| /* | 
 |  *  Copyright (c) 2015 The WebRTC project authors. All Rights Reserved. | 
 |  * | 
 |  *  Use of this source code is governed by a BSD-style license | 
 |  *  that can be found in the LICENSE file in the root of the source | 
 |  *  tree. An additional intellectual property rights grant can be found | 
 |  *  in the file PATENTS.  All contributing project authors may | 
 |  *  be found in the AUTHORS file in the root of the source tree. | 
 |  */ | 
 | #ifndef SYSTEM_WRAPPERS_INCLUDE_NTP_TIME_H_ | 
 | #define SYSTEM_WRAPPERS_INCLUDE_NTP_TIME_H_ | 
 |  | 
 | #include <cmath> | 
 | #include <cstdint> | 
 | #include <limits> | 
 |  | 
 | #include "rtc_base/numerics/safe_conversions.h" | 
 |  | 
 | namespace webrtc { | 
 |  | 
 | class NtpTime { | 
 |  public: | 
 |   static constexpr uint64_t kFractionsPerSecond = 0x100000000; | 
 |   NtpTime() : value_(0) {} | 
 |   explicit NtpTime(uint64_t value) : value_(value) {} | 
 |   NtpTime(uint32_t seconds, uint32_t fractions) | 
 |       : value_(seconds * kFractionsPerSecond + fractions) {} | 
 |  | 
 |   NtpTime(const NtpTime&) = default; | 
 |   NtpTime& operator=(const NtpTime&) = default; | 
 |   explicit operator uint64_t() const { return value_; } | 
 |  | 
 |   void Set(uint32_t seconds, uint32_t fractions) { | 
 |     value_ = seconds * kFractionsPerSecond + fractions; | 
 |   } | 
 |   void Reset() { value_ = 0; } | 
 |  | 
 |   int64_t ToMs() const { | 
 |     static constexpr double kNtpFracPerMs = 4.294967296E6;  // 2^32 / 1000. | 
 |     const double frac_ms = static_cast<double>(fractions()) / kNtpFracPerMs; | 
 |     return 1000 * static_cast<int64_t>(seconds()) + | 
 |            static_cast<int64_t>(frac_ms + 0.5); | 
 |   } | 
 |   // NTP standard (RFC1305, section 3.1) explicitly state value 0 is invalid. | 
 |   bool Valid() const { return value_ != 0; } | 
 |  | 
 |   uint32_t seconds() const { | 
 |     return dchecked_cast<uint32_t>(value_ / kFractionsPerSecond); | 
 |   } | 
 |   uint32_t fractions() const { | 
 |     return dchecked_cast<uint32_t>(value_ % kFractionsPerSecond); | 
 |   } | 
 |  | 
 |  private: | 
 |   uint64_t value_; | 
 | }; | 
 |  | 
 | inline bool operator==(const NtpTime& n1, const NtpTime& n2) { | 
 |   return static_cast<uint64_t>(n1) == static_cast<uint64_t>(n2); | 
 | } | 
 | inline bool operator!=(const NtpTime& n1, const NtpTime& n2) { | 
 |   return !(n1 == n2); | 
 | } | 
 |  | 
 | // Converts `int64_t` milliseconds to Q32.32-formatted fixed-point seconds. | 
 | // Performs clamping if the result overflows or underflows. | 
 | inline int64_t Int64MsToQ32x32(int64_t milliseconds) { | 
 |   // TODO(bugs.webrtc.org/10893): Change to use `rtc::saturated_cast` once the | 
 |   // bug has been fixed. | 
 |   double result = | 
 |       std::round(milliseconds * (NtpTime::kFractionsPerSecond / 1000.0)); | 
 |  | 
 |   // Explicitly cast values to double to avoid implicit conversion warnings | 
 |   // The conversion of the std::numeric_limits<int64_t>::max() triggers | 
 |   // -Wimplicit-int-float-conversion warning in clang 10.0.0 without explicit | 
 |   // cast | 
 |   if (result <= static_cast<double>(std::numeric_limits<int64_t>::min())) { | 
 |     return std::numeric_limits<int64_t>::min(); | 
 |   } | 
 |  | 
 |   if (result >= static_cast<double>(std::numeric_limits<int64_t>::max())) { | 
 |     return std::numeric_limits<int64_t>::max(); | 
 |   } | 
 |  | 
 |   return dchecked_cast<int64_t>(result); | 
 | } | 
 |  | 
 | // Converts `int64_t` milliseconds to UQ32.32-formatted fixed-point seconds. | 
 | // Performs clamping if the result overflows or underflows. | 
 | inline uint64_t Int64MsToUQ32x32(int64_t milliseconds) { | 
 |   // TODO(bugs.webrtc.org/10893): Change to use `rtc::saturated_cast` once the | 
 |   // bug has been fixed. | 
 |   double result = | 
 |       std::round(milliseconds * (NtpTime::kFractionsPerSecond / 1000.0)); | 
 |  | 
 |   // Explicitly cast values to double to avoid implicit conversion warnings | 
 |   // The conversion of the std::numeric_limits<int64_t>::max() triggers | 
 |   // -Wimplicit-int-float-conversion warning in clang 10.0.0 without explicit | 
 |   // cast | 
 |   if (result <= static_cast<double>(std::numeric_limits<uint64_t>::min())) { | 
 |     return std::numeric_limits<uint64_t>::min(); | 
 |   } | 
 |  | 
 |   if (result >= static_cast<double>(std::numeric_limits<uint64_t>::max())) { | 
 |     return std::numeric_limits<uint64_t>::max(); | 
 |   } | 
 |  | 
 |   return dchecked_cast<uint64_t>(result); | 
 | } | 
 |  | 
 | // Converts Q32.32-formatted fixed-point seconds to `int64_t` milliseconds. | 
 | inline int64_t Q32x32ToInt64Ms(int64_t q32x32) { | 
 |   return dchecked_cast<int64_t>( | 
 |       std::round(q32x32 * (1000.0 / NtpTime::kFractionsPerSecond))); | 
 | } | 
 |  | 
 | // Converts UQ32.32-formatted fixed-point seconds to `int64_t` milliseconds. | 
 | inline int64_t UQ32x32ToInt64Ms(uint64_t q32x32) { | 
 |   return dchecked_cast<int64_t>( | 
 |       std::round(q32x32 * (1000.0 / NtpTime::kFractionsPerSecond))); | 
 | } | 
 |  | 
 | // Converts UQ32.32-formatted fixed-point seconds to `int64_t` microseconds. | 
 | inline int64_t UQ32x32ToInt64Us(uint64_t q32x32) { | 
 |   return dchecked_cast<int64_t>( | 
 |       std::round(q32x32 * (1'000'000.0 / NtpTime::kFractionsPerSecond))); | 
 | } | 
 |  | 
 | // Converts Q32.32-formatted fixed-point seconds to `int64_t` microseconds. | 
 | inline int64_t Q32x32ToInt64Us(int64_t q32x32) { | 
 |   return dchecked_cast<int64_t>( | 
 |       std::round(q32x32 * (1'000'000.0 / NtpTime::kFractionsPerSecond))); | 
 | } | 
 |  | 
 | }  // namespace webrtc | 
 | #endif  // SYSTEM_WRAPPERS_INCLUDE_NTP_TIME_H_ |