Files
gomobile-java/bind/genobjc.go

1014 lines
25 KiB
Go
Raw Normal View History

// Copyright 2015 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 bind
import (
"fmt"
"go/constant"
"go/types"
"math"
"strings"
"unicode"
"unicode/utf8"
)
// TODO(hyangah): handle method name conflicts.
// - struct with SetF method and exported F field.
// - method names conflicting with NSObject methods. e.g. Init
// - interface type with InitWithRef.
// TODO(hyangah): error code/domain propagation
type objcGen struct {
prefix string // prefix arg passed by flag.
// fields set by init.
namePrefix string
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
*generator
}
type interfaceInfo struct {
obj *types.TypeName
t *types.Interface
summary ifaceSummary
}
type structInfo struct {
obj *types.TypeName
t *types.Struct
}
func (g *objcGen) init() {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.generator.init()
g.namePrefix = g.prefix + strings.Title(g.pkgName)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
func (g *objcGen) genGoH() error {
g.Printf(objcPreamble, g.pkg.Path(), g.gobindOpts(), g.pkg.Path())
g.Printf("#ifndef __%s_H__\n", g.pkgName)
g.Printf("#define __%s_H__\n\n", g.pkgName)
g.Printf("#include <stdint.h>\n")
g.Printf("#include <objc/objc.h>\n")
for _, i := range g.interfaces {
if !i.summary.implementable {
continue
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
for _, m := range i.summary.callable {
g.genInterfaceMethodSignature(m, i.obj.Name(), true)
g.Printf("\n")
}
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("#endif\n")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
if len(g.err) > 0 {
return g.err
}
return nil
}
func (g *objcGen) genH() error {
g.Printf(objcPreamble, g.pkg.Path(), g.gobindOpts(), g.pkg.Path())
g.Printf("#ifndef __%s%s_H__\n", g.prefix, strings.Title(g.pkgName))
g.Printf("#define __%s%s_H__\n", g.prefix, strings.Title(g.pkgName))
g.Printf("\n")
g.Printf("#include <Foundation/Foundation.h>\n")
g.Printf("\n")
// Forward declaration of @class and @protocol
for _, s := range g.structs {
g.Printf("@class %s%s;\n", g.namePrefix, s.obj.Name())
}
for _, i := range g.interfaces {
g.Printf("@protocol %s%s;\n", g.namePrefix, i.obj.Name())
if i.summary.implementable {
g.Printf("@class %s%s;\n", g.namePrefix, i.obj.Name())
// Forward declaration for other cases will be handled at the beginning of genM.
}
}
if len(g.structs) > 0 || len(g.interfaces) > 0 {
g.Printf("\n")
}
// @interfaces
for _, s := range g.structs {
g.genStructH(s.obj, s.t)
g.Printf("\n")
}
for _, i := range g.interfaces {
g.genInterfaceH(i.obj, i.t)
g.Printf("\n")
}
// const
// TODO: prefix with k?, or use a class method?
for _, obj := range g.constants {
switch b := obj.Type().(*types.Basic); b.Kind() {
case types.String, types.UntypedString:
g.Printf("FOUNDATION_EXPORT NSString* const %s%s;\n", g.namePrefix, obj.Name())
default:
g.Printf("FOUNDATION_EXPORT const %s %s%s;\n", g.objcType(obj.Type()), g.namePrefix, obj.Name())
}
}
if len(g.constants) > 0 {
g.Printf("\n")
}
// var
if len(g.vars) > 0 {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("@interface %s : NSObject\n", g.namePrefix)
for _, obj := range g.vars {
objcType := g.objcType(obj.Type())
g.Printf("+ (%s) %s;\n", objcType, lowerFirst(obj.Name()))
g.Printf("+ (void) set%s:(%s)v;\n", obj.Name(), objcType)
g.Printf("\n")
}
g.Printf("@end\n\n")
}
// static functions.
for _, obj := range g.funcs {
g.genFuncH(obj)
g.Printf("\n")
}
// declare all named types first.
g.Printf("#endif\n")
if len(g.err) > 0 {
return g.err
}
return nil
}
func (g *objcGen) gobindOpts() string {
opts := []string{"-lang=objc"}
if g.prefix != "Go" {
opts = append(opts, "-prefix="+g.prefix)
}
return strings.Join(opts, " ")
}
func (g *objcGen) genM() error {
g.Printf(objcPreamble, g.pkg.Path(), g.gobindOpts(), g.pkg.Path())
g.Printf("#include %q\n", g.namePrefix+".h")
g.Printf("#include <Foundation/Foundation.h>\n")
g.Printf("#include \"seq.h\"\n")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("#include \"_cgo_export.h\"\n")
g.Printf("\n")
g.Printf("static NSString* errDomain = @\"go.%s\";\n", g.pkg.Path())
g.Printf("\n")
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf(`@protocol goSeqRefInterface
-(GoSeqRef*) _ref;
@end
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
`)
// forward declarations skipped from genH
for _, i := range g.interfaces {
if i.summary.implementable {
g.Printf("@class %s%s;\n\n", g.namePrefix, i.obj.Name())
}
}
for _, i := range g.interfaces {
if i.summary.implementable {
// @interface Interface -- similar to what genStructH does.
g.genInterfaceInterface(i.obj, i.summary, true)
g.Printf("\n")
}
}
// struct
for _, s := range g.structs {
g.genStructM(s.obj, s.t)
g.Printf("\n")
}
// interface
var needProxy []*types.TypeName
for _, i := range g.interfaces {
if g.genInterfaceM(i.obj, i.t) {
needProxy = append(needProxy, i.obj)
}
g.Printf("\n")
}
// const
for _, o := range g.constants {
g.genConstM(o)
}
if len(g.constants) > 0 {
g.Printf("\n")
}
// vars
if len(g.vars) > 0 {
g.Printf("@implementation %s\n", g.namePrefix)
for _, o := range g.vars {
g.genVarM(o)
}
g.Printf("@end\n\n")
}
g.Printf("\n")
for _, obj := range g.funcs {
g.genFuncM(obj)
g.Printf("\n")
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
for _, i := range g.interfaces {
for _, m := range i.summary.callable {
g.genInterfaceMethodProxy(i.obj, m)
}
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("__attribute__((constructor)) static void init() {\n")
g.Indent()
g.Printf("init_seq();\n")
g.Outdent()
g.Printf("}\n")
if len(g.err) > 0 {
return g.err
}
return nil
}
func (g *objcGen) genVarM(o *types.Var) {
objcType := g.objcType(o.Type())
// setter
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("+ (void) set%s:(%s)v {\n", o.Name(), objcType)
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genWrite("v", o.Type(), modeRetained)
g.Printf("var_set%s_%s(_v);\n", g.pkgPrefix, o.Name())
g.genRelease("v", o.Type(), modeRetained)
g.Outdent()
g.Printf("}\n\n")
// getter
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("+ (%s) %s {\n", objcType, lowerFirst(o.Name()))
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("%s r0 = ", g.cgoType(o.Type()))
g.Printf("var_get%s_%s();\n", g.pkgPrefix, o.Name())
g.genRead("_r0", "r0", o.Type(), modeReturned)
g.Printf("return _r0;\n")
g.Outdent()
g.Printf("}\n\n")
}
func (g *objcGen) genConstM(o *types.Const) {
cName := fmt.Sprintf("%s%s", g.namePrefix, o.Name())
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
objcType := g.objcType(o.Type())
switch b := o.Type().(*types.Basic); b.Kind() {
case types.Bool, types.UntypedBool:
v := "NO"
if constant.BoolVal(o.Val()) {
v = "YES"
}
g.Printf("const BOOL %s = %s;\n", cName, v)
case types.String, types.UntypedString:
g.Printf("NSString* const %s = @%s;\n", cName, o.Val().ExactString())
case types.Int, types.Int8, types.Int16, types.Int32:
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("const %s %s = %s;\n", objcType, cName, o.Val())
case types.Int64, types.UntypedInt:
i, exact := constant.Int64Val(o.Val())
if !exact {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.errorf("const value %s for %s cannot be represented as %s", o.Val(), o.Name(), objcType)
return
}
if i == math.MinInt64 {
// -9223372036854775808LL does not work because 922337203685477508 is
// larger than max int64.
