search

CMS-HGCAL / rpi-daq

Deprecated. Use gitlab version instead:
https://gitlab.cern.ch/cms-hgcal-tb/rpi-daq
0 stars 4 forks source link

Improves local fifo read performance by 30% #8

Closed adavidzh closed 6 years ago

adavidzh commented 6 years ago
  1. Performance gain based on reading all GPIO lines in one call and demangling bits and writing all bits in paired SET/CLR calls on the full 32 bits.
  2. Benchmarking performed using test_fifo.py, added as make testfifo.
  3. Added the ability to compile gpiohb.c with full C++ support in case we want to use complex data structures in the future.
  4. Reproduced problems reported when the strobe (ST) signal is not set multiple times or not followed with the ACK read.

This PR is work in progress and should be tested carefully before merging.

Idea for further improvement: perform multiple reads in a loop, avoiding setting the address.