Add vertical loopfilter SIMD.

lpf_v_sb128uv_w4_8bpc_c: 568.2
lpf_v_sb128uv_w4_8bpc_avx2: 60.8
lpf_v_sb128uv_w6_8bpc_c: 744.2
lpf_v_sb128uv_w6_8bpc_avx2: 135.7
lpf_v_sb128y_w4_8bpc_c: 1046.4
lpf_v_sb128y_w4_8bpc_avx2: 118.0
lpf_v_sb128y_w8_8bpc_c: 1441.8
lpf_v_sb128y_w8_8bpc_avx2: 292.9
lpf_v_sb128y_w16_8bpc_c: 1778.7
lpf_v_sb128y_w16_8bpc_avx2: 577.5

    $ time tools/dav1d -i ~/Downloads/Chimera-AV1-8bit-1920x1080-6736kbps.ivf --limit=1000 -o /dev/null --muxer=yuv
    before:
    real        0m30.496s
    user        0m30.230s
    after:
    real        0m29.818s
    user        0m29.556s

changed the description

Maybe split the framework changes and the simd addition into separate patches? It would make bisecting potential issues easier.

added performance label

added 1 commit

• 4204671f - Rewrite long loopfilter.

Compare with previous version

added 5 commits

• 24b9ba08 - Add width argument to lpf_sb.
• 84a848b0 - Merge unit tests functions.
• 3ebcc004 - Rewrite horizontal loopfilter
• 8a0e38fd - Add horizontal loopfilter to unit test
• 4c69847d - Add horizontal loopfilter AVX2 SIMD.

Compare with previous version

added 7 commits

• 971430e9 - Add vertical loopfilter SIMD.
• f6d7ed95 - Rewrite long loopfilter.
• 18f9d4e2 - Add width argument to lpf_sb.
• 658bbad5 - Merge unit tests functions.
• a88b1656 - Rewrite horizontal loopfilter
• ff4661b8 - Add horizontal loopfilter to unit test
• 65b8a7c3 - Add horizontal loopfilter AVX2 SIMD.

Compare with previous version

added 9 commits

• 0df51097 - Vertical loopfilter simplifications.
• 4dfdbe78 - Add vertical loopfilter unit test
• 952a9385 - Add vertical loopfilter AVX2 SIMD
• f8533c50 - Rewrite long loopfilter
• 49a20c09 - Add width argument to lpf_sb
• dbf76e07 - Merge unit tests functions
• 03cb8c18 - Rewrite horizontal loopfilter
• c80c0a4a - Add horizontal loopfilter to unit test
• 5760d231 - Add horizontal loopfilter AVX2 SIMD.

Compare with previous version

@jamrial done

added 1 commit

• 554ae1a3 - Get rid of 8kB pb_mask table

Compare with previous version

mentioned in merge request !157 (merged)

added 26 commits

• 554ae1a3...540c0a7a - 16 commits from branch videolan:master
• 045bf507 - Vertical loopfilter simplifications
• 290d01d5 - Add vertical loopfilter unit test
• 76cfb887 - Add vertical loopfilter AVX2 SIMD
• 7b2d347b - Rewrite long loopfilter
• 280d4dda - Add width argument to lpf_sb
• b4229043 - Merge unit tests functions
• bbfa9461 - Rewrite horizontal loopfilter
• ae90a081 - Add horizontal loopfilter to unit test
• 45d66e87 - Add horizontal loopfilter AVX2 SIMD
• 6c7550ec - Get rid of 8kB pb_mask table

Compare with previous version

resolved all discussions

added 17 commits

• 6c7550ec...76646c7d - 11 commits from branch videolan:master
• 36647aaa - Vertical loopfilter simplifications
• a146437d - Add vertical loopfilter unit test
• 04b70ea5 - Add vertical loopfilter AVX2 SIMD
• bf7a4786 - Rewrite horizontal loopfilter
• 24effd47 - Add horizontal loopfilter to unit test
• aec3d25c - Add horizontal loopfilter AVX2 SIMD

Compare with previous version

merged

changed the description

added x86 label