diff --git a/mfbt/Saturate.h b/mfbt/Saturate.h
new file mode 100644
index 00000000000..5d3315e17be
--- /dev/null
+++ b/mfbt/Saturate.h
@@ -0,0 +1,287 @@
+/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
+/* vim: set ts=8 sts=2 et sw=2 tw=80: */
+/* 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/. */
+
+/* Provides saturation arithmetics for scalar types. */
+
+#ifndef mozilla_Saturate_h
+#define mozilla_Saturate_h
+
+#include "mozilla/Attributes.h"
+#include "mozilla/Move.h"
+#include "mozilla/NumericLimits.h"
+#include "mozilla/TypeTraits.h"
+
+namespace mozilla {
+namespace detail {
+
+/**
+ * |SaturateOp<T>| wraps scalar values for saturation arithmetics. Usage:
+ *
+ *    uint32_t value = 1;
+ *
+ *    ++SaturateOp<uint32_t>(value); // value is 2
+ *    --SaturateOp<uint32_t>(value); // value is 1
+ *    --SaturateOp<uint32_t>(value); // value is 0
+ *    --SaturateOp<uint32_t>(value); // value is still 0
+ *
+ * Please add new operators when required.
+ *
+ * |SaturateOp<T>| will saturate at the minimum and maximum values of
+ * type T. If you need other bounds, implement a clamped-type class and
+ * specialize the type traits accordingly.
+ */
+template <typename T>
+class SaturateOp
+{
+public:
+  explicit SaturateOp(T& aValue)
+    : mValue(aValue)
+  {
+    // We should actually check for |std::is_scalar<T>::value| to be
+    // true, but this type trait is not available everywhere. Relax
+    // this assertion if you want to use floating point values as well.
+    static_assert(IsIntegral<T>::value,
+                  "Integral type required in instantiation");
+  }
+
+  // Add and subtract operators
+
+  T operator+(const T& aRhs) const
+  {
+    return T(mValue) += aRhs;
+  }
+
+  T operator-(const T& aRhs) const
+  {
+    return T(mValue) -= aRhs;
+  }
+
+  // Compound operators
+
+  const T& operator+=(const T& aRhs) const
+  {
+    const T min = NumericLimits<T>::min();
+    const T max = NumericLimits<T>::max();
+
+    if (aRhs > static_cast<T>(0)) {
+      mValue = (max - aRhs) < mValue ? max : mValue + aRhs;
+    } else {
+      mValue = (min - aRhs) > mValue ? min : mValue + aRhs;
+    }
+    return mValue;
+  }
+
+  const T& operator-=(const T& aRhs) const
+  {
+    const T min = NumericLimits<T>::min();
+    const T max = NumericLimits<T>::max();
+
+    if (aRhs > static_cast<T>(0)) {
+      mValue = (min + aRhs) > mValue ? min : mValue - aRhs;
+    } else {
+      mValue = (max + aRhs) < mValue ? max : mValue - aRhs;
+    }
+    return mValue;
+  }
+
+  // Increment and decrement operators
+
+  const T& operator++() const // prefix
+  {
+    return operator+=(static_cast<T>(1));
+  }
+
+  T operator++(int) const // postfix
+  {
+    const T value(mValue);
+    operator++();
+    return value;
+  }
+
+  const T& operator--() const // prefix
+  {
+    return operator-=(static_cast<T>(1));
+  }
+
+  T operator--(int) const // postfix
+  {
+    const T value(mValue);
+    operator--();
+    return value;
+  }
+
+private:
+  SaturateOp(const SaturateOp<T>&) = delete;
+  SaturateOp(SaturateOp<T>&&) = delete;
+  SaturateOp& operator=(const SaturateOp<T>&) = delete;
+  SaturateOp& operator=(SaturateOp<T>&&) = delete;
+
+  T& mValue;
+};
+
+/**
+ * |Saturate<T>| is a value type for saturation arithmetics. It's
+ * build on top of |SaturateOp<T>|.
