Do not use host floating point types when emitting

ASCII IR; loading and storing these can change the
bits of NaNs on some hosts.  Remove or add warnings
at a few other places using host floating point;
this is a bad thing to do in general.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@62712 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 6 insertions, 2 deletions

diff --git a/include/llvm/Support/MathExtras.h b/include/llvm/Support/MathExtras.h
index 8a89d85cd5..d55fb80268 100644
--- a/include/llvm/Support/MathExtras.h
+++ b/include/llvm/Support/MathExtras.h
@@ -361,7 +361,9 @@ inline float BitsToFloat(uint32_t Bits) {
 }
 
 /// DoubleToBits - This function takes a double and returns the bit
-/// equivalent 64-bit integer.
+/// equivalent 64-bit integer.  Note that copying doubles around
+/// changes the bits of NaNs on some hosts, notably x86, so this
+/// routine cannot be used if these bits are needed.
 inline uint64_t DoubleToBits(double Double) {
   union {
     uint64_t L;
@@ -372,7 +374,9 @@ inline uint64_t DoubleToBits(double Double) {
 }
 
 /// FloatToBits - This function takes a float and returns the bit
-/// equivalent 32-bit integer.
+/// equivalent 32-bit integer.  Note that copying floats around
+/// changes the bits of NaNs on some hosts, notably x86, so this
+/// routine cannot be used if these bits are needed.
 inline uint32_t FloatToBits(float Float) {
   union {
     uint32_t I;