protobuf-go/internal/impl/encode.go

184 lines
4.6 KiB
Go
Raw Normal View History

internal/impl: add fast-path marshal implementation This is a port of the v1 table marshaler, with some substantial cleanup and refactoring. Benchstat results from the protobuf reference benchmark data comparing the v1 package with v2, with AllowPartial:true set for the new package. This is not an apples-to-apples comparison, since v1 doesn't have a way to disable required field checks. Required field checks in v2 package currently go through reflection, which performs terribly; my initial experimentation indicates that fast-path required field checks will not add a large amount of cost; these results are incomplete but not wholly inaccurate. name old time/op new time/op delta /dataset.google_message3_1.pb/Marshal-12 219ms ± 1% 232ms ± 1% +5.85% (p=0.004 n=6+5) /dataset.google_message2.pb/Marshal-12 261µs ± 3% 248µs ± 1% -5.14% (p=0.002 n=6+6) /dataset.google_message1_proto2.pb/Marshal-12 681ns ± 2% 637ns ± 3% -6.53% (p=0.002 n=6+6) /dataset.google_message1_proto3.pb/Marshal-12 1.10µs ± 8% 0.99µs ± 3% -9.63% (p=0.002 n=6+6) /dataset.google_message3_3.pb/Marshal-12 44.2ms ± 3% 35.2ms ± 1% -20.28% (p=0.004 n=6+5) /dataset.google_message4.pb/Marshal-12 91.4ms ± 2% 94.9ms ± 2% +3.78% (p=0.002 n=6+6) /dataset.google_message3_2.pb/Marshal-12 78.7ms ± 6% 80.8ms ± 4% ~ (p=0.310 n=6+6) /dataset.google_message3_4.pb/Marshal-12 10.6ms ± 3% 10.6ms ± 8% ~ (p=0.662 n=5+6) /dataset.google_message3_5.pb/Marshal-12 675ms ± 4% 510ms ± 2% -24.40% (p=0.002 n=6+6) /dataset.google_message3_1.pb/Marshal 219ms ± 1% 236ms ± 7% +8.06% (p=0.004 n=5+6) /dataset.google_message2.pb/Marshal 257µs ± 1% 250µs ± 3% ~ (p=0.052 n=5+6) /dataset.google_message1_proto2.pb/Marshal 685ns ± 1% 628ns ± 1% -8.41% (p=0.008 n=5+5) /dataset.google_message1_proto3.pb/Marshal 1.08µs ± 1% 0.98µs ± 2% -9.31% (p=0.004 n=5+6) /dataset.google_message3_3.pb/Marshal 43.7ms ± 1% 35.1ms ± 1% -19.76% (p=0.002 n=6+6) /dataset.google_message4.pb/Marshal 93.4ms ± 4% 94.9ms ± 2% ~ (p=0.180 n=6+6) /dataset.google_message3_2.pb/Marshal 105ms ± 2% 98ms ± 7% -6.81% (p=0.009 n=5+6) /dataset.google_message3_4.pb/Marshal 16.3ms ± 6% 15.7ms ± 3% -3.44% (p=0.041 n=6+6) /dataset.google_message3_5.pb/Marshal 676ms ± 4% 504ms ± 2% -25.50% (p=0.004 n=6+5) Change-Id: I72cc4597117f4cf5d236ef505777d49dd4a5f75d Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/171020 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
2019-04-01 20:49:56 +00:00
// Copyright 2019 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package impl
import (
"sort"
"sync/atomic"
"google.golang.org/protobuf/internal/errors"
proto "google.golang.org/protobuf/proto"
pref "google.golang.org/protobuf/reflect/protoreflect"
piface "google.golang.org/protobuf/runtime/protoiface"
)
type marshalOptions uint
const (
marshalAllowPartial marshalOptions = 1 << iota
marshalDeterministic
marshalUseCachedSize
)
func newMarshalOptions(opts piface.MarshalOptions) marshalOptions {
var o marshalOptions
if opts.AllowPartial {
o |= marshalAllowPartial
}
if opts.Deterministic {
o |= marshalDeterministic
}
if opts.UseCachedSize {
o |= marshalUseCachedSize
}
return o
}
func (o marshalOptions) Options() proto.MarshalOptions {
return proto.MarshalOptions{
AllowPartial: o.AllowPartial(),
Deterministic: o.Deterministic(),
UseCachedSize: o.UseCachedSize(),
}
}
func (o marshalOptions) AllowPartial() bool { return o&marshalAllowPartial != 0 }
func (o marshalOptions) Deterministic() bool { return o&marshalDeterministic != 0 }
func (o marshalOptions) UseCachedSize() bool { return o&marshalUseCachedSize != 0 }
// size is protoreflect.Methods.Size.
