View Bug Details

IDProjectCategoryView StatusLast Update
0001754DCP-o-maticFeaturespublic2022-01-16 20:59
Reportercarl Assigned Tocarl  
PrioritynormalSeverityminorReproducibilityhave not tried
Status acknowledgedResolutionopen 
Summary0001754: Optimisation
Description

A general catch-all for optimisation thoughts/to-dos

  • opj_int_fix_mul is called a lot and the assert()s seem to show up in the profile, strangely; is it being inlined?
  • mct_encode_real looks like a candidate for SSE, although it's doing fixed point which isn't ideal.
  • See how far the fixed point stuff goes back as it's probably quicker to do it in floating point with SSE; grep fix_mul suggests dwt.c, t1.c, mct.c.

A callgrind on 2.15.76 suggests that of the J2K encode it's 52% dwt_encode, 29% t1_encode and 9% mct_encode.

Some CI-ish benchmarking would be nice. Maybe ccheck can be resurrected for this.

TagsNo tags attached.
Branch
Estimated weeks required
Estimated work requiredUndecided

Activities

There are no notes attached to this issue.

Bug History

Date Modified Username Field Change
2020-06-02 20:53 carl New Bug
2020-06-02 20:55 carl Description Updated
2020-06-02 20:55 carl Estimated work required => Undecided
2020-06-02 20:57 carl Description Updated
2020-06-02 21:03 carl Description Updated
2020-06-02 22:50 carl Description Updated
2020-06-03 00:12 carl Description Updated
2020-06-03 00:12 carl Description Updated
2022-01-16 20:59 carl Assigned To => carl
2022-01-16 20:59 carl Status new => acknowledged