runtime: eliminate contention during stack allocation

Standard-sized stack frames use plain malloc/free
instead of centralized lock-protected FixAlloc.
Benchmark results on HP Z600 (2 x Xeon E5620, 8 HT cores, 2.40GHz)
are as follows:
benchmark                                        old ns/op    new ns/op    delta
BenchmarkStackGrowth                               1045.00       949.00   -9.19%
BenchmarkStackGrowth-2                             3450.00       800.00  -76.81%
BenchmarkStackGrowth-4                             5076.00       513.00  -89.89%
BenchmarkStackGrowth-8                             7805.00       471.00  -93.97%
BenchmarkStackGrowth-16                           11751.00       321.00  -97.27%

R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4657091

src/pkg/runtime/proc_test.go

@@ -6,6 +6,7 @@ package runtime_test
 
 import (
 	"runtime"
+	"sync/atomic"
 	"testing"
 )
 
@@ -44,3 +45,31 @@ func TestStopTheWorldDeadlock(t *testing.T) {
 	stop <- true
 	runtime.GOMAXPROCS(maxprocs)
 }
+
+func stackGrowthRecursive(i int) {
+	var pad [128]uint64
+	if i != 0 && pad[0] == 0 {
+		stackGrowthRecursive(i - 1)
+	}
+}
+
+func BenchmarkStackGrowth(b *testing.B) {
+	const CallsPerSched = 1000
+	procs := runtime.GOMAXPROCS(-1)
+	N := int32(b.N / CallsPerSched)
+	c := make(chan bool, procs)
+	for p := 0; p < procs; p++ {
+		go func() {
+			for atomic.AddInt32(&N, -1) >= 0 {
+				runtime.Gosched()
+				for g := 0; g < CallsPerSched; g++ {
+					stackGrowthRecursive(10)
+				}
+			}
+			c <- true
+		}()
+	}
+	for p := 0; p < procs; p++ {
+		<-c
+	}
+}