C++17 added a new header which includes the ability to safely convert floating points to/from chars in a completely reproducible way. The mechanism also guarantees the shortest possible representation of the chars which will completely reproduce the floating point number. Such behavior is exactly what we'd like to encode/decode doubles in ParameterSets.
The problem is none of the compilers we use implement that part of the standard. The trunk of gcc evidently does and there is an open source implementation from Microsoft (which is said to be blazingly fast).
C++17 added a new header which includes the ability to safely convert floating points to/from chars in a completely reproducible way. The mechanism also guarantees the shortest possible representation of the chars which will completely reproduce the floating point number. Such behavior is exactly what we'd like to encode/decode doubles in ParameterSets.
The problem is none of the compilers we use implement that part of the standard. The trunk of gcc evidently does and there is an open source implementation from Microsoft (which is said to be blazingly fast).