CL_INVALID_BINARY error trying to run on arria10 FPGA (Intel Devcloud)

[jland]

I believe this issue is a somewhere between High Level Design and the Intel Devcloud environment. If this post is better suited to be asked elsewhere, feel free to let me know. 

Below is the error I am a bit stuck on:

terminate called after throwing an instance of 'cl::sycl::runtime_error'
  what():  Native API failed. Native API returns: -42 (CL_INVALID_BINARY) -42 (CL_INVALID_BINARY)
Aborted

I have been using the "Hough_Transform_on_FPGAs_Using_oneAPI" jupyter notebook collection as a reference for this process. From what I have read, this error seems to usually be related to trying to run an FPGA bitstream on the wrong type of device. I am fairly certain everything I have done should be aimed an an arria10 device, though I could be wrong. Everything appears to work fine on an emulator or a normal program compilation. I only run into trouble running on the actual FPGA hardware.

This is the command I have been submitting to an fpga_compile node. My understanding is that this should default to be arria10 compatible.

dpcpp -v -fintelfpga -Xshardware main.cpp -o pseudo.fpga

I have tried running this on different fpga_runtime / fpga_runtime:arria10 nodes with no success. Below I am including the section of my code where the error is occurring incase it is helpful, but I do not believe it is currently the cause of my issue. (I might be very wrong though!)

device_queue.submit([&greyscale_buffer, &image_buffer, image, width, height, out](cl::sycl::handler& cgh) {
    // A discard_write is a write access that doesn't need to preserve existing memory contents
    auto data = greyscale_buffer.get_access<cl::sycl::access::mode::discard_write>(cgh);
    auto image_data = image_buffer.get_access<cl::sycl::access::mode::read>(cgh);
    // THIS PARALLEL_FOR IS WHERE THE ERROR OCCURS
    cgh.parallel_for(cl::sycl::range<1>(width * height), [image_data, out](cl::sycl::id<1> idx) {
        //     Just commented out code inside currently
    });
}).wait();

If there's anyone who could help point me in the right direction it would be much appreciated. Thanks!

[hareesh]

Hi,

pls follow bellow guidance document.

https://www.intel.com/content/www/us/en/develop/documentation/oneapi-programming-guide/top/programming-interface/fpga-flow/fpga-bsps-and-boards/fpga-board-initialization.html

reference:

https://community.intel.com/t5/Intel-DevCloud/Invalid-Binary-for-FPGA-Stratix-10-Nodes/m-p/1300748#M2604

Thanks,

[jland]

Unfortunately this does not seem to resolve the issue. I see this documentation states, "the compiler chooses the default FPGA board variant pac_a10 from the intel_a10gx_pac BSP". I connected to an fpga_runtime:arria10 node and ran the following command:

> aocl initialize acl0 pac_a10
aocl initialize: Running initialize from /glob/development-tools/versions/oneapi/2022.2/oneapi/intelfpgadpcpp/2022.1.0/board/intel_a10gx_pac/linux64/libexec
Program succeed.

After doing this, I still run into the same error trying to run the program on an FPGA. I also tried initializing pac_a10_usm which fails, but since it is not the default board variant I don't believe it is a factor here.

[hareesh]

Hi,

can you pls share what steps you are following for initialization and compilation

[jland]

I've been using the "Hough_Transform_on_FPGAs_Using_oneAPI" jupyter notebook as a guide for the process. Not too sure on how to share it since it was just included in my user folder. The specific notebook I am referencing should be located at "~/Hough_Transform_on_FPGAs_Using_oneAPI/03_FPGA_Bitstream_Compilation_Using_the_Intel_FPGA_Add-On_for_oneAPI_Base_Toolkit/FPGA_Bitstream_Compilation_Using_the_Intel_FPGA_Add-On_for_oneAPI_Base_Toolkit.ipynb".

I think have shared the steps I am doing for compilation and the nodes I have been trying to run on above. If there's some piece which seems missing feel free to let me know.

[hareesh]

Hi,

When are you getting problem at initialization or compilation time?

[hareesh]

Hi,

we don't received any response from you. can you pls respond?

[hareesh]

Hi,

We do not receive any response from you to the previous question/reply/answer that I have provided. This thread will be transitioned to community support. If you have a new question, feel free to open a new thread to get the support from Intel experts. Otherwise, the community users will continue to help you on this thread.

Thank you.