Updated README to mention compiling for Windows

This metadata is emitted by Clang and I found it is important for source
level debugging on MacOS. This patch does not get source level debugging
to work yet (for that, it seems like packages need to be built
separately), but it is a step in the right direction.

.dockerignore

@@ -1,3 +1,5 @@
 build/
 llvm-*/
+.github
+.circleci
 

.github/workflows/build-macos.yml

@@ -29,11 +29,11 @@ jobs:
         with:
           go-version: '1.20'
           cache: true
-      - name: Cache LLVM source
-        uses: actions/cache@v3
+      - name: Restore LLVM source cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-source
         with:
-          key: llvm-source-15-macos-v2
+          key: llvm-source-15-macos-v3
           path: |
             llvm-project/clang/lib/Headers
             llvm-project/clang/include
@@ -43,11 +43,22 @@ jobs:
       - name: Download LLVM source
         if: steps.cache-llvm-source.outputs.cache-hit != 'true'
         run: make llvm-source
-      - name: Cache LLVM build
-        uses: actions/cache@v3
+      - name: Save LLVM source cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-source.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-source.outputs.cache-primary-key }}
+          path: |
+            llvm-project/clang/lib/Headers
+            llvm-project/clang/include
+            llvm-project/compiler-rt
+            llvm-project/lld/include
+            llvm-project/llvm/include
+      - name: Restore LLVM build cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-build
         with:
-          key: llvm-build-15-macos-v3
+          key: llvm-build-15-macos-v4
           path: llvm-build
       - name: Build LLVM
         if: steps.cache-llvm-build.outputs.cache-hit != 'true'
@@ -61,6 +72,12 @@ jobs:
           # build!
           make llvm-build
           find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
+      - name: Save LLVM build cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-build.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-build.outputs.cache-primary-key }}
+          path: llvm-build
       - name: Cache wasi-libc sysroot
         uses: actions/cache@v3
         id: cache-wasi-libc
@@ -72,7 +89,7 @@ jobs:
         run: make wasi-libc
       - name: Test TinyGo
         shell: bash
-        run: make test GOTESTFLAGS="-v -short"
+        run: make test GOTESTFLAGS="-short"
       - name: Build TinyGo release tarball
         run: make release -j3
       - name: Test stdlib packages

.github/workflows/docker.yml

@@ -53,6 +53,7 @@ jobs:
           push: true
           tags: ${{ steps.meta.outputs.tags }}
           labels: ${{ steps.meta.outputs.labels }}
+          build-contexts: tinygo-llvm-build=docker-image://tinygo/llvm-15
           cache-from: type=gha
           cache-to: type=gha,mode=max
       - name: Trigger Drivers repo build on Github Actions
@@ -69,21 +70,31 @@ jobs:
           -H "Accept: application/vnd.github.v3+json" \
           https://api.github.com/repos/tinygo-org/bluetooth/actions/workflows/linux.yml/dispatches \
           -d '{"ref": "dev"}'
-      - name: Trigger TinyFS repo build on CircleCI
+      - name: Trigger TinyFS repo build on Github Actions
         run: |
-          curl --location --request POST 'https://circleci.com/api/v2/project/github/tinygo-org/tinyfs/pipeline' \
-            --header 'Content-Type: application/json' \
-            -d '{"branch": "dev"}' \
-            -u "${{ secrets.CIRCLECI_API_TOKEN }}"
-      - name: Trigger TinyFont repo build on CircleCI
+          curl -X POST \
+          -H "Authorization: Bearer ${{secrets.GHA_ACCESS_TOKEN}}" \
+          -H "Accept: application/vnd.github.v3+json" \
+          https://api.github.com/repos/tinygo-org/tinyfs/actions/workflows/build.yml/dispatches \
+          -d '{"ref": "dev"}'
+      - name: Trigger TinyFont repo build on Github Actions
         run: |
-          curl --location --request POST 'https://circleci.com/api/v2/project/github/tinygo-org/tinyfont/pipeline' \
-            --header 'Content-Type: application/json' \
-            -d '{"branch": "dev"}' \
-            -u "${{ secrets.CIRCLECI_API_TOKEN }}"
-      - name: Trigger TinyDraw repo build on CircleCI
+          curl -X POST \
+          -H "Authorization: Bearer ${{secrets.GHA_ACCESS_TOKEN}}" \
+          -H "Accept: application/vnd.github.v3+json" \
+          https://api.github.com/repos/tinygo-org/tinyfont/actions/workflows/build.yml/dispatches \
+          -d '{"ref": "dev"}'
+      - name: Trigger TinyDraw repo build on Github Actions
         run: |
-          curl --location --request POST 'https://circleci.com/api/v2/project/github/tinygo-org/tinydraw/pipeline' \
-            --header 'Content-Type: application/json' \
-            -d '{"branch": "dev"}' \
-            -u "${{ secrets.CIRCLECI_API_TOKEN }}"
+          curl -X POST \
+          -H "Authorization: Bearer ${{secrets.GHA_ACCESS_TOKEN}}" \
+          -H "Accept: application/vnd.github.v3+json" \
+          https://api.github.com/repos/tinygo-org/tinydraw/actions/workflows/build.yml/dispatches \
+          -d '{"ref": "dev"}'
+      - name: Trigger TinyTerm repo build on Github Actions
+        run: |
+          curl -X POST \
+          -H "Authorization: Bearer ${{secrets.GHA_ACCESS_TOKEN}}" \
+          -H "Accept: application/vnd.github.v3+json" \
+          https://api.github.com/repos/tinygo-org/tinyterm/actions/workflows/build.yml/dispatches \
+          -d '{"ref": "dev"}'

.github/workflows/linux.yml

@@ -39,11 +39,11 @@ jobs:
           path: |
             ~/.cache/go-build
             ~/go/pkg/mod
-      - name: Cache LLVM source
-        uses: actions/cache@v3
+      - name: Restore LLVM source cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-source
         with:
-          key: llvm-source-15-linux-alpine-v2
+          key: llvm-source-15-linux-alpine-v3
           path: |
             llvm-project/clang/lib/Headers
             llvm-project/clang/include
@@ -53,11 +53,22 @@ jobs:
       - name: Download LLVM source
         if: steps.cache-llvm-source.outputs.cache-hit != 'true'
         run: make llvm-source
-      - name: Cache LLVM build
-        uses: actions/cache@v3
+      - name: Save LLVM source cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-source.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-source.outputs.cache-primary-key }}
+          path: |
+            llvm-project/clang/lib/Headers
+            llvm-project/clang/include
+            llvm-project/compiler-rt
+            llvm-project/lld/include
+            llvm-project/llvm/include
+      - name: Restore LLVM build cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-build
         with:
-          key: llvm-build-15-linux-alpine-v3
+          key: llvm-build-15-linux-alpine-v4
           path: llvm-build
       - name: Build LLVM
         if: steps.cache-llvm-build.outputs.cache-hit != 'true'
@@ -71,6 +82,12 @@ jobs:
           make llvm-build
           # Remove unnecessary object files (to reduce cache size).
           find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
+      - name: Save LLVM build cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-build.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-build.outputs.cache-primary-key }}
+          path: llvm-build
       - name: Cache Binaryen
         uses: actions/cache@v3
         id: cache-binaryen
@@ -122,7 +139,9 @@ jobs:
           cache: true
       - name: Install wasmtime
         run: |
-          curl https://wasmtime.dev/install.sh -sSf | bash
+          mkdir -p $HOME/.wasmtime $HOME/.wasmtime/bin
+          curl https://github.com/bytecodealliance/wasmtime/releases/download/v5.0.0/wasmtime-v5.0.0-x86_64-linux.tar.xz -o wasmtime-v5.0.0-x86_64-linux.tar.xz -SfL
+          tar -C $HOME/.wasmtime/bin --wildcards -xf wasmtime-v5.0.0-x86_64-linux.tar.xz --strip-components=1 wasmtime-v5.0.0-x86_64-linux/*
           echo "$HOME/.wasmtime/bin" >> $GITHUB_PATH
       - name: Download release artifact
         uses: actions/download-artifact@v3
@@ -166,13 +185,15 @@ jobs:
           node-version: '14'
       - name: Install wasmtime
         run: |
-          curl https://wasmtime.dev/install.sh -sSf | bash
+          mkdir -p $HOME/.wasmtime $HOME/.wasmtime/bin
+          curl -L https://github.com/bytecodealliance/wasmtime/releases/download/v5.0.0/wasmtime-v5.0.0-x86_64-linux.tar.xz -o wasmtime-v5.0.0-x86_64-linux.tar.xz -SfL
+          tar -C $HOME/.wasmtime/bin --wildcards -xf wasmtime-v5.0.0-x86_64-linux.tar.xz --strip-components=1 wasmtime-v5.0.0-x86_64-linux/*
           echo "$HOME/.wasmtime/bin" >> $GITHUB_PATH
-      - name: Cache LLVM source
-        uses: actions/cache@v3
+      - name: Restore LLVM source cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-source
         with:
-          key: llvm-source-15-linux-asserts-v2
+          key: llvm-source-15-linux-asserts-v3
           path: |
             llvm-project/clang/lib/Headers
             llvm-project/clang/include
@@ -182,11 +203,22 @@ jobs:
       - name: Download LLVM source
         if: steps.cache-llvm-source.outputs.cache-hit != 'true'
         run: make llvm-source
-      - name: Cache LLVM build
-        uses: actions/cache@v3
+      - name: Save LLVM source cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-source.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-source.outputs.cache-primary-key }}
+          path: |
+            llvm-project/clang/lib/Headers
+            llvm-project/clang/include
+            llvm-project/compiler-rt
+            llvm-project/lld/include
+            llvm-project/llvm/include
+      - name: Restore LLVM build cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-build
         with:
-          key: llvm-build-15-linux-asserts-v3
+          key: llvm-build-15-linux-asserts-v4
           path: llvm-build
       - name: Build LLVM
         if: steps.cache-llvm-build.outputs.cache-hit != 'true'
@@ -198,6 +230,12 @@ jobs:
           make llvm-build ASSERT=1
           # Remove unnecessary object files (to reduce cache size).
           find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
+      - name: Save LLVM build cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-build.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-build.outputs.cache-primary-key }}
+          path: llvm-build
       - name: Cache Binaryen
         uses: actions/cache@v3
         id: cache-binaryen
@@ -254,11 +292,11 @@ jobs:
         with:
           go-version: '1.20'
           cache: true
-      - name: Cache LLVM source
-        uses: actions/cache@v3
+      - name: Restore LLVM source cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-source
         with:
-          key: llvm-source-15-linux-v2
+          key: llvm-source-15-linux-v3
           path: |
             llvm-project/clang/lib/Headers
             llvm-project/clang/include
@@ -268,11 +306,22 @@ jobs:
       - name: Download LLVM source
         if: steps.cache-llvm-source.outputs.cache-hit != 'true'
         run: make llvm-source
-      - name: Cache LLVM build
-        uses: actions/cache@v3
+      - name: Save LLVM source cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-source.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-source.outputs.cache-primary-key }}
+          path: |
+            llvm-project/clang/lib/Headers
+            llvm-project/clang/include
+            llvm-project/compiler-rt
+            llvm-project/lld/include
+            llvm-project/llvm/include
+      - name: Restore LLVM build cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-build
         with:
-          key: llvm-build-15-linux-arm-v3
+          key: llvm-build-15-linux-arm-v4
           path: llvm-build
       - name: Build LLVM
         if: steps.cache-llvm-build.outputs.cache-hit != 'true'
@@ -286,6 +335,12 @@ jobs:
           make llvm-build CROSS=arm-linux-gnueabihf
           # Remove unnecessary object files (to reduce cache size).
           find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
+      - name: Save LLVM build cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-build.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-build.outputs.cache-primary-key }}
+          path: llvm-build
       - name: Cache Binaryen
         uses: actions/cache@v3
         id: cache-binaryen
@@ -354,11 +409,11 @@ jobs:
         with:
           go-version: '1.20'
           cache: true
-      - name: Cache LLVM source
-        uses: actions/cache@v3
+      - name: Restore LLVM source cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-source
         with:
-          key: llvm-source-15-linux-v2
+          key: llvm-source-15-linux-v3
           path: |
             llvm-project/clang/lib/Headers
             llvm-project/clang/include
@@ -368,11 +423,22 @@ jobs:
       - name: Download LLVM source
         if: steps.cache-llvm-source.outputs.cache-hit != 'true'
         run: make llvm-source
-      - name: Cache LLVM build
-        uses: actions/cache@v3
+      - name: Save LLVM source cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-source.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-source.outputs.cache-primary-key }}
+          path: |
+            llvm-project/clang/lib/Headers
+            llvm-project/clang/include
+            llvm-project/compiler-rt
+            llvm-project/lld/include
+            llvm-project/llvm/include
+      - name: Restore LLVM build cache
+        uses: actions/cache/restore@v3
         id: cache-llvm-build
         with:
-          key: llvm-build-15-linux-arm64-v3
+          key: llvm-build-15-linux-arm64-v4
           path: llvm-build
       - name: Build LLVM
         if: steps.cache-llvm-build.outputs.cache-hit != 'true'
@@ -384,6 +450,12 @@ jobs:
           make llvm-build CROSS=aarch64-linux-gnu
           # Remove unnecessary object files (to reduce cache size).
           find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
+      - name: Save LLVM build cache
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-build.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-build.outputs.cache-primary-key }}
+          path: llvm-build
       - name: Cache Binaryen
         uses: actions/cache@v3
         id: cache-binaryen

.github/workflows/windows.yml

@@ -15,6 +15,12 @@ jobs:
   build-windows:
     runs-on: windows-2022
     steps:
+      - name: Configure pagefile
+        uses: al-cheb/configure-pagefile-action@v1.3
+        with:
+          minimum-size: 8GB
+          maximum-size: 24GB
+          disk-root: "C:"
       - uses: brechtm/setup-scoop@v2
         with:
           scoop_update: 'false'
@@ -31,11 +37,11 @@ jobs:
         with:
           go-version: '1.20'
           cache: true
-      - name: Cache LLVM source
-        uses: actions/cache@v3
+      - name: Restore cached LLVM source
+        uses: actions/cache/restore@v3
         id: cache-llvm-source
         with:
-          key: llvm-source-15-windows-v2
+          key: llvm-source-15-windows-v4
           path: |
             llvm-project/clang/lib/Headers
             llvm-project/clang/include
@@ -45,11 +51,22 @@ jobs:
       - name: Download LLVM source
         if: steps.cache-llvm-source.outputs.cache-hit != 'true'
         run: make llvm-source
-      - name: Cache LLVM build
-        uses: actions/cache@v3
+      - name: Save cached LLVM source
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-source.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-source.outputs.cache-primary-key }}
+          path: |
+            llvm-project/clang/lib/Headers
+            llvm-project/clang/include
+            llvm-project/compiler-rt
+            llvm-project/lld/include
+            llvm-project/llvm/include
+      - name: Restore cached LLVM build
+        uses: actions/cache/restore@v3
         id: cache-llvm-build
         with:
-          key: llvm-build-15-windows-v3
+          key: llvm-build-15-windows-v5
           path: llvm-build
       - name: Build LLVM
         if: steps.cache-llvm-build.outputs.cache-hit != 'true'
@@ -62,6 +79,12 @@ jobs:
           make llvm-build CCACHE=OFF
           # Remove unnecessary object files (to reduce cache size).
           find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
+      - name: Save cached LLVM build
+        uses: actions/cache/save@v3
+        if: steps.cache-llvm-build.outputs.cache-hit != 'true'
+        with:
+          key: ${{ steps.cache-llvm-build.outputs.cache-primary-key }}
+          path: llvm-build
       - name: Cache wasi-libc sysroot
         uses: actions/cache@v3
         id: cache-wasi-libc
@@ -76,7 +99,7 @@ jobs:
           scoop install wasmtime
       - name: Test TinyGo
         shell: bash
-        run: make test GOTESTFLAGS="-v -short"
+        run: make test GOTESTFLAGS="-short"
       - name: Build TinyGo release tarball
         shell: bash
         run: make build/release -j4
@@ -100,6 +123,12 @@ jobs:
     runs-on: windows-2022
     needs: build-windows
     steps:
+      - name: Configure pagefile
+        uses: al-cheb/configure-pagefile-action@v1.3
+        with:
+          minimum-size: 8GB
+          maximum-size: 24GB
+          disk-root: "C:"
       - uses: brechtm/setup-scoop@v2
         with:
           scoop_update: 'false'
@@ -131,6 +160,12 @@ jobs:
     runs-on: windows-2022
     needs: build-windows
     steps:
+      - name: Configure pagefile
+        uses: al-cheb/configure-pagefile-action@v1.3
+        with:
+          minimum-size: 8GB
+          maximum-size: 24GB
+          disk-root: "C:"
       - name: Checkout
         uses: actions/checkout@v3
       - name: Install Go
@@ -154,6 +189,12 @@ jobs:
     runs-on: windows-2022
     needs: build-windows
     steps:
+      - name: Configure pagefile
+        uses: al-cheb/configure-pagefile-action@v1.3
+        with:
+          minimum-size: 8GB
+          maximum-size: 24GB
+          disk-root: "C:"
       - uses: brechtm/setup-scoop@v2
         with:
           scoop_update: 'false'

Makefile

@@ -30,7 +30,7 @@ GO ?= go
 export GOROOT = $(shell $(GO) env GOROOT)
 
 # Flags to pass to go test.
-GOTESTFLAGS ?= -v
+GOTESTFLAGS ?=
 
 # md5sum binary
 MD5SUM = md5sum
@@ -296,6 +296,7 @@ TEST_PACKAGES_FAST = \
 	encoding \
 	encoding/ascii85 \
 	encoding/base32 \
+	encoding/base64 \
 	encoding/csv \
 	encoding/hex \
 	go/scanner \
@@ -331,26 +332,34 @@ TEST_PACKAGES_FAST += crypto/elliptic/internal/fiat
 endif
 
 # archive/zip requires os.ReadAt, which is not yet supported on windows
+# bytes requires mmap
 # compress/flate appears to hang on wasi
 # compress/lzw appears to hang on wasi
 # crypto/hmac fails on wasi, it exits with a "slice out of range" panic
 # debug/plan9obj requires os.ReadAt, which is not yet supported on windows
+# image requires recover(), which is not  yet supported on wasi
 # io/ioutil requires os.ReadDir, which is not yet supported on windows or wasi
+# mime/quotedprintable requires syscall.Faccessat
 # strconv requires recover() which is not yet supported on wasi
+# text/tabwriter requries recover(), which is not  yet supported on wasi
 # text/template/parse requires recover(), which is not yet supported on wasi
 # testing/fstest requires os.ReadDir, which is not yet supported on windows or wasi
 
 # Additional standard library packages that pass tests on individual platforms
 TEST_PACKAGES_LINUX := \
 	archive/zip \
+	bytes \
 	compress/flate \
 	compress/lzw \
 	crypto/hmac \
 	debug/dwarf \
 	debug/plan9obj \
+	image \
 	io/ioutil \
+	mime/quotedprintable \
 	strconv \
 	testing/fstest \
+	text/tabwriter \
 	text/template/parse
 
 TEST_PACKAGES_DARWIN := $(TEST_PACKAGES_LINUX)
@@ -465,6 +474,8 @@ smoketest:
 	@$(MD5SUM) test.hex
 	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/pininterrupt
 	@$(MD5SUM) test.hex
+	$(TINYGO) build -size short -o test.hex -target=nano-rp2040         examples/rtcinterrupt
+	@$(MD5SUM) test.hex
 	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/serial
 	@$(MD5SUM) test.hex
 	$(TINYGO) build -size short -o test.hex -target=pca10040            examples/systick

README.md

@@ -41,7 +41,7 @@ See the [getting started instructions](https://tinygo.org/getting-started/) for
 
 ## Supported boards/targets
 
-You can compile TinyGo programs for microcontrollers, WebAssembly and Linux.
+You can compile TinyGo programs for microcontrollers, WebAssembly, Windows, and Linux.
 
 The following 94 microcontroller boards are currently supported:
 

builder/build.go

@@ -169,6 +169,7 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 		CodeModel:       config.CodeModel(),
 		RelocationModel: config.RelocationModel(),
 		SizeLevel:       sizeLevel,
+		TinyGoVersion:   goenv.Version,
 
 		Scheduler:          config.Scheduler(),
 		AutomaticStackSize: config.AutomaticStackSize(),
@@ -190,6 +191,9 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 	lprogram, err := loader.Load(config, pkgName, config.ClangHeaders, types.Config{
 		Sizes: compiler.Sizes(machine),
 	})
+	if err != nil {
+		return BuildResult{}, err
+	}
 	result := BuildResult{
 		ModuleRoot: lprogram.MainPkg().Module.Dir,
 		MainDir:    lprogram.MainPkg().Dir,
@@ -199,9 +203,6 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 		// If there is no module root, just the regular root.
 		result.ModuleRoot = lprogram.MainPkg().Root
 	}
-	if err != nil { // failed to load AST
-		return result, err
-	}
 	err = lprogram.Parse()
 	if err != nil {
 		return result, err
@@ -305,7 +306,6 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 				actionID := packageAction{
 					ImportPath:       pkg.ImportPath,
 					CompilerBuildID:  string(compilerBuildID),
-					TinyGoVersion:    goenv.Version,
 					LLVMVersion:      llvm.Version,
 					Config:           compilerConfig,
 					CFlags:           pkg.CFlags,
@@ -594,12 +594,7 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 			defer llvmBuf.Dispose()
 			return result, os.WriteFile(outpath, llvmBuf.Bytes(), 0666)
 		case ".bc":
-			var buf llvm.MemoryBuffer
-			if config.UseThinLTO() {
-				buf = llvm.WriteThinLTOBitcodeToMemoryBuffer(mod)
-			} else {
-				buf = llvm.WriteBitcodeToMemoryBuffer(mod)
-			}
+			buf := llvm.WriteThinLTOBitcodeToMemoryBuffer(mod)
 			defer buf.Dispose()
 			return result, os.WriteFile(outpath, buf.Bytes(), 0666)
 		case ".ll":
@@ -621,16 +616,7 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 		dependencies: []*compileJob{programJob},
 		result:       objfile,
 		run: func(*compileJob) error {
-			var llvmBuf llvm.MemoryBuffer
-			if config.UseThinLTO() {
-				llvmBuf = llvm.WriteThinLTOBitcodeToMemoryBuffer(mod)
-			} else {
-				var err error
-				llvmBuf, err = machine.EmitToMemoryBuffer(mod, llvm.ObjectFile)
-				if err != nil {
-					return err
-				}
-			}
+			llvmBuf := llvm.WriteThinLTOBitcodeToMemoryBuffer(mod)
 			defer llvmBuf.Dispose()
 			return os.WriteFile(objfile, llvmBuf.Bytes(), 0666)
 		},
@@ -664,7 +650,7 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 		job := &compileJob{
 			description: "compile extra file " + path,
 			run: func(job *compileJob) error {
-				result, err := compileAndCacheCFile(abspath, tmpdir, config.CFlags(), config.UseThinLTO(), config.Options.PrintCommands)
+				result, err := compileAndCacheCFile(abspath, tmpdir, config.CFlags(), config.Options.PrintCommands)
 				job.result = result
 				return err
 			},
@@ -682,7 +668,7 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 			job := &compileJob{
 				description: "compile CGo file " + abspath,
 				run: func(job *compileJob) error {
-					result, err := compileAndCacheCFile(abspath, tmpdir, pkg.CFlags, config.UseThinLTO(), config.Options.PrintCommands)
+					result, err := compileAndCacheCFile(abspath, tmpdir, pkg.CFlags, config.Options.PrintCommands)
 					job.result = result
 					return err
 				},
@@ -741,36 +727,34 @@ func Build(pkgName, outpath, tmpdir string, config *compileopts.Config) (BuildRe
 				}
 				ldflags = append(ldflags, dependency.result)
 			}
-			if config.UseThinLTO() {
-				ldflags = append(ldflags, "-mllvm", "-mcpu="+config.CPU())
-				if config.GOOS() == "windows" {
-					// Options for the MinGW wrapper for the lld COFF linker.
-					ldflags = append(ldflags,
-						"-Xlink=/opt:lldlto="+strconv.Itoa(optLevel),
-						"--thinlto-cache-dir="+filepath.Join(cacheDir, "thinlto"))
-				} else if config.GOOS() == "darwin" {
-					// Options for the ld64-compatible lld linker.
-					ldflags = append(ldflags,
-						"--lto-O"+strconv.Itoa(optLevel),
-						"-cache_path_lto", filepath.Join(cacheDir, "thinlto"))
-				} else {
-					// Options for the ELF linker.
-					ldflags = append(ldflags,
-						"--lto-O"+strconv.Itoa(optLevel),
-						"--thinlto-cache-dir="+filepath.Join(cacheDir, "thinlto"),
-					)
-				}
-				if config.CodeModel() != "default" {
-					ldflags = append(ldflags,
-						"-mllvm", "-code-model="+config.CodeModel())
-				}
-				if sizeLevel >= 2 {
-					// Workaround with roughly the same effect as
-					// https://reviews.llvm.org/D119342.
-					// Can hopefully be removed in LLVM 15.
-					ldflags = append(ldflags,
-						"-mllvm", "--rotation-max-header-size=0")
-				}
+			ldflags = append(ldflags, "-mllvm", "-mcpu="+config.CPU())
+			if config.GOOS() == "windows" {
+				// Options for the MinGW wrapper for the lld COFF linker.
+				ldflags = append(ldflags,
+					"-Xlink=/opt:lldlto="+strconv.Itoa(optLevel),
+					"--thinlto-cache-dir="+filepath.Join(cacheDir, "thinlto"))
+			} else if config.GOOS() == "darwin" {
+				// Options for the ld64-compatible lld linker.
+				ldflags = append(ldflags,
+					"--lto-O"+strconv.Itoa(optLevel),
+					"-cache_path_lto", filepath.Join(cacheDir, "thinlto"))
+			} else {
+				// Options for the ELF linker.
+				ldflags = append(ldflags,
+					"--lto-O"+strconv.Itoa(optLevel),
+					"--thinlto-cache-dir="+filepath.Join(cacheDir, "thinlto"),
+				)
+			}
+			if config.CodeModel() != "default" {
+				ldflags = append(ldflags,
+					"-mllvm", "-code-model="+config.CodeModel())
+			}
+			if sizeLevel >= 2 {
+				// Workaround with roughly the same effect as
+				// https://reviews.llvm.org/D119342.
+				// Can hopefully be removed in LLVM 15.
+				ldflags = append(ldflags,
+					"-mllvm", "--rotation-max-header-size=0")
 			}
 			if config.Options.PrintCommands != nil {
 				config.Options.PrintCommands(config.Target.Linker, ldflags...)
@@ -1069,10 +1053,6 @@ func optimizeProgram(mod llvm.Module, config *compileopts.Config) error {
 		}
 	}
 
-	if config.GOOS() != "darwin" && !config.UseThinLTO() {
-		transform.ApplyFunctionSections(mod) // -ffunction-sections
-	}
-
 	// Insert values from -ldflags="-X ..." into the IR.
 	err = setGlobalValues(mod, config.Options.GlobalValues)
 	if err != nil {

builder/cc.go

@@ -56,7 +56,7 @@ import (
 //     depfile but without invalidating its name. For this reason, the depfile is
 //     written on each new compilation (even when it seems unnecessary). However, it
 //     could in rare cases lead to a stale file fetched from the cache.
-func compileAndCacheCFile(abspath, tmpdir string, cflags []string, thinlto bool, printCommands func(string, ...string)) (string, error) {
+func compileAndCacheCFile(abspath, tmpdir string, cflags []string, printCommands func(string, ...string)) (string, error) {
 	// Hash input file.
 	fileHash, err := hashFile(abspath)
 	if err != nil {
@@ -67,11 +67,6 @@ func compileAndCacheCFile(abspath, tmpdir string, cflags []string, thinlto bool,
 	unlock := lock(filepath.Join(goenv.Get("GOCACHE"), fileHash+".c.lock"))
 	defer unlock()
 
-	ext := ".o"
-	if thinlto {
-		ext = ".bc"
-	}
-
 	// Create cache key for the dependencies file.
 	buf, err := json.Marshal(struct {
 		Path        string
@@ -104,7 +99,7 @@ func compileAndCacheCFile(abspath, tmpdir string, cflags []string, thinlto bool,
 		}
 
 		// Obtain hashes of all the files listed as a dependency.
-		outpath, err := makeCFileCachePath(dependencies, depfileNameHash, ext)
+		outpath, err := makeCFileCachePath(dependencies, depfileNameHash)
 		if err == nil {
 			if _, err := os.Stat(outpath); err == nil {
 				return outpath, nil
@@ -117,7 +112,7 @@ func compileAndCacheCFile(abspath, tmpdir string, cflags []string, thinlto bool,
 		return "", err
 	}
 
-	objTmpFile, err := os.CreateTemp(goenv.Get("GOCACHE"), "tmp-*"+ext)
+	objTmpFile, err := os.CreateTemp(goenv.Get("GOCACHE"), "tmp-*.bc")
 	if err != nil {
 		return "", err
 	}
@@ -127,11 +122,8 @@ func compileAndCacheCFile(abspath, tmpdir string, cflags []string, thinlto bool,
 		return "", err
 	}
 	depTmpFile.Close()
-	flags := append([]string{}, cflags...)                                   // copy cflags
-	flags = append(flags, "-MD", "-MV", "-MTdeps", "-MF", depTmpFile.Name()) // autogenerate dependencies
-	if thinlto {
-		flags = append(flags, "-flto=thin")
-	}
+	flags := append([]string{}, cflags...)                                                 // copy cflags
+	flags = append(flags, "-MD", "-MV", "-MTdeps", "-MF", depTmpFile.Name(), "-flto=thin") // autogenerate dependencies
 	flags = append(flags, "-c", "-o", objTmpFile.Name(), abspath)
 	if strings.ToLower(filepath.Ext(abspath)) == ".s" {
 		// If this is an assembly file (.s or .S, lowercase or uppercase), then
@@ -189,7 +181,7 @@ func compileAndCacheCFile(abspath, tmpdir string, cflags []string, thinlto bool,
 	}
 
 	// Move temporary object file to final location.
-	outpath, err := makeCFileCachePath(dependencySlice, depfileNameHash, ext)
+	outpath, err := makeCFileCachePath(dependencySlice, depfileNameHash)
 	if err != nil {
 		return "", err
 	}
@@ -204,7 +196,7 @@ func compileAndCacheCFile(abspath, tmpdir string, cflags []string, thinlto bool,
 // Create a cache path (a path in GOCACHE) to store the output of a compiler
 // job. This path is based on the dep file name (which is a hash of metadata
 // including compiler flags) and the hash of all input files in the paths slice.
-func makeCFileCachePath(paths []string, depfileNameHash, ext string) (string, error) {
+func makeCFileCachePath(paths []string, depfileNameHash string) (string, error) {
 	// Hash all input files.
 	fileHashes := make(map[string]string, len(paths))
 	for _, path := range paths {
@@ -229,7 +221,7 @@ func makeCFileCachePath(paths []string, depfileNameHash, ext string) (string, er
 	outFileNameBuf := sha512.Sum512_224(buf)
 	cacheKey := hex.EncodeToString(outFileNameBuf[:])
 
-	outpath := filepath.Join(goenv.Get("GOCACHE"), "obj-"+cacheKey+ext)
+	outpath := filepath.Join(goenv.Get("GOCACHE"), "obj-"+cacheKey+".bc")
 	return outpath, nil
 }
 

builder/sizes.go

@@ -118,10 +118,6 @@ var (
 	//   pack:   data created when storing a constant in an interface for example
 	//   string: buffer behind strings
 	packageSymbolRegexp = regexp.MustCompile(`\$(alloc|embedfsfiles|embedfsslice|embedslice|pack|string)(\.[0-9]+)?$`)
-
-	// Reflect sidetables. Created by the reflect lowering pass.
-	// See src/reflect/sidetables.go.
-	reflectDataRegexp = regexp.MustCompile(`^reflect\.[a-zA-Z]+Sidetable$`)
 )
 
 // readProgramSizeFromDWARF reads the source location for each line of code and
@@ -375,7 +371,7 @@ func loadProgramSize(path string, packagePathMap map[string]string) (*programSiz
 			if section.Flags&elf.SHF_ALLOC == 0 {
 				continue
 			}
-			if packageSymbolRegexp.MatchString(symbol.Name) || reflectDataRegexp.MatchString(symbol.Name) {
+			if packageSymbolRegexp.MatchString(symbol.Name) {
 				addresses = append(addresses, addressLine{
 					Address:    symbol.Value,
 					Length:     symbol.Size,
@@ -836,9 +832,8 @@ func findPackagePath(path string, packagePathMap map[string]string) string {
 		} else if packageSymbolRegexp.MatchString(path) {
 			// Parse symbol names like main$alloc or runtime$string.
 			packagePath = path[:strings.LastIndex(path, "$")]
-		} else if reflectDataRegexp.MatchString(path) {
-			// Parse symbol names like reflect.structTypesSidetable.
-			packagePath = "Go reflect data"
+		} else if path == "<Go type>" {
+			packagePath = "Go types"
 		} else if path == "<Go interface assert>" {
 			// Interface type assert, generated by the interface lowering pass.
 			packagePath = "Go interface assert"

compileopts/config.go

@@ -75,7 +75,8 @@ func (c *Config) GOARM() string {
 
 // BuildTags returns the complete list of build tags used during this build.
 func (c *Config) BuildTags() []string {
-	tags := append(c.Target.BuildTags, []string{"tinygo", "math_big_pure_go", "gc." + c.GC(), "scheduler." + c.Scheduler(), "serial." + c.Serial()}...)
+	targetTags := filterTags(c.Target.BuildTags, c.Options.Tags)
+	tags := append(targetTags, []string{"tinygo", "math_big_pure_go", "gc." + c.GC(), "scheduler." + c.Scheduler(), "serial." + c.Serial()}...)
 	for i := 1; i <= c.GoMinorVersion; i++ {
 		tags = append(tags, fmt.Sprintf("go1.%d", i))
 	}
@@ -190,14 +191,6 @@ func (c *Config) StackSize() uint64 {
 	return c.Target.DefaultStackSize
 }
 
-// UseThinLTO returns whether ThinLTO should be used for the given target.
-func (c *Config) UseThinLTO() bool {
-	// All architectures support ThinLTO now. However, this code is kept for the
-	// time being in case there are regressions. The non-ThinLTO code support
-	// should be removed when it is proven to work reliably.
-	return true
-}
-
 // RP2040BootPatch returns whether the RP2040 boot patch should be applied that
 // calculates and patches in the checksum for the 2nd stage bootloader.
 func (c *Config) RP2040BootPatch() bool {
@@ -550,3 +543,27 @@ type TestConfig struct {
 	CompileTestBinary bool
 	// TODO: Filter the test functions to run, include verbose flag, etc
 }
+
+// filterTags removes predefined build tags for a target if a conflicting option
+// is provided by the user.
+func filterTags(targetTags []string, userTags []string) []string {
+	var filtered []string
+	for _, t := range targetTags {
+		switch {
+		case strings.HasPrefix(t, "runtime_memhash_"):
+			overridden := false
+			for _, ut := range userTags {
+				if strings.HasPrefix(ut, "runtime_memhash_") {
+					overridden = true
+					break
+				}
+			}
+			if !overridden {
+				filtered = append(filtered, t)
+			}
+		default:
+			filtered = append(filtered, t)
+		}
+	}
+	return filtered
+}

compileopts/config_test.go

@@ -0,0 +1,132 @@
+package compileopts
+
+import (
+	"fmt"
+	"strings"
+	"testing"
+)
+
+func TestBuildTags(t *testing.T) {
+	tests := []struct {
+		targetTags []string
+		userTags   []string
+		result     []string
+	}{
+		{
+			targetTags: []string{},
+			userTags:   []string{},
+			result: []string{
+				"tinygo",
+				"math_big_pure_go",
+				"gc.conservative",
+				"scheduler.none",
+				"serial.none",
+			},
+		},
+		{
+			targetTags: []string{"bear"},
+			userTags:   []string{},
+			result: []string{
+				"bear",
+				"tinygo",
+				"math_big_pure_go",
+				"gc.conservative",
+				"scheduler.none",
+				"serial.none",
+			},
+		},
+		{
+			targetTags: []string{},
+			userTags:   []string{"cat"},
+			result: []string{
+				"tinygo",
+				"math_big_pure_go",
+				"gc.conservative",
+				"scheduler.none",
+				"serial.none",
+				"cat",
+			},
+		},
+		{
+			targetTags: []string{"bear"},
+			userTags:   []string{"cat"},
+			result: []string{
+				"bear",
+				"tinygo",
+				"math_big_pure_go",
+				"gc.conservative",
+				"scheduler.none",
+				"serial.none",
+				"cat",
+			},
+		},
+		{
+			targetTags: []string{"bear", "runtime_memhash_leveldb"},
+			userTags:   []string{"cat"},
+			result: []string{
+				"bear",
+				"runtime_memhash_leveldb",
+				"tinygo",
+				"math_big_pure_go",
+				"gc.conservative",
+				"scheduler.none",
+				"serial.none",
+				"cat",
+			},
+		},
+		{
+			targetTags: []string{"bear", "runtime_memhash_leveldb"},
+			userTags:   []string{"cat", "runtime_memhash_leveldb"},
+			result: []string{
+				"bear",
+				"tinygo",
+				"math_big_pure_go",
+				"gc.conservative",
+				"scheduler.none",
+				"serial.none",
+				"cat",
+				"runtime_memhash_leveldb",
+			},
+		},
+		{
+			targetTags: []string{"bear", "runtime_memhash_leveldb"},
+			userTags:   []string{"cat", "runtime_memhash_sip"},
+			result: []string{
+				"bear",
+				"tinygo",
+				"math_big_pure_go",
+				"gc.conservative",
+				"scheduler.none",
+				"serial.none",
+				"cat",
+				"runtime_memhash_sip",
+			},
+		},
+	}
+
+	for _, tc := range tests {
+		tt := tc
+		t.Run(fmt.Sprintf("%s+%s", strings.Join(tt.targetTags, ","), strings.Join(tt.userTags, ",")), func(t *testing.T) {
+			c := &Config{
+				Target: &TargetSpec{
+					BuildTags: tt.targetTags,
+				},
+				Options: &Options{
+					Tags: tt.userTags,
+				},
+			}
+
+			res := c.BuildTags()
+
+			if len(res) != len(tt.result) {
+				t.Errorf("expected %d tags, got %d", len(tt.result), len(res))
+			}
+
+			for i, tag := range tt.result {
+				if tag != res[i] {
+					t.Errorf("tag %d: expected %s, got %s", i, tt.result[i], tag)
+				}
+			}
+		})
+	}
+}

compileopts/target.go

@@ -326,9 +326,9 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) {
 	}
 	if goarch != "wasm" {
 		suffix := ""
-		if goos == "windows" {
-			// Windows uses a different calling convention from other operating
-			// systems so we need separate assembly files.
+		if goos == "windows" && goarch == "amd64" {
+			// Windows uses a different calling convention on amd64 from other
+			// operating systems so we need separate assembly files.
 			suffix = "_windows"
 		}
 		spec.ExtraFiles = append(spec.ExtraFiles, "src/runtime/asm_"+goarch+suffix+".S")

compiler/compiler.go

@@ -47,6 +47,7 @@ type Config struct {
 	CodeModel       string
 	RelocationModel string
 	SizeLevel       int
+	TinyGoVersion   string // for llvm.ident
 
