maps,runtime: improve maps.Keys

name     old time/op    new time/op    delta
Keys-10    8.65ms ± 0%    7.06ms ± 0%  -18.41%  (p=0.000 n=9+9)

name     old alloc/op   new alloc/op   delta
Keys-10    58.2kB ± 1%    47.4kB ± 2%  -18.54%  (p=0.000 n=10+10)

name     old allocs/op  new allocs/op  delta
Keys-10      0.00           0.00          ~     (all equal)

Change-Id: Ia7061c37be89e906e79bc3ef3bb1ef0d7913f9f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/481435
Reviewed-by: Keith Randall <[email protected]>
TryBot-Result: Gopher Robot <[email protected]>
Reviewed-by: Keith Randall <[email protected]>
Reviewed-by: Heschi Kreinick <[email protected]>
Run-TryBot: xie cui <[email protected]>

src/go/build/deps_test.go

@@ -46,7 +46,7 @@ var depsRules = `
 	  internal/goexperiment, internal/goos,
 	  internal/goversion, internal/nettrace, internal/platform,
 	  log/internal,
-	  maps, unicode/utf8, unicode/utf16, unicode,
+	  unicode/utf8, unicode/utf16, unicode,
 	  unsafe;
 
 	# slices depends on unsafe for overlapping check.
@@ -57,6 +57,8 @@ var depsRules = `
 	internal/goarch, unsafe
 	< internal/abi;
 
+	unsafe < maps;
+
 	# RUNTIME is the core runtime group of packages, all of them very light-weight.
 	internal/abi, internal/cpu, internal/goarch,
 	internal/coverage/rtcov, internal/godebugs, internal/goexperiment,

src/maps/maps.go

@@ -5,16 +5,25 @@
 // Package maps defines various functions useful with maps of any type.
 package maps
 
+import "unsafe"
+
+// keys is implemented in the runtime package.
+//
+//go:noescape
+func keys(m any, slice unsafe.Pointer)
+
 // Keys returns the keys of the map m.
 // The keys will be in an indeterminate order.
 func Keys[M ~map[K]V, K comparable, V any](m M) []K {
 	r := make([]K, 0, len(m))
-	for k := range m {
-		r = append(r, k)
-	}
+	keys(m, unsafe.Pointer(&r))
 	return r
 }
 
+func keysForBenchmarking[M ~map[K]V, K comparable, V any](m M, s []K) {
+	keys(m, unsafe.Pointer(&s))
+}
+
 // Values returns the values of the map m.
 // The values will be in an indeterminate order.
 func Values[M ~map[K]V, K comparable, V any](m M) []V {

src/maps/maps_test.go

@@ -29,6 +29,23 @@ func TestKeys(t *testing.T) {
 	if !slices.Equal(got2, want) {
 		t.Errorf("Keys(%v) = %v, want %v", m2, got2, want)
 	}
+
+	// test for oldbucket code path
+	// We grow from 128 to 256 buckets at size 832 (6.5 * 128).
+	// Then we have to evacuate 128 buckets, which means we'll be done evacuation at 832+128=960 elements inserted.
+	// so 840 is a good number to test for oldbucket code path.
+	var want3 []int
+	var m = make(map[int]int)
+	for i := 0; i < 840; i++ {
+		want3 = append(want3, i)
+		m[i] = i
+	}
+
+	got3 := Keys(m)
+	sort.Ints(got3)
+	if !slices.Equal(got3, want3) {
+		t.Errorf("Keys(%v) = %v, want %v", m, got3, want3)
+	}
 }
 
 func TestValues(t *testing.T) {
@@ -216,3 +233,18 @@ func TestCloneWithMapAssign(t *testing.T) {
 		}
 	}
 }
+
+var keysArr []int
+
+func BenchmarkKeys(b *testing.B) {
+	m := make(map[int]int, 1000000)
+	for i := 0; i < 1000000; i++ {
+		m[i] = i
+	}
+	b.ResetTimer()
+
+	slice := make([]int, 0, len(m))
+	for i := 0; i < b.N; i++ {
+		keysForBenchmarking(m, slice)
+	}
+}

src/runtime/map.go

@@ -1593,3 +1593,67 @@ func mapclone2(t *maptype, src *hmap) *hmap {
 	}
 	return dst
 }
+
+// keys for implementing maps.keys
+//
+//go:linkname keys maps.keys
+func keys(m any, p unsafe.Pointer) {
+	e := efaceOf(&m)
+	t := (*maptype)(unsafe.Pointer(e._type))
+	h := (*hmap)(e.data)
+
+	if h == nil || h.count == 0 {
+		return
+	}
+	s := (*slice)(p)
+	r := int(fastrand())
+	offset := uint8(r >> h.B & (bucketCnt - 1))
+	if h.B == 0 {
+		copyKeys(t, h, (*bmap)(h.buckets), s, offset)
+		return
+	}
+	arraySize := int(bucketShift(h.B))
+	buckets := h.buckets
+	for i := 0; i < arraySize; i++ {
+		bucket := (i + r) & (arraySize - 1)
+		b := (*bmap)(add(buckets, uintptr(bucket)*uintptr(t.BucketSize)))
+		copyKeys(t, h, b, s, offset)
+	}
+
+	if h.growing() {
+		oldArraySize := int(h.noldbuckets())
+		for i := 0; i < oldArraySize; i++ {
+			bucket := (i + r) & (oldArraySize - 1)
+			b := (*bmap)(add(h.oldbuckets, uintptr(bucket)*uintptr(t.BucketSize)))
+			if evacuated(b) {
+				continue
+			}
+			copyKeys(t, h, b, s, offset)
+		}
+	}
+	return
+}
+
+func copyKeys(t *maptype, h *hmap, b *bmap, s *slice, offset uint8) {
+	for b != nil {
+		for i := uintptr(0); i < bucketCnt; i++ {
+			offi := (i + uintptr(offset)) & (bucketCnt - 1)
+			if isEmpty(b.tophash[offi]) {
+				continue
+			}
+			if h.flags&hashWriting != 0 {
+				fatal("concurrent map read and map write")
+			}
+			k := add(unsafe.Pointer(b), dataOffset+offi*uintptr(t.KeySize))
+			if t.IndirectKey() {
+				k = *((*unsafe.Pointer)(k))
+			}
+			if s.len >= s.cap {
+				fatal("concurrent map read and map write")
+			}
+			typedmemmove(t.Key, add(s.array, uintptr(s.len)*uintptr(t.KeySize)), k)
+			s.len++
+		}
+		b = b.overflow(t)
+	}
+}