[Schedule] Allow partitioning the same array on multiple dimensions (#547)

[zzzDavid]

Summary

• Relaxes the duplicate partition check in Schedule.partition() to allow partitioning the same array on different dimensions, fixing [Bug] Cannot partition the same array on multiple dimensions #547
• The MLIR backend already supports composing multiple partition layout maps, and the HLS emitter already emits per-dimension #pragma HLS array_partition directives — the error was an overly restrictive Python-level check
• Updates the FEATHER example to use selective per-dimension partitioning instead of the full-partition workaround
• Adds test_partition_multi_dim covering complete, block, cyclic, and mixed-type multi-dim partitions with HLS codegen verification

Test plan

• python -m pytest tests/test_schedule_memory.py::test_partition_multi_dim -v — new test passes
• python -m pytest tests/test_schedule_memory.py -k partition -v — all 11 partition tests pass
• Lint checks pass (license headers, black, clang-format)

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Relaxes Schedule.partition()’s duplicate-partition guard to allow partitioning the same array across different dimensions (fixing #547), aligns the FEATHER example with the intended per-dimension partitioning strategy, and adds a regression test covering multi-dimension partition behavior with HLS codegen checks.

Changes:

• Update Schedule.partition() to permit multiple partitions on the same array as long as they target different dimensions.
• Update FEATHER example to partition weights on dimensions 2 and 3 instead of fully partitioning all dimensions.
• Add test_partition_multi_dim to validate multi-dim partition combinations and expected HLS pragmas, plus error cases.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 9 comments.

File	Description
allo/customize.py	Relaxes duplicate-partition validation to allow per-dimension composition while still blocking conflicting cases.
examples/feather/feather.py	Switches from full-partition workaround to explicit per-dimension partitions on weights.
tests/test_schedule_memory.py	Adds a new test covering multi-dimension partitioning and HLS pragma emission expectations.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[chhzh123]

LGTM, thanks!