Skip to content

Task 12: batch fixed-order SU(4) construction (3.91x) - #17

Open
QingyunQian wants to merge 2 commits into
sxzgroup:v2from
QingyunQian:codex/task-12-final-expert-optimization
Open

Task 12: batch fixed-order SU(4) construction (3.91x)#17
QingyunQian wants to merge 2 commits into
sxzgroup:v2from
QingyunQian:codex/task-12-final-expert-optimization

Conversation

@QingyunQian

@QingyunQian QingyunQian commented Jul 29, 2026

Copy link
Copy Markdown

Take-home insight

Batch the fixed-order Padé construction of all SU(4) gates so JAX/TensorCircuit compile and execute one structured kernel instead of many small matrix-exponential calls.

Measured result

Comparison Runtime Speedup
Public expert 9.082742 s 1.000x
Optimized 2.320613 s 3.914003x

The small pair-fusion variant was removed from this upstream PR because it does not explain the result. The README keeps the decision-relevant factor table and one figure.

Reviewer-facing files

  • optimized_solutions/challenge-12/README.md
  • optimized_solutions/challenge-12/solution_12_batched_su4.py
  • optimized_solutions/challenge-12/factor-ablation.svg

Full profiles and research chronology remain in hmyuuu/OrbitBreakersExpertBenchmarks.

@QingyunQian QingyunQian changed the title Task 12: publish batched-SU4 and pair-fused optimizations (3.91x/4.25x local) Task 12: batch fixed-order SU(4) construction (3.91x) Jul 30, 2026
@QingyunQian
QingyunQian changed the base branch from main to v2 August 3, 2026 06:08
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