g.Printf("const int64_t %s = %dLL-1;\n", cName, i+1)
} else {
g.Printf("const int64_t %s = %dLL;\n", cName, i)
}
case types.Float32, types.Float64, types.UntypedFloat:
f, _ := constant.Float64Val(o.Val())
if math.IsInf(f, 0) || math.Abs(f) > math.MaxFloat64 {
g.errorf("const value %s for %s cannot be represented as double", o.Val(), o.Name())
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("const %s %s = %g;\n", objcType, cName, f)
default:
g.errorf("unsupported const type %s for %s", b, o.Name())
}
}
type funcSummary struct {
name string
ret string
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
sig *types.Signature
params, retParams []paramInfo
}
type paramInfo struct {
typ types.Type
name string
}
func (g *objcGen) funcSummary(obj *types.Func) *funcSummary {
sig := obj.Type().(*types.Signature)
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
s := &funcSummary{name: obj.Name(), sig: sig}
params := sig.Params()
for i := 0; i < params.Len(); i++ {
p := params.At(i)
v := paramInfo{
typ: p.Type(),
name: paramName(params, i),
}
s.params = append(s.params, v)
}
res := sig.Results()
switch res.Len() {
case 0:
s.ret = "void"
case 1:
p := res.At(0)
if isErrorType(p.Type()) {
s.retParams = append(s.retParams, paramInfo{
typ: p.Type(),
name: "error",
})
s.ret = "BOOL"
} else {
name := p.Name()
if name == "" || paramRE.MatchString(name) {
name = "ret0_"
}
typ := p.Type()
s.retParams = append(s.retParams, paramInfo{typ: typ, name: name})
s.ret = g.objcType(typ)
}
case 2:
name := res.At(0).Name()
if name == "" || paramRE.MatchString(name) {
name = "ret0_"
}
s.retParams = append(s.retParams, paramInfo{
typ: res.At(0).Type(),
name: name,
})
if !isErrorType(res.At(1).Type()) {
g.errorf("second result value must be of type error: %s", obj)
return nil
}
s.retParams = append(s.retParams, paramInfo{
typ: res.At(1).Type(),
name: "error", // TODO(hyangah): name collision check.
})
s.ret = "BOOL"
default:
// TODO(hyangah): relax the constraint on multiple return params.
g.errorf("too many result values: %s", obj)
return nil
}
return s
}
func (s *funcSummary) asFunc(g *objcGen) string {
var params []string
for _, p := range s.params {
params = append(params, g.objcType(p.typ)+" "+p.name)
}
if !s.returnsVal() {
for _, p := range s.retParams {
params = append(params, g.objcType(p.typ)+"* "+p.name)
}
}
return fmt.Sprintf("%s %s%s(%s)", s.ret, g.namePrefix, s.name, strings.Join(params, ", "))
}
func (s *funcSummary) asMethod(g *objcGen) string {
var params []string
for i, p := range s.params {
var key string
if i != 0 {
key = p.name
}
params = append(params, fmt.Sprintf("%s:(%s)%s", key, g.objcType(p.typ), p.name))
}
if !s.returnsVal() {
for _, p := range s.retParams {
var key string
if len(params) > 0 {
key = p.name
}
params = append(params, fmt.Sprintf("%s:(%s)%s", key, g.objcType(p.typ)+"*", p.name))
}
}
return fmt.Sprintf("(%s)%s%s", s.ret, lowerFirst(s.name), strings.Join(params, " "))
}
func (s *funcSummary) callMethod(g *objcGen) string {
var params []string
for i, p := range s.params {
var key string
if i != 0 {
key = p.name
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
params = append(params, fmt.Sprintf("%s:_%s", key, p.name))
}
if !s.returnsVal() {
for _, p := range s.retParams {
var key string
if len(params) > 0 {
key = p.name
}
params = append(params, fmt.Sprintf("%s:&%s", key, p.name))
}
}
return fmt.Sprintf("%s%s", lowerFirst(s.name), strings.Join(params, " "))
}
func (s *funcSummary) returnsVal() bool {
return len(s.retParams) == 1 && !isErrorType(s.retParams[0].typ)
}
func (g *objcGen) genFuncH(obj *types.Func) {
if s := g.funcSummary(obj); s != nil {
g.Printf("FOUNDATION_EXPORT %s;\n", s.asFunc(g))
}
}
func (g *objcGen) genFuncM(obj *types.Func) {
s := g.funcSummary(obj)
if s == nil {
return
}
g.Printf("%s {\n", s.asFunc(g))
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genFunc(s, "")
g.Outdent()
g.Printf("}\n")
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *objcGen) genGetter(oName string, f *types.Var) {
t := f.Type()
if isErrorType(t) {
t = types.Typ[types.String]
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("- (%s)%s {\n", g.objcType(t), lowerFirst(f.Name()))
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("int32_t refnum = go_seq_go_to_refnum(self._ref);\n")
g.Printf("%s r0 = ", g.cgoType(f.Type()))
g.Printf("proxy%s_%s_%s_Get(refnum);\n", g.pkgPrefix, oName, f.Name())
g.genRead("_r0", "r0", f.Type(), modeReturned)
g.Printf("return _r0;\n")
g.Outdent()
g.Printf("}\n\n")
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *objcGen) genSetter(oName string, f *types.Var) {
t := f.Type()
if isErrorType(t) {
t = types.Typ[types.String]
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("- (void)set%s:(%s)v {\n", f.Name(), g.objcType(t))
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("int32_t refnum = go_seq_go_to_refnum(self._ref);\n")
g.genWrite("v", f.Type(), modeRetained)
g.Printf("proxy%s_%s_%s_Set(refnum, _v);\n", g.pkgPrefix, oName, f.Name())
g.genRelease("v", f.Type(), modeRetained)
g.Outdent()
g.Printf("}\n\n")
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *objcGen) genWrite(varName string, t types.Type, mode varMode) {
if isErrorType(t) {
g.genWrite(varName, types.Typ[types.String], mode)
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
switch t := t.(type) {
case *types.Basic:
switch t.Kind() {
case types.String:
g.Printf("nstring _%s = go_seq_from_objc_string(%s);\n", varName, varName)
default:
g.Printf("%s _%s = (%s)%s;\n", g.cgoType(t), varName, g.cgoType(t), varName)
}
case *types.Slice:
switch e := t.Elem().(type) {
case *types.Basic:
switch e.Kind() {
case types.Uint8: // Byte.
