runtime/pprof: include example uses of pprof with PMU.

Add four examples loop fusion, matrix transpose, concurrent gorotuines, and a weighted serial
execution of functions to demonstrate profiling use cases with different metrics.

src/net/http/pprof/examples/fusion.go

@@ -0,0 +1,61 @@
+// +build ignore
+
+package main
+
+import (
+	"log"
+	_ "math/rand"
+	"net/http"
+	_ "net/http/pprof"
+)
+
+const N = (1 << 20)
+
+var a []float32
+var b []float32
+var c []float32
+var d []float32
+
+func init() {
+	a = make([]float32, N)
+	b = make([]float32, N)
+	c = make([]float32, N)
+	d = make([]float32, N)
+	//	for i := 0; i < N; i++ {
+	//		a[i] = rand.Float32()
+	//		b[i] = rand.Float32()
+	//	}
+}
+
+//go:noinline
+func addSub() {
+	for i := 0; i < N; i++ {
+		c[i] = b[i] + c[i]
+	}
+	for i := 0; i < N; i++ {
+		d[i] = b[i] - c[i]
+	}
+}
+
+//go:noinline
+func addSubFuse() {
+	for i := 0; i < N; i++ {
+		c[i] = b[i] + c[i]
+		d[i] = b[i] - c[i]
+	}
+}
+
+func run() error {
+	addSub()
+	addSubFuse()
+	return nil
+}
+
+func main() {
+	go func() { log.Println(http.ListenAndServe("localhost:6060", nil)) }()
+	for {
+		if err := run(); err != nil {
+			log.Fatal(err)
+		}
+	}
+}

src/net/http/pprof/examples/goroutine.go

@@ -0,0 +1,190 @@
+// +build ignore
+
+package main
+
+import (
+	"log"
+	"net/http"
+	_ "net/http/pprof"
+	"sync"
+)
+
+var wg sync.WaitGroup
+
+//go:noinline
+func f1() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f2() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f3() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f4() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f5() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f6() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f7() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f8() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f9() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func f10() int {
+	defer wg.Done()
+
+	var sum int
+	for i := 0; i < 500000000; i++ {
+		sum -= i / 2
+		sum *= i
+		sum /= i/3 + 1
+		sum -= i / 4
+	}
+
+	return sum
+}
+
+//go:noinline
+func run() error {
+	wg.Add(10)
+	defer wg.Wait()
+
+	go f1()
+	go f2()
+	go f3()
+	go f4()
+	go f5()
+	go f6()
+	go f7()
+	go f8()
+	go f9()
+	go f10()
+
+	return nil
+}
+
+func main() {
+	go func() { log.Println(http.ListenAndServe("localhost:6060", nil)) }()
+	for {
+		if err := run(); err != nil {
+			log.Fatal(err)
+		}
+	}
+}

src/net/http/pprof/examples/run.sh

@@ -0,0 +1,28 @@
+set -ex 
+#GOROOT=<set me>
+#PATH=<set me>
+#if needed and supported
+#GO_PPROF_PRECISION=3
+
+HOST=localhost
+PORT=6060
+L1MISS="r08d1"
+
+for f in fusion.go  goroutine.go  serial.go  transpose.go
+do
+name=`basename $f .go`
+go run $f &
+pid=$!
+sleep 2
+# simple timer profile
+curl -o ${name}.timer.prof ${HOST}:${PORT}/debug/pprof/profile?seconds=10
+# CPU cycles @ 10M
+curl -o ${name}.cycles.prof ${HOST}:${PORT}/debug/pprof/profile?event=cycles\&period=10000000\&seconds=20
+# CPU instructions @ 1M
+curl -o ${name}.ins.prof ${HOST}:${PORT}/debug/pprof/profile?event=instructions\&period=1000000\&seconds=20
+# CPU L1 D-cache miss @ 10000 
+curl -o ${name}.l1miss.prof ${HOST}:${PORT}/debug/pprof/profile?event=${L1MISS}\&period=10000\&seconds=20
+# CPU LLC cache miss @ 10000 
+curl -o ${name}.llcmiss.prof ${HOST}:${PORT}/debug/pprof/profile?event=cacheMisses\&period=10000\&seconds=20
+killall $name || "cannot kill"
+done