[BUG] Wrong worker computation for single-CPU systems

[msanft]

Which component has the problem?

CuTe DSL

Bug Report

Describe the bug
On single-core systems, the following line produces max_workers=0:

cutlass/python/CuTeDSL/cutlass/cutlass_dsl/cutlass.py

Lines 411 to 413 in 4ca61d0

	# 16 threads max to avoid context switching overhead
	# To avoid oversubscription, we use half of cpu_count()
	max_workers = min(16, (os.cpu_count() or 8) // 2)

This is then unconditionally passed into ThreadPoolExecutor, which errors for a worker amount of 0.

Steps/Code to reproduce bug
Run cute.compile on a single-core system.

Expected behavior
max_workers=1 should be computed.

Environment details (please complete the following information):

• Environment location: Single-core VM

[brandon-yujie-sun]

Thanks for reporting this. Would address that in upcoming release.