Nuke old singlepass compiler, enable optimizing compiler by default (#…

…2130)

Signed-off-by: Takeshi Yoneda <[email protected]>

.github/workflows/integration.yaml

@@ -78,7 +78,7 @@ jobs:
 
   zig:
     needs: build_zig_test_binary
-    name: Zig (${{ matrix.os.name }}, ${{ matrix.arch }}, ${{ matrix.compiler }})
+    name: Zig (${{ matrix.os.name }}, ${{ matrix.arch }})
     runs-on: ${{ matrix.os.version }}
     strategy:
       fail-fast: false # don't fail fast as sometimes failures are arch/OS specific
@@ -92,18 +92,15 @@ jobs:
             name: macOS
           - version: windows-2022
             name: Windows
-        compiler: [baseline]
         arch:     [amd64]
         include:
           - os:
               version: macos-14
               name: macOS
-            compiler: optimizing
             arch:     "arm64"
           - os:
               version: ubuntu-22.04
               name: Ubuntu
-            compiler:   optimizing
             arch:       "amd64"
             go-version: "1.21"
 
@@ -129,7 +126,7 @@ jobs:
       - name: Run built test binaries
         run: |
           cd ${{ env.STDLIB_TESTS }}
-          go test -bench='BenchmarkZig/${{ matrix.compiler }}'
+          go test -bench='BenchmarkZig' -timeout=20m
 
   build_tinygo_test_binary:
     name: Build TinyGo test binary
@@ -168,7 +165,7 @@ jobs:
 
   tinygo:
     needs: build_tinygo_test_binary
-    name: TinyGo (${{ matrix.os.name }}, ${{ matrix.arch }}, ${{ matrix.compiler }})
+    name: TinyGo (${{ matrix.os.name }}, ${{ matrix.arch }})
     runs-on: ${{ matrix.os.version }}
     strategy:
       fail-fast: false # don't fail fast as sometimes failures are arch/OS specific
@@ -182,18 +179,15 @@ jobs:
             name: macOS
           - version: windows-2022
             name: Windows
-        compiler: [baseline]
         arch:     [amd64]
         include:
           - os:
               version: macos-14
               name: macOS
-            compiler: optimizing
             arch:     "arm64"
           - os:
               version: ubuntu-22.04
               name: Ubuntu
-            compiler:   optimizing
             arch:       "amd64"
             go-version: "1.21"
 
@@ -219,7 +213,7 @@ jobs:
       - name: Run test binaries
         run: |
           cd ${{ env.STDLIB_TESTS }}
-          go test -bench='BenchmarkTinyGo/${{ matrix.compiler }}'
+          go test -bench='BenchmarkTinyGo' -timeout=20m
 
   wasi-testsuite:
     name: wasi-testsuite
@@ -280,7 +274,7 @@ jobs:
   go_tests:
     # Due to the embedding of the GOROOT of the building env(https://github.com/golang/go/blob/3c59639b902fada0a2e5a6a35bafd10fc9183b89/src/os/os_test.go#L112),
     # we have to build and cache on each OS unlike others in this file.
-    name: Go-${{ matrix.go-version }} (${{ matrix.os.name }}, ${{ matrix.arch }}, ${{ matrix.compiler }})
+    name: Go-${{ matrix.go-version }} (${{ matrix.os.name }}, ${{ matrix.arch }})
     runs-on: ${{ matrix.os.version }}
     strategy:
       fail-fast: false # don't fail fast as sometimes failures are arch/OS specific
@@ -294,7 +288,6 @@ jobs:
           name: macOS
         - version: windows-2022
           name: Windows
-        compiler: [baseline]
         arch:     [amd64]
         go-version:
           - "1.21"
@@ -303,25 +296,21 @@ jobs:
           - os:
               version: macos-14
               name: macOS
-            compiler:   optimizing
             arch:       "arm64"
             go-version: "1.21"
           - os:
               version: macos-14
               name: macOS
-            compiler:   optimizing
             arch:       "arm64"
             go-version: "1.22"
           - os:
               version: ubuntu-22.04
               name: Ubuntu
-            compiler:   optimizing
             arch:       "amd64"
             go-version: "1.21"
           - os:
               version: ubuntu-22.04
               name: Ubuntu
-            compiler:   optimizing
             arch:       "amd64"
             go-version: "1.22"
 
