Version 0.8.1 Release Candidate 1: CUDA 12.x compatibility improvements & minor fixes

[eyalroz]

Changes since v0.8.0:

Minor bug fixes

• context_t::flags() should not be virtual #706 : Made context_t::flags() non-virtual
• Fix C++20-implicit and pre-C++20 comparison operators for launch_configuration_t #710 : Fixed the comparison operators for launch configurations
• Span-to-c-array copy function ignores stream of choice, acts synchronously #709 : Span-to-C-array copy no longer ignoring the designated stream
• Infinite recursion in one variant of link_t::add_file() #708 : Avoiding infinite recursion in link_t::add_file()

To the unique_span & unique_region classes

• swap method of unique_span doesn't swap the deleters #703 : unique_span swap now correctly swaps the deleters as well
• unique_region_t move constructor and assigment operator doesn't compile #713 : Move ctor and assignment operator of unique_region_t
• Fix typo in --no-source-include NVRTC option #702 : Fixed a typo when passing the --no-source-include option to NVRTC

CUDA 12.x compatibility

• Update static tables with Blackwell architecture info #711 : Added preliminary information regarding Blackwell cards and micro-architecture
  *Respect lack of support for --version-ident with CUDA 12.2 and later #701 : The --version-ident compilation option to NVRTC was dropped in CUDA 12.2; this is now respected by the wrappers and the option is not exposed for 12.2 and newer versions of CUDA.
• Fix typo in --no-source-include NVRTC option #702 : Fixed handling of --version-ident (we had a spacing issue)
• Support new JIT compilation flags #635, Respect lack of support for --version-ident with CUDA 12.2 and later #701 : Added support for the --fdevice_syntax_only and --minimal options for NVRTC compilation

Other changes

• Compiler warnings coming from release 0.8.0, when using clang #704 : Limited the clang warning flags (no -pedantic) to avoid warnings we can't resolve
• Avoid accidental infinite recursion in library_t::get_managed() #705 : Made some methods of library_t be const

---

This discussion was created from the release Version 0.8.1 Release Candidate 1: CUDA 12.x compatibility improvements & minor fixes.