eigen/test/rand.cpp
2024-02-24 13:16:23 +00:00

248 lines
10 KiB
C++

// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2015 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#include <cstdlib>
#include "main.h"
// For GCC-6, if this function is inlined then there seems to be an optimization
// bug that triggers a failure. This failure goes away if you access `r` in
// in any way, and for any other compiler.
template <typename Scalar>
EIGEN_DONT_INLINE Scalar check_in_range(Scalar x, Scalar y) {
Scalar r = internal::random<Scalar>(x, y);
VERIFY(r >= x);
if (y >= x) {
VERIFY(r <= y);
}
return r;
}
template <typename Scalar>
void check_all_in_range(Scalar x, Scalar y) {
Array<int, 1, Dynamic> mask(y - x + 1);
mask.fill(0);
int64_t n = (y - x + 1) * 32;
for (int64_t k = 0; k < n; ++k) {
mask(check_in_range(x, y) - x)++;
}
for (Index i = 0; i < mask.size(); ++i)
if (mask(i) == 0) std::cout << "WARNING: value " << x + i << " not reached." << std::endl;
VERIFY((mask > 0).all());
}
template <typename Scalar, typename EnableIf = void>
class HistogramHelper {
public:
HistogramHelper(int nbins) : HistogramHelper(Scalar(-1), Scalar(1), nbins) {}
HistogramHelper(Scalar lower, Scalar upper, int nbins) {
lower_ = static_cast<double>(lower);
upper_ = static_cast<double>(upper);
num_bins_ = nbins;
bin_width_ = (upper_ - lower_) / static_cast<double>(nbins);
}
int bin(Scalar v) {
double result = (static_cast<double>(v) - lower_) / bin_width_;
return std::min<int>(static_cast<int>(result), num_bins_ - 1);
}
double uniform_bin_probability(int bin) {
double range = upper_ - lower_;
if (bin < num_bins_ - 1) {
return bin_width_ / range;
}
return (upper_ - (lower_ + double(bin) * bin_width_)) / range;
}
private:
double lower_;
double upper_;
int num_bins_;
double bin_width_;
};
template <typename Scalar>
class HistogramHelper<Scalar, std::enable_if_t<Eigen::NumTraits<Scalar>::IsInteger>> {
public:
using RangeType = typename Eigen::internal::make_unsigned<Scalar>::type;
HistogramHelper(int nbins)
: HistogramHelper(Eigen::NumTraits<Scalar>::lowest(), Eigen::NumTraits<Scalar>::highest(), nbins) {}
HistogramHelper(Scalar lower, Scalar upper, int nbins)
: lower_{lower}, upper_{upper}, num_bins_{nbins}, bin_width_{bin_width(lower, upper, nbins)} {}
int bin(Scalar v) { return static_cast<int>(RangeType(RangeType(v) - RangeType(lower_)) / bin_width_); }
double uniform_bin_probability(int bin) {
// The full range upper - lower + 1 might overflow the RangeType by one.
// So instead, we know we have (nbins - 1) bins of width bin_width_,
// and the last bin of width:
RangeType last_bin_width =
RangeType(upper_) - (RangeType(lower_) + RangeType(num_bins_ - 1) * bin_width_) + RangeType(1);
double last_bin_ratio = static_cast<double>(last_bin_width) / static_cast<double>(bin_width_);
// Total probability = (nbins - 1) * p + last_bin_ratio * p = 1.0
// p = 1.0 / (nbins - 1 + last_bin_ratio)
double p = 1.0 / (last_bin_ratio + num_bins_ - 1);
if (bin < num_bins_ - 1) {
return p;
}
return last_bin_ratio * p;
}
private:
static constexpr RangeType bin_width(Scalar lower, Scalar upper, int nbins) {
// Avoid overflow in computing the full range.
// floor( (upper - lower + 1) / nbins) )
// = floor( (upper- nbins - lower + 1 + nbins) / nbins) )
return RangeType(RangeType(upper - nbins) - RangeType(lower) + 1) / nbins + 1;
}
Scalar lower_;
Scalar upper_;
int num_bins_;
RangeType bin_width_;
};
template <typename Scalar>
void check_histogram(Scalar x, Scalar y, int bins) {
Eigen::VectorXd hist = Eigen::VectorXd::Zero(bins);
HistogramHelper<Scalar> hist_helper(x, y, bins);
int64_t n = static_cast<int64_t>(bins) * 10000; // Approx 10000 per bin.
for (int64_t k = 0; k < n; ++k) {
Scalar r = check_in_range(x, y);
int bin = hist_helper.bin(r);
hist(bin)++;
}
// Normalize bins by probability.
