[PATCH] D18320: APFloat: Fix signalling nans for scalbn

Matt Arsenault via llvm-commits llvm-commits at lists.llvm.org
Tue Mar 22 10:11:48 PDT 2016

arsenm added inline comments.

Comment at: lib/Support/APFloat.cpp:3987
@@ -3986,1 +3986,3 @@
   X.normalize(RoundingMode, lfExactlyZero);
+  if (X.isNaN())
+    X.makeQuiet();
scanon wrote:
> Long-term, it is likely better to have `normalize` quiet SNaN.  Can you give that approach a try and see how much short-term fallout there would be?  If it's too much of a mess, we can use your current approach instead.
I tried doing this and expected the other places that do the same thing would happen to work out, but I get some APFloatTest failures. It's quite likely I added this to normalize incorrectly though


More information about the llvm-commits mailing list