Jerome Kieffer
Jerome Kieffer
`setup.py` is gone
Here are the `valgrind` "massif" profile for two calls of the program when running on a limited number of images (2000), with and without profiling activated. Valgrind still suggests to...
I run it several other times and it looks like the profiling in OpenCL prevents the memory from being freed.
So I tired to collect only timestamps for each event instead of the complete process. The patch is for now implemented in: https://github.com/silx-kit/silx/pull/3690 The memory profile looks like this, now....
I got struck by something similar in another project ... but profiling was not involved this time. https://github.com/kif/multianalyzer/blob/main/multianalyzer/opencl.py The pattern was similar: read data from an HDF5 file with large...
Issue: what if the number of calls to the kernel differ ... use NaN ?
One should document this new environment variable ... beside this, the code LGTM
Tested on PoCL 3, 4 and 5: # POCL3: OpenCL 3.0 PoCL 3.1+debian Linux, None+Asserts, RELOC, SPIR, LLVM 15.0.6, SLEEF, DISTRO, POCL_DEBUG ``` lintaillefer:~/workspace/silx % PYOPENCL_CTX=portable POCL_DEBUG=all POCL_WORK_GROUP_METHOD=cbs ./run_tests.py src/silx/image/test/test_medianfilter.py...
I managed to get it crashing with: `OpenCL 3.0 PoCL 5.0 Linux, Debug+Asserts, RELOC, SPIR, SPIR-V, LLVM 16.0.6, SLEEF, POCL_DEBUG`
So the regression appears to be more in LLVM than in PoCL ...