Change List

2024/01/21

Added suppport for enumerating devices on MacOSX, see Vulkan SDK 1.3.216.0 Release Notes.

Implemented proper fallback for internal memory types. Fixes mac OS allocation errors.
Defining VUDA_STD_LAYER_ENABLED on MacOS now enables VK_LAYER_KHRONOS_validation instead of VK_LAYER_LUNARG_standard_validation.
cudaHostAlloc with cudaHostAllocWriteCombined now correctly allocates memory with HOST_VISIBLE and HOST_COHERENT, while cudaMallocHost/cudaHostAlloc now correctly allocates HOST_CACHED memory (if it exists).
If a memory type with HOST_CACHED exists it is now used on systems where it is available even though it is not HOST_VISIBLE, i.e. flush and invalidate are now in place.

VUDA Wiki introduced.
CUDA function wrapper (vuda_runtime.hpp), enabling porting of and comparison with CUDA.
VUDA now looks for queue families dedicated to compute and transfer.
Deleting a node in the binary search tree that holds device pointers is now a copy free operation.
Device query sample.

Date	Changes
2019/11/23	support for embedded kernels, embedded kernel sample
2019/10/15	internal state and memory allocation improvements, shared memory sample
2018/12/20	data structure for managing descriptor set and command buffer pool allocations
2018/11/25	Makefiles and gcc conformity
2018/11/13	virtual alloc for local device mem, one buffer per mem alloc, vuda::events, sync memcpy conformity, bandwidthtest, julia set
2018/10/25	memory allocator introduced: mallocHost, hostAlloc, optimized memory transfers, (comparable speeds with cuda in simple vector addition)
2018/10/17	kernel interface updated: kernel specialization, arbitrary arguments
2018/10/06	initial commit

The VUDA source code is released under the MIT license, see LICENSE.md for the full text.