Use finite math. #2290
Use finite math. #2290
Conversation
ktf
commented
Feb 4, 2014
ktf
commented
- Fixes issue with gcc 4.9.0.
- std::numeric_limit::infinity is non-finite.
|
A new Pull Request was created by @ktf (Giulio Eulisse) for CMSSW_7_1_X. Use finite math. It involves the following packages: DataFormats/EgammaCandidates @nclopezo, @cmsbuild, @anton-a, @thspeer, @slava77, @Degano can you please review it and eventually sign? Thanks. |
| dist(std::numeric_limits<float>::infinity()), | ||
| dcot(std::numeric_limits<float>::infinity()), | ||
| radius(std::numeric_limits<float>::infinity()) | ||
| : flags(-std::numeric_limits<float>::max()), |
There was a problem hiding this comment.
what's the meaning of
(int)(-std::numeric_limits<float>::max())We discussed this with Florian back in July and a better choice was -1 (another alternative is -9999, the "native"
undefined value for the algorithm used here https://cmssdt.cern.ch/SDT/lxr/source/RecoEgamma/EgammaTools/src/ConversionFinder.cc#461)
There was a problem hiding this comment.
I do not have any preference. Any value would do. The advantage of
-max() is that it will behave like -inf for small numbers (i.e. X / -max
-> ~0) in case some bit of the code depended on that (which it
shouldn't, but that's another thing).
- Fixes issue with gcc 4.9.0. - std::numeric_limit<float>::infinity is non-finite.
|
I changed it for |
|
Just for the flags. |
This is two hours after +1 from cmsbuild. |
|
I wrote @nclopezo about this… Not sure what is going on... |
|
Hi, The comparison is not being triggered correctly. I am looking into the configuration of the jenkins rules to see why. For this one, I triggered it again manually. It is in the queue now. |
|
@slava77 |