+ */
+template <typename T>
+class Saturate
+{
+public:
+  Saturate() = default;
+  MOZ_IMPLICIT Saturate(const Saturate<T>&) = default;
+
+  MOZ_IMPLICIT Saturate(Saturate<T>&& aValue)
+  {
+    mValue = Move(aValue.mValue);
+  }
+
+  explicit Saturate(const T& aValue)
+    : mValue(aValue)
+  { }
+
+  const T& value() const
+  {
+    return mValue;
+  }
+
+  // Compare operators
+
+  bool operator==(const Saturate<T>& aRhs) const
+  {
+    return mValue == aRhs.mValue;
+  }
+
+  bool operator!=(const Saturate<T>& aRhs) const
+  {
+    return !operator==(aRhs);
+  }
+
+  bool operator==(const T& aRhs) const
+  {
+    return mValue == aRhs;
+  }
+
+  bool operator!=(const T& aRhs) const
+  {
+    return !operator==(aRhs);
+  }
+
+  // Assignment operators
+
+  Saturate<T>& operator=(const Saturate<T>&) = default;
+
+  Saturate<T>& operator=(Saturate<T>&& aRhs)
+  {
+    mValue = Move(aRhs.mValue);
+    return *this;
+  }
+
+  // Add and subtract operators
+
+  Saturate<T> operator+(const Saturate<T>& aRhs) const
+  {
+    Saturate<T> lhs(mValue);
+    return lhs += aRhs.mValue;
+  }
+
+  Saturate<T> operator+(const T& aRhs) const
+  {
+    Saturate<T> lhs(mValue);
+    return lhs += aRhs;
+  }
+
+  Saturate<T> operator-(const Saturate<T>& aRhs) const
+  {
+    Saturate<T> lhs(mValue);
+    return lhs -= aRhs.mValue;
+  }
+
+  Saturate<T> operator-(const T& aRhs) const
+  {
+    Saturate<T> lhs(mValue);
+    return lhs -= aRhs;
+  }
+
+  // Compound operators
+
+  Saturate<T>& operator+=(const Saturate<T>& aRhs)
+  {
+    SaturateOp<T>(mValue) += aRhs.mValue;
+    return *this;
+  }
+
+  Saturate<T>& operator+=(const T& aRhs)
+  {
+    SaturateOp<T>(mValue) += aRhs;
+    return *this;
+  }
+
+  Saturate<T>& operator-=(const Saturate<T>& aRhs)
+  {
+    SaturateOp<T>(mValue) -= aRhs.mValue;
+    return *this;
+  }
+
+  Saturate<T>& operator-=(const T& aRhs)
+  {
+    SaturateOp<T>(mValue) -= aRhs;
+    return *this;
+  }
+
+  // Increment and decrement operators
+
+  Saturate<T>& operator++() // prefix
+  {
+    ++SaturateOp<T>(mValue);
+    return *this;
+  }
+
+  Saturate<T> operator++(int) // postfix
+  {
+    return Saturate<T>(SaturateOp<T>(mValue)++);
+  }
+
+  Saturate<T>& operator--() // prefix
+  {
+    --SaturateOp<T>(mValue);
+    return *this;
+  }
+
+  Saturate<T> operator--(int) // postfix
+  {
+    return Saturate<T>(SaturateOp<T>(mValue)--);
+  }
+
+private:
