sync: scalable Pool

Introduce fixed-size P-local caches.
When local caches overflow/underflow a batch of items
is transferred to/from global mutex-protected cache.

benchmark                    old ns/op    new ns/op    delta
BenchmarkPool                    50554        22423  -55.65%
BenchmarkPool-4                 400359         5904  -98.53%
BenchmarkPool-16                403311         1598  -99.60%
BenchmarkPool-32                367310         1526  -99.58%

BenchmarkPoolOverlflow            5214         3633  -30.32%
BenchmarkPoolOverlflow-4         42663         9539  -77.64%
BenchmarkPoolOverlflow-8         46919        11385  -75.73%
BenchmarkPoolOverlflow-16        39454        13048  -66.93%

BenchmarkSprintfEmpty                    84           63  -25.68%
BenchmarkSprintfEmpty-2                 371           32  -91.13%
BenchmarkSprintfEmpty-4                 465           22  -95.25%
BenchmarkSprintfEmpty-8                 565           12  -97.77%
BenchmarkSprintfEmpty-16                498            5  -98.87%
BenchmarkSprintfEmpty-32                492            4  -99.04%

BenchmarkSprintfString                  259          229  -11.58%
BenchmarkSprintfString-2                574          144  -74.91%
BenchmarkSprintfString-4                651           77  -88.05%
BenchmarkSprintfString-8                868           47  -94.48%
BenchmarkSprintfString-16               825           33  -95.96%
BenchmarkSprintfString-32               825           30  -96.28%

BenchmarkSprintfInt                     213          188  -11.74%
BenchmarkSprintfInt-2                   448          138  -69.20%
BenchmarkSprintfInt-4                   624           52  -91.63%
BenchmarkSprintfInt-8                   691           31  -95.43%
BenchmarkSprintfInt-16                  724           18  -97.46%
BenchmarkSprintfInt-32                  718           16  -97.70%

BenchmarkSprintfIntInt                  311          282   -9.32%
BenchmarkSprintfIntInt-2                333          145  -56.46%
BenchmarkSprintfIntInt-4                642          110  -82.87%
BenchmarkSprintfIntInt-8                832           42  -94.90%
BenchmarkSprintfIntInt-16               817           24  -97.00%
BenchmarkSprintfIntInt-32               805           22  -97.17%

BenchmarkSprintfPrefixedInt             309          269  -12.94%
BenchmarkSprintfPrefixedInt-2           245          168  -31.43%
BenchmarkSprintfPrefixedInt-4           598           99  -83.36%
BenchmarkSprintfPrefixedInt-8           770           67  -91.23%
BenchmarkSprintfPrefixedInt-16          829           54  -93.49%
BenchmarkSprintfPrefixedInt-32          824           50  -93.83%

BenchmarkSprintfFloat                   418          398   -4.78%
BenchmarkSprintfFloat-2                 295          203  -31.19%
BenchmarkSprintfFloat-4                 585          128  -78.12%
BenchmarkSprintfFloat-8                 873           60  -93.13%
BenchmarkSprintfFloat-16                884           33  -96.24%
BenchmarkSprintfFloat-32                881           29  -96.62%

BenchmarkManyArgs                      1097         1069   -2.55%
BenchmarkManyArgs-2                     705          567  -19.57%
BenchmarkManyArgs-4                     792          319  -59.72%
BenchmarkManyArgs-8                     963          172  -82.14%
BenchmarkManyArgs-16                   1115          103  -90.76%
BenchmarkManyArgs-32                   1133           90  -92.03%

LGTM=rsc
R=golang-codereviews, bradfitz, minux.ma, gobot, rsc
CC=golang-codereviews
https://golang.org/cl/46010043

Author: dvyukov

src/pkg/go/build/deps_test.go

@@ -29,7 +29,7 @@ var pkgDeps = map[string][]string{
 	"errors":      {},
 	"io":          {"errors", "sync"},
 	"runtime":     {"unsafe"},
-	"sync":        {"sync/atomic", "unsafe"},
+	"sync":        {"runtime", "sync/atomic", "unsafe"},
 	"sync/atomic": {"unsafe"},
 	"unsafe":      {},
 

src/pkg/runtime/mgc0.c

@@ -68,15 +68,19 @@ clearpools(void)
 {
 	void **pool, **next;
 	P *p, **pp;
+	uintptr off;
 	int32 i;
 