g.Printf("nbyteslice _%s = go_seq_from_objc_bytearray(%s, %d);\n", varName, varName, g.toCFlag(mode.copySlice()))
default:
g.errorf("unsupported type: %s", t)
}
default:
g.errorf("unsupported type: %s", t)
}
case *types.Named:
switch u := t.Underlying().(type) {
case *types.Interface:
g.genRefWrite(varName, t)
default:
panic(fmt.Sprintf("unsupported named type: %s / %T", u, u))
}
case *types.Pointer:
g.genRefWrite(varName, t)
default:
g.Printf("%s _%s = (%s)%s;\n", g.cgoType(t), varName, g.cgoType(t), varName)
}
}
func (g *objcGen) genRefWrite(varName string, t types.Type) {
g.Printf("int32_t _%s;\n", varName)
g.Printf("if ([(id<NSObject>)(%s) isKindOfClass:[%s class]]) {\n", varName, g.refTypeBase(t))
g.Indent()
g.Printf("id<goSeqRefInterface> %[1]s_proxy = (id<goSeqRefInterface>)(%[1]s);\n", varName)
g.Printf("_%s = go_seq_go_to_refnum(%s_proxy._ref);\n", varName, varName)
g.Outdent()
g.Printf("} else {\n")
g.Indent()
g.Printf("_%s = go_seq_to_refnum(%s);\n", varName, varName)
g.Outdent()
g.Printf("}\n")
}
func (g *objcGen) genRefRead(toName, fromName string, t types.Type) {
ptype := g.objcType(t)
g.Printf("%s %s = nil;\n", ptype, toName)
g.Printf("GoSeqRef* %s_ref = go_seq_from_refnum(%s);\n", toName, fromName)
g.Printf("if (%s_ref != NULL) {\n", toName)
g.Printf(" %s = %s_ref.obj;\n", toName, toName)
g.Printf(" if (%s == nil) {\n", toName)
g.Printf(" %s = [[%s alloc] initWithRef:%s_ref];\n", toName, g.refTypeBase(t), toName)
g.Printf(" }\n")
g.Printf("}\n")
}
func (g *objcGen) genRead(toName, fromName string, t types.Type, mode varMode) {
if isErrorType(t) {
g.genRead(toName, fromName, types.Typ[types.String], mode)
return
}
switch t := t.(type) {
case *types.Basic:
switch t.Kind() {
case types.String:
g.Printf("NSString *%s = go_seq_to_objc_string(%s);\n", toName, fromName)
case types.Bool:
g.Printf("BOOL %s = %s ? YES : NO;\n", toName, fromName)
default:
g.Printf("%s %s = (%s)%s;\n", g.objcType(t), toName, g.objcType(t), fromName)
}
case *types.Slice:
switch e := t.Elem().(type) {
case *types.Basic:
switch e.Kind() {
case types.Uint8: // Byte.
