I am attempting to use the SIRF superbuild Docker image on a MacBook Pro with Apple Silicon (M-series). While the container starts successfully and VS Code can attach to it, execution inside the container is extremely slow and often appears to hang (e.g., Python execution or imports stall indefinitely).
This appears to be related to architecture mismatch and/or filesystem performance in the devcontainer setup.
Host OS: macOS (Apple Silicon, M-series)
Image used: ghcr.io/synerbi/sirf:latest
Image architecture: amd64/linux
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
note #956 suggests emulation - I think @epapoutsellis runs CIL docker via similar emulation.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()
For CIL there is no problem at all once you installed the standard libraries for macOS(arm), including Astra and RTK for tomography. In fact, the cpu recons (since there is no gpu) are actually faster than cpu on linux and windows. Same applies with pytorch and the mps device.
For SIRF and CIL inside a docker again there is no problem. It works fine and I can confirm that is slow. Slow because of
Screenshot.2026-03-19.at.14.52.38.png (view on web) and slow because we are using no-cuda AcquisitionProjection operators, e.g., [ParallelProj](https://github.com/gschramm/parallelproj).I think we can easily have a CIL docker image for macOS. I have never tried building SIRF on a macOS.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.![]()