Version 0.6.9 Beta 1: Memory handling streamlined with unique_span rather unique_ptr

[eyalroz]

Changes since v0.6.8:

Memory allocation & copying-related changes

• Support copying from standard-library containers with contiguous storage #606 Can now copy directly to and from containers with contiguous storage - without going through pointers or specifying the size

Unique spans (unique_span_t)

• Add a unique_region_t class or some equivalent mechanism #291 Added a unique_span_t<T> template class, combining the functionality of cuda::unique_ptr and cuda::span (and being somewhat similar to std::dynarray which almost made it into C++14). Many CUDA programs want to represent both the ownership of allocated memory, and the range of that memory for actual use, in the same variable - without the on-the-fly reallocation behavior of std::vector. This is now possible. Also implemented an untyped version of this, named unique_region.
• Add an empty() method to span() #601 Added an empty() method to cuda::span (to match that of std::span - as it is no sometimes used)
• Use our unique_span's instead of the faux-dynarray class #603 Use unique_span_t instead of our cuda::dynarray (which had been an std::vector under the hood), in various places in the API, especially RTC

Memory regions (memory::region_t)

• Use region_t's for mapped memory #594 Now using memory::region_t for mapped memory rather than a different, mapped-memory specific region class
• Make cuda::memory::region_t more constexpr-friendly #602 Make memory::region_t more constexpr-friendly
• Make memory::region_t CUDA-independent #604 memory::region_t's are now CUDA-independent, i.e. do not utilize any CUDA-specific definitions
• Support for constructing const_region_t's from region_t&& #605 Can now construct const_region_t's from rvalue references to regions

Documentation & comments

• correct the documentation of supports_memory_pools #595 Correct the documentation for supports_memory_pools

Launch configuration & launch config builder changes

• Checking against the optional rather than the value in launch_config_builder_t #596 Corrected a check against the associated device in the kernel-setting method of the launch config builder

CUDA libraries and in-library, non-associated kernel support

• library kernel get_attribute and set_attribute have return issues #598 Corrected the API and implementation of get_attribute() and set_attribute() for library kernels

Refactoring

• Split off a type_traits.hpp header from types.hpp #607 Split off a detail/type_traits.hpp from types.hpp

Other changes

• An invalid file, name_caching_program.hpp, has snuck into our code #599 An invalid file, name_caching_program.hpp, had snuck into our code - removed it
• Stick to an underscore suffix for proxy class field name #600 Adhere to underscore suffix for proxy class field names

---

This discussion was created from the release Version 0.6.9 Beta 1: Memory handling streamlined with unique_span rather unique_ptr.