1

So I need to have some form of low latency hardware acceleration for Tensorflow inferencing on a raspberry-pi. I looked into using leflow but it cannot support large tensorflow models on small FPGAs.

I have figured out how to build tensorflow from source with sycl support. Where I am getting stuck though is on how to configure sycl (or OpenCL which sycl wraps) to work with an fpga.

I am also not sure how to physically wire my fpga to the raspis gpios such that it accelerates opencl.

Any help, or links regarding this matter would be appreciated.

Tamara Koliada
  • 1,200
  • 2
  • 14
  • 31
dadrake
  • 112
  • 7
  • 1
    Using ComputeCpp, an implementation of SYCL, you can only target devices that provide driver support for SPIR OpenCL or SPIR-V. You can check if a device is supported by copying the ComputeCpp binaries onto the device then executing the "computecpp_info" tool (in the bin folder). This will output what hardware is detected and if there is SPIR/SPIR-V support. – Rod Burns Feb 25 '19 at 10:02
  • What FPGA do you want to use? Does it support OpenCL 1.2 with SPIR-V? That’s a requirement for the SYCL compilers I know about that support FPGAs. Also, to run well on an FPGA, the code needs to be written for an FPGA, not a GPU. I think TensorFlow SYCL targets GPUs. – Jeff Hammond Nov 08 '19 at 04:00

0 Answers0