g.Printf("NSData *%s = go_seq_to_objc_bytearray(%s, %d);\n", toName, fromName, g.toCFlag(mode.copySlice()))
default:
g.errorf("unsupported type: %s", t)
}
default:
g.errorf("unsupported type: %s", t)
}
case *types.Pointer:
switch t := t.Elem().(type) {
case *types.Named:
g.genRefRead(toName, fromName, types.NewPointer(t))
default:
g.errorf("unsupported type %s", t)
}
case *types.Named:
switch t.Underlying().(type) {
case *types.Interface, *types.Pointer:
g.genRefRead(toName, fromName, t)
default:
g.errorf("unsupported, direct named type %s", t)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
default:
g.Printf("%s %s = (%s)%s;\n", g.objcType(t), toName, g.objcType(t), fromName)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *objcGen) genFunc(s *funcSummary, objName string) {
if objName != "" {
g.Printf("int32_t refnum = go_seq_go_to_refnum(self._ref);\n")
}
for _, p := range s.params {
g.genWrite(p.name, p.typ, modeTransient)
}
resPrefix := ""
if len(s.retParams) > 0 {
if len(s.retParams) == 1 {
g.Printf("%s r0 = ", g.cgoType(s.retParams[0].typ))
} else {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
resPrefix = "res."
g.Printf("struct proxy%s_%s_%s_return res = ", g.pkgPrefix, objName, s.name)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
g.Printf("proxy%s_%s_%s(", g.pkgPrefix, objName, s.name)
if objName != "" {
g.Printf("refnum")
}
for i, p := range s.params {
if i > 0 || objName != "" {
g.Printf(", ")
}
g.Printf("_%s", p.name)
}
g.Printf(");\n")
for _, p := range s.params {
g.genRelease(p.name, p.typ, modeTransient)
}
for i, r := range s.retParams {
g.genRead("_"+r.name, fmt.Sprintf("%sr%d", resPrefix, i), r.typ, modeReturned)
}
if !s.returnsVal() {
for _, p := range s.retParams {
if isErrorType(p.typ) {
g.Printf("if ([_%s length] != 0 && %s != nil) {\n", p.name, p.name)
g.Indent()
g.Printf("NSMutableDictionary* details = [NSMutableDictionary dictionary];\n")
g.Printf("[details setValue:_%s forKey:NSLocalizedDescriptionKey];\n", p.name)
g.Printf("*%s = [NSError errorWithDomain:errDomain code:1 userInfo:details];\n", p.name)
g.Outdent()
g.Printf("}\n")
} else {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("*%s = _%s;\n", p.name, p.name)
}
}
}
if n := len(s.retParams); n > 0 {
p := s.retParams[n-1]
if isErrorType(p.typ) {
g.Printf("return ([_%s length] == 0);\n", p.name)
} else {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("return _%s;\n", p.name)
}
}
}
func (g *objcGen) genInterfaceInterface(obj *types.TypeName, summary ifaceSummary, isProtocol bool) {
g.Printf("@interface %[1]s%[2]s : NSObject", g.namePrefix, obj.Name())
if isProtocol {
g.Printf(" <%[1]s%[2]s>", g.namePrefix, obj.Name())
}
g.Printf(" {\n}\n")
g.Printf("@property(strong, readonly) id _ref;\n")
g.Printf("\n")
g.Printf("- (id)initWithRef:(id)ref;\n")
for _, m := range summary.callable {
s := g.funcSummary(m)
g.Printf("- %s;\n", s.asMethod(g))
}
g.Printf("@end\n")
}
func (g *objcGen) genInterfaceH(obj *types.TypeName, t *types.Interface) {
summary := makeIfaceSummary(t)
if !summary.implementable {
g.genInterfaceInterface(obj, summary, false)
return
}
g.Printf("@protocol %s%s\n", g.namePrefix, obj.Name())
for _, m := range makeIfaceSummary(t).callable {
s := g.funcSummary(m)
g.Printf("- %s;\n", s.asMethod(g))
}
g.Printf("@end\n")
}
func (g *objcGen) genInterfaceM(obj *types.TypeName, t *types.Interface) bool {
summary := makeIfaceSummary(t)
// @implementation Interface -- similar to what genStructM does.
