Possible Low dma_afu Bandwidth Due to memcpy Function
fpgabist may report lower bandwidth for the dma_afu but not the native loopback 3 (NLB3) due to the use of the memcpy function in the dma_afu driver.
You can workaround this erratum by removing memcpy from the dma_afu driver code and adding code to accept buffers from the user that have been pre-pinned. For use with OpenCL* , there is no current workaround.
Affects: Intel® Acceleration Stack 1.0 Beta and Production
Status: Planned fix in Intel® Acceleration Stack 1.1
Did you find the information on this page useful?