Skip to content

x86: Add high bit-depth ipred z1 SSSE3 asm

Henrik Gramner requested to merge gramner/dav1d:ipred_z1_16bpc_ssse3 into master
                                  x86-64 (clang)      x86-32 (gcc)

intra_pred_z1_w4_16bpc_c:         219.0 ( 1.00x)     198.6 ( 1.00x)
intra_pred_z1_w4_16bpc_ssse3:      39.3 ( 5.57x)      39.6 ( 5.02x)
intra_pred_z1_w4_16bpc_avx2:       31.4 ( 6.97x)

intra_pred_z1_w8_16bpc_c:         501.2 ( 1.00x)     497.5 ( 1.00x)
intra_pred_z1_w8_16bpc_ssse3:      69.5 ( 7.21x)      70.1 ( 7.10x)
intra_pred_z1_w8_16bpc_avx2:       45.6 (10.99x)

intra_pred_z1_w16_16bpc_c:       1193.0 ( 1.00x)    1272.6 ( 1.00x)
intra_pred_z1_w16_16bpc_ssse3:    135.3 ( 8.82x)     136.9 ( 9.29x)
intra_pred_z1_w16_16bpc_avx2:      76.3 (15.64x)

intra_pred_z1_w32_16bpc_c:       1847.0 ( 1.00x)    3825.8 ( 1.00x)
intra_pred_z1_w32_16bpc_ssse3:    278.1 ( 6.64x)     287.4 (13.31x)
intra_pred_z1_w32_16bpc_avx2:     173.6 (10.64x)

intra_pred_z1_w64_16bpc_c:       2913.0 ( 1.00x)    8076.0 ( 1.00x)
intra_pred_z1_w64_16bpc_ssse3:    616.7 ( 4.72x)     646.6 (12.49x)
intra_pred_z1_w64_16bpc_avx2:     341.1 ( 8.54x)

Merge request reports