Alejandro Saucedo
Alejandro Saucedo
#176 adds support for timestamps, and is covered in a test that currently runs only in GPU but should also be mentioned in the documentation
We currently have an example that shows how to run GPU accelerated applications with [Android via NDK](https://kompute.cc/overview/mobile-android.html), but we would like to also add an example for iPhone showing how...
Since #136, all eDevice tensors wil have a primary and a staging memory - the latter is created upon initialisation, resulting in two copies of the memory maintained consistently. This...
Currently when creating tensor buffers, these are created with `vk::SharingMode::eExclusive`. It would be important to explore usecases where enabling for `vk::SharingMode::eConcurrent` would be required.
Currently Tensors have a constructor that allows to be created with full data array. This is sub-optimal when staging tensors or output tensors are created given that the data would...
Explore creating an export for vcpkg and contribute it back to the listing so it's simpler to get started with vulkan kompute
Currently the OpCreateTensor performs a recordCopy command to all the tensors provided regardless whether they are hostvisible or devicevisible. The options for hosvisible would be to copy with a recordcopy...
Currently all the examples show how to use Vulkan Kompute from scratch, but it will be important to show how vulkan kompute can be used on an existing vulkan application....
Currently there is no support for multidimensional vectors, it would be important to explore how this could be added through either abstraction of flattened vectors, or through image2D / image3D...
Currently the shaders are built using the add_custom_target, which results in the first pass only building the files but the variables not storing them, which means that the first run...