diff --git a/doc/sf/ccmath.qbk b/doc/sf/ccmath.qbk
index d0d6a41a5a..c53b063e57 100644
--- a/doc/sf/ccmath.qbk
+++ b/doc/sf/ccmath.qbk
@@ -149,6 +149,12 @@ All of the following functions require C++17 or greater.
         template <typename Arithmetic1, typename Arithmetic2>
         inline constexpr Promoted hypot(Arithmetic1 x, Arithmetic2 y) noexcept
 
+        template <typename Real>
+        inline constexpr Real fdim(Real x, Real y) noexcept
+
+        template <typename Arithmetic1, typename Arithmetic2>
+        inline constexpr Promoted fdim(Arithmetic1 x, Arithmetic2 y) noexcept
+
     } // Namespaces
 
 [endsect] [/section:ccmath Constexpr CMath]
diff --git a/include/boost/math/ccmath/ccmath.hpp b/include/boost/math/ccmath/ccmath.hpp
index b2ef415d09..6f02ca6022 100644
--- a/include/boost/math/ccmath/ccmath.hpp
+++ b/include/boost/math/ccmath/ccmath.hpp
@@ -30,5 +30,6 @@
 #include <boost/math/ccmath/remainder.hpp>
 #include <boost/math/ccmath/copysign.hpp>
 #include <boost/math/ccmath/hypot.hpp>
+#include <boost/math/ccmath/fdim.hpp>
 
