internal/trace: add "per-P" MMU analysis

The current MMU analysis considers all Ps together, so if, for
example, one of four Ps is blocked, mutator utilization is 75%.
However, this is less useful for understanding the impact on
individual goroutines because that one blocked goroutine could be
blocked for a very long time, but we still appear to have good
utilization.

Hence, this introduces a new flag that does a "per-P" analysis where
the utilization of each P is considered independently. The MMU is then
the combination of the MMU for each P's utilization function.

Change-Id: Id67b980d4d82b511d28300cdf92ccbb5ae8f0c78
Reviewed-on: https://go-review.googlesource.com/c/60797
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>

src/cmd/trace/mmu.go

@@ -27,12 +27,12 @@ func init() {
 
 var mmuCache struct {
 	init     sync.Once
-	util     []trace.MutatorUtil
+	util     [][]trace.MutatorUtil
 	mmuCurve *trace.MMUCurve
 	err      error
 }
 
-func getMMUCurve() ([]trace.MutatorUtil, *trace.MMUCurve, error) {
+func getMMUCurve() ([][]trace.MutatorUtil, *trace.MMUCurve, error) {
 	mmuCache.init.Do(func() {
 		tr, err := parseTrace()
 		if err != nil {
@@ -69,7 +69,16 @@ func httpMMUPlot(w http.ResponseWriter, r *http.Request) {
 	// Cover six orders of magnitude.
 	xMax := xMin * 1e6
 	// But no more than the length of the trace.
-	if maxMax := time.Duration(mu[len(mu)-1].Time - mu[0].Time); xMax > maxMax {
+	minEvent, maxEvent := mu[0][0].Time, mu[0][len(mu[0])-1].Time
+	for _, mu1 := range mu[1:] {
+		if mu1[0].Time < minEvent {
+			minEvent = mu1[0].Time
+		}
+		if mu1[len(mu1)-1].Time > maxEvent {
+			maxEvent = mu1[len(mu1)-1].Time
+		}
+	}
+	if maxMax := time.Duration(maxEvent - minEvent); xMax > maxMax {
 		xMax = maxMax
 	}
 	// Compute MMU curve.

src/internal/traceparser/gc_test.go

@@ -29,14 +29,14 @@ func TestMMU(t *testing.T) {
 	// 0.5      *   *   *   *
 	// 0.0      *****   *****
 	//      0   1   2   3   4   5
-	util := []MutatorUtil{
+	util := [][]MutatorUtil{{
 		{0e9, 1},
 		{1e9, 0},
 		{2e9, 1},
 		{3e9, 0},
 		{4e9, 1},
 		{5e9, 0},
-	}
+	}}
 	mmuCurve := NewMMUCurve(util)
 
 	for _, test := range []struct {
@@ -90,7 +90,7 @@ func TestMMUTrace(t *testing.T) {
 	// Test the optimized implementation against the "obviously
 	// correct" implementation.
 	for window := time.Nanosecond; window < 10*time.Second; window *= 10 {
-		want := mmuSlow(mu, window)
+		want := mmuSlow(mu[0], window)
 		got := mmuCurve.MMU(window)
 		if !aeq(want, got) {
 			t.Errorf("want %f, got %f mutator utilization in window %s", want, got, window)