Skip to content

arm64: msac: Add handwritten versions of msac_decode_bool functions

Martin Storsjö requested to merge mstorsjo/dav1d:arm64-msac-bool into master
GCC                     Cortex A53   A72   A73
msac_decode_bool_c:           29.9  17.9  23.2
msac_decode_bool_neon:        28.4  16.0  20.8
msac_decode_bool_adapt_c:     50.1  27.5  32.2
msac_decode_bool_adapt_neon:  39.2  21.4  29.2
msac_decode_bool_equi_c:      26.6  16.9  19.4
msac_decode_bool_equi_neon:   24.9  12.5  15.4

Clang                   Cortex A53   A72   A73
msac_decode_bool_c:           28.0  16.6  23.1
msac_decode_bool_neon:        27.9  16.0  20.9
msac_decode_bool_adapt_c:     48.4  26.5  31.9
msac_decode_bool_adapt_neon:  38.9  20.4  29.2
msac_decode_bool_equi_c:      23.7  13.4  18.8
msac_decode_bool_equi_neon:   23.7  12.5  14.9

This is as fast as, or faster than, what either GCC or Clang produces.

Merge request reports