runtime: change osyield to use Windows SwitchToThread

It appears that windows osyield is just 15ms sleep on my computer
(see benchmarks below). Replace NtWaitForSingleObject in osyield
with SwitchToThread (as suggested by Dmitry).

Also add issue #14790 related benchmarks, so we can track perfomance
changes in CL 20834 and CL 20835 and beyond.

Update #14790

benchmark                             old ns/op     new ns/op     delta
BenchmarkChanToSyscallPing1ms         1953200       1953000       -0.01%
BenchmarkChanToSyscallPing15ms        31562904      31248400      -1.00%
BenchmarkSyscallToSyscallPing1ms      5247          4202          -19.92%
BenchmarkSyscallToSyscallPing15ms     5260          4374          -16.84%
BenchmarkChanToChanPing1ms            474           494           +4.22%
BenchmarkChanToChanPing15ms           468           489           +4.49%
BenchmarkOsYield1ms                   980018        75.5          -99.99%
BenchmarkOsYield15ms                  15625200      75.8          -100.00%

Change-Id: I1b4cc7caca784e2548ee3c846ca07ef152ebedce
Reviewed-on: https://go-review.googlesource.com/21294
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

1 files changed, 144 insertions, 0 deletions

diff --git a/src/runtime/syscall_windows_test.go b/src/runtime/syscall_windows_test.go
index e069eec64c..730b6d6d9c 100644
--- a/src/runtime/syscall_windows_test.go
+++ b/src/runtime/syscall_windows_test.go
@@ -864,3 +864,147 @@ func TestLoadLibraryEx(t *testing.T) {
 	t.Skipf("LoadLibraryEx not usable, but not expected. (LoadLibraryEx=%v; flags=%v)",
 		have, flags)
 }
+
+var (
+	modwinmm    = syscall.NewLazyDLL("winmm.dll")
+	modkernel32 = syscall.NewLazyDLL("kernel32.dll")
+
+	proctimeBeginPeriod = modwinmm.NewProc("timeBeginPeriod")
+	proctimeEndPeriod   = modwinmm.NewProc("timeEndPeriod")
+
+	procCreateEvent = modkernel32.NewProc("CreateEventW")
+	procSetEvent    = modkernel32.NewProc("SetEvent")
+)
+
+func timeBeginPeriod(period uint32) {
+	syscall.Syscall(proctimeBeginPeriod.Addr(), 1, uintptr(period), 0, 0)
+}
+
+func timeEndPeriod(period uint32) {
+	syscall.Syscall(proctimeEndPeriod.Addr(), 1, uintptr(period), 0, 0)
+}
+
+func createEvent() (syscall.Handle, error) {
+	r0, _, e0 := syscall.Syscall6(procCreateEvent.Addr(), 4, 0, 0, 0, 0, 0, 0)
+	if r0 == 0 {
+		return 0, syscall.Errno(e0)
+	}
+	return syscall.Handle(r0), nil
+}
+
+func setEvent(h syscall.Handle) error {
+	r0, _, e0 := syscall.Syscall(procSetEvent.Addr(), 1, uintptr(h), 0, 0)
+	if r0 == 0 {
+		return syscall.Errno(e0)
+	}
+	return nil
+}
+
+func benchChanToSyscallPing(b *testing.B) {
+	ch := make(chan int)
+	event, err := createEvent()
+	if err != nil {
+		b.Fatal(err)
+	}
+	go func() {
+		for i := 0; i < b.N; i++ {
+			syscall.WaitForSingleObject(event, syscall.INFINITE)
+			ch <- 1
+		}
+	}()
+	for i := 0; i < b.N; i++ {
+		err := setEvent(event)
+		if err != nil {
+			b.Fatal(err)
+		}
+		<-ch
+	}
+}
+
+func BenchmarkChanToSyscallPing1ms(b *testing.B) {
+	timeBeginPeriod(1)
+	benchChanToSyscallPing(b)
+	timeEndPeriod(1)
+}
+
+func BenchmarkChanToSyscallPing15ms(b *testing.B) {
+	benchChanToSyscallPing(b)
+}
+
+func benchSyscallToSyscallPing(b *testing.B) {
+	event1, err := createEvent()
+	if err != nil {
+		b.Fatal(err)
+	}
+	event2, err := createEvent()
+	if err != nil {
+		b.Fatal(err)
+	}
+	go func() {
+		for i := 0; i < b.N; i++ {
+			syscall.WaitForSingleObject(event1, syscall.INFINITE)
+			err := setEvent(event2)
+			if err != nil {
+				b.Fatal(err)
+			}
+		}
+	}()
+	for i := 0; i < b.N; i++ {
+		err := setEvent(event1)
+		if err != nil {
+			b.Fatal(err)
+		}
+		syscall.WaitForSingleObject(event2, syscall.INFINITE)
+	}
+}
+
+func BenchmarkSyscallToSyscallPing1ms(b *testing.B) {
+	timeBeginPeriod(1)
+	benchSyscallToSyscallPing(b)
+	timeEndPeriod(1)
+}
+
+func BenchmarkSyscallToSyscallPing15ms(b *testing.B) {
+	benchSyscallToSyscallPing(b)
+}
+
+func benchChanToChanPing(b *testing.B) {
+	ch1 := make(chan int)
+	ch2 := make(chan int)
+	go func() {
+		for i := 0; i < b.N; i++ {
+			<-ch1
+			ch2 <- 1
+		}
+	}()
+	for i := 0; i < b.N; i++ {
+		ch1 <- 1
+		<-ch2
+	}
+}
+
+func BenchmarkChanToChanPing1ms(b *testing.B) {
+	timeBeginPeriod(1)
+	benchChanToChanPing(b)
+	timeEndPeriod(1)
+}
+
+func BenchmarkChanToChanPing15ms(b *testing.B) {
+	benchChanToChanPing(b)
+}
+
+func benchOsYield(b *testing.B) {
+	for i := 0; i < b.N; i++ {
+		runtime.OsYield()
+	}
+}
+
+func BenchmarkOsYield1ms(b *testing.B) {
+	timeBeginPeriod(1)
+	benchOsYield(b)
+	timeEndPeriod(1)
+}
+
+func BenchmarkOsYield15ms(b *testing.B) {
+	benchOsYield(b)
+}