Skip to content

x86: Add 10bpc 16x32/32x16/32x32 itx AVX-512 (Ice Lake) asm

Henrik Gramner requested to merge gramner/dav1d:itx16_32x32_avx512icl into master
inv_txfm_add_16x32_dct_dct_0_10bpc_avx2:       40.0
inv_txfm_add_16x32_dct_dct_0_10bpc_avx512icl:  37.2
inv_txfm_add_16x32_dct_dct_1_10bpc_avx2:      349.0
inv_txfm_add_16x32_dct_dct_1_10bpc_avx512icl: 218.9
inv_txfm_add_16x32_dct_dct_2_10bpc_avx2:      477.2
inv_txfm_add_16x32_dct_dct_2_10bpc_avx512icl: 359.2
inv_txfm_add_16x32_dct_dct_4_10bpc_avx2:      746.0
inv_txfm_add_16x32_dct_dct_4_10bpc_avx512icl: 568.9

inv_txfm_add_16x32_identity_identity_2_10bpc_avx2:       59.6
inv_txfm_add_16x32_identity_identity_2_10bpc_avx512icl:  53.2
inv_txfm_add_16x32_identity_identity_4_10bpc_avx2:      134.8
inv_txfm_add_16x32_identity_identity_4_10bpc_avx512icl: 102.6

inv_txfm_add_32x16_dct_dct_0_10bpc_avx2:       41.2
inv_txfm_add_32x16_dct_dct_0_10bpc_avx512icl:  33.2
inv_txfm_add_32x16_dct_dct_1_10bpc_avx2:      509.4
inv_txfm_add_32x16_dct_dct_1_10bpc_avx512icl: 283.2
inv_txfm_add_32x16_dct_dct_2_10bpc_avx2:      817.9
inv_txfm_add_32x16_dct_dct_2_10bpc_avx512icl: 579.3
inv_txfm_add_32x16_dct_dct_4_10bpc_avx2:      814.3
inv_txfm_add_32x16_dct_dct_4_10bpc_avx512icl: 662.9

inv_txfm_add_32x16_identity_identity_2_10bpc_avx2:       64.7
inv_txfm_add_32x16_identity_identity_2_10bpc_avx512icl:  54.1
inv_txfm_add_32x16_identity_identity_4_10bpc_avx2:      135.3
inv_txfm_add_32x16_identity_identity_4_10bpc_avx512icl: 105.8

inv_txfm_add_32x32_dct_dct_0_10bpc_avx2:        68.1
inv_txfm_add_32x32_dct_dct_0_10bpc_avx512icl:   56.9
inv_txfm_add_32x32_dct_dct_1_10bpc_avx2:       779.2
inv_txfm_add_32x32_dct_dct_1_10bpc_avx512icl:  415.1
inv_txfm_add_32x32_dct_dct_2_10bpc_avx2:      1026.0
inv_txfm_add_32x32_dct_dct_2_10bpc_avx512icl:  673.7
inv_txfm_add_32x32_dct_dct_4_10bpc_avx2:      1551.6
inv_txfm_add_32x32_dct_dct_4_10bpc_avx512icl: 1215.6

inv_txfm_add_32x32_identity_identity_2_10bpc_avx2:       40.6
inv_txfm_add_32x32_identity_identity_2_10bpc_avx512icl:  39.8
inv_txfm_add_32x32_identity_identity_4_10bpc_avx2:      126.9
inv_txfm_add_32x32_identity_identity_4_10bpc_avx512icl: 117.0

Merge request reports