@@ -358,10 +347,10 @@ jobs:
       - name: Run built test binaries
         run: |
           cd ${{ env.STDLIB_TESTS }}
-          go test -bench='BenchmarkWasip1/${{ matrix.compiler }}'
+          go test -bench='BenchmarkWasip1' -timeout=20m
 
   libsodium:
-    name: libsodium (${{ matrix.os.name }}, ${{ matrix.os.arch }}, optimizing)
+    name: libsodium (${{ matrix.os.name }}, ${{ matrix.os.arch }})
     runs-on: ${{ matrix.os.version }}
     strategy:
       fail-fast: false # don't fail fast as sometimes failures are arch/OS specific

Makefile

@@ -23,7 +23,6 @@ go_test_options ?= -timeout 300s
 ensureCompilerFastest := -ldflags '-X github.com/tetratelabs/wazero/internal/integration_test/vs.ensureCompilerFastest=true'
 .PHONY: bench
 bench:
-	@go test -run=NONE -benchmem -bench=. ./internal/engine/compiler/...
 	@go build ./internal/integration_test/bench/...
 	@# Don't use -test.benchmem as it isn't accurate when comparing against CGO libs
 	@for d in vs/time vs/wasmedge vs/wasmtime ; do \

RATIONALE.md

@@ -1536,7 +1536,38 @@ If a module reaches this limit, an error is returned at the compilation phase.
 
 ## Compiler engine implementation
 
-See [compiler/RATIONALE.md](internal/engine/compiler/RATIONALE.md).
+### Why it's safe to execute runtime-generated machine codes against async Goroutine preemption
+
+Goroutine preemption is the mechanism of the Go runtime to switch goroutines contexts on an OS thread.
+There are two types of preemption: cooperative preemption and async preemption. The former happens, for example,
+when making a function call, and it is not an issue for our runtime-generated functions as they do not make
+direct function calls to Go-implemented functions. On the other hand, the latter, async preemption, can be problematic
+since it tries to interrupt the execution of Goroutine at any point of function, and manipulates CPU register states.
+
+Fortunately, our runtime-generated machine codes do not need to take the async preemption into account.
+All the assembly codes are entered via the trampoline implemented as Go Assembler Function (e.g. [arch_amd64.s](./arch_amd64.s)),
+and as of Go 1.20, these assembler functions are considered as _unsafe_ for async preemption:
+- https://github.com/golang/go/blob/go1.20rc1/src/runtime/preempt.go#L406-L407
+- https://github.com/golang/go/blob/9f0234214473dfb785a5ad84a8fc62a6a395cbc3/src/runtime/traceback.go#L227
+
+From the Go runtime point of view, the execution of runtime-generated machine codes is considered as a part of
+that trampoline function. Therefore, runtime-generated machine code is also correctly considered unsafe for async preemption.
+
+## Why context cancellation is handled in Go code rather than native code
+
+Since [wazero v1.0.0-pre.9](https://github.com/tetratelabs/wazero/releases/tag/v1.0.0-pre.9), the runtime
+supports integration with Go contexts to interrupt execution after a timeout, or in response to explicit cancellation.
+This support is internally implemented as a special opcode `builtinFunctionCheckExitCode` that triggers the execution of
+a Go function (`ModuleInstance.FailIfClosed`) that atomically checks a sentinel value at strategic points in the code
+(e.g. [within loops][checkexitcode_loop]).
+
+[It _is indeed_ possible to check the sentinel value directly, without leaving the native world][native_check], thus sparing some cycles;
+however, because native code never preempts (see section above), this may lead to a state where the other goroutines
+never get the chance to run, and thus never get the chance to set the sentinel value; effectively preventing
+cancellation from taking place.
+
+[checkexitcode_loop]: https://github.com/tetratelabs/wazero/blob/86444c67a37dbf9e693ae5b365901f64968d9025/internal/wazeroir/compiler.go#L467-L476
+[native_check]: https://github.com/tetratelabs/wazero/issues/1409
 
 ## Golang patterns
 

README.md

@@ -41,8 +41,6 @@ during `Runtime.CompileModule`. This means your WebAssembly functions execute
 natively at runtime. Compiler is faster than Interpreter, often by order of
 magnitude (10x) or more. This is done without host-specific dependencies.
 
