Error when attempting profiling in tensorflow-rocm: "aqlprofile API table load failed: HSA_STATUS_ERROR: A generic error has occurred."

[tarcey]

When attempting to profile a custom training loop in tensorflow following the documentation in tensorflow's trace.py,

        tf.profiler.experimental.start('logdir')
        for step in range(num_steps):
          # Creates a trace event for each training step with the
          # step number.
          with tf.profiler.experimental.Trace("Train", step_num=step):
            train_fn()
        tf.profiler.experimental.stop()

the error aqlprofile API table load failed: HSA_STATUS_ERROR: A generic error has occurred. is thrown and the program halts. I have hsa-amd-aqlprofile-bin, rocprofile and roctracer installed, built from source from the PKGBUILDs provided here. I use a gfx900 card, and tensorflow 2.6.0 also built from source with a customized PKGBUILD based on tensorflow-rocm also provided in rocm-arch.

Would someone try to use profiling in tensorflow+rocm and see if they encounter the same error?