Change default # of digits for APFloat::toString

This is a re-commit of r189442; I'll follow up with clang changes.

The previous default was almost, but not quite enough digits to
represent a floating-point value in a manner which preserves the
representation when it's read back in.  The larger default is much
less confusing.

I spent some time looking into printing exactly the right number of
digits if a precision isn't specified, but it's kind of complicated,
and I'm not really sure I understand what APFloat::toString is supposed
to output for FormatPrecision != 0 (or maybe the current API specification
is just silly, not sure which).  I have a WIP patch if anyone is interested.

llvm-svn: 189624

1 files changed, 8 insertions, 5 deletions

diff --git a/llvm/lib/Support/APFloat.cpp b/llvm/lib/Support/APFloat.cpp
index 34bc6b6dd6b..676e2d4ba00 100644
--- a/llvm/lib/Support/APFloat.cpp
+++ b/llvm/lib/Support/APFloat.cpp
@@ -3546,11 +3546,14 @@ void APFloat::toString(SmallVectorImpl<char> &Str,
   // Set FormatPrecision if zero.  We want to do this before we
   // truncate trailing zeros, as those are part of the precision.
   if (!FormatPrecision) {
-    // It's an interesting question whether to use the nominal
-    // precision or the active precision here for denormals.
-
-    // FormatPrecision = ceil(significandBits / lg_2(10))
-    FormatPrecision = (semantics->precision * 59 + 195) / 196;
+    // We use enough digits so the number can be round-tripped back to an
+    // APFloat. The formula comes from "How to Print Floating-Point Numbers
+    // Accurately" by Steele and White.
+    // FIXME: Using a formula based purely on the precision is conservative;
+    // we can print fewer digits depending on the actual value being printed.
+
+    // FormatPrecision = 2 + floor(significandBits / lg_2(10))
+    FormatPrecision = 2 + semantics->precision * 59 / 196;
   }
 
   // Ignore trailing binary zeros.