mbox series

[FFmpeg-devel,0/4] AArch64 NEON for HEVC

Message ID 20210107121020.86179-1-josh@itanimul.li
Headers show
Series AArch64 NEON for HEVC
Related show

Message

Josh Dekker Jan. 7, 2021, 12:10 p.m. UTC
checkasm: all 657 tests passed
hevc_add_res_4x4_8_c: 49.7
hevc_add_res_4x4_8_neon: 20.5
hevc_add_res_4x4_10_c: 45.7
hevc_add_res_4x4_10_neon: 18.7
hevc_add_res_8x8_8_c: 211.0
hevc_add_res_8x8_8_neon: 24.5
hevc_add_res_8x8_10_c: 195.7
hevc_add_res_8x8_10_neon: 24.0
hevc_add_res_16x16_8_c: 787.2
hevc_add_res_16x16_8_neon: 79.0
hevc_add_res_16x16_10_c: 714.7
hevc_add_res_16x16_10_neon: 77.7
hevc_add_res_32x32_8_c: 3444.2
hevc_add_res_32x32_8_neon: 306.5
hevc_add_res_32x32_10_c: 3820.7
hevc_add_res_32x32_10_neon: 299.5
hevc_idct_4x4_dc_8_c: 16.2
hevc_idct_4x4_dc_8_neon: 13.7
hevc_idct_4x4_dc_10_c: 16.2
hevc_idct_4x4_dc_10_neon: 14.5
hevc_idct_8x8_dc_8_c: 40.7
hevc_idct_8x8_dc_8_neon: 18.5
hevc_idct_8x8_dc_10_c: 39.2
hevc_idct_8x8_dc_10_neon: 19.2
hevc_idct_16x16_dc_8_c: 136.7
hevc_idct_16x16_dc_8_neon: 35.7
hevc_idct_16x16_dc_10_c: 136.0
hevc_idct_16x16_dc_10_neon: 36.0
hevc_idct_32x32_dc_8_c: 1386.7
hevc_idct_32x32_dc_8_neon: 132.0
hevc_idct_32x32_dc_10_c: 1366.2
hevc_idct_32x32_dc_10_neon: 132.0
hevc_sao_band_8x8_8_c: 230.7
hevc_sao_band_8x8_8_neon: 92.7

Please disregard my previous email with subject 'lavc/aarch64: add HEVC
add_residual NEON', the patch was split incorrectly.

IDCT (first) and QPEL functions in the works, then SAO edge, and
whatever is left for parity with ARM NEON.