 	// Various compiler options that determine how code is generated.
 	Scheduler          string
@@ -321,6 +322,14 @@ func CompilePackage(moduleName string, pkg *loader.Package, ssaPkg *ssa.Package,
 				llvm.ConstInt(c.ctx.Int32Type(), 4, false).ConstantAsMetadata(),
 			}),
 		)
+		if c.TinyGoVersion != "" {
+			// It is necessary to set llvm.ident, otherwise debugging on MacOS
+			// won't work.
+			c.mod.AddNamedMetadataOperand("llvm.ident",
+				c.ctx.MDNode(([]llvm.Metadata{
+					c.ctx.MDString("TinyGo version " + c.TinyGoVersion),
+				})))
+		}
 		c.dibuilder.Finalize()
 		c.dibuilder.Destroy()
 	}
@@ -340,12 +349,15 @@ func CompilePackage(moduleName string, pkg *loader.Package, ssaPkg *ssa.Package,
 	return c.mod, c.diagnostics
 }
 
+func (c *compilerContext) getRuntimeType(name string) types.Type {
+	return c.runtimePkg.Scope().Lookup(name).(*types.TypeName).Type()
+}
+
 // getLLVMRuntimeType obtains a named type from the runtime package and returns
 // it as a LLVM type, creating it if necessary. It is a shorthand for
 // getLLVMType(getRuntimeType(name)).
 func (c *compilerContext) getLLVMRuntimeType(name string) llvm.Type {
-	typ := c.runtimePkg.Scope().Lookup(name).(*types.TypeName).Type()
-	return c.getLLVMType(typ)
+	return c.getLLVMType(c.getRuntimeType(name))
 }
 
 // getLLVMType returns a LLVM type for a Go type. It doesn't recreate already

compiler/compiler_test.go

@@ -49,6 +49,7 @@ func TestCompiler(t *testing.T) {
 		{"goroutine.go", "cortex-m-qemu", "tasks"},
 		{"channel.go", "", ""},
 		{"gc.go", "", ""},
+		{"zeromap.go", "", ""},
 	}
 	if goMinor >= 20 {
 		tests = append(tests, testCase{"go1.20.go", "", ""})

compiler/defer.go

@@ -271,7 +271,7 @@ func (b *builder) createDefer(instr *ssa.Defer) {
 		typecode := b.CreateExtractValue(itf, 0, "invoke.func.typecode")
 		receiverValue := b.CreateExtractValue(itf, 1, "invoke.func.receiver")
 		values = []llvm.Value{callback, next, typecode, receiverValue}
-		valueTypes = append(valueTypes, b.uintptrType, b.i8ptrType)
+		valueTypes = append(valueTypes, b.i8ptrType, b.i8ptrType)
 		for _, arg := range instr.Call.Args {
 			val := b.getValue(arg)
 			values = append(values, val)
@@ -476,7 +476,7 @@ func (b *builder) createRunDefers() {
 				valueTypes = append(valueTypes, b.getFuncType(callback.Signature()))
 			} else {
 				//Expect typecode
-				valueTypes = append(valueTypes, b.uintptrType, b.i8ptrType)
+				valueTypes = append(valueTypes, b.i8ptrType, b.i8ptrType)
 			}
 
 			for _, arg := range callback.Args {

compiler/interface.go

@@ -6,6 +6,7 @@ package compiler
 // interface-lowering.go for more details.
 
 import (
+	"fmt"
 	"go/token"
 	"go/types"
 	"strconv"
@@ -15,6 +16,49 @@ import (
 	"tinygo.org/x/go-llvm"
 )
 
+// Type kinds for basic types.
+// They must match the constants for the Kind type in src/reflect/type.go.
+var basicTypes = [...]uint8{
+	types.Bool:          1,
+	types.Int:           2,
+	types.Int8:          3,
+	types.Int16:         4,
+	types.Int32:         5,
+	types.Int64:         6,
+	types.Uint:          7,
+	types.Uint8:         8,
+	types.Uint16:        9,
+	types.Uint32:        10,
+	types.Uint64:        11,
+	types.Uintptr:       12,
+	types.Float32:       13,
+	types.Float64:       14,
+	types.Complex64:     15,
+	types.Complex128:    16,
+	types.String:        17,
+	types.UnsafePointer: 18,
+}
+
+// These must also match the constants for the Kind type in src/reflect/type.go.
+const (
+	typeKindChan      = 19
+	typeKindInterface = 20
+	typeKindPointer   = 21
+	typeKindSlice     = 22
+	typeKindArray     = 23
+	typeKindSignature = 24
+	typeKindMap       = 25
+	typeKindStruct    = 26
+)
+
+// Flags stored in the first byte of the struct field byte array. Must be kept
+// up to date with src/reflect/type.go.
+const (
+	structFieldFlagAnonymous = 1 << iota
+	structFieldFlagHasTag
+	structFieldFlagIsExported
+)
+
 // createMakeInterface emits the LLVM IR for the *ssa.MakeInterface instruction.
 // It tries to put the type in the interface value, but if that's not possible,
 // it will do an allocation of the right size and put that in the interface
@@ -23,10 +67,9 @@ import (
 // An interface value is a {typecode, value} tuple named runtime._interface.
 func (b *builder) createMakeInterface(val llvm.Value, typ types.Type, pos token.Pos) llvm.Value {
 	itfValue := b.emitPointerPack([]llvm.Value{val})
-	itfTypeCodeGlobal := b.getTypeCode(typ)
-	itfTypeCode := b.CreatePtrToInt(itfTypeCodeGlobal, b.uintptrType, "")
+	itfType := b.getTypeCode(typ)
 	itf := llvm.Undef(b.getLLVMRuntimeType("_interface"))
-	itf = b.CreateInsertValue(itf, itfTypeCode, 0, "")
+	itf = b.CreateInsertValue(itf, itfType, 0, "")
 	itf = b.CreateInsertValue(itf, itfValue, 1, "")
 	return itf
 }
@@ -41,118 +84,240 @@ func (b *builder) extractValueFromInterface(itf llvm.Value, llvmType llvm.Type)
 }
 
 // getTypeCode returns a reference to a type code.
-// It returns a pointer to an external global which should be replaced with the
-// real type in the interface lowering pass.
+// A type code is a pointer to a constant global that describes the type.
+// This function returns a pointer to the 'kind' field (which might not be the
+// first field in the struct).
 func (c *compilerContext) getTypeCode(typ types.Type) llvm.Value {
+	ms := c.program.MethodSets.MethodSet(typ)
+	hasMethodSet := ms.Len() != 0
+	if _, ok := typ.Underlying().(*types.Interface); ok {
+		hasMethodSet = false
+	}
 	globalName := "reflect/types.type:" + getTypeCodeName(typ)
 	global := c.mod.NamedGlobal(globalName)
 	if global.IsNil() {
-		// Create a new typecode global.
-		global = llvm.AddGlobal(c.mod, c.getLLVMRuntimeType("typecodeID"), globalName)
-		// Some type classes contain more information for underlying types or
-		// element types. Store it directly in the typecode global to make
-		// reflect lowering simpler.
-		var references llvm.Value
-		var length int64
-		var methodSet llvm.Value
-		var ptrTo llvm.Value
-		var typeAssert llvm.Value
+		var typeFields []llvm.Value
+		// Define the type fields. These must match the structs in
+		// src/reflect/type.go (ptrType, arrayType, etc). See the comment at the
+		// top of src/reflect/type.go for more information on the layout of these structs.
+		typeFieldTypes := []*types.Var{
+			types.NewVar(token.NoPos, nil, "kind", types.Typ[types.Int8]),
+		}
 		switch typ := typ.(type) {
+		case *types.Basic:
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "ptrTo", types.Typ[types.UnsafePointer]),
+			)
 		case *types.Named:
-			references = c.getTypeCode(typ.Underlying())
-		case *types.Chan:
-			references = c.getTypeCode(typ.Elem())
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "ptrTo", types.Typ[types.UnsafePointer]),
+				types.NewVar(token.NoPos, nil, "underlying", types.Typ[types.UnsafePointer]),
+			)
+		case *types.Chan, *types.Slice:
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "ptrTo", types.Typ[types.UnsafePointer]),
+				types.NewVar(token.NoPos, nil, "elementType", types.Typ[types.UnsafePointer]),
+			)
 		case *types.Pointer:
-			references = c.getTypeCode(typ.Elem())
-		case *types.Slice:
-			references = c.getTypeCode(typ.Elem())
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "elementType", types.Typ[types.UnsafePointer]),
+			)
 		case *types.Array:
-			references = c.getTypeCode(typ.Elem())
-			length = typ.Len()
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "ptrTo", types.Typ[types.UnsafePointer]),
+				types.NewVar(token.NoPos, nil, "elementType", types.Typ[types.UnsafePointer]),
+				types.NewVar(token.NoPos, nil, "length", types.Typ[types.Uintptr]),
+			)
+		case *types.Map:
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "ptrTo", types.Typ[types.UnsafePointer]),
+				types.NewVar(token.NoPos, nil, "elementType", types.Typ[types.UnsafePointer]),
+				types.NewVar(token.NoPos, nil, "keyType", types.Typ[types.UnsafePointer]),
+			)
 		case *types.Struct:
-			// Take a pointer to the typecodeID of the first field (if it exists).
-			structGlobal := c.makeStructTypeFields(typ)
-			references = llvm.ConstBitCast(structGlobal, global.Type())
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "numFields", types.Typ[types.Uint16]),
+				types.NewVar(token.NoPos, nil, "ptrTo", types.Typ[types.UnsafePointer]),
+				types.NewVar(token.NoPos, nil, "fields", types.NewArray(c.getRuntimeType("structField"), int64(typ.NumFields()))),
+			)
 		case *types.Interface:
-			methodSetGlobal := c.getInterfaceMethodSet(typ)
-			references = llvm.ConstBitCast(methodSetGlobal, global.Type())
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "ptrTo", types.Typ[types.UnsafePointer]),
+			)
+			// TODO: methods
+		case *types.Signature:
+			typeFieldTypes = append(typeFieldTypes,
+				types.NewVar(token.NoPos, nil, "ptrTo", types.Typ[types.UnsafePointer]),
+			)
+			// TODO: signature params and return values
 		}
-		if _, ok := typ.Underlying().(*types.Interface); !ok {
-			methodSet = c.getTypeMethodSet(typ)
-		} else {
-			typeAssert = c.getInterfaceImplementsFunc(typ)
-			typeAssert = llvm.ConstPtrToInt(typeAssert, c.uintptrType)
+		if hasMethodSet {
+			// This method set is appended at the start of the struct. It is
+			// removed in the interface lowering pass.
+			// TODO: don't remove these and instead do what upstream Go is doing
+			// instead. See: https://research.swtch.com/interfaces. This can
+			// likely be optimized in LLVM using
+			// https://llvm.org/docs/TypeMetadata.html.
+			typeFieldTypes = append([]*types.Var{
+				types.NewVar(token.NoPos, nil, "methodSet", types.Typ[types.UnsafePointer]),
+			}, typeFieldTypes...)
 		}
-		if _, ok := typ.Underlying().(*types.Pointer); !ok {
-			ptrTo = c.getTypeCode(types.NewPointer(typ))
+		globalType := types.NewStruct(typeFieldTypes, nil)
+		global = llvm.AddGlobal(c.mod, c.getLLVMType(globalType), globalName)
+		metabyte := getTypeKind(typ)
+		switch typ := typ.(type) {
+		case *types.Basic:
+			typeFields = []llvm.Value{c.getTypeCode(types.NewPointer(typ))}
+		case *types.Named:
+			typeFields = []llvm.Value{
+				c.getTypeCode(types.NewPointer(typ)), // ptrTo
+				c.getTypeCode(typ.Underlying()),      // underlying
+			}
+			metabyte |= 1 << 5 // "named" flag
+		case *types.Chan:
+			typeFields = []llvm.Value{
+				c.getTypeCode(types.NewPointer(typ)), // ptrTo
+				c.getTypeCode(typ.Elem()),            // elementType
+			}
+		case *types.Slice:
+			typeFields = []llvm.Value{
+				c.getTypeCode(types.NewPointer(typ)), // ptrTo
+				c.getTypeCode(typ.Elem()),            // elementType
+			}
+		case *types.Pointer:
+			typeFields = []llvm.Value{c.getTypeCode(typ.Elem())}
+		case *types.Array:
+			typeFields = []llvm.Value{
+				c.getTypeCode(types.NewPointer(typ)),                   // ptrTo
+				c.getTypeCode(typ.Elem()),                              // elementType
+				llvm.ConstInt(c.uintptrType, uint64(typ.Len()), false), // length
+			}
+		case *types.Map:
+			typeFields = []llvm.Value{
+				c.getTypeCode(types.NewPointer(typ)), // ptrTo
+				c.getTypeCode(typ.Elem()),            // elem
+				c.getTypeCode(typ.Key()),             // key
+			}
+		case *types.Struct:
+			typeFields = []llvm.Value{
+				llvm.ConstInt(c.ctx.Int16Type(), uint64(typ.NumFields()), false), // numFields
+				c.getTypeCode(types.NewPointer(typ)),                             // ptrTo
+			}
+			structFieldType := c.getLLVMRuntimeType("structField")
+			var fields []llvm.Value
+			for i := 0; i < typ.NumFields(); i++ {
+				field := typ.Field(i)
+				var flags uint8
+				if field.Anonymous() {
+					flags |= structFieldFlagAnonymous
+				}
+				if typ.Tag(i) != "" {
+					flags |= structFieldFlagHasTag
+				}
+				if token.IsExported(field.Name()) {
+					flags |= structFieldFlagIsExported
+				}
+				data := string(flags) + field.Name() + "\x00"
+				if typ.Tag(i) != "" {
+					if len(typ.Tag(i)) > 0xff {
+						c.addError(field.Pos(), fmt.Sprintf("struct tag is %d bytes which is too long, max is 255", len(typ.Tag(i))))
+					}
+					data += string([]byte{byte(len(typ.Tag(i)))}) + typ.Tag(i)
+				}
+				dataInitializer := c.ctx.ConstString(data, false)
+				dataGlobal := llvm.AddGlobal(c.mod, dataInitializer.Type(), globalName+"."+field.Name())
+				dataGlobal.SetInitializer(dataInitializer)
+				dataGlobal.SetAlignment(1)
+				dataGlobal.SetUnnamedAddr(true)
+				dataGlobal.SetLinkage(llvm.InternalLinkage)
+				dataGlobal.SetGlobalConstant(true)
+				fieldType := c.getTypeCode(field.Type())
+				fields = append(fields, llvm.ConstNamedStruct(structFieldType, []llvm.Value{
+					fieldType,
+					llvm.ConstGEP(dataGlobal.GlobalValueType(), dataGlobal, []llvm.Value{
+						llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+						llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+					}),
+				}))
+			}
+			typeFields = append(typeFields, llvm.ConstArray(structFieldType, fields))
+		case *types.Interface:
+			typeFields = []llvm.Value{c.getTypeCode(types.NewPointer(typ))}
+			// TODO: methods
+		case *types.Signature:
+			typeFields = []llvm.Value{c.getTypeCode(types.NewPointer(typ))}
+			// TODO: params, return values, etc
 		}
-		globalValue := llvm.ConstNull(global.GlobalValueType())
-		if !references.IsNil() {
-			globalValue = c.builder.CreateInsertValue(globalValue, references, 0, "")
-		}
-		if length != 0 {
-			lengthValue := llvm.ConstInt(c.uintptrType, uint64(length), false)
-			globalValue = c.builder.CreateInsertValue(globalValue, lengthValue, 1, "")
-		}
-		if !methodSet.IsNil() {
-			globalValue = c.builder.CreateInsertValue(globalValue, methodSet, 2, "")
-		}
-		if !ptrTo.IsNil() {
-			globalValue = c.builder.CreateInsertValue(globalValue, ptrTo, 3, "")
-		}
-		if !typeAssert.IsNil() {
-			globalValue = c.builder.CreateInsertValue(globalValue, typeAssert, 4, "")
+		// Prepend metadata byte.
+		typeFields = append([]llvm.Value{
+			llvm.ConstInt(c.ctx.Int8Type(), uint64(metabyte), false),
+		}, typeFields...)
+		if hasMethodSet {
+			typeFields = append([]llvm.Value{
+				llvm.ConstBitCast(c.getTypeMethodSet(typ), c.i8ptrType),
+			}, typeFields...)
 		}
+		alignment := c.targetData.TypeAllocSize(c.i8ptrType)
+		globalValue := c.ctx.ConstStruct(typeFields, false)
 		global.SetInitializer(globalValue)
 		global.SetLinkage(llvm.LinkOnceODRLinkage)
 		global.SetGlobalConstant(true)
-	}
-	return global
-}
-
-// makeStructTypeFields creates a new global that stores all type information
-// related to this struct type, and returns the resulting global. This global is
-// actually an array of all the fields in the structs.
-func (c *compilerContext) makeStructTypeFields(typ *types.Struct) llvm.Value {
-	// The global is an array of runtime.structField structs.
-	runtimeStructField := c.getLLVMRuntimeType("structField")
-	structGlobalType := llvm.ArrayType(runtimeStructField, typ.NumFields())
-	structGlobal := llvm.AddGlobal(c.mod, structGlobalType, "reflect/types.structFields")
-	structGlobalValue := llvm.ConstNull(structGlobalType)
-	for i := 0; i < typ.NumFields(); i++ {
-		fieldGlobalValue := llvm.ConstNull(runtimeStructField)
-		fieldGlobalValue = c.builder.CreateInsertValue(fieldGlobalValue, c.getTypeCode(typ.Field(i).Type()), 0, "")
-		fieldNameType, fieldName := c.makeGlobalArray([]byte(typ.Field(i).Name()), "reflect/types.structFieldName", c.ctx.Int8Type())
-		fieldName.SetLinkage(llvm.PrivateLinkage)
-		fieldName.SetUnnamedAddr(true)
-		fieldName = llvm.ConstGEP(fieldNameType, fieldName, []llvm.Value{
-			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
-			llvm.ConstInt(c.ctx.Int32Type(), 0, false),
-		})
-		fieldGlobalValue = c.builder.CreateInsertValue(fieldGlobalValue, fieldName, 1, "")
-		if typ.Tag(i) != "" {
-			fieldTagType, fieldTag := c.makeGlobalArray([]byte(typ.Tag(i)), "reflect/types.structFieldTag", c.ctx.Int8Type())
-			fieldTag.SetLinkage(llvm.PrivateLinkage)
-			fieldTag.SetUnnamedAddr(true)
-			fieldTag = llvm.ConstGEP(fieldTagType, fieldTag, []llvm.Value{
-				llvm.ConstInt(c.ctx.Int32Type(), 0, false),
-				llvm.ConstInt(c.ctx.Int32Type(), 0, false),
+		global.SetAlignment(int(alignment))
+		if c.Debug {
+			file := c.getDIFile("<Go type>")
+			diglobal := c.dibuilder.CreateGlobalVariableExpression(file, llvm.DIGlobalVariableExpression{
+				Name:        "type " + typ.String(),
+				File:        file,
+				Line:        1,
+				Type:        c.getDIType(globalType),
+				LocalToUnit: false,
+				Expr:        c.dibuilder.CreateExpression(nil),
+				AlignInBits: uint32(alignment * 8),
 			})
-			fieldGlobalValue = c.builder.CreateInsertValue(fieldGlobalValue, fieldTag, 2, "")
+			global.AddMetadata(0, diglobal)
 		}
-		if typ.Field(i).Embedded() {
-			fieldEmbedded := llvm.ConstInt(c.ctx.Int1Type(), 1, false)
-			fieldGlobalValue = c.builder.CreateInsertValue(fieldGlobalValue, fieldEmbedded, 3, "")
-		}
-		structGlobalValue = c.builder.CreateInsertValue(structGlobalValue, fieldGlobalValue, i, "")
 	}
-	structGlobal.SetInitializer(structGlobalValue)
-	structGlobal.SetUnnamedAddr(true)
-	structGlobal.SetLinkage(llvm.PrivateLinkage)
-	return structGlobal
+	offset := uint64(0)
+	if hasMethodSet {
+		// The pointer to the method set is always the first element of the
+		// global (if there is a method set). However, the pointer we return
+		// should point to the 'kind' field not the method set.
+		offset = 1
+	}
+	return llvm.ConstGEP(global.GlobalValueType(), global, []llvm.Value{
+		llvm.ConstInt(llvm.Int32Type(), 0, false),
+		llvm.ConstInt(llvm.Int32Type(), offset, false),
+	})
 }
 
-var basicTypes = [...]string{
+// getTypeKind returns the type kind for the given type, as defined by
+// reflect.Kind.
+func getTypeKind(t types.Type) uint8 {
+	switch t := t.Underlying().(type) {
+	case *types.Basic:
+		return basicTypes[t.Kind()]
+	case *types.Chan:
+		return typeKindChan
+	case *types.Interface:
+		return typeKindInterface
+	case *types.Pointer:
+		return typeKindPointer
+	case *types.Slice:
+		return typeKindSlice
+	case *types.Array:
+		return typeKindArray
+	case *types.Signature:
+		return typeKindSignature
+	case *types.Map:
+		return typeKindMap
+	case *types.Struct:
+		return typeKindStruct
+	default:
+		panic("unknown type")
+	}
+}
+
+var basicTypeNames = [...]string{
 	types.Bool:          "bool",
 	types.Int:           "int",
 	types.Int8:          "int8",
@@ -183,7 +348,7 @@ func getTypeCodeName(t types.Type) string {
 	case *types.Array:
 		return "array:" + strconv.FormatInt(t.Len(), 10) + ":" + getTypeCodeName(t.Elem())
 	case *types.Basic:
-		return "basic:" + basicTypes[t.Kind()]
+		return "basic:" + basicTypeNames[t.Kind()]
 	case *types.Chan:
 		return "chan:" + getTypeCodeName(t.Elem())
 	case *types.Interface:
@@ -235,75 +400,40 @@ func getTypeCodeName(t types.Type) string {
 // getTypeMethodSet returns a reference (GEP) to a global method set. This
 // method set should be unreferenced after the interface lowering pass.
 func (c *compilerContext) getTypeMethodSet(typ types.Type) llvm.Value {
-	global := c.mod.NamedGlobal(typ.String() + "$methodset")
-	zero := llvm.ConstInt(c.ctx.Int32Type(), 0, false)
-	if !global.IsNil() {
-		// the method set already exists
-		return llvm.ConstGEP(global.GlobalValueType(), global, []llvm.Value{zero, zero})
-	}
+	globalName := typ.String() + "$methodset"
+	global := c.mod.NamedGlobal(globalName)
+	if global.IsNil() {
+		ms := c.program.MethodSets.MethodSet(typ)
 
-	ms := c.program.MethodSets.MethodSet(typ)
-	if ms.Len() == 0 {
-		// no methods, so can leave that one out
-		return llvm.ConstPointerNull(llvm.PointerType(c.getLLVMRuntimeType("interfaceMethodInfo"), 0))
-	}
-
-	methods := make([]llvm.Value, ms.Len())
-	interfaceMethodInfoType := c.getLLVMRuntimeType("interfaceMethodInfo")
-	for i := 0; i < ms.Len(); i++ {
-		method := ms.At(i)
-		signatureGlobal := c.getMethodSignature(method.Obj().(*types.Func))
-		fn := c.program.MethodValue(method)
-		llvmFnType, llvmFn := c.getFunction(fn)
-		if llvmFn.IsNil() {
-			// compiler error, so panic
-			panic("cannot find function: " + c.getFunctionInfo(fn).linkName)
+		// Create method set.
+		var signatures, wrappers []llvm.Value
+		for i := 0; i < ms.Len(); i++ {
+			method := ms.At(i)
+			signatureGlobal := c.getMethodSignature(method.Obj().(*types.Func))
+			signatures = append(signatures, signatureGlobal)
+			fn := c.program.MethodValue(method)
+			llvmFnType, llvmFn := c.getFunction(fn)
+			if llvmFn.IsNil() {
+				// compiler error, so panic
+				panic("cannot find function: " + c.getFunctionInfo(fn).linkName)
+			}
+			wrapper := c.getInterfaceInvokeWrapper(fn, llvmFnType, llvmFn)
+			wrappers = append(wrappers, wrapper)
 		}
-		wrapper := c.getInterfaceInvokeWrapper(fn, llvmFnType, llvmFn)
-		methodInfo := llvm.ConstNamedStruct(interfaceMethodInfoType, []llvm.Value{
-			signatureGlobal,
-			llvm.ConstPtrToInt(wrapper, c.uintptrType),
-		})
-		methods[i] = methodInfo
-	}
-	arrayType := llvm.ArrayType(interfaceMethodInfoType, len(methods))
-	value := llvm.ConstArray(interfaceMethodInfoType, methods)
-	global = llvm.AddGlobal(c.mod, arrayType, typ.String()+"$methodset")
-	global.SetInitializer(value)
-	global.SetGlobalConstant(true)
-	global.SetLinkage(llvm.LinkOnceODRLinkage)
-	return llvm.ConstGEP(arrayType, global, []llvm.Value{zero, zero})
-}
 
-// getInterfaceMethodSet returns a global variable with the method set of the
-// given named interface type. This method set is used by the interface lowering
-// pass.
-func (c *compilerContext) getInterfaceMethodSet(typ types.Type) llvm.Value {
-	name := typ.String()
-	if _, ok := typ.(*types.Named); !ok {
-		// Anonymous interface.
-		name = "reflect/types.interface:" + name
+		// Construct global value.
+		globalValue := c.ctx.ConstStruct([]llvm.Value{
+			llvm.ConstInt(c.uintptrType, uint64(ms.Len()), false),
+			llvm.ConstArray(c.i8ptrType, signatures),
+			c.ctx.ConstStruct(wrappers, false),
+		}, false)
+		global = llvm.AddGlobal(c.mod, globalValue.Type(), globalName)
+		global.SetInitializer(globalValue)
+		global.SetGlobalConstant(true)
+		global.SetUnnamedAddr(true)
+		global.SetLinkage(llvm.LinkOnceODRLinkage)
 	}
-	global := c.mod.NamedGlobal(name + "$interface")
-	zero := llvm.ConstInt(c.ctx.Int32Type(), 0, false)
-	if !global.IsNil() {
-		// method set already exist, return it
-		return llvm.ConstGEP(global.GlobalValueType(), global, []llvm.Value{zero, zero})
-	}
-
-	// Every method is a *i8 reference indicating the signature of this method.
-	methods := make([]llvm.Value, typ.Underlying().(*types.Interface).NumMethods())
-	for i := range methods {
-		method := typ.Underlying().(*types.Interface).Method(i)
-		methods[i] = c.getMethodSignature(method)
-	}
-
-	value := llvm.ConstArray(c.i8ptrType, methods)
-	global = llvm.AddGlobal(c.mod, value.Type(), name+"$interface")
-	global.SetInitializer(value)
-	global.SetGlobalConstant(true)
-	global.SetLinkage(llvm.LinkOnceODRLinkage)
-	return llvm.ConstGEP(value.Type(), global, []llvm.Value{zero, zero})
+	return global
 }
 
 // getMethodSignatureName returns a unique name (that can be used as the name of
@@ -443,7 +573,7 @@ func (c *compilerContext) getInterfaceImplementsFunc(assertedType types.Type) ll
 	fnName := getTypeCodeName(assertedType.Underlying()) + ".$typeassert"
 	llvmFn := c.mod.NamedFunction(fnName)
 	if llvmFn.IsNil() {
-		llvmFnType := llvm.FunctionType(c.ctx.Int1Type(), []llvm.Type{c.uintptrType}, false)
+		llvmFnType := llvm.FunctionType(c.ctx.Int1Type(), []llvm.Type{c.i8ptrType}, false)
 		llvmFn = llvm.AddFunction(c.mod, fnName, llvmFnType)
 		c.addStandardDeclaredAttributes(llvmFn)
 		methods := c.getMethodsString(assertedType.Underlying().(*types.Interface))
@@ -464,7 +594,7 @@ func (c *compilerContext) getInvokeFunction(instr *ssa.CallCommon) llvm.Value {
 		for i := 0; i < sig.Params().Len(); i++ {
 			paramTuple = append(paramTuple, sig.Params().At(i))
 		}
-		paramTuple = append(paramTuple, types.NewVar(token.NoPos, nil, "$typecode", types.Typ[types.Uintptr]))
+		paramTuple = append(paramTuple, types.NewVar(token.NoPos, nil, "$typecode", types.Typ[types.UnsafePointer]))
 		llvmFnType := c.getRawFuncType(types.NewSignature(sig.Recv(), types.NewTuple(paramTuple...), sig.Results(), false))
 		llvmFn = llvm.AddFunction(c.mod, fnName, llvmFnType)
 		c.addStandardDeclaredAttributes(llvmFn)
@@ -601,7 +731,7 @@ func typestring(t types.Type) string {
 	case *types.Array:
 		return "[" + strconv.FormatInt(t.Len(), 10) + "]" + typestring(t.Elem())
 	case *types.Basic:
-		return basicTypes[t.Kind()]
+		return basicTypeNames[t.Kind()]
 	case *types.Chan:
 		switch t.Dir() {
 		case types.SendRecv:

compiler/intrinsics.go

@@ -24,6 +24,8 @@ func (b *builder) defineIntrinsicFunction() {
 		b.createMemoryCopyImpl()
 	case name == "runtime.memzero":
 		b.createMemoryZeroImpl()
+	case name == "runtime.KeepAlive":
+		b.createKeepAliveImpl()
 	case strings.HasPrefix(name, "runtime/volatile.Load"):
 		b.createVolatileLoad()
 	case strings.HasPrefix(name, "runtime/volatile.Store"):
@@ -87,6 +89,29 @@ func (b *builder) createMemoryZeroImpl() {
 	b.CreateRetVoid()
 }
 
+// createKeepAlive creates the runtime.KeepAlive function. It is implemented
+// using inline assembly.
+func (b *builder) createKeepAliveImpl() {
+	b.createFunctionStart(true)
+
+	// Get the underlying value of the interface value.
+	interfaceValue := b.getValue(b.fn.Params[0])
+	pointerValue := b.CreateExtractValue(interfaceValue, 1, "")
+
+	// Create an equivalent of the following C code, which is basically just a
+	// nop but ensures the pointerValue is kept alive:
+	//
+	//     __asm__ __volatile__("" : : "r"(pointerValue))
+	//
+	// It should be portable to basically everything as the "r" register type
+	// exists basically everywhere.
+	asmType := llvm.FunctionType(b.ctx.VoidType(), []llvm.Type{b.i8ptrType}, false)
+	asmFn := llvm.InlineAsm(asmType, "", "r", true, false, 0, false)
+	b.createCall(asmType, asmFn, []llvm.Value{pointerValue}, "")
+
+	b.CreateRetVoid()
+}
+
 var mathToLLVMMapping = map[string]string{
 	"math.Ceil":  "llvm.ceil.f64",
 	"math.Exp":   "llvm.exp.f64",

compiler/map.go

@@ -89,6 +89,7 @@ func (b *builder) createMapLookup(keyType, valueType types.Type, m, key llvm.Val
 		// growth.
 		mapKeyAlloca, mapKeyPtr, mapKeySize := b.createTemporaryAlloca(key.Type(), "hashmap.key")
 		b.CreateStore(key, mapKeyAlloca)
+		b.zeroUndefBytes(b.getLLVMType(keyType), mapKeyAlloca)
 		// Fetch the value from the hashmap.
 		params := []llvm.Value{m, mapKeyPtr, mapValuePtr, mapValueSize}
 		commaOkValue = b.createRuntimeCall("hashmapBinaryGet", params, "")
@@ -133,6 +134,7 @@ func (b *builder) createMapUpdate(keyType types.Type, m, key, value llvm.Value,
 		// key can be compared with runtime.memequal
 		keyAlloca, keyPtr, keySize := b.createTemporaryAlloca(key.Type(), "hashmap.key")
 		b.CreateStore(key, keyAlloca)
+		b.zeroUndefBytes(b.getLLVMType(keyType), keyAlloca)
 		params := []llvm.Value{m, keyPtr, valuePtr}
 		b.createRuntimeCall("hashmapBinarySet", params, "")
 		b.emitLifetimeEnd(keyPtr, keySize)
@@ -161,6 +163,7 @@ func (b *builder) createMapDelete(keyType types.Type, m, key llvm.Value, pos tok
 	} else if hashmapIsBinaryKey(keyType) {
 		keyAlloca, keyPtr, keySize := b.createTemporaryAlloca(key.Type(), "hashmap.key")
 		b.CreateStore(key, keyAlloca)
+		b.zeroUndefBytes(b.getLLVMType(keyType), keyAlloca)
 		params := []llvm.Value{m, keyPtr}
 		b.createRuntimeCall("hashmapBinaryDelete", params, "")
 		b.emitLifetimeEnd(keyPtr, keySize)
@@ -240,7 +243,8 @@ func (b *builder) createMapIteratorNext(rangeVal ssa.Value, llvmRangeVal, it llv
 }
 
 // Returns true if this key type does not contain strings, interfaces etc., so
-// can be compared with runtime.memequal.
+// can be compared with runtime.memequal.  Note that padding bytes are undef
+// and can alter two "equal" structs being equal when compared with memequal.
 func hashmapIsBinaryKey(keyType types.Type) bool {
 	switch keyType := keyType.(type) {
 	case *types.Basic:
@@ -263,3 +267,76 @@ func hashmapIsBinaryKey(keyType types.Type) bool {
 		return false
 	}
 }
+
+func (b *builder) zeroUndefBytes(llvmType llvm.Type, ptr llvm.Value) error {
+	// We know that hashmapIsBinaryKey is true, so we only have to handle those types that can show up there.
+	// To zero all undefined bytes, we iterate over all the fields in the type.  For each element, compute the
+	// offset of that element.  If it's Basic type, there are no internal padding bytes.  For compound types, we recurse to ensure
+	// we handle nested types.  Next, we determine if there are any padding bytes before the next
+	// element and zero those as well.
+
+	zero := llvm.ConstInt(b.ctx.Int32Type(), 0, false)
+
+	switch llvmType.TypeKind() {
+	case llvm.IntegerTypeKind:
+		// no padding bytes
+		return nil
+	case llvm.PointerTypeKind:
+		// mo padding bytes
+		return nil
+	case llvm.ArrayTypeKind:
+		llvmArrayType := llvmType
+		llvmElemType := llvmType.ElementType()
+
+		for i := 0; i < llvmArrayType.ArrayLength(); i++ {
+			idx := llvm.ConstInt(b.uintptrType, uint64(i), false)
+			elemPtr := b.CreateInBoundsGEP(llvmArrayType, ptr, []llvm.Value{zero, idx}, "")
+
+			// zero any padding bytes in this element
+			b.zeroUndefBytes(llvmElemType, elemPtr)
+		}
+
+	case llvm.StructTypeKind:
+		llvmStructType := llvmType
+		numFields := llvmStructType.StructElementTypesCount()
+		llvmElementTypes := llvmStructType.StructElementTypes()
+
+		for i := 0; i < numFields; i++ {
+			idx := llvm.ConstInt(b.ctx.Int32Type(), uint64(i), false)
+			elemPtr := b.CreateInBoundsGEP(llvmStructType, ptr, []llvm.Value{zero, idx}, "")
+
+			// zero any padding bytes in this field
+			llvmElemType := llvmElementTypes[i]
+			b.zeroUndefBytes(llvmElemType, elemPtr)
+
+			// zero any padding bytes before the next field, if any
+			offset := b.targetData.ElementOffset(llvmStructType, i)
+			storeSize := b.targetData.TypeStoreSize(llvmElemType)
+			fieldEndOffset := offset + storeSize
+
+			var nextOffset uint64
+			if i < numFields-1 {
+				nextOffset = b.targetData.ElementOffset(llvmStructType, i+1)
+			} else {
+				// Last field?  Next offset is the total size of the allcoate struct.
+				nextOffset = b.targetData.TypeAllocSize(llvmStructType)
+			}
+
+			if fieldEndOffset != nextOffset {
+				n := llvm.ConstInt(b.uintptrType, nextOffset-fieldEndOffset, false)
+				llvmStoreSize := llvm.ConstInt(b.uintptrType, storeSize, false)
+				gepPtr := elemPtr
+				if gepPtr.Type() != b.i8ptrType {
+					gepPtr = b.CreateBitCast(gepPtr, b.i8ptrType, "") // LLVM 14
+				}
+				paddingStart := b.CreateInBoundsGEP(b.ctx.Int8Type(), gepPtr, []llvm.Value{llvmStoreSize}, "")
+				if paddingStart.Type() != b.i8ptrType {
+					paddingStart = b.CreateBitCast(paddingStart, b.i8ptrType, "") // LLVM 14
+				}
+				b.createRuntimeCall("memzero", []llvm.Value{paddingStart, n}, "")
+			}
+		}
+	}
+
+	return nil
+}

compiler/testdata/defer-cortex-m-qemu.ll

@@ -4,7 +4,7 @@ target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
 target triple = "thumbv7m-unknown-unknown-eabi"
 
 %runtime.deferFrame = type { ptr, ptr, [0 x ptr], ptr, i1, %runtime._interface }
-%runtime._interface = type { i32, ptr }
+%runtime._interface = type { ptr, ptr }
 %runtime._defer = type { i32, ptr }
 
 declare noalias nonnull ptr @runtime.alloc(i32, ptr, ptr) #0

compiler/testdata/gc.ll

@@ -3,8 +3,7 @@ source_filename = "gc.go"
 target datalayout = "e-m:e-p:32:32-p10:8:8-p20:8:8-i64:64-n32:64-S128-ni:1:10:20"
 target triple = "wasm32-unknown-wasi"
 
-%runtime.typecodeID = type { ptr, i32, ptr, ptr, i32 }
-%runtime._interface = type { i32, ptr }
+%runtime._interface = type { ptr, ptr }
 
 @main.scalar1 = hidden global ptr null, align 4
 @main.scalar2 = hidden global ptr null, align 4
@@ -22,8 +21,8 @@ target triple = "wasm32-unknown-wasi"
 @main.slice3 = hidden global { ptr, i32, i32 } zeroinitializer, align 8
 @"runtime/gc.layout:62-2000000000000001" = linkonce_odr unnamed_addr constant { i32, [8 x i8] } { i32 62, [8 x i8] c"\01\00\00\00\00\00\00 " }
 @"runtime/gc.layout:62-0001" = linkonce_odr unnamed_addr constant { i32, [8 x i8] } { i32 62, [8 x i8] c"\01\00\00\00\00\00\00\00" }
-@"reflect/types.type:basic:complex128" = linkonce_odr constant %runtime.typecodeID { ptr null, i32 0, ptr null, ptr @"reflect/types.type:pointer:basic:complex128", i32 0 }
-@"reflect/types.type:pointer:basic:complex128" = linkonce_odr constant %runtime.typecodeID { ptr @"reflect/types.type:basic:complex128", i32 0, ptr null, ptr null, i32 0 }
+@"reflect/types.type:basic:complex128" = linkonce_odr constant { i8, ptr } { i8 16, ptr @"reflect/types.type:pointer:basic:complex128" }, align 4
+@"reflect/types.type:pointer:basic:complex128" = linkonce_odr constant { i8, ptr } { i8 21, ptr @"reflect/types.type:basic:complex128" }, align 4
 
 declare noalias nonnull ptr @runtime.alloc(i32, ptr, ptr) #0
 
@@ -129,7 +128,8 @@ entry:
   store double %v.r, ptr %0, align 8
   %.repack1 = getelementptr inbounds { double, double }, ptr %0, i32 0, i32 1
   store double %v.i, ptr %.repack1, align 8
-  %1 = insertvalue %runtime._interface { i32 ptrtoint (ptr @"reflect/types.type:basic:complex128" to i32), ptr undef }, ptr %0, 1
+  %1 = insertvalue %runtime._interface { ptr @"reflect/types.type:basic:complex128", ptr undef }, ptr %0, 1
+  call void @runtime.trackPointer(ptr nonnull @"reflect/types.type:basic:complex128", ptr nonnull %stackalloc, ptr undef) #2
   call void @runtime.trackPointer(ptr nonnull %0, ptr nonnull %stackalloc, ptr undef) #2
   ret %runtime._interface %1
 }

compiler/testdata/goroutine-cortex-m-qemu-tasks.ll

@@ -145,34 +145,34 @@ entry:
 declare void @runtime.chanClose(ptr dereferenceable_or_null(32), ptr) #0
 
 ; Function Attrs: nounwind
-define hidden void @main.startInterfaceMethod(i32 %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
+define hidden void @main.startInterfaceMethod(ptr %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
 entry:
   %0 = call ptr @runtime.alloc(i32 16, ptr null, ptr undef) #8
   store ptr %itf.value, ptr %0, align 4
-  %1 = getelementptr inbounds { ptr, %runtime._string, i32 }, ptr %0, i32 0, i32 1
+  %1 = getelementptr inbounds { ptr, %runtime._string, ptr }, ptr %0, i32 0, i32 1
   store ptr @"main$string", ptr %1, align 4
-  %.repack1 = getelementptr inbounds { ptr, %runtime._string, i32 }, ptr %0, i32 0, i32 1, i32 1
+  %.repack1 = getelementptr inbounds { ptr, %runtime._string, ptr }, ptr %0, i32 0, i32 1, i32 1
   store i32 4, ptr %.repack1, align 4
-  %2 = getelementptr inbounds { ptr, %runtime._string, i32 }, ptr %0, i32 0, i32 2
-  store i32 %itf.typecode, ptr %2, align 4
+  %2 = getelementptr inbounds { ptr, %runtime._string, ptr }, ptr %0, i32 0, i32 2
+  store ptr %itf.typecode, ptr %2, align 4
   %stacksize = call i32 @"internal/task.getGoroutineStackSize"(i32 ptrtoint (ptr @"interface:{Print:func:{basic:string}{}}.Print$invoke$gowrapper" to i32), ptr undef) #8
   call void @"internal/task.start"(i32 ptrtoint (ptr @"interface:{Print:func:{basic:string}{}}.Print$invoke$gowrapper" to i32), ptr nonnull %0, i32 %stacksize, ptr undef) #8
   ret void
 }
 
-declare void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(ptr, ptr, i32, i32, ptr) #6
+declare void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(ptr, ptr, i32, ptr, ptr) #6
 
 ; Function Attrs: nounwind
 define linkonce_odr void @"interface:{Print:func:{basic:string}{}}.Print$invoke$gowrapper"(ptr %0) unnamed_addr #7 {
 entry:
   %1 = load ptr, ptr %0, align 4
-  %2 = getelementptr inbounds { ptr, ptr, i32, i32 }, ptr %0, i32 0, i32 1
+  %2 = getelementptr inbounds { ptr, ptr, i32, ptr }, ptr %0, i32 0, i32 1
   %3 = load ptr, ptr %2, align 4
-  %4 = getelementptr inbounds { ptr, ptr, i32, i32 }, ptr %0, i32 0, i32 2
+  %4 = getelementptr inbounds { ptr, ptr, i32, ptr }, ptr %0, i32 0, i32 2
   %5 = load i32, ptr %4, align 4
-  %6 = getelementptr inbounds { ptr, ptr, i32, i32 }, ptr %0, i32 0, i32 3
-  %7 = load i32, ptr %6, align 4
-  call void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(ptr %1, ptr %3, i32 %5, i32 %7, ptr undef) #8
+  %6 = getelementptr inbounds { ptr, ptr, i32, ptr }, ptr %0, i32 0, i32 3
+  %7 = load ptr, ptr %6, align 4
+  call void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(ptr %1, ptr %3, i32 %5, ptr %7, ptr undef) #8
   ret void
 }
 

compiler/testdata/goroutine-wasm-asyncify.ll

@@ -154,35 +154,35 @@ entry:
 declare void @runtime.chanClose(ptr dereferenceable_or_null(32), ptr) #0
 
 ; Function Attrs: nounwind
-define hidden void @main.startInterfaceMethod(i32 %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
+define hidden void @main.startInterfaceMethod(ptr %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
 entry:
   %stackalloc = alloca i8, align 1
   %0 = call ptr @runtime.alloc(i32 16, ptr null, ptr undef) #8
   call void @runtime.trackPointer(ptr nonnull %0, ptr nonnull %stackalloc, ptr undef) #8
   store ptr %itf.value, ptr %0, align 4
-  %1 = getelementptr inbounds { ptr, %runtime._string, i32 }, ptr %0, i32 0, i32 1
+  %1 = getelementptr inbounds { ptr, %runtime._string, ptr }, ptr %0, i32 0, i32 1
   store ptr @"main$string", ptr %1, align 4
-  %.repack1 = getelementptr inbounds { ptr, %runtime._string, i32 }, ptr %0, i32 0, i32 1, i32 1
+  %.repack1 = getelementptr inbounds { ptr, %runtime._string, ptr }, ptr %0, i32 0, i32 1, i32 1
   store i32 4, ptr %.repack1, align 4
-  %2 = getelementptr inbounds { ptr, %runtime._string, i32 }, ptr %0, i32 0, i32 2
-  store i32 %itf.typecode, ptr %2, align 4
+  %2 = getelementptr inbounds { ptr, %runtime._string, ptr }, ptr %0, i32 0, i32 2
+  store ptr %itf.typecode, ptr %2, align 4
   call void @"internal/task.start"(i32 ptrtoint (ptr @"interface:{Print:func:{basic:string}{}}.Print$invoke$gowrapper" to i32), ptr nonnull %0, i32 16384, ptr undef) #8
   ret void
 }
 
-declare void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(ptr, ptr, i32, i32, ptr) #6
+declare void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(ptr, ptr, i32, ptr, ptr) #6
 
 ; Function Attrs: nounwind
 define linkonce_odr void @"interface:{Print:func:{basic:string}{}}.Print$invoke$gowrapper"(ptr %0) unnamed_addr #7 {
 entry:
   %1 = load ptr, ptr %0, align 4
-  %2 = getelementptr inbounds { ptr, ptr, i32, i32 }, ptr %0, i32 0, i32 1
+  %2 = getelementptr inbounds { ptr, ptr, i32, ptr }, ptr %0, i32 0, i32 1
   %3 = load ptr, ptr %2, align 4
-  %4 = getelementptr inbounds { ptr, ptr, i32, i32 }, ptr %0, i32 0, i32 2
+  %4 = getelementptr inbounds { ptr, ptr, i32, ptr }, ptr %0, i32 0, i32 2
   %5 = load i32, ptr %4, align 4
-  %6 = getelementptr inbounds { ptr, ptr, i32, i32 }, ptr %0, i32 0, i32 3
-  %7 = load i32, ptr %6, align 4
-  call void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(ptr %1, ptr %3, i32 %5, i32 %7, ptr undef) #8
+  %6 = getelementptr inbounds { ptr, ptr, i32, ptr }, ptr %0, i32 0, i32 3
+  %7 = load ptr, ptr %6, align 4
+  call void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(ptr %1, ptr %3, i32 %5, ptr %7, ptr undef) #8
   call void @runtime.deadlock(ptr undef) #8
   unreachable
 }

compiler/testdata/interface.ll

@@ -3,22 +3,17 @@ source_filename = "interface.go"
 target datalayout = "e-m:e-p:32:32-p10:8:8-p20:8:8-i64:64-n32:64-S128-ni:1:10:20"
 target triple = "wasm32-unknown-wasi"
 
-%runtime.typecodeID = type { ptr, i32, ptr, ptr, i32 }
-%runtime._interface = type { i32, ptr }
+%runtime._interface = type { ptr, ptr }
 %runtime._string = type { ptr, i32 }
 
-@"reflect/types.type:basic:int" = linkonce_odr constant %runtime.typecodeID { ptr null, i32 0, ptr null, ptr @"reflect/types.type:pointer:basic:int", i32 0 }
-@"reflect/types.type:pointer:basic:int" = linkonce_odr constant %runtime.typecodeID { ptr @"reflect/types.type:basic:int", i32 0, ptr null, ptr null, i32 0 }
-@"reflect/types.type:pointer:named:error" = linkonce_odr constant %runtime.typecodeID { ptr @"reflect/types.type:named:error", i32 0, ptr null, ptr null, i32 0 }
-@"reflect/types.type:named:error" = linkonce_odr constant %runtime.typecodeID { ptr @"reflect/types.type:interface:{Error:func:{}{basic:string}}", i32 0, ptr null, ptr @"reflect/types.type:pointer:named:error", i32 ptrtoint (ptr @"interface:{Error:func:{}{basic:string}}.$typeassert" to i32) }
-@"reflect/types.type:interface:{Error:func:{}{basic:string}}" = linkonce_odr constant %runtime.typecodeID { ptr @"reflect/types.interface:interface{Error() string}$interface", i32 0, ptr null, ptr @"reflect/types.type:pointer:interface:{Error:func:{}{basic:string}}", i32 ptrtoint (ptr @"interface:{Error:func:{}{basic:string}}.$typeassert" to i32) }
-@"reflect/methods.Error() string" = linkonce_odr constant i8 0, align 1
-@"reflect/types.interface:interface{Error() string}$interface" = linkonce_odr constant [1 x ptr] [ptr @"reflect/methods.Error() string"]
-@"reflect/types.type:pointer:interface:{Error:func:{}{basic:string}}" = linkonce_odr constant %runtime.typecodeID { ptr @"reflect/types.type:interface:{Error:func:{}{basic:string}}", i32 0, ptr null, ptr null, i32 0 }
-@"reflect/types.type:pointer:interface:{String:func:{}{basic:string}}" = linkonce_odr constant %runtime.typecodeID { ptr @"reflect/types.type:interface:{String:func:{}{basic:string}}", i32 0, ptr null, ptr null, i32 0 }
-@"reflect/types.type:interface:{String:func:{}{basic:string}}" = linkonce_odr constant %runtime.typecodeID { ptr @"reflect/types.interface:interface{String() string}$interface", i32 0, ptr null, ptr @"reflect/types.type:pointer:interface:{String:func:{}{basic:string}}", i32 ptrtoint (ptr @"interface:{String:func:{}{basic:string}}.$typeassert" to i32) }
-@"reflect/methods.String() string" = linkonce_odr constant i8 0, align 1
-@"reflect/types.interface:interface{String() string}$interface" = linkonce_odr constant [1 x ptr] [ptr @"reflect/methods.String() string"]
+@"reflect/types.type:basic:int" = linkonce_odr constant { i8, ptr } { i8 2, ptr @"reflect/types.type:pointer:basic:int" }, align 4
+@"reflect/types.type:pointer:basic:int" = linkonce_odr constant { i8, ptr } { i8 21, ptr @"reflect/types.type:basic:int" }, align 4
+@"reflect/types.type:pointer:named:error" = linkonce_odr constant { i8, ptr } { i8 21, ptr @"reflect/types.type:named:error" }, align 4
+@"reflect/types.type:named:error" = linkonce_odr constant { i8, ptr, ptr } { i8 52, ptr @"reflect/types.type:pointer:named:error", ptr @"reflect/types.type:interface:{Error:func:{}{basic:string}}" }, align 4
+@"reflect/types.type:interface:{Error:func:{}{basic:string}}" = linkonce_odr constant { i8, ptr } { i8 20, ptr @"reflect/types.type:pointer:interface:{Error:func:{}{basic:string}}" }, align 4
+@"reflect/types.type:pointer:interface:{Error:func:{}{basic:string}}" = linkonce_odr constant { i8, ptr } { i8 21, ptr @"reflect/types.type:interface:{Error:func:{}{basic:string}}" }, align 4
+@"reflect/types.type:pointer:interface:{String:func:{}{basic:string}}" = linkonce_odr constant { i8, ptr } { i8 21, ptr @"reflect/types.type:interface:{String:func:{}{basic:string}}" }, align 4
+@"reflect/types.type:interface:{String:func:{}{basic:string}}" = linkonce_odr constant { i8, ptr } { i8 20, ptr @"reflect/types.type:pointer:interface:{String:func:{}{basic:string}}" }, align 4
 @"reflect/types.typeid:basic:int" = external constant i8
 
 declare noalias nonnull ptr @runtime.alloc(i32, ptr, ptr) #0
@@ -35,42 +30,42 @@ entry:
 define hidden %runtime._interface @main.simpleType(ptr %context) unnamed_addr #1 {
 entry:
   %stackalloc = alloca i8, align 1
+  call void @runtime.trackPointer(ptr nonnull @"reflect/types.type:basic:int", ptr nonnull %stackalloc, ptr undef) #6
   call void @runtime.trackPointer(ptr null, ptr nonnull %stackalloc, ptr undef) #6
-  ret %runtime._interface { i32 ptrtoint (ptr @"reflect/types.type:basic:int" to i32), ptr null }
+  ret %runtime._interface { ptr @"reflect/types.type:basic:int", ptr null }
 }
 
 ; Function Attrs: nounwind
 define hidden %runtime._interface @main.pointerType(ptr %context) unnamed_addr #1 {
 entry:
   %stackalloc = alloca i8, align 1
+  call void @runtime.trackPointer(ptr nonnull @"reflect/types.type:pointer:basic:int", ptr nonnull %stackalloc, ptr undef) #6
   call void @runtime.trackPointer(ptr null, ptr nonnull %stackalloc, ptr undef) #6
-  ret %runtime._interface { i32 ptrtoint (ptr @"reflect/types.type:pointer:basic:int" to i32), ptr null }
+  ret %runtime._interface { ptr @"reflect/types.type:pointer:basic:int", ptr null }
 }
 
 ; Function Attrs: nounwind
 define hidden %runtime._interface @main.interfaceType(ptr %context) unnamed_addr #1 {
 entry:
   %stackalloc = alloca i8, align 1
+  call void @runtime.trackPointer(ptr nonnull @"reflect/types.type:pointer:named:error", ptr nonnull %stackalloc, ptr undef) #6
   call void @runtime.trackPointer(ptr null, ptr nonnull %stackalloc, ptr undef) #6
-  ret %runtime._interface { i32 ptrtoint (ptr @"reflect/types.type:pointer:named:error" to i32), ptr null }
+  ret %runtime._interface { ptr @"reflect/types.type:pointer:named:error", ptr null }
 }
 
-declare i1 @"interface:{Error:func:{}{basic:string}}.$typeassert"(i32) #2
-
 ; Function Attrs: nounwind
 define hidden %runtime._interface @main.anonymousInterfaceType(ptr %context) unnamed_addr #1 {
 entry:
   %stackalloc = alloca i8, align 1
+  call void @runtime.trackPointer(ptr nonnull @"reflect/types.type:pointer:interface:{String:func:{}{basic:string}}", ptr nonnull %stackalloc, ptr undef) #6
   call void @runtime.trackPointer(ptr null, ptr nonnull %stackalloc, ptr undef) #6
-  ret %runtime._interface { i32 ptrtoint (ptr @"reflect/types.type:pointer:interface:{String:func:{}{basic:string}}" to i32), ptr null }
+  ret %runtime._interface { ptr @"reflect/types.type:pointer:interface:{String:func:{}{basic:string}}", ptr null }
 }
 
-declare i1 @"interface:{String:func:{}{basic:string}}.$typeassert"(i32) #3
-
 ; Function Attrs: nounwind
-define hidden i1 @main.isInt(i32 %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
+define hidden i1 @main.isInt(ptr %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
 entry:
-  %typecode = call i1 @runtime.typeAssert(i32 %itf.typecode, ptr nonnull @"reflect/types.typeid:basic:int", ptr undef) #6
+  %typecode = call i1 @runtime.typeAssert(ptr %itf.typecode, ptr nonnull @"reflect/types.typeid:basic:int", ptr undef) #6
   br i1 %typecode, label %typeassert.ok, label %typeassert.next
 
 typeassert.next:                                  ; preds = %typeassert.ok, %entry
@@ -80,12 +75,12 @@ typeassert.ok:                                    ; preds = %entry
   br label %typeassert.next
 }
 
-declare i1 @runtime.typeAssert(i32, ptr dereferenceable_or_null(1), ptr) #0
+declare i1 @runtime.typeAssert(ptr, ptr dereferenceable_or_null(1), ptr) #0
 
 ; Function Attrs: nounwind
-define hidden i1 @main.isError(i32 %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
+define hidden i1 @main.isError(ptr %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
 entry:
-  %0 = call i1 @"interface:{Error:func:{}{basic:string}}.$typeassert"(i32 %itf.typecode) #6
+  %0 = call i1 @"interface:{Error:func:{}{basic:string}}.$typeassert"(ptr %itf.typecode) #6
   br i1 %0, label %typeassert.ok, label %typeassert.next
 
 typeassert.next:                                  ; preds = %typeassert.ok, %entry
@@ -95,10 +90,12 @@ typeassert.ok:                                    ; preds = %entry
   br label %typeassert.next
 }
 
+declare i1 @"interface:{Error:func:{}{basic:string}}.$typeassert"(ptr) #2
+
 ; Function Attrs: nounwind
-define hidden i1 @main.isStringer(i32 %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
+define hidden i1 @main.isStringer(ptr %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
 entry:
-  %0 = call i1 @"interface:{String:func:{}{basic:string}}.$typeassert"(i32 %itf.typecode) #6
+  %0 = call i1 @"interface:{String:func:{}{basic:string}}.$typeassert"(ptr %itf.typecode) #6
   br i1 %0, label %typeassert.ok, label %typeassert.next
 
 typeassert.next:                                  ; preds = %typeassert.ok, %entry
@@ -108,26 +105,28 @@ typeassert.ok:                                    ; preds = %entry
   br label %typeassert.next
 }
 
+declare i1 @"interface:{String:func:{}{basic:string}}.$typeassert"(ptr) #3
+
 ; Function Attrs: nounwind
-define hidden i8 @main.callFooMethod(i32 %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
+define hidden i8 @main.callFooMethod(ptr %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
 entry:
-  %0 = call i8 @"interface:{String:func:{}{basic:string},main.foo:func:{basic:int}{basic:uint8}}.foo$invoke"(ptr %itf.value, i32 3, i32 %itf.typecode, ptr undef) #6
+  %0 = call i8 @"interface:{String:func:{}{basic:string},main.foo:func:{basic:int}{basic:uint8}}.foo$invoke"(ptr %itf.value, i32 3, ptr %itf.typecode, ptr undef) #6
   ret i8 %0
 }
 
-declare i8 @"interface:{String:func:{}{basic:string},main.foo:func:{basic:int}{basic:uint8}}.foo$invoke"(ptr, i32, i32, ptr) #4
+declare i8 @"interface:{String:func:{}{basic:string},main.foo:func:{basic:int}{basic:uint8}}.foo$invoke"(ptr, i32, ptr, ptr) #4
 
 ; Function Attrs: nounwind
-define hidden %runtime._string @main.callErrorMethod(i32 %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
+define hidden %runtime._string @main.callErrorMethod(ptr %itf.typecode, ptr %itf.value, ptr %context) unnamed_addr #1 {
 entry:
   %stackalloc = alloca i8, align 1
-  %0 = call %runtime._string @"interface:{Error:func:{}{basic:string}}.Error$invoke"(ptr %itf.value, i32 %itf.typecode, ptr undef) #6
+  %0 = call %runtime._string @"interface:{Error:func:{}{basic:string}}.Error$invoke"(ptr %itf.value, ptr %itf.typecode, ptr undef) #6
   %1 = extractvalue %runtime._string %0, 0
   call void @runtime.trackPointer(ptr %1, ptr nonnull %stackalloc, ptr undef) #6
   ret %runtime._string %0
 }
 
-declare %runtime._string @"interface:{Error:func:{}{basic:string}}.Error$invoke"(ptr, i32, ptr) #5
+declare %runtime._string @"interface:{Error:func:{}{basic:string}}.Error$invoke"(ptr, ptr, ptr) #5
 
 attributes #0 = { "target-features"="+bulk-memory,+nontrapping-fptoint,+sign-ext" }
 attributes #1 = { nounwind "target-features"="+bulk-memory,+nontrapping-fptoint,+sign-ext" }

compiler/testdata/zeromap.go

@@ -0,0 +1,37 @@
+package main
+
+type hasPadding struct {
+	b1 bool
+	i  int
+	b2 bool
+}
+
+type nestedPadding struct {
+	b bool
+	hasPadding
+	i int
+}
+
+//go:noinline
+func testZeroGet(m map[hasPadding]int, s hasPadding) int {
+	return m[s]
+}
+
+//go:noinline
+func testZeroSet(m map[hasPadding]int, s hasPadding) {
+	m[s] = 5
+}
+
+//go:noinline
+func testZeroArrayGet(m map[[2]hasPadding]int, s [2]hasPadding) int {
+	return m[s]
+}
+
+//go:noinline
+func testZeroArraySet(m map[[2]hasPadding]int, s [2]hasPadding) {
+	m[s] = 5
+}
+
+func main() {
+
+}

compiler/testdata/zeromap.ll

@@ -0,0 +1,170 @@
+; ModuleID = 'zeromap.go'
+source_filename = "zeromap.go"
+target datalayout = "e-m:e-p:32:32-p10:8:8-p20:8:8-i64:64-n32:64-S128-ni:1:10:20"
+target triple = "wasm32-unknown-wasi"
+
+%main.hasPadding = type { i1, i32, i1 }
+
+declare noalias nonnull ptr @runtime.alloc(i32, ptr, ptr) #0
+
+declare void @runtime.trackPointer(ptr nocapture readonly, ptr, ptr) #0
+
+; Function Attrs: nounwind
+define hidden void @main.init(ptr %context) unnamed_addr #1 {
+entry:
+  ret void
+}
+
+; Function Attrs: noinline nounwind
+define hidden i32 @main.testZeroGet(ptr dereferenceable_or_null(40) %m, i1 %s.b1, i32 %s.i, i1 %s.b2, ptr %context) unnamed_addr #2 {
+entry:
+  %hashmap.key = alloca %main.hasPadding, align 8
+  %hashmap.value = alloca i32, align 4
+  %s = alloca %main.hasPadding, align 8
+  %0 = insertvalue %main.hasPadding zeroinitializer, i1 %s.b1, 0
+  %1 = insertvalue %main.hasPadding %0, i32 %s.i, 1
+  %2 = insertvalue %main.hasPadding %1, i1 %s.b2, 2
+  %stackalloc = alloca i8, align 1
+  store %main.hasPadding zeroinitializer, ptr %s, align 8
+  call void @runtime.trackPointer(ptr nonnull %s, ptr nonnull %stackalloc, ptr undef) #4
+  store %main.hasPadding %2, ptr %s, align 8
+  call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %hashmap.value)
+  call void @llvm.lifetime.start.p0(i64 12, ptr nonnull %hashmap.key)
+  store %main.hasPadding %2, ptr %hashmap.key, align 8
+  %3 = getelementptr inbounds i8, ptr %hashmap.key, i32 1
+  call void @runtime.memzero(ptr nonnull %3, i32 3, ptr undef) #4
+  %4 = getelementptr inbounds i8, ptr %hashmap.key, i32 9
+  call void @runtime.memzero(ptr nonnull %4, i32 3, ptr undef) #4
+  %5 = call i1 @runtime.hashmapBinaryGet(ptr %m, ptr nonnull %hashmap.key, ptr nonnull %hashmap.value, i32 4, ptr undef) #4
+  call void @llvm.lifetime.end.p0(i64 12, ptr nonnull %hashmap.key)
+  %6 = load i32, ptr %hashmap.value, align 4
+  call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %hashmap.value)
+  ret i32 %6
+}
+
+; Function Attrs: argmemonly nocallback nofree nosync nounwind willreturn
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #3
+
+declare void @runtime.memzero(ptr, i32, ptr) #0
+
+declare i1 @runtime.hashmapBinaryGet(ptr dereferenceable_or_null(40), ptr, ptr, i32, ptr) #0
+
+; Function Attrs: argmemonly nocallback nofree nosync nounwind willreturn
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #3
+
+; Function Attrs: noinline nounwind
+define hidden void @main.testZeroSet(ptr dereferenceable_or_null(40) %m, i1 %s.b1, i32 %s.i, i1 %s.b2, ptr %context) unnamed_addr #2 {
+entry:
+  %hashmap.key = alloca %main.hasPadding, align 8
+  %hashmap.value = alloca i32, align 4
+  %s = alloca %main.hasPadding, align 8
+  %0 = insertvalue %main.hasPadding zeroinitializer, i1 %s.b1, 0
+  %1 = insertvalue %main.hasPadding %0, i32 %s.i, 1
+  %2 = insertvalue %main.hasPadding %1, i1 %s.b2, 2
+  %stackalloc = alloca i8, align 1
+  store %main.hasPadding zeroinitializer, ptr %s, align 8
+  call void @runtime.trackPointer(ptr nonnull %s, ptr nonnull %stackalloc, ptr undef) #4
+  store %main.hasPadding %2, ptr %s, align 8
+  call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %hashmap.value)
+  store i32 5, ptr %hashmap.value, align 4
+  call void @llvm.lifetime.start.p0(i64 12, ptr nonnull %hashmap.key)
+  store %main.hasPadding %2, ptr %hashmap.key, align 8
+  %3 = getelementptr inbounds i8, ptr %hashmap.key, i32 1
+  call void @runtime.memzero(ptr nonnull %3, i32 3, ptr undef) #4
+  %4 = getelementptr inbounds i8, ptr %hashmap.key, i32 9
+  call void @runtime.memzero(ptr nonnull %4, i32 3, ptr undef) #4
+  call void @runtime.hashmapBinarySet(ptr %m, ptr nonnull %hashmap.key, ptr nonnull %hashmap.value, ptr undef) #4
+  call void @llvm.lifetime.end.p0(i64 12, ptr nonnull %hashmap.key)
+  call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %hashmap.value)
+  ret void
+}
+
+declare void @runtime.hashmapBinarySet(ptr dereferenceable_or_null(40), ptr, ptr, ptr) #0
+
+; Function Attrs: noinline nounwind
+define hidden i32 @main.testZeroArrayGet(ptr dereferenceable_or_null(40) %m, [2 x %main.hasPadding] %s, ptr %context) unnamed_addr #2 {
+entry:
+  %hashmap.key = alloca [2 x %main.hasPadding], align 8
+  %hashmap.value = alloca i32, align 4
+  %s1 = alloca [2 x %main.hasPadding], align 8
+  %stackalloc = alloca i8, align 1
+  store %main.hasPadding zeroinitializer, ptr %s1, align 8
+  %s1.repack2 = getelementptr inbounds [2 x %main.hasPadding], ptr %s1, i32 0, i32 1
+  store %main.hasPadding zeroinitializer, ptr %s1.repack2, align 4
+  call void @runtime.trackPointer(ptr nonnull %s1, ptr nonnull %stackalloc, ptr undef) #4
+  %s.elt = extractvalue [2 x %main.hasPadding] %s, 0
+  store %main.hasPadding %s.elt, ptr %s1, align 8
+  %s1.repack3 = getelementptr inbounds [2 x %main.hasPadding], ptr %s1, i32 0, i32 1
+  %s.elt4 = extractvalue [2 x %main.hasPadding] %s, 1
+  store %main.hasPadding %s.elt4, ptr %s1.repack3, align 4
+  call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %hashmap.value)
+  call void @llvm.lifetime.start.p0(i64 24, ptr nonnull %hashmap.key)
+  %s.elt7 = extractvalue [2 x %main.hasPadding] %s, 0
+  store %main.hasPadding %s.elt7, ptr %hashmap.key, align 8
+  %hashmap.key.repack8 = getelementptr inbounds [2 x %main.hasPadding], ptr %hashmap.key, i32 0, i32 1
+  %s.elt9 = extractvalue [2 x %main.hasPadding] %s, 1
+  store %main.hasPadding %s.elt9, ptr %hashmap.key.repack8, align 4
+  %0 = getelementptr inbounds i8, ptr %hashmap.key, i32 1
+  call void @runtime.memzero(ptr nonnull %0, i32 3, ptr undef) #4
+  %1 = getelementptr inbounds i8, ptr %hashmap.key, i32 9
+  call void @runtime.memzero(ptr nonnull %1, i32 3, ptr undef) #4
+  %2 = getelementptr inbounds i8, ptr %hashmap.key, i32 13
+  call void @runtime.memzero(ptr nonnull %2, i32 3, ptr undef) #4
+  %3 = getelementptr inbounds i8, ptr %hashmap.key, i32 21
+  call void @runtime.memzero(ptr nonnull %3, i32 3, ptr undef) #4
+  %4 = call i1 @runtime.hashmapBinaryGet(ptr %m, ptr nonnull %hashmap.key, ptr nonnull %hashmap.value, i32 4, ptr undef) #4
+  call void @llvm.lifetime.end.p0(i64 24, ptr nonnull %hashmap.key)
+  %5 = load i32, ptr %hashmap.value, align 4
+  call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %hashmap.value)
+  ret i32 %5
+}
+
+; Function Attrs: noinline nounwind
+define hidden void @main.testZeroArraySet(ptr dereferenceable_or_null(40) %m, [2 x %main.hasPadding] %s, ptr %context) unnamed_addr #2 {
+entry:
+  %hashmap.key = alloca [2 x %main.hasPadding], align 8
+  %hashmap.value = alloca i32, align 4
+  %s1 = alloca [2 x %main.hasPadding], align 8
+  %stackalloc = alloca i8, align 1
+  store %main.hasPadding zeroinitializer, ptr %s1, align 8
+  %s1.repack2 = getelementptr inbounds [2 x %main.hasPadding], ptr %s1, i32 0, i32 1
+  store %main.hasPadding zeroinitializer, ptr %s1.repack2, align 4
+  call void @runtime.trackPointer(ptr nonnull %s1, ptr nonnull %stackalloc, ptr undef) #4
+  %s.elt = extractvalue [2 x %main.hasPadding] %s, 0
+  store %main.hasPadding %s.elt, ptr %s1, align 8
+  %s1.repack3 = getelementptr inbounds [2 x %main.hasPadding], ptr %s1, i32 0, i32 1
+  %s.elt4 = extractvalue [2 x %main.hasPadding] %s, 1
+  store %main.hasPadding %s.elt4, ptr %s1.repack3, align 4
+  call void @llvm.lifetime.start.p0(i64 4, ptr nonnull %hashmap.value)
+  store i32 5, ptr %hashmap.value, align 4
+  call void @llvm.lifetime.start.p0(i64 24, ptr nonnull %hashmap.key)
+  %s.elt7 = extractvalue [2 x %main.hasPadding] %s, 0
+  store %main.hasPadding %s.elt7, ptr %hashmap.key, align 8
+  %hashmap.key.repack8 = getelementptr inbounds [2 x %main.hasPadding], ptr %hashmap.key, i32 0, i32 1
+  %s.elt9 = extractvalue [2 x %main.hasPadding] %s, 1
+  store %main.hasPadding %s.elt9, ptr %hashmap.key.repack8, align 4
+  %0 = getelementptr inbounds i8, ptr %hashmap.key, i32 1
+  call void @runtime.memzero(ptr nonnull %0, i32 3, ptr undef) #4
+  %1 = getelementptr inbounds i8, ptr %hashmap.key, i32 9
+  call void @runtime.memzero(ptr nonnull %1, i32 3, ptr undef) #4
+  %2 = getelementptr inbounds i8, ptr %hashmap.key, i32 13
+  call void @runtime.memzero(ptr nonnull %2, i32 3, ptr undef) #4
+  %3 = getelementptr inbounds i8, ptr %hashmap.key, i32 21
+  call void @runtime.memzero(ptr nonnull %3, i32 3, ptr undef) #4
+  call void @runtime.hashmapBinarySet(ptr %m, ptr nonnull %hashmap.key, ptr nonnull %hashmap.value, ptr undef) #4
+  call void @llvm.lifetime.end.p0(i64 24, ptr nonnull %hashmap.key)
+  call void @llvm.lifetime.end.p0(i64 4, ptr nonnull %hashmap.value)
+  ret void
+}
+
+; Function Attrs: nounwind
+define hidden void @main.main(ptr %context) unnamed_addr #1 {
+entry:
+  ret void
+}
+
+attributes #0 = { "target-features"="+bulk-memory,+nontrapping-fptoint,+sign-ext" }
+attributes #1 = { nounwind "target-features"="+bulk-memory,+nontrapping-fptoint,+sign-ext" }
+attributes #2 = { noinline nounwind "target-features"="+bulk-memory,+nontrapping-fptoint,+sign-ext" }
+attributes #3 = { argmemonly nocallback nofree nosync nounwind willreturn }
+attributes #4 = { nounwind }

goenv/version.go

@@ -12,7 +12,7 @@ import (
 
 // Version of TinyGo.
 // Update this value before release of new version of software.
-const Version = "0.27.0"
+const Version = "0.28.0-dev"
 
 var (
 	// This variable is set at build time using -ldflags parameters.

interp/interpreter.go

@@ -238,7 +238,7 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 				// which case this call won't even get to this point but will
 				// already be emitted in initAll.
 				continue
-			case strings.HasPrefix(callFn.name, "runtime.print") || callFn.name == "runtime._panic" || callFn.name == "runtime.hashmapGet" ||
+			case strings.HasPrefix(callFn.name, "runtime.print") || callFn.name == "runtime._panic" || callFn.name == "runtime.hashmapGet" || callFn.name == "runtime.hashmapInterfaceHash" ||
 				callFn.name == "os.runtime_args" || callFn.name == "internal/task.start" || callFn.name == "internal/task.Current":
 				// These functions should be run at runtime. Specifically:
 				//   * Print and panic functions are best emitted directly without
@@ -378,42 +378,6 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 				copy(dstBuf.buf[dst.offset():dst.offset()+nBytes], srcBuf.buf[src.offset():])
 				dstObj.buffer = dstBuf
 				mem.put(dst.index(), dstObj)
-			case callFn.name == "(reflect.rawType).elem":
-				if r.debug {
-					fmt.Fprintln(os.Stderr, indent+"call (reflect.rawType).elem:", operands[1:])
-				}
-				// Extract the type code global from the first parameter.
-				typecodeIDPtrToInt, err := operands[1].toLLVMValue(inst.llvmInst.Operand(0).Type(), &mem)
-				if err != nil {
-					return nil, mem, r.errorAt(inst, err)
-				}
-				typecodeID := typecodeIDPtrToInt.Operand(0)
-
-				// Get the type class.
-				// See also: getClassAndValueFromTypeCode in transform/reflect.go.
-				typecodeName := typecodeID.Name()
-				const prefix = "reflect/types.type:"
-				if !strings.HasPrefix(typecodeName, prefix) {
-					panic("unexpected typecode name: " + typecodeName)
-				}
-				id := typecodeName[len(prefix):]
-				class := id[:strings.IndexByte(id, ':')]
-				value := id[len(class)+1:]
-				if class == "named" {
-					// Get the underlying type.
-					class = value[:strings.IndexByte(value, ':')]
-					value = value[len(class)+1:]
-				}
-
-				// Elem() is only valid for certain type classes.
-				switch class {
-				case "chan", "pointer", "slice", "array":
-					elementType := r.builder.CreateExtractValue(typecodeID.Initializer(), 0, "")
-					uintptrType := r.mod.Context().IntType(int(mem.r.pointerSize) * 8)
-					locals[inst.localIndex] = r.getValue(llvm.ConstPtrToInt(elementType, uintptrType))
-				default:
-					return nil, mem, r.errorAt(inst, fmt.Errorf("(reflect.Type).Elem() called on %s type", class))
-				}
 			case callFn.name == "runtime.typeAssert":
 				// This function must be implemented manually as it is normally
 				// implemented by the interface lowering pass.
@@ -424,15 +388,22 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 				if err != nil {
 					return nil, mem, r.errorAt(inst, err)
 				}
-				actualTypePtrToInt, err := operands[1].toLLVMValue(inst.llvmInst.Operand(0).Type(), &mem)
+				actualType, err := operands[1].toLLVMValue(inst.llvmInst.Operand(0).Type(), &mem)
 				if err != nil {
 					return nil, mem, r.errorAt(inst, err)
 				}
-				if !actualTypePtrToInt.IsAConstantInt().IsNil() && actualTypePtrToInt.ZExtValue() == 0 {
+				if !actualType.IsAConstantInt().IsNil() && actualType.ZExtValue() == 0 {
 					locals[inst.localIndex] = literalValue{uint8(0)}
 					break
 				}
-				actualType := actualTypePtrToInt.Operand(0)
+				// Strip pointer casts (bitcast, getelementptr).
+				for !actualType.IsAConstantExpr().IsNil() {
+					opcode := actualType.Opcode()
+					if opcode != llvm.GetElementPtr && opcode != llvm.BitCast {
+						break
+					}
+					actualType = actualType.Operand(0)
+				}
 				if strings.TrimPrefix(actualType.Name(), "reflect/types.type:") == strings.TrimPrefix(assertedType.Name(), "reflect/types.typeid:") {
 					locals[inst.localIndex] = literalValue{uint8(1)}
 				} else {
@@ -448,11 +419,12 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 				if err != nil {
 					return nil, mem, r.errorAt(inst, err)
 				}
-				methodSetPtr, err := mem.load(typecodePtr.addOffset(r.pointerSize*2), r.pointerSize).asPointer(r)
+				methodSetPtr, err := mem.load(typecodePtr.addOffset(-int64(r.pointerSize)), r.pointerSize).asPointer(r)
 				if err != nil {
 					return nil, mem, r.errorAt(inst, err)
 				}
 				methodSet := mem.get(methodSetPtr.index()).llvmGlobal.Initializer()
+				numMethods := int(r.builder.CreateExtractValue(methodSet, 0, "").ZExtValue())
 				llvmFn := inst.llvmInst.CalledValue()
 				methodSetAttr := llvmFn.GetStringAttributeAtIndex(-1, "tinygo-methods")
 				methodSetString := methodSetAttr.GetStringValue()
@@ -460,9 +432,9 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 				// Make a set of all the methods on the concrete type, for
 				// easier checking in the next step.
 				concreteTypeMethods := map[string]struct{}{}
-				for i := 0; i < methodSet.Type().ArrayLength(); i++ {
-					methodInfo := r.builder.CreateExtractValue(methodSet, i, "")
-					name := r.builder.CreateExtractValue(methodInfo, 0, "").Name()
+				for i := 0; i < numMethods; i++ {
+					methodInfo := r.builder.CreateExtractValue(methodSet, 1, "")
+					name := r.builder.CreateExtractValue(methodInfo, i, "").Name()
 					concreteTypeMethods[name] = struct{}{}
 				}
 
@@ -488,15 +460,16 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 					fmt.Fprintln(os.Stderr, indent+"invoke method:", operands[1:])
 				}
 
-				// Load the type code of the interface value.
-				typecodeIDBitCast, err := operands[len(operands)-2].toLLVMValue(inst.llvmInst.Operand(len(operands)-3).Type(), &mem)
+				// Load the type code and method set of the interface value.
+				typecodePtr, err := operands[len(operands)-2].asPointer(r)
 				if err != nil {
 					return nil, mem, r.errorAt(inst, err)
 				}
-				typecodeID := typecodeIDBitCast.Operand(0).Initializer()
-
-				// Load the method set, which is part of the typecodeID object.
-				methodSet := stripPointerCasts(r.builder.CreateExtractValue(typecodeID, 2, "")).Initializer()
+				methodSetPtr, err := mem.load(typecodePtr.addOffset(-int64(r.pointerSize)), r.pointerSize).asPointer(r)
+				if err != nil {
+					return nil, mem, r.errorAt(inst, err)
+				}
+				methodSet := mem.get(methodSetPtr.index()).llvmGlobal.Initializer()
 
 				// We don't need to load the interface method set.
 
@@ -508,13 +481,14 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 
 				// Iterate through all methods, looking for the one method that
 				// should be returned.
-				numMethods := methodSet.Type().ArrayLength()
+				numMethods := int(r.builder.CreateExtractValue(methodSet, 0, "").ZExtValue())
 				var method llvm.Value
 				for i := 0; i < numMethods; i++ {
-					methodSignatureAgg := r.builder.CreateExtractValue(methodSet, i, "")
-					methodSignature := r.builder.CreateExtractValue(methodSignatureAgg, 0, "")
+					methodSignatureAgg := r.builder.CreateExtractValue(methodSet, 1, "")
+					methodSignature := r.builder.CreateExtractValue(methodSignatureAgg, i, "")
 					if methodSignature == signature {
-						method = r.builder.CreateExtractValue(methodSignatureAgg, 1, "").Operand(0)
+						methodAgg := r.builder.CreateExtractValue(methodSet, 2, "")
+						method = r.builder.CreateExtractValue(methodAgg, i, "")
 					}
 				}
 				if method.IsNil() {
@@ -685,7 +659,7 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 				}
 				continue
 			}
-			ptr = ptr.addOffset(uint32(offset))
+			ptr = ptr.addOffset(int64(offset))
 			locals[inst.localIndex] = ptr
 			if r.debug {
 				fmt.Fprintln(os.Stderr, indent+"gep:", operands, "->", ptr)
@@ -784,7 +758,7 @@ func (r *runner) run(fn *function, params []value, parentMem *memoryView, indent
 				case llvm.Add:
 					// This likely means this is part of a
 					// unsafe.Pointer(uintptr(ptr) + offset) pattern.
-					lhsPtr = lhsPtr.addOffset(uint32(rhs.Uint()))
+					lhsPtr = lhsPtr.addOffset(int64(rhs.Uint()))
 					locals[inst.localIndex] = lhsPtr
 					continue
 				case llvm.Xor:

interp/memory.go

@@ -501,7 +501,7 @@ func (v pointerValue) offset() uint32 {
 // addOffset essentially does a GEP operation (pointer arithmetic): it adds the
 // offset to the pointer. It also checks that the offset doesn't overflow the
 // maximum offset size (which is 4GB).
-func (v pointerValue) addOffset(offset uint32) pointerValue {
+func (v pointerValue) addOffset(offset int64) pointerValue {
 	result := pointerValue{v.pointer + uint64(offset)}
 	if checks && v.index() != result.index() {
 		panic("interp: offset out of range")
@@ -815,7 +815,7 @@ func (v rawValue) rawLLVMValue(mem *memoryView) (llvm.Value, error) {
 					// as a ptrtoint, so that they can be used in certain
 					// optimizations.
 					name := elementType.StructName()
-					if name == "runtime.typecodeID" || name == "runtime.funcValueWithSignature" {
+					if name == "runtime.funcValueWithSignature" {
 						uintptrType := ctx.IntType(int(mem.r.pointerSize) * 8)
 						field = llvm.ConstPtrToInt(field, uintptrType)
 					}

interp/testdata/interface.ll

@@ -1,17 +1,16 @@
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64--linux"
 
-%runtime.typecodeID = type { %runtime.typecodeID*, i64, %runtime.interfaceMethodInfo* }
-%runtime.interfaceMethodInfo = type { i8*, i64 }
-
 @main.v1 = global i1 0
 @main.v2 = global i1 0
-@"reflect/types.type:named:main.foo" = private constant %runtime.typecodeID { %runtime.typecodeID* @"reflect/types.type:basic:int", i64 0, %runtime.interfaceMethodInfo* null }
+@"reflect/types.type:named:main.foo" = private constant { i8, i8*, i8* } { i8 34, i8* getelementptr inbounds ({ i8, i8* }, { i8, i8* }* @"reflect/types.type:pointer:named:main.foo", i32 0, i32 0), i8* getelementptr inbounds ({ i8, i8* }, { i8, i8* }* @"reflect/types.type:basic:int", i32 0, i32 0) }, align 4
+@"reflect/types.type:pointer:named:main.foo" = external constant { i8, i8* }
 @"reflect/types.typeid:named:main.foo" = external constant i8
-@"reflect/types.type:basic:int" = external constant %runtime.typecodeID
+@"reflect/types.type:basic:int" = private constant { i8, i8* } { i8 2, i8* getelementptr inbounds ({ i8, i8* }, { i8, i8* }* @"reflect/types.type:pointer:basic:int", i32 0, i32 0) }, align 4
+@"reflect/types.type:pointer:basic:int" = external constant { i8, i8* }
 
 
-declare i1 @runtime.typeAssert(i64, i8*, i8*, i8*)
+declare i1 @runtime.typeAssert(i8*, i8*, i8*, i8*)
 
 define void @runtime.initAll() unnamed_addr {
 entry:
@@ -22,9 +21,9 @@ entry:
 define internal void @main.init() unnamed_addr {
 entry:
   ; Test type asserts.
-  %typecode = call i1 @runtime.typeAssert(i64 ptrtoint (%runtime.typecodeID* @"reflect/types.type:named:main.foo" to i64), i8* @"reflect/types.typeid:named:main.foo", i8* undef, i8* null)
+  %typecode = call i1 @runtime.typeAssert(i8* getelementptr inbounds ({ i8, i8*, i8* }, { i8, i8*, i8* }* @"reflect/types.type:named:main.foo", i32 0, i32 0), i8* @"reflect/types.typeid:named:main.foo", i8* undef, i8* null)
   store i1 %typecode, i1* @main.v1
-  %typecode2 = call i1 @runtime.typeAssert(i64 0, i8* @"reflect/types.typeid:named:main.foo", i8* undef, i8* null)
+  %typecode2 = call i1 @runtime.typeAssert(i8* null, i8* @"reflect/types.typeid:named:main.foo", i8* undef, i8* null)
   store i1 %typecode2, i1* @main.v2
   ret void
 }

main.go

@@ -644,7 +644,7 @@ func Debug(debugger, pkgName string, ocdOutput bool, options *compileopts.Option
 	case "qemu-user":
 		port = ":1234"
 		// Run in an emulator.
-		args := append(emulator[1:], "-g", "1234")
+		args := append([]string{"-g", "1234"}, emulator[1:]...)
 		daemon = executeCommand(config.Options, emulator[0], args...)
 		daemon.Stdout = os.Stdout
 		daemon.Stderr = os.Stderr

main_test.go

@@ -180,7 +180,8 @@ func runPlatTests(options compileopts.Options, tests []string, t *testing.T) {
 			// Skip the ones that aren't.
 			switch name {
 			case "reflect.go":
-				// Reflect tests do not work due to type code issues.
+				// Reflect tests do not run correctly, probably because of the
+				// limited amount of memory.
 				continue
 
 			case "gc.go":
@@ -188,20 +189,16 @@ func runPlatTests(options compileopts.Options, tests []string, t *testing.T) {
 				continue
 
 			case "json.go", "stdlib.go", "testing.go":
-				// Breaks interp.
+				// Too big for AVR. Doesn't fit in flash/RAM.
 				continue
 
 			case "math.go":
-				// Stuck somewhere, not sure what's happening.
+				// Needs newer picolibc version (for sqrt).
 				continue
 
 			case "cgo/":
-				// CGo does not work on AVR.
-				continue
-
-			case "timers.go":
-				// Doesn't compile:
-				//   panic: compiler: could not store type code number inside interface type code
+				// CGo function pointers don't work on AVR (needs LLVM 16 and
+				// some compiler changes).
 				continue
 
 			default:

src/examples/flash/main.go

@@ -0,0 +1,57 @@
+package main
+
+import (
+	"machine"
+	"time"
+)
+
+var (
+	err     error
+	message = "1234567887654321123456788765432112345678876543211234567887654321"
+)
+
+func main() {
+	time.Sleep(3 * time.Second)
+
+	// Print out general information
+	println("Flash data start:      ", machine.FlashDataStart())
+	println("Flash data end:        ", machine.FlashDataEnd())
+	println("Flash data size, bytes:", machine.Flash.Size())
+	println("Flash write block size:", machine.Flash.WriteBlockSize())
+	println("Flash erase block size:", machine.Flash.EraseBlockSize())
+	println()
+
+	flash := machine.OpenFlashBuffer(machine.Flash, machine.FlashDataStart())
+	original := make([]byte, len(message))
+	saved := make([]byte, len(message))
+
+	// Read flash contents on start (data shall survive power off)
+	print("Reading data from flash: ")
+	_, err = flash.Read(original)
+	checkError(err)
+	println(string(original))
+
+	// Write the message to flash
+	print("Writing data to flash: ")
+	flash.Seek(0, 0) // rewind back to beginning
+	_, err = flash.Write([]byte(message))
+	checkError(err)
+	println(string(message))
+
+	// Read back flash contents after write (verify data is the same as written)
+	print("Reading data back from flash: ")
+	flash.Seek(0, 0) // rewind back to beginning
+	_, err = flash.Read(saved)
+	checkError(err)
+	println(string(saved))
+	println()
+}
+
+func checkError(err error) {
+	if err != nil {
+		for {
+			println(err.Error())
+			time.Sleep(time.Second)
+		}
+	}
+}

src/examples/rtcinterrupt/rtcinterrupt.go

@@ -0,0 +1,35 @@
+//go:build rp2040
+
+package main
+
+// This example demonstrates scheduling a delayed interrupt by real time clock.
+//
+// An interrupt may execute user callback function or used for its side effects
+// like waking up from sleep or dormant states.
+//
+// The interrupt can be configured to repeat.
+//
+// There is no separate method to disable interrupt, use 0 delay for that.
+//
+// Unfortunately, it is not possible to use time.Duration to work with RTC directly,
+// that would introduce a circular dependency between "machine" and "time" packages.
+
+import (
+	"fmt"
+	"machine"
+	"time"
+)
+
+func main() {
+
+	// Schedule and enable recurring interrupt.
+	// The callback function is executed in the context of an interrupt handler,
+	// so regular restructions for this sort of code apply: no blocking, no memory allocation, etc.
+	delay := time.Minute + 12*time.Second
+	machine.RTC.SetInterrupt(uint32(delay.Seconds()), true, func() { println("Peekaboo!") })
+
+	for {
+		fmt.Printf("%v\r\n", time.Now().Format(time.RFC3339))
+		time.Sleep(1 * time.Second)
+	}
+}

src/internal/task/task_stack.go

@@ -2,7 +2,10 @@
 
 package task
 
-import "unsafe"
+import (
+	"runtime/interrupt"
+	"unsafe"
+)
 
 //go:linkname runtimePanic runtime.runtimePanic
 func runtimePanic(str string)
@@ -45,6 +48,9 @@ func Pause() {
 	if *currentTask.state.canaryPtr != stackCanary {
 		runtimePanic("goroutine stack overflow")
 	}
+	if interrupt.In() {
+		runtimePanic("blocked inside interrupt")
+	}
 	currentTask.state.pause()
 }
 

src/internal/task/task_stack_arm64.S

@@ -4,7 +4,6 @@ _tinygo_startTask:
 #else
 .section .text.tinygo_startTask
 .global  tinygo_startTask
-.type    tinygo_startTask, %function
 tinygo_startTask:
 #endif
     .cfi_startproc
@@ -35,7 +34,6 @@ tinygo_startTask:
 #endif
     .cfi_endproc
 #ifndef __MACH__
-.size tinygo_startTask, .-tinygo_startTask
 #endif
 
 
@@ -44,7 +42,6 @@ tinygo_startTask:
 _tinygo_swapTask:
 #else
 .global tinygo_swapTask
-.type tinygo_swapTask, %function
 tinygo_swapTask:
 #endif
     // This function gets the following parameters:
@@ -52,12 +49,16 @@ tinygo_swapTask:
     // x1 = oldStack *uintptr
 
     // Save all callee-saved registers:
-    stp     x19, x20, [sp, #-96]!
+    stp     x19, x20, [sp, #-160]!
     stp     x21, x22, [sp, #16]
     stp     x23, x24, [sp, #32]
     stp     x25, x26, [sp, #48]
     stp     x27, x28, [sp, #64]
     stp     x29, x30, [sp, #80]
+    stp     d8,  d9,  [sp, #96]
+    stp     d10, d11, [sp, #112]
+    stp     d12, d13, [sp, #128]
+    stp     d14, d15, [sp, #144]
 
     // Save the current stack pointer in oldStack.
     mov x8, sp
@@ -67,10 +68,14 @@ tinygo_swapTask:
     mov sp, x0
 
     // Restore stack state and return.
+    ldp     d14, d15, [sp, #144]
+    ldp     d12, d13, [sp, #128]
+    ldp     d10, d11, [sp, #112]
+    ldp     d8, d9, [sp, #96]
     ldp     x29, x30, [sp, #80]
     ldp     x27, x28, [sp, #64]
     ldp     x25, x26, [sp, #48]
     ldp     x23, x24, [sp, #32]
     ldp     x21, x22, [sp, #16]
-    ldp     x19, x20, [sp], #96
+    ldp     x19, x20, [sp], #160
     ret

src/internal/task/task_stack_arm64.go

@@ -1,4 +1,4 @@
-//go:build scheduler.tasks && arm64 && !windows
+//go:build scheduler.tasks && arm64
 
 package task
 
@@ -21,8 +21,16 @@ type calleeSavedRegs struct {
 	x27 uintptr
 	x28 uintptr
 	x29 uintptr
+	pc  uintptr // aka x30 aka LR
 
-	pc uintptr // aka x30 aka LR
+	d8  uintptr
+	d9  uintptr
+	d10 uintptr
+	d11 uintptr
+	d12 uintptr
+	d13 uintptr
+	d14 uintptr
+	d15 uintptr
 }
 
 // archInit runs architecture-specific setup for the goroutine startup.

src/internal/task/task_stack_arm64_windows.S

@@ -1,65 +0,0 @@
-.section .text.tinygo_startTask,"ax"
-.global  tinygo_startTask
-tinygo_startTask:
-    .cfi_startproc
-    // Small assembly stub for starting a goroutine. This is already run on the
-    // new stack, with the callee-saved registers already loaded.
-    // Most importantly, x19 contains the pc of the to-be-started function and
-    // x20 contains the only argument it is given. Multiple arguments are packed
-    // into one by storing them in a new allocation.
-
-    // Indicate to the unwinder that there is nothing to unwind, this is the
-    // root frame. It avoids the following (bogus) error message in GDB:
-    //     Backtrace stopped: previous frame identical to this frame (corrupt stack?)
-    .cfi_undefined lr
-
-    // Set the first argument of the goroutine start wrapper, which contains all
-    // the arguments.
-    mov   x0, x20
-
-    // Branch to the "goroutine start" function. By using blx instead of bx,
-    // we'll return here instead of tail calling.
-    blr   x19
-
-    // After return, exit this goroutine. This is a tail call.
-    b     tinygo_pause
-    .cfi_endproc
-
-
-.global tinygo_swapTask
-tinygo_swapTask:
-    // This function gets the following parameters:
-    // x0 = newStack uintptr
-    // x1 = oldStack *uintptr
-
-    // Save all callee-saved registers:
-    stp     x19, x20, [sp, #-160]!
-    stp     x21, x22, [sp, #16]
-    stp     x23, x24, [sp, #32]
-    stp     x25, x26, [sp, #48]
-    stp     x27, x28, [sp, #64]
-    stp     x29, x30, [sp, #80]
-    stp     d8,  d9,  [sp, #96]
-    stp     d10, d11, [sp, #112]
-    stp     d12, d13, [sp, #128]
-    stp     d14, d15, [sp, #144]
-
-    // Save the current stack pointer in oldStack.
-    mov x8, sp
-    str x8, [x1]
-
-    // Switch to the new stack pointer.
-    mov sp, x0
-
-    // Restore stack state and return.
-    ldp     d14, d15, [sp, #144]
-    ldp     d12, d13, [sp, #128]
-    ldp     d10, d11, [sp, #112]
-    ldp     d8, d9, [sp, #96]
-    ldp     x29, x30, [sp, #80]
-    ldp     x27, x28, [sp, #64]
-    ldp     x25, x26, [sp, #48]
-    ldp     x23, x24, [sp, #32]
-    ldp     x21, x22, [sp, #16]
-    ldp     x19, x20, [sp], #160
-    ret

src/internal/task/task_stack_arm64_windows.go

@@ -1,72 +0,0 @@
-//go:build scheduler.tasks && arm64 && windows
-
-package task
-
-import "unsafe"
-
-var systemStack uintptr
-
-// calleeSavedRegs is the list of registers that must be saved and restored
-// when switching between tasks. Also see task_stack_arm64_windows.S that
-// relies on the exact layout of this struct.
-type calleeSavedRegs struct {
-	x19 uintptr
-	x20 uintptr
-	x21 uintptr
-	x22 uintptr
-	x23 uintptr
-	x24 uintptr
-	x25 uintptr
-	x26 uintptr
-	x27 uintptr
-	x28 uintptr
-	x29 uintptr
-	pc  uintptr // aka x30 aka LR
-
-	d8  uintptr
-	d9  uintptr
-	d10 uintptr
-	d11 uintptr
-	d12 uintptr
-	d13 uintptr
-	d14 uintptr
-	d15 uintptr
-}
-
-// archInit runs architecture-specific setup for the goroutine startup.
-func (s *state) archInit(r *calleeSavedRegs, fn uintptr, args unsafe.Pointer) {
-	// Store the initial sp for the startTask function (implemented in assembly).
-	s.sp = uintptr(unsafe.Pointer(r))
-
-	// Initialize the registers.
-	// These will be popped off of the stack on the first resume of the goroutine.
-
-	// Start the function at tinygo_startTask (defined in src/internal/task/task_stack_arm64_windows.S).
-	// This assembly code calls a function (passed in x19) with a single argument
-	// (passed in x20). After the function returns, it calls Pause().
-	r.pc = uintptr(unsafe.Pointer(&startTask))
-
-	// Pass the function to call in x19.
-	// This function is a compiler-generated wrapper which loads arguments out of a struct pointer.
-	// See createGoroutineStartWrapper (defined in compiler/goroutine.go) for more information.
-	r.x19 = fn
-
-	// Pass the pointer to the arguments struct in x20.
-	r.x20 = uintptr(args)
-}
-
-func (s *state) resume() {
-	swapTask(s.sp, &systemStack)
-}
-
-func (s *state) pause() {
-	newStack := systemStack
-	systemStack = 0
-	swapTask(newStack, &s.sp)
-}
-
-// SystemStack returns the system stack pointer when called from a task stack.
-// When called from the system stack, it returns 0.
-func SystemStack() uintptr {
-	return systemStack
-}

src/machine/board_lorae5.go

@@ -56,6 +56,18 @@ var (
 	}
 	DefaultUART = UART0
 
+	// Since we treat UART1 as zero, let's also call it by the real name
+	UART1 = UART0
+
+	// UART2
+	UART2  = &_UART2
+	_UART2 = UART{
+		Buffer:            NewRingBuffer(),
+		Bus:               stm32.USART2,
+		TxAltFuncSelector: AF7_USART1_2,
+		RxAltFuncSelector: AF7_USART1_2,
+	}
+
 	// I2C Busses
 	I2C1 = &I2C{
 		Bus:             stm32.I2C1,
@@ -72,4 +84,5 @@ var (
 func init() {
 	// Enable UARTs Interrupts
 	UART0.Interrupt = interrupt.New(stm32.IRQ_USART1, _UART0.handleInterrupt)
+	UART2.Interrupt = interrupt.New(stm32.IRQ_USART2, _UART2.handleInterrupt)
 }

src/machine/board_qtpy.go

@@ -29,7 +29,7 @@ const (
 
 // Analog pins
 const (
-	A0 = D1
+	A0 = D0
 	A1 = D1
 	A2 = D2
 	A3 = D3

src/machine/board_xiao-rp2040.go

@@ -50,11 +50,11 @@ const (
 
 // I2C pins
 const (
-	I2C0_SDA_PIN Pin = D4
-	I2C0_SCL_PIN Pin = D5
+	I2C0_SDA_PIN Pin = D2
+	I2C0_SCL_PIN Pin = D3
 
-	I2C1_SDA_PIN Pin = NoPin
-	I2C1_SCL_PIN Pin = NoPin
+	I2C1_SDA_PIN Pin = D4
+	I2C1_SCL_PIN Pin = D5
 )
 
 // SPI pins

src/machine/flash.go

@@ -0,0 +1,144 @@
+//go:build nrf || nrf51 || nrf52 || nrf528xx || stm32f4 || stm32l4 || stm32wlx
+
+package machine
+
+import (
+	"errors"
+	"io"
+	"unsafe"
+)
+
+//go:extern __flash_data_start
+var flashDataStart [0]byte
+
+//go:extern __flash_data_end
+var flashDataEnd [0]byte
+
+// Return the start of the writable flash area, aligned on a page boundary. This
+// is usually just after the program and static data.
+func FlashDataStart() uintptr {
+	pagesize := uintptr(eraseBlockSize())
+	return (uintptr(unsafe.Pointer(&flashDataStart)) + pagesize - 1) &^ (pagesize - 1)
+}
+
+// Return the end of the writable flash area. Usually this is the address one
+// past the end of the on-chip flash.
+func FlashDataEnd() uintptr {
+	return uintptr(unsafe.Pointer(&flashDataEnd))
+}
+
+var (
+	errFlashCannotErasePage     = errors.New("cannot erase flash page")
+	errFlashInvalidWriteLength  = errors.New("write flash data must align to correct number of bits")
+	errFlashNotAllowedWriteData = errors.New("not allowed to write flash data")
+	errFlashCannotWriteData     = errors.New("cannot write flash data")
+	errFlashCannotReadPastEOF   = errors.New("cannot read beyond end of flash data")
+	errFlashCannotWritePastEOF  = errors.New("cannot write beyond end of flash data")
+)
+
+// BlockDevice is the raw device that is meant to store flash data.
+type BlockDevice interface {
+	// ReadAt reads the given number of bytes from the block device.
+	io.ReaderAt
+
+	// WriteAt writes the given number of bytes to the block device.
+	io.WriterAt
+
+	// Size returns the number of bytes in this block device.
+	Size() int64
+
+	// WriteBlockSize returns the block size in which data can be written to
+	// memory. It can be used by a client to optimize writes, non-aligned writes
+	// should always work correctly.
+	WriteBlockSize() int64
+
+	// EraseBlockSize returns the smallest erasable area on this particular chip
+	// in bytes. This is used for the block size in EraseBlocks.
+	// It must be a power of two, and may be as small as 1. A typical size is 4096.
+	EraseBlockSize() int64
+
+	// EraseBlocks erases the given number of blocks. An implementation may
+	// transparently coalesce ranges of blocks into larger bundles if the chip
+	// supports this. The start and len parameters are in block numbers, use
+	// EraseBlockSize to map addresses to blocks.
+	EraseBlocks(start, len int64) error
+}
+
+// FlashBuffer implements the ReadWriteCloser interface using the BlockDevice interface.
+type FlashBuffer struct {
+	b       BlockDevice
+	start   uintptr
+	current uintptr
+}
+
+// OpenFlashBuffer opens a FlashBuffer.
+func OpenFlashBuffer(b BlockDevice, address uintptr) *FlashBuffer {
+	return &FlashBuffer{b: b, start: address, current: address}
+}
+
+// Read data from a FlashBuffer.
+func (fl *FlashBuffer) Read(p []byte) (n int, err error) {
+	fl.b.ReadAt(p, int64(fl.current))
+
+	fl.current += uintptr(len(p))
+
+	return len(p), nil
+}
+
+// Write data to a FlashBuffer.
+func (fl *FlashBuffer) Write(p []byte) (n int, err error) {
+	// any new pages needed?
+	// NOTE probably will not work as expected if you try to write over page boundary
+	// of pages with different sizes.
+	pagesize := uintptr(fl.b.EraseBlockSize())
+	currentPageCount := (fl.current - fl.start + pagesize - 1) / pagesize
+	totalPagesNeeded := (fl.current - fl.start + uintptr(len(p)) + pagesize - 1) / pagesize
+	if currentPageCount == totalPagesNeeded {
+		// just write the data
+		n, err := fl.b.WriteAt(p, int64(fl.current))
+		if err != nil {
+			return 0, err
+		}
+		fl.current += uintptr(n)
+		return n, nil
+	}
+
+	// erase enough blocks to hold the data
+	page := fl.flashPageFromAddress(fl.start + (currentPageCount * pagesize))
+	fl.b.EraseBlocks(page, int64(totalPagesNeeded-currentPageCount))
+
+	// write the data
+	for i := 0; i < len(p); i += int(pagesize) {
+		var last int = i + int(pagesize)
+		if i+int(pagesize) > len(p) {
+			last = len(p)
+		}
+
+		_, err := fl.b.WriteAt(p[i:last], int64(fl.current))
+		if err != nil {
+			return 0, err
+		}
+		fl.current += uintptr(last - i)
+	}
+
+	return len(p), nil
+}
+
+// Close the FlashBuffer.
+func (fl *FlashBuffer) Close() error {
+	return nil
+}
+
+// Seek implements io.Seeker interface, but with limitations.
+// You can only seek relative to the start.
+// Also, you cannot use seek before write operations, only read.
+func (fl *FlashBuffer) Seek(offset int64, whence int) (int64, error) {
+	fl.current = fl.start + uintptr(offset)
+
+	return offset, nil
+}
+
+// calculate page number from address
+func (fl *FlashBuffer) flashPageFromAddress(address uintptr) int64 {
+	return int64(address-memoryStart) / fl.b.EraseBlockSize()
+}

src/machine/machine_atsamd21g18.go

@@ -406,61 +406,61 @@ func (p Pin) getPinCfg() uint8 {
 		return uint8(sam.PORT.PINCFG1_0.Get()>>16) & 0xff
 	case 35: // PB03
 		return uint8(sam.PORT.PINCFG1_0.Get()>>24) & 0xff
-	case 37: // PB04
+	case 36: // PB04
 		return uint8(sam.PORT.PINCFG1_4.Get()>>0) & 0xff
-	case 38: // PB05
+	case 37: // PB05
 		return uint8(sam.PORT.PINCFG1_4.Get()>>8) & 0xff
-	case 39: // PB06
+	case 38: // PB06
 		return uint8(sam.PORT.PINCFG1_4.Get()>>16) & 0xff
-	case 40: // PB07
+	case 39: // PB07
 		return uint8(sam.PORT.PINCFG1_4.Get()>>24) & 0xff
-	case 41: // PB08
+	case 40: // PB08
 		return uint8(sam.PORT.PINCFG1_8.Get()>>0) & 0xff
-	case 42: // PB09
+	case 41: // PB09
 		return uint8(sam.PORT.PINCFG1_8.Get()>>8) & 0xff
-	case 43: // PB10
+	case 42: // PB10
 		return uint8(sam.PORT.PINCFG1_8.Get()>>16) & 0xff
-	case 44: // PB11
+	case 43: // PB11
 		return uint8(sam.PORT.PINCFG1_8.Get()>>24) & 0xff
-	case 45: // PB12
+	case 44: // PB12
 		return uint8(sam.PORT.PINCFG1_12.Get()>>0) & 0xff
-	case 46: // PB13
+	case 45: // PB13
 		return uint8(sam.PORT.PINCFG1_12.Get()>>8) & 0xff
-	case 47: // PB14
+	case 46: // PB14
 		return uint8(sam.PORT.PINCFG1_12.Get()>>16) & 0xff
-	case 48: // PB15
+	case 47: // PB15
 		return uint8(sam.PORT.PINCFG1_12.Get()>>24) & 0xff
-	case 49: // PB16
+	case 48: // PB16
 		return uint8(sam.PORT.PINCFG1_16.Get()>>0) & 0xff
-	case 50: // PB17
+	case 49: // PB17
 		return uint8(sam.PORT.PINCFG1_16.Get()>>8) & 0xff
-	case 51: // PB18
+	case 50: // PB18
 		return uint8(sam.PORT.PINCFG1_16.Get()>>16) & 0xff
-	case 52: // PB19
+	case 51: // PB19
 		return uint8(sam.PORT.PINCFG1_16.Get()>>24) & 0xff
-	case 53: // PB20
+	case 52: // PB20
 		return uint8(sam.PORT.PINCFG1_20.Get()>>0) & 0xff
-	case 54: // PB21
+	case 53: // PB21
 		return uint8(sam.PORT.PINCFG1_20.Get()>>8) & 0xff
-	case 55: // PB22
+	case 54: // PB22
 		return uint8(sam.PORT.PINCFG1_20.Get()>>16) & 0xff
-	case 56: // PB23
+	case 55: // PB23
 		return uint8(sam.PORT.PINCFG1_20.Get()>>24) & 0xff
-	case 57: // PB24
+	case 56: // PB24
 		return uint8(sam.PORT.PINCFG1_24.Get()>>0) & 0xff
-	case 58: // PB25
+	case 57: // PB25
 		return uint8(sam.PORT.PINCFG1_24.Get()>>8) & 0xff
-	case 59: // PB26
+	case 58: // PB26
 		return uint8(sam.PORT.PINCFG1_24.Get()>>16) & 0xff
-	case 60: // PB27
+	case 59: // PB27
 		return uint8(sam.PORT.PINCFG1_24.Get()>>24) & 0xff
-	case 61: // PB28
+	case 60: // PB28
 		return uint8(sam.PORT.PINCFG1_28.Get()>>0) & 0xff
-	case 62: // PB29
+	case 61: // PB29
 		return uint8(sam.PORT.PINCFG1_28.Get()>>8) & 0xff
-	case 63: // PB30
+	case 62: // PB30
 		return uint8(sam.PORT.PINCFG1_28.Get()>>16) & 0xff
-	case 64: // PB31
+	case 63: // PB31
 		return uint8(sam.PORT.PINCFG1_28.Get()>>24) & 0xff
 	default:
 		return 0

src/machine/machine_nrf.go

@@ -3,7 +3,9 @@
 package machine
 
 import (
+	"bytes"
 	"device/nrf"
+	"encoding/binary"
 	"runtime/interrupt"
 	"unsafe"
 )
@@ -382,3 +384,109 @@ func ReadTemperature() int32 {
 	nrf.TEMP.EVENTS_DATARDY.Set(0)
 	return temp
 }
+
+const memoryStart = 0x0
+
+// compile-time check for ensuring we fulfill BlockDevice interface
+var _ BlockDevice = flashBlockDevice{}
+
+var Flash flashBlockDevice
+
+type flashBlockDevice struct {
+}
+
+// ReadAt reads the given number of bytes from the block device.
+func (f flashBlockDevice) ReadAt(p []byte, off int64) (n int, err error) {
+	if FlashDataStart()+uintptr(off)+uintptr(len(p)) > FlashDataEnd() {
+		return 0, errFlashCannotReadPastEOF
+	}
+
+	data := unsafe.Slice((*byte)(unsafe.Pointer(FlashDataStart()+uintptr(off))), len(p))
+	copy(p, data)
+
+	return len(p), nil
+}
+
+// WriteAt writes the given number of bytes to the block device.
+// Only double-word (64 bits) length data can be programmed. See rm0461 page 78.
+// If the length of p is not long enough it will be padded with 0xFF bytes.
+// This method assumes that the destination is already erased.
+func (f flashBlockDevice) WriteAt(p []byte, off int64) (n int, err error) {
+	if FlashDataStart()+uintptr(off)+uintptr(len(p)) > FlashDataEnd() {
+		return 0, errFlashCannotWritePastEOF
+	}
+
+	address := FlashDataStart() + uintptr(off)
+	padded := f.pad(p)
+
+	waitWhileFlashBusy()
+
+	nrf.NVMC.SetCONFIG_WEN(nrf.NVMC_CONFIG_WEN_Wen)
+	defer nrf.NVMC.SetCONFIG_WEN(nrf.NVMC_CONFIG_WEN_Ren)
+
+	for j := 0; j < len(padded); j += int(f.WriteBlockSize()) {
+		// write word
+		*(*uint32)(unsafe.Pointer(address)) = binary.LittleEndian.Uint32(padded[j : j+int(f.WriteBlockSize())])
+		address += uintptr(f.WriteBlockSize())
+		waitWhileFlashBusy()
+	}
+
+	return len(padded), nil
+}
+
+// Size returns the number of bytes in this block device.
+func (f flashBlockDevice) Size() int64 {
+	return int64(FlashDataEnd() - FlashDataStart())
+}
+
+const writeBlockSize = 4
+
+// WriteBlockSize returns the block size in which data can be written to
+// memory. It can be used by a client to optimize writes, non-aligned writes
+// should always work correctly.
+func (f flashBlockDevice) WriteBlockSize() int64 {
+	return writeBlockSize
+}
+
+// EraseBlockSize returns the smallest erasable area on this particular chip
+// in bytes. This is used for the block size in EraseBlocks.
+// It must be a power of two, and may be as small as 1. A typical size is 4096.
+func (f flashBlockDevice) EraseBlockSize() int64 {
+	return eraseBlockSize()
+}
+
+// EraseBlocks erases the given number of blocks. An implementation may
+// transparently coalesce ranges of blocks into larger bundles if the chip
+// supports this. The start and len parameters are in block numbers, use
+// EraseBlockSize to map addresses to blocks.
+func (f flashBlockDevice) EraseBlocks(start, len int64) error {
+	address := FlashDataStart() + uintptr(start*f.EraseBlockSize())
+	waitWhileFlashBusy()
+
+	nrf.NVMC.SetCONFIG_WEN(nrf.NVMC_CONFIG_WEN_Een)
+	defer nrf.NVMC.SetCONFIG_WEN(nrf.NVMC_CONFIG_WEN_Ren)
+
+	for i := start; i < start+len; i++ {
+		nrf.NVMC.ERASEPAGE.Set(uint32(address))
+		waitWhileFlashBusy()
+		address += uintptr(f.EraseBlockSize())
+	}
+
+	return nil
+}
+
+// pad data if needed so it is long enough for correct byte alignment on writes.
+func (f flashBlockDevice) pad(p []byte) []byte {
+	paddingNeeded := f.WriteBlockSize() - (int64(len(p)) % f.WriteBlockSize())
+	if paddingNeeded == 0 {
+		return p
+	}
+
+	padding := bytes.Repeat([]byte{0xff}, int(paddingNeeded))
+	return append(p, padding...)
+}
+
+func waitWhileFlashBusy() {
+	for nrf.NVMC.GetREADY() != nrf.NVMC_READY_READY_Ready {
+	}
+}

src/machine/machine_nrf51.go

@@ -6,6 +6,12 @@ import (
 	"device/nrf"
 )
 
+const eraseBlockSizeValue = 1024
+
+func eraseBlockSize() int64 {
+	return eraseBlockSizeValue
+}
+
 // Get peripheral and pin number for this GPIO pin.
 func (p Pin) getPortPin() (*nrf.GPIO_Type, uint32) {
 	return nrf.GPIO, uint32(p)

src/machine/machine_nrf52.go

@@ -63,3 +63,9 @@ var (
 	PWM1 = &PWM{PWM: nrf.PWM1}
 	PWM2 = &PWM{PWM: nrf.PWM2}
 )
+
+const eraseBlockSizeValue = 4096
+
+func eraseBlockSize() int64 {
+	return eraseBlockSizeValue
+}

src/machine/machine_nrf52833.go

@@ -84,3 +84,9 @@ var (
 	PWM2 = &PWM{PWM: nrf.PWM2}
 	PWM3 = &PWM{PWM: nrf.PWM3}
 )
+
+const eraseBlockSizeValue = 4096
+
+func eraseBlockSize() int64 {
+	return eraseBlockSizeValue
+}

src/machine/machine_nrf52840.go

@@ -102,3 +102,9 @@ func (pdm *PDM) Read(buf []int16) (uint32, error) {
 
 	return uint32(len(buf)), nil
 }
+
+const eraseBlockSizeValue = 4096
+
+func eraseBlockSize() int64 {
+	return eraseBlockSizeValue
+}

src/machine/machine_rp2040_i2c.go

@@ -57,6 +57,8 @@ var (
 	ErrInvalidTgtAddr     = errors.New("invalid target i2c address not in 0..0x80 or is reserved")
 	ErrI2CGeneric         = errors.New("i2c error")
 	ErrRP2040I2CDisable   = errors.New("i2c rp2040 peripheral timeout in disable")
+	errInvalidI2CSDA      = errors.New("invalid I2C SDA pin")
+	errInvalidI2CSCL      = errors.New("invalid I2C SCL pin")
 )
 
 // Tx performs a write and then a read transfer placing the result in
@@ -90,7 +92,7 @@ func (i2c *I2C) Tx(addr uint16, w, r []byte) error {
 //	SCL: 3, 7, 11, 15, 19, 27
 func (i2c *I2C) Configure(config I2CConfig) error {
 	const defaultBaud uint32 = 100_000 // 100kHz standard mode
-	if config.SCL == 0 {
+	if config.SCL == 0 && config.SDA == 0 {
 		// If config pins are zero valued or clock pin is invalid then we set default values.
 		switch i2c.Bus {
 		case rp.I2C0:
@@ -101,6 +103,23 @@ func (i2c *I2C) Configure(config I2CConfig) error {
 			config.SDA = I2C1_SDA_PIN
 		}
 	}
+	var okSCL, okSDA bool
+	switch i2c.Bus {
+	case rp.I2C0:
+		okSCL = (config.SCL+3)%4 == 0
+		okSDA = (config.SDA+4)%4 == 0
+	case rp.I2C1:
+		okSCL = (config.SCL+1)%4 == 0
+		okSDA = (config.SDA+2)%4 == 0
+	}
+
+	switch {
+	case !okSCL:
+		return errInvalidI2CSCL
+	case !okSDA:
+		return errInvalidI2CSDA
+	}
+
 	if config.Frequency == 0 {
 		config.Frequency = defaultBaud
 	}

src/machine/machine_rp2040_rtc.go

@@ -0,0 +1,240 @@
+//go:build rp2040
+
+// Implementation based on code located here:
+// https://github.com/raspberrypi/pico-sdk/blob/master/src/rp2_common/hardware_rtc/rtc.c
+
+package machine
+
+import (
+	"device/rp"
+	"errors"
+	"runtime/interrupt"
+	"unsafe"
+)
+
+type rtcType rp.RTC_Type
+
+type rtcTime struct {
+	Year  int16
+	Month int8
+	Day   int8
+	Dotw  int8
+	Hour  int8
+	Min   int8
+	Sec   int8
+}
+
+var RTC = (*rtcType)(unsafe.Pointer(rp.RTC))
+
+const (
+	second = 1
+	minute = 60 * second
+	hour   = 60 * minute
+	day    = 24 * hour
+)
+
+var (
+	rtcAlarmRepeats bool
+	rtcCallback     func()
+	rtcEpoch        = rtcTime{
+		Year: 1970, Month: 1, Day: 1, Dotw: 4, Hour: 0, Min: 0, Sec: 0,
+	}
+)
+
+var (
+	ErrRtcDelayTooSmall = errors.New("RTC interrupt deplay is too small, shall be at least 1 second")
+	ErrRtcDelayTooLarge = errors.New("RTC interrupt deplay is too large, shall be no more than 1 day")
+)
+
+// SetInterrupt configures delayed and optionally recurring interrupt by real time clock.
+//
+// Delay is specified in whole seconds, allowed range depends on platform.
+// Zero delay disables previously configured interrupt, if any.
+//
+// RP2040 implementation allows delay to be up to 1 day, otherwise a respective error is emitted.
+func (rtc *rtcType) SetInterrupt(delay uint32, repeat bool, callback func()) error {
+
+	// Verify delay range
+	if delay > day {
+		return ErrRtcDelayTooLarge
+	}
+
+	// De-configure delayed interrupt if delay is zero
+	if delay == 0 {
+		rtc.disableInterruptMatch()
+		return nil
+	}
+
+	// Configure delayed interrupt
+	rtc.setDivider()
+
+	rtcAlarmRepeats = repeat
+	rtcCallback = callback
+
+	err := rtc.setTime(rtcEpoch)
+	if err != nil {
+		return err
+	}
+	rtc.setAlarm(toAlarmTime(delay), callback)
+
+	return nil
+}
+
+func toAlarmTime(delay uint32) rtcTime {
+	result := rtcEpoch
+	remainder := delay + 1 // needed "+1", otherwise alarm fires one second too early
+	if remainder >= hour {
+		result.Hour = int8(remainder / hour)
+		remainder %= hour
+	}
+	if remainder >= minute {
+		result.Min = int8(remainder / minute)
+		remainder %= minute
+	}
+	result.Sec = int8(remainder)
+	return result
+}
+
+func (rtc *rtcType) setDivider() {
+	// Get clk_rtc freq and make sure it is running
+	rtcFreq := configuredFreq[clkRTC]
+	if rtcFreq == 0 {
+		panic("can not set RTC divider, clock is not running")
+	}
+
+	// Take rtc out of reset now that we know clk_rtc is running
+	resetBlock(rp.RESETS_RESET_RTC)
+	unresetBlockWait(rp.RESETS_RESET_RTC)
+
+	// Set up the 1 second divider.
+	// If rtc_freq is 400 then clkdiv_m1 should be 399
+	rtcFreq -= 1
+
+	// Check the freq is not too big to divide
+	if rtcFreq > rp.RTC_CLKDIV_M1_CLKDIV_M1_Msk {
+		panic("can not set RTC divider, clock frequency is too big to divide")
+	}
+
+	// Write divide value
+	rtc.CLKDIV_M1.Set(rtcFreq)
+}
+
+// setTime configures RTC with supplied time, initialises and activates it.
+func (rtc *rtcType) setTime(t rtcTime) error {
+
+	// Disable RTC and wait while it is still running
+	rtc.CTRL.Set(0)
+	for rtc.isActive() {
+	}
+
+	rtc.SETUP_0.Set((uint32(t.Year) << rp.RTC_SETUP_0_YEAR_Pos) |
+		(uint32(t.Month) << rp.RTC_SETUP_0_MONTH_Pos) |
+		(uint32(t.Day) << rp.RTC_SETUP_0_DAY_Pos))
+
+	rtc.SETUP_1.Set((uint32(t.Dotw) << rp.RTC_SETUP_1_DOTW_Pos) |
+		(uint32(t.Hour) << rp.RTC_SETUP_1_HOUR_Pos) |
+		(uint32(t.Min) << rp.RTC_SETUP_1_MIN_Pos) |
+		(uint32(t.Sec) << rp.RTC_SETUP_1_SEC_Pos))
+
+	// Load setup values into RTC clock domain
+	rtc.CTRL.SetBits(rp.RTC_CTRL_LOAD)
+
+	// Enable RTC and wait for it to be running
+	rtc.CTRL.SetBits(rp.RTC_CTRL_RTC_ENABLE)
+	for !rtc.isActive() {
+	}
+
+	return nil
+}
+
+func (rtc *rtcType) isActive() bool {
+	return rtc.CTRL.HasBits(rp.RTC_CTRL_RTC_ACTIVE)
+}
+
+// setAlarm configures alarm in RTC and arms it.
+// The callback is executed in the context of an interrupt handler,
+// so regular restructions for this sort of code apply: no blocking, no memory allocation, etc.
+func (rtc *rtcType) setAlarm(t rtcTime, callback func()) {
+
+	rtc.disableInterruptMatch()
+
+	// Clear all match enable bits
+	rtc.IRQ_SETUP_0.ClearBits(rp.RTC_IRQ_SETUP_0_YEAR_ENA | rp.RTC_IRQ_SETUP_0_MONTH_ENA | rp.RTC_IRQ_SETUP_0_DAY_ENA)
+	rtc.IRQ_SETUP_1.ClearBits(rp.RTC_IRQ_SETUP_1_DOTW_ENA | rp.RTC_IRQ_SETUP_1_HOUR_ENA | rp.RTC_IRQ_SETUP_1_MIN_ENA | rp.RTC_IRQ_SETUP_1_SEC_ENA)
+
+	// Only add to setup if it isn't -1 and set the match enable bits for things we care about
+	if t.Year >= 0 {
+		rtc.IRQ_SETUP_0.SetBits(uint32(t.Year) << rp.RTC_SETUP_0_YEAR_Pos)
+		rtc.IRQ_SETUP_0.SetBits(rp.RTC_IRQ_SETUP_0_YEAR_ENA)
+	}
+
+	if t.Month >= 0 {
+		rtc.IRQ_SETUP_0.SetBits(uint32(t.Month) << rp.RTC_SETUP_0_MONTH_Pos)
+		rtc.IRQ_SETUP_0.SetBits(rp.RTC_IRQ_SETUP_0_MONTH_ENA)
+	}
+
+	if t.Day >= 0 {
+		rtc.IRQ_SETUP_0.SetBits(uint32(t.Day) << rp.RTC_SETUP_0_DAY_Pos)
+		rtc.IRQ_SETUP_0.SetBits(rp.RTC_IRQ_SETUP_0_DAY_ENA)
+	}
+
+	if t.Dotw >= 0 {
+		rtc.IRQ_SETUP_1.SetBits(uint32(t.Dotw) << rp.RTC_SETUP_1_DOTW_Pos)
+		rtc.IRQ_SETUP_1.SetBits(rp.RTC_IRQ_SETUP_1_DOTW_ENA)
+	}
+
+	if t.Hour >= 0 {
+		rtc.IRQ_SETUP_1.SetBits(uint32(t.Hour) << rp.RTC_SETUP_1_HOUR_Pos)
+		rtc.IRQ_SETUP_1.SetBits(rp.RTC_IRQ_SETUP_1_HOUR_ENA)
+	}
+
+	if t.Min >= 0 {
+		rtc.IRQ_SETUP_1.SetBits(uint32(t.Min) << rp.RTC_SETUP_1_MIN_Pos)
+		rtc.IRQ_SETUP_1.SetBits(rp.RTC_IRQ_SETUP_1_MIN_ENA)
+	}
+
+	if t.Sec >= 0 {
+		rtc.IRQ_SETUP_1.SetBits(uint32(t.Sec) << rp.RTC_SETUP_1_SEC_Pos)
+		rtc.IRQ_SETUP_1.SetBits(rp.RTC_IRQ_SETUP_1_SEC_ENA)
+	}
+
+	// Enable the IRQ at the proc
+	interrupt.New(rp.IRQ_RTC_IRQ, rtcHandleInterrupt).Enable()
+
+	// Enable the IRQ at the peri
+	rtc.INTE.Set(rp.RTC_INTE_RTC)
+
+	rtc.enableInterruptMatch()
+}
+
+func (rtc *rtcType) enableInterruptMatch() {
+	// Set matching and wait for it to be enabled
+	rtc.IRQ_SETUP_0.SetBits(rp.RTC_IRQ_SETUP_0_MATCH_ENA)
+	for !rtc.IRQ_SETUP_0.HasBits(rp.RTC_IRQ_SETUP_0_MATCH_ACTIVE) {
+	}
+}
+
+func (rtc *rtcType) disableInterruptMatch() {
+	// Disable matching and wait for it to stop being active
+	rtc.IRQ_SETUP_0.ClearBits(rp.RTC_IRQ_SETUP_0_MATCH_ENA)
+	for rtc.IRQ_SETUP_0.HasBits(rp.RTC_IRQ_SETUP_0_MATCH_ACTIVE) {
+	}
+}
+
+func rtcHandleInterrupt(itr interrupt.Interrupt) {
+	// Always disable the alarm to clear the current IRQ.
+	// Even if it is a repeatable alarm, we don't want it to keep firing.
+	// If it matches on a second it can keep firing for that second.
+	RTC.disableInterruptMatch()
+
+	// Call user callback function
+	if rtcCallback != nil {
+		rtcCallback()
+	}
+
+	if rtcAlarmRepeats {
+		// If it is a repeatable alarm, reset time and re-enable the alarm.
+		RTC.setTime(rtcEpoch)
+		RTC.enableInterruptMatch()
+	}
+}

src/machine/machine_rp2040_spi.go

@@ -40,6 +40,9 @@ var (
 	ErrLSBNotSupported = errors.New("SPI LSB unsupported on PL022")
 	ErrSPITimeout      = errors.New("SPI timeout")
 	ErrSPIBaud         = errors.New("SPI baud too low or above 66.5Mhz")
+	errSPIInvalidSDI   = errors.New("invalid SPI SDI pin")
+	errSPIInvalidSDO   = errors.New("invalid SPI SDO pin")
+	errSPIInvalidSCK   = errors.New("invalid SPI SCK pin")
 )
 
 type SPI struct {
@@ -162,7 +165,7 @@ func (spi SPI) GetBaudRate() uint32 {
 // No pin configuration is needed of SCK, SDO and SDI needed after calling Configure.
 func (spi SPI) Configure(config SPIConfig) error {
 	const defaultBaud uint32 = 115200
-	if config.SCK == 0 {
+	if config.SCK == 0 && config.SDO == 0 && config.SDI == 0 {
 		// set default pins if config zero valued or invalid clock pin supplied.
 		switch spi.Bus {
 		case rp.SPI0:
@@ -175,6 +178,27 @@ func (spi SPI) Configure(config SPIConfig) error {
 			config.SDI = SPI1_SDI_PIN
 		}
 	}
+	var okSDI, okSDO, okSCK bool
+	switch spi.Bus {
+	case rp.SPI0:
+		okSDI = config.SDI == 0 || config.SDI == 4 || config.SDI == 16
+		okSDO = config.SDO == 3 || config.SDO == 7 || config.SDO == 19
+		okSCK = config.SCK == 2 || config.SCK == 6 || config.SCK == 18
+	case rp.SPI1:
+		okSDI = config.SDI == 8 || config.SDI == 12 || config.SDI == 28
+		okSDO = config.SDO == 11 || config.SDO == 15 || config.SDO == 27
+		okSCK = config.SCK == 10 || config.SCK == 14 || config.SCK == 26
+	}
+
+	switch {
+	case !okSDI:
+		return errSPIInvalidSDI
+	case !okSDO:
+		return errSPIInvalidSDO
+	case !okSCK:
+		return errSPIInvalidSCK
+	}
+
 	if config.DataBits < 4 || config.DataBits > 16 {
 		config.DataBits = 8
 	}

src/machine/machine_rp2040_usb.go

@@ -127,10 +127,10 @@ func handleUSBIRQ(intr interrupt.Interrupt) {
 	// Bus is reset
 	if (status & rp.USBCTRL_REGS_INTS_BUS_RESET) > 0 {
 		rp.USBCTRL_REGS.SIE_STATUS.Set(rp.USBCTRL_REGS_SIE_STATUS_BUS_RESET)
-		rp.USBCTRL_REGS.ADDR_ENDP.Set(0)
-
-		initEndpoint(0, usb.ENDPOINT_TYPE_CONTROL)
 		fixRP2040UsbDeviceEnumeration()
+
+		rp.USBCTRL_REGS.ADDR_ENDP.Set(0)
+		initEndpoint(0, usb.ENDPOINT_TYPE_CONTROL)
 	}
 }
 

src/machine/machine_stm32_flash.go

@@ -0,0 +1,122 @@
+//go:build stm32f4 || stm32l4 || stm32wlx
+
+package machine
+
+import (
+	"device/stm32"
+
+	"bytes"
+	"unsafe"
+)
+
+// compile-time check for ensuring we fulfill BlockDevice interface
+var _ BlockDevice = flashBlockDevice{}
+
+var Flash flashBlockDevice
+
+type flashBlockDevice struct {
+}
+
+// ReadAt reads the given number of bytes from the block device.
+func (f flashBlockDevice) ReadAt(p []byte, off int64) (n int, err error) {
+	if FlashDataStart()+uintptr(off)+uintptr(len(p)) > FlashDataEnd() {
+		return 0, errFlashCannotReadPastEOF
+	}
+
+	data := unsafe.Slice((*byte)(unsafe.Pointer(FlashDataStart()+uintptr(off))), len(p))
+	copy(p, data)
+
+	return len(p), nil
+}
+
+// WriteAt writes the given number of bytes to the block device.
+// Only double-word (64 bits) length data can be programmed. See rm0461 page 78.
+// If the length of p is not long enough it will be padded with 0xFF bytes.
+// This method assumes that the destination is already erased.
+func (f flashBlockDevice) WriteAt(p []byte, off int64) (n int, err error) {
+	if FlashDataStart()+uintptr(off)+uintptr(len(p)) > FlashDataEnd() {
+		return 0, errFlashCannotWritePastEOF
+	}
+
+	unlockFlash()
+	defer lockFlash()
+
+	return writeFlashData(FlashDataStart()+uintptr(off), f.pad(p))
+}
+
+// Size returns the number of bytes in this block device.
+func (f flashBlockDevice) Size() int64 {
+	return int64(FlashDataEnd() - FlashDataStart())
+}
+
+// WriteBlockSize returns the block size in which data can be written to
+// memory. It can be used by a client to optimize writes, non-aligned writes
+// should always work correctly.
+func (f flashBlockDevice) WriteBlockSize() int64 {
+	return writeBlockSize
+}
+
+func eraseBlockSize() int64 {
+	return eraseBlockSizeValue
+}
+
+// EraseBlockSize returns the smallest erasable area on this particular chip
+// in bytes. This is used for the block size in EraseBlocks.
+// It must be a power of two, and may be as small as 1. A typical size is 4096.
+// TODO: correctly handle processors that have differently sized blocks
+// in different areas of memory like the STM32F40x and STM32F1x.
+func (f flashBlockDevice) EraseBlockSize() int64 {
+	return eraseBlockSize()
+}
+
+// EraseBlocks erases the given number of blocks. An implementation may
+// transparently coalesce ranges of blocks into larger bundles if the chip
+// supports this. The start and len parameters are in block numbers, use
+// EraseBlockSize to map addresses to blocks.
+func (f flashBlockDevice) EraseBlocks(start, len int64) error {
+	unlockFlash()
+	defer lockFlash()
+
+	for i := start; i < start+len; i++ {
+		if err := eraseBlock(uint32(i)); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+// pad data if needed so it is long enough for correct byte alignment on writes.
+func (f flashBlockDevice) pad(p []byte) []byte {
+	paddingNeeded := f.WriteBlockSize() - (int64(len(p)) % f.WriteBlockSize())
+	if paddingNeeded == 0 {
+		return p
+	}
+
+	padded := bytes.Repeat([]byte{0xff}, int(paddingNeeded))
+	return append(p, padded...)
+}
+
+const memoryStart = 0x08000000
+
+func unlockFlash() {
+	// keys as described rm0461 page 76
+	var fkey1 uint32 = 0x45670123
+	var fkey2 uint32 = 0xCDEF89AB
+
+	// Wait for the flash memory not to be busy
+	for stm32.FLASH.GetSR_BSY() != 0 {
+	}
+
+	// Check if the controller is unlocked already
+	if stm32.FLASH.GetCR_LOCK() != 0 {
+		// Write the first key
+		stm32.FLASH.SetKEYR(fkey1)
+		// Write the second key
+		stm32.FLASH.SetKEYR(fkey2)
+	}
+}
+
+func lockFlash() {
+	stm32.FLASH.SetCR_LOCK(1)
+}

src/machine/machine_stm32f4.go

@@ -6,6 +6,8 @@ package machine
 
 import (
 	"device/stm32"
+	"encoding/binary"
+	"errors"
 	"math/bits"
 	"runtime/interrupt"
 	"runtime/volatile"
@@ -791,3 +793,142 @@ func (i2c *I2C) getSpeed(config I2CConfig) uint32 {
 		}
 	}
 }
+
+//---------- Flash related code
+
+// the block size actually depends on the sector.
+// TODO: handle this correctly for sectors > 3
+const eraseBlockSizeValue = 16384
+
+// see RM0090 page 75
+func sectorNumber(address uintptr) uint32 {
+	switch {
+	// 0x0800 0000 - 0x0800 3FFF
+	case address >= 0x08000000 && address <= 0x08003FFF:
+		return 0
+	// 0x0800 4000 - 0x0800 7FFF
+	case address >= 0x08004000 && address <= 0x08007FFF:
+		return 1
+	// 0x0800 8000 - 0x0800 BFFF
+	case address >= 0x08008000 && address <= 0x0800BFFF:
+		return 2
+	// 0x0800 C000 - 0x0800 FFFF
+	case address >= 0x0800C000 && address <= 0x0800FFFF:
+		return 3
+	// 0x0801 0000 - 0x0801 FFFF
+	case address >= 0x08010000 && address <= 0x0801FFFF:
+		return 4
+	// 0x0802 0000 - 0x0803 FFFF
+	case address >= 0x08020000 && address <= 0x0803FFFF:
+		return 5
+	// 0x0804 0000 - 0x0805 FFFF
+	case address >= 0x08040000 && address <= 0x0805FFFF:
+		return 6
+	case address >= 0x08060000 && address <= 0x0807FFFF:
+		return 7
+	case address >= 0x08080000 && address <= 0x0809FFFF:
+		return 8
+	case address >= 0x080A0000 && address <= 0x080BFFFF:
+		return 9
+	case address >= 0x080C0000 && address <= 0x080DFFFF:
+		return 10
+	case address >= 0x080E0000 && address <= 0x080FFFFF:
+		return 11
+	default:
+		return 0
+	}
+}
+
+// calculate sector number from address
+// var sector uint32 = sectorNumber(address)
+
+// see RM0090 page 85
+// eraseBlock at the passed in block number
+func eraseBlock(block uint32) error {
+	waitUntilFlashDone()
+
+	// clear any previous errors
+	stm32.FLASH.SR.SetBits(0xF0)
+
+	// set SER bit
+	stm32.FLASH.SetCR_SER(1)
+	defer stm32.FLASH.SetCR_SER(0)
+
+	// set the block (aka sector) to be erased
+	stm32.FLASH.SetCR_SNB(block)
+	defer stm32.FLASH.SetCR_SNB(0)
+
+	// start the page erase
+	stm32.FLASH.SetCR_STRT(1)
+
+	waitUntilFlashDone()
+
+	if err := checkError(); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+const writeBlockSize = 2
+
+// see RM0090 page 86
+// must write data in word-length
+func writeFlashData(address uintptr, data []byte) (int, error) {
+	if len(data)%writeBlockSize != 0 {
+		return 0, errFlashInvalidWriteLength
+	}
+
+	waitUntilFlashDone()
+
+	// clear any previous errors
+	stm32.FLASH.SR.SetBits(0xF0)
+
+	// set parallelism to x32
+	stm32.FLASH.SetCR_PSIZE(2)
+
+	for i := 0; i < len(data); i += writeBlockSize {
+		// start write operation
+		stm32.FLASH.SetCR_PG(1)
+
+		*(*uint16)(unsafe.Pointer(address)) = binary.BigEndian.Uint16(data[i : i+writeBlockSize])
+
+		waitUntilFlashDone()
+
+		if err := checkError(); err != nil {
+			return i, err
+		}
+
+		// end write operation
+		stm32.FLASH.SetCR_PG(0)
+	}
+
+	return len(data), nil
+}
+
+func waitUntilFlashDone() {
+	for stm32.FLASH.GetSR_BSY() != 0 {
+	}
+}
+
+var (
+	errFlashPGS = errors.New("errFlashPGS")
+	errFlashPGP = errors.New("errFlashPGP")
+	errFlashPGA = errors.New("errFlashPGA")
+	errFlashWRP = errors.New("errFlashWRP")
+)
+
+func checkError() error {
+	switch {
+	case stm32.FLASH.GetSR_PGSERR() != 0:
+		return errFlashPGS
+	case stm32.FLASH.GetSR_PGPERR() != 0:
+		return errFlashPGP
+	case stm32.FLASH.GetSR_PGAERR() != 0:
+		return errFlashPGA
+	case stm32.FLASH.GetSR_WRPERR() != 0:
+		return errFlashWRP
+	}
+
+	return nil
+}

src/machine/machine_stm32l4.go

@@ -4,6 +4,8 @@ package machine
 
 import (
 	"device/stm32"
+	"encoding/binary"
+	"errors"
 	"runtime/interrupt"
 	"runtime/volatile"
 	"unsafe"
@@ -543,3 +545,104 @@ func initRNG() {
 	stm32.RCC.AHB2ENR.SetBits(stm32.RCC_AHB2ENR_RNGEN)
 	stm32.RNG.CR.SetBits(stm32.RNG_CR_RNGEN)
 }
+
+//---------- Flash related code
+
+const eraseBlockSizeValue = 2048
+
+// see RM0394 page 83
+// eraseBlock of the passed in block number
+func eraseBlock(block uint32) error {
+	waitUntilFlashDone()
+
+	// clear any previous errors
+	stm32.FLASH.SR.SetBits(0x3FA)
+
+	// page erase operation
+	stm32.FLASH.SetCR_PER(1)
+	defer stm32.FLASH.SetCR_PER(0)
+
+	// set the page to be erased
+	stm32.FLASH.SetCR_PNB(block)
+
+	// start the page erase
+	stm32.FLASH.SetCR_START(1)
+
+	waitUntilFlashDone()
+
+	if err := checkError(); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+const writeBlockSize = 8
+
+// see RM0394 page 84
+// It is only possible to program double word (2 x 32-bit data).
+func writeFlashData(address uintptr, data []byte) (int, error) {
+	if len(data)%writeBlockSize != 0 {
+		return 0, errFlashInvalidWriteLength
+	}
+
+	waitUntilFlashDone()
+
+	// clear any previous errors
+	stm32.FLASH.SR.SetBits(0x3FA)
+
+	for j := 0; j < len(data); j += writeBlockSize {
+		// start page write operation
+		stm32.FLASH.SetCR_PG(1)
+
+		// write first word using double-word low order word
+		*(*uint32)(unsafe.Pointer(address)) = binary.BigEndian.Uint32(data[j+writeBlockSize/2 : j+writeBlockSize])
+
+		address += writeBlockSize / 2
+
+		// write second word using double-word high order word
+		*(*uint32)(unsafe.Pointer(address)) = binary.BigEndian.Uint32(data[j : j+writeBlockSize/2])
+
+		waitUntilFlashDone()
+
+		if err := checkError(); err != nil {
+			return j, err
+		}
+
+		// end flash write
+		stm32.FLASH.SetCR_PG(0)
+		address += writeBlockSize / 2
+	}
+
+	return len(data), nil
+}
+
+func waitUntilFlashDone() {
+	for stm32.FLASH.GetSR_BSY() != 0 {
+	}
+}
+
+var (
+	errFlashPGS  = errors.New("errFlashPGS")
+	errFlashSIZE = errors.New("errFlashSIZE")
+	errFlashPGA  = errors.New("errFlashPGA")
+	errFlashWRP  = errors.New("errFlashWRP")
+	errFlashPROG = errors.New("errFlashPROG")
+)
+
+func checkError() error {
+	switch {
+	case stm32.FLASH.GetSR_PGSERR() != 0:
+		return errFlashPGS
+	case stm32.FLASH.GetSR_SIZERR() != 0:
+		return errFlashSIZE
+	case stm32.FLASH.GetSR_PGAERR() != 0:
+		return errFlashPGA
+	case stm32.FLASH.GetSR_WRPERR() != 0:
+		return errFlashWRP
+	case stm32.FLASH.GetSR_PROGERR() != 0:
+		return errFlashPROG
+	}
+
+	return nil
+}

src/machine/machine_stm32wlx.go

@@ -6,6 +6,8 @@ package machine
 
 import (
 	"device/stm32"
+	"encoding/binary"
+	"errors"
 	"math/bits"
 	"runtime/interrupt"
 	"runtime/volatile"
@@ -424,3 +426,115 @@ const (
 	ARR_MAX = 0x10000
 	PSC_MAX = 0x10000
 )
+
+//---------- Flash related code
+
+const eraseBlockSizeValue = 2048
+
+// eraseBlock of the passed in block number
+func eraseBlock(block uint32) error {
+	waitUntilFlashDone()
+
+	// check if operation is allowed.
+	if stm32.FLASH.GetSR_PESD() != 0 {
+		return errFlashCannotErasePage
+	}
+
+	// clear any previous errors
+	stm32.FLASH.SR.SetBits(0x3FA)
+
+	// page erase operation
+	stm32.FLASH.SetCR_PER(1)
+	defer stm32.FLASH.SetCR_PER(0)
+
+	// set the address to the page to be written
+	stm32.FLASH.SetCR_PNB(block)
+	defer stm32.FLASH.SetCR_PNB(0)
+
+	// start the page erase
+	stm32.FLASH.SetCR_STRT(1)
+
+	waitUntilFlashDone()
+
+	if err := checkError(); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+const writeBlockSize = 8
+
+func writeFlashData(address uintptr, data []byte) (int, error) {
+	if len(data)%writeBlockSize != 0 {
+		return 0, errFlashInvalidWriteLength
+	}
+
+	waitUntilFlashDone()
+
+	// check if operation is allowed
+	if stm32.FLASH.GetSR_PESD() != 0 {
+		return 0, errFlashNotAllowedWriteData
+	}
+
+	// clear any previous errors
+	stm32.FLASH.SR.SetBits(0x3FA)
+
+	for j := 0; j < len(data); j += writeBlockSize {
+		// start page write operation
+		stm32.FLASH.SetCR_PG(1)
+
+		// write first word using double-word low order word
+		*(*uint32)(unsafe.Pointer(address)) = binary.BigEndian.Uint32(data[j+writeBlockSize/2 : j+writeBlockSize])
+
+		address += writeBlockSize / 2
+
+		// write second word using double-word high order word
+		*(*uint32)(unsafe.Pointer(address)) = binary.BigEndian.Uint32(data[j : j+writeBlockSize/2])
+
+		waitUntilFlashDone()
+
+		if err := checkError(); err != nil {
+			return j, err
+		}
+
+		// end flash write
+		stm32.FLASH.SetCR_PG(0)
+		address += writeBlockSize / 2
+	}
+
+	return len(data), nil
+}
+
+func waitUntilFlashDone() {
+	for stm32.FLASH.GetSR_BSY() != 0 {
+	}
+
+	for stm32.FLASH.GetSR_CFGBSY() != 0 {
+	}
+}
+
+var (
+	errFlashPGS  = errors.New("errFlashPGS")
+	errFlashSIZE = errors.New("errFlashSIZE")
+	errFlashPGA  = errors.New("errFlashPGA")
+	errFlashWRP  = errors.New("errFlashWRP")
+	errFlashPROG = errors.New("errFlashPROG")
+)
+
+func checkError() error {
+	switch {
+	case stm32.FLASH.GetSR_PGSERR() != 0:
+		return errFlashPGS
+	case stm32.FLASH.GetSR_SIZERR() != 0:
+		return errFlashSIZE
+	case stm32.FLASH.GetSR_PGAERR() != 0:
+		return errFlashPGA
+	case stm32.FLASH.GetSR_WRPERR() != 0:
+		return errFlashWRP
+	case stm32.FLASH.GetSR_PROGERR() != 0:
+		return errFlashPROG
+	}
+
+	return nil
+}

src/machine/usb/descriptor.go

@@ -67,7 +67,7 @@ var DescriptorCDCHID = Descriptor{
 		0x07, 0x05, 0x02, 0x02, 0x40, 0x00, 0x00,
 		0x07, 0x05, 0x83, 0x02, 0x40, 0x00, 0x00,
 		0x09, 0x04, 0x02, 0x00, 0x01, 0x03, 0x00, 0x00, 0x00,
-		0x09, 0x21, 0x01, 0x01, 0x00, 0x01, 0x22, 0x65, 0x00,
+		0x09, 0x21, 0x01, 0x01, 0x00, 0x01, 0x22, 0x7E, 0x00,
 		0x07, 0x05, 0x84, 0x03, 0x40, 0x00, 0x01,
 	},
 	HID: map[uint16][]byte{
@@ -80,6 +80,19 @@ var DescriptorCDCHID = Descriptor{
 			0x03, 0x15, 0x00, 0x25, 0x01, 0x95, 0x03, 0x75, 0x01, 0x81, 0x02, 0x95, 0x01, 0x75, 0x05, 0x81,
 			0x03, 0x05, 0x01, 0x09, 0x30, 0x09, 0x31, 0x09, 0x38, 0x15, 0x81, 0x25, 0x7f, 0x75, 0x08, 0x95,
 			0x03, 0x81, 0x06, 0xc0, 0xc0,
+
+			0x05, 0x0C, //       Usage Page (Consumer)
+			0x09, 0x01, //       Usage (Consumer Control)
+			0xA1, 0x01, //       Collection (Application)
+			0x85, 0x03, //         Report ID (3)
+			0x15, 0x00, //         Logical Minimum (0)
+			0x26, 0xFF, 0x1F, //   Logical Maximum (8191)
+			0x19, 0x00, //         Usage Minimum (Unassigned)
+			0x2A, 0xFF, 0x1F, //   Usage Maximum (0x1FFF)
+			0x75, 0x10, //         Report Size (16)
+			0x95, 0x01, //         Report Count (1)
+			0x81, 0x00, //         Input (Data,Array,Abs,No Wrap,Linear,Preferred State,No Null Position)
+			0xC0, //             End       Collection
 		},
 	},
 }

src/machine/usb/hid/keyboard/keyboard.go

@@ -238,16 +238,26 @@ func (kb *keyboard) sendKey(consumer bool, b []byte) bool {
 
 func (kb *keyboard) keyboardSendKeys(consumer bool) bool {
 	var b [9]byte
-	b[0] = 0x02
-	b[1] = kb.mod
-	b[2] = 0x02
-	b[3] = kb.key[0]
-	b[4] = kb.key[1]
-	b[5] = kb.key[2]
-	b[6] = kb.key[3]
-	b[7] = kb.key[4]
-	b[8] = kb.key[5]
-	return kb.sendKey(consumer, b[:])
+
+	if !consumer {
+		b[0] = 0x02 // REPORT_ID
+		b[1] = kb.mod
+		b[2] = 0x02
+		b[3] = kb.key[0]
+		b[4] = kb.key[1]
+		b[5] = kb.key[2]
+		b[6] = kb.key[3]
+		b[7] = kb.key[4]
+		b[8] = kb.key[5]
+		return kb.sendKey(consumer, b[:])
+
+	} else {
+		b[0] = 0x03 // REPORT_ID
+		b[1] = uint8(kb.con[0])
+		b[2] = uint8((kb.con[0] & 0x0300) >> 8)
+
+		return kb.sendKey(consumer, b[:3])
+	}
 }
 
 // Down transmits a key-down event for the given Keycode.

src/machine/usb/hid/keyboard/keycode.go

@@ -77,6 +77,7 @@ const (
 	KeyModifierRightAlt   Keycode = 0x40 | 0xE000
 	KeyModifierRightGUI   Keycode = 0x80 | 0xE000
 
+	// KeySystemXXX is not supported now
 	KeySystemPowerDown Keycode = 0x81 | 0xE200
 	KeySystemSleep     Keycode = 0x82 | 0xE200
 	KeySystemWakeUp    Keycode = 0x83 | 0xE200

src/reflect/all_test.go

@@ -71,7 +71,7 @@ var deepEqualTests = []DeepEqualTest{
 	{&[3]int{1, 2, 3}, &[3]int{1, 2, 3}, true},
 	{Basic{1, 0.5}, Basic{1, 0.5}, true},
 	{error(nil), error(nil), true},
-	//{map[int]string{1: "one", 2: "two"}, map[int]string{2: "two", 1: "one"}, true},
+	{map[int]string{1: "one", 2: "two"}, map[int]string{2: "two", 1: "one"}, true},
 	{fn1, fn2, true},
 	{[]byte{1, 2, 3}, []byte{1, 2, 3}, true},
 	{[]MyByte{1, 2, 3}, []MyByte{1, 2, 3}, true},
@@ -87,10 +87,10 @@ var deepEqualTests = []DeepEqualTest{
 	{&[3]int{1, 2, 3}, &[3]int{1, 2, 4}, false},
 	{Basic{1, 0.5}, Basic{1, 0.6}, false},
 	{Basic{1, 0}, Basic{2, 0}, false},
-	//{map[int]string{1: "one", 3: "two"}, map[int]string{2: "two", 1: "one"}, false},
-	//{map[int]string{1: "one", 2: "txo"}, map[int]string{2: "two", 1: "one"}, false},
-	//{map[int]string{1: "one"}, map[int]string{2: "two", 1: "one"}, false},
-	//{map[int]string{2: "two", 1: "one"}, map[int]string{1: "one"}, false},
+	{map[int]string{1: "one", 3: "two"}, map[int]string{2: "two", 1: "one"}, false},
+	{map[int]string{1: "one", 2: "txo"}, map[int]string{2: "two", 1: "one"}, false},
+	{map[int]string{1: "one"}, map[int]string{2: "two", 1: "one"}, false},
+	{map[int]string{2: "two", 1: "one"}, map[int]string{1: "one"}, false},
 	{nil, 1, false},
 	{1, nil, false},
 	{fn1, fn3, false},
@@ -104,16 +104,16 @@ var deepEqualTests = []DeepEqualTest{
 	{&[1]float64{math.NaN()}, self{}, true},
 	{[]float64{math.NaN()}, []float64{math.NaN()}, false},
 	{[]float64{math.NaN()}, self{}, true},
-	//{map[float64]float64{math.NaN(): 1}, map[float64]float64{1: 2}, false},
-	//{map[float64]float64{math.NaN(): 1}, self{}, true},
+	{map[float64]float64{math.NaN(): 1}, map[float64]float64{1: 2}, false},
+	{map[float64]float64{math.NaN(): 1}, self{}, true},
 
 	// Nil vs empty: not the same.
 	{[]int{}, []int(nil), false},
 	{[]int{}, []int{}, true},
 	{[]int(nil), []int(nil), true},
-	//{map[int]int{}, map[int]int(nil), false},
-	//{map[int]int{}, map[int]int{}, true},
-	//{map[int]int(nil), map[int]int(nil), true},
+	{map[int]int{}, map[int]int(nil), false},
+	{map[int]int{}, map[int]int{}, true},
+	{map[int]int(nil), map[int]int(nil), true},
 
 	// Mismatched types
 	{1, 1.0, false},
@@ -130,8 +130,8 @@ var deepEqualTests = []DeepEqualTest{
 	// Possible loops.
 	{&loopy1, &loopy1, true},
 	{&loopy1, &loopy2, true},
-	//{&cycleMap1, &cycleMap2, true},
-	//{&cycleMap1, &cycleMap3, false},
+	{&cycleMap1, &cycleMap2, true},
+	{&cycleMap1, &cycleMap3, false},
 }
 
 func TestDeepEqual(t *testing.T) {

src/reflect/deepequal.go

@@ -15,7 +15,7 @@ import "unsafe"
 type visit struct {
 	a1  unsafe.Pointer
 	a2  unsafe.Pointer
-	typ rawType
+	typ *rawType
 }
 
 // Tests for deep equality using reflected types. The map argument tracks

src/reflect/sidetables.go

@@ -1,61 +0,0 @@
-package reflect
-
-import (
-	"unsafe"
-)
-
-// This stores a varint for each named type. Named types are identified by their
-// name instead of by their type. The named types stored in this struct are
-// non-basic types: pointer, struct, and channel.
-//
-//go:extern reflect.namedNonBasicTypesSidetable
-var namedNonBasicTypesSidetable uintptr
-
-//go:extern reflect.structTypesSidetable
-var structTypesSidetable byte
-
-//go:extern reflect.structNamesSidetable
-var structNamesSidetable byte
-
-//go:extern reflect.arrayTypesSidetable
-var arrayTypesSidetable byte
-
-// readStringSidetable reads a string from the given table (like
-// structNamesSidetable) and returns this string. No heap allocation is
-// necessary because it makes the string point directly to the raw bytes of the
-// table.
-func readStringSidetable(table unsafe.Pointer, index uintptr) string {
-	nameLen, namePtr := readVarint(unsafe.Pointer(uintptr(table) + index))
-	return *(*string)(unsafe.Pointer(&stringHeader{
-		data: namePtr,
-		len:  nameLen,
-	}))
-}
-
-// readVarint decodes a varint as used in the encoding/binary package.
-// It has an input pointer and returns the read varint and the pointer
-// incremented to the next field in the data structure, just after the varint.
-//
-// Details:
-// https://github.com/golang/go/blob/e37a1b1c/src/encoding/binary/varint.go#L7-L25
-func readVarint(buf unsafe.Pointer) (uintptr, unsafe.Pointer) {
-	var n uintptr
-	shift := uintptr(0)
-	for {
-		// Read the next byte in the buffer.
-		c := *(*byte)(buf)
-
-		// Decode the bits from this byte and add them to the output number.
-		n |= uintptr(c&0x7f) << shift
-		shift += 7
-
-		// Increment the buf pointer (pointer arithmetic!).
-		buf = unsafe.Pointer(uintptr(buf) + 1)
-
-		// Check whether this is the last byte of this varint. The upper bit
-		// (msb) indicates whether any bytes follow.
-		if c>>7 == 0 {
-			return n, buf
-		}
-	}
-}

src/reflect/type.go

@@ -2,36 +2,74 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+// Type information of an interface is stored as a pointer to a global in the
+// interface type (runtime._interface). This is called a type struct.
+// It always starts with a byte that contains both the type kind and a few
+// flags. In most cases it also contains a pointer to another type struct
+// (ptrTo), that is the pointer type of the current type (for example, type int
+// also has a pointer to the type *int). The exception is pointer types, to
+// avoid infinite recursion.
+//
+// The layouts specifically look like this:
+// - basic types (Bool..UnsafePointer):
+//     meta         uint8 // actually: kind + flags
+//     ptrTo        *typeStruct
+// - named types (see elemType):
+//     meta         uint8
+//     ptrTo        *typeStruct
+//     underlying   *typeStruct // the underlying, non-named type
+// - channels and slices (see elemType):
+//     meta          uint8
+//     ptrTo        *typeStruct
+//     elementType  *typeStruct // the type that you get with .Elem()
+// - pointer types (see ptrType, this doesn't include chan, map, etc):
+//     meta         uint8
+//     elementType  *typeStruct
+// - array types (see arrayType)
+//     meta         uint8
+//     ptrTo        *typeStruct
+//     elem         *typeStruct // element type of the array
+//     arrayLen     uintptr     // length of the array (this is part of the type)
+// - map types (this is still missing the key and element types)
+//     meta         uint8
+//     ptrTo        *typeStruct
+//     elem         *typeStruct
+//     key          *typeStruct
+// - struct types (see structType):
+//     meta         uint8
+//     numField     uint16
+//     ptrTo        *typeStruct
+//     fields       [...]structField // the remaining fields are all of type structField
+// - interface types (this is missing the interface methods):
+//     meta         uint8
+//     ptrTo        *typeStruct
+// - signature types (this is missing input and output parameters):
+//     meta         uint8
+//     ptrTo        *typeStruct
+//
+// The type struct is essentially a union of all the above types. Which it is,
+// can be determined by looking at the meta byte.
+
 package reflect
 
 import (
 	"unsafe"
 )
 
-// The compiler uses a compact encoding to store type information. Unlike the
-// main Go compiler, most of the types are stored directly in the type code.
-//
-// Type code bit allocation:
-// xxxxx0: basic types, where xxxxx is the basic type number (never 0).
-//         The higher bits indicate the named type, if any.
-//  nxxx1: complex types, where n indicates whether this is a named type (named
-//         if set) and xxx contains the type kind number:
-//             0 (0001): Chan
-//             1 (0011): Interface
-//             2 (0101): Pointer
-//             3 (0111): Slice
-//             4 (1001): Array
-//             5 (1011): Func
-//             6 (1101): Map
-//             7 (1111): Struct
-//         The higher bits are either the contents of the type depending on the
-//         type (if n is clear) or indicate the number of the named type (if n
-//         is set).
+// Flags stored in the first byte of the struct field byte array. Must be kept
+// up to date with compiler/interface.go.
+const (
+	structFieldFlagAnonymous = 1 << iota
+	structFieldFlagHasTag
+	structFieldFlagIsExported
+)
 
-type Kind uintptr
+type Kind uint8
 
 // Copied from reflect/type.go
 // https://golang.org/src/reflect/type.go?s=8302:8316#L217
+// These constants must match basicTypes and the typeKind* constants in
+// compiler/interface.go
 const (
 	Invalid Kind = iota
 	Bool
@@ -124,11 +162,6 @@ func (k Kind) String() string {
 	}
 }
 
-// basicType returns a new Type for this kind if Kind is a basic type.
-func (k Kind) basicType() rawType {
-	return rawType(k << 1)
-}
-
 // Copied from reflect/type.go
 // https://go.dev/src/reflect/type.go?#L348
 
@@ -346,8 +379,71 @@ type Type interface {
 	Out(i int) Type
 }
 
-// The typecode as used in an interface{}.
-type rawType uintptr
+// Constants for the 'meta' byte.
+const (
+	kindMask  = 31 // mask to apply to the meta byte to get the Kind value
+	flagNamed = 32 // flag that is set if this is a named type
+)
+
+// The base type struct. All type structs start with this.
+type rawType struct {
+	meta uint8 // metadata byte, contains kind and flags (see contants above)
+}
+
+// All types that have an element type: named, chan, slice, array, map (but not
+// pointer because it doesn't have ptrTo).
+type elemType struct {
+	rawType
+	ptrTo *rawType
+	elem  *rawType
+}
+
+type ptrType struct {
+	rawType
+	elem *rawType
+}
+
+type arrayType struct {
+	rawType
+	ptrTo    *rawType
+	elem     *rawType
+	arrayLen uintptr
+}
+
+type mapType struct {
+	rawType
+	ptrTo *rawType
+	elem  *rawType
+	key   *rawType
+}
+
+// Type for struct types. The numField value is intentionally put before ptrTo
+// for better struct packing on 32-bit and 64-bit architectures. On these
+// architectures, the ptrTo field still has the same offset as in all the other
+// type structs.
+// The fields array isn't necessarily 1 structField long, instead it is as long
+// as numFields. The array is given a length of 1 to satisfy the Go type
+// checker.
+type structType struct {
+	rawType
+	numField uint16
+	ptrTo    *rawType
+	fields   [1]structField // the remaining fields are all of type structField
+}
+
+type structField struct {
+	fieldType *rawType
+	data      unsafe.Pointer // various bits of information, packed in a byte array
+}
+
+// Equivalent to (go/types.Type).Underlying(): if this is a named type return
+// the underlying type, else just return the type itself.
+func (t *rawType) underlying() *rawType {
+	if t.meta&flagNamed != 0 {
+		return (*elemType)(unsafe.Pointer(t)).elem
+	}
+	return t
+}
 
 func TypeOf(i interface{}) Type {
 	return ValueOf(i).typecode
@@ -356,70 +452,53 @@ func TypeOf(i interface{}) Type {
 func PtrTo(t Type) Type { return PointerTo(t) }
 
 func PointerTo(t Type) Type {
-	if t.Kind() == Pointer {
+	switch t.Kind() {
+	case Pointer:
 		panic("reflect: cannot make **T type")
+	case Struct:
+		return (*structType)(unsafe.Pointer(t.(*rawType))).ptrTo
+	default:
+		return (*elemType)(unsafe.Pointer(t.(*rawType))).ptrTo
 	}
-	ptrType := t.(rawType)<<5 | 5 // 0b0101 == 5
-	if ptrType>>5 != t {
-		panic("reflect: PointerTo type does not fit")
-	}
-	return ptrType
 }
 
-func (t rawType) String() string {
+func (t *rawType) String() string {
 	return "T"
 }
 
-func (t rawType) Kind() Kind {
-	if t%2 == 0 {
-		// basic type
-		return Kind((t >> 1) % 32)
-	} else {
-		return Kind(t>>1)%8 + 19
-	}
+func (t *rawType) Kind() Kind {
+	return Kind(t.meta & kindMask)
 }
 
 // Elem returns the element type for channel, slice and array types, the
 // pointed-to value for pointer types, and the key type for map types.
-func (t rawType) Elem() Type {
+func (t *rawType) Elem() Type {
 	return t.elem()
 }
 
-func (t rawType) elem() rawType {
-	switch t.Kind() {
-	case Chan, Pointer, Slice:
-		return t.stripPrefix()
-	case Array:
-		index := t.stripPrefix()
-		elem, _ := readVarint(unsafe.Pointer(uintptr(unsafe.Pointer(&arrayTypesSidetable)) + uintptr(index)))
-		return rawType(elem)
-	default: // not implemented: Map
-		panic("unimplemented: (reflect.Type).Elem()")
+func (t *rawType) elem() *rawType {
+	underlying := t.underlying()
+	switch underlying.Kind() {
+	case Pointer:
+		return (*ptrType)(unsafe.Pointer(underlying)).elem
+	case Chan, Slice, Array, Map:
+		return (*elemType)(unsafe.Pointer(underlying)).elem
+	default:
+		panic(&TypeError{"Elem"})
 	}
 }
 
-// stripPrefix removes the "prefix" (the low 5 bits of the type code) from
-// the type code. If this is a named type, it will resolve the underlying type
-// (which is the data for this named type). If it is not, the lower bits are
-// simply shifted off.
-//
-// The behavior is only defined for non-basic types.
-func (t rawType) stripPrefix() rawType {
-	// Look at the 'n' bit in the type code (see the top of this file) to see
-	// whether this is a named type.
-	if (t>>4)%2 != 0 {
-		// This is a named type. The data is stored in a sidetable.
-		namedTypeNum := t >> 5
-		n := *(*uintptr)(unsafe.Pointer(uintptr(unsafe.Pointer(&namedNonBasicTypesSidetable)) + uintptr(namedTypeNum)*unsafe.Sizeof(uintptr(0))))
-		return rawType(n)
+func (t *rawType) key() *rawType {
+	underlying := t.underlying()
+	if underlying.Kind() != Map {
+		panic(&TypeError{"Key"})
 	}
-	// Not a named type, so the value is stored directly in the type code.
-	return t >> 5
+	return (*mapType)(unsafe.Pointer(underlying)).key
 }
 
 // Field returns the type of the i'th field of this struct type. It panics if t
 // is not a struct type.
-func (t rawType) Field(i int) StructField {
+func (t *rawType) Field(i int) StructField {
 	field := t.rawField(i)
 	return StructField{
 		Name:      field.Name,
@@ -435,82 +514,87 @@ func (t rawType) Field(i int) StructField {
 // Type member to an interface.
 //
 // For internal use only.
-func (t rawType) rawField(i int) rawStructField {
+func (t *rawType) rawField(n int) rawStructField {
 	if t.Kind() != Struct {
 		panic(&TypeError{"Field"})
 	}
-	structIdentifier := t.stripPrefix()
-
-	numField, p := readVarint(unsafe.Pointer(uintptr(unsafe.Pointer(&structTypesSidetable)) + uintptr(structIdentifier)))
-	if uint(i) >= uint(numField) {
+	descriptor := (*structType)(unsafe.Pointer(t.underlying()))
+	if uint(n) >= uint(descriptor.numField) {
 		panic("reflect: field index out of range")
 	}
 
-	// Iterate over every field in the struct and update the StructField each
-	// time, until the target field has been reached. This is very much not
-	// efficient, but it is easy to implement.
-	// Adding a jump table at the start to jump to the field directly would
-	// make this much faster, but that would also impact code size.
-	field := rawStructField{}
-	offset := uintptr(0)
-	for fieldNum := 0; fieldNum <= i; fieldNum++ {
-		// Read some flags of this field, like whether the field is an
-		// embedded field.
-		flagsByte := *(*uint8)(p)
-		p = unsafe.Pointer(uintptr(p) + 1)
+	// Iterate over all the fields to calculate the offset.
+	// This offset could have been stored directly in the array (to make the
+	// lookup faster), but by calculating it on-the-fly a bit of storage can be
+	// saved.
+	field := &descriptor.fields[0]
+	var offset uintptr = 0
+	for i := 0; i < n; i++ {
+		offset += field.fieldType.Size()
 
-		// Read the type of this struct field.
-		var fieldTypeVal uintptr
-		fieldTypeVal, p = readVarint(p)
-		fieldType := rawType(fieldTypeVal)
-		field.Type = fieldType
+		// Increment pointer to the next field.
+		field = (*structField)(unsafe.Add(unsafe.Pointer(field), unsafe.Sizeof(structField{})))
 
-		// Move Offset forward to align it to this field's alignment.
-		// Assume alignment is a power of two.
-		offset = align(offset, uintptr(fieldType.Align()))
-		field.Offset = offset
-		offset += fieldType.Size() // starting (unaligned) offset for next field
-
-		// Read the field name.
-		var nameNum uintptr
-		nameNum, p = readVarint(p)
-		field.Name = readStringSidetable(unsafe.Pointer(&structNamesSidetable), nameNum)
-
-		// The first bit in the flagsByte indicates whether this is an embedded
-		// field.
-		field.Anonymous = flagsByte&1 != 0
-
-		// The second bit indicates whether there is a tag.
-		if flagsByte&2 != 0 {
-			// There is a tag.
-			var tagNum uintptr
-			tagNum, p = readVarint(p)
-			field.Tag = StructTag(readStringSidetable(unsafe.Pointer(&structNamesSidetable), tagNum))
-		} else {
-			// There is no tag.
-			field.Tag = ""
-		}
-
-		// The third bit indicates whether this field is exported.
-		if flagsByte&4 != 0 {
-			// This field is exported.
-			field.PkgPath = ""
-		} else {
-			// This field is unexported.
-			// TODO: list the real package path here. Storing it should not
-			// significantly impact binary size as there is only a limited
-			// number of packages in any program.
-			field.PkgPath = "<unimplemented>"
-		}
+		// Align the offset for the next field.
+		offset = align(offset, uintptr(field.fieldType.Align()))
 	}
 
-	return field
+	data := field.data
+
+	// Read some flags of this field, like whether the field is an embedded
+	// field. See structFieldFlagAnonymous and similar flags.
+	flagsByte := *(*byte)(data)
+	data = unsafe.Add(data, 1)
+
+	// Read the field name.
+	nameStart := data
+	var nameLen uintptr
+	for *(*byte)(data) != 0 {
+		nameLen++
+		data = unsafe.Add(data, 1) // C: data++
+	}
+	name := *(*string)(unsafe.Pointer(&stringHeader{
+		data: nameStart,
+		len:  nameLen,
+	}))
+
+	// Read the field tag, if there is one.
+	var tag string
+	if flagsByte&structFieldFlagHasTag != 0 {
+		data = unsafe.Add(data, 1) // C: data+1
+		tagLen := uintptr(*(*byte)(data))
+		data = unsafe.Add(data, 1) // C: data+1
+		tag = *(*string)(unsafe.Pointer(&stringHeader{
+			data: data,
+			len:  tagLen,
+		}))
+	}
+
+	// Set the PkgPath to some (arbitrary) value if the package path is not
+	// exported.
+	pkgPath := ""
+	if flagsByte&structFieldFlagIsExported == 0 {
+		// This field is unexported.
+		// TODO: list the real package path here. Storing it should not
+		// significantly impact binary size as there is only a limited
+		// number of packages in any program.
+		pkgPath = "<unimplemented>"
+	}
+
+	return rawStructField{
+		Name:      name,
+		PkgPath:   pkgPath,
+		Type:      field.fieldType,
+		Tag:       StructTag(tag),
+		Anonymous: flagsByte&structFieldFlagAnonymous != 0,
+		Offset:    offset,
+	}
 }
 
 // Bits returns the number of bits that this type uses. It is only valid for
 // arithmetic types (integers, floats, and complex numbers). For other types, it
 // will panic.
-func (t rawType) Bits() int {
+func (t *rawType) Bits() int {
 	kind := t.Kind()
 	if kind >= Int && kind <= Complex128 {
 		return int(t.Size()) * 8
@@ -520,34 +604,26 @@ func (t rawType) Bits() int {
 
 // Len returns the number of elements in this array. It panics of the type kind
 // is not Array.
-func (t rawType) Len() int {
+func (t *rawType) Len() int {
 	if t.Kind() != Array {
 		panic(TypeError{"Len"})
 	}
 
-	// skip past the element type
-	arrayIdentifier := t.stripPrefix()
-	_, p := readVarint(unsafe.Pointer(uintptr(unsafe.Pointer(&arrayTypesSidetable)) + uintptr(arrayIdentifier)))
-
-	// Read the array length.
-	arrayLen, _ := readVarint(p)
-	return int(arrayLen)
+	return int((*arrayType)(unsafe.Pointer(t.underlying())).arrayLen)
 }
 
 // NumField returns the number of fields of a struct type. It panics for other
 // type kinds.
-func (t rawType) NumField() int {
+func (t *rawType) NumField() int {
 	if t.Kind() != Struct {
 		panic(&TypeError{"NumField"})
 	}
-	structIdentifier := t.stripPrefix()
-	n, _ := readVarint(unsafe.Pointer(uintptr(unsafe.Pointer(&structTypesSidetable)) + uintptr(structIdentifier)))
-	return int(n)
+	return int((*structType)(unsafe.Pointer(t.underlying())).numField)
 }
 
 // Size returns the size in bytes of a given type. It is similar to
 // unsafe.Sizeof.
-func (t rawType) Size() uintptr {
+func (t *rawType) Size() uintptr {
 	switch t.Kind() {
 	case Bool, Int8, Uint8:
 		return 1
@@ -596,7 +672,7 @@ func (t rawType) Size() uintptr {
 
 // Align returns the alignment of this type. It is similar to calling
 // unsafe.Alignof.
-func (t rawType) Align() int {
+func (t *rawType) Align() int {
 	switch t.Kind() {
 	case Bool, Int8, Uint8:
 		return int(unsafe.Alignof(int8(0)))
@@ -648,14 +724,14 @@ func (t rawType) Align() int {
 
 // FieldAlign returns the alignment if this type is used in a struct field. It
 // is currently an alias for Align() but this might change in the future.
-func (t rawType) FieldAlign() int {
+func (t *rawType) FieldAlign() int {
 	return t.Align()
 }
 
 // AssignableTo returns whether a value of type t can be assigned to a variable
 // of type u.
-func (t rawType) AssignableTo(u Type) bool {
-	if t == u.(rawType) {
+func (t *rawType) AssignableTo(u Type) bool {
+	if t == u.(*rawType) {
 		return true
 	}
 	if u.Kind() == Interface {
@@ -664,7 +740,7 @@ func (t rawType) AssignableTo(u Type) bool {
 	return false
 }
 
-func (t rawType) Implements(u Type) bool {
+func (t *rawType) Implements(u Type) bool {
 	if u.Kind() != Interface {
 		panic("reflect: non-interface type passed to Type.Implements")
 	}
@@ -672,7 +748,7 @@ func (t rawType) Implements(u Type) bool {
 }
 
 // Comparable returns whether values of this type can be compared to each other.
-func (t rawType) Comparable() bool {
+func (t *rawType) Comparable() bool {
 	switch t.Kind() {
 	case Bool, Int, Int8, Int16, Int32, Int64, Uint, Uint8, Uint16, Uint32, Uint64, Uintptr:
 		return true
@@ -709,36 +785,59 @@ func (t rawType) Comparable() bool {
 	}
 }
 
+// isbinary() returns if the hashmapAlgorithmBinary functions can be used on this type
+func (t *rawType) isBinary() bool {
+	switch t.Kind() {
+	case Bool, Int, Int8, Int16, Int32, Int64, Uint, Uint8, Uint16, Uint32, Uint64, Uintptr:
+		return true
+	case Float32, Float64, Complex64, Complex128:
+		return true
+	case Pointer:
+		return true
+	case Array:
+		return t.elem().isBinary()
+	case Struct:
+		numField := t.NumField()
+		for i := 0; i < numField; i++ {
+			if !t.rawField(i).Type.isBinary() {
+				return false
+			}
+		}
+		return true
+	}
+	return false
+}
+
 func (t rawType) ChanDir() ChanDir {
 	panic("unimplemented: (reflect.Type).ChanDir()")
 }
 
-func (t rawType) ConvertibleTo(u Type) bool {
+func (t *rawType) ConvertibleTo(u Type) bool {
 	panic("unimplemented: (reflect.Type).ConvertibleTo()")
 }
 
-func (t rawType) IsVariadic() bool {
+func (t *rawType) IsVariadic() bool {
 	panic("unimplemented: (reflect.Type).IsVariadic()")
 }
 
-func (t rawType) NumIn() int {
+func (t *rawType) NumIn() int {
 	panic("unimplemented: (reflect.Type).NumIn()")
 }
 
-func (t rawType) NumOut() int {
+func (t *rawType) NumOut() int {
 	panic("unimplemented: (reflect.Type).NumOut()")
 }
 
-func (t rawType) NumMethod() int {
+func (t *rawType) NumMethod() int {
 	panic("unimplemented: (reflect.Type).NumMethod()")
 }
 
-func (t rawType) Name() string {
+func (t *rawType) Name() string {
 	panic("unimplemented: (reflect.Type).Name()")
 }
 
-func (t rawType) Key() Type {
-	panic("unimplemented: (reflect.Type).Key()")
+func (t *rawType) Key() Type {
+	return t.key()
 }
 
 func (t rawType) In(i int) Type {
@@ -792,7 +891,7 @@ func (f StructField) IsExported() bool {
 type rawStructField struct {
 	Name      string
 	PkgPath   string
-	Type      rawType
+	Type      *rawType
 	Tag       StructTag
 	Anonymous bool
 	Offset    uintptr

src/reflect/value.go

@@ -17,7 +17,7 @@ const (
 )
 
 type Value struct {
-	typecode rawType
+	typecode *rawType
 	value    unsafe.Pointer
 	flags    valueFlags
 }
@@ -44,15 +44,15 @@ func Indirect(v Value) Value {
 }
 
 //go:linkname composeInterface runtime.composeInterface
-func composeInterface(rawType, unsafe.Pointer) interface{}
+func composeInterface(unsafe.Pointer, unsafe.Pointer) interface{}
 
 //go:linkname decomposeInterface runtime.decomposeInterface
-func decomposeInterface(i interface{}) (rawType, unsafe.Pointer)
+func decomposeInterface(i interface{}) (unsafe.Pointer, unsafe.Pointer)
 
 func ValueOf(i interface{}) Value {
 	typecode, value := decomposeInterface(i)
 	return Value{
-		typecode: typecode,
+		typecode: (*rawType)(typecode),
 		value:    value,
 		flags:    valueFlagExported,
 	}
@@ -85,7 +85,7 @@ func valueInterfaceUnsafe(v Value) interface{} {
 		}
 		v.value = unsafe.Pointer(value)
 	}
-	return composeInterface(v.typecode, v.value)
+	return composeInterface(unsafe.Pointer(v.typecode), v.value)
 }
 
 func (v Value) Type() Type {
@@ -136,7 +136,7 @@ func (v Value) IsZero() bool {
 //
 // RawType returns the raw, underlying type code. It is used in the runtime
 // package and needs to be exported for the runtime package to access it.
-func (v Value) RawType() rawType {
+func (v Value) RawType() *rawType {
 	return v.typecode
 }
 
@@ -205,7 +205,7 @@ func (v Value) pointer() unsafe.Pointer {
 }
 
 func (v Value) IsValid() bool {
-	return v.typecode != 0
+	return v.typecode != nil
 }
 
 func (v Value) CanInterface() bool {
@@ -453,7 +453,7 @@ func (v Value) Elem() Value {
 	case Interface:
 		typecode, value := decomposeInterface(*(*interface{})(v.value))
 		return Value{
-			typecode: typecode,
+			typecode: (*rawType)(typecode),
 			value:    value,
 			flags:    v.flags &^ valueFlagIndirect,
 		}
@@ -523,6 +523,8 @@ func (v Value) Field(i int) Value {
 	}
 }
 
+var uint8Type = TypeOf(uint8(0)).(*rawType)
+
 func (v Value) Index(i int) Value {
 	switch v.Kind() {
 	case Slice:
@@ -550,7 +552,7 @@ func (v Value) Index(i int) Value {
 			panic("reflect: string index out of range")
 		}
 		return Value{
-			typecode: Uint8.basicType(),
+			typecode: uint8Type,
 			value:    unsafe.Pointer(uintptr(*(*uint8)(unsafe.Pointer(uintptr(s.data) + uintptr(i))))),
 			flags:    v.flags & valueFlagExported,
 		}
@@ -639,30 +641,119 @@ func (v Value) OverflowFloat(x float64) bool {
 }
 
 func (v Value) MapKeys() []Value {
-	panic("unimplemented: (reflect.Value).MapKeys()")
+	if v.Kind() != Map {
+		panic(&ValueError{Method: "MapKeys", Kind: v.Kind()})
+	}
+
+	// empty map
+	if v.Len() == 0 {
+		return nil
+	}
+
+	keys := make([]Value, 0, v.Len())
+
+	it := hashmapNewIterator()
+	k := New(v.typecode.Key())
+	e := New(v.typecode.Elem())
+
+	for hashmapNext(v.pointer(), it, k.value, e.value) {
+		keys = append(keys, k.Elem())
+		k = New(v.typecode.Key())
+	}
+
+	return keys
 }
 
+//go:linkname hashmapStringGet runtime.hashmapStringGetUnsafePointer
+func hashmapStringGet(m unsafe.Pointer, key string, value unsafe.Pointer, valueSize uintptr) bool
+
+//go:linkname hashmapBinaryGet runtime.hashmapBinaryGetUnsafePointer
+func hashmapBinaryGet(m unsafe.Pointer, key, value unsafe.Pointer, valueSize uintptr) bool
+
 func (v Value) MapIndex(key Value) Value {
+	if v.Kind() != Map {
+		panic(&ValueError{Method: "MapIndex", Kind: v.Kind()})
+	}
+
+	// compare key type with actual key type of map
+	if key.typecode != v.typecode.key() {
+		// type error?
+		panic("reflect.Value.MapIndex: incompatible types for key")
+	}
+
+	elemType := v.typecode.Elem()
+	elem := New(elemType)
+
+	if key.Kind() == String {
+		if ok := hashmapStringGet(v.pointer(), *(*string)(key.value), elem.value, elemType.Size()); !ok {
+			return Value{}
+		}
+		return elem.Elem()
+	} else if key.typecode.isBinary() {
+		var keyptr unsafe.Pointer
+		if key.isIndirect() || key.typecode.Size() > unsafe.Sizeof(uintptr(0)) {
+			keyptr = key.value
+		} else {
+			keyptr = unsafe.Pointer(&key.value)
+		}
+		//TODO(dgryski): zero out padding bytes in key, if any
+		if ok := hashmapBinaryGet(v.pointer(), keyptr, elem.value, elemType.Size()); !ok {
+			return Value{}
+		}
+		return elem.Elem()
+	}
+
+	// TODO(dgryski): Add other map types.  For now, just string and binary types are supported.
 	panic("unimplemented: (reflect.Value).MapIndex()")
 }
 
