceFloat.setData(input, numInputFrames);
sourceFloat.output.connect(&limiter.input);
limiter.output.connect(&sinkFloat.input);
const int numOutputFrames = std::size(output);
int32_t numRead = sinkFloat.read(output, numOutputFrames);
ASSERT_EQ(numInputFrames, numRead);
for (int i = 0; i < numRead; i++) {
// limiter must be symmetric wrt 0.
EXPECT_NEAR(output[i], -output[kNumSamples - i - 1], kTolerance);
if (i > 0) {
EXPECT_GE(output[i], output[i - 1]); // limiter must be monotonic
}
if (input[i] == 0.f) {
EXPECT_EQ(0.f, output[i]);
} else if (input[i] > 0.0f) {
EXPECT_GE(output[i], 0.0f);
EXPECT_LE(output[i], M_SQRT2); // limiter actually limits
EXPECT_LE(output[i], input[i]); // a limiter, gain <= 1
} else {
EXPECT_LE(output[i], 0.0f);
EXPECT_GE(output[i], -M_SQRT2); // limiter actually limits
EXPECT_GE(output[i], input[i]); // a limiter, gain <= 1
}
if (-1.f <= input[i] && input[i] <= 1.f) {
EXPECT_EQ(input[i], output[i]);
}
}
}
TEST(test_flowgraph, module_limiter_nan) {
constexpr int kArbitraryOutputSize = 100;
static const float input[] = {NAN, 0.5f, NAN, NAN, -10.0f, NAN};
static const float expected[] = {0.0f, 0.5f, 0.5f, 0.5f, -M_SQRT2, -M_SQRT2};
constexpr float tolerance = 0.00001f;
float output[kArbitraryOutputSize];
SourceFloat sourceFloat{1};
Limiter limiter{1};
SinkFloat sinkFloat{1};
const int numInputFrames = std::size(input);
sourceFloat.setData(input, numInputFrames);
sourceFloat.output.connect(&limiter.input);
limiter.output.connect(&sinkFloat.input);
const int numOutputFrames = std::size(output);
int32_t numRead = sinkFloat.read(output, numOutputFrames);
ASSERT_EQ(numInputFrames, numRead);
for (int i = 0; i < numRead; i++) {
EXPECT_NEAR(expected[i], output[i], tolerance);
}
}
TEST(test_flowgraph, module_sinki16_multiple_reads) {
static constexpr int kNumSamples = 8;
std::array<int16_t, kNumSamples + 10> output; // larger than input
SourceFloat sourceFloat{1};
SinkI16 sinkI16{1};
sourceFloat.setData(kInputFloat.data(), kNumSamples);
sourceFloat.output.connect(&sinkI16.input);
output.fill(777);
// Read the first half of the data
int32_t numRead = sinkI16.read(output.data(), kNumSamples / 2);
ASSERT_EQ(kNumSamples / 2, numRead);
for (int i = 0; i < numRead; i++) {
EXPECT_EQ(kExpectedI16.at(i), output.at(i)) << ", i = " << i;
}
// Read the rest of the data
numRead = sinkI16.read(output.data(), output.size());
ASSERT_EQ(kNumSamples / 2, numRead);
for (int i = 0; i < numRead; i++) {
EXPECT_EQ(kExpectedI16.at(i + kNumSamples / 2), output.at(i)) << ", i = " << i;
}
}
// =================================== FLOAT to Q8.23 ==============
__attribute__((noinline))
static int32_t clamp24FromFloat(float f)
{
static const float scale = 1 << 23;
return (int32_t) lroundf(fmaxf(fminf(f * scale, scale - 1.f), -scale));
}
void local_convert_float_to_i8_24(const float *input,
int32_t *output,
int count) {
for (int i = 0; i < count; i++) {
*output++ = clamp24FromFloat(*input++);
}
}
TEST(test_flowgraph, local_convert_float_to_i8_24) {
std::array<int32_t, kNumSamples> output;
// Convert audio signal using the function.
