bytes: add padding in bytes.Buffer

Add padding in bytes.Buffer for avoid false sharing.
BufferPaddingSize has also been added for this into internal/cpu,
so that the padding is specific to each architecture.

Signed-off-by: Maxim Eryomenko <[email protected]>

Author: moeryomenko

src/bytes/buffer.go

@@ -8,6 +8,7 @@ package bytes
 
 import (
 	"errors"
+	"internal/cpu"
 	"io"
 	"unicode/utf8"
 )
@@ -22,15 +23,15 @@ type Buffer struct {
 	off      int    // read at &buf[off], write at &buf[len(buf)]
 	lastRead readOp // last read operation, so that Unread* can work correctly.
 
-	// FIXME: it would be advisable to align Buffer to cachelines to avoid false
-	// sharing.
+	// This is padding for Buffer to cache lines to avoid false sharing.
+	_ [cpu.BufferPaddingSize]byte
 }
 
 // The readOp constants describe the last action performed on
 // the buffer, so that UnreadRune and UnreadByte can check for
 // invalid usage. opReadRuneX constants are chosen such that
 // converted to int they correspond to the rune size that was read.
-type readOp int8
+type readOp int
 
 // Don't use iota for these, as the values need to correspond with the
 // names and comments, which is easier to see when being explicit.

src/internal/cpu/cpu.go

@@ -19,6 +19,8 @@ type CacheLinePad struct{ _ [CacheLinePadSize]byte }
 // so we use the constant per GOARCH CacheLinePadSize as an approximation.
 var CacheLineSize uintptr = CacheLinePadSize
 
+var BufferPaddingSize uintptr = BufferCacheLinePadSize
+
 var X86 x86
 
 // The booleans in x86 contain the correspondingly named cpuid feature bit.

src/internal/cpu/cpu_386.go

@@ -4,4 +4,7 @@
 
 package cpu
 
-const GOARCH = "386"
+const (
+	GOARCH                 = "386"
+	BufferCacheLinePadSize = 44
+)

src/internal/cpu/cpu_amd64.go

@@ -4,4 +4,7 @@
 
 package cpu
 
-const GOARCH = "amd64"
+const (
+	GOARCH                 = "amd64"
+	BufferCacheLinePadSize = 24
+)

src/internal/cpu/cpu_amd64p32.go

@@ -4,4 +4,7 @@
 
 package cpu
 
-const GOARCH = "amd64p32"
+const (
+	GOARCH                 = "amd64p32"
+	BufferCacheLinePadSize = 44
+)

src/internal/cpu/cpu_arm.go

@@ -4,7 +4,10 @@
 
 package cpu
 
-const CacheLinePadSize = 32
+const (
+	CacheLinePadSize       = 32
+	BufferCacheLinePadSize = 12
+)
 
 // arm doesn't have a 'cpuid' equivalent, so we rely on HWCAP/HWCAP2.
 // These are linknamed in runtime/os_(linux|freebsd)_arm.go and are

src/internal/cpu/cpu_arm64.go

@@ -4,7 +4,10 @@
 
 package cpu
 
-const CacheLinePadSize = 64
+const (
+	CacheLinePadSize       = 64
+	BufferCacheLinePadSize = 24
+)
 
 // arm64 doesn't have a 'cpuid' equivalent, so we rely on HWCAP/HWCAP2.
 // These are initialized by archauxv in runtime/os_linux_arm64.go.

src/internal/cpu/cpu_mips.go

@@ -4,4 +4,7 @@
 
 package cpu
 
-const CacheLinePadSize = 32
+const (
+	CacheLinePadSize       = 32
+	BufferCacheLinePadSize = 12
+)

src/internal/cpu/cpu_mips64.go

@@ -4,4 +4,7 @@
 
 package cpu
 
-const CacheLinePadSize = 32
+const (
+	CacheLinePadSize       = 32
+	BufferCacheLinePadSize = 24
+)

src/internal/cpu/cpu_mips64le.go

@@ -4,4 +4,7 @@
 
 package cpu
 
-const CacheLinePadSize = 32
+const (
+	CacheLinePadSize       = 32
+	BufferCacheLinePadSize = 24
+)

src/internal/cpu/cpu_mipsle.go

@@ -4,4 +4,7 @@
 
 package cpu
 
-const CacheLinePadSize = 32
+const (
+	CacheLinePadSize       = 32
+	BufferCacheLinePadSize = 12
+)

src/internal/cpu/cpu_ppc64x.go

@@ -6,7 +6,10 @@
 
 package cpu
 
-const CacheLinePadSize = 128
+const (
+	CacheLinePadSize       = 128
+	BufferCacheLinePadSize = 88
+)
 
 // ppc64x doesn't have a 'cpuid' equivalent, so we rely on HWCAP/HWCAP2.
 // These are initialized by archauxv in runtime/os_linux_ppc64x.go.

src/internal/cpu/cpu_s390x.go

@@ -4,7 +4,10 @@
 
 package cpu
 
-const CacheLinePadSize = 256
+const (
+	CacheLinePadSize       = 256
+	BufferCacheLinePadSize = 216
+)
 
 // bitIsSet reports whether the bit at index is set. The bit index
 // is in big endian order, so bit index 0 is the leftmost bit.

src/internal/cpu/cpu_wasm.go

@@ -4,4 +4,7 @@
 
 package cpu
 
-const CacheLinePadSize = 64
+const (
+	CacheLinePadSize       = 64
+	BufferCacheLinePadSize = 44
+)