g.Printf("@implementation %s%s {\n", g.namePrefix, obj.Name())
g.Printf("}\n")
g.Printf("\n")
g.Printf("- (id)initWithRef:(id)ref {\n")
g.Indent()
g.Printf("self = [super init];\n")
g.Printf("if (self) { __ref = ref; }\n")
g.Printf("return self;\n")
g.Outdent()
g.Printf("}\n")
g.Printf("\n")
for _, m := range summary.callable {
s := g.funcSummary(m)
g.Printf("- %s {\n", s.asMethod(g))
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genFunc(s, obj.Name())
g.Outdent()
g.Printf("}\n\n")
}
g.Printf("@end\n")
g.Printf("\n")
return summary.implementable
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
func (g *objcGen) genInterfaceMethodProxy(obj *types.TypeName, m *types.Func) {
oName := obj.Name()
s := g.funcSummary(m)
g.genInterfaceMethodSignature(m, oName, false)
g.Indent()
g.Printf("%s o = go_seq_objc_from_refnum(refnum);\n", g.objcType(obj.Type()))
for _, p := range s.params {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genRead("_"+p.name, p.name, p.typ, modeTransient)
}
// call method
if !s.returnsVal() {
for _, p := range s.retParams {
if isErrorType(p.typ) {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.Printf("NSError* %s = nil;\n", p.name)
} else {
g.Printf("%s %s;\n", g.objcType(p.typ), p.name)
}
}
}
if s.ret == "void" {
g.Printf("[o %s];\n", s.callMethod(g))
} else {
g.Printf("%s returnVal = [o %s];\n", s.ret, s.callMethod(g))
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
if len(s.retParams) > 0 {
if s.returnsVal() { // len(s.retParams) == 1 && s.retParams[0] != error
p := s.retParams[0]
g.genWrite("returnVal", p.typ, modeReturned)
g.Printf("return _returnVal;\n")
} else {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
var rets []string
for i, p := range s.retParams {
if isErrorType(p.typ) {
g.Printf("NSString *%s_str = nil;\n", p.name)
if i == len(s.retParams)-1 { // last param.
g.Printf("if (!returnVal) {\n")
} else {
g.Printf("if (%s != nil) {\n", p.name)
}
g.Indent()
g.Printf("%[1]s_str = [%[1]s localizedDescription];\n", p.name)
g.Printf("if (%[1]s_str == nil || %[1]s_str.length == 0) {\n", p.name)
g.Indent()
g.Printf("%[1]s_str = @\"gobind: unknown error\";\n", p.name)
g.Outdent()
g.Printf("}\n")
g.Outdent()
g.Printf("}\n")
g.genWrite(p.name+"_str", p.typ, modeReturned)
rets = append(rets, fmt.Sprintf("_%s_str", p.name))
} else {
g.genWrite(p.name, p.typ, modeReturned)
rets = append(rets, "_"+p.name)
}
}
if len(rets) > 1 {
g.Printf("cproxy%s_%s_%s_return _sres = {\n", g.pkgPrefix, oName, m.Name())
g.Printf(" %s\n", strings.Join(rets, ", "))
g.Printf("};\n")
g.Printf("return _sres;\n")
} else {
g.Printf("return %s;\n", rets[0])
}
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
}
g.Outdent()
g.Printf("}\n\n")
}
// genRelease cleans up arguments that weren't copied in genWrite.
func (g *objcGen) genRelease(varName string, t types.Type, mode varMode) {
if isErrorType(t) {
g.genRelease(varName, types.Typ[types.String], mode)
return
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
switch t := t.(type) {
case *types.Slice:
switch e := t.Elem().(type) {
case *types.Basic:
switch e.Kind() {
case types.Uint8: // Byte.
if !mode.copySlice() {
// If the argument was not mutable, go_seq_from_objc_bytearray created a copy.
// Free it here.
g.Printf("if (![%s isKindOfClass:[NSMutableData class]]) {\n", varName)
g.Printf(" free(_%s.ptr);\n", varName)
g.Printf("}\n")
}
}
}
}
}
func (g *objcGen) genStructH(obj *types.TypeName, t *types.Struct) {
g.Printf("@interface %s%s : NSObject {\n", g.namePrefix, obj.Name())
g.Printf("}\n")
g.Printf("@property(strong, readonly) id _ref;\n")
g.Printf("\n")
g.Printf("- (id)initWithRef:(id)ref;\n")
// accessors to exported fields.
for _, f := range exportedFields(t) {
name, typ := f.Name(), g.objcFieldType(f.Type())
g.Printf("- (%s)%s;\n", typ, lowerFirst(name))
g.Printf("- (void)set%s:(%s)v;\n", name, typ)
}
// exported methods
for _, m := range exportedMethodSet(types.NewPointer(obj.Type())) {
s := g.funcSummary(m)
g.Printf("- %s;\n", lowerFirst(s.asMethod(g)))
}
g.Printf("@end\n")
}
func (g *objcGen) genStructM(obj *types.TypeName, t *types.Struct) {
fields := exportedFields(t)
methods := exportedMethodSet(types.NewPointer(obj.Type()))
g.Printf("\n")
g.Printf("@implementation %s%s {\n", g.namePrefix, obj.Name())
g.Printf("}\n\n")
g.Printf("- (id)initWithRef:(id)ref {\n")
g.Indent()
g.Printf("self = [super init];\n")
g.Printf("if (self) { __ref = ref; }\n")
g.Printf("return self;\n")
g.Outdent()
g.Printf("}\n\n")
for _, f := range fields {
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genGetter(obj.Name(), f)
g.genSetter(obj.Name(), f)
}
for _, m := range methods {
s := g.funcSummary(m)
g.Printf("- %s {\n", s.asMethod(g))
g.Indent()
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
g.genFunc(s, obj.Name())
g.Outdent()
g.Printf("}\n\n")
}
g.Printf("@end\n")
}
func (g *objcGen) errorf(format string, args ...interface{}) {
g.err = append(g.err, fmt.Errorf(format, args...))