func (mi *MessageInfo) size(msg pref.ProtoMessage) (size int) {
internal/impl: add fast-path marshal implementation This is a port of the v1 table marshaler, with some substantial cleanup and refactoring. Benchstat results from the protobuf reference benchmark data comparing the v1 package with v2, with AllowPartial:true set for the new package. This is not an apples-to-apples comparison, since v1 doesn't have a way to disable required field checks. Required field checks in v2 package currently go through reflection, which performs terribly; my initial experimentation indicates that fast-path required field checks will not add a large amount of cost; these results are incomplete but not wholly inaccurate. name old time/op new time/op delta /dataset.google_message3_1.pb/Marshal-12 219ms ± 1% 232ms ± 1% +5.85% (p=0.004 n=6+5) /dataset.google_message2.pb/Marshal-12 261µs ± 3% 248µs ± 1% -5.14% (p=0.002 n=6+6) /dataset.google_message1_proto2.pb/Marshal-12 681ns ± 2% 637ns ± 3% -6.53% (p=0.002 n=6+6) /dataset.google_message1_proto3.pb/Marshal-12 1.10µs ± 8% 0.99µs ± 3% -9.63% (p=0.002 n=6+6) /dataset.google_message3_3.pb/Marshal-12 44.2ms ± 3% 35.2ms ± 1% -20.28% (p=0.004 n=6+5) /dataset.google_message4.pb/Marshal-12 91.4ms ± 2% 94.9ms ± 2% +3.78% (p=0.002 n=6+6) /dataset.google_message3_2.pb/Marshal-12 78.7ms ± 6% 80.8ms ± 4% ~ (p=0.310 n=6+6) /dataset.google_message3_4.pb/Marshal-12 10.6ms ± 3% 10.6ms ± 8% ~ (p=0.662 n=5+6) /dataset.google_message3_5.pb/Marshal-12 675ms ± 4% 510ms ± 2% -24.40% (p=0.002 n=6+6) /dataset.google_message3_1.pb/Marshal 219ms ± 1% 236ms ± 7% +8.06% (p=0.004 n=5+6) /dataset.google_message2.pb/Marshal 257µs ± 1% 250µs ± 3% ~ (p=0.052 n=5+6) /dataset.google_message1_proto2.pb/Marshal 685ns ± 1% 628ns ± 1% -8.41% (p=0.008 n=5+5) /dataset.google_message1_proto3.pb/Marshal 1.08µs ± 1% 0.98µs ± 2% -9.31% (p=0.004 n=5+6) /dataset.google_message3_3.pb/Marshal 43.7ms ± 1% 35.1ms ± 1% -19.76% (p=0.002 n=6+6) /dataset.google_message4.pb/Marshal 93.4ms ± 4% 94.9ms ± 2% ~ (p=0.180 n=6+6) /dataset.google_message3_2.pb/Marshal 105ms ± 2% 98ms ± 7% -6.81% (p=0.009 n=5+6) /dataset.google_message3_4.pb/Marshal 16.3ms ± 6% 15.7ms ± 3% -3.44% (p=0.041 n=6+6) /dataset.google_message3_5.pb/Marshal 676ms ± 4% 504ms ± 2% -25.50% (p=0.004 n=6+5) Change-Id: I72cc4597117f4cf5d236ef505777d49dd4a5f75d Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/171020 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
2019-04-01 20:49:56 +00:00
return mi.sizePointer(pointerOfIface(msg), 0)
}
func (mi *MessageInfo) sizePointer(p pointer, opts marshalOptions) (size int) {
