Home - Waterfall Grid T-Grid Console Builders Recent Builds Buildslaves Changesources - JSON API - About

Builder ffmpeg64-solaris10-i386 Build #14098

Results:

Failed

SourceStamp:

Projectffmpeg
Repositoryhttps://git.ffmpeg.org/ffmpeg.git
Branchmaster
Revisiona8a7a0619694629d21339a950ae5dcd0ed135cf7
Changes1 change

BuildSlave:

unstable10x

Reason:

The SingleBranchScheduler scheduler named 'schedule-ffmpeg64-solaris10-i386' triggered this build

Steps and Logfiles:

  1. git updating ( 5 secs )
    1. stdio
  2. shell  
    1. - no logs -
  3. shell_1  
    1. - no logs -
  4. shell_2  
    1. - no logs -
  5. shell_3  
    1. - no logs -
  6. shell_4  
    1. - no logs -
  7. shell_5  
    1. - no logs -

Build Properties:

NameValueSource
branch master Build
builddir /export/home/buildbot/slave/ffmpeg64-solaris10-i386 slave
buildername ffmpeg64-solaris10-i386 Builder
buildnumber 14098 Build
codebase Build
project ffmpeg Build
repository https://git.ffmpeg.org/ffmpeg.git Build
revision a8a7a0619694629d21339a950ae5dcd0ed135cf7 Build
scheduler schedule-ffmpeg64-solaris10-i386 Scheduler
slavename unstable10x BuildSlave
workdir /export/home/buildbot/slave/ffmpeg64-solaris10-i386 slave (deprecated)

Forced Build Properties:

NameLabelValue

Responsible Users:

  1. Marcos Ashton Iglesias

Timing:

StartWed Sep 9 15:27:25 2026
EndWed Sep 9 15:27:30 2026
Elapsed5 secs

All Changes:

:

  1. Change #281275

    Category ffmpeg
    Changed by Marcos Ashton Iglesias <marcosashiglesiasohnoyoudont@gmail.com>
    Changed at Wed 09 Sep 2026 16:58:33
    Repository https://git.ffmpeg.org/ffmpeg.git
    Project ffmpeg
    Branch master
    Revision a8a7a0619694629d21339a950ae5dcd0ed135cf7

    Comments

    avcodec/x86/hevc: add SSE2 ref_filter_3tap
    Smooths the intra reference samples for 8 bit, the second of the HEVC
    intra prediction groups x86 was missing; aarch64 NEON already has it.
    
    out[i] = (in[i-1] + 2*in[i] + in[i+1] + 2) >> 2 is exact in bytes.
    pavgb rounds up, so clearing the low bit of in[i-1]+in[i+1] before the
    second average gives the wanted result for every triple. The last
    sample is copied rather than filtered and is selected into the final
    store, since the destination has no room past it. The corner, which
    both arrays share, goes out as a dword because x86-32 has no byte
    register to spare here.
    
    SSE2 rather than SSSE3: nothing can be assumed aligned, so three
    unaligned loads beat a load plus two palignr.
    
    checkasm --bench on a Core Ultra 7 155H:
    hevc_ref_filter_3tap_8x8_8_c:         75.5
    hevc_ref_filter_3tap_8x8_8_sse2:       7.1 (10.49x)
    hevc_ref_filter_3tap_16x16_8_c:      187.7
    hevc_ref_filter_3tap_16x16_8_sse2:    12.0 (15.57x)
    hevc_ref_filter_3tap_32x32_8_c:      348.0
    hevc_ref_filter_3tap_32x32_8_sse2:    22.2 (15.37x)
    
    Part of #23022.

    Changed files

    • libavcodec/x86/hevc/pred.asm
    • libavcodec/x86/hevc/pred_init.c