output.fill(777);
local_convert_float_to_i8_24(kInputFloat.data(), output.data(), kNumSamples);
for (int i = 0; i < kNumSamples; i++) {
EXPECT_EQ(kExpectedI8_24.at(i), output.at(i)) << ", i = " << i;
}
}
TEST(test_flowgraph, module_sinkI8_24) {
std::array<int32_t, kNumSamples + 10> output; // larger than input
SourceFloat sourceFloat{2};
SinkI8_24 sinkI8_24{2};
sourceFloat.setData(kInputFloat.data(), kNumSamples);
sourceFloat.output.connect(&sinkI8_24.input);
output.fill(777);
int32_t numRead = sinkI8_24.read(output.data(), output.size());
ASSERT_EQ(kNumSamples, numRead);
for (int i = 0; i < numRead; i++) {
EXPECT_EQ(kExpectedI8_24.at(i), output.at(i)) << ", i = " << i;
}
}
TEST(test_flowgraph, module_sourceI8_24) {
static const int32_t input[] = {1 << 23, 1 << 22, -(1 << 21), -(1 << 23), 0, 1 << 25,
-(1 << 25)};
static const float expected[] = {1.0f, 0.5f, -0.25f, -1.0f, 0.0f, 4.0f, -4.0f};
float output[100];
SourceI8_24 sourceI8_24{1};
SinkFloat sinkFloat{1};
int numSamples = std::size(input);
sourceI8_24.setData(input, numSamples);
sourceI8_24.output.connect(&sinkFloat.input);
int32_t numRead = sinkFloat.read(output, numSamples);
ASSERT_EQ(numSamples, numRead);
for (int i = 0; i < numRead; i++) {
EXPECT_EQ(expected[i], output[i]) << ", i = " << i;
}
}
void checkSampleRateConversionVariedSizes(int32_t sourceSampleRate,
int32_t sinkSampleRate,
MultiChannelResampler::Quality resamplerQuality) {
AAudioFlowGraph flowgraph;
aaudio_result_t result = flowgraph.configure(AUDIO_FORMAT_PCM_FLOAT /* sourceFormat */,
1 /* sourceChannelCount */,
sourceSampleRate,
AUDIO_FORMAT_PCM_FLOAT /* sinkFormat */,
1 /* sinkChannelCount */,
sinkSampleRate,
false /* useMonoBlend */,
false /* useVolumeRamps */,
0.0f /* audioBalance */,
resamplerQuality);
IntegerRatio ratio(sourceSampleRate, sinkSampleRate);
ratio.reduce();
ASSERT_EQ(AAUDIO_OK, result);
const int inputSize = ratio.getNumerator();
const int outputSize = ratio.getDenominator();
float input[inputSize];
float output[outputSize];
for (int i = 0; i < inputSize; i++) {
input[i] = i * 1.0f / inputSize;
}
int inputUsed = 0;
int outputRead = 0;
int curInputSize = 1;
// Process the data with larger and larger input buffer sizes.
while (inputUsed < inputSize) {
outputRead += flowgraph.process((void *) (input + inputUsed),
curInputSize,
(void *) (output + outputRead),
outputSize - outputRead);
inputUsed += curInputSize;
curInputSize = std::min(curInputSize + 5, inputSize - inputUsed);
}
ASSERT_EQ(outputSize, outputRead);
for (int i = 1; i < outputSize; i++) {
// The first values of the flowgraph will be close to zero.
// Besides those, the values should be strictly increasing.
if (output[i - 1] > 0.01f) {
EXPECT_GT(output[i], output[i - 1]);
}
}
}
TEST(test_flowgraph, flowgraph_varied_sizes_all) {
const int rates[] = {8000, 11025, 22050, 32000, 44100, 48000, 64000, 88200, 96000};
const MultiChannelResampler::Quality qualities[] =
{
MultiChannelResampler::Quality::Fastest,
MultiChannelResampler::Quality::Low,
MultiChannelResampler::Quality::Medium,
MultiChannelResampler::Quality::High,
MultiChannelResampler::Quality::Best
};
for (int srcRate : rates) {
for (int destRate : rates) {
for (auto quality : qualities) {
if (srcRate != destRate) {
checkSampleRateConversionVariedSizes(srcRate, destRate, quality);
}
}
}
}
}
void checkSampleRateConversionPullLater(int32_t sourceSampleRate,
int32_t sinkSampleRate,
MultiChannelResampler::Quality resamplerQuality) {
AAudioFlowGraph flowgraph;
aaudio_result_t result = flowgraph.configure(AUDIO_FORMAT_PCM_FLOAT /* sourceFormat */,
1 /* sourceChannelCount */,
sourceSampleRate,
AUDIO_FORMAT_PCM_FLOAT /* sinkFormat */,
1 /* sinkChannelCount */,
sinkSampleRate,
false /* useMonoBlend */,
false /* useVolumeRamps */,
0.0f /* audioBalance */,
resamplerQuality);
IntegerRatio ratio(sourceSampleRate, sinkSampleRate);
ratio.reduce();
ASSERT_EQ(AAUDIO_OK, result);
const int inputSize = ratio.getNumerator();
const int outputSize = ratio.getDenominator();
float input[inputSize];
float output[outputSize];
for (int i = 0; i < inputSize; i++) {
input[i] = i * 1.0f / inputSize;
}
// Read half the data with process.
int outputRead = flowgraph.process((void *) input,
inputSize,
(void *) output,
outputSize / 2);
ASSERT_EQ(outputSize / 2, outputRead);
// Now read the other half of the data with pull.
outputRead += flowgraph.pull(
(void *) (output + outputRead),
outputSize - outputRead);
ASSERT_EQ(outputSize, outputRead);
for (int i = 1; i < outputSize; i++) {
// The first values of the flowgraph will be close to zero.
