Skip to content

x86: Add high bit-depth ipred z2 SSSE3 asm

Henrik Gramner requested to merge gramner/dav1d:ipred_z2_16bpc_ssse3 into master
                                 x86-64 (clang)     x86-32 (gcc)

intra_pred_z2_w4_16bpc_c:         225.0 ( 1.00x)    250.7 ( 1.00x)
intra_pred_z2_w4_16bpc_ssse3:      70.0 ( 3.21x)     78.5 ( 3.20x)
intra_pred_z2_w4_16bpc_avx2:       51.1 ( 4.40x)

intra_pred_z2_w8_16bpc_c:         629.3 ( 1.00x)    660.7 ( 1.00x)
intra_pred_z2_w8_16bpc_ssse3:     121.9 ( 5.16x)    134.4 ( 4.91x)
intra_pred_z2_w8_16bpc_avx2:      102.7 ( 6.13x)

intra_pred_z2_w16_16bpc_c:       1779.6 ( 1.00x)   1861.0 ( 1.00x)
intra_pred_z2_w16_16bpc_ssse3:    270.3 ( 6.58x)    286.7 ( 6.49x)
intra_pred_z2_w16_16bpc_avx2:     202.4 ( 8.79x)

intra_pred_z2_w32_16bpc_c:       4234.7 ( 1.00x)   4341.4 ( 1.00x)
intra_pred_z2_w32_16bpc_ssse3:    562.5 ( 7.53x)    585.8 ( 7.41x)
intra_pred_z2_w32_16bpc_avx2:     399.0 (10.61x)

intra_pred_z2_w64_16bpc_c:       9255.7 ( 1.00x)   9284.7 ( 1.00x)
intra_pred_z2_w64_16bpc_ssse3:   1284.1 ( 7.21x)   1358.1 ( 6.84x)
intra_pred_z2_w64_16bpc_avx2:     811.4 (11.41x)

Merge request reports