+  T mValue;
+};
+
+} // namespace detail
+
+typedef detail::Saturate<int8_t> SaturateInt8;
+typedef detail::Saturate<int16_t> SaturateInt16;
+typedef detail::Saturate<int32_t> SaturateInt32;
+typedef detail::Saturate<uint8_t> SaturateUint8;
+typedef detail::Saturate<uint16_t> SaturateUint16;
+typedef detail::Saturate<uint32_t> SaturateUint32;
+
+} // namespace mozilla
+
+template<typename LhsT, typename RhsT>
+bool
+operator==(LhsT aLhs, const mozilla::detail::Saturate<RhsT>& aRhs)
+{
+  return aRhs.operator==(static_cast<RhsT>(aLhs));
+}
+
+template<typename LhsT, typename RhsT>
+bool
+operator!=(LhsT aLhs, const mozilla::detail::Saturate<RhsT>& aRhs)
+{
+  return !(aLhs == aRhs);
+}
+
+#endif // mozilla_Saturate_h
diff --git a/mfbt/moz.build b/mfbt/moz.build
index f6acbe8d5c5..38d424dd2ed 100644
--- a/mfbt/moz.build
+++ b/mfbt/moz.build
@@ -73,6 +73,7 @@ EXPORTS.mozilla = [
     'RefPtr.h',
     'ReverseIterator.h',
     'RollingMean.h',
+    'Saturate.h',
     'Scoped.h',
     'ScopeExit.h',
     'SegmentedVector.h',
diff --git a/mfbt/tests/TestSaturate.cpp b/mfbt/tests/TestSaturate.cpp
new file mode 100644
index 00000000000..cd1208ff8f3
--- /dev/null
+++ b/mfbt/tests/TestSaturate.cpp
@@ -0,0 +1,215 @@
+/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
+/* vim: set ts=8 sts=2 et sw=2 tw=80: */
+/* 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 <mozilla/Saturate.h>
+
+#include <mozilla/Assertions.h>
+#include <mozilla/NumericLimits.h>
+
+using mozilla::detail::Saturate;
+using mozilla::NumericLimits;
+
+#define A(a) MOZ_RELEASE_ASSERT(a, "Test \'" #a "\'  failed.")
+
+static const unsigned long sNumOps = 32;
+
+template<typename T>
+static T
+StartValue()
+{
+  // Specialize |StartValue| for the given type.
+  A(false);
+}
+
+template<>
+int8_t
+StartValue<int8_t>()
+{
+  return 0;
+}
+
+template<>
+int16_t
+StartValue<int16_t>()
+{
+  return 0;
+}
+
+template<>
+int32_t
+StartValue<int32_t>()
+{
+  return 0;
+}
+
+template<>
+uint8_t
+StartValue<uint8_t>()
+{
+  // Picking a value near middle of uint8_t's range.
+  return static_cast<uint8_t>(NumericLimits<int8_t>::max());
+}
+
+template<>
+uint16_t
+StartValue<uint16_t>()
+{
+  // Picking a value near middle of uint16_t's range.
+  return static_cast<uint8_t>(NumericLimits<int16_t>::max());
+}
+
+template<>
+uint32_t
+StartValue<uint32_t>()
+{
+  // Picking a value near middle of uint32_t's range.
