The monotonic time source had two independent problems:

• On Linux, runtimeNano passed clite's CLOCK_MONOTONIC, whose value is Darwin's clock id (6). Linux interprets 6 as CLOCK_MONOTONIC_COARSE, a millisecond-granularity clock: consecutive time.Now() readings were identical 100% of the time and the smallest nonzero delta was 1ms. Timer precision, sync spin timing and any benchmark that trusts time.Since were all millisecond-quantized.
• On Darwin, clock_gettime(CLOCK_MONOTONIC) itself only has microsecond granularity (96% identical consecutive readings, 1µs minimum delta).

Changes

Mirror Go's runtime structure with a per-OS nanotime1 in the runtime package itself, keeping the hot path free of clite indirection and clite unchanged:

• Darwin reads CLOCK_UPTIME_RAW through clock_gettime_nsec_np — the same clock Go's nanotime uses there — via a direct C linkname.
• Linux uses clock_gettime with the OS-correct CLOCK_MONOTONIC id as a local constant.
• Remaining platforms keep the previous behavior.

Validation

Consecutive time.Now() delta probe (min nonzero delta / fraction of zero deltas), 200k iterations:

time.Sleep / Timer / Ticker smoke: 50ms sleep → 51ms, 30ms timer → 31ms, 3×10ms ticker → 32ms, identical before and after. go test ./test/go -run 'Select|Channel' passes on macOS; builds and probes verified in the Linux arm64 container.

Found while benchmarking #2012: cold-path metrics on Linux read as 0ns because the monotonic clock could not resolve sub-millisecond durations.

🤖 Generated with Claude Code