Axel Huebl

No rush, the release 2.10.2 is out now and contains the fix :)

Hi @jank324 @cr-xu, In your original paper (https://doi.org/10.1103/PhysRevAccelBeams.27.054601, Table 1) and your 2024 MaLAPA slides you have this nice comparison of a small benchmark you ran with Cheetah against other...

Hm, while we got this to work to compile and got a speed bump a few weeks ago, I see a regression with CPU compilation since the other week. Both...

So for the last month, I tried on and off a non trivial beamline and compilation on CPU with the 0.7.4 release of Cheetah. I cannot find a single way...

@cr-xu gets this on the example on his notebook: > I tried running the htu part in notebook and got: > > Running with compile mode: none > Track: 3_581_916_000ns...

@jank324 @cr-xu I figured it out. If I use the last tagged release `0.7.4`, all compiled code is like an _order of magitude slower_. If I use the `master` branch,...

Ok, results look sensible and will be awesome for Cheetah users (CPU and GPU) over previously published performance, I'll share them tomorrow after I rerun and make the plots understandable....

# Chromatic Push, No Space Charge _Preliminary data_ from my laptop, comparing speedups for using 1 CPU core (1 performance core on a 12th Gen Intel(R) Core(TM) i9-12900H) vs. 1...

For our IGF space charge implementation in Cheetah, I see about a 2.6x (DP) and 3.3x (SP) speedup from compiling on CPU, in a preliminary test.

Hi @jank324, thank you for the quick review and recommendations! I will run a few tests and report back here.