+  return static_cast<uint8_t>(NumericLimits<int32_t>::max());
+}
+
+// Add
+//
+
+template<typename T>
+static void
+TestPrefixIncr()
+{
+  T value = StartValue<T>();
+  Saturate<T> satValue(value);
+
+  for (T i = 0; i < static_cast<T>(sNumOps); ++i) {
+    A(++value == ++satValue);
+  }
+}
+
+template<typename T>
+static void
+TestPostfixIncr()
+{
+  T value = StartValue<T>();
+  Saturate<T> satValue(value);
+
+  for (T i = 0; i < static_cast<T>(sNumOps); ++i) {
+    A(value++ == satValue++);
+  }
+}
+
+template<typename T>
+static void
+TestAdd()
+{
+  T value = StartValue<T>();
+  Saturate<T> satValue(value);
+
+  for (T i = 0; i < static_cast<T>(sNumOps); ++i) {
+    A((value + i) == (satValue + i));
+  }
+}
+
+// Subtract
+//
+
+template<typename T>
+static void
+TestPrefixDecr()
+{
+  T value = StartValue<T>();
+  Saturate<T> satValue(value);
+
+  for (T i = 0; i < static_cast<T>(sNumOps); ++i) {
+    A(--value == --satValue);
+  }
+}
+
+template<typename T>
+static void
+TestPostfixDecr()
+{
+  T value = StartValue<T>();
+  Saturate<T> satValue(value);
+
+  for (T i = 0; i < static_cast<T>(sNumOps); ++i) {
+    A(value-- == satValue--);
+  }
+}
+
+template<typename T>
+static void
+TestSub()
+{
+  T value = StartValue<T>();
+  Saturate<T> satValue(value);
+
+  for (T i = 0; i < static_cast<T>(sNumOps); ++i) {
+    A((value - i) == (satValue - i));
+  }
+}
+
+// Corner cases near bounds
+//
+
+template<typename T>
+static void
+TestUpperBound()
+{
+  Saturate<T> satValue(NumericLimits<T>::max());
+
+  A(--satValue == (NumericLimits<T>::max() - 1));
+  A(++satValue == (NumericLimits<T>::max()));
+  A(++satValue == (NumericLimits<T>::max())); // don't overflow here
+  A(++satValue == (NumericLimits<T>::max())); // don't overflow here
+  A(--satValue == (NumericLimits<T>::max() - 1)); // back at (max - 1)
+  A(--satValue == (NumericLimits<T>::max() - 2));
+}
+
+template<typename T>
+static void
+TestLowerBound()
+{
+  Saturate<T> satValue(NumericLimits<T>::min());
+
+  A(++satValue == (NumericLimits<T>::min() + 1));
+  A(--satValue == (NumericLimits<T>::min()));
+  A(--satValue == (NumericLimits<T>::min())); // don't overflow here
+  A(--satValue == (NumericLimits<T>::min())); // don't overflow here
+  A(++satValue == (NumericLimits<T>::min() + 1)); // back at (max + 1)
+  A(++satValue == (NumericLimits<T>::min() + 2));
+}
+
+// Framework
+//
+
+template<typename T>
+static void
+TestAll()
+{
+  // Assert that we don't accidently hit type's range limits in tests.
+  const T value = StartValue<T>();
+  A(NumericLimits<T>::min() + static_cast<T>(sNumOps) <= value);
+  A(NumericLimits<T>::max() - static_cast<T>(sNumOps) >= value);
+
+  TestPrefixIncr<T>();
+  TestPostfixIncr<T>();
+  TestAdd<T>();
+
+  TestPrefixDecr<T>();
+  TestPostfixDecr<T>();
+  TestSub<T>();
+
+  TestUpperBound<T>();
+  TestLowerBound<T>();
+}
+
+int
+main()
+{
+  TestAll<int8_t>();
+  TestAll<int16_t>();
+  TestAll<int32_t>();
+  TestAll<uint8_t>();
+  TestAll<uint16_t>();
+  TestAll<uint32_t>();
+  return 0;
+}
diff --git a/mfbt/tests/moz.build b/mfbt/tests/moz.build
index c75ee8eec15..499cd730de4 100644
--- a/mfbt/tests/moz.build
+++ b/mfbt/tests/moz.build
@@ -31,6 +31,7 @@ CppUnitTests([
     'TestPair',
     'TestRefPtr',
     'TestRollingMean',
+    'TestSaturate',
     'TestScopeExit',
     'TestSegmentedVector',
     'TestSHA1',
diff --git a/testing/cppunittest.ini b/testing/cppunittest.ini
index 05899358c4b..a3e5da5d317 100644
--- a/testing/cppunittest.ini
+++ b/testing/cppunittest.ini
@@ -67,6 +67,7 @@ skip-if = os == 'android' # Bug 1147630
 [TestRollingMean]
 [TestSHA1]
 [TestSTSParser]
+[TestSaturate]
 [TestSplayTree]
 [TestStartupCache]
 skip-if = os == 'b2g' || os == 'android'  # Bug 929655