BLAST-ImpactX/impactx - CodSpeed

impactx

Blog Docs Changelog

Performance History

Latest Results

Lattice: Remove What Nothing Uses A selection tracked its own validity twice. Every structural edit bumps the lattice's edit count, and a selection compares that count, so the registry of live selections and the flag it set were a second mechanism doing the first one's job: making the invalidation a no-op failed no test. Keep the count. `PolygonVertices` and `particle_pusher_t` are leftovers of designs this branch replaced, and `Owners::list()` and `Lattice::emplace()` never had a caller. The last is worth noting: a member template is only type-checked where it is instantiated, so an unused one is not merely unused, it is unverified. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014Au2ZAQ2jjQ74U4cXhjcaT

ax3l:fix-lattice-shared

17 minutes ago

Lattice: Remove What Nothing Uses A selection tracked its own validity twice. Every structural edit bumps the lattice's edit count, and a selection compares that count, so the registry of live selections and the flag it set were a second mechanism doing the first one's job: making the invalidation a no-op failed no test. Keep the count. `PolygonVertices` and `particle_pusher_t` are leftovers of designs this branch replaced, and `Owners::list()` and `Lattice::emplace()` never had a caller. The last is worth noting: a member template is only type-checked where it is instantiated, so an unused one is not merely unused, it is unverified. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014Au2ZAQ2jjQ74U4cXhjcaT

ax3l:fix-lattice-shared

1 hour ago

Pass user wake data from RFCavity to HandleWakefield.

cemitch99:user_specified_wake

2 hours ago

Lattice: Remove What Nothing Uses A selection tracked its own validity twice. Every structural edit bumps the lattice's edit count, and a selection compares that count, so the registry of live selections and the flag it set were a second mechanism doing the first one's job: making the invalidation a no-op failed no test. Keep the count. `PolygonVertices` and `particle_pusher_t` are leftovers of designs this branch replaced, and `Owners::list()` and `Lattice::emplace()` never had a caller. The last is worth noting: a member template is only type-checked where it is instantiated, so an unused one is not merely unused, it is unverified. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014Au2ZAQ2jjQ74U4cXhjcaT

ax3l:fix-lattice-shared

2 hours ago

Data API Update

cemitch99:user_specified_wake

3 hours ago

ax3l:topic-igf-greens-cache

4 hours ago

Space Charge: bound the dynamically resized mesh ResizeMesh fitted the box to the exact particle extent on every slice step, so the resolution drifted continuously and the FFT solver had to rebuild its Green's function each time. Give the padding a band instead: geometry.prob_relative stays the smallest the mesh may be and the new geometry.prob_relative_max the largest. Within that band the mesh is chosen from a fixed set of lengths, so a beam of nearly the same size lands on exactly the same mesh and the Green's function can be reused. The allowed lengths are 2^(k/m) meters, with m = ceil(1/log2(r)) for a band r = prob_relative_max/prob_relative[0]. Rounding m up keeps the mesh inside the band, and keeping it a whole number of lengths per doubling makes two meshes a factor of two apart differ by exactly two, which the solver relies on to reuse a Green's function built at another scale. The longitudinal axis is treated in the frame the solver works in, so a changing reference energy resizes the mesh by itself. Setting prob_relative_max equal to prob_relative[0] fits the beam exactly, which stays bit-identical to the previous behavior. Also consolidates the two diverging copies of read_mr_prob_relative, which had to be kept in sync by hand for every new geometry parameter. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

ax3l:topic-igf-greens-cache

22 hours ago

Space Charge: bound the dynamically resized mesh ResizeMesh fitted the box to the exact particle extent on every slice step, so the resolution drifted continuously and the FFT solver had to rebuild its Green's function each time. Give the padding a band instead: geometry.prob_relative stays the smallest the mesh may be and the new geometry.prob_relative_max the largest. Within that band the mesh is chosen from a fixed set of lengths, so a beam of nearly the same size lands on exactly the same mesh and the Green's function can be reused. The allowed lengths are 2^(k/m) meters, with m = ceil(1/log2(r)) for a band r = prob_relative_max/prob_relative[0]. Rounding m up keeps the mesh inside the band, and keeping it a whole number of lengths per doubling makes two meshes a factor of two apart differ by exactly two, which the solver relies on to reuse a Green's function built at another scale. The longitudinal axis is treated in the frame the solver works in, so a changing reference energy resizes the mesh by itself. Setting prob_relative_max equal to prob_relative[0] fits the beam exactly, which stays bit-identical to the previous behavior. Also consolidates the two diverging copies of read_mr_prob_relative, which had to be kept in sync by hand for every new geometry parameter. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

ax3l:topic-igf-greens-cache

22 hours ago

Latest Branches

-1%

Pythonic `sim.lattice` List#1381

17 days ago

d368086

ax3l:fix-lattice-shared

0%

[WIP] Preparation for user-specified wakefields#1635

2 hours ago

3078b6f

cemitch99:user_specified_wake

0%

[WIP] Space Charge: bound the dynamically resized mesh#1621

5 hours ago

f5f22e1

ax3l:topic-igf-greens-cache

© 2026 CodSpeed Technology

Home Terms Privacy Docs