SmemAllocator and expr/vector.py are thin re-exports of MLIR upstream. Both leak MLIR dialect-level noise
into kernel code.
Delete smem_allocator.py and expr/vector.py; move check_smem_capacity onto SharedAllocator or KernelFunction, and route all vector use through typing.Vector methods instead of upstream dialect ops.
SmemAllocatorand expr/vector.py are thin re-exports of MLIR upstream. Both leak MLIR dialect-level noiseinto kernel code.
Delete smem_allocator.py and expr/vector.py; move
check_smem_capacityontoSharedAllocatororKernelFunction, and route all vector use through typing.Vector methods instead of upstream dialect ops.