Nicholas Christensen

Results 21 comments of Nicholas Christensen

If this is implemented, it may also be good to consider allowing uploads to [Radiocells](https://radiocells.org/) (Cells and Wifi) and [OpenWLANMap](https://github.com/openwifi-su) (Wifi only) (edit: possibly defunct now).

I also have this issue in on Android 8.1. After encrypting the phone the device admin enables correctly.

In Android Pie raw `vdc cryptfs` commands are no longer supported. https://source.android.com/security/encryption/file-based https://android.stackexchange.com/questions/212370/how-to-change-the-encryption-key-on-rooted-android-9 There is an ugly workaround however. https://forum.xda-developers.com/general/security/guide-separate-passwords-encrypted-t3048072/page6

I no longer see kernel execution times of 1e-9 with this pull request so I think it is an improvement.

From https://stackoverflow.com/questions/39924433/opencl-events-ambiguity it seems to me that `CL_PROFILING_COMMAND_SUBMIT - CL_PROFILING_COMMAND_START` is the pre-execution latency. CL_PROFILING_COMMAND_COMPLETE was added in OpenCL 2.0. I'm guessing `CL_PROFILING_COMMAND_COMPLETE - CL_PROFILING_COMMAND_END` is the post-execution latency. There...

The error message for anyone else who may stumble upon this: `AttributeError: 'numpy.ndarray' object has no attribute 'offset' `

Attemping to run `svm.py` on a Crusher MI250X with Rocm 5.2.0 results in a segmentation fault. ``` Device 'gfx90a:sramecc-:xnack-' on platform 'AMD Accelerated Parallel Processing (OpenCL 2.1 AMD-APP (3452.0))' has...

After pulling in the lastest changes and recompiling I am no longer seeing segmentation faults on Rocm 5.2.0 for `svm.py` or `demo_array_svm.py`.

The idea was to autotune with [ytopt](https://github.com/ytopt-team/ytopt) using certain fixed values that might change later on. For example, autotuning the multiplication of a 32 x 32 matrix and a 32...