improve use of preprocessor macros in CUDA kernels - Redmine #1211
With extensions and optimizations gradually added to the CUDA kernels the code became quite hard to read doe to the following aspects:
- poorly documented #ifdef
’s (especially no comment mark at closing
#endif
’s
- use of catch-all #else
- use of
#if X..#endif #if Y..#endif
instead of#if X #elif Y #endif
when X and Y are anyway mutually exclusive.
More concrete comments are on the gerrit page of change 2232.
(from redmine: issue id 1211, created on 2013-03-27 by pszilard)
- Relations:
- relates #1666 (closed)