Support: Add specializations for reverseBits to use builtin

[aaron.ballman]

This LGTM, but I'm not certain how we could add test coverage for the change. Maybe this is sufficiently obvious to be okay without it, though.

[arsenm]

In D77413#1960124, @aaron.ballman wrote:

This LGTM, but I'm not certain how we could add test coverage for the change. Maybe this is sufficiently obvious to be okay without it, though.

It's already covered by unit tests

[aaron.ballman]

In D77413#1960156, @arsenm wrote:

In D77413#1960124, @aaron.ballman wrote:

This LGTM, but I'm not certain how we could add test coverage for the change. Maybe this is sufficiently obvious to be okay without it, though.

It's already covered by unit tests

That's true -- I was thinking in terms of checking use of the fallback vs the builtin, but I suppose we don't care which gets used, just that the results are correct. LGTM fully now.

[arsenm]

ea397a76f56fda7883f0104789d2084001409a49

llvm/include/llvm/Support/MathExtras.h

T reverseBits(T Val) {
   unsigned char out[sizeof(Val)];
   std::memcpy(in, &Val, sizeof(Val));
   for (unsigned i = 0; i < sizeof(Val); ++i)
     out[(sizeof(Val) - i) - 1] = BitReverseTable256[in[i]];
   std::memcpy(&Val, out, sizeof(Val));
   return Val;
 }
 
+#if __has_builtin(__builtin_bitreverse8)
+template<>
+inline uint8_t reverseBits<uint8_t>(uint8_t Val) {
+  return __builtin_bitreverse8(Val);
+}
+#endif
+
+#if __has_builtin(__builtin_bitreverse16)
+template<>
+inline uint16_t reverseBits<uint16_t>(uint16_t Val) {
+  return __builtin_bitreverse16(Val);
+}
+#endif
+
+#if __has_builtin(__builtin_bitreverse32)
+template<>
+inline uint32_t reverseBits<uint32_t>(uint32_t Val) {
+  return __builtin_bitreverse32(Val);
+}
+#endif
+
+#if __has_builtin(__builtin_bitreverse64)
+template<>
+inline uint64_t reverseBits<uint64_t>(uint64_t Val) {
+  return __builtin_bitreverse64(Val);
+}
+#endif
+
 // NOTE: The following support functions use the _32/_64 extensions instead of
 // type overloading so that signed and unsigned integers can be used without
 // ambiguity.
 
 /// Return the high 32 bits of a 64 bit value.
 constexpr inline uint32_t Hi_32(uint64_t Value) {
   return static_cast<uint32_t>(Value >> 32);
 }