Peter Willendrup

@farhi great with this detailed input. I will formulate individual tickets for the mxgui/doc-oriented things. @ebknudsen and I just agreed to try to do a common work-camp for McXtrace in...

This can be rectified locally by 1. replacing your `$MCXTRACE/tools/Python/mccodelib/utils.py` by this one: https://raw.githubusercontent.com/McStasMcXtrace/McCode/a51267381588b8c29fd06e89d436a1d0d40341ca/tools/Python/mccodelib/utils.py 2. (as root) run mxdoc -i An updated McXtrace will be shot out later in the...

Transitioning to `r'rawstring`' or double-escaping `\\` should likely do the trick

A quick search indicates that we need to at least have a look at these ``` ./mccodelib/fcparticleparser.py ./mccodelib/mcplotloader.py ./mccodelib/pipetools.py ./mcgui/viewclasses.py ./mctest/mctest.py ./mcdisplay/mantid_xml/mcdisplay.py ./mcdoc/mcdoc.py ./mcrun/mccode.py ./mcrun/mcsplit.py ```

It all makes sense and I believe you are absolutely right in your analysis, and a supplemental simulation dataset point in the same direction: In http://tmp.mcstas.org/oneoff_2024-02-09 CPU and GPU seem...

CPU run overview-plots, same order as in the instrument file: GPU run overview-plots, same order as in the instrument file: - So now they at least numerically agree, but still...

I am now collecting old datasets from historical tests to check if the flipping actually ever worked on GPU... - Which it did, e.g. in December 2023 (Unfortunately we've lost...

@sq-meng something is still iffy somewhere: Using kdtree does not yield the same results as regular in the case of the constant field - neither on CPU or GPU. With...

So in conclusion, at the moment: * kdtree does not function on GPU - less of a problem 🟡 * regular provides the same output on CPU and GPU, which...

Right! I was missing the last block! I will rerun and see where this takes us on my side...