Skip to content

Optimize native MCT and Metal irreversible IDWT#56

Merged
jcwal1516 merged 8 commits into
mainfrom
perf/post-0.7-metal-wavelet-kernels
Jul 22, 2026
Merged

Optimize native MCT and Metal irreversible IDWT#56
jcwal1516 merged 8 commits into
mainfrom
perf/post-0.7-metal-wavelet-kernels

Conversation

@jcwal1516

Copy link
Copy Markdown
Member

Summary

  • fuse the native inverse MCT level shift and correct inverse-MCT addend application
  • stage the Metal irreversible 9/7 inverse wavelet transform
  • add and harden an isolated Metal IDWT profiling/performance guard
  • remove the superseded manual GPU-capture harness and rejected-experiment narrative

Why

This branch preserves the accepted native and staged Metal improvements while dropping diagnostic capture material that no longer belongs in the product source or current benchmark evidence.

Validation

The Metal correctness and performance checks associated with this branch were completed before publication. This remains a separate draft performance line because it predates substantial current main work and will need an explicit rebase/conflict review before merge.

…avelet-kernels

# Conflicts:
#	crates/j2k-metal/src/compute/abi.rs
#	crates/j2k-metal/src/compute/decode_dispatch.rs
#	crates/j2k-metal/src/compute/decode_dispatch/idwt.rs
#	crates/j2k-metal/src/compute/symbol_inventory.rs
#	crates/j2k-metal/src/idwt.metal
#	crates/j2k-metal/src/idwt.rs
#	crates/j2k-native/src/j2c/decode.rs
@jcwal1516
jcwal1516 marked this pull request as ready for review July 22, 2026 04:29
@jcwal1516
jcwal1516 merged commit 7b03047 into main Jul 22, 2026
37 of 41 checks passed
@jcwal1516
jcwal1516 deleted the perf/post-0.7-metal-wavelet-kernels branch July 22, 2026 04:29
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant