Fix digits10 bug #4281
Merged
Fix digits10 bug #4281
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Co-Authored-By: Chris Marsh <chrismarsh.c2@gmail.com> Co-Authored-By: Daniel Arndt <arndtd@ornl.gov>
c8beb93 to
bb72d32
Compare
|
(I permuted the two commits so that both will compile so it is not annoying when doing git bisections) |
|
This didn't pass. Maybe really just hardcoding those values is the way to go. |
Chrismarsh
added a commit
to Chrismarsh/kokkos
that referenced
this pull request
Sep 1, 2021
|
What about diff --git a/core/src/Kokkos_NumericTraits.hpp b/core/src/Kokkos_NumericTraits.hpp
index a0b96b8cb..1b0460370 100644
--- a/core/src/Kokkos_NumericTraits.hpp
+++ b/core/src/Kokkos_NumericTraits.hpp
@@ -166,9 +166,29 @@ template <class> struct max_digits10_helper {};
//template <> struct max_digits10_helper<float> { static constexpr int value = FLT_DECIMAL_DIG; };
//template <> struct max_digits10_helper<double> { static constexpr int value = DBL_DECIMAL_DIG; };
//template <> struct max_digits10_helper<long double> { static constexpr int value = LDBL_DECIMAL_DIG; };
-template <> struct max_digits10_helper<float> { static constexpr int value = 9; };
-template <> struct max_digits10_helper<double> { static constexpr int value = 17; };
-template <> struct max_digits10_helper<long double> { static constexpr int value = 21; };
+template <> struct max_digits10_helper<float> {
+#ifdef FLT_DECIMAL_DIG
+ static constexpr int value = FLT_DECIMAL_DIG;
+#else
+ static constexpr int value = (digits_helper<float>::value * 643L+2135)/2136+1;
+#endif
+};
+template <> struct max_digits10_helper<double> {
+#ifdef DBL_DECIMAL_DIG
+ static constexpr int value = DBL_DECIMAL_DIG;
+#else
+ static constexpr int value = (digits_helper<double>::value * 643L+2135)/2136+1;
+#endif
+};
+template <> struct max_digits10_helper<long double> {
+#ifdef DECIMAL_DIG
+ static constexpr int value = DECIMAL_DIG;
+#elif defined(LDBL_DECIMAL_DIG)
+ static constexpr int value = LDBL_DECIMAL_DIG;
+#else
+ static constexpr int value = (digits_helper<long double>::value * 643L+2135)/2136+1;
+#endif
+};(untested) for trying to fix the failing tests? |
380b55a to
8addcda
Compare
Discovered issue with long double specialization. Replaced hardcoded values by formula suggested by cppreference.com ``` std::ceil(std::numeric_limits<T>::digits * std::log10(2) + 1) ``` Co-Authored-By: Daniel Arndt <arndtd@ornl.gov>
8addcda to
2a12ea7
Compare
|
|
crtrott
approved these changes
Sep 1, 2021
crtrott
pushed a commit
that referenced
this pull request
Sep 2, 2021
* Fixes constexpr errors with frounding-math on gcc < 10. Use std::numeric_limits digits10 instead of bespoke version trilinos/Trilinos/issues/9056 /issues/4267 * Remove limits + std::numeric_limits as this is not available on all backends and follow gcc's constexpr-valid approximation of log10(2) * Update the log10_2 type to int * Revert log10_2 changes en lieu of /pull/4281 * declare const Co-authored-by: Damien L-G <dalg24+github@gmail.com> * update comment * remove const as it causes various ctors to be implicitly deleted Co-authored-by: Damien L-G <dalg24+github@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fix #4280
Conflicts with #4278