-If interested, check out the [RATIONALE.md][8] and help us optimize further!
-
 ### Conformance
 
 Both runtimes pass WebAssembly Core [1.0][7] and [2.0][14] specification tests
@@ -123,7 +121,6 @@ wazero is a registered trademark of Tetrate.io, Inc. in the United States and/or
 [5]: https://github.com/WebAssembly/WASI
 [6]: https://pkg.go.dev/golang.org/x/sys/unix
 [7]: https://github.com/WebAssembly/spec/tree/wg-1.0/test/core
-[8]: internal/engine/compiler/RATIONALE.md
 [9]: https://github.com/tetratelabs/wazero/issues/506
 [10]: https://go.dev/doc/devel/release
 [11]: https://github.com/actions/virtual-environments

builder_test.go

@@ -308,9 +308,7 @@ func TestNewHostModuleBuilder_Compile(t *testing.T) {
 		tc := tt
 
 		t.Run(tc.name, func(t *testing.T) {
-			cfg := NewRuntimeConfig()
-			cfg.(*runtimeConfig).EnableOptimizingCompiler()
-			b := tc.input(NewRuntimeWithConfig(testCtx, cfg)).(*hostModuleBuilder)
+			b := tc.input(NewRuntime(testCtx)).(*hostModuleBuilder)
 			compiled, err := b.Compile(testCtx)
 			require.NoError(t, err)
 			m := compiled.(*compiledModule)

cmd/wazero/wazero.go

@@ -19,7 +19,6 @@ import (
 	"github.com/tetratelabs/wazero/api"
 	"github.com/tetratelabs/wazero/experimental"
 	"github.com/tetratelabs/wazero/experimental/logging"
-	"github.com/tetratelabs/wazero/experimental/opt"
 	"github.com/tetratelabs/wazero/experimental/sock"
 	"github.com/tetratelabs/wazero/experimental/sysfs"
 	"github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1"
@@ -160,10 +159,6 @@ func doRun(args []string, stdOut io.Writer, stdErr logging.Writer) int {
 	flags.BoolVar(&useInterpreter, "interpreter", false,
 		"Interprets WebAssembly modules instead of compiling them into native code.")
 
-	var useOptimizingCompiler bool
-	flags.BoolVar(&useOptimizingCompiler, "optimizing-compiler", false,
-		"[Experimental] Compiles WebAssembly modules using the optimizing compiler.")
-
 	var envs sliceFlag
 	flags.Var(&envs, "env", "key=value pair of environment variable to expose to the binary. "+
 		"Can be specified multiple times.")
@@ -272,8 +267,6 @@ func doRun(args []string, stdOut io.Writer, stdErr logging.Writer) int {
 	var rtc wazero.RuntimeConfig
 	if useInterpreter {
 		rtc = wazero.NewRuntimeConfigInterpreter()
-	} else if useOptimizingCompiler {
-		rtc = opt.NewRuntimeConfigOptimizingCompiler()
 	} else {
 		rtc = wazero.NewRuntimeConfig()
 	}

config.go

@@ -12,7 +12,6 @@ import (
 
 	"github.com/tetratelabs/wazero/api"
 	experimentalsys "github.com/tetratelabs/wazero/experimental/sys"
-	"github.com/tetratelabs/wazero/internal/engine/compiler"
 	"github.com/tetratelabs/wazero/internal/engine/interpreter"
 	"github.com/tetratelabs/wazero/internal/engine/wazevo"
 	"github.com/tetratelabs/wazero/internal/filecache"
@@ -158,7 +157,7 @@ type RuntimeConfig interface {
 	// This is especially useful when one wants to run untrusted Wasm binaries since otherwise, any invocation of
 	// api.Function can potentially block the corresponding Goroutine forever. Moreover, it might block the
 	// entire underlying OS thread which runs the api.Function call. See "Why it's safe to execute runtime-generated
-	// machine codes against async Goroutine preemption" section in internal/engine/compiler/RATIONALE.md for detail.
+	// machine codes against async Goroutine preemption" section in RATIONALE.md for detail.
 	//
 	// Note that this comes with a bit of extra cost when enabled. The reason is that internally this forces
 	// interpreter and compiler runtimes to insert the periodical checks on the conditions above. For that reason,
@@ -191,11 +190,6 @@ type runtimeConfig struct {
 	ensureTermination     bool
 }
 
-// EnableOptimizingCompiler implements experimental/opt/enabler.EnableOptimizingCompiler.
-func (c *runtimeConfig) EnableOptimizingCompiler() {
-	c.newEngine = wazevo.NewEngine
-}
-
 // engineLessConfig helps avoid copy/pasting the wrong defaults.
 var engineLessConfig = &runtimeConfig{
 	enabledFeatures:       api.CoreFeaturesV2,
@@ -229,7 +223,7 @@ const (
 func NewRuntimeConfigCompiler() RuntimeConfig {
 	ret := engineLessConfig.clone()
 	ret.engineKind = engineKindCompiler
-	ret.newEngine = compiler.NewEngine
+	ret.newEngine = wazevo.NewEngine
 	return ret
 }
 

experimental/checkpoint_example_test.go

@@ -23,7 +23,8 @@ type snapshotsKey struct{}
 func Example_enableSnapshotterKey() {
 	ctx := context.Background()
 
-	rt := wazero.NewRuntime(ctx)
+	// TODO: currently, only the interpreter is supported for snapshotting.
+	rt := wazero.NewRuntimeWithConfig(ctx, wazero.NewRuntimeConfigInterpreter())
 	defer rt.Close(ctx) // This closes everything this Runtime created.
 
 	// Enable experimental snapshotting functionality by setting it to context. We use this

experimental/checkpoint_test.go

@@ -13,7 +13,8 @@ import (
 func TestSnapshotNestedWasmInvocation(t *testing.T) {
 	ctx := context.Background()
 
-	rt := wazero.NewRuntime(ctx)
+	// TODO: currently, only the interpreter is supported for snapshotting.
+	rt := wazero.NewRuntimeWithConfig(ctx, wazero.NewRuntimeConfigInterpreter())
 	defer rt.Close(ctx)
 
 	sidechannel := 0
@@ -69,7 +70,8 @@ func TestSnapshotNestedWasmInvocation(t *testing.T) {
 func TestSnapshotMultipleWasmInvocations(t *testing.T) {
 	ctx := context.Background()
 
-	rt := wazero.NewRuntime(ctx)
+	// TODO: currently, only the interpreter is supported for snapshotting.
+	rt := wazero.NewRuntimeWithConfig(ctx, wazero.NewRuntimeConfigInterpreter())
 	defer rt.Close(ctx)
 
 	_, err := rt.NewHostModuleBuilder("example").

experimental/features_example_test.go

@@ -9,7 +9,7 @@ import (
 	"sync"
 	"sync/atomic"
 
-	wazero "github.com/tetratelabs/wazero"
+	"github.com/tetratelabs/wazero"
 	"github.com/tetratelabs/wazero/api"
 	"github.com/tetratelabs/wazero/experimental"
 	"github.com/tetratelabs/wazero/imports/wasi_snapshot_preview1"
@@ -34,7 +34,9 @@ func ExampleCoreFeaturesThreads() {
 	ctx := context.Background()
 
 	// Threads support must be enabled explicitly in addition to standard V2 features.
-	cfg := wazero.NewRuntimeConfig().WithCoreFeatures(api.CoreFeaturesV2 | experimental.CoreFeaturesThreads)
+
+	// TODO: currently, only the interpreter is supported for snapshotting.
+	cfg := wazero.NewRuntimeConfigInterpreter().WithCoreFeatures(api.CoreFeaturesV2 | experimental.CoreFeaturesThreads)
 
 	r := wazero.NewRuntimeWithConfig(ctx, cfg)
 	defer r.Close(ctx)

experimental/listener_test.go

@@ -8,9 +8,7 @@ import (
 	"github.com/tetratelabs/wazero"
 	"github.com/tetratelabs/wazero/api"
 	"github.com/tetratelabs/wazero/experimental"
-	"github.com/tetratelabs/wazero/experimental/opt"
 	"github.com/tetratelabs/wazero/experimental/wazerotest"
-	"github.com/tetratelabs/wazero/internal/platform"
 	"github.com/tetratelabs/wazero/internal/testing/binaryencoding"
 	"github.com/tetratelabs/wazero/internal/testing/require"
 	"github.com/tetratelabs/wazero/internal/wasm"
@@ -75,12 +73,7 @@ func TestFunctionListenerFactory(t *testing.T) {
 		},
 	})
 
-	var r wazero.Runtime
-	if platform.CompilerSupported() {
-		r = wazero.NewRuntimeWithConfig(ctx, opt.NewRuntimeConfigOptimizingCompiler())
-	} else {
-		r = wazero.NewRuntime(ctx)
-	}
+	r := wazero.NewRuntime(ctx)
 	defer r.Close(ctx) // This closes everything this Runtime created.
 
 	_, err := r.NewHostModuleBuilder("host").NewFunctionBuilder().WithFunc(func() {}).Export("").Instantiate(ctx)