Michael Pratt

Filed #74111.

In case it wasn't already clear, I'm fairly certain this case in `do_wp_page` is doing COW of a read-only mapping due to a write. I'm not 100% sure, but [`ptep_clear_flush`...

> > I'd guess that the RO page being COW'd is the zero page? > > Oh... are you saying that `memclrNoHeapPointers` is writing the zero page? Some of the...

I commented on golang/go#73193 about feature detection, but one more comment about breakage. > Is there any plan to make sure automaxprocs doesn't break if the runtime implementation changes? What...

In the new trace format, we do usually store absolute (monotonic) timestamps. The start of each batch has the full absolute timestamp. Each event within the batch contains the time...

In theory, I believe this field just needs to be in absolute time: https://cs.opensource.google/go/go/+/master:src/internal/trace/traceviewer/format/format.go;l=25;drc=70f4717e96cf9d8ca8a5f439d7d6487ec822ce49;bpv=1;bpt=1 The current values tend to just come from computations like https://cs.opensource.google/go/go/+/master:src/cmd/trace/goroutinegen.go;l=137;drc=70f4717e96cf9d8ca8a5f439d7d6487ec822ce49;bpv=1;bpt=1, which currently are `event.Time -...

> The absolute value can thus be calculated if we knew the time since boot. But I haven't been able to find that anywhere in a trace. Did I miss...

That makes sense. There is a caveat that it will probably be difficult to precisely line up events using wall time, but I agree it would be useful even just...

(cc @mknyszek for thoughts on including a real time timestamp in each generation)

`runtime.walltime`