for (int i = 0; i < bins; ++i) {
hist(i) = hist(i) / n / hist_helper.uniform_bin_probability(i);
}
VERIFY(((hist.array() - 1.0).abs() < 0.05).all());
}
template <typename Scalar>
void check_histogram(int bins) {
Eigen::VectorXd hist = Eigen::VectorXd::Zero(bins);
HistogramHelper<Scalar> hist_helper(bins);
int64_t n = static_cast<int64_t>(bins) * 10000; // Approx 10000 per bin.
for (int64_t k = 0; k < n; ++k) {
Scalar r = Eigen::internal::random<Scalar>();
int bin = hist_helper.bin(r);
hist(bin)++;
}
// Normalize bins by probability.
for (int i = 0; i < bins; ++i) {
hist(i) = hist(i) / n / hist_helper.uniform_bin_probability(i);
}
VERIFY(((hist.array() - 1.0).abs() < 0.05).all());
}
EIGEN_DECLARE_TEST(rand) {
int64_t int64_ref = NumTraits<int64_t>::highest() / 10;
// the minimum guarantees that these conversions are safe
int8_t int8t_offset = static_cast<int8_t>((std::min)(g_repeat, 64));
int16_t int16t_offset = static_cast<int16_t>((std::min)(g_repeat, 8000));
EIGEN_UNUSED_VARIABLE(int64_ref);
EIGEN_UNUSED_VARIABLE(int8t_offset);
EIGEN_UNUSED_VARIABLE(int16t_offset);
for (int i = 0; i < g_repeat * 10000; i++) {
CALL_SUBTEST_1(check_in_range<float>(10.0f, 11.0f));
CALL_SUBTEST_1(check_in_range<float>(1.24234523f, 1.24234523f));
CALL_SUBTEST_1(check_in_range<float>(-1.0f, 1.0f));
CALL_SUBTEST_1(check_in_range<float>(-1432.2352f, -1432.2352f));
CALL_SUBTEST_2(check_in_range<double>(10.0, 11.0));
CALL_SUBTEST_2(check_in_range<double>(1.24234523, 1.24234523));
CALL_SUBTEST_2(check_in_range<double>(-1.0, 1.0));
CALL_SUBTEST_2(check_in_range<double>(-1432.2352, -1432.2352));
CALL_SUBTEST_3(check_in_range<long double>(10.0L, 11.0L));
CALL_SUBTEST_3(check_in_range<long double>(1.24234523L, 1.24234523L));
CALL_SUBTEST_3(check_in_range<long double>(-1.0L, 1.0L));
CALL_SUBTEST_3(check_in_range<long double>(-1432.2352L, -1432.2352L));
CALL_SUBTEST_4(check_in_range<half>(half(10.0f), half(11.0f)));
CALL_SUBTEST_4(check_in_range<half>(half(1.24234523f), half(1.24234523f)));
CALL_SUBTEST_4(check_in_range<half>(half(-1.0f), half(1.0f)));
CALL_SUBTEST_4(check_in_range<half>(half(-1432.2352f), half(-1432.2352f)));
CALL_SUBTEST_5(check_in_range<bfloat16>(bfloat16(10.0f), bfloat16(11.0f)));
CALL_SUBTEST_5(check_in_range<bfloat16>(bfloat16(1.24234523f), bfloat16(1.24234523f)));
CALL_SUBTEST_5(check_in_range<bfloat16>(bfloat16(-1.0f), bfloat16(1.0f)));
CALL_SUBTEST_5(check_in_range<bfloat16>(bfloat16(-1432.2352f), bfloat16(-1432.2352f)));
CALL_SUBTEST_6(check_in_range<int32_t>(0, -1));
CALL_SUBTEST_6(check_in_range<int16_t>(0, -1));
CALL_SUBTEST_6(check_in_range<int64_t>(0, -1));
CALL_SUBTEST_6(check_in_range<int32_t>(-673456, 673456));
CALL_SUBTEST_6(check_in_range<int32_t>(-RAND_MAX + 10, RAND_MAX - 10));
CALL_SUBTEST_6(check_in_range<int16_t>(-24345, 24345));
CALL_SUBTEST_6(check_in_range<int64_t>(-int64_ref, int64_ref));
}
CALL_SUBTEST_7(check_all_in_range<int8_t>(11, 11));
CALL_SUBTEST_7(check_all_in_range<int8_t>(11, 11 + int8t_offset));
CALL_SUBTEST_7(check_all_in_range<int8_t>(-5, 5));