// Besides those, the values should be strictly increasing.
if (output[i - 1] > 0.01f) {
EXPECT_GT(output[i], output[i - 1]);
}
}
}
// TODO: b/289508408 - Remove non-parameterized tests if they get noisy.
TEST(test_flowgraph, flowgraph_pull_later_all) {
const int rates[] = {8000, 11025, 22050, 32000, 44100, 48000, 64000, 88200, 96000};
const MultiChannelResampler::Quality qualities[] =
{
MultiChannelResampler::Quality::Fastest,
MultiChannelResampler::Quality::Low,
MultiChannelResampler::Quality::Medium,
MultiChannelResampler::Quality::High,
MultiChannelResampler::Quality::Best
};
for (int srcRate : rates) {
for (int destRate : rates) {
for (auto quality : qualities) {
if (srcRate != destRate) {
checkSampleRateConversionPullLater(srcRate, destRate, quality);
}
}
}
}
}
class TestFlowgraphSampleRateConversion : public ::testing::Test,
public ::testing::WithParamInterface<TestFlowgraphResamplerParams> {
};
const char* resamplerQualityToString(MultiChannelResampler::Quality quality) {
switch (quality) {
case MultiChannelResampler::Quality::Fastest: return "FASTEST";
case MultiChannelResampler::Quality::Low: return "LOW";
case MultiChannelResampler::Quality::Medium: return "MEDIUM";
case MultiChannelResampler::Quality::High: return "HIGH";
case MultiChannelResampler::Quality::Best: return "BEST";
}
return "UNKNOWN";
}
static std::string getTestName(
const ::testing::TestParamInfo<TestFlowgraphResamplerParams>& info) {
return std::string()
+ std::to_string(std::get<PARAM_SOURCE_SAMPLE_RATE>(info.param))
+ "__" + std::to_string(std::get<PARAM_SINK_SAMPLE_RATE>(info.param))
+ "__" + resamplerQualityToString(std::get<PARAM_RESAMPLER_QUALITY>(info.param));
}
TEST_P(TestFlowgraphSampleRateConversion, test_flowgraph_pull_later) {
checkSampleRateConversionPullLater(std::get<PARAM_SOURCE_SAMPLE_RATE>(GetParam()),
std::get<PARAM_SINK_SAMPLE_RATE>(GetParam()),
std::get<PARAM_RESAMPLER_QUALITY>(GetParam()));
}
TEST_P(TestFlowgraphSampleRateConversion, test_flowgraph_varied_sizes) {
checkSampleRateConversionVariedSizes(std::get<PARAM_SOURCE_SAMPLE_RATE>(GetParam()),
std::get<PARAM_SINK_SAMPLE_RATE>(GetParam()),
std::get<PARAM_RESAMPLER_QUALITY>(GetParam()));
}
INSTANTIATE_TEST_SUITE_P(
test_flowgraph,
TestFlowgraphSampleRateConversion,
::testing::Values(
TestFlowgraphResamplerParams({8000, 11025, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({8000, 48000, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({8000, 44100, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({11025, 24000, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({11025, 48000,
MultiChannelResampler::Quality::Fastest}),
TestFlowgraphResamplerParams({11025, 48000, MultiChannelResampler::Quality::Low}),
TestFlowgraphResamplerParams({11025, 48000,
MultiChannelResampler::Quality::Medium}),
TestFlowgraphResamplerParams({11025, 48000, MultiChannelResampler::Quality::High}),
TestFlowgraphResamplerParams({11025, 48000, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({11025, 44100, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({11025, 88200, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({16000, 48000, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({44100, 48000, MultiChannelResampler::Quality::Low}),
TestFlowgraphResamplerParams({44100, 48000, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({48000, 11025, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({48000, 44100, MultiChannelResampler::Quality::Best}),
TestFlowgraphResamplerParams({44100, 11025, MultiChannelResampler::Quality::Best})),
&getTestName
);