Add Python latency benchmarks that ensure we are achieving our target of < 1us overhead when comparing against the equivalent C++ implementation.
Activity
leofang commentedon Feb 10, 2026
@danielfrg will take over the learnings from #1579 (comment) and work out a benchmark suite that can target
to cover a range of most common tasks.
2 remaining items
leofang commentedon Apr 7, 2026
Push out to the milestone since this is not critical/blocking.
cuda.coreimport time #2045