Skip to content

[pull] master from fraunhoferhhi:master#166

Merged
pull[bot] merged 3 commits into
chenm001:masterfrom
fraunhoferhhi:master
Jun 18, 2026
Merged

[pull] master from fraunhoferhhi:master#166
pull[bot] merged 3 commits into
chenm001:masterfrom
fraunhoferhhi:master

Conversation

@pull

@pull pull Bot commented Jun 18, 2026

Copy link
Copy Markdown

See Commits and Changes for more details.


Created by pull[bot] (v2.0.0-alpha.4)

Can you help keep this open source service alive? 💖 Please sponsor : )

athulya-arm and others added 3 commits June 10, 2026 11:20
Simplify the `foldingRequired` calculation by deriving the four row
distances from the first row in each 4 row group. Since `startHeight`
and loop row offset are `STEP_Y` aligned, any CTU wrap can only occur
after those four rows.

This new version improves performance by approximately ~2% when
compared to the existing Neon version when benchmarked on a Neoverse
V2 with LLVM 22.
Skip clipping for fixed filter sets in Filter7x7Blk_neon.
Fixed ALF luma filter sets use the default clip value, so clipping does
not affect the filtered differences. Add a `isFixedFilterSet` flag
through the ALF filter callbacks, including 5x5, scalar, Neon, and x86
paths. Use `isFixedFilterSet` in the Neon 7x7 path to select a no-clip
implementation.

The fixed-filter no-clip case shows an approximate 45% reduction when
benchmarked on a Neoverse V2 with LLVM 21. Non-fixed filter sets
continue to use the clipped path.

Extend filterBlk unit test to check both clipped and fixed-filter 7x7
modes.
Arm: Update Neon implementation for Filter7x7Blk
@pull pull Bot locked and limited conversation to collaborators Jun 18, 2026
@pull pull Bot added the ⤵️ pull label Jun 18, 2026
@pull pull Bot merged commit 4046482 into chenm001:master Jun 18, 2026
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants