compiler: refactor public interface

This commit merges NewCompiler and Compile into one simplifying the
external interface. More importantly, it does away with the entire
Compiler object so the public API becomes a lot smaller.

The refactor is not complete: eventually, the compiler should just
compile a single package without trying to load it first (that should be
done by the builder package).

builder/build.go

@@ -17,6 +17,7 @@ import (
 	"github.com/tinygo-org/tinygo/goenv"
 	"github.com/tinygo-org/tinygo/interp"
 	"github.com/tinygo-org/tinygo/transform"
+	"tinygo.org/x/go-llvm"
 )
 
 // Build performs a single package to executable Go build. It takes in a package
@@ -26,34 +27,34 @@ import (
 // The error value may be of type *MultiError. Callers will likely want to check
 // for this case and print such errors individually.
 func Build(pkgName, outpath string, config *compileopts.Config, action func(string) error) error {
-	c, err := compiler.NewCompiler(pkgName, config)
+	// Compile Go code to IR.
+	machine, err := compiler.NewTargetMachine(config)
 	if err != nil {
 		return err
 	}
-
-	// Compile Go code to IR.
-	errs := c.Compile(pkgName)
-	if len(errs) != 0 {
+	mod, extraFiles, errs := compiler.Compile(pkgName, machine, config)
+	if errs != nil {
 		return newMultiError(errs)
 	}
+
 	if config.Options.PrintIR {
 		fmt.Println("; Generated LLVM IR:")
-		fmt.Println(c.IR())
+		fmt.Println(mod.String())
 	}
-	if err := c.Verify(); err != nil {
+	if err := llvm.VerifyModule(mod, llvm.PrintMessageAction); err != nil {
 		return errors.New("verification error after IR construction")
 	}
 
-	err = interp.Run(c.Module(), config.DumpSSA())
+	err = interp.Run(mod, config.DumpSSA())
 	if err != nil {
 		return err
 	}
-	if err := c.Verify(); err != nil {
+	if err := llvm.VerifyModule(mod, llvm.PrintMessageAction); err != nil {
 		return errors.New("verification error after interpreting runtime.initAll")
 	}
 
 	if config.GOOS() != "darwin" {
-		transform.ApplyFunctionSections(c.Module()) // -ffunction-sections
+		transform.ApplyFunctionSections(mod) // -ffunction-sections
 	}
 
 	// Browsers cannot handle external functions that have type i64 because it
@@ -62,7 +63,7 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(stri
 	// stack-allocated values.
 	// Use -wasm-abi=generic to disable this behaviour.
 	if config.Options.WasmAbi == "js" && strings.HasPrefix(config.Triple(), "wasm") {
-		err := transform.ExternalInt64AsPtr(c.Module())
+		err := transform.ExternalInt64AsPtr(mod)
 		if err != nil {
 			return err
 		}
@@ -73,22 +74,22 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(stri
 	errs = nil
 	switch config.Options.Opt {
 	case "none", "0":
-		errs = transform.Optimize(c.Module(), config, 0, 0, 0) // -O0
+		errs = transform.Optimize(mod, config, 0, 0, 0) // -O0
 	case "1":
-		errs = transform.Optimize(c.Module(), config, 1, 0, 0) // -O1
+		errs = transform.Optimize(mod, config, 1, 0, 0) // -O1
 	case "2":
-		errs = transform.Optimize(c.Module(), config, 2, 0, 225) // -O2
+		errs = transform.Optimize(mod, config, 2, 0, 225) // -O2
 	case "s":
-		errs = transform.Optimize(c.Module(), config, 2, 1, 225) // -Os
+		errs = transform.Optimize(mod, config, 2, 1, 225) // -Os
 	case "z":
-		errs = transform.Optimize(c.Module(), config, 2, 2, 5) // -Oz, default
+		errs = transform.Optimize(mod, config, 2, 2, 5) // -Oz, default
 	default:
 		errs = []error{errors.New("unknown optimization level: -opt=" + config.Options.Opt)}
 	}
 	if len(errs) > 0 {
 		return newMultiError(errs)
 	}
-	if err := c.Verify(); err != nil {
+	if err := llvm.VerifyModule(mod, llvm.PrintMessageAction); err != nil {
 		return errors.New("verification failure after LLVM optimization passes")
 	}
 
@@ -98,8 +99,8 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(stri
 	// pointers are flash and which are in RAM so that pointers can have a
 	// correct address space parameter (address space 1 is for flash).
 	if strings.HasPrefix(config.Triple(), "avr") {
-		transform.NonConstGlobals(c.Module())
-		if err := c.Verify(); err != nil {
+		transform.NonConstGlobals(mod)
+		if err := llvm.VerifyModule(mod, llvm.PrintMessageAction); err != nil {
 			return errors.New("verification error after making all globals non-constant on AVR")
 		}
 	}
@@ -108,11 +109,17 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(stri
 	outext := filepath.Ext(outpath)
 	switch outext {
 	case ".o":
-		return c.EmitObject(outpath)
+		llvmBuf, err := machine.EmitToMemoryBuffer(mod, llvm.ObjectFile)
+		if err != nil {
+			return err
+		}
+		return ioutil.WriteFile(outpath, llvmBuf.Bytes(), 0666)
 	case ".bc":
-		return c.EmitBitcode(outpath)
+		data := llvm.WriteBitcodeToMemoryBuffer(mod).Bytes()
+		return ioutil.WriteFile(outpath, data, 0666)
 	case ".ll":
-		return c.EmitText(outpath)
+		data := []byte(mod.String())
+		return ioutil.WriteFile(outpath, data, 0666)
 	default:
 		// Act as a compiler driver.
 
@@ -125,7 +132,11 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(stri
 
 		// Write the object file.
 		objfile := filepath.Join(dir, "main.o")
-		err = c.EmitObject(objfile)
+		llvmBuf, err := machine.EmitToMemoryBuffer(mod, llvm.ObjectFile)
+		if err != nil {
+			return err
+		}
+		err = ioutil.WriteFile(objfile, llvmBuf.Bytes(), 0666)
 		if err != nil {
 			return err
 		}
@@ -167,16 +178,13 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(stri
 		}
 
 		// Compile C files in packages.
-		for i, pkg := range c.Packages() {
-			for _, file := range pkg.CFiles {
-				path := filepath.Join(pkg.Package.Dir, file)
-				outpath := filepath.Join(dir, "pkg"+strconv.Itoa(i)+"-"+file+".o")
-				err := runCCompiler(config.Target.Compiler, append(config.CFlags(), "-c", "-o", outpath, path)...)
-				if err != nil {
-					return &commandError{"failed to build", path, err}
-				}
-				ldflags = append(ldflags, outpath)
+		for i, file := range extraFiles {
+			outpath := filepath.Join(dir, "pkg"+strconv.Itoa(i)+"-"+filepath.Base(file)+".o")
+			err := runCCompiler(config.Target.Compiler, append(config.CFlags(), "-c", "-o", outpath, file)...)
+			if err != nil {
+				return &commandError{"failed to build", file, err}
 			}
+			ldflags = append(ldflags, outpath)
 		}
 
 		// Link the object files together.

compiler/compiler.go

@@ -56,12 +56,6 @@ type compilerContext struct {
 	astComments      map[string]*ast.CommentGroup
 }
 
-type Compiler struct {
-	compilerContext
-	builder   llvm.Builder
-	initFuncs []llvm.Value
-}
-
 // builder contains all information relevant to build a single function.
 type builder struct {
 	*compilerContext
@@ -88,28 +82,40 @@ type Phi struct {
 	llvm llvm.Value
 }
 
-func NewCompiler(pkgName string, config *compileopts.Config) (*Compiler, error) {
-	c := &Compiler{
-		compilerContext: compilerContext{
-			Config:  config,
-			difiles: make(map[string]llvm.Metadata),
-			ditypes: make(map[types.Type]llvm.Metadata),
-		},
-	}
-
+// NewTargetMachine returns a new llvm.TargetMachine based on the passed-in
+// configuration. It is used by the compiler and is needed for machine code
+// emission.
+func NewTargetMachine(config *compileopts.Config) (llvm.TargetMachine, error) {
 	target, err := llvm.GetTargetFromTriple(config.Triple())
 	if err != nil {
-		return nil, err
+		return llvm.TargetMachine{}, err
 	}
 	features := strings.Join(config.Features(), ",")
-	c.machine = target.CreateTargetMachine(config.Triple(), config.CPU(), features, llvm.CodeGenLevelDefault, llvm.RelocStatic, llvm.CodeModelDefault)
-	c.targetData = c.machine.CreateTargetData()
+	machine := target.CreateTargetMachine(config.Triple(), config.CPU(), features, llvm.CodeGenLevelDefault, llvm.RelocStatic, llvm.CodeModelDefault)
+	return machine, nil
+}
+
+// Compile the given package path or .go file path. Return an error when this
+// fails (in any stage). If successful it returns the LLVM module and a list of
+// extra C files to be compiled. If not, one or more errors will be returned.
+//
+// The fact that it returns a list of filenames to compile is a layering
+// violation. Eventually, this Compile function should only compile a single
+// package and not the whole program, and loading of the program (including CGo
+// processing) should be moved outside the compiler package.
+func Compile(pkgName string, machine llvm.TargetMachine, config *compileopts.Config) (llvm.Module, []string, []error) {
+	c := &compilerContext{
+		Config:     config,
+		difiles:    make(map[string]llvm.Metadata),
+		ditypes:    make(map[types.Type]llvm.Metadata),
+		machine:    machine,
+		targetData: machine.CreateTargetData(),
+	}
 
 	c.ctx = llvm.NewContext()
 	c.mod = c.ctx.NewModule(pkgName)
 	c.mod.SetTarget(config.Triple())
 	c.mod.SetDataLayout(c.targetData.String())
-	c.builder = c.ctx.NewBuilder()
 	if c.Debug() {
 		c.dibuilder = llvm.NewDIBuilder(c.mod)
 	}
@@ -131,21 +137,6 @@ func NewCompiler(pkgName string, config *compileopts.Config) (*Compiler, error)
 	c.funcPtrAddrSpace = dummyFunc.Type().PointerAddressSpace()
 	dummyFunc.EraseFromParentAsFunction()
 
-	return c, nil
-}
-
-func (c *Compiler) Packages() []*loader.Package {
-	return c.ir.LoaderProgram.Sorted()
-}
-
-// Return the LLVM module. Only valid after a successful compile.
-func (c *Compiler) Module() llvm.Module {
-	return c.mod
-}
-
-// Compile the given package path or .go file path. Return an error when this
-// fails (in any stage).
-func (c *Compiler) Compile(mainPath string) []error {
 	// Prefix the GOPATH with the system GOROOT, as GOROOT is already set to
 	// the TinyGo root.
 	overlayGopath := goenv.Get("GOPATH")
@@ -157,7 +148,7 @@ func (c *Compiler) Compile(mainPath string) []error {
 
 	wd, err := os.Getwd()
 	if err != nil {
-		return []error{err}
+		return c.mod, nil, []error{err}
 	}
 	lprogram := &loader.Program{
 		Build: &build.Context{
@@ -217,17 +208,17 @@ func (c *Compiler) Compile(mainPath string) []error {
 		ClangHeaders: c.ClangHeaders,
 	}
 
-	if strings.HasSuffix(mainPath, ".go") {
-		_, err = lprogram.ImportFile(mainPath)
+	if strings.HasSuffix(pkgName, ".go") {
+		_, err = lprogram.ImportFile(pkgName)
 		if err != nil {
-			return []error{err}
+			return c.mod, nil, []error{err}
 		}
 	} else {
-		_, err = lprogram.Import(mainPath, wd, token.Position{
+		_, err = lprogram.Import(pkgName, wd, token.Position{
 			Filename: "build command-line-arguments",
 		})
 		if err != nil {
-			return []error{err}
+			return c.mod, nil, []error{err}
 		}
 	}
 
@@ -235,15 +226,15 @@ func (c *Compiler) Compile(mainPath string) []error {
 		Filename: "build default import",
 	})
 	if err != nil {
-		return []error{err}
+		return c.mod, nil, []error{err}
 	}
 
 	err = lprogram.Parse(c.TestConfig.CompileTestBinary)
 	if err != nil {
-		return []error{err}
+		return c.mod, nil, []error{err}
 	}
 
-	c.ir = ir.NewProgram(lprogram, mainPath)
+	c.ir = ir.NewProgram(lprogram, pkgName)
 
 	// Run a simple dead code elimination pass.
 	c.ir.SimpleDCE()
@@ -252,7 +243,7 @@ func (c *Compiler) Compile(mainPath string) []error {
 	if c.Debug() {
 		c.cu = c.dibuilder.CreateCompileUnit(llvm.DICompileUnit{
 			Language:  0xb, // DW_LANG_C99 (0xc, off-by-one?)
-			File:      mainPath,
+			File:      pkgName,
 			Dir:       "",
 			Producer:  "TinyGo",
 			Optimized: true,
@@ -281,9 +272,12 @@ func (c *Compiler) Compile(mainPath string) []error {
 	}
 
 	// Add definitions to declarations.
+	var initFuncs []llvm.Value
+	irbuilder := c.ctx.NewBuilder()
+	defer irbuilder.Dispose()
 	for _, f := range c.ir.Functions {
 		if f.Synthetic == "package initializer" {
-			c.initFuncs = append(c.initFuncs, f.LLVMFn)
+			initFuncs = append(initFuncs, f.LLVMFn)
 		}
 		if f.CName() != "" {
 			continue
@@ -294,8 +288,8 @@ func (c *Compiler) Compile(mainPath string) []error {
 
 		// Create the function definition.
 		b := builder{
-			compilerContext: &c.compilerContext,
-			Builder:         c.builder,
+			compilerContext: c,
+			Builder:         irbuilder,
 			fn:              f,
 			locals:          make(map[ssa.Value]llvm.Value),
 			dilocals:        make(map[*types.Var]llvm.Metadata),
@@ -313,14 +307,14 @@ func (c *Compiler) Compile(mainPath string) []error {
 	if c.Debug() {
 		difunc := c.attachDebugInfo(initFn)
 		pos := c.ir.Program.Fset.Position(initFn.Pos())
-		c.builder.SetCurrentDebugLocation(uint(pos.Line), uint(pos.Column), difunc, llvm.Metadata{})
+		irbuilder.SetCurrentDebugLocation(uint(pos.Line), uint(pos.Column), difunc, llvm.Metadata{})
 	}
 	block := c.ctx.AddBasicBlock(initFn.LLVMFn, "entry")
-	c.builder.SetInsertPointAtEnd(block)
-	for _, fn := range c.initFuncs {
-		c.builder.CreateCall(fn, []llvm.Value{llvm.Undef(c.i8ptrType), llvm.Undef(c.i8ptrType)}, "")
+	irbuilder.SetInsertPointAtEnd(block)
+	for _, fn := range initFuncs {
+		irbuilder.CreateCall(fn, []llvm.Value{llvm.Undef(c.i8ptrType), llvm.Undef(c.i8ptrType)}, "")
 	}
-	c.builder.CreateRetVoid()
+	irbuilder.CreateRetVoid()
 
 	// Conserve for goroutine lowering. Without marking these as external, they
 	// would be optimized away.
@@ -392,7 +386,15 @@ func (c *Compiler) Compile(mainPath string) []error {
 		c.dibuilder.Finalize()
 	}
 
-	return c.diagnostics
+	// Gather the list of (C) file paths that should be included in the build.
+	var extraFiles []string
+	for _, pkg := range c.ir.LoaderProgram.Sorted() {
+		for _, file := range pkg.CFiles {
+			extraFiles = append(extraFiles, filepath.Join(pkg.Package.Dir, file))
+		}
+	}
+
+	return c.mod, extraFiles, c.diagnostics
 }
 
 // getLLVMRuntimeType obtains a named type from the runtime package and returns
@@ -2577,47 +2579,3 @@ func (b *builder) createUnOp(unop *ssa.UnOp) (llvm.Value, error) {
 		return llvm.Value{}, b.makeError(unop.Pos(), "todo: unknown unop")
 	}
 }
-
-// IR returns the whole IR as a human-readable string.
-func (c *Compiler) IR() string {
-	return c.mod.String()
-}
-
-func (c *Compiler) Verify() error {
-	return llvm.VerifyModule(c.mod, llvm.PrintMessageAction)
-}
-
-// Emit object file (.o).
-func (c *Compiler) EmitObject(path string) error {
-	llvmBuf, err := c.machine.EmitToMemoryBuffer(c.mod, llvm.ObjectFile)
-	if err != nil {
-		return err
-	}
-	return c.writeFile(llvmBuf.Bytes(), path)
-}
-
-// Emit LLVM bitcode file (.bc).
-func (c *Compiler) EmitBitcode(path string) error {
-	data := llvm.WriteBitcodeToMemoryBuffer(c.mod).Bytes()
-	return c.writeFile(data, path)
-}
-
-// Emit LLVM IR source file (.ll).
-func (c *Compiler) EmitText(path string) error {
-	data := []byte(c.mod.String())
-	return c.writeFile(data, path)
-}
-
-// Write the data to the file specified by path.
-func (c *Compiler) writeFile(data []byte, path string) error {
-	// Write output to file
-	f, err := os.OpenFile(path, os.O_RDWR|os.O_CREATE|os.O_TRUNC, 0666)
-	if err != nil {
-		return err
-	}
-	_, err = f.Write(data)
-	if err != nil {
-		return err
-	}
-	return f.Close()
-}

compiler/symbol.go

@@ -26,7 +26,7 @@ type globalInfo struct {
 
 // loadASTComments loads comments on globals from the AST, for use later in the
 // program. In particular, they are required for //go:extern pragmas on globals.
-func (c *Compiler) loadASTComments(lprogram *loader.Program) {
+func (c *compilerContext) loadASTComments(lprogram *loader.Program) {
 	c.astComments = map[string]*ast.CommentGroup{}
 	for _, pkgInfo := range lprogram.Sorted() {
 		for _, file := range pkgInfo.Files {