Tested to pass FATE on Linux and Windows.

Checkasm numbers vs the existing SSE2 code on Zen 5 (Strix Halo):
vp9_inv_adst_adst_16x16_sub16_add_10_sse2:       1041.8 ( 1.92x)
vp9_inv_adst_adst_16x16_sub16_add_10_avx512icl:   132.5 (15.06x)

vp9_inv_dct_adst_16x16_sub16_add_10_sse2:         901.0 ( 1.98x)
vp9_inv_dct_adst_16x16_sub16_add_10_avx512icl:    120.8 (14.79x)

vp9_inv_dct_dct_16x16_sub16_add_10_sse2:          750.6 ( 2.10x)
vp9_inv_dct_dct_16x16_sub16_add_10_avx512icl:     110.9 (14.18x)

vp9_inv_dct_dct_32x32_sub32_add_10_sse2:         3922.6 ( 2.24x)
vp9_inv_dct_dct_32x32_sub32_add_10_avx512icl:     506.6 (17.37x)

Attachment: vp9_itx_10_avx512.patch
Description: Binary data

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".

Reply via email to