Skip to content

x86: Add high bit-depth ipred z3 SSSE3 asm

Henrik Gramner requested to merge gramner/dav1d:ipred_z3_16bpc_ssse3 into master
                                  x86-64 (clang)      x86-32 (gcc)

intra_pred_z3_w4_16bpc_c:         205.8 ( 1.00x)     200.1 ( 1.00x)
intra_pred_z3_w4_16bpc_ssse3:      45.0 ( 4.57x)      49.4 ( 4.05x)
intra_pred_z3_w4_16bpc_avx2:       33.6 ( 6.13x)

intra_pred_z3_w8_16bpc_c:         575.3 ( 1.00x)     547.7 ( 1.00x)
intra_pred_z3_w8_16bpc_ssse3:      81.9 ( 7.03x)      87.8 ( 6.24x)
intra_pred_z3_w8_16bpc_avx2:       53.3 (10.79x)

intra_pred_z3_w16_16bpc_c:       1616.8 ( 1.00x)    1494.9 ( 1.00x)
intra_pred_z3_w16_16bpc_ssse3:    195.7 ( 8.26x)     199.3 ( 7.50x)
intra_pred_z3_w16_16bpc_avx2:     104.7 (15.44x)

intra_pred_z3_w32_16bpc_c:       3614.9 ( 1.00x)    3269.7 ( 1.00x)
intra_pred_z3_w32_16bpc_ssse3:    418.6 ( 8.64x)     420.7 ( 7.77x)
intra_pred_z3_w32_16bpc_avx2:     206.9 (17.47x)

intra_pred_z3_w64_16bpc_c:       8635.0 ( 1.00x)    7635.4 ( 1.00x)
intra_pred_z3_w64_16bpc_ssse3:    961.2 ( 8.98x)     959.3 ( 7.98x)
intra_pred_z3_w64_16bpc_avx2:     461.1 (18.73x)

Merge request reports