+//go:linkname hashmapNewIterator runtime.hashmapNewIterator
+func hashmapNewIterator() unsafe.Pointer
+
+//go:linkname hashmapNext runtime.hashmapNextUnsafePointer
+func hashmapNext(m unsafe.Pointer, it unsafe.Pointer, key, value unsafe.Pointer) bool
+
 func (v Value) MapRange() *MapIter {
-	panic("unimplemented: (reflect.Value).MapRange()")
+	if v.Kind() != Map {
+		panic(&ValueError{Method: "MapRange", Kind: v.Kind()})
+	}
+
+	return &MapIter{
+		m:   v,
+		it:  hashmapNewIterator(),
+		key: New(v.typecode.Key()),
+		val: New(v.typecode.Elem()),
+	}
 }
 
 type MapIter struct {
+	m   Value
+	it  unsafe.Pointer
+	key Value
+	val Value
+
+	valid bool
 }
 
 func (it *MapIter) Key() Value {
-	panic("unimplemented: (*reflect.MapIter).Key()")
+	if !it.valid {
+		panic("reflect.MapIter.Key called on invalid iterator")
+	}
+
+	return it.key.Elem()
 }
 
 func (it *MapIter) Value() Value {
-	panic("unimplemented: (*reflect.MapIter).Value()")
+	if !it.valid {
+		panic("reflect.MapIter.Value called on invalid iterator")
+	}
+
+	return it.val.Elem()
 }
 
 func (it *MapIter) Next() bool {
-	panic("unimplemented: (*reflect.MapIter).Next()")
+	it.valid = hashmapNext(it.m.pointer(), it.it, it.key.value, it.val.value)
+	return it.valid
 }
 
 func (v Value) Set(x Value) {
@@ -803,7 +894,7 @@ func Zero(typ Type) Value {
 // new value of the given type.
 func New(typ Type) Value {
 	return Value{
-		typecode: PtrTo(typ).(rawType),
+		typecode: PtrTo(typ).(*rawType),
 		value:    alloc(typ.Size(), nil),
 		flags:    valueFlagExported,
 	}
@@ -901,8 +992,70 @@ func AppendSlice(s, t Value) Value {
 	}
 }
 
+//go:linkname hashmapStringSet runtime.hashmapStringSetUnsafePointer
+func hashmapStringSet(m unsafe.Pointer, key string, value unsafe.Pointer)
+
+//go:linkname hashmapBinarySet runtime.hashmapBinarySetUnsafePointer
+func hashmapBinarySet(m unsafe.Pointer, key, value unsafe.Pointer)
+
+//go:linkname hashmapStringDelete runtime.hashmapStringDeleteUnsafePointer
+func hashmapStringDelete(m unsafe.Pointer, key string)
+
+//go:linkname hashmapBinaryDelete runtime.hashmapBinaryDeleteUnsafePointer
+func hashmapBinaryDelete(m unsafe.Pointer, key unsafe.Pointer)
+
 func (v Value) SetMapIndex(key, elem Value) {
-	panic("unimplemented: (reflect.Value).SetMapIndex()")
+	if v.Kind() != Map {
+		panic(&ValueError{Method: "SetMapIndex", Kind: v.Kind()})
+	}
+
+	// compare key type with actual key type of map
+	if key.typecode != v.typecode.key() {
+		panic("reflect.Value.SetMapIndex: incompatible types for key")
+	}
+
+	// if elem is the zero Value, it means delete
+	del := elem == Value{}
+
+	if !del && elem.typecode != v.typecode.elem() {
+		panic("reflect.Value.SetMapIndex: incompatible types for value")
+	}
+
+	if key.Kind() == String {
+		if del {
+			hashmapStringDelete(v.pointer(), *(*string)(key.value))
+		} else {
+			var elemptr unsafe.Pointer
+			if elem.isIndirect() || elem.typecode.Size() > unsafe.Sizeof(uintptr(0)) {
+				elemptr = elem.value
+			} else {
+				elemptr = unsafe.Pointer(&elem.value)
+			}
+			hashmapStringSet(v.pointer(), *(*string)(key.value), elemptr)
+		}
+
+	} else if key.typecode.isBinary() {
+		var keyptr unsafe.Pointer
+		if key.isIndirect() || key.typecode.Size() > unsafe.Sizeof(uintptr(0)) {
+			keyptr = key.value
+		} else {
+			keyptr = unsafe.Pointer(&key.value)
+		}
+
+		if del {
+			hashmapBinaryDelete(v.pointer(), keyptr)
+		} else {
+			var elemptr unsafe.Pointer
+			if elem.isIndirect() || elem.typecode.Size() > unsafe.Sizeof(uintptr(0)) {
+				elemptr = elem.value
+			} else {
+				elemptr = unsafe.Pointer(&elem.value)
+			}
+			hashmapBinarySet(v.pointer(), keyptr, elemptr)
+		}
+	} else {
+		panic("unimplemented: (reflect.Value).MapIndex()")
+	}
 }
 
 // FieldByIndex returns the nested field corresponding to index.
@@ -919,9 +1072,53 @@ func (v Value) FieldByName(name string) Value {
 	panic("unimplemented: (reflect.Value).FieldByName()")
 }
 
+//go:linkname hashmapMake runtime.hashmapMakeUnsafePointer
+func hashmapMake(keySize, valueSize uintptr, sizeHint uintptr, alg uint8) unsafe.Pointer
+
+// MakeMapWithSize creates a new map with the specified type and initial space
+// for approximately n elements.
+func MakeMapWithSize(typ Type, n int) Value {
+
+	// TODO(dgryski): deduplicate these?  runtime and reflect both need them.
+	const (
+		hashmapAlgorithmBinary uint8 = iota
+		hashmapAlgorithmString
+		hashmapAlgorithmInterface
+	)
+
+	if typ.Kind() != Map {
+		panic(&ValueError{"MakeMap", typ.Kind()})
+	}
+
+	if n < 0 {
+		panic("reflect.MakeMapWithSize: negative size hint")
+	}
+
+	key := typ.Key().(*rawType)
+	val := typ.Elem().(*rawType)
+
+	var alg uint8
+
+	if key.Kind() == String {
+		alg = hashmapAlgorithmString
+	} else if key.isBinary() {
+		alg = hashmapAlgorithmBinary
+	} else {
+		panic("reflect.MakeMap: unimplemented key type")
+	}
+
+	m := hashmapMake(key.Size(), val.Size(), uintptr(n), alg)
+
+	return Value{
+		typecode: typ.(*rawType),
+		value:    m,
+		flags:    valueFlagExported,
+	}
+}
+
 // MakeMap creates a new map with the specified type.
 func MakeMap(typ Type) Value {
-	panic("unimplemented: reflect.MakeMap()")
+	return MakeMapWithSize(typ, 8)
 }
 
 func (v Value) Call(in []Value) []Value {

src/reflect/value_test.go

@@ -2,6 +2,7 @@ package reflect_test
 
 import (
 	. "reflect"
+	"sort"
 	"testing"
 )
 
@@ -30,3 +31,101 @@ func TestIndirectPointers(t *testing.T) {
 		t.Errorf("bad indirect array index via reflect")
 	}
 }
+
+func TestMap(t *testing.T) {
+
+	m := make(map[string]int)
+
+	mtyp := TypeOf(m)
+
+	if got, want := mtyp.Key().Kind().String(), "string"; got != want {
+		t.Errorf("m.Type().Key().String()=%q, want %q", got, want)
+	}
+
+	if got, want := mtyp.Elem().Kind().String(), "int"; got != want {
+		t.Errorf("m.Elem().String()=%q, want %q", got, want)
+	}
+
+	m["foo"] = 2
+
+	mref := ValueOf(m)
+	two := mref.MapIndex(ValueOf("foo"))
+
+	if got, want := two.Interface().(int), 2; got != want {
+		t.Errorf("MapIndex(`foo`)=%v, want %v", got, want)
+	}
+
+	m["bar"] = 3
+	m["baz"] = 4
+	m["qux"] = 5
+
+	it := mref.MapRange()
+
+	var gotKeys []string
+	for it.Next() {
+		k := it.Key()
+		v := it.Value()
+
+		kstr := k.Interface().(string)
+		vint := v.Interface().(int)
+
+		gotKeys = append(gotKeys, kstr)
+
+		if m[kstr] != vint {
+			t.Errorf("m[%v]=%v, want %v", kstr, vint, m[kstr])
+		}
+	}
+	var wantKeys []string
+	for k := range m {
+		wantKeys = append(wantKeys, k)
+	}
+	sort.Strings(gotKeys)
+	sort.Strings(wantKeys)
+
+	if !equal(gotKeys, wantKeys) {
+		t.Errorf("MapRange return unexpected keys: got %v, want %v", gotKeys, wantKeys)
+	}
+
+	refMapKeys := mref.MapKeys()
+	gotKeys = gotKeys[:0]
+	for _, v := range refMapKeys {
+		gotKeys = append(gotKeys, v.Interface().(string))
+	}
+
+	sort.Strings(gotKeys)
+	if !equal(gotKeys, wantKeys) {
+		t.Errorf("MapKeys return unexpected keys: got %v, want %v", gotKeys, wantKeys)
+	}
+
+	mref.SetMapIndex(ValueOf("bar"), Value{})
+	if _, ok := m["bar"]; ok {
+		t.Errorf("SetMapIndex failed to delete `bar`")
+	}
+
+	mref.SetMapIndex(ValueOf("baz"), ValueOf(6))
+	if got, want := m["baz"], 6; got != want {
+		t.Errorf("SetMapIndex(bar, 6) got %v, want %v", got, want)
+	}
+
+	m2ref := MakeMap(mref.Type())
+	m2ref.SetMapIndex(ValueOf("foo"), ValueOf(2))
+
+	m2 := m2ref.Interface().(map[string]int)
+
+	if m2["foo"] != 2 {
+		t.Errorf("MakeMap failed to create map")
+	}
+}
+
+func equal[T comparable](a, b []T) bool {
+	if len(a) != len(b) {
+		return false
+	}
+
+	for i, aa := range a {
+		if b[i] != aa {
+			return false
+		}
+	}
+	return true
+}

src/runtime/asm_arm64.S

@@ -4,7 +4,6 @@ _tinygo_scanCurrentStack:
 #else
 .section .text.tinygo_scanCurrentStack
 .global tinygo_scanCurrentStack
-.type tinygo_scanCurrentStack, %function
 tinygo_scanCurrentStack:
 #endif
     // Sources:
@@ -12,12 +11,16 @@ tinygo_scanCurrentStack:
     //   * https://godbolt.org/z/qrvrEh
 
     // Save callee-saved registers.
-    stp     x29, x30, [sp, #-96]!
+    stp     x29, x30, [sp, #-160]!
     stp     x28, x27, [sp, #16]
     stp     x26, x25, [sp, #32]
     stp     x24, x23, [sp, #48]
     stp     x22, x21, [sp, #64]
     stp     x20, x19, [sp, #80]
+    stp     d8,  d9,  [sp, #96]
+    stp     d10, d11, [sp, #112]
+    stp     d12, d13, [sp, #128]
+    stp     d14, d15, [sp, #144]
 
     // Scan the stack.
     mov     x0, sp
@@ -28,7 +31,7 @@ tinygo_scanCurrentStack:
 #endif
 
     // Restore stack state and return.
-    ldp     x29, x30, [sp], #96
+    ldp     x29, x30, [sp], #160
     ret
 
 
@@ -38,7 +41,6 @@ _tinygo_longjmp:
 #else
 .section .text.tinygo_longjmp
 .global tinygo_longjmp
-.type tinygo_longjmp, %function
 tinygo_longjmp:
 #endif
     // Note: the code we jump to assumes x0 is set to a non-zero value if we

src/runtime/asm_arm64_windows.S

@@ -1,36 +0,0 @@
-.section .text.tinygo_scanCurrentStack,"ax"
-.global tinygo_scanCurrentStack
-tinygo_scanCurrentStack:
-    // Sources:
-    //   * https://learn.microsoft.com/en-us/cpp/build/arm64-windows-abi-conventions?view=msvc-170
-    //   * https://godbolt.org/z/foc1xncvb
-
-    // Save callee-saved registers.
-    stp     x29, x30, [sp, #-160]!
-    stp     x28, x27, [sp, #16]
-    stp     x26, x25, [sp, #32]
-    stp     x24, x23, [sp, #48]
-    stp     x22, x21, [sp, #64]
-    stp     x20, x19, [sp, #80]
-    stp     d8,  d9,  [sp, #96]
-    stp     d10, d11, [sp, #112]
-    stp     d12, d13, [sp, #128]
-    stp     d14, d15, [sp, #144]
-
-    // Scan the stack.
-    mov     x0, sp
-    bl      tinygo_scanstack
-
-    // Restore stack state and return.
-    ldp     x29, x30, [sp], #160
-    ret
-
-
-.section .text.tinygo_longjmp,"ax"
-.global tinygo_longjmp
-tinygo_longjmp:
-    // Note: the code we jump to assumes x0 is set to a non-zero value if we
-    // jump from here (which is conveniently already the case).
-    ldp x1, x2, [x0] // jumpSP, jumpPC
-    mov sp, x1
-    br  x2

src/runtime/gc_blocks.go

@@ -277,6 +277,10 @@ func alloc(size uintptr, layout unsafe.Pointer) unsafe.Pointer {
 		size += align(unsafe.Sizeof(layout))
 	}
 
+	if interrupt.In() {
+		runtimePanic("alloc in interrupt")
+	}
+
 	gcTotalAlloc += uint64(size)
 	gcMallocs++
 
@@ -687,3 +691,7 @@ func ReadMemStats(m *MemStats) {
 	m.Frees = gcFrees
 	m.Sys = uint64(heapEnd - heapStart)
 }
+
+func SetFinalizer(obj interface{}, finalizer interface{}) {
+	// Unimplemented.
+}

src/runtime/gc_custom.go

@@ -20,6 +20,7 @@ package runtime
 // - func free(ptr unsafe.Pointer)
 // - func markRoots(start, end uintptr)
 // - func GC()
+// - func SetFinalizer(obj interface{}, finalizer interface{})
 // - func ReadMemStats(ms *runtime.MemStats)
 //
 //
@@ -51,6 +52,9 @@ func markRoots(start, end uintptr)
 // GC is called to explicitly run garbage collection.
 func GC()
 
+// SetFinalizer registers a finalizer.
+func SetFinalizer(obj interface{}, finalizer interface{})
+
 // ReadMemStats populates m with memory statistics.
 func ReadMemStats(ms *MemStats)
 

src/runtime/gc_leaking.go

@@ -85,6 +85,10 @@ func GC() {
 	// No-op.
 }
 
+func SetFinalizer(obj interface{}, finalizer interface{}) {
+	// No-op.
+}
+
 func initHeap() {
 	// preinit() may have moved heapStart; reset heapptr
 	heapptr = heapStart

src/runtime/gc_none.go

@@ -26,6 +26,10 @@ func GC() {
 	// Unimplemented.
 }
 
+func SetFinalizer(obj interface{}, finalizer interface{}) {
+	// Unimplemented.
+}
+
 func initHeap() {
 	// Nothing to initialize.
 }

src/runtime/hashmap.go

@@ -49,6 +49,10 @@ type hashmapIterator struct {
 	bucketIndex  uint8          // current index into bucket
 }
 
+func hashmapNewIterator() unsafe.Pointer {
+	return unsafe.Pointer(new(hashmapIterator))
+}
+
 // Get the topmost 8 bits of the hash, without using a special value (like 0).
 func hashmapTopHash(hash uint32) uint8 {
 	tophash := uint8(hash >> 24)
@@ -84,6 +88,10 @@ func hashmapMake(keySize, valueSize uintptr, sizeHint uintptr, alg uint8) *hashm
 	}
 }
 
+func hashmapMakeUnsafePointer(keySize, valueSize uintptr, sizeHint uintptr, alg uint8) unsafe.Pointer {
+	return (unsafe.Pointer)(hashmapMake(keySize, valueSize, sizeHint, alg))
+}
+
 func hashmapKeyEqualAlg(alg hashmapAlgorithm) func(x, y unsafe.Pointer, n uintptr) bool {
 	switch alg {
 	case hashmapAlgorithmBinary:
@@ -142,10 +150,8 @@ func hashmapLen(m *hashmap) int {
 	return int(m.count)
 }
 
-// wrapper for use in reflect
-func hashmapLenUnsafePointer(p unsafe.Pointer) int {
-	m := (*hashmap)(p)
-	return hashmapLen(m)
+func hashmapLenUnsafePointer(m unsafe.Pointer) int {
+	return hashmapLen((*hashmap)(m))
 }
 
 // Set a specified key to a given value. Grow the map if necessary.
@@ -208,6 +214,10 @@ func hashmapSet(m *hashmap, key unsafe.Pointer, value unsafe.Pointer, hash uint3
 	*emptySlotTophash = tophash
 }
 
+func hashmapSetUnsafePointer(m unsafe.Pointer, key unsafe.Pointer, value unsafe.Pointer, hash uint32) {
+	hashmapSet((*hashmap)(m), key, value, hash)
+}
+
 // hashmapInsertIntoNewBucket creates a new bucket, inserts the given key and
 // value into the bucket, and returns a pointer to this bucket.
 func hashmapInsertIntoNewBucket(m *hashmap, key, value unsafe.Pointer, tophash uint8) *hashmapBucket {
@@ -299,6 +309,10 @@ func hashmapGet(m *hashmap, key, value unsafe.Pointer, valueSize uintptr, hash u
 	return false
 }
 
+func hashmapGetUnsafePointer(m unsafe.Pointer, key, value unsafe.Pointer, valueSize uintptr, hash uint32) bool {
+	return hashmapGet((*hashmap)(m), key, value, valueSize, hash)
+}
+
 // Delete a given key from the map. No-op when the key does not exist in the
 // map.
 //
@@ -409,6 +423,10 @@ func hashmapNext(m *hashmap, it *hashmapIterator, key, value unsafe.Pointer) boo
 	}
 }
 
+func hashmapNextUnsafePointer(m unsafe.Pointer, it unsafe.Pointer, key, value unsafe.Pointer) bool {
+	return hashmapNext((*hashmap)(m), (*hashmapIterator)(it), key, value)
+}
+
 // Hashmap with plain binary data keys (not containing strings etc.).
 func hashmapBinarySet(m *hashmap, key, value unsafe.Pointer) {
 	if m == nil {
@@ -418,6 +436,10 @@ func hashmapBinarySet(m *hashmap, key, value unsafe.Pointer) {
 	hashmapSet(m, key, value, hash)
 }
 
+func hashmapBinarySetUnsafePointer(m unsafe.Pointer, key, value unsafe.Pointer) {
+	hashmapBinarySet((*hashmap)(m), key, value)
+}
+
 func hashmapBinaryGet(m *hashmap, key, value unsafe.Pointer, valueSize uintptr) bool {
 	if m == nil {
 		memzero(value, uintptr(valueSize))
@@ -427,6 +449,10 @@ func hashmapBinaryGet(m *hashmap, key, value unsafe.Pointer, valueSize uintptr)
 	return hashmapGet(m, key, value, valueSize, hash)
 }
 
+func hashmapBinaryGetUnsafePointer(m unsafe.Pointer, key, value unsafe.Pointer, valueSize uintptr) bool {
+	return hashmapBinaryGet((*hashmap)(m), key, value, valueSize)
+}
+
 func hashmapBinaryDelete(m *hashmap, key unsafe.Pointer) {
 	if m == nil {
 		return
@@ -435,6 +461,10 @@ func hashmapBinaryDelete(m *hashmap, key unsafe.Pointer) {
 	hashmapDelete(m, key, hash)
 }
 
+func hashmapBinaryDeleteUnsafePointer(m unsafe.Pointer, key unsafe.Pointer) {
+	hashmapBinaryDelete((*hashmap)(m), key)
+}
+
 // Hashmap with string keys (a common case).
 
 func hashmapStringEqual(x, y unsafe.Pointer, n uintptr) bool {
@@ -459,6 +489,10 @@ func hashmapStringSet(m *hashmap, key string, value unsafe.Pointer) {
 	hashmapSet(m, unsafe.Pointer(&key), value, hash)
 }
 
+func hashmapStringSetUnsafePointer(m unsafe.Pointer, key string, value unsafe.Pointer) {
+	hashmapStringSet((*hashmap)(m), key, value)
+}
+
 func hashmapStringGet(m *hashmap, key string, value unsafe.Pointer, valueSize uintptr) bool {
 	if m == nil {
 		memzero(value, uintptr(valueSize))
@@ -468,6 +502,10 @@ func hashmapStringGet(m *hashmap, key string, value unsafe.Pointer, valueSize ui
 	return hashmapGet(m, unsafe.Pointer(&key), value, valueSize, hash)
 }
 
+func hashmapStringGetUnsafePointer(m unsafe.Pointer, key string, value unsafe.Pointer, valueSize uintptr) bool {
+	return hashmapStringGet((*hashmap)(m), key, value, valueSize)
+}
+
 func hashmapStringDelete(m *hashmap, key string) {
 	if m == nil {
 		return
@@ -476,6 +514,10 @@ func hashmapStringDelete(m *hashmap, key string) {
 	hashmapDelete(m, unsafe.Pointer(&key), hash)
 }
 
+func hashmapStringDeleteUnsafePointer(m unsafe.Pointer, key string) {
+	hashmapStringDelete((*hashmap)(m), key)
+}
+
 // Hashmap with interface keys (for everything else).
 
 // This is a method that is intentionally unexported in the reflect package. It
@@ -506,7 +548,7 @@ func hashmapFloat64Hash(ptr unsafe.Pointer, seed uintptr) uint32 {
 
 func hashmapInterfaceHash(itf interface{}, seed uintptr) uint32 {
 	x := reflect.ValueOf(itf)
-	if x.RawType() == 0 {
+	if x.RawType() == nil {
 		return 0 // nil interface
 	}
 
@@ -563,7 +605,7 @@ func hashmapInterfaceHash(itf interface{}, seed uintptr) uint32 {
 }
 
 func hashmapInterfacePtrHash(iptr unsafe.Pointer, size uintptr, seed uintptr) uint32 {
-	_i := *(*_interface)(iptr)
+	_i := *(*interface{})(iptr)
 	return hashmapInterfaceHash(_i, seed)
 }
 

src/runtime/interface.go

@@ -11,17 +11,17 @@ import (
 )
 
 type _interface struct {
-	typecode uintptr
+	typecode unsafe.Pointer
 	value    unsafe.Pointer
 }
 
 //go:inline
-func composeInterface(typecode uintptr, value unsafe.Pointer) _interface {
+func composeInterface(typecode, value unsafe.Pointer) _interface {
 	return _interface{typecode, value}
 }
 
 //go:inline
-func decomposeInterface(i _interface) (uintptr, unsafe.Pointer) {
+func decomposeInterface(i _interface) (unsafe.Pointer, unsafe.Pointer) {
 	return i.typecode, i.value
 }
 
@@ -34,7 +34,7 @@ func reflectValueEqual(x, y reflect.Value) bool {
 	// Note: doing a x.Type() == y.Type() comparison would not work here as that
 	// would introduce an infinite recursion: comparing two reflect.Type values
 	// is done with this reflectValueEqual runtime call.
-	if x.RawType() == 0 || y.RawType() == 0 {
+	if x.RawType() == nil || y.RawType() == nil {
 		// One of them is nil.
 		return x.RawType() == y.RawType()
 	}
@@ -94,48 +94,13 @@ func interfaceTypeAssert(ok bool) {
 // lowered to inline IR in the interface lowering pass.
 // See compiler/interface-lowering.go for details.
 
-type interfaceMethodInfo struct {
-	signature *uint8  // external *i8 with a name identifying the Go function signature
-	funcptr   uintptr // bitcast from the actual function pointer
-}
-
-type typecodeID struct {
-	// Depending on the type kind of this typecodeID, this pointer is something
-	// different:
-	// * basic types: null
-	// * named type: the underlying type
-	// * interface: null
-	// * chan/pointer/slice/array: the element type
-	// * struct: bitcast of global with structField array
-	// * func/map: TODO
-	references *typecodeID
-
-	// The array length, for array types.
-	length uintptr
-
-	methodSet *interfaceMethodInfo // nil or a GEP of an array
-
-	// The type that's a pointer to this type, nil if it is already a pointer.
-	// Keeping the type struct alive here is important so that values from
-	// reflect.New (which uses reflect.PtrTo) can be used in type asserts etc.
-	ptrTo *typecodeID
-
-	// typeAssert is a ptrtoint of a declared interface assert function.
-	// It only exists to make the rtcalls pass easier.
-	typeAssert uintptr
-}
-
-// structField is used by the compiler to pass information to the interface
-// lowering pass. It is not used in the final binary.
 type structField struct {
-	typecode *typecodeID // type of this struct field
-	name     *uint8      // pointer to char array
-	tag      *uint8      // pointer to char array, or nil
-	embedded bool
+	typecode unsafe.Pointer // type of this struct field
+	data     *uint8         // pointer to byte array containing name, tag, and 'embedded' flag
 }
 
 // Pseudo function call used during a type assert. It is used during interface
 // lowering, to assign the lowest type numbers to the types with the most type
 // asserts. Also, it is replaced with const false if this type assert can never
 // happen.
-func typeAssert(actualType uintptr, assertedType *uint8) bool
+func typeAssert(actualType unsafe.Pointer, assertedType *uint8) bool

src/runtime/interrupt/interrupt_avr.go

@@ -34,3 +34,12 @@ func Restore(state State) {
 		"state": state,
 	})
 }
+
+// In returns whether the system is currently in an interrupt.
+//
+// Warning: this always returns false on AVR, as there does not appear to be a
+// reliable way to determine whether we're currently running inside an interrupt
+// handler.
+func In() bool {
+	return false
+}

src/runtime/interrupt/interrupt_cortexm.go

@@ -50,3 +50,13 @@ func Disable() (state State) {
 func Restore(state State) {
 	arm.EnableInterrupts(uintptr(state))
 }
+
+// In returns whether the system is currently in an interrupt.
+func In() bool {
+	// The VECTACTIVE field gives the instruction vector that is currently
+	// active (in handler mode), or 0 if not in an interrupt.
+	// Documentation:
+	// https://developer.arm.com/documentation/dui0497/a/cortex-m0-peripherals/system-control-block/interrupt-control-and-state-register
+	vectactive := uint8(arm.SCB.ICSR.Get())
+	return vectactive != 0
+}

src/runtime/interrupt/interrupt_gameboyadvance.go

@@ -2,6 +2,8 @@
 
 package interrupt
 
+// This is good documentation of the GBA: https://www.akkit.org/info/gbatek.htm
+
 import (
 	"runtime/volatile"
 	"unsafe"
@@ -36,8 +38,11 @@ func (irq Interrupt) Enable() {
 	regInterruptEnable.SetBits(1 << uint(irq.num))
 }
 
+var inInterrupt bool
+
 //export handleInterrupt
 func handleInterrupt() {
+	inInterrupt = true
 	flags := regInterruptRequestFlags.Get()
 	for i := 0; i < 14; i++ {
 		if flags&(1<<uint(i)) != 0 {
@@ -45,6 +50,7 @@ func handleInterrupt() {
 			callInterruptHandler(i)
 		}
 	}
+	inInterrupt = false
 }
 
 // Pseudo function call that is replaced by the compiler with the actual
@@ -115,3 +121,8 @@ func Restore(state State) {
 	// Restore interrupts to the previous state.
 	regGlobalInterruptEnable.Set(uint16(state))
 }
+
+// In returns whether the system is currently in an interrupt.
+func In() bool {
+	return inInterrupt
+}

src/runtime/interrupt/interrupt_none.go

@@ -23,3 +23,9 @@ func Disable() (state State) {
 // calling Disable, this will not re-enable interrupts, allowing for nested
 // cricital sections.
 func Restore(state State) {}
+
+// In returns whether the system is currently in an interrupt.
+func In() bool {
+	// There are no interrupts, so it can't be in one.
+	return false
+}

src/runtime/interrupt/interrupt_tinygoriscv.go

@@ -27,3 +27,12 @@ func Disable() (state State) {
 func Restore(state State) {
 	riscv.EnableInterrupts(uintptr(state))
 }
+
+// In returns whether the system is currently in an interrupt.
+func In() bool {
+	// There is one exception that has the value 0 (instruction address
+	// misaligned), but it's not very likely and even if it happens, it's not
+	// really something that can be recovered from. Therefore I think it's safe
+	// to ignore it. It's handled specially (in handleException).
+	return riscv.MCAUSE.Get() != 0
+}

src/runtime/interrupt/interrupt_xtensa.go

@@ -29,3 +29,11 @@ func Restore(state State) {
 		"state": state,
 	})
 }
+
+// In returns whether the system is currently in an interrupt.
+//
+// Warning: interrupts have not been implemented for Xtensa yet so this always
+// returns false.
+func In() bool {
+	return false
+}

src/runtime/runtime.go

@@ -88,13 +88,9 @@ func LockOSThread() {
 func UnlockOSThread() {
 }
 
-func KeepAlive(x interface{}) {
-	// Unimplemented. Only required with SetFinalizer().
-}
-
-func SetFinalizer(obj interface{}, finalizer interface{}) {
-	// Unimplemented.
-}
+// KeepAlive makes sure the value in the interface is alive until at least the
+// point of the call.
+func KeepAlive(x interface{})
 
 var godebugUpdate func(string, string)
 

src/runtime/runtime_atsamd51.go

@@ -27,6 +27,7 @@ func init() {
 	initSERCOMClocks()
 	initUSBClock()
 	initADCClock()
+	enableCache()
 
 	cdc.EnableUSBCDC()
 	machine.USBDev.Configure(machine.UARTConfig{})
@@ -367,6 +368,10 @@ func initADCClock() {
 		sam.GCLK_PCHCTRL_CHEN)
 }
 
+func enableCache() {
+	sam.CMCC.CTRL.SetBits(sam.CMCC_CTRL_CEN)
+}
+
 func waitForEvents() {
 	arm.Asm("wfe")
 }

src/runtime/runtime_fe310.go

@@ -25,6 +25,12 @@ func main() {
 	// Zero the threshold value to allow all priorities of interrupts.
 	sifive.PLIC.THRESHOLD.Set(0)
 
+	// Zero MCAUSE, which is set to the reset reason on reset. It must be zeroed
+	// to make interrupt.In() work.
+	// This would also be a good time to save the reset reason, but that hasn't
+	// been implemented yet.
+	riscv.MCAUSE.Set(0)
+
 	// Set the interrupt address.
 	// Note that this address must be aligned specially, otherwise the MODE bits
 	// of MTVEC won't be zero.
@@ -73,6 +79,10 @@ func handleInterrupt() {
 		// misaligned loads). However, for now we'll just print a fatal error.
 		handleException(code)
 	}
+
+	// Zero MCAUSE so that it can later be used to see whether we're in an
+	// interrupt or not.
+	riscv.MCAUSE.Set(0)
 }
 
 // initPeripherals configures periperhals the way the runtime expects them.

src/runtime/runtime_k210.go

@@ -31,6 +31,12 @@ func main() {
 		kendryte.PLIC.PRIORITY[i].Set(0)
 	}
 
+	// Zero MCAUSE, which is set to the reset reason on reset. It must be zeroed
+	// to make interrupt.In() work.
+	// This would also be a good time to save the reset reason, but that hasn't
+	// been implemented yet.
+	riscv.MCAUSE.Set(0)
+
 	// Set the interrupt address.
 	// Note that this address must be aligned specially, otherwise the MODE bits
 	// of MTVEC won't be zero.
@@ -93,6 +99,10 @@ func handleInterrupt() {
 		// misaligned loads). However, for now we'll just print a fatal error.
 		handleException(code)
 	}
+
+	// Zero MCAUSE so that it can later be used to see whether we're in an
+	// interrupt or not.
+	riscv.MCAUSE.Set(0)
 }
 
 // initPeripherals configures periperhals the way the runtime expects them.

targets/arm.ld

@@ -69,3 +69,7 @@ _heap_start = _ebss;
 _heap_end = ORIGIN(RAM) + LENGTH(RAM);
 _globals_start = _sdata;
 _globals_end = _ebss;
+
+/* For the flash API */
+__flash_data_start = LOADADDR(.data) + SIZEOF(.data);
+__flash_data_end = ORIGIN(FLASH_TEXT) + LENGTH(FLASH_TEXT);

testdata/gc.go

@@ -1,5 +1,7 @@
 package main
 
+import "runtime"
+
 var xorshift32State uint32 = 1
 
 func xorshift32(x uint32) uint32 {
@@ -17,6 +19,7 @@ func randuint32() uint32 {
 
 func main() {
 	testNonPointerHeap()
+	testKeepAlive()
 }
 
 var scalarSlices [4][]byte
@@ -64,3 +67,10 @@ func testNonPointerHeap() {
 	}
 	println("ok")
 }
+
+func testKeepAlive() {
+	// There isn't much we can test, but at least we can test that
+	// runtime.KeepAlive compiles correctly.
+	var x int
+	runtime.KeepAlive(&x)
+}

testdata/map.go

@@ -129,6 +129,8 @@ func main() {
 	floatcmplx()
 
 	mapgrow()
+
+	interfacerehash()
 }
 
 func floatcmplx() {
@@ -274,3 +276,35 @@ func mapgrow() {
 	}
 	println("done")
 }
+
+type Counter interface {
+	count() int
+}
+
+type counter struct {
+	i int
+}
+
+func (c *counter) count() int {
+	return c.i
+}
+
+func interfacerehash() {
+	m := make(map[Counter]int)
+
+	for i := 0; i < 20; i++ {
+		c := &counter{i}
+		m[c] = i
+	}
+
+	var failures int
+	for k, v := range m {
+		if got := m[k]; got != v {
+			println("lookup failure got", got, "want", v)
+			failures++
+		}
+	}
+	if failures == 0 {
+		println("no interface lookup failures")
+	}
+}

testdata/map.txt

@@ -80,3 +80,4 @@ tested growing of a map
 2
 2
 done
+no interface lookup failures

testdata/reflect.go

@@ -3,6 +3,7 @@ package main
 import (
 	"errors"
 	"reflect"
+	"strconv"
 	"unsafe"
 )
 
@@ -17,8 +18,8 @@ type (
 		Y int16
 	}
 	mystruct struct {
-		n    int `foo:"bar"`
-		some point
+		n    int   `foo:"bar"`
+		some point "some\x00tag"
 		zero struct{}
 		buf  []byte
 		Buf  []byte
@@ -480,7 +481,7 @@ func showValue(rv reflect.Value, indent string) {
 		for i := 0; i < rv.NumField(); i++ {
 			field := rt.Field(i)
 			println(indent+"  field:", i, field.Name)
-			println(indent+"  tag:", field.Tag)
+			println(indent+"  tag:", strconv.Quote(string(field.Tag)))
 			println(indent+"  embedded:", field.Anonymous)
 			println(indent+"  exported:", field.IsExported())
 			showValue(rv.Field(i), indent+"  ")

testdata/reflect.txt

@@ -233,7 +233,7 @@ reflect type: struct
 reflect type: struct
   struct: 1
   field: 0 error
-  tag: 
+  tag: ""
   embedded: true
   exported: false
   reflect type: interface caninterface=false
@@ -242,19 +242,19 @@ reflect type: struct
 reflect type: struct
   struct: 3
   field: 0 a
-  tag: 
+  tag: ""
   embedded: false
   exported: false
   reflect type: uint8 caninterface=false
     uint: 42
   field: 1 b
-  tag: 
+  tag: ""
   embedded: false
   exported: false
   reflect type: int16 caninterface=false
     int: 321
   field: 2 c
-  tag: 
+  tag: ""
   embedded: false
   exported: false
   reflect type: int8 caninterface=false
@@ -262,37 +262,37 @@ reflect type: struct
 reflect type: struct comparable=false
   struct: 5
   field: 0 n
-  tag: foo:"bar"
+  tag: "foo:\"bar\""
   embedded: false
   exported: false
   reflect type: int caninterface=false
     int: 5
   field: 1 some
-  tag: 
+  tag: "some\x00tag"
   embedded: false
   exported: false
   reflect type: struct caninterface=false
     struct: 2
     field: 0 X
-    tag: 
+    tag: ""
     embedded: false
     exported: true
     reflect type: int16 caninterface=false
       int: -5
     field: 1 Y
-    tag: 
+    tag: ""
     embedded: false
     exported: true
     reflect type: int16 caninterface=false
       int: 3
   field: 2 zero
-  tag: 
+  tag: ""
   embedded: false
   exported: false
   reflect type: struct caninterface=false
     struct: 0
   field: 3 buf
-  tag: 
+  tag: ""
   embedded: false
   exported: false
   reflect type: slice caninterface=false comparable=false
@@ -306,7 +306,7 @@ reflect type: struct comparable=false
     reflect type: uint8 addrable=true caninterface=false
       uint: 111
   field: 4 Buf
-  tag: 
+  tag: ""
   embedded: false
   exported: true
   reflect type: slice comparable=false
@@ -322,14 +322,14 @@ reflect type: ptr
   reflect type: struct settable=true addrable=true
     struct: 2
     field: 0 next
-    tag: description:"chain"
+    tag: "description:\"chain\""
     embedded: false
     exported: false
     reflect type: ptr addrable=true caninterface=false
       pointer: false struct
       nil: true
     field: 1 foo
-    tag: 
+    tag: ""
     embedded: false
     exported: false
     reflect type: int addrable=true caninterface=false
@@ -337,13 +337,13 @@ reflect type: ptr
 reflect type: struct
   struct: 2
   field: 0 A
-  tag: 
+  tag: ""
   embedded: false
   exported: true
   reflect type: uintptr
     uint: 2
   field: 1 B
-  tag: 
+  tag: ""
   embedded: false
   exported: true
   reflect type: uintptr

tools/gen-device-svd/gen-device-svd.go

@@ -90,9 +90,10 @@ type SVDCluster struct {
 }
 
 type Device struct {
-	Metadata    *Metadata
-	Interrupts  []*Interrupt
-	Peripherals []*Peripheral
+	Metadata       *Metadata
+	Interrupts     []*Interrupt
+	Peripherals    []*Peripheral
+	PeripheralDict map[string]*Peripheral
 }
 
 type Metadata struct {
@@ -191,6 +192,142 @@ func cleanName(text string) string {
 	return text
 }
 