 	// clear sync.Pool's
 	for(pool = pools.head; pool != nil; pool = next) {
 		next = pool[0];
 		pool[0] = nil; // next
-		pool[1] = nil; // slice
-		pool[2] = nil;
-		pool[3] = nil;
+		pool[1] = nil; // local
+		pool[2] = nil; // localSize
+		off = (uintptr)pool[3] / sizeof(void*);
+		pool[off+0] = nil; // global slice
+		pool[off+1] = nil;
+		pool[off+2] = nil;
 	}
 	pools.head = nil;
 

src/pkg/runtime/proc.c

@@ -3046,3 +3046,23 @@ haveexperiment(int8 *name)
 	}
 	return 0;
 }
+
+// func runtime_procPin() int
+void
+sync·runtime_procPin(intgo p)
+{
+	M *mp;
+
+	mp = m;
+	// Disable preemption.
+	mp->locks++;
+	p = mp->p->id;
+	FLUSH(&p);
+}
+
+// func runtime_procUnpin()
+void
+sync·runtime_procUnpin(void)
+{
+	m->locks--;
+}

src/pkg/sync/pool.go

@@ -4,6 +4,18 @@
 
 package sync
 
+import (
+	"runtime"
+	"sync/atomic"
+	"unsafe"
+)
+
+const (
+	cacheLineSize = 128
+	poolLocalSize = 2 * cacheLineSize
+	poolLocalCap  = poolLocalSize/unsafe.Sizeof(*(*interface{})(nil)) - 1
+)
+
 // A Pool is a set of temporary objects that may be individually saved
 // and retrieved.
 //
@@ -26,29 +38,52 @@ package sync
 //
 // This is an experimental type and might not be released.
 type Pool struct {
-	next *Pool         // for use by runtime. must be first.
-	list []interface{} // offset known to runtime
-	mu   Mutex         // guards list
+	// The following fields are known to runtime.
+	next         *Pool      // for use by runtime
+	local        *poolLocal // local fixed-size per-P pool, actually an array
+	localSize    uintptr    // size of the local array
+	globalOffset uintptr    // offset of global
+	// The rest is not known to runtime.
 
 	// New optionally specifies a function to generate
 	// a value when Get would otherwise return nil.
 	// It may not be changed concurrently with calls to Get.
 	New func() interface{}
+
+	pad [cacheLineSize]byte
+	// Read-mostly date above this point, mutable data follows.
+	mu     Mutex
+	global []interface{} // global fallback pool
 }
 
-func runtime_registerPool(*Pool)
+// Local per-P Pool appendix.
+type poolLocal struct {
+	tail   int
+	unused int
+	buf    [poolLocalCap]interface{}
+}
+
+func init() {
+	var v poolLocal
+	if unsafe.Sizeof(v) != poolLocalSize {
+		panic("sync: incorrect pool size")
+	}
+}
 
 // Put adds x to the pool.
 func (p *Pool) Put(x interface{}) {
 	if x == nil {
 		return
 	}
-	p.mu.Lock()
-	if p.list == nil {
-		runtime_registerPool(p)
+	l := p.pin()
+	t := l.tail
+	if t < int(poolLocalCap) {
+		l.buf[t] = x
+		l.tail = t + 1
+		runtime_procUnpin()
+		return
 	}
-	p.list = append(p.list, x)
-	p.mu.Unlock()
+	p.putSlow(l, x)
 }
 
