Bert Wesarg
Bert Wesarg
And it is still broken on ROCm 4.3
@all As roctracer will be replaced with rocprofiler SDK, I don't see a reason to continue here, do you?
> Especially, [this](https://github.com/SX-Aurora/veda/blob/master/src/smi/veda-smi.cpp) looks helpful. the ASCII art?
better ask "how should this be visualized"? from this we might figure a way to map it to OTF2
how is the `SWITCH` event modeled into OTF2? If it is a function, why not use an attribute to the enter event, whether the bit is set or not. In...
I will give this PR a last test in my use case and will approve then
> I will give this PR a last test in my use case and will approve then LGTM
that first showed up over year ago and I was just coming back to it, but I can still reproduce this with 1.0.6 (9bff68a4ed9975054f2c0ac8da6f0ac5691221a5). But yes, the `libOpenCL.so` is using...
btw, if I call `clGetExtensionFunctionAddress("")` before `gotcha_wrap`, it also solves the deadlock, but I consider this a rather fragile workaround
I tested it with the AMD one, which’s `libOpenCL.so` ICD dispatcher is from the [Khronos group](https://github.com/KhronosGroup/OpenCL-ICD-Loader), and the deadlock happens only in their code. I checked with the Intel oneAPI...