Module sparse_fft_gpu_memory 

Memory management for GPU-accelerated sparse FFT

This module provides memory management utilities for GPU-accelerated sparse FFT implementations, including buffer allocation, reuse, and transfer optimization.

Structs

BufferDescriptor

Memory buffer descriptor

GPUMemoryManager

GPU Memory manager for sparse FFT operations

Enums

AllocationStrategy

Buffer allocation strategy

BufferLocation

Memory buffer location

BufferType

Memory buffer type

Functions

get_global_memory_manager

Get global memory manager

init_cuda_device

Initialize CUDA device (no-op without CUDA feature)

init_global_memory_manager

Initialize global memory manager

init_gpu_backend

Initialize the best available GPU backend

init_hip_device

Initialize HIP device (no-op without HIP feature)

init_sycl_device

Initialize SYCL device (no-op without SYCL feature)

is_cuda_available

Check if CUDA is available (always false without CUDA feature)

is_gpu_available

Check if any GPU backend is available

is_hip_available

Check if HIP is available (always false without HIP feature)

is_sycl_available

Check if SYCL is available (always false without SYCL feature)

memory_efficient_gpu_sparse_fft

Memory-efficient GPU sparse FFT computation