[Libre-soc-dev] [RFC] SVP64 Vertical-First Mode loops
programmerjake at gmail.com
Wed Aug 18 17:53:26 BST 2021
On Wed, Aug 18, 2021, 09:15 lkcl <luke.leighton at gmail.com> wrote:
> instead of pre-calculating the entire table, which itself results in
> considerably more LDs, and in strip-mining of the L1 Cache, Vertical-First
> Mode allows each cosine value to be calculated *on demand* as a scalar
> element, for a SPECIFIC src/dststep at the EXACT moment it is needed.
Even if we get a HW cos pipeline, it will almost always be much faster to
load the constant from memory...additionally some codecs may specify using
specific rounded values of cos (for repeatability across implementations)
and we have to use those exact values, not recalculate our own.
More information about the Libre-soc-dev