You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

277 lines
9.2 KiB

// Copyright 2019 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.
#pragma once
#include <gtest/gtest.h>
#include <algorithm>
#include <cassert>
#include <cstddef>
#include <cstdlib>
#include <functional>
#include <random>
#include <vector>
#include <fp16.h>
#include <xnnpack.h>
#include <xnnpack/AlignedAllocator.h>
#include <xnnpack/pack.h>
#include <xnnpack/params-init.h>
#include <xnnpack/params.h>
class VMulCAddCMicrokernelTester {
public:
enum class Variant {
Native,
Scalar,
};
inline VMulCAddCMicrokernelTester& channel_tile(size_t channel_tile) {
this->channel_tile_ = channel_tile;
return *this;
}
inline size_t channel_tile() const {
return this->channel_tile_;
}
inline VMulCAddCMicrokernelTester& channels(size_t channels) {
assert(channels != 0);
this->channels_ = channels;
return *this;
}
inline size_t channels() const {
return this->channels_;
}
inline size_t packed_channels() const {
return channels() % channel_tile() == 0 ? channels() : (channels() / channel_tile() + 1) * channel_tile();
}
inline VMulCAddCMicrokernelTester& rows(size_t rows) {
assert(rows != 0);
this->rows_ = rows;
return *this;
}
inline size_t rows() const {
return this->rows_;
}
inline VMulCAddCMicrokernelTester& input_stride(size_t input_stride) {
this->input_stride_ = input_stride;
return *this;
}
inline size_t input_stride() const {
return this->input_stride_ == 0 ? channels() : this->input_stride_;
}
inline VMulCAddCMicrokernelTester& output_stride(size_t output_stride) {
this->output_stride_ = output_stride;
return *this;
}
inline size_t output_stride() const {
return this->output_stride_ == 0 ? channels() : this->output_stride_;
}
inline VMulCAddCMicrokernelTester& inplace(bool inplace) {
this->inplace_ = inplace;
return *this;
}
inline bool inplace() const {
return this->inplace_;
}
inline VMulCAddCMicrokernelTester& qmin(uint8_t qmin) {
this->qmin_ = qmin;
return *this;
}
inline uint8_t qmin() const {
return this->qmin_;
}
inline VMulCAddCMicrokernelTester& qmax(uint8_t qmax) {
this->qmax_ = qmax;
return *this;
}
inline uint8_t qmax() const {
return this->qmax_;
}
inline VMulCAddCMicrokernelTester& iterations(size_t iterations) {
this->iterations_ = iterations;
return *this;
}
inline size_t iterations() const {
return this->iterations_;
}
void Test(xnn_f16_vmulcaddc_ukernel_function vmulcaddc, Variant variant = Variant::Native) const {
std::random_device random_device;
auto rng = std::mt19937(random_device());
auto f32rng = std::bind(std::uniform_real_distribution<float>(0.0f, 1.0f), rng);
auto f16rng = std::bind(fp16_ieee_from_fp32_value, f32rng);
if (inplace()) {
ASSERT_EQ(input_stride(), output_stride());
}
std::vector<uint16_t> x((rows() - 1) * input_stride() + channels() + XNN_EXTRA_BYTES / sizeof(uint16_t));
std::vector<uint16_t> scale(channels());
std::vector<uint16_t> bias(channels());
std::vector<uint16_t, AlignedAllocator<uint16_t, 64>> packed_w(packed_channels() * 2);
std::vector<uint16_t> y((rows() - 1) * output_stride() + channels() + (inplace() ? XNN_EXTRA_BYTES / sizeof(uint16_t) : 0));
std::vector<float> y_ref(rows() * channels());
for (size_t iteration = 0; iteration < iterations(); iteration++) {
std::generate(scale.begin(), scale.end(), std::ref(f16rng));
std::generate(bias.begin(), bias.end(), std::ref(f16rng));
std::generate(x.begin(), x.end(), std::ref(f16rng));
if (inplace()) {
std::copy(x.cbegin(), x.cend(), y.begin());
} else {
std::fill(y.begin(), y.end(), UINT16_C(0x7E00) /* NaN */);
}
const uint16_t* x_data = inplace() ? y.data() : x.data();
std::fill(packed_w.begin(), packed_w.end(), UINT16_C(0x7E00) /* NaN */);
xnn_pack_f16_vmulcaddc_w(channels(), channel_tile(),
scale.data(), bias.data(), packed_w.data(), nullptr);
// Compute reference results.
for (size_t i = 0; i < rows(); i++) {
for (size_t j = 0; j < channels(); j++) {
y_ref[i * channels() + j] = fp16_ieee_to_fp32_value(x_data[i * input_stride() + j]) * fp16_ieee_to_fp32_value(scale[j]) + fp16_ieee_to_fp32_value(bias[j]);
}
}
const float accumulated_min = *std::min_element(y_ref.cbegin(), y_ref.cend());
const float accumulated_max = *std::max_element(y_ref.cbegin(), y_ref.cend());
const float accumulated_range = accumulated_max - accumulated_min;
const float y_max = fp16_ieee_to_fp32_value(fp16_ieee_from_fp32_value(accumulated_max - accumulated_range / 255.0f * float(255 - qmax())));
const float y_min = fp16_ieee_to_fp32_value(fp16_ieee_from_fp32_value(accumulated_min + accumulated_range / 255.0f * float(qmin())));
for (float& y_value : y_ref) {
y_value = std::max(std::min(y_value, y_max), y_min);
}
// Prepare parameters.
