While refactoring a medium complex kernel, I started hitting this error when calling enqueueTask.I have no idea what causes this. And I cannot go back to before the refactoring.
Please help, I am totally blocked on this.
[XRT] ERROR: bad mem_data index '65'../src/SortGW.cpp:299 Error calling err = queue.enqueueTask(kernel, NULL, &event1), error code is: -6
C++ kernel targeting Alveo U280 in Emu-SW on Linux ubuntu
Xilinx Vitis IDE v2019.2 (64-bit)SW Build 2708876 on Wed Nov 6 21:40:25 MST 2019
This issue seems related to resource exahustion on the FPGA.
Right now I have 4 kernels with total of 8 CU (kernel1 x 4, kernel2 x1, kernel3 x 4, kernel4 x1).
Playing with the number of kernels/CU the issue goes away.However, I cannot reduce the number of kernels and/or CUs without drastically change the program.
Q: How can I have more information on what is the actual issue so that I can try to workaround it without refactoring the whole world?
A general advice for you guys is that you can improve your runtime diagnostics :)