runtime: remove unnecessary wakeups of worker threads

Currently we wake up new worker threads whenever we pass
through the scheduler with nmspinning==0. This leads to
lots of unnecessary thread wake ups.
Instead let only spinning threads wake up new spinning threads.

For the following program:

package main
import "runtime"
func main() {
	for i := 0; i < 1e7; i++ {
		runtime.Gosched()
	}
}

Before:
$ time ./test
real	0m4.278s
user	0m7.634s
sys	0m1.423s

$ strace -c ./test
% time     seconds  usecs/call     calls    errors syscall
 99.93    9.314936           3   2685009     17536 futex

After:
$ time ./test
real	0m1.200s
user	0m1.181s
sys	0m0.024s

$ strace -c ./test
% time     seconds  usecs/call     calls    errors syscall
  3.11    0.000049          25         2           futex

Fixes #13527

Change-Id: Ia1f5bf8a896dcc25d8b04beb1f4317aa9ff16f74
Reviewed-on: https://go-review.googlesource.com/17540
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

src/runtime/proc_test.go

@@ -6,6 +6,7 @@ package runtime_test
 
 import (
 	"math"
+	"net"
 	"runtime"
 	"runtime/debug"
 	"sync"
@@ -132,6 +133,79 @@ func TestGoroutineParallelism(t *testing.T) {
 	}
 }
 
+// Test that all runnable goroutines are scheduled at the same time.
+func TestGoroutineParallelism2(t *testing.T) {
+	//testGoroutineParallelism2(t, false, false)
+	testGoroutineParallelism2(t, true, false)
+	testGoroutineParallelism2(t, false, true)
+	testGoroutineParallelism2(t, true, true)
+}
+
+func testGoroutineParallelism2(t *testing.T, load, netpoll bool) {
+	if runtime.NumCPU() == 1 {
+		// Takes too long, too easy to deadlock, etc.
+		t.Skip("skipping on uniprocessor")
+	}
+	P := 4
+	N := 10
+	if testing.Short() {
+		N = 3
+	}
+	defer runtime.GOMAXPROCS(runtime.GOMAXPROCS(P))
+	// If runtime triggers a forced GC during this test then it will deadlock,
+	// since the goroutines can't be stopped/preempted.
+	// Disable GC for this test (see issue #10958).
+	defer debug.SetGCPercent(debug.SetGCPercent(-1))
+	for try := 0; try < N; try++ {
+		if load {
+			// Create P goroutines and wait until they all run.
+			// When we run the actual test below, worker threads
+			// running the goroutines will start parking.
+			done := make(chan bool)
+			x := uint32(0)
+			for p := 0; p < P; p++ {
+				go func() {
+					if atomic.AddUint32(&x, 1) == uint32(P) {
+						done <- true
+						return
+					}
+					for atomic.LoadUint32(&x) != uint32(P) {
+					}
+				}()
+			}
+			<-done
+		}
+		if netpoll {
+			// Enable netpoller, affects schedler behavior.
+			ln, err := net.Listen("tcp", "localhost:0")
+			if err != nil {
+				defer ln.Close() // yup, defer in a loop
+			}
+		}
+		done := make(chan bool)
+		x := uint32(0)
+		// Spawn P goroutines in a nested fashion just to differ from TestGoroutineParallelism.
+		for p := 0; p < P/2; p++ {
+			go func(p int) {
+				for p2 := 0; p2 < 2; p2++ {
+					go func(p2 int) {
+						for i := 0; i < 3; i++ {
+							expected := uint32(P*i + p*2 + p2)
+							for atomic.LoadUint32(&x) != expected {
+							}
+							atomic.StoreUint32(&x, expected+1)
+						}
+						done <- true
+					}(p2)
+				}
+			}(p)
+		}
+		for p := 0; p < P; p++ {
+			<-done
+		}
+	}
+}
+
 func TestBlockLocked(t *testing.T) {
 	const N = 10
 	c := make(chan bool)