+func processSubCluster(p *Peripheral, cluster *SVDCluster, clusterOffset uint64, clusterName string, peripheralDict map[string]*Peripheral) []*Peripheral {
+	var peripheralsList []*Peripheral
+	clusterPrefix := clusterName + "_"
+	cpRegisters := []*PeripheralField{}
+
+	for _, regEl := range cluster.Registers {
+		cpRegisters = append(cpRegisters, parseRegister(p.GroupName, regEl, p.BaseAddress+clusterOffset, clusterPrefix)...)
+	}
+	// handle sub-clusters of registers
+	for _, subClusterEl := range cluster.Clusters {
+		subclusterName := strings.ReplaceAll(subClusterEl.Name, "[%s]", "")
+		subclusterPrefix := subclusterName + "_"
+		subclusterOffset, err := strconv.ParseUint(subClusterEl.AddressOffset, 0, 32)
+		if err != nil {
+			panic(err)
+		}
+		subdim := *subClusterEl.Dim
+		subdimIncrement, err := strconv.ParseInt(subClusterEl.DimIncrement, 0, 32)
+		if err != nil {
+			panic(err)
+		}
+
+		if subdim > 1 {
+			subcpRegisters := []*PeripheralField{}
+			for _, regEl := range subClusterEl.Registers {
+				subcpRegisters = append(subcpRegisters, parseRegister(p.GroupName, regEl, p.BaseAddress+clusterOffset+subclusterOffset, subclusterPrefix)...)
+			}
+
+			cpRegisters = append(cpRegisters, &PeripheralField{
+				Name:        subclusterName,
+				Address:     p.BaseAddress + clusterOffset + subclusterOffset,
+				Description: subClusterEl.Description,
+				Registers:   subcpRegisters,
+				Array:       subdim,
+				ElementSize: int(subdimIncrement),
+				ShortName:   clusterPrefix + subclusterName,
+			})
+		} else {
+			for _, regEl := range subClusterEl.Registers {
+				cpRegisters = append(cpRegisters, parseRegister(regEl.Name, regEl, p.BaseAddress+clusterOffset+subclusterOffset, subclusterPrefix)...)
+			}
+		}
+	}
+
+	sort.SliceStable(cpRegisters, func(i, j int) bool {
+		return cpRegisters[i].Address < cpRegisters[j].Address
+	})
+	clusterPeripheral := &Peripheral{
+		Name:        p.Name + "_" + clusterName,
+		GroupName:   p.GroupName + "_" + clusterName,
+		Description: p.Description + " - " + clusterName,
+		ClusterName: clusterName,
+		BaseAddress: p.BaseAddress + clusterOffset,
+		Registers:   cpRegisters,
+	}
+	peripheralsList = append(peripheralsList, clusterPeripheral)
+	peripheralDict[clusterPeripheral.Name] = clusterPeripheral
+	p.Subtypes = append(p.Subtypes, clusterPeripheral)
+
+	return peripheralsList
+}
+
+func processCluster(p *Peripheral, clusters []*SVDCluster, peripheralDict map[string]*Peripheral) []*Peripheral {
+	var peripheralsList []*Peripheral
+	for _, cluster := range clusters {
+		clusterName := strings.ReplaceAll(cluster.Name, "[%s]", "")
+		if cluster.DimIndex != nil {
+			clusterName = strings.ReplaceAll(clusterName, "%s", "")
+		}
+		clusterPrefix := clusterName + "_"
+		clusterOffset, err := strconv.ParseUint(cluster.AddressOffset, 0, 32)
+		if err != nil {
+			panic(err)
+		}
+		var dim, dimIncrement int
+		if cluster.Dim == nil {
+			// Nordic SVD have sub-clusters with another sub-clusters.
+			if clusterOffset == 0 || len(cluster.Clusters) > 0 {
+				peripheralsList = append(peripheralsList, processSubCluster(p, cluster, clusterOffset, clusterName, peripheralDict)...)
+				continue
+			}
+			dim = -1
+			dimIncrement = -1
+		} else {
+			dim = *cluster.Dim
+			if dim == 1 {
+				dimIncrement = -1
+			} else {
+				inc, err := strconv.ParseUint(cluster.DimIncrement, 0, 32)
+				if err != nil {
+					panic(err)
+				}
+				dimIncrement = int(inc)
+			}
+		}
+		clusterRegisters := []*PeripheralField{}
+		for _, regEl := range cluster.Registers {
+			regName := p.GroupName
+			if regName == "" {
+				regName = p.Name
+			}
+			clusterRegisters = append(clusterRegisters, parseRegister(regName, regEl, p.BaseAddress+clusterOffset, clusterPrefix)...)
+		}
+		sort.SliceStable(clusterRegisters, func(i, j int) bool {
+			return clusterRegisters[i].Address < clusterRegisters[j].Address
+		})
+		if dimIncrement == -1 && len(clusterRegisters) > 0 {
+			lastReg := clusterRegisters[len(clusterRegisters)-1]
+			lastAddress := lastReg.Address
+			if lastReg.Array != -1 {
+				lastAddress = lastReg.Address + uint64(lastReg.Array*lastReg.ElementSize)
+			}
+			firstAddress := clusterRegisters[0].Address
+			dimIncrement = int(lastAddress - firstAddress)
+		}
+
+		if !unicode.IsUpper(rune(clusterName[0])) && !unicode.IsDigit(rune(clusterName[0])) {
+			clusterName = strings.ToUpper(clusterName)
+		}
+
+		p.Registers = append(p.Registers, &PeripheralField{
+			Name:        clusterName,
+			Address:     p.BaseAddress + clusterOffset,
+			Description: cluster.Description,
+			Registers:   clusterRegisters,
+			Array:       dim,
+			ElementSize: dimIncrement,
+			ShortName:   clusterName,
+		})
+	}
+	sort.SliceStable(p.Registers, func(i, j int) bool {
+		return p.Registers[i].Address < p.Registers[j].Address
+	})
+	return peripheralsList
+}
+
 // Read ARM SVD files.
 func readSVD(path, sourceURL string) (*Device, error) {
 	// Open the XML file.
@@ -293,133 +430,7 @@ func readSVD(path, sourceURL string) (*Device, error) {
 			}
 			p.Registers = append(p.Registers, parseRegister(regName, register, baseAddress, "")...)
 		}
-		for _, cluster := range periphEl.Clusters {
-			clusterName := strings.ReplaceAll(cluster.Name, "[%s]", "")
-			if cluster.DimIndex != nil {
-				clusterName = strings.ReplaceAll(clusterName, "%s", "")
-			}
-			clusterPrefix := clusterName + "_"
-			clusterOffset, err := strconv.ParseUint(cluster.AddressOffset, 0, 32)
-			if err != nil {
-				panic(err)
-			}
-			var dim, dimIncrement int
-			if cluster.Dim == nil {
-				if clusterOffset == 0 {
-					// make this a separate peripheral
-					cpRegisters := []*PeripheralField{}
-					for _, regEl := range cluster.Registers {
-						cpRegisters = append(cpRegisters, parseRegister(groupName, regEl, baseAddress, clusterName+"_")...)
-					}
-					// handle sub-clusters of registers
-					for _, subClusterEl := range cluster.Clusters {
-						subclusterName := strings.ReplaceAll(subClusterEl.Name, "[%s]", "")
-						subclusterPrefix := subclusterName + "_"
-						subclusterOffset, err := strconv.ParseUint(subClusterEl.AddressOffset, 0, 32)
-						if err != nil {
-							panic(err)
-						}
-						subdim := *subClusterEl.Dim
-						subdimIncrement, err := strconv.ParseInt(subClusterEl.DimIncrement, 0, 32)
-						if err != nil {
-							panic(err)
-						}
-
-						if subdim > 1 {
-							subcpRegisters := []*PeripheralField{}
-							subregSize := 0
-							for _, regEl := range subClusterEl.Registers {
-								size, err := strconv.ParseInt(*regEl.Size, 0, 32)
-								if err != nil {
-									panic(err)
-								}
-								subregSize += int(size)
-								subcpRegisters = append(subcpRegisters, parseRegister(groupName, regEl, baseAddress+subclusterOffset, subclusterPrefix)...)
-							}
-							cpRegisters = append(cpRegisters, &PeripheralField{
-								Name:        subclusterName,
-								Address:     baseAddress + subclusterOffset,
-								Description: subClusterEl.Description,
-								Registers:   subcpRegisters,
-								Array:       subdim,
-								ElementSize: int(subdimIncrement),
-								ShortName:   clusterPrefix + subclusterName,
-							})
-						} else {
-							for _, regEl := range subClusterEl.Registers {
-								cpRegisters = append(cpRegisters, parseRegister(regEl.Name, regEl, baseAddress+subclusterOffset, subclusterPrefix)...)
-							}
-						}
-					}
-
-					sort.SliceStable(cpRegisters, func(i, j int) bool {
-						return cpRegisters[i].Address < cpRegisters[j].Address
-					})
-					clusterPeripheral := &Peripheral{
-						Name:        periphEl.Name + "_" + clusterName,
-						GroupName:   groupName + "_" + clusterName,
-						Description: description + " - " + clusterName,
-						ClusterName: clusterName,
-						BaseAddress: baseAddress,
-						Registers:   cpRegisters,
-					}
-					peripheralsList = append(peripheralsList, clusterPeripheral)
-					peripheralDict[clusterPeripheral.Name] = clusterPeripheral
-					p.Subtypes = append(p.Subtypes, clusterPeripheral)
-					continue
-				}
-				dim = -1
-				dimIncrement = -1
-			} else {
-				dim = *cluster.Dim
-				if dim == 1 {
-					dimIncrement = -1
-				} else {
-					inc, err := strconv.ParseUint(cluster.DimIncrement, 0, 32)
-					if err != nil {
-						panic(err)
-					}
-					dimIncrement = int(inc)
-				}
-			}
-			clusterRegisters := []*PeripheralField{}
-			for _, regEl := range cluster.Registers {
-				regName := groupName
-				if regName == "" {
-					regName = periphEl.Name
-				}
-				clusterRegisters = append(clusterRegisters, parseRegister(regName, regEl, baseAddress+clusterOffset, clusterPrefix)...)
-			}
-			sort.SliceStable(clusterRegisters, func(i, j int) bool {
-				return clusterRegisters[i].Address < clusterRegisters[j].Address
-			})
-			if dimIncrement == -1 && len(clusterRegisters) > 0 {
-				lastReg := clusterRegisters[len(clusterRegisters)-1]
-				lastAddress := lastReg.Address
-				if lastReg.Array != -1 {
-					lastAddress = lastReg.Address + uint64(lastReg.Array*lastReg.ElementSize)
-				}
-				firstAddress := clusterRegisters[0].Address
-				dimIncrement = int(lastAddress - firstAddress)
-			}
-
-			if !unicode.IsUpper(rune(clusterName[0])) && !unicode.IsDigit(rune(clusterName[0])) {
-				clusterName = strings.ToUpper(clusterName)
-			}
-
-			p.Registers = append(p.Registers, &PeripheralField{
-				Name:        clusterName,
-				Address:     baseAddress + clusterOffset,
-				Description: cluster.Description,
-				Registers:   clusterRegisters,
-				Array:       dim,
-				ElementSize: dimIncrement,
-				ShortName:   clusterName,
-			})
-		}
-		sort.SliceStable(p.Registers, func(i, j int) bool {
-			return p.Registers[i].Address < p.Registers[j].Address
-		})
+		peripheralsList = append(peripheralsList, processCluster(p, periphEl.Clusters, peripheralDict)...)
 	}
 
 	// Make a sorted list of interrupts.
@@ -459,9 +470,10 @@ func readSVD(path, sourceURL string) (*Device, error) {
 		metadata.NVICPrioBits = device.CPU.NVICPrioBits
 	}
 	return &Device{
-		Metadata:    metadata,
-		Interrupts:  interruptList,
-		Peripherals: peripheralsList,
+		Metadata:       metadata,
+		Interrupts:     interruptList,
+		Peripherals:    peripheralsList,
+		PeripheralDict: peripheralDict,
 	}, nil
 }
 
@@ -979,10 +991,11 @@ var (
 
 		address := peripheral.BaseAddress
 		type clusterInfo struct {
-			name      string
-			address   uint64
-			size      uint64
-			registers []*PeripheralField
+			name        string
+			description string
+			address     uint64
+			size        uint64
+			registers   []*PeripheralField
 		}
 		clusters := []clusterInfo{}
 		for _, register := range peripheral.Registers {
@@ -1024,7 +1037,7 @@ var (
 			if register.Registers != nil {
 				// This is a cluster, not a register. Create the cluster type.
 				regType = peripheral.GroupName + "_" + register.Name
-				clusters = append(clusters, clusterInfo{regType, register.Address, uint64(register.ElementSize), register.Registers})
+				clusters = append(clusters, clusterInfo{regType, register.Description, register.Address, uint64(register.ElementSize), register.Registers})
 				regType = regType + "_Type"
 				subaddress := register.Address
 				for _, subregister := range register.Registers {
@@ -1075,7 +1088,16 @@ var (
 				continue
 			}
 
+			if _, ok := device.PeripheralDict[cluster.name]; ok {
+				continue
+			}
+
 			fmt.Fprintln(w)
+			if cluster.description != "" {
+				for _, l := range splitLine(cluster.description) {
+					fmt.Fprintf(w, "// %s\n", l)
+				}
+			}
 			fmt.Fprintf(w, "type %s_Type struct {\n", cluster.name)
 
 			address := cluster.address
@@ -1116,7 +1138,7 @@ var (
 				if register.Registers != nil {
 					// This is a cluster, not a register. Create the cluster type.
 					regType = peripheral.GroupName + "_" + register.Name
-					clusters = append(clusters, clusterInfo{regType, register.Address, uint64(register.ElementSize), register.Registers})
+					clusters = append(clusters, clusterInfo{regType, register.Description, register.Address, uint64(register.ElementSize), register.Registers})
 					regType = regType + "_Type"
 
 					subaddress := register.Address

transform/globals.go

@@ -1,20 +0,0 @@
-package transform
-
-import "tinygo.org/x/go-llvm"
-
-// This file implements small transformations on globals (functions and global
-// variables) for specific ABIs/architectures.
-
-// ApplyFunctionSections puts every function in a separate section. This makes
-// it possible for the linker to remove dead code. It is the equivalent of
-// passing -ffunction-sections to a C compiler.
-func ApplyFunctionSections(mod llvm.Module) {
-	llvmFn := mod.FirstFunction()
-	for !llvmFn.IsNil() {
-		if !llvmFn.IsDeclaration() && llvmFn.Section() == "" {
-			name := llvmFn.Name()
-			llvmFn.SetSection(".text." + name)
-		}
-		llvmFn = llvm.NextFunction(llvmFn)
-	}
-}

transform/globals_test.go

@@ -1,15 +0,0 @@
-package transform_test
-
-import (
-	"testing"
-
-	"github.com/tinygo-org/tinygo/transform"
-	"tinygo.org/x/go-llvm"
-)
-
-func TestApplyFunctionSections(t *testing.T) {
-	t.Parallel()
-	testTransform(t, "testdata/globals-function-sections", func(mod llvm.Module) {
-		transform.ApplyFunctionSections(mod)
-	})
-}

transform/interface-lowering.go

@@ -13,8 +13,7 @@ package transform
 //
 // typeAssert:
 //     Replaced with an icmp instruction so it can be directly used in a type
-//     switch. This is very easy to optimize for LLVM: it will often translate a
-//     type switch into a regular switch statement.
+//     switch.
 //
 // interface type assert:
 //     These functions are defined by creating a big type switch over all the
@@ -54,10 +53,11 @@ type methodInfo struct {
 // typeInfo describes a single concrete Go type, which can be a basic or a named
 // type. If it is a named type, it may have methods.
 type typeInfo struct {
-	name      string
-	typecode  llvm.Value
-	methodSet llvm.Value
-	methods   []*methodInfo
+	name        string
+	typecode    llvm.Value
+	typecodeGEP llvm.Value
+	methodSet   llvm.Value
+	methods     []*methodInfo
 }
 
 // getMethod looks up the method on this type with the given signature and
@@ -91,6 +91,8 @@ type lowerInterfacesPass struct {
 	difiles     map[string]llvm.Metadata
 	ctx         llvm.Context
 	uintptrType llvm.Type
+	targetData  llvm.TargetData
+	i8ptrType   llvm.Type
 	types       map[string]*typeInfo
 	signatures  map[string]*signatureInfo
 	interfaces  map[string]*interfaceInfo
@@ -101,14 +103,17 @@ type lowerInterfacesPass struct {
 // before LLVM can work on them. This is done so that a few cleanup passes can
 // run before assigning the final type codes.
 func LowerInterfaces(mod llvm.Module, config *compileopts.Config) error {
+	ctx := mod.Context()
 	targetData := llvm.NewTargetData(mod.DataLayout())
 	defer targetData.Dispose()
 	p := &lowerInterfacesPass{
 		mod:         mod,
 		config:      config,
-		builder:     mod.Context().NewBuilder(),
-		ctx:         mod.Context(),
+		builder:     ctx.NewBuilder(),
+		ctx:         ctx,
+		targetData:  targetData,
 		uintptrType: mod.Context().IntType(targetData.PointerSize() * 8),
+		i8ptrType:   llvm.PointerType(ctx.Int8Type(), 0),
 		types:       make(map[string]*typeInfo),
 		signatures:  make(map[string]*signatureInfo),
 		interfaces:  make(map[string]*interfaceInfo),
@@ -151,11 +156,26 @@ func (p *lowerInterfacesPass) run() error {
 				}
 				p.types[name] = t
 				initializer := global.Initializer()
-				if initializer.IsNil() {
-					continue
+				firstField := p.builder.CreateExtractValue(initializer, 0, "")
+				if firstField.Type() != p.ctx.Int8Type() {
+					// This type has a method set at index 0. Change the GEP to
+					// point to index 1 (the meta byte).
+					t.typecodeGEP = llvm.ConstGEP(global.GlobalValueType(), global, []llvm.Value{
+						llvm.ConstInt(p.ctx.Int32Type(), 0, false),
+						llvm.ConstInt(p.ctx.Int32Type(), 1, false),
+					})
+					methodSet := stripPointerCasts(firstField)
+					if !strings.HasSuffix(methodSet.Name(), "$methodset") {
+						panic("expected method set")
+					}
+					p.addTypeMethods(t, methodSet)
+				} else {
+					// This type has no method set.
+					t.typecodeGEP = llvm.ConstGEP(global.GlobalValueType(), global, []llvm.Value{
+						llvm.ConstInt(p.ctx.Int32Type(), 0, false),
+						llvm.ConstInt(p.ctx.Int32Type(), 0, false),
+					})
 				}
-				methodSet := p.builder.CreateExtractValue(initializer, 2, "")
-				p.addTypeMethods(t, methodSet)
 			}
 		}
 	}
@@ -266,10 +286,10 @@ func (p *lowerInterfacesPass) run() error {
 		actualType := use.Operand(0)
 		name := strings.TrimPrefix(use.Operand(1).Name(), "reflect/types.typeid:")
 		if t, ok := p.types[name]; ok {
-			// The type exists in the program, so lower to a regular integer
+			// The type exists in the program, so lower to a regular pointer
 			// comparison.
 			p.builder.SetInsertPointBefore(use)
-			commaOk := p.builder.CreateICmp(llvm.IntEQ, llvm.ConstPtrToInt(t.typecode, p.uintptrType), actualType, "typeassert.ok")
+			commaOk := p.builder.CreateICmp(llvm.IntEQ, t.typecodeGEP, actualType, "typeassert.ok")
 			use.ReplaceAllUsesWith(commaOk)
 		} else {
 			// The type does not exist in the program, so lower to a constant
@@ -283,15 +303,45 @@ func (p *lowerInterfacesPass) run() error {
 	}
 
 	// Remove all method sets, which are now unnecessary and inhibit later
-	// optimizations if they are left in place. Also remove references to the
-	// interface type assert functions just to be sure.
-	zeroUintptr := llvm.ConstNull(p.uintptrType)
+	// optimizations if they are left in place.
+	zero := llvm.ConstInt(p.ctx.Int32Type(), 0, false)
 	for _, t := range p.types {
-		initializer := t.typecode.Initializer()
-		methodSet := p.builder.CreateExtractValue(initializer, 2, "")
-		initializer = p.builder.CreateInsertValue(initializer, llvm.ConstNull(methodSet.Type()), 2, "")
-		initializer = p.builder.CreateInsertValue(initializer, zeroUintptr, 4, "")
-		t.typecode.SetInitializer(initializer)
+		if !t.methodSet.IsNil() {
+			initializer := t.typecode.Initializer()
+			var newInitializerFields []llvm.Value
+			for i := 1; i < initializer.Type().StructElementTypesCount(); i++ {
+				newInitializerFields = append(newInitializerFields, p.builder.CreateExtractValue(initializer, i, ""))
+			}
+			newInitializer := p.ctx.ConstStruct(newInitializerFields, false)
+			typecodeName := t.typecode.Name()
+			newGlobal := llvm.AddGlobal(p.mod, newInitializer.Type(), typecodeName+".tmp")
+			newGlobal.SetInitializer(newInitializer)
+			newGlobal.SetLinkage(t.typecode.Linkage())
+			newGlobal.SetGlobalConstant(true)
+			newGlobal.SetAlignment(t.typecode.Alignment())
+			for _, use := range getUses(t.typecode) {
+				if !use.IsAConstantExpr().IsNil() {
+					opcode := use.Opcode()
+					if opcode == llvm.GetElementPtr && use.OperandsCount() == 3 {
+						if use.Operand(1).ZExtValue() == 0 && use.Operand(2).ZExtValue() == 1 {
+							gep := p.builder.CreateInBoundsGEP(newGlobal.GlobalValueType(), newGlobal, []llvm.Value{zero, zero}, "")
+							use.ReplaceAllUsesWith(gep)
+						}
+					}
+				}
+			}
+			// Fallback.
+			if hasUses(t.typecode) {
+				bitcast := llvm.ConstBitCast(newGlobal, p.i8ptrType)
+				negativeOffset := -int64(p.targetData.TypeAllocSize(p.i8ptrType))
+				gep := p.builder.CreateInBoundsGEP(p.ctx.Int8Type(), bitcast, []llvm.Value{llvm.ConstInt(p.ctx.Int32Type(), uint64(negativeOffset), true)}, "")
+				bitcast2 := llvm.ConstBitCast(gep, t.typecode.Type())
+				t.typecode.ReplaceAllUsesWith(bitcast2)
+			}
+			t.typecode.EraseFromParentAsGlobal()
+			newGlobal.SetName(typecodeName)
+			t.typecode = newGlobal
+		}
 	}
 
 	return nil
@@ -301,22 +351,22 @@ func (p *lowerInterfacesPass) run() error {
 // retrieves the signatures and the references to the method functions
 // themselves for later type<->interface matching.
 func (p *lowerInterfacesPass) addTypeMethods(t *typeInfo, methodSet llvm.Value) {
-	if !t.methodSet.IsNil() || methodSet.IsNull() {
+	if !t.methodSet.IsNil() {
 		// no methods or methods already read
 		return
 	}
-	if !methodSet.IsAConstantExpr().IsNil() && methodSet.Opcode() == llvm.GetElementPtr {
-		methodSet = methodSet.Operand(0) // get global from GEP, for LLVM 14 (non-opaque pointers)
-	}
 
 	// This type has methods, collect all methods of this type.
 	t.methodSet = methodSet
 	set := methodSet.Initializer() // get value from global
-	for i := 0; i < set.Type().ArrayLength(); i++ {
-		methodData := p.builder.CreateExtractValue(set, i, "")
-		signatureGlobal := p.builder.CreateExtractValue(methodData, 0, "")
+	signatures := p.builder.CreateExtractValue(set, 1, "")
+	wrappers := p.builder.CreateExtractValue(set, 2, "")
+	numMethods := signatures.Type().ArrayLength()
+	for i := 0; i < numMethods; i++ {
+		signatureGlobal := p.builder.CreateExtractValue(signatures, i, "")
+		function := p.builder.CreateExtractValue(wrappers, i, "")
+		function = stripPointerCasts(function) // strip bitcasts
 		signatureName := signatureGlobal.Name()
-		function := p.builder.CreateExtractValue(methodData, 1, "").Operand(0)
 		signature := p.getSignature(signatureName)
 		method := &methodInfo{
 			function:      function,
@@ -401,7 +451,7 @@ func (p *lowerInterfacesPass) defineInterfaceImplementsFunc(fn llvm.Value, itf *
 	actualType := fn.Param(0)
 	for _, typ := range itf.types {
 		nextBlock := p.ctx.AddBasicBlock(fn, typ.name+".next")
-		cmp := p.builder.CreateICmp(llvm.IntEQ, actualType, llvm.ConstPtrToInt(typ.typecode, p.uintptrType), typ.name+".icmp")
+		cmp := p.builder.CreateICmp(llvm.IntEQ, actualType, typ.typecodeGEP, typ.name+".icmp")
 		p.builder.CreateCondBr(cmp, thenBlock, nextBlock)
 		p.builder.SetInsertPointAtEnd(nextBlock)
 	}
@@ -440,7 +490,7 @@ func (p *lowerInterfacesPass) defineInterfaceMethodFunc(fn llvm.Value, itf *inte
 		params[i] = fn.Param(i + 1)
 	}
 	params = append(params,
-		llvm.Undef(llvm.PointerType(p.ctx.Int8Type(), 0)),
+		llvm.Undef(p.i8ptrType),
 	)
 
 	// Start chain in the entry block.
@@ -472,7 +522,7 @@ func (p *lowerInterfacesPass) defineInterfaceMethodFunc(fn llvm.Value, itf *inte
 		// Create type check (if/else).
 		bb := p.ctx.AddBasicBlock(fn, typ.name)
 		next := p.ctx.AddBasicBlock(fn, typ.name+".next")
-		cmp := p.builder.CreateICmp(llvm.IntEQ, actualType, llvm.ConstPtrToInt(typ.typecode, p.uintptrType), typ.name+".icmp")
+		cmp := p.builder.CreateICmp(llvm.IntEQ, actualType, typ.typecodeGEP, typ.name+".icmp")
 		p.builder.CreateCondBr(cmp, bb, next)
 
 		// The function we will redirect to when the interface has this type.
@@ -522,7 +572,7 @@ func (p *lowerInterfacesPass) defineInterfaceMethodFunc(fn llvm.Value, itf *inte
 	// method on a nil interface.
 	nilPanic := p.mod.NamedFunction("runtime.nilPanic")
 	p.builder.CreateCall(nilPanic.GlobalValueType(), nilPanic, []llvm.Value{
-		llvm.Undef(llvm.PointerType(p.ctx.Int8Type(), 0)),
+		llvm.Undef(p.i8ptrType),
 	}, "")
 	p.builder.CreateUnreachable()
 }

transform/optimizer.go

@@ -116,7 +116,6 @@ func Optimize(mod llvm.Module, config *compileopts.Config, optLevel, sizeLevel i
 		goPasses.Run(mod)
 
 		// Run TinyGo-specific interprocedural optimizations.
-		LowerReflect(mod)
 		OptimizeAllocs(mod, config.Options.PrintAllocs, func(pos token.Position, msg string) {
 			fmt.Fprintln(os.Stderr, pos.String()+": "+msg)
 		})
@@ -129,7 +128,6 @@ func Optimize(mod llvm.Module, config *compileopts.Config, optLevel, sizeLevel i
 		if err != nil {
 			return []error{err}
 		}
-		LowerReflect(mod)
 		errs := LowerInterrupts(mod)
 		if len(errs) > 0 {
 			return errs

transform/reflect.go

@@ -1,567 +0,0 @@
-package transform
-
-// This file has some compiler support for run-time reflection using the reflect
-// package. In particular, it encodes type information in type codes in such a
-// way that the reflect package can decode the type from this information.
-// Where needed, it also adds some side tables for looking up more information
-// about a type, when that information cannot be stored directly in the type
-// code.
-//
-// Go has 26 different type kinds.
-//
-// Type kinds are subdivided in basic types (see the list of basicTypes below)
-// that are mostly numeric literals and non-basic (or "complex") types that are
-// more difficult to encode. These non-basic types come in two forms:
-//   * Prefix types (pointer, slice, interface, channel): these just add
-//     something to an existing type. For example, a pointer like *int just adds
-//     the fact that it's a pointer to an existing type (int).
-//     These are encoded efficiently by adding a prefix to a type code.
-//   * Types with multiple fields (struct, array, func, map). All of these have
-//     multiple fields contained within. Most obviously structs can contain many
-//     types as fields. Also arrays contain not just the element type but also
-//     the length parameter which can be any arbitrary number and thus may not
-//     fit in a type code.
-//     These types are encoded using side tables.
-//
-// This distinction is also important for how named types are encoded. At the
-// moment, named basic type just get a unique number assigned while named
-// non-basic types have their underlying type stored in a sidetable.
-
-import (
-	"encoding/binary"
-	"go/ast"
-	"math/big"
-	"sort"
-	"strings"
-
-	"tinygo.org/x/go-llvm"
-)
-
-// A list of basic types and their numbers. This list should be kept in sync
-// with the list of Kind constants of type.go in the reflect package.
-var basicTypes = map[string]int64{
-	"bool":           1,
-	"int":            2,
-	"int8":           3,
-	"int16":          4,
-	"int32":          5,
-	"int64":          6,
-	"uint":           7,
-	"uint8":          8,
-	"uint16":         9,
-	"uint32":         10,
-	"uint64":         11,
-	"uintptr":        12,
-	"float32":        13,
-	"float64":        14,
-	"complex64":      15,
-	"complex128":     16,
-	"string":         17,
-	"unsafe.Pointer": 18,
-}
-
-// A list of non-basic types. Adding 19 to this number will give the Kind as
-// used in src/reflect/types.go, and it must be kept in sync with that list.
-var nonBasicTypes = map[string]int64{
-	"chan":      0,
-	"interface": 1,
-	"pointer":   2,
-	"slice":     3,
-	"array":     4,
-	"func":      5,
-	"map":       6,
-	"struct":    7,
-}
-
-// typeCodeAssignmentState keeps some global state around for type code
-// assignments, used to assign one unique type code to each Go type.
-type typeCodeAssignmentState struct {
-	// Builder used purely for constant operations (because LLVM 15 removed many
-	// llvm.Const* functions).
-	builder llvm.Builder
-
-	// An integer that's incremented each time it's used to give unique IDs to
-	// type codes that are not yet fully supported otherwise by the reflect
-	// package (or are simply unused in the compiled program).
-	fallbackIndex int
-
-	// This is the length of an uintptr. Only used occasionally to know whether
-	// a given number can be encoded as a varint.
-	uintptrLen int
-
-	// Map of named types to their type code. It is important that named types
-	// get unique IDs for each type.
-	namedBasicTypes    map[string]int
-	namedNonBasicTypes map[string]int
-
-	// Map of array types to their type code.
-	arrayTypes               map[string]int
-	arrayTypesSidetable      []byte
-	needsArrayTypesSidetable bool
-
-	// Map of struct types to their type code.
-	structTypes               map[string]int
-	structTypesSidetable      []byte
-	needsStructNamesSidetable bool
-
-	// Map of struct names and tags to their name string.
-	structNames               map[string]int
-	structNamesSidetable      []byte
-	needsStructTypesSidetable bool
-
-	// This byte array is stored in reflect.namedNonBasicTypesSidetable and is
-	// used at runtime to get details about a named non-basic type.
-	// Entries are varints (see makeVarint below and readVarint in
-	// reflect/sidetables.go for the encoding): one varint per entry. The
-	// integers in namedNonBasicTypes are indices into this array. Because these
-	// are varints, most type codes are really small (just one byte).
-	//
-	// Note that this byte buffer is not created when it is not needed
-	// (reflect.namedNonBasicTypesSidetable has no uses), see
-	// needsNamedTypesSidetable.
-	namedNonBasicTypesSidetable []uint64
-
-	// This indicates whether namedNonBasicTypesSidetable needs to be created at
-	// all. If it is false, namedNonBasicTypesSidetable will contain simple
-	// monotonically increasing numbers.
-	needsNamedNonBasicTypesSidetable bool
-}
-
-// LowerReflect is used to assign a type code to each type in the program
-// that is ever stored in an interface. It tries to use the smallest possible
-// numbers to make the code that works with interfaces as small as possible.
-func LowerReflect(mod llvm.Module) {
-	// if reflect were not used, we could skip generating the sidetable
-	// this does not help in practice, and is difficult to do correctly
-
-	// Obtain slice of all types in the program.
-	type typeInfo struct {
-		typecode llvm.Value
-		name     string
-		numUses  int
-	}
-	var types []*typeInfo
-	for global := mod.FirstGlobal(); !global.IsNil(); global = llvm.NextGlobal(global) {
-		if strings.HasPrefix(global.Name(), "reflect/types.type:") {
-			types = append(types, &typeInfo{
-				typecode: global,
-				name:     global.Name(),
-				numUses:  len(getUses(global)),
-			})
-		}
-	}
-
-	// Sort the slice in a way that often used types are assigned a type code
-	// first.
-	sort.Slice(types, func(i, j int) bool {
-		if types[i].numUses != types[j].numUses {
-			return types[i].numUses < types[j].numUses
-		}
-		// It would make more sense to compare the name in the other direction,
-		// but for some reason that increases binary size. Could be a fluke, but
-		// could also have some good reason (and possibly hint at a small
-		// optimization).
-		return types[i].name > types[j].name
-	})
-
-	// Assign typecodes the way the reflect package expects.
-	targetData := llvm.NewTargetData(mod.DataLayout())
-	defer targetData.Dispose()
-	uintptrType := mod.Context().IntType(targetData.PointerSize() * 8)
-	state := typeCodeAssignmentState{
-		builder:                          mod.Context().NewBuilder(),
-		fallbackIndex:                    1,
-		uintptrLen:                       targetData.PointerSize() * 8,
-		namedBasicTypes:                  make(map[string]int),
-		namedNonBasicTypes:               make(map[string]int),
-		arrayTypes:                       make(map[string]int),
-		structTypes:                      make(map[string]int),
-		structNames:                      make(map[string]int),
-		needsNamedNonBasicTypesSidetable: len(getUses(mod.NamedGlobal("reflect.namedNonBasicTypesSidetable"))) != 0,
-		needsStructTypesSidetable:        len(getUses(mod.NamedGlobal("reflect.structTypesSidetable"))) != 0,
-		needsStructNamesSidetable:        len(getUses(mod.NamedGlobal("reflect.structNamesSidetable"))) != 0,
-		needsArrayTypesSidetable:         len(getUses(mod.NamedGlobal("reflect.arrayTypesSidetable"))) != 0,
-	}
-	defer state.builder.Dispose()
-	for _, t := range types {
-		num := state.getTypeCodeNum(t.typecode)
-		if num.BitLen() > state.uintptrLen || !num.IsUint64() {
-			// TODO: support this in some way, using a side table for example.
-			// That's less efficient but better than not working at all.
-			// Particularly important on systems with 16-bit pointers (e.g.
-			// AVR).
-			panic("compiler: could not store type code number inside interface type code")
-		}
-
-		// Replace each use of the type code global with the constant type code.
-		for _, use := range getUses(t.typecode) {
-			if use.IsAConstantExpr().IsNil() {
-				continue
-			}
-			typecode := llvm.ConstInt(uintptrType, num.Uint64(), false)
-			switch use.Opcode() {
-			case llvm.PtrToInt:
-				// Already of the correct type.
-			case llvm.BitCast:
-				// Could happen when stored in an interface (which is of type
-				// i8*).
-				typecode = llvm.ConstIntToPtr(typecode, use.Type())
-			default:
-				panic("unexpected constant expression")
-			}
-			use.ReplaceAllUsesWith(typecode)
-		}
-	}
-
-	// Only create this sidetable when it is necessary.
-	if state.needsNamedNonBasicTypesSidetable {
-		global := replaceGlobalIntWithArray(mod, "reflect.namedNonBasicTypesSidetable", state.namedNonBasicTypesSidetable)
-		global.SetLinkage(llvm.InternalLinkage)
-		global.SetUnnamedAddr(true)
-		global.SetGlobalConstant(true)
-	}
-	if state.needsArrayTypesSidetable {
-		global := replaceGlobalIntWithArray(mod, "reflect.arrayTypesSidetable", state.arrayTypesSidetable)
-		global.SetLinkage(llvm.InternalLinkage)
-		global.SetUnnamedAddr(true)
-		global.SetGlobalConstant(true)
-	}
-	if state.needsStructTypesSidetable {
-		global := replaceGlobalIntWithArray(mod, "reflect.structTypesSidetable", state.structTypesSidetable)
-		global.SetLinkage(llvm.InternalLinkage)
-		global.SetUnnamedAddr(true)
-		global.SetGlobalConstant(true)
-	}
-	if state.needsStructNamesSidetable {
-		global := replaceGlobalIntWithArray(mod, "reflect.structNamesSidetable", state.structNamesSidetable)
-		global.SetLinkage(llvm.