}
func (g *objcGen) refTypeBase(typ types.Type) string {
switch typ := typ.(type) {
case *types.Pointer:
if _, ok := typ.Elem().(*types.Named); ok {
return g.objcType(typ.Elem())
}
case *types.Named:
n := typ.Obj()
if n.Pkg() == g.pkg {
switch typ.Underlying().(type) {
case *types.Interface, *types.Struct:
return g.namePrefix + n.Name()
}
}
}
// fallback to whatever objcType returns. This must not happen.
return g.objcType(typ)
}
func (g *objcGen) objcFieldType(t types.Type) string {
if isErrorType(t) {
return "NSString*"
}
return g.objcType(t)
}
func (g *objcGen) objcType(typ types.Type) string {
if isErrorType(typ) {
return "NSError*"
}
switch typ := typ.(type) {
case *types.Basic:
switch typ.Kind() {
case types.Bool, types.UntypedBool:
return "BOOL"
case types.Int:
return "int"
case types.Int8:
return "int8_t"
case types.Int16:
return "int16_t"
case types.Int32, types.UntypedRune: // types.Rune
return "int32_t"
case types.Int64, types.UntypedInt:
return "int64_t"
case types.Uint8:
// byte is an alias of uint8, and the alias is lost.
return "byte"
case types.Uint16:
return "uint16_t"
case types.Uint32:
return "uint32_t"
case types.Uint64:
return "uint64_t"
case types.Float32:
return "float"
case types.Float64, types.UntypedFloat:
return "double"
case types.String, types.UntypedString:
return "NSString*"
default:
g.errorf("unsupported type: %s", typ)
return "TODO"
}
case *types.Slice:
elem := g.objcType(typ.Elem())
// Special case: NSData seems to be a better option for byte slice.
if elem == "byte" {
return "NSData*"
}
// TODO(hyangah): support other slice types: NSArray or CFArrayRef.
// Investigate the performance implication.
g.errorf("unsupported type: %s", typ)
return "TODO"
case *types.Pointer:
if _, ok := typ.Elem().(*types.Named); ok {
return g.objcType(typ.Elem()) + "*"
}
g.errorf("unsupported pointer to type: %s", typ)
return "TODO"
case *types.Named:
n := typ.Obj()
if n.Pkg() != g.pkg {
g.errorf("type %s is in package %s; only types defined in package %s is supported", n.Name(), n.Pkg().Name(), g.pkg.Name())
return "TODO"
}
switch t := typ.Underlying().(type) {
case *types.Interface:
if makeIfaceSummary(t).implementable {
return "id<" + g.namePrefix + n.Name() + ">"
} else {
return g.namePrefix + n.Name() + "*"
}
case *types.Struct:
return g.namePrefix + n.Name()
}
g.errorf("unsupported, named type %s", typ)
return "TODO"
default:
g.errorf("unsupported type: %#+v, %s", typ, typ)
return "TODO"
}
}
func lowerFirst(s string) string {
if s == "" {
return ""
}
var conv []rune
for len(s) > 0 {
r, n := utf8.DecodeRuneInString(s)
if !unicode.IsUpper(r) {
if l := len(conv); l > 1 {
conv[l-1] = unicode.ToUpper(conv[l-1])
}
return string(conv) + s
}
conv = append(conv, unicode.ToLower(r))
s = s[n:]
}
return string(conv)
}
mobile/bind: replace seq serialization with direct calls The seq serialization machinery is a historic artifact from when Go mobile code had to run in a separate process. Now that Go code is running in-process, replace the explicit serialization with direct calls and pass arguments on the stack. The benefits are a much smaller bind runtime, much less garbage (and, in Java, fewer objects with finalizers), less argument copying, and faster cross-language calls. The cost is a more complex generator, because some of the work from the bind runtime is moved to generated code. Generated code now handles conversion between Go and Java/ObjC types, multiple return values and memory management of byte slice and string arguments. To overcome the