internal/impl: add fast-path marshal implementation This is a port of the v1 table marshaler, with some substantial cleanup and refactoring. Benchstat results from the protobuf reference benchmark data comparing the v1 package with v2, with AllowPartial:true set for the new package. This is not an apples-to-apples comparison, since v1 doesn't have a way to disable required field checks. Required field checks in v2 package currently go through reflection, which performs terribly; my initial experimentation indicates that fast-path required field checks will not add a large amount of cost; these results are incomplete but not wholly inaccurate. name old time/op new time/op delta /dataset.google_message3_1.pb/Marshal-12 219ms ± 1% 232ms ± 1% +5.85% (p=0.004 n=6+5) /dataset.google_message2.pb/Marshal-12 261µs ± 3% 248µs ± 1% -5.14% (p=0.002 n=6+6) /dataset.google_message1_proto2.pb/Marshal-12 681ns ± 2% 637ns ± 3% -6.53% (p=0.002 n=6+6) /dataset.google_message1_proto3.pb/Marshal-12 1.10µs ± 8% 0.99µs ± 3% -9.63% (p=0.002 n=6+6) /dataset.google_message3_3.pb/Marshal-12 44.2ms ± 3% 35.2ms ± 1% -20.28% (p=0.004 n=6+5) /dataset.google_message4.pb/Marshal-12 91.4ms ± 2% 94.9ms ± 2% +3.78% (p=0.002 n=6+6) /dataset.google_message3_2.pb/Marshal-12 78.7ms ± 6% 80.8ms ± 4% ~ (p=0.310 n=6+6) /dataset.google_message3_4.pb/Marshal-12 10.6ms ± 3% 10.6ms ± 8% ~ (p=0.662 n=5+6) /dataset.google_message3_5.pb/Marshal-12 675ms ± 4% 510ms ± 2% -24.40% (p=0.002 n=6+6) /dataset.google_message3_1.pb/Marshal 219ms ± 1% 236ms ± 7% +8.06% (p=0.004 n=5+6) /dataset.google_message2.pb/Marshal 257µs ± 1% 250µs ± 3% ~ (p=0.052 n=5+6) /dataset.google_message1_proto2.pb/Marshal 685ns ± 1% 628ns ± 1% -8.41% (p=0.008 n=5+5) /dataset.google_message1_proto3.pb/Marshal 1.08µs ± 1% 0.98µs ± 2% -9.31% (p=0.004 n=5+6) /dataset.google_message3_3.pb/Marshal 43.7ms ± 1% 35.1ms ± 1% -19.76% (p=0.002 n=6+6) /dataset.google_message4.pb/Marshal 93.4ms ± 4% 94.9ms ± 2% ~ (p=0.180 n=6+6) /dataset.google_message3_2.pb/Marshal 105ms ± 2% 98ms ± 7% -6.81% (p=0.009 n=5+6) /dataset.google_message3_4.pb/Marshal 16.3ms ± 6% 15.7ms ± 3% -3.44% (p=0.041 n=6+6) /dataset.google_message3_5.pb/Marshal 676ms ± 4% 504ms ± 2% -25.50% (p=0.004 n=6+5) Change-Id: I72cc4597117f4cf5d236ef505777d49dd4a5f75d Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/171020 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
2019-04-01 20:49:56 +00:00
mi.init()
if p.IsNil() {
return 0
}
if opts.UseCachedSize() && mi.sizecacheOffset.IsValid() {
return int(atomic.LoadInt32(p.Apply(mi.sizecacheOffset).Int32()))
}
return mi.sizePointerSlow(p, opts)
}
func (mi *MessageInfo) sizePointerSlow(p pointer, opts marshalOptions) (size int) {
internal/impl: add fast-path marshal implementation This is a port of the v1 table marshaler, with some substantial cleanup and refactoring. Benchstat results from the protobuf reference benchmark data comparing the v1 package with v2, with AllowPartial:true set for the new package. This is not an apples-to-apples comparison, since v1 doesn't have a way to disable required field checks. Required field checks in v2 package currently go through reflection, which performs terribly; my initial experimentation indicates that fast-path required field checks will not add a large amount of cost; these results are incomplete but not wholly inaccurate. name old time/op new time/op delta /dataset.google_message3_1.pb/Marshal-12 219ms ± 1% 232ms ± 1% +5.85% (p=0.004 n=6+5) /dataset.google_message2.pb/Marshal-12 261µs ± 3% 248µs ± 1% -5.14% (p=0.002 n=6+6) /dataset.google_message1_proto2.pb/Marshal-12 681ns ± 2% 637ns ± 3% -6.53% (p=0.002 n=6+6) /dataset.google_message1_proto3.pb/Marshal-12 1.10µs ± 8% 0.99µs ± 3% -9.63% (p=0.002 n=6+6) /dataset.google_message3_3.pb/Marshal-12 44.2ms ± 3% 35.2ms ± 1% -20.28% (p=0.004 n=6+5) /dataset.google_message4.pb/Marshal-12 91.4ms ± 2% 94.9ms ± 2% +3.78% (p=0.002 n=6+6) /dataset.google_message3_2.pb/Marshal-12 78.7ms ± 6% 80.8ms ± 4% ~ (p=0.310 n=6+6) /dataset.google_message3_4.pb/Marshal-12 10.6ms ± 3% 10.6ms ± 8% ~ (p=0.662 n=5+6) /dataset.google_message3_5.pb/Marshal-12 675ms ± 4% 510ms ± 2% -24.40% (p=0.002 n=6+6) /dataset.google_message3_1.pb/Marshal 219ms ± 1% 236ms ± 7% +8.06% (p=0.004 n=5+6) /dataset.google_message2.pb/Marshal 257µs ± 1% 250µs ± 3% ~ (p=0.052 n=5+6) /dataset.google_message1_proto2.pb/Marshal 685ns ± 1% 628ns ± 1% -8.41% (p=0.008 n=5+5) /dataset.google_message1_proto3.pb/Marshal 1.08µs ± 1% 0.98µs ± 2% -9.31% (p=0.004 n=5+6) /dataset.google_message3_3.pb/Marshal 43.7ms ± 1% 35.1ms ± 1% -19.76% (p=0.002 n=6+6) /dataset.google_message4.pb/Marshal 93.4ms ± 4% 94.9ms ± 2% ~ (p=0.180 n=6+6) /dataset.google_message3_2.pb/Marshal 105ms ± 2% 98ms ± 7% -6.81% (p=0.009 n=5+6) /dataset.google_message3_4.pb/Marshal 16.3ms ± 6% 15.7ms ± 3% -3.44% (p=0.041 n=6+6) /dataset.google_message3_5.pb/Marshal 676ms ± 4% 504ms ± 2% -25.50% (p=0.004 n=6+5) Change-Id: I72cc4597117f4cf5d236ef505777d49dd4a5f75d Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/171020 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
2019-04-01 20:49:56 +00:00
if mi.extensionOffset.IsValid() {
e := p.Apply(mi.extensionOffset).Extensions()
size += mi.sizeExtensions(e, opts)
}
for _, f := range mi.fieldsOrdered {
fptr := p.Apply(f.offset)
if f.isPointer && fptr.Elem().IsNil() {
continue
}
if f.funcs.size == nil {
continue
}
size += f.funcs.size(fptr, f.tagsize, opts)
}
if mi.unknownOffset.IsValid() {
u := *p.Apply(mi.unknownOffset).Bytes()
size += len(u)
}
if mi.sizecacheOffset.IsValid() {
atomic.StoreInt32(p.Apply(mi.sizecacheOffset).Int32(), int32(size))
}
return size
}
// marshalAppend is protoreflect.Methods.MarshalAppend.
func (mi *MessageInfo) marshalAppend(b []byte, msg pref.ProtoMessage, opts piface.MarshalOptions) ([]byte, error) {
internal/impl: add fast-path marshal implementation This is a port of the v1 table marshaler, with some substantial cleanup and refactoring. Benchstat results from the protobuf reference benchmark data comparing the v1 package with v2, with AllowPartial:true set for the new package. This is not an apples-to-apples comparison, since v1 doesn't have a way to disable required field checks. Required field checks in v2 package currently go through reflection, which performs terribly; my initial experimentation indicates that fast-path required field checks will not add a large amount of cost; these results are incomplete but not wholly inaccurate. name old time/op new time/op delta /dataset.google_message3_1.pb/Marshal-12 219ms ± 1% 232ms ± 1% +5.85% (p=0.004 n=6+5) /dataset.google_message2.pb/Marshal-12 261µs ± 3% 248µs ± 1% -5.14% (p=0.002 n=6+6) /dataset.google_message1_proto2.pb/Marshal-12 681ns ± 2% 637ns ± 3% -6.53% (p=0.002 n=6+6) /dataset.google_message1_proto3.pb/Marshal-12 1.10µs ± 8% 0.99µs ± 3% -9.63% (p=0.002 n=6+6) /dataset.google_message3_3.pb/Marshal-12 44.2ms ± 3% 35.2ms ± 1% -20.28% (p=0.004 n=6+5) /dataset.google_message4.pb/Marshal-12 91.4ms ± 2% 94.9ms ± 2% +3.78% (p=0.002 n=6+6) /dataset.google_message3_2.pb/Marshal-12 78.7ms ± 6% 80.8ms ± 4% ~ (p=0.310 n=6+6) /dataset.google_message3_4.pb/Marshal-12 10.6ms ± 3% 10.6ms ± 8% ~ (p=0.662 n=5+6) /dataset.google_message3_5.pb/Marshal-12 675ms ± 4% 510ms ± 2% -24.40% (p=0.002 n=6+6) /dataset.google_message3_1.pb/Marshal 219ms ± 1% 236ms ± 7% +8.06% (p=0.004 n=5+6) /dataset.google_message2.pb/Marshal 257µs ± 1% 250µs ± 3% ~ (p=0.052 n=5+6) /dataset.google_message1_proto2.pb/Marshal 685ns ± 1% 628ns ± 1% -8.41% (p=0.008 n=5+5) /dataset.google_message1_proto3.pb/Marshal 1.08µs ± 1% 0.98µs ± 2% -9.31% (p=0.004 n=5+6) /dataset.google_message3_3.pb/Marshal 43.7ms ± 1% 35.1ms ± 1% -19.76% (p=0.002 n=6+6) /dataset.google_message4.pb/Marshal 93.4ms ± 4% 94.9ms ± 2% ~ (p=0.180 n=6+6) /dataset.google_message3_2.pb/Marshal 105ms ± 2% 98ms ± 7% -6.81% (p=0.009 n=5+6) /dataset.google_message3_4.pb/Marshal 16.3ms ± 6% 15.7ms ± 3% -3.44% (p=0.041 n=6+6) /dataset.google_message3_5.pb/Marshal 676ms ± 4% 504ms ± 2% -25.50% (p=0.004 n=6+5) Change-Id: I72cc4597117f4cf5d236ef505777d49dd4a5f75d Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/171020 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
2019-04-01 20:49:56 +00:00
return mi.marshalAppendPointer(b, pointerOfIface(msg), newMarshalOptions(opts))
}
func (mi *MessageInfo) marshalAppendPointer(b []byte, p pointer, opts marshalOptions) ([]byte, error) {