CALL_SUBTEST_7(check_all_in_range<int8_t>(-11 - int8t_offset, -11));
CALL_SUBTEST_7(check_all_in_range<int8_t>(-126, -126 + int8t_offset));
CALL_SUBTEST_7(check_all_in_range<int8_t>(126 - int8t_offset, 126));
CALL_SUBTEST_7(check_all_in_range<int8_t>(-126, 126));
CALL_SUBTEST_8(check_all_in_range<int16_t>(11, 11));
CALL_SUBTEST_8(check_all_in_range<int16_t>(11, 11 + int16t_offset));
CALL_SUBTEST_8(check_all_in_range<int16_t>(-5, 5));
CALL_SUBTEST_8(check_all_in_range<int16_t>(-11 - int16t_offset, -11));
CALL_SUBTEST_8(check_all_in_range<int16_t>(-24345, -24345 + int16t_offset));
CALL_SUBTEST_8(check_all_in_range<int16_t>(24345, 24345 + int16t_offset));
CALL_SUBTEST_9(check_all_in_range<int32_t>(11, 11));
CALL_SUBTEST_9(check_all_in_range<int32_t>(11, 11 + g_repeat));
CALL_SUBTEST_9(check_all_in_range<int32_t>(-5, 5));
CALL_SUBTEST_9(check_all_in_range<int32_t>(-11 - g_repeat, -11));
CALL_SUBTEST_9(check_all_in_range<int32_t>(-673456, -673456 + g_repeat));
CALL_SUBTEST_9(check_all_in_range<int32_t>(673456, 673456 + g_repeat));
CALL_SUBTEST_10(check_all_in_range<int64_t>(11, 11));
CALL_SUBTEST_10(check_all_in_range<int64_t>(11, 11 + g_repeat));
CALL_SUBTEST_10(check_all_in_range<int64_t>(-5, 5));
CALL_SUBTEST_10(check_all_in_range<int64_t>(-11 - g_repeat, -11));
CALL_SUBTEST_10(check_all_in_range<int64_t>(-int64_ref, -int64_ref + g_repeat));
CALL_SUBTEST_10(check_all_in_range<int64_t>(int64_ref, int64_ref + g_repeat));
CALL_SUBTEST_11(check_histogram<int32_t>(-5, 5, 11));
int bins = 100;
EIGEN_UNUSED_VARIABLE(bins)
CALL_SUBTEST_11(check_histogram<int32_t>(-3333, -3333 + bins * (3333 / bins) - 1, bins));
bins = 1000;
CALL_SUBTEST_11(check_histogram<int32_t>(-RAND_MAX + 10, -RAND_MAX + 10 + bins * (RAND_MAX / bins) - 1, bins));
CALL_SUBTEST_11(check_histogram<int32_t>(-RAND_MAX + 10,
-int64_t(RAND_MAX) + 10 + bins * (2 * int64_t(RAND_MAX) / bins) - 1, bins));
CALL_SUBTEST_12(check_histogram<uint8_t>(/*bins=*/16));
CALL_SUBTEST_12(check_histogram<uint16_t>(/*bins=*/1024));
CALL_SUBTEST_12(check_histogram<uint32_t>(/*bins=*/1024));
CALL_SUBTEST_12(check_histogram<uint64_t>(/*bins=*/1024));
CALL_SUBTEST_13(check_histogram<int8_t>(/*bins=*/16));
CALL_SUBTEST_13(check_histogram<int16_t>(/*bins=*/1024));
CALL_SUBTEST_13(check_histogram<int32_t>(/*bins=*/1024));
CALL_SUBTEST_13(check_histogram<int64_t>(/*bins=*/1024));
CALL_SUBTEST_14(check_histogram<float>(-10.0f, 10.0f, /*bins=*/1024));
CALL_SUBTEST_14(check_histogram<double>(-10.0, 10.0, /*bins=*/1024));
CALL_SUBTEST_14(check_histogram<long double>(-10.0L, 10.0L, /*bins=*/1024));
CALL_SUBTEST_14(check_histogram<half>(half(-10.0f), half(10.0f), /*bins=*/512));
CALL_SUBTEST_14(check_histogram<bfloat16>(bfloat16(-10.0f), bfloat16(10.0f), /*bins=*/64));
CALL_SUBTEST_15(check_histogram<float>(/*bins=*/1024));
CALL_SUBTEST_15(check_histogram<double>(/*bins=*/1024));
CALL_SUBTEST_15(check_histogram<long double>(/*bins=*/1024));
CALL_SUBTEST_15(check_histogram<half>(/*bins=*/512));
CALL_SUBTEST_15(check_histogram<bfloat16>(/*bins=*/64));
}