axsaucedo
Follow
Create your own GitHub profile
Sign up for your own profile on GitHub, the best place to host code, manage projects, and build software alongside 50 million developers.
Sign up
Pinned
3,138 contributions in the last year
Activity overview
Contributed to
SeldonIO/seldon-core,
SeldonIO/jenkins-x-seldon-core-sandbox,
axsaucedo/vulkan-kompute
and 5 other
repositories
Contribution activity
September 2020
Created a pull request in SeldonIO/seldon-core that received 21 comments
Created an issue in SeldonIO/seldon-core that received 4 comments
semverComapre broken in some kubernetes flavours resulting in CRDs not being installed
semverCompare breaks in kubernetes flavours that don't have the exact versioning convention when running kubectl version. This results in the CRDs …
4
comments
- Remove the template params from OpAlgoBase for dispatch layout
- Remove ability to copy output parameters from OpAlgoBase now that there's OpTensorSync
- Expose push constants in OpAlgo
- Document the utilities to convert shaders into C++ header files
- Explore / discuss for potential ideas or improvements
- Add explicit multi-threading interfaces to ensure correctness when running in parallel
- Add tests and documentation for loops passing data to/from device
- Add function to reinitialise / resize / recreate tensor
- Add preSubmit function to OpBase to account for multiple eval commands in parallel
- Evaluat performance of copy command on tensor (recordCopy vs map)
- Expose ability to create barriers in OpTensor operations
- Add noexcept to all constructors to explicitly state no exception thrown
- OpCreateTensor doesn't map data into GPU with OpCreateTensor for host tensors
- Remove vulkan commandbuffer from Tensor
- Add example showing how existing vulkan graphical engine can integrate Kompute
- Provide further granularity on handling staging tensors
- Create operation to copy data from local to device memory with staging
- Migrate to GTest
- Create delete function in manager to free / destroy sequence
- Allow for extending memory allocations for optimization
- Add vulkan memory allocator capabilities (or extension capabilities)
- Move all todos in the code into github issues
- Handle situation where empty vector of tensors passed to AlgoOpBase initialiser
- Document the three types of memory ownership in classes - never, optional and always
- Remove spdlog as a required dependency
- Some issues not shown.
8
contributions
in private repositories
Sep 2 – Sep 4