Using Burst Data Transfers

Transferring data in bursts hides the memory access latency as well as improves bandwidth utilization and efficiency of the memory controller. It is recommended to infer burst transfers from successive requests of data from consecutive address locations. Refer to Inferring Burst Transfer from/to Global Memory for more details. Using memcpy() (C/C++ kernels) should be avoided because it cannot be pipelined with the computation in the current SDAccel™ Environment. A for-loop should be used in place of the memcpy().