hal: vsx: further optimize v_signmask
authorPaul E. Murphy <pmur@users.noreply.github.com>
Fri, 2 Aug 2019 20:06:47 +0000 (15:06 -0500)
committerPaul E. Murphy <pmur@users.noreply.github.com>
Mon, 5 Aug 2019 14:00:22 +0000 (09:00 -0500)
commit1031b7f4bc8eddcaeb67b11a764da386cdd0d08d
tree42411c485cfb916e2ec3217c3bd2f528f4ea5ac7
parent7295983964044c280484469d73d6b8f59dbc5a4f
hal: vsx: further optimize v_signmask

Use the quadword bit permutation instruction to creatively move
the sign bits to create the mask. Note that values above 127 will
result in 0.
modules/core/include/opencv2/core/hal/intrin_vsx.hpp