xnn_f16_minmax_params params = xnn_init_f16_minmax_params(
fp16_ieee_from_fp32_value(y_min),
fp16_ieee_from_fp32_value(y_max));
// Call optimized micro-kernel.
vmulcaddc(rows(), channels() * sizeof(uint16_t),
x_data, input_stride() * sizeof(uint16_t),
packed_w.data(),
y.data(), output_stride() * sizeof(uint16_t),
&params);
// Verify results.
for (size_t i = 0; i < rows(); i++) {
for (size_t j = 0; j < channels(); j++) {
ASSERT_NEAR(fp16_ieee_to_fp32_value(y[i * output_stride() + j]), y_ref[i * channels() + j], std::max(1.0e-4f, std::abs(y_ref[i * channels() + j]) * 1.0e-2f))
<< "at pixel " << i << " / " << rows()
<< ", channel = " << j << " / " << channels();
}
}
}
}
void Test(xnn_f32_vmulcaddc_ukernel_function vmulcaddc, Variant variant = Variant::Native) const {
std::random_device random_device;
auto rng = std::mt19937(random_device());
auto f32rng = std::bind(std::uniform_real_distribution<float>(0.0f, 1.0f), rng);
if (inplace()) {
ASSERT_EQ(input_stride(), output_stride());
}
std::vector<float> x((rows() - 1) * input_stride() + channels() + XNN_EXTRA_BYTES / sizeof(float));
std::vector<float> scale(channels());
std::vector<float> bias(channels());
std::vector<float, AlignedAllocator<float, 64>> packed_w(packed_channels() * 2);
std::vector<float> y((rows() - 1) * output_stride() + channels() + (inplace() ? XNN_EXTRA_BYTES / sizeof(float) : 0));
std::vector<float> y_ref(rows() * channels());
for (size_t iteration = 0; iteration < iterations(); iteration++) {
std::generate(scale.begin(), scale.end(), std::ref(f32rng));
std::generate(bias.begin(), bias.end(), std::ref(f32rng));
std::generate(x.begin(), x.end(), std::ref(f32rng));
if (inplace()) {
std::copy(x.cbegin(), x.cend(), y.begin());
} else {
std::fill(y.begin(), y.end(), nanf(""));
}
const float* x_data = inplace() ? y.data() : x.data();
std::fill(packed_w.begin(), packed_w.end(), nanf(""));
xnn_pack_f32_vmulcaddc_w(channels(), channel_tile(),
scale.data(), bias.data(), packed_w.data(), nullptr);
// Compute reference results.
for (size_t i = 0; i < rows(); i++) {
for (size_t j = 0; j < channels(); j++) {
y_ref[i * channels() + j] = x_data[i * input_stride() + j] * scale[j] + bias[j];
}
}
const float accumulated_min = *std::min_element(y_ref.cbegin(), y_ref.cend());
const float accumulated_max = *std::max_element(y_ref.cbegin(), y_ref.cend());
const float accumulated_range = accumulated_max - accumulated_min;
const float y_max = accumulated_max - accumulated_range / 255.0f * float(255 - qmax());
const float y_min = accumulated_min + accumulated_range / 255.0f * float(qmin());
for (float& y_value : y_ref) {
y_value = std::max<float>(std::min<float>(y_value, y_max), y_min);
}
// Prepare parameters.
xnn_f32_minmax_params params = { };
switch (variant) {
case Variant::Native:
params = xnn_init_f32_minmax_params(y_min, y_max);
break;
case Variant::Scalar:
params = xnn_init_scalar_f32_minmax_params(y_min, y_max);
break;
}
// Call optimized micro-kernel.
vmulcaddc(rows(), channels() * sizeof(float),
x_data, input_stride() * sizeof(float),
packed_w.data(),
y.data(), output_stride() * sizeof(float),
&params);
// Verify results.
for (size_t i = 0; i < rows(); i++) {
for (size_t j = 0; j < channels(); j++) {
ASSERT_NEAR(y[i * output_stride() + j], y_ref[i * channels() + j], std::abs(y_ref[i * channels() + j]) * 1.0e-6f)
<< "at pixel " << i << " / " << rows()
<< ", channel = " << j << " / " << channels();
}
}
}
}
private:
size_t channel_tile_{1};
size_t channels_{1};
size_t rows_{1};
size_t input_stride_{0};
size_t output_stride_{0};
bool inplace_{false};
uint8_t qmin_{0};
uint8_t qmax_{255};
size_t iterations_{15};
};