FUKernel
The FUKernel class extends the Kernel class to manage a fusion of multiple compute kernels across different devices. It provides methods to set and retrieve a device mask, track the number of fused kernels, and access individual kernels within the fusion. This design is useful for high-performance computing workflows that require coordinated execution of compute operations across multiple devices.
#include <platform/TellusimKernel.h>
Inheritance
- Inherits from Kernel
Constructors
FUKernel()
explicit FUKernel(Array<Kernel> kernels, bool owner = false)
Methods
Device mask.
void setMask(uint32_t mask)
uint32_t getMask() const
Fusion kernels.
uint32_t getNumKernels() const
const Kernel getKernel(uint32_t index) const
Kernel getKernel(uint32_t index)