 #endif // BOOST_MATH_CCMATH_HPP
diff --git a/include/boost/math/ccmath/fdim.hpp b/include/boost/math/ccmath/fdim.hpp
new file mode 100644
index 0000000000..5b03d05c89
--- /dev/null
+++ b/include/boost/math/ccmath/fdim.hpp
@@ -0,0 +1,100 @@
+//  (C) Copyright Matt Borland 2022.
+//  Use, modification and distribution are subject to the
+//  Boost Software License, Version 1.0. (See accompanying file
+//  LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
+
+#ifndef BOOST_MATH_CCMATH_FDIM_HPP
+#define BOOST_MATH_CCMATH_FDIM_HPP
+
+#include <cmath>
+#include <limits>
+#include <type_traits>
+#include <boost/math/tools/is_constant_evaluated.hpp>
+#include <boost/math/ccmath/isnan.hpp>
+
+namespace boost::math::ccmath {
+
+namespace detail {
+
+template <typename T>
+inline constexpr T fdim_impl(const T x, const T y) noexcept
+{
+    if (x <= y)
+    {
+        return 0;
+    }
+    else if ((y < 0) && (x > (std::numeric_limits<T>::max)() + y))
+    {
+        return std::numeric_limits<T>::infinity();
+    }
+    else
+    {
+        return x - y;
+    }
+}
+
+} // Namespace detail
+
+template <typename Real, std::enable_if_t<!std::is_integral_v<Real>, bool> = true>
+inline constexpr Real fdim(Real x, Real y) noexcept
+{
+    if (BOOST_MATH_IS_CONSTANT_EVALUATED(x))
+    {
+        return boost::math::ccmath::isnan(x) ? std::numeric_limits<Real>::quiet_NaN() :
+               boost::math::ccmath::isnan(y) ? std::numeric_limits<Real>::quiet_NaN() :
+               boost::math::ccmath::detail::fdim_impl(x, y);
+    }
+    else
+    {
+        using std::fdim;
+        return fdim(x, y);
+    }
+}
+
+template <typename T1, typename T2>
+inline constexpr auto fdim(T1 x, T2 y) noexcept
+{
+    if (BOOST_MATH_IS_CONSTANT_EVALUATED(x))
+    {
+        // If the type is an integer (e.g. epsilon == 0) then set the epsilon value to 1 so that type is at a minimum 
+        // cast to double
+        constexpr auto T1p = std::numeric_limits<T1>::epsilon() > 0 ? std::numeric_limits<T1>::epsilon() : 1;
+        constexpr auto T2p = std::numeric_limits<T2>::epsilon() > 0 ? std::numeric_limits<T2>::epsilon() : 1;
+        
+        using promoted_type = 
+                              #ifndef BOOST_MATH_NO_LONG_DOUBLE_MATH_FUNCTIONS
+                              std::conditional_t<T1p <= LDBL_EPSILON && T1p <= T2p, T1,
+                              std::conditional_t<T2p <= LDBL_EPSILON && T2p <= T1p, T2,
+                              #endif
+                              std::conditional_t<T1p <= DBL_EPSILON && T1p <= T2p, T1,
+                              std::conditional_t<T2p <= DBL_EPSILON && T2p <= T1p, T2, double
+                              #ifndef BOOST_MATH_NO_LONG_DOUBLE_MATH_FUNCTIONS
+                              >>>>;
+                              #else
+                              >>;
+                              #endif
+
+        return boost::math::ccmath::fdim(promoted_type(x), promoted_type(y));
+    }
+    else
+    {
+        using std::fdim;
+        return fdim(x, y);
+    }
+}
+
+inline constexpr float fdimf(float x, float y) noexcept
+{
+    return boost::math::ccmath::fdim(x, y);
+}
+
+#ifndef BOOST_MATH_NO_LONG_DOUBLE_MATH_FUNCTIONS
+inline constexpr long double fdiml(long double x, long double y) noexcept
+{
+    return boost::math::ccmath::fdim(x, y);
+}
+#endif
+
+} // Namespace boost::math::ccmath
+
+#endif // BOOST_MATH_CCMATH_FDIM_HPP
diff --git a/test/Jamfile.v2 b/test/Jamfile.v2
index f6b633c860..63a3086a80 100644
--- a/test/Jamfile.v2
+++ b/test/Jamfile.v2
@@ -144,6 +144,7 @@ test-suite special_fun :
    [ run ccmath_remainder_test.cpp ../../test/build//boost_unit_test_framework : : : [ requires cxx17_if_constexpr ] ]
    [ run ccmath_copysign_test.cpp ../../test/build//boost_unit_test_framework : : : [ requires cxx17_if_constexpr ] ]
    [ run ccmath_hypot_test.cpp ../../test/build//boost_unit_test_framework : : : [ requires cxx17_if_constexpr ] ]
+   [ run ccmath_fdim_test.cpp ../../test/build//boost_unit_test_framework : : : [ requires cxx17_if_constexpr ] ]
    [ run log1p_expm1_test.cpp test_instances//test_instances pch_light ../../test/build//boost_unit_test_framework  ]
    [ run powm1_sqrtp1m1_test.cpp test_instances//test_instances pch_light ../../test/build//boost_unit_test_framework  ]
    [ run special_functions_test.cpp ../../test/build//boost_unit_test_framework  ]
diff --git a/test/ccmath_fdim_test.cpp b/test/ccmath_fdim_test.cpp
new file mode 100644
index 0000000000..f5f688d534
--- /dev/null
+++ b/test/ccmath_fdim_test.cpp
@@ -0,0 +1,54 @@
+//  (C) Copyright Matt Borland 2022.
+//  Use, modification and distribution are subject to the
+//  Boost Software License, Version 1.0. (See accompanying file
+//  LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
+
+#include <cmath>
+#include <cfloat>
+#include <cstdint>
+#include <limits>
+#include <type_traits>
+#include <boost/math/ccmath/fdim.hpp>
+#include <boost/math/ccmath/isnan.hpp>
+#include <boost/math/ccmath/isinf.hpp>
+
+#ifdef BOOST_HAS_FLOAT128
+#include <boost/multiprecision/float128.hpp>
+#endif
+
+template <typename T>
+constexpr void test()
+{
+    static_assert(boost::math::ccmath::isnan(boost::math::ccmath::fdim(std::numeric_limits<T>::quiet_NaN(), T(1))), "If x is NaN, NaN is returned");
+    static_assert(boost::math::ccmath::isnan(boost::math::ccmath::fdim(T(1), std::numeric_limits<T>::quiet_NaN())), "If y is NaN, NaN is returned");
+
+    static_assert(boost::math::ccmath::fdim(T(4), T(1)) == T(3));
+    static_assert(boost::math::ccmath::fdim(T(1), T(4)) == T(0));
+    static_assert(boost::math::ccmath::fdim(T(4), T(-1)) == T(5));
+    static_assert(boost::math::ccmath::fdim(T(1), T(-4)) == T(5));
+
+    static_assert(boost::math::ccmath::isinf(boost::math::ccmath::fdim(std::numeric_limits<T>::infinity(), T(-1))));
+}
+
+#if !defined(BOOST_MATH_NO_CONSTEXPR_DETECTION) && !defined(BOOST_MATH_USING_BUILTIN_CONSTANT_P)
+int main()
+{
+    test<float>();
+    test<double>();
+
+    #ifndef BOOST_MATH_NO_LONG_DOUBLE_MATH_FUNCTIONS
+    test<long double>();
+    #endif
+    
+    #ifdef BOOST_HAS_FLOAT128
+    test<boost::multiprecision::float128>();
+    #endif
+
+    return 0;
+}
+#else
+int main()
+{
+    return 0;
+}
+#endif
diff --git a/test/compile_test/ccmath_fdim_incl_test.cpp b/test/compile_test/ccmath_fdim_incl_test.cpp
new file mode 100644
index 0000000000..39e8874d2e
--- /dev/null
+++ b/test/compile_test/ccmath_fdim_incl_test.cpp
@@ -0,0 +1,16 @@
+//  (C) Copyright Matt Borland 2022.
+//  Use, modification and distribution are subject to the
+//  Boost Software License, Version 1.0. (See accompanying file
+//  LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
+
+#include <boost/math/ccmath/fdim.hpp>
+#include "test_compile_result.hpp"
+
+void compile_and_link_test()
+{
+   check_result<float>(boost::math::ccmath::fdim(1.0f, 1.0f));
+   check_result<double>(boost::math::ccmath::fdim(1.0, 1.0));
+#ifndef BOOST_MATH_NO_LONG_DOUBLE_MATH_FUNCTIONS
+   check_result<long double>(boost::math::ccmath::fdim(1.0l, 1.0l));
+#endif
+}