Skip to content

ops_todo

astewart-xmos edited this page Nov 2, 2021 · 2 revisions

This page is a backlog of operations that may be added to lib_xs3_math in future updates. This list is a work-in-progress. Not all ops listed below are guaranteed to be added, and not all ops added will come from this list.

This page was created because it seems preferable to either creating dozens of GitHub issues for individual ops or to having a single GitHub issue with dozens of ops. Instead, issues will be created for ops as they're selected for implementation, and they will be removed from this list.

Except where otherwise stated, the ops below are intended to be vectorized.

  • arctan2
    • Compute the argument (phase angle) associated with a complex vector
  • discrete cosine transform
  • 16- and 32-bit matrix-vector multiplies
    • Matrix-matrix multiply as well?
  • Matrix transposition
  • Vectorized comparisons
    • a[] <-- (b[] >= 0)? 1 : 0
    • a[] <-- (b[] == c[])? 1 : 0
    • a[] <-- (b[] >= c[])? 1 : 0
  • Extract individual elements from BFP vectors
    • Get float_s*_t from bfp_s*_t
    • Get float_complex_s*_t from bfp_complex_s*_t
    • Get float or double from bfp_s32_t
    • Get float from bfp_s16_t (double is unnecessary as single-precision floats have more than 16 mantissa bits)
Clone this wiki locally