Files
TRIMSP/CHANGELOG.md
salman 0ad68e1e57 Exclude build-id links from rpm to avoid electron conflicts
Two Electron applications built from the identical electron binary set
package the same ELF build-id links under /usr/lib/.build-id, so
installing trimsp next to musrflow failed the transaction test. Define
_build_id_links none in the rpm spec so the package no longer claims
those paths, regardless of the el9 build host.
2026-09-24 14:46:48 +02:00

1.8 KiB

Changelog

Unreleased

Fixed

  • The RPM no longer packages ELF build-id links under /usr/lib/.build-id, which previously conflicted when another Electron application built from the same binary set (e.g. musrflow) was installed alongside trimsp.

1.4.0 - 2026-08-27

Publication release of TRIMSP-NL Workbench corresponding to the modernization and validation described in the accompanying manuscript.

Added

  • Generalized target handling for layers containing more than five elements.
  • Runtime-selectable RANLUX, PCG32, and xoshiro256** random-number backends.
  • Explicit release, debug, profiling, and static Fortran build modes.
  • Standalone packaging support while retaining the browser-based workflow.
  • A named software version in the Fortran output.

Changed

  • Refactored the legacy Fortran source incrementally, including extraction of GUI input handling and output-writing routines from the main transport flow.
  • Improved comments, variable documentation, and use of named constants.
  • Retained RANLUX as the default backend and preserved the established input-deck format for backward compatibility.

Fixed

  • Corrected localized depth-bin index handling.
  • Corrected transmitted-energy accumulation.
  • Explicitly initialized recoil counters.
  • Corrected selected layer-index updates.

Validation

The publication test suite contains 13 representative target configurations and 143 energy-configuration points. The alternative RNG backends were compared with the historical RANLUX reference over 429 simulations in total. In that suite, the fast backends reduced the summed runtime by approximately a factor of two without a systematic backend-dependent trend in the reported implantation, backscattering, transmission, range, straggling, or representative profile observables.