Phase 4 — multi-chip large-N FFT (N > 2^20)

[scttfrdmn]

Roadmap phase tracker

This issue tracks trnfft's work on Phase 4 of the trnsci roadmap.

See the suite-level roadmap
for the full phase matrix and cross-project dependencies. A reader-friendly
version lives at trnsci.dev/roadmap/.

What this phase means for trnfft:
Large-N FFT (N > 2^20 samples per dim) doesn't fit in a single
NeuronCore's SBUF + HBM working set. Phase 4 shards the butterfly
stages across NeuronCores and chips.

Done means:

• N-chip decomposition of the butterfly: twiddle factors and stage partitioning that overlap compute and cross-chip DMA.
• Validated end-to-end against scipy FFT on N = 2^22, 2^24 (single and multi-dim).
• Benchmarks at trnsci.dev/trnfft/benchmarks/ with 1-chip vs N-chip scaling curves.

Coordination

• Label: phase-4-multichip — matches the same label in every sub-project.
• Cross-project dependencies, if any, are called out in the "Done means"
  list above. Link any PRs / child issues here for tracking.

Close this issue when all "Done means" items are satisfied.