Version 0.6.3 Release Candidate 2: Memory pools, minor fixes #496

eyalroz · 2023-04-17T09:01:09Z

eyalroz
Apr 17, 2023
Maintainer

Added a cuda::memory::pool_t proxy class
Memory pools are created using cuda::memory::pool::create() or via device methods
IPC: Can import (and export) pools and their allocations to/from other processes
Move physical allocation namespace out of virtual_ and up into memory #485 Moved the physical_allocation namespace up from memory::virtual_ into memory, as it is used also for memory pools

cuda::memory::host::allocate should return a region_t #488 cuda::memory::host::allocate() now returns a cuda::memory::region_t, for better consistency.
Adopting code to cuda::kernel::wrap API changes #486 Some changes to cuda::kernel::wrap.
Renamed cuda::memory::attribute_value_type_t -> cuda::memory::attribute_value_t
cuda::span improvements #482 Improvements to the built-in cuda::span (which is used when std::span is unavailable) - making it somewhat more compatible with std::span
Make more comparison operators constexpr and noexcept
Support converting regions into spans - at least explicitly #483 It's now easier to convert memory::region_t's into typed spans.

Wrappers now build correctly (again) with --std=c++20.
Try to build an executable for each C++ version after 11 supported by the compiler #480 Add example program built with each C++ version after 11 supported by the compiler

Now requiring CMake version 3.25. You can download an up-to-date version from Kitware's website; it doesn't require any special installation.
Switch from CUDA::nvToolkitExt to CUDA::nvtx3 (and to CMake 3.25) #490 Switched from depending on CUDA::nvToolkitExt to depending on CUDA::nvtx, for CUDA versions 10.0 and above.
fatbin targets placed in "wrong" target folder #495 Now using correct output root dir for fatbin files used in example programs

This discussion was created from the release Version 0.6.3 Release Candidate 2: Memory pools, minor fixes.