runtime: use multiplication with overflow check for makeslice

This improves performance for slices with an element size larger
than 32 bytes and removes loading a value from the maxElems
array for smaller element sizes.

name                 old time/op  new time/op  delta
MakeSlice/Byte       18.0ns ± 4%  18.0ns ± 2%     ~     (p=0.575 n=20+17)
MakeSlice/Int16      21.8ns ± 2%  21.6ns ± 1%   -0.63%  (p=0.035 n=20+19)
MakeSlice/Int        42.0ns ± 2%  41.6ns ± 1%     ~     (p=0.121 n=20+18)
MakeSlice/Ptr        62.6ns ± 2%  62.4ns ± 2%     ~     (p=0.491 n=20+18)
MakeSlice/Struct/24  57.4ns ± 3%  56.0ns ± 2%   -2.40%  (p=0.000 n=19+19)
MakeSlice/Struct/32  62.1ns ± 2%  60.6ns ± 3%   -2.43%  (p=0.000 n=20+20)
MakeSlice/Struct/40  77.3ns ± 3%  68.9ns ± 3%  -10.91%  (p=0.000 n=20+20)

Updates #21588

Change-Id: Ie12807bf8f77c0e15453413f47e3d7de771b798f
Reviewed-on: https://go-review.googlesource.com/c/142377
Run-TryBot: Martin Möhrmann <martisch@uos.de>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

src/runtime/slice.go

@@ -54,21 +54,21 @@ func panicmakeslicecap() {
 }
 
 func makeslice(et *_type, len, cap int) slice {
-	// NOTE: The len > maxElements check here is not strictly necessary,
-	// but it produces a 'len out of range' error instead of a 'cap out of range' error
-	// when someone does make([]T, bignumber). 'cap out of range' is true too,
-	// but since the cap is only being supplied implicitly, saying len is clearer.
-	// See issue 4085.
-	maxElements := maxSliceCap(et.size)
-	if len < 0 || uintptr(len) > maxElements {
-		panicmakeslicelen()
-	}
-
-	if cap < len || uintptr(cap) > maxElements {
+	mem, overflow := math.MulUintptr(et.size, uintptr(cap))
+	if overflow || mem > maxAlloc || len < 0 || len > cap {
+		// NOTE: Produce a 'len out of range' error instead of a
+		// 'cap out of range' error when someone does make([]T, bignumber).
+		// 'cap out of range' is true too, but since the cap is only being
+		// supplied implicitly, saying len is clearer.
+		// See golang.org/issue/4085.
+		mem, overflow := math.MulUintptr(et.size, uintptr(len))
+		if overflow || mem > maxAlloc || len < 0 {
+			panicmakeslicelen()
+		}
 		panicmakeslicecap()
 	}
+	p := mallocgc(mem, et, true)
 
-	p := mallocgc(et.size*uintptr(cap), et, true)
 	return slice{p, len, cap}
 }
 

src/runtime/slice_test.go

@@ -10,20 +10,68 @@ import (
 
 const N = 20
 
-func BenchmarkMakeSlice(b *testing.B) {
-	var x []byte
-	for i := 0; i < b.N; i++ {
-		x = make([]byte, 32)
-		_ = x
-	}
-}
-
 type (
 	struct24 struct{ a, b, c int64 }
 	struct32 struct{ a, b, c, d int64 }
 	struct40 struct{ a, b, c, d, e int64 }
 )
 
+func BenchmarkMakeSlice(b *testing.B) {
+	const length = 2
+	b.Run("Byte", func(b *testing.B) {
+		var x []byte
+		for i := 0; i < b.N; i++ {
+			x = make([]byte, length, 2*length)
+			_ = x
+		}
+	})
+	b.Run("Int16", func(b *testing.B) {
+		var x []int16
+		for i := 0; i < b.N; i++ {
+			x = make([]int16, length, 2*length)
+			_ = x
+		}
+	})
+	b.Run("Int", func(b *testing.B) {
+		var x []int
+		for i := 0; i < b.N; i++ {
+			x = make([]int, length, 2*length)
+			_ = x
+		}
+	})
+	b.Run("Ptr", func(b *testing.B) {
+		var x []*byte
+		for i := 0; i < b.N; i++ {
+			x = make([]*byte, length, 2*length)
+			_ = x
+		}
+	})
+	b.Run("Struct", func(b *testing.B) {
+		b.Run("24", func(b *testing.B) {
+			var x []struct24
+			for i := 0; i < b.N; i++ {
+				x = make([]struct24, length, 2*length)
+				_ = x
+			}
+		})
+		b.Run("32", func(b *testing.B) {
+			var x []struct32
+			for i := 0; i < b.N; i++ {
+				x = make([]struct32, length, 2*length)
+				_ = x
+			}
+		})
+		b.Run("40", func(b *testing.B) {
+			var x []struct40
+			for i := 0; i < b.N; i++ {
+				x = make([]struct40, length, 2*length)
+				_ = x
+			}
+		})
+
+	})
+}
+
 func BenchmarkGrowSlice(b *testing.B) {
 	b.Run("Byte", func(b *testing.B) {
 		x := make([]byte, 9)