lack of calling C code between Go packages, all bound packages now end up in the same (fake) package, "gomobile_bind", instead of separate packages (go_<pkgname>). To avoid name clashes, the package name is added as a prefix to generated functions and types. Also, don't copy byte arrays passed to Go, saving call time and allowing read([]byte)-style interfaces to foreign callers (#12113). Finally, add support for nil interfaces and struct pointers to objc. This is a large CL, but most of the changes stem from changing testdata. The full benchcmp output on the CL/20095 benchmarks on my Nexus 5 is reproduced below. Note that the savings for the JavaSlice* benchmarks are skewed because byte slices are no longer copied before passing them to Go. benchmark old ns/op new ns/op delta BenchmarkJavaEmpty 26.0 19.0 -26.92% BenchmarkJavaEmptyDirect 23.0 22.0 -4.35% BenchmarkJavaNoargs 7685 2339 -69.56% BenchmarkJavaNoargsDirect 17405 8041 -53.80% BenchmarkJavaOnearg 26887 2366 -91.20% BenchmarkJavaOneargDirect 34266 7910 -76.92% BenchmarkJavaOneret 38325 2245 -94.14% BenchmarkJavaOneretDirect 46265 7708 -83.34% BenchmarkJavaManyargs 41720 2535 -93.92% BenchmarkJavaManyargsDirect 51026 8373 -83.59% BenchmarkJavaRefjava 38139 21260 -44.26% BenchmarkJavaRefjavaDirect 42706 28150 -34.08% BenchmarkJavaRefgo 34403 6843 -80.11% BenchmarkJavaRefgoDirect 40193 16582 -58.74% BenchmarkJavaStringShort 32366 9323 -71.20% BenchmarkJavaStringShortDirect 41973 19118 -54.45% BenchmarkJavaStringLong 127879 94420 -26.16% BenchmarkJavaStringLongDirect 133776 114760 -14.21% BenchmarkJavaStringShortUnicode 32562 9221 -71.68% BenchmarkJavaStringShortUnicodeDirect 41464 19094 -53.95% BenchmarkJavaStringLongUnicode 131015 89401 -31.76% BenchmarkJavaStringLongUnicodeDirect 134130 90786 -32.31% BenchmarkJavaSliceShort 42462 7538 -82.25% BenchmarkJavaSliceShortDirect 52940 17017 -67.86% BenchmarkJavaSliceLong 138391 8466 -93.88% BenchmarkJavaSliceLongDirect 205804 15666 -92.39% BenchmarkGoEmpty 3.00 3.00 +0.00% BenchmarkGoEmptyDirect 3.00 3.00 +0.00% BenchmarkGoNoarg 40342 13716 -66.00% BenchmarkGoNoargDirect 46691 13569 -70.94% BenchmarkGoOnearg 43529 13757 -68.40% BenchmarkGoOneargDirect 44867 14078 -68.62% BenchmarkGoOneret 45456 13559 -70.17% BenchmarkGoOneretDirect 44694 13442 -69.92% BenchmarkGoRefjava 55111 28071 -49.06% BenchmarkGoRefjavaDirect 60883 26872 -55.86% BenchmarkGoRefgo 57038 29223 -48.77% BenchmarkGoRefgoDirect 56153 27812 -50.47% BenchmarkGoManyargs 67967 17398 -74.40% BenchmarkGoManyargsDirect 60617 16998 -71.96% BenchmarkGoStringShort 57538 22600 -60.72% BenchmarkGoStringShortDirect 52627 22704 -56.86% BenchmarkGoStringLong 128485 52530 -59.12% BenchmarkGoStringLongDirect 138377 52079 -62.36% BenchmarkGoStringShortUnicode 57062 22994 -59.70% BenchmarkGoStringShortUnicodeDirect 62563 22938 -63.34% BenchmarkGoStringLongUnicode 139913 55553 -60.29% BenchmarkGoStringLongUnicodeDirect 150863 57791 -61.69% BenchmarkGoSliceShort 59279 20215 -65.90% BenchmarkGoSliceShortDirect 60160 21136 -64.87% BenchmarkGoSliceLong 411225 301870 -26.59% BenchmarkGoSliceLongDirect 399029 298915 -25.09% Fixes golang/go#12619 Fixes golang/go#12113 Fixes golang/go#13033 Change-Id: I2b45e9e98a1248e3c23a5137f775f7364908bec7 Reviewed-on: https://go-review.googlesource.com/19821 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
2016-02-12 18:50:33 +01:00
const (
objcPreamble = `// Objective-C API for talking to %[1]s Go package.
// gobind %[2]s %[3]s
//
// File is generated by gobind. Do not edit.
`
)