 // Get selects an arbitrary item from the Pool, removes it from the
@@ -60,16 +95,116 @@ func (p *Pool) Put(x interface{}) {
 // If Get would otherwise return nil and p.New is non-nil, Get returns
 // the result of calling p.New.
 func (p *Pool) Get() interface{} {
+	l := p.pin()
+	t := l.tail
+	if t > 0 {
+		t -= 1
+		x := l.buf[t]
+		l.tail = t
+		runtime_procUnpin()
+		return x
+	}
+	return p.getSlow()
+}
+
+func (p *Pool) putSlow(l *poolLocal, x interface{}) {
+	// Grab half of items from local pool and put to global pool.
+	// Can not lock the mutex while pinned.
+	const N = int(poolLocalCap/2 + 1)
+	var buf [N]interface{}
+	buf[0] = x
+	for i := 1; i < N; i++ {
+		l.tail--
+		buf[i] = l.buf[l.tail]
+	}
+	runtime_procUnpin()
+
 	p.mu.Lock()
-	var x interface{}
-	if n := len(p.list); n > 0 {
-		x = p.list[n-1]
-		p.list[n-1] = nil // Just to be safe
-		p.list = p.list[:n-1]
+	p.global = append(p.global, buf[:]...)
+	p.mu.Unlock()
+}
+
+func (p *Pool) getSlow() (x interface{}) {
+	// Grab a batch of items from global pool and put to local pool.
+	// Can not lock the mutex while pinned.
+	runtime_procUnpin()
+	p.mu.Lock()
+	pid := runtime_procPin()
+	s := p.localSize
+	l := p.local
+	if uintptr(pid) < s {
+		l = indexLocal(l, pid)
+		// Get the item to return.
+		last := len(p.global) - 1
+		if last >= 0 {
+			x = p.global[last]
+			p.global = p.global[:last]
+		}
+		// Try to refill local pool, we may have been rescheduled to another P.
+		if last > 0 && l.tail == 0 {
+			n := int(poolLocalCap / 2)
+			gl := len(p.global)
+			if n > gl {
+				n = gl
+			}
+			copy(l.buf[:], p.global[gl-n:])
+			p.global = p.global[:gl-n]
+			l.tail = n
+		}
 	}
+	runtime_procUnpin()
 	p.mu.Unlock()
+
 	if x == nil && p.New != nil {
 		x = p.New()
 	}
-	return x
+	return
 }
+
+// pin pins current goroutine to P, disables preemption and returns poolLocal pool for the P.
+// Caller must call runtime_procUnpin() when done with the pool.
+func (p *Pool) pin() *poolLocal {
+	pid := runtime_procPin()
+	// In pinSlow we store to localSize and then to local, here we load in opposite order.
+	// Since we've disabled preemption, GC can not happen in between.
+	// Thus here we must observe local at least as large localSize.
+	// We can observe a newer/larger local, it is fine (we must observe its zero-initialized-ness).
+	s := atomic.LoadUintptr(&p.localSize) // load-acquire
+	l := p.local                          // load-consume
+	if uintptr(pid) < s {
+		return indexLocal(l, pid)
+	}
+	return p.pinSlow()
+}
+
+func (p *Pool) pinSlow() *poolLocal {
+	// Retry under the mutex.
+	runtime_procUnpin()
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	pid := runtime_procPin()
+	s := p.localSize
+	l := p.local
+	if uintptr(pid) < s {
+		return indexLocal(l, pid)
+	}
+	if p.local == nil {
+		p.globalOffset = unsafe.Offsetof(p.global)
+		runtime_registerPool(p)
+	}
+	// If GOMAXPROCS changes between GCs, we re-allocate the array and lose the old one.
+	size := runtime.GOMAXPROCS(0)
+	local := make([]poolLocal, size)
+	atomic.StorePointer((*unsafe.Pointer)(unsafe.Pointer(&p.local)), unsafe.Pointer(&local[0])) // store-release
+	atomic.StoreUintptr(&p.localSize, uintptr(size))                                            // store-release
+	return &local[pid]
+}
+
+func indexLocal(l *poolLocal, i int) *poolLocal {
+	return (*poolLocal)(unsafe.Pointer(uintptr(unsafe.Pointer(l)) + unsafe.Sizeof(*l)*uintptr(i))) // uh...
+}
+
+// Implemented in runtime.
+func runtime_registerPool(*Pool)
+func runtime_procPin() int
+func runtime_procUnpin()

src/pkg/sync/pool_test.go

@@ -11,7 +11,6 @@ import (
 	"sync/atomic"
 	"testing"
 	"time"
-	"unsafe"
 )
 
 func TestPool(t *testing.T) {
@@ -125,28 +124,41 @@ func TestPoolStress(t *testing.T) {
 }
 
 func BenchmarkPool(b *testing.B) {
-	procs := runtime.GOMAXPROCS(-1)
-	var dec func() bool
-	if unsafe.Sizeof(b.N) == 8 {
-		n := int64(b.N)
-		dec = func() bool {
-			return atomic.AddInt64(&n, -1) >= 0
-		}
-	} else {
-		n := int32(b.N)
-		dec = func() bool {
-			return atomic.AddInt32(&n, -1) >= 0
-		}
+	var p Pool
+	var wg WaitGroup
+	n0 := uintptr(b.N)
+	n := n0
+	for i := 0; i < runtime.GOMAXPROCS(0); i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			for atomic.AddUintptr(&n, ^uintptr(0)) < n0 {
+				for b := 0; b < 100; b++ {
+					p.Put(1)
+					p.Get()
+				}
+			}
+		}()
 	}
+	wg.Wait()
+}
+
+func BenchmarkPoolOverlflow(b *testing.B) {
 	var p Pool
 	var wg WaitGroup
-	for i := 0; i < procs; i++ {
+	n0 := uintptr(b.N)
+	n := n0
+	for i := 0; i < runtime.GOMAXPROCS(0); i++ {
 		wg.Add(1)
 		go func() {
 			defer wg.Done()
-			for dec() {
-				p.Put(1)
-				p.Get()
+			for atomic.AddUintptr(&n, ^uintptr(0)) < n0 {
+				for b := 0; b < 100; b++ {
+					p.Put(1)
+				}
+				for b := 0; b < 100; b++ {
+					p.Get()
+				}
 			}
 		}()
 	}