Release Version 0.6.6: Minor bug fixes + compatibility improvements · eyalroz/cuda-api-wrappers

Changes since v0.6.4:

Functionality improvements

#545 Now checking, and throwing, errors due to cudaGetLastError() after kernel launches (mostly grid errors)
#547 When compiling in debug mode, now performing more launch configuration validity checks before launching a kernel
#549 Avoiding some excessive device property querying.

#539, #544 NVRTC compilation logs now returned without a trailing nul ('\0') character.
#542 More robust use of namespace in the library's macros, so they don't trigger compilation errors regardless of the namespace of the code you use them in
#543 Now retrieving correct error strings again for Runtime-API-only errors
#550 Fixed a wrong side of comparison in some block configuration logic of the launch config builder
#553 Replaced inappropriate use of cbegin() and cend() in favor of begin() and end() in rtc::program::add_headers() code which may take inputs without these two methods.

#546 Resolved a build on Windows with rtc.hpp - fixed an overload resolution issue regarding compilation parameter marshalling.

Want to help me with testing 0.7? Drop me a line... (it will have CUDA execution graph support)