diff options
author | Erik Schnetter <schnetter@gmail.com> | 2013-07-03 21:18:00 -0400 |
---|---|---|
committer | Erik Schnetter <schnetter@gmail.com> | 2013-07-03 21:18:00 -0400 |
commit | 59efbc01e582f2bb257c6218be3bf07d7063c138 (patch) | |
tree | 58f1794f56c6b35ebb248a3479b18de381d11be4 | |
parent | 9be7b822d1d3733a4b5305fc9c649cb6e1215220 (diff) | |
download | vecmathlib-59efbc01e582f2bb257c6218be3bf07d7063c138.zip vecmathlib-59efbc01e582f2bb257c6218be3bf07d7063c138.tar.gz |
Clarify that we won't change rounding modes during optimization
-rw-r--r-- | vecmathlib.h | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/vecmathlib.h b/vecmathlib.h index 77bb859..8314905 100644 --- a/vecmathlib.h +++ b/vecmathlib.h @@ -19,7 +19,7 @@ // Possible effects of not having VML_HAVE_FP_CONTRACT: // - can re-associate // - can replace division by reciprocal -// - can break ties differently when rounding (?) +// - (can break ties differently when rounding) no, this seems too invasive // - can evaluate functions with reduced precision (80% of significant digits) // default settings |