perf(ClusterFinderCUDA): FP32 device pedestal and bulk memcpy drain

Browse Source

- Device pedestal arrays (mean/sum/sum2) are now float instead of
  double: halves global-memory bandwidth for pedestal reads/writes and
  eliminates FP64 arithmetic in the kernel (3.3x kernel speedup,
  15µs -> 4.6µs).

- Replace the per-cluster push_back loop in the D2H drain with a
  single resize()+memcpy().

...

This commit is contained in:

kferjaoui

2026-05-21 14:12:02 +02:00

parent 6a12e3de24

commit 4c66802980

4 changed files with 107 additions and 87 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

python/tests/ClusterFinderCUDA.ipynb

+49 -28

File diff suppressed because one or more lines are too long