diff --git a/include/fast_float/ascii_number.h b/include/fast_float/ascii_number.h
index ee37664..6f3e69d 100644
--- a/include/fast_float/ascii_number.h
+++ b/include/fast_float/ascii_number.h
@@ -12,9 +12,11 @@ namespace fast_float {
 
 // Next function can be micro-optimized, but compilers are entirely
 // able to optimize it well.
-fastfloat_really_inline bool is_integer(char c)  noexcept  { return c >= '0' && c <= '9'; }
+fastfloat_really_inline constexpr bool is_integer(char c) noexcept {
+  return c >= '0' && c <= '9';
+}
 
-fastfloat_really_inline uint64_t byteswap(uint64_t val) {
+fastfloat_really_inline constexpr uint64_t byteswap(uint64_t val) {
   return (val & 0xFF00000000000000) >> 56
     | (val & 0x00FF000000000000) >> 40
     | (val & 0x0000FF0000000000) >> 24
@@ -44,7 +46,8 @@ fastfloat_really_inline void write_u64(uint8_t *chars, uint64_t val) {
 }
 
 // credit  @aqrit
-fastfloat_really_inline uint32_t  parse_eight_digits_unrolled(uint64_t val) {
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14
+uint32_t parse_eight_digits_unrolled(uint64_t val) {
   const uint64_t mask = 0x000000FF000000FF;
   const uint64_t mul1 = 0x000F424000000064; // 100 + (1000000ULL << 32)
   const uint64_t mul2 = 0x0000271000000001; // 1 + (10000ULL << 32)
@@ -59,7 +62,7 @@ fastfloat_really_inline uint32_t parse_eight_digits_unrolled(const char *chars)
 }
 
 // credit @aqrit
-fastfloat_really_inline bool is_made_of_eight_digits_fast(uint64_t val)  noexcept  {
+fastfloat_really_inline constexpr bool is_made_of_eight_digits_fast(uint64_t val)  noexcept  {
   return !((((val + 0x4646464646464646) | (val - 0x3030303030303030)) &
      0x8080808080808080));
 }
diff --git a/include/fast_float/bigint.h b/include/fast_float/bigint.h
index 220e057..073ddf8 100644
--- a/include/fast_float/bigint.h
+++ b/include/fast_float/bigint.h
@@ -54,23 +54,23 @@ struct stackvec {
     FASTFLOAT_ASSERT(try_extend(s));
   }
 
-  limb& operator[](size_t index) noexcept {
+  FASTFLOAT_CONSTEXPR14 limb& operator[](size_t index) noexcept {
     FASTFLOAT_DEBUG_ASSERT(index < length);
     return data[index];
   }
-  const limb& operator[](size_t index) const noexcept {
+  FASTFLOAT_CONSTEXPR14 const limb& operator[](size_t index) const noexcept {
     FASTFLOAT_DEBUG_ASSERT(index < length);
     return data[index];
   }
   // index from the end of the container
-  const limb& rindex(size_t index) const noexcept {
+  FASTFLOAT_CONSTEXPR14 const limb& rindex(size_t index) const noexcept {
     FASTFLOAT_DEBUG_ASSERT(index < length);
     size_t rindex = length - index - 1;
     return data[rindex];
   }
 
   // set the length, without bounds checking.
-  void set_len(size_t len) noexcept {
+  FASTFLOAT_CONSTEXPR14 void set_len(size_t len) noexcept {
     length = uint16_t(len);
   }
   constexpr size_t len() const noexcept {
@@ -83,12 +83,12 @@ struct stackvec {
     return size;
   }
   // append item to vector, without bounds checking
-  void push_unchecked(limb value) noexcept {
+  FASTFLOAT_CONSTEXPR14 void push_unchecked(limb value) noexcept {
     data[length] = value;
     length++;
   }
   // append item to vector, returning if item was added
-  bool try_push(limb value) noexcept {
+  FASTFLOAT_CONSTEXPR14 bool try_push(limb value) noexcept {
     if (len() < capacity()) {
       push_unchecked(value);
       return true;
@@ -137,7 +137,7 @@ struct stackvec {
   // check if any limbs are non-zero after the given index.
   // this needs to be done in reverse order, since the index
   // is relative to the most significant limbs.
-  bool nonzero(size_t index) const noexcept {
+  FASTFLOAT_CONSTEXPR14 bool nonzero(size_t index) const noexcept {
     while (index < len()) {
       if (rindex(index) != 0) {
         return true;
@@ -147,14 +147,14 @@ struct stackvec {
     return false;
   }
   // normalize the big integer, so most-significant zero limbs are removed.
-  void normalize() noexcept {
+  FASTFLOAT_CONSTEXPR14 void normalize() noexcept {
     while (len() > 0 && rindex(0) == 0) {
       length--;
     }
   }
 };
 
-fastfloat_really_inline
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14
 uint64_t empty_hi64(bool& truncated) noexcept {
   truncated = false;
   return 0;
diff --git a/include/fast_float/decimal_to_binary.h b/include/fast_float/decimal_to_binary.h
index 6da6c66..2ce43ac 100644
--- a/include/fast_float/decimal_to_binary.h
+++ b/include/fast_float/decimal_to_binary.h
@@ -63,7 +63,7 @@ namespace detail {
 // create an adjusted mantissa, biased by the invalid power2
 // for significant digits already multiplied by 10 ** q.
 template <typename binary>
-fastfloat_really_inline
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14
 adjusted_mantissa compute_error_scaled(int64_t q, uint64_t w, int lz) noexcept  {
   int hilz = int(w >> 63) ^ 1;
   adjusted_mantissa answer;
diff --git a/include/fast_float/digit_comparison.h b/include/fast_float/digit_comparison.h
index 4c43433..4e0e6a8 100644
--- a/include/fast_float/digit_comparison.h
+++ b/include/fast_float/digit_comparison.h
@@ -23,7 +23,8 @@ constexpr static uint64_t powers_of_ten_uint64[] = {
 // this algorithm is not even close to optimized, but it has no practical
 // effect on performance: in order to have a faster algorithm, we'd need
 // to slow down performance for faster algorithms, and this is still fast.
-fastfloat_really_inline int32_t scientific_exponent(parsed_number_string& num) noexcept {
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14
+int32_t scientific_exponent(parsed_number_string& num) noexcept {
   uint64_t mantissa = num.mantissa;
   int32_t exponent = int32_t(num.exponent);
   while (mantissa >= 10000) {
@@ -81,7 +82,8 @@ fastfloat_really_inline adjusted_mantissa to_extended_halfway(T value) noexcept
 
 // round an extended-precision float to the nearest machine float.
 template <typename T, typename callback>
-fastfloat_really_inline void round(adjusted_mantissa& am, callback cb) noexcept {
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14
+void round(adjusted_mantissa& am, callback cb) noexcept {
   int32_t mantissa_shift = 64 - binary_format<T>::mantissa_explicit_bits() - 1;
   if (-am.power2 >= mantissa_shift) {
     // have a denormal float
@@ -110,20 +112,16 @@ fastfloat_really_inline void round(adjusted_mantissa& am, callback cb) noexcept
 }
 
 template <typename callback>
-fastfloat_really_inline
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14
 void round_nearest_tie_even(adjusted_mantissa& am, int32_t shift, callback cb) noexcept {
-  uint64_t mask;
-  uint64_t halfway;
-  if (shift == 64) {
-    mask = UINT64_MAX;
-  } else {
-    mask = (uint64_t(1) << shift) - 1;
-  }
-  if (shift == 0) {
-    halfway = 0;
-  } else {
-    halfway = uint64_t(1) << (shift - 1);
-  }
+  const uint64_t mask
+  = (shift == 64)
+    ? UINT64_MAX
+    : (uint64_t(1) << shift) - 1;
+  const uint64_t halfway
+  = (shift == 0)
+    ? 0
+    : uint64_t(1) << (shift - 1);
   uint64_t truncated_bits = am.mantissa & mask;
   bool is_above = truncated_bits > halfway;
   bool is_halfway = truncated_bits == halfway;
@@ -140,7 +138,8 @@ void round_nearest_tie_even(adjusted_mantissa& am, int32_t shift, callback cb) n
   am.mantissa += uint64_t(cb(is_odd, is_halfway, is_above));
 }
 
-fastfloat_really_inline void round_down(adjusted_mantissa& am, int32_t shift) noexcept {
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14
+void round_down(adjusted_mantissa& am, int32_t shift) noexcept {
   if (shift == 64) {
     am.mantissa = 0;
   } else {
@@ -199,7 +198,7 @@ void parse_eight_digits(const char*& p, limb& value, size_t& counter, size_t& co
   count += 8;
 }
 
-fastfloat_really_inline
+fastfloat_really_inline FASTFLOAT_CONSTEXPR14
 void parse_one_digit(const char*& p, limb& value, size_t& counter, size_t& count) noexcept {
   value = value * 10 + limb(*p - '0');
   p++;
diff --git a/include/fast_float/fast_table.h b/include/fast_float/fast_table.h
index aa7804b..a3225c2 100644
--- a/include/fast_float/fast_table.h
+++ b/include/fast_float/fast_table.h
@@ -40,7 +40,7 @@ static const uint64_t power_of_five_128[number_of_entries];
 };
 
 template <class unused>
-const uint64_t powers_template<unused>::power_of_five_128[number_of_entries] = {
+constexpr uint64_t powers_template<unused>::power_of_five_128[number_of_entries] = {
         0xeef453d6923bd65a,0x113faa2906a13b3f,
         0x9558b4661b6565f8,0x4ac7ca59a424c507,
         0xbaaee17fa23ebf76,0x5d79bcf00d2df649,
diff --git a/include/fast_float/float_common.h b/include/fast_float/float_common.h
index 037cb04..4e4f1e7 100644
--- a/include/fast_float/float_common.h
+++ b/include/fast_float/float_common.h
@@ -91,11 +91,17 @@
 // rust style `try!()` macro, or `?` operator
 #define FASTFLOAT_TRY(x) { if (!(x)) return false; }
 
+#if __cpp_constexpr >= 201304
+#define FASTFLOAT_CONSTEXPR14 constexpr
+#else
+#define FASTFLOAT_CONSTEXPR14
+#endif
+
 namespace fast_float {
 
 // Compares two ASCII strings in a case insensitive manner.
-inline bool fastfloat_strncasecmp(const char *input1, const char *input2,
-                                  size_t length) {
+inline FASTFLOAT_CONSTEXPR14 bool
+fastfloat_strncasecmp(const char *input1, const char *input2, size_t length) {
   char running_diff{0};
   for (size_t i = 0; i < length; i++) {
     running_diff |= (input1[i] ^ input2[i]);
@@ -112,14 +118,14 @@ template <typename T>
 struct span {
   const T* ptr;
   size_t length;
-  span(const T* _ptr, size_t _length) : ptr(_ptr), length(_length) {}
-  span() : ptr(nullptr), length(0) {}
+  constexpr span(const T* _ptr, size_t _length) : ptr(_ptr), length(_length) {}
+  constexpr span() : ptr(nullptr), length(0) {}
 
   constexpr size_t len() const noexcept {
     return length;
   }
 
-  const T& operator[](size_t index) const noexcept {
+  FASTFLOAT_CONSTEXPR14 const T& operator[](size_t index) const noexcept {
     FASTFLOAT_DEBUG_ASSERT(index < length);
     return ptr[index];
   }
@@ -128,8 +134,8 @@ struct span {
 struct value128 {
   uint64_t low;
   uint64_t high;
-  value128(uint64_t _low, uint64_t _high) : low(_low), high(_high) {}
-  value128() : low(0), high(0) {}
+  constexpr value128(uint64_t _low, uint64_t _high) : low(_low), high(_high) {}
+  constexpr value128() : low(0), high(0) {}
 };
 
 /* result might be undefined when input_num is zero */
@@ -160,14 +166,14 @@ fastfloat_really_inline int leading_zeroes(uint64_t input_num) {
 #ifdef FASTFLOAT_32BIT
 
 // slow emulation routine for 32-bit
-fastfloat_really_inline uint64_t emulu(uint32_t x, uint32_t y) {
+fastfloat_really_inline constexpr uint64_t emulu(uint32_t x, uint32_t y) {
     return x * (uint64_t)y;
 }
 
 // slow emulation routine for 32-bit
 #if !defined(__MINGW64__)
-fastfloat_really_inline uint64_t _umul128(uint64_t ab, uint64_t cd,
-                                          uint64_t *hi) {
+fastfloat_really_inline constexpr uint64_t _umul128(uint64_t ab, uint64_t cd,
+                                                    uint64_t *hi) {
   uint64_t ad = emulu((uint32_t)(ab >> 32), (uint32_t)cd);
   uint64_t bd = emulu((uint32_t)ab, (uint32_t)cd);
   uint64_t adbc = ad + emulu((uint32_t)ab, (uint32_t)(cd >> 32));
@@ -207,10 +213,10 @@ struct adjusted_mantissa {
   uint64_t mantissa{0};
   int32_t power2{0}; // a negative value indicates an invalid result
   adjusted_mantissa() = default;
-  bool operator==(const adjusted_mantissa &o) const {
+  constexpr bool operator==(const adjusted_mantissa &o) const {
     return mantissa == o.mantissa && power2 == o.power2;
   }
-  bool operator!=(const adjusted_mantissa &o) const {
+  constexpr bool operator!=(const adjusted_mantissa &o) const {
     return mantissa != o.mantissa || power2 != o.power2;
   }
 };
@@ -453,21 +459,26 @@ fastfloat_really_inline void to_float(bool negative, adjusted_mantissa am, T &va
 }
 
 #if FASTFLOAT_SKIP_WHITE_SPACE // disabled by default
-inline bool is_space(uint8_t c) {
-   static const bool table[] = {
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
-   return table[c];
- }
+template <typename = void>
+struct space_lut {
+  static const bool value[];
+};
+
+template <typename T>
+constexpr bool space_lut<T>::value[] = {
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+
+inline constexpr bool is_space(uint8_t c) { return space_lut<>::value[c]; }
 #endif
 } // namespace fast_float