internal/impl: add fast-path marshal implementation This is a port of the v1 table marshaler, with some substantial cleanup and refactoring. Benchstat results from the protobuf reference benchmark data comparing the v1 package with v2, with AllowPartial:true set for the new package. This is not an apples-to-apples comparison, since v1 doesn't have a way to disable required field checks. Required field checks in v2 package currently go through reflection, which performs terribly; my initial experimentation indicates that fast-path required field checks will not add a large amount of cost; these results are incomplete but not wholly inaccurate. name old time/op new time/op delta /dataset.google_message3_1.pb/Marshal-12 219ms ± 1% 232ms ± 1% +5.85% (p=0.004 n=6+5) /dataset.google_message2.pb/Marshal-12 261µs ± 3% 248µs ± 1% -5.14% (p=0.002 n=6+6) /dataset.google_message1_proto2.pb/Marshal-12 681ns ± 2% 637ns ± 3% -6.53% (p=0.002 n=6+6) /dataset.google_message1_proto3.pb/Marshal-12 1.10µs ± 8% 0.99µs ± 3% -9.63% (p=0.002 n=6+6) /dataset.google_message3_3.pb/Marshal-12 44.2ms ± 3% 35.2ms ± 1% -20.28% (p=0.004 n=6+5) /dataset.google_message4.pb/Marshal-12 91.4ms ± 2% 94.9ms ± 2% +3.78% (p=0.002 n=6+6) /dataset.google_message3_2.pb/Marshal-12 78.7ms ± 6% 80.8ms ± 4% ~ (p=0.310 n=6+6) /dataset.google_message3_4.pb/Marshal-12 10.6ms ± 3% 10.6ms ± 8% ~ (p=0.662 n=5+6) /dataset.google_message3_5.pb/Marshal-12 675ms ± 4% 510ms ± 2% -24.40% (p=0.002 n=6+6) /dataset.google_message3_1.pb/Marshal 219ms ± 1% 236ms ± 7% +8.06% (p=0.004 n=5+6) /dataset.google_message2.pb/Marshal 257µs ± 1% 250µs ± 3% ~ (p=0.052 n=5+6) /dataset.google_message1_proto2.pb/Marshal 685ns ± 1% 628ns ± 1% -8.41% (p=0.008 n=5+5) /dataset.google_message1_proto3.pb/Marshal 1.08µs ± 1% 0.98µs ± 2% -9.31% (p=0.004 n=5+6) /dataset.google_message3_3.pb/Marshal 43.7ms ± 1% 35.1ms ± 1% -19.76% (p=0.002 n=6+6) /dataset.google_message4.pb/Marshal 93.4ms ± 4% 94.9ms ± 2% ~ (p=0.180 n=6+6) /dataset.google_message3_2.pb/Marshal 105ms ± 2% 98ms ± 7% -6.81% (p=0.009 n=5+6) /dataset.google_message3_4.pb/Marshal 16.3ms ± 6% 15.7ms ± 3% -3.44% (p=0.041 n=6+6) /dataset.google_message3_5.pb/Marshal 676ms ± 4% 504ms ± 2% -25.50% (p=0.004 n=6+5) Change-Id: I72cc4597117f4cf5d236ef505777d49dd4a5f75d Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/171020 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
2019-04-01 20:49:56 +00:00
mi.init()
if p.IsNil() {
return b, nil
}
var err error
var nerr errors.NonFatal
// The old marshaler encodes extensions at beginning.
if mi.extensionOffset.IsValid() {
e := p.Apply(mi.extensionOffset).Extensions()
// TODO: Special handling for MessageSet?
b, err = mi.appendExtensions(b, e, opts)
if !nerr.Merge(err) {
return b, err
}
}
for _, f := range mi.fieldsOrdered {
fptr := p.Apply(f.offset)
if f.isPointer && fptr.Elem().IsNil() {
continue
}
if f.funcs.marshal == nil {
continue
}
b, err = f.funcs.marshal(b, fptr, f.wiretag, opts)
if !nerr.Merge(err) {
return b, err
}
}
if mi.unknownOffset.IsValid() {
u := *p.Apply(mi.unknownOffset).Bytes()
b = append(b, u...)
}
return b, nerr.E
}
func (mi *MessageInfo) sizeExtensions(ext *legacyExtensionMap, opts marshalOptions) (n int) {
internal/impl: add fast-path marshal implementation This is a port of the v1 table marshaler, with some substantial cleanup and refactoring. Benchstat results from the protobuf reference benchmark data comparing the v1 package with v2, with AllowPartial:true set for the new package. This is not an apples-to-apples comparison, since v1 doesn't have a way to disable required field checks. Required field checks in v2 package currently go through reflection, which performs terribly; my initial experimentation indicates that fast-path required field checks will not add a large amount of cost; these results are incomplete but not wholly inaccurate. name old time/op new time/op delta /dataset.google_message3_1.pb/Marshal-12 219ms ± 1% 232ms ± 1% +5.85% (p=0.004 n=6+5) /dataset.google_message2.pb/Marshal-12 261µs ± 3% 248µs ± 1% -5.14% (p=0.002 n=6+6) /dataset.google_message1_proto2.pb/Marshal-12 681ns ± 2% 637ns ± 3% -6.53% (p=0.002 n=6+6) /dataset.google_message1_proto3.pb/Marshal-12 1.10µs ± 8% 0.99µs ± 3% -9.63% (p=0.002 n=6+6) /dataset.google_message3_3.pb/Marshal-12 44.2ms ± 3% 35.2ms ± 1% -20.28% (p=0.004 n=6+5) /dataset.google_message4.pb/Marshal-12 91.4ms ± 2% 94.9ms ± 2% +3.78% (p=0.002 n=6+6) /dataset.google_message3_2.pb/Marshal-12 78.7ms ± 6% 80.8ms ± 4% ~ (p=0.310 n=6+6) /dataset.google_message3_4.pb/Marshal-12 10.6ms ± 3% 10.6ms ± 8% ~ (p=0.662 n=5+6) /dataset.google_message3_5.pb/Marshal-12 675ms ± 4% 510ms ± 2% -24.40% (p=0.002 n=6+6) /dataset.google_message3_1.pb/Marshal 219ms ± 1% 236ms ± 7% +8.06% (p=0.004 n=5+6) /dataset.google_message2.pb/Marshal 257µs ± 1% 250µs ± 3% ~ (p=0.052 n=5+6) /dataset.google_message1_proto2.pb/Marshal 685ns ± 1% 628ns ± 1% -8.41% (p=0.008 n=5+5) /dataset.google_message1_proto3.pb/Marshal 1.08µs ± 1% 0.98µs ± 2% -9.31% (p=0.004 n=5+6) /dataset.google_message3_3.pb/Marshal 43.7ms ± 1% 35.1ms ± 1% -19.76% (p=0.002 n=6+6) /dataset.google_message4.pb/Marshal 93.4ms ± 4% 94.9ms ± 2% ~ (p=0.180 n=6+6) /dataset.google_message3_2.pb/Marshal 105ms ± 2% 98ms ± 7% -6.81% (p=0.009 n=5+6) /dataset.google_message3_4.pb/Marshal 16.3ms ± 6% 15.7ms ± 3% -3.44% (p=0.041 n=6+6) /dataset.google_message3_5.pb/Marshal 676ms ± 4% 504ms ± 2% -25.50% (p=0.004 n=6+5) Change-Id: I72cc4597117f4cf5d236ef505777d49dd4a5f75d Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/171020 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
2019-04-01 20:49:56 +00:00
if ext == nil {
return 0
}
for _, e := range *ext {
ei := mi.extensionFieldInfo(e.Desc)
if ei.funcs.size == nil {
continue
}
n += ei.funcs.size(e.value, ei.tagsize, opts)
}
return n
}
func (mi *MessageInfo) appendExtensions(b []byte, ext *legacyExtensionMap, opts marshalOptions) ([]byte, error) {
internal/impl: add fast-path marshal implementation This is a port of the v1 table marshaler, with some substantial cleanup and refactoring. Benchstat results from the protobuf reference benchmark data comparing the v1 package with v2, with AllowPartial:true set for the new package. This is not an apples-to-apples comparison, since v1 doesn't have a way to disable required field checks. Required field checks in v2 package currently go through reflection, which performs terribly; my initial experimentation indicates that fast-path required field checks will not add a large amount of cost; these results are incomplete but not wholly inaccurate. name old time/op new time/op delta /dataset.google_message3_1.pb/Marshal-12 219ms ± 1% 232ms ± 1% +5.85% (p=0.004 n=6+5) /dataset.google_message2.pb/Marshal-12 261µs ± 3% 248µs ± 1% -5.14% (p=0.002 n=6+6) /dataset.google_message1_proto2.pb/Marshal-12 681ns ± 2% 637ns ± 3% -6.53% (p=0.002 n=6+6) /dataset.google_message1_proto3.pb/Marshal-12 1.10µs ± 8% 0.99µs ± 3% -9.63% (p=0.002 n=6+6) /dataset.google_message3_3.pb/Marshal-12 44.2ms ± 3% 35.2ms ± 1% -20.28% (p=0.004 n=6+5) /dataset.google_message4.pb/Marshal-12 91.4ms ± 2% 94.9ms ± 2% +3.78% (p=0.002 n=6+6) /dataset.google_message3_2.pb/Marshal-12 78.7ms ± 6% 80.8ms ± 4% ~ (p=0.310 n=6+6) /dataset.google_message3_4.pb/Marshal-12 10.6ms ± 3% 10.6ms ± 8% ~ (p=0.662 n=5+6) /dataset.google_message3_5.pb/Marshal-12 675ms ± 4% 510ms ± 2% -24.40% (p=0.002 n=6+6) /dataset.google_message3_1.pb/Marshal 219ms ± 1% 236ms ± 7% +8.06% (p=0.004 n=5+6) /dataset.google_message2.pb/Marshal 257µs ± 1% 250µs ± 3% ~ (p=0.052 n=5+6) /dataset.google_message1_proto2.pb/Marshal 685ns ± 1% 628ns ± 1% -8.41% (p=0.008 n=5+5) /dataset.google_message1_proto3.pb/Marshal 1.08µs ± 1% 0.98µs ± 2% -9.31% (p=0.004 n=5+6) /dataset.google_message3_3.pb/Marshal 43.7ms ± 1% 35.1ms ± 1% -19.76% (p=0.002 n=6+6) /dataset.google_message4.pb/Marshal 93.4ms ± 4% 94.9ms ± 2% ~ (p=0.180 n=6+6) /dataset.google_message3_2.pb/Marshal 105ms ± 2% 98ms ± 7% -6.81% (p=0.009 n=5+6) /dataset.google_message3_4.pb/Marshal 16.3ms ± 6% 15.7ms ± 3% -3.44% (p=0.041 n=6+6) /dataset.google_message3_5.pb/Marshal 676ms ± 4% 504ms ± 2% -25.50% (p=0.004 n=6+5) Change-Id: I72cc4597117f4cf5d236ef505777d49dd4a5f75d Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/171020 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
2019-04-01 20:49:56 +00:00
if ext == nil {
return b, nil
}
switch len(*ext) {
case 0:
return b, nil
case 1:
// Fast-path for one extension: Don't bother sorting the keys.
var err error
for _, e := range *ext {
ei := mi.extensionFieldInfo(e.Desc)
b, err = ei.funcs.marshal(b, e.value, ei.wiretag, opts)
}
return b, err
default:
// Sort the keys to provide a deterministic encoding.
// Not sure this is required, but the old code does it.
keys := make([]int, 0, len(*ext))
for k := range *ext {
keys = append(keys, int(k))
}
sort.Ints(keys)
var err error
var nerr errors.NonFatal
for _, k := range keys {
e := (*ext)[int32(k)]
ei := mi.extensionFieldInfo(e.Desc)
b, err = ei.funcs.marshal(b, e.value, ei.wiretag, opts)
if !nerr.Merge(err) {
return b, err
}
}
return b, nerr.E
}
}