Template std::shared_ptr record (std::vector> tensors, TArgs&. Op: Object derived from kp::BaseOp that will be recoreded by the sequence which will be used when the operation is evaluated. Shared_ptr of the Sequence class itself Parameters This function also requires the Sequence to be recording, otherwise it will not be able to add the operation. This template requires classes to be derived from the OpBase class. Record function for operation to be added to the GPU queue in batch. std::shared_ptr record (std::shared_ptr op ) ¶ TotalTimestamps: Maximum number of timestamps to allocateĭestructor for sequence which is responsible for cleaning all subsequent owned operations. QueueIndex: Vulkan compute queue index in device Main constructor for sequence which requires core vulkan components to generate all dependent resources. Public Functions Sequence (std::shared_ptr physicalDevice, std::shared_ptr device, std::shared_ptr computeQueue, uint32_t queueIndex, uint32_t totalTimestamps = 0 ) ¶ class kp :: Sequence : public std::enable_shared_from_this ¶Ĭontainer of operations that can be sent to GPU as batch The execution of sequences can be synchronous or asynchronous, and it can be coordinated through its respective vk::Fence. The Kompute Sequence consists of batches of kp::Operations, which are executed on a respective GPU queue. Vector of physical devices containing their respective properties Sequence ¶ List the devices available in the current vulkan instance. Vk::PhysicalDeviceProperties containing information about the device std::vector listDevices ( ) const ¶ vk::PhysicalDeviceProperties getDeviceProperties ( ) const ¶ Run a pseudo-garbage collection to release all the managed resources that have been already freed due to these reaching to zero ref count. PushConstants: (optional) templatable vector parameter to use for push constants, and defaults to an empty constantĭestroy the GPU resources and all managed resources by manager. SpecializationConstants: (optional) templatable vector parameter to use for specialization constants, and defaults to an empty constant Template std::shared_ptr algorithm ( const std::vector> & tensors, const std::vector & spirv, const Workgroup & workgroup, const std::vector & specializationConstants, const std::vector & pushConstants ) ¶Ĭreate a managed algorithm that will be destroyed by this manager if it hasn’t been destroyed by its reference count going to zero. PushConstants: (optional) float vector to use for push constants, and defaults to an empty constant SpecializationConstants: (optional) float vector to use for specialization constants, and defaults to an empty constant Workgroup: (optional) kp::Workgroup for algorithm to use, and defaults to (tensor.size(), 1, 1) Spirv: (optional) The SPIRV bytes for the algorithm to dispatch Tensors: (optional) The tensors to initialise the algorithm with Shared pointer with initialised algorithm Parameters Manager (uint32_t physicalDeviceIndex, const std::vector & familyQueueIndices = ) ¶ĭefault non-template function that can be used to create algorithm objects which provides default types to the push and spec constants as floats. Variable Types for Tensors, and Push/Spec Constantsīase constructor and default used which creates the base resources including choosing the device 0 by default.Convolutional Neural Network (CNN) Simple Upscale.Example Benchmark with Matrix Multiplication.C++ Class Documentation & Reference C++ Class Documentation & Reference Contents.Extending Kompute with Custom Operations.

Doxygen param type python