| // Copyright 2011 The Chromium Authors |
| // Use of this source code is governed by a BSD-style license that can be |
| // found in the LICENSE file. |
| |
| #include "base/rand_util.h" |
| |
| #include <limits.h> |
| #include <math.h> |
| #include <stdint.h> |
| |
| #include <algorithm> |
| #include <atomic> |
| #include <limits> |
| |
| #include "base/check.h" |
| #include "base/check_op.h" |
| #include "base/containers/span.h" |
| #include "base/time/time.h" |
| |
| namespace base { |
| |
| namespace { |
| |
| // A MetricSubsampler instance is not thread-safe. However, the global |
| // sampling state may be read concurrently with writing it via testing |
| // scopers, hence the need to use atomics. All operations use |
| // memory_order_relaxed because there are no dependent memory accesses. |
| std::atomic<bool> g_subsampling_always_sample = false; |
| std::atomic<bool> g_subsampling_never_sample = false; |
| |
| MetricsSubSampler* GetSharedSubsampler() { |
| static thread_local MetricsSubSampler g_shared_subsampler; |
| return &g_shared_subsampler; |
| } |
| |
| } // namespace |
| |
| uint64_t RandUint64() { |
| uint64_t number; |
| RandBytes(base::byte_span_from_ref(number)); |
| return number; |
| } |
| |
| int RandInt(int min, int max) { |
| DCHECK_LE(min, max); |
| |
| uint64_t range = static_cast<uint64_t>(max) - static_cast<uint64_t>(min) + 1; |
| // |range| is at most UINT_MAX + 1, so the result of RandGenerator(range) |
| // is at most UINT_MAX. Hence it's safe to cast it from uint64_t to int64_t. |
| int result = |
| static_cast<int>(min + static_cast<int64_t>(base::RandGenerator(range))); |
| DCHECK_GE(result, min); |
| DCHECK_LE(result, max); |
| return result; |
| } |
| |
| double RandDouble() { |
| return BitsToOpenEndedUnitInterval(base::RandUint64()); |
| } |
| |
| float RandFloat() { |
| return BitsToOpenEndedUnitIntervalF(base::RandUint64()); |
| } |
| |
| bool RandBool() { |
| uint8_t number; |
| RandBytes(span_from_ref(number)); |
| return number & 1; |
| } |
| |
| TimeDelta RandTimeDelta(TimeDelta start, TimeDelta limit) { |
| // We must have a finite, non-empty, non-reversed interval. |
| CHECK_LT(start, limit); |
| CHECK(!start.is_min()); |
| CHECK(!limit.is_max()); |
| |
| const int64_t range = (limit - start).InMicroseconds(); |
| // Because of the `CHECK_LT()` above, range > 0, so this cast is safe. |
| const uint64_t delta_us = base::RandGenerator(static_cast<uint64_t>(range)); |
| // ...and because `range` fit in an `int64_t`, so will `delta_us`. |
| return start + Microseconds(static_cast<int64_t>(delta_us)); |
| } |
| |
| TimeDelta RandTimeDeltaUpTo(TimeDelta limit) { |
| return RandTimeDelta(TimeDelta(), limit); |
| } |
| |
| double BitsToOpenEndedUnitInterval(uint64_t bits) { |
| // We try to get maximum precision by masking out as many bits as will fit |
| // in the target type's mantissa, and raising it to an appropriate power to |
| // produce output in the range [0, 1). For IEEE 754 doubles, the mantissa |
| // is expected to accommodate 53 bits (including the implied bit). |
| static_assert(std::numeric_limits<double>::radix == 2, |
| "otherwise use scalbn"); |
| constexpr int kBits = std::numeric_limits<double>::digits; |
| return ldexp(bits & ((UINT64_C(1) << kBits) - 1u), -kBits); |
| } |
| |
| float BitsToOpenEndedUnitIntervalF(uint64_t bits) { |
| // We try to get maximum precision by masking out as many bits as will fit |
| // in the target type's mantissa, and raising it to an appropriate power to |
| // produce output in the range [0, 1). For IEEE 754 floats, the mantissa is |
| // expected to accommodate 12 bits (including the implied bit). |
| static_assert(std::numeric_limits<float>::radix == 2, "otherwise use scalbn"); |
| constexpr int kBits = std::numeric_limits<float>::digits; |
| return ldexpf(bits & ((UINT64_C(1) << kBits) - 1u), -kBits); |
| } |
| |
| uint64_t RandGenerator(uint64_t range) { |
| DCHECK_GT(range, 0u); |
| // We must discard random results above this number, as they would |
| // make the random generator non-uniform (consider e.g. if |
| // MAX_UINT64 was 7 and |range| was 5, then a result of 1 would be twice |
| // as likely as a result of 3 or 4). |
| uint64_t max_acceptable_value = |
| (std::numeric_limits<uint64_t>::max() / range) * range - 1; |
| |
| uint64_t value; |
| do { |
| value = base::RandUint64(); |
| } while (value > max_acceptable_value); |
| |
| return value % range; |
| } |
| |
| std::string RandBytesAsString(size_t length) { |
| std::string result(length, '\0'); |
| RandBytes(as_writable_byte_span(result)); |
| return result; |
| } |
| |
| std::vector<uint8_t> RandBytesAsVector(size_t length) { |
| std::vector<uint8_t> result(length); |
| RandBytes(result); |
| return result; |
| } |
| |
| InsecureRandomGenerator::InsecureRandomGenerator() { |
| a_ = base::RandUint64(); |
| b_ = base::RandUint64(); |
| } |
| |
| void InsecureRandomGenerator::ReseedForTesting(uint64_t seed) { |
| a_ = seed; |
| b_ = seed; |
| } |
| |
| uint64_t InsecureRandomGenerator::RandUint64() const { |
| // Using XorShift128+, which is simple and widely used. See |
| // https://en.wikipedia.org/wiki/Xorshift#xorshift+ for details. |
| uint64_t t = a_; |
| const uint64_t s = b_; |
| |
| a_ = s; |
| t ^= t << 23; |
| t ^= t >> 17; |
| t ^= s ^ (s >> 26); |
| b_ = t; |
| |
| return t + s; |
| } |
| |
| uint32_t InsecureRandomGenerator::RandUint32() const { |
| // The generator usually returns an uint64_t, truncate it. |
| // |
| // It is noted in this paper (https://arxiv.org/abs/1810.05313) that the |
| // lowest 32 bits fail some statistical tests from the Big Crush |
| // suite. Use the higher ones instead. |
| return this->RandUint64() >> 32; |
| } |
| |
| double InsecureRandomGenerator::RandDouble() const { |
| uint64_t x = RandUint64(); |
| // From https://vigna.di.unimi.it/xorshift/. |
| // 53 bits of mantissa, hence the "hexadecimal exponent" 1p-53. |
| return (x >> 11) * 0x1.0p-53; |
| } |
| |
| MetricsSubSampler::MetricsSubSampler() = default; |
| bool MetricsSubSampler::ShouldSample(double probability) const { |
| if (g_subsampling_always_sample.load(std::memory_order_relaxed)) { |
| return true; |
| } |
| if (g_subsampling_never_sample.load(std::memory_order_relaxed)) { |
| return false; |
| } |
| |
| DCHECK(probability >= 0 && probability <= 1); |
| return generator_.RandDouble() < probability; |
| } |
| |
| void MetricsSubSampler::Reseed() { |
| generator_ = InsecureRandomGenerator(); |
| } |
| |
| MetricsSubSampler::ScopedAlwaysSampleForTesting:: |
| ScopedAlwaysSampleForTesting() { |
| DCHECK(!g_subsampling_always_sample.load(std::memory_order_relaxed)); |
| DCHECK(!g_subsampling_never_sample.load(std::memory_order_relaxed)); |
| g_subsampling_always_sample.store(true, std::memory_order_relaxed); |
| } |
| |
| MetricsSubSampler::ScopedAlwaysSampleForTesting:: |
| ~ScopedAlwaysSampleForTesting() { |
| DCHECK(g_subsampling_always_sample.load(std::memory_order_relaxed)); |
| DCHECK(!g_subsampling_never_sample.load(std::memory_order_relaxed)); |
| g_subsampling_always_sample.store(false, std::memory_order_relaxed); |
| } |
| |
| MetricsSubSampler::ScopedNeverSampleForTesting::ScopedNeverSampleForTesting() { |
| DCHECK(!g_subsampling_always_sample.load(std::memory_order_relaxed)); |
| DCHECK(!g_subsampling_never_sample.load(std::memory_order_relaxed)); |
| g_subsampling_never_sample.store(true, std::memory_order_relaxed); |
| } |
| |
| MetricsSubSampler::ScopedNeverSampleForTesting::~ScopedNeverSampleForTesting() { |
| DCHECK(!g_subsampling_always_sample); |
| DCHECK(g_subsampling_never_sample); |
| g_subsampling_never_sample.store(false, std::memory_order_relaxed); |
| } |
| |
| bool ShouldRecordSubsampledMetric(double probability) { |
| return GetSharedSubsampler()->ShouldSample(probability); |
| } |
| |
| void ReseedSharedMetricsSubsampler() { |
| GetSharedSubsampler()->Reseed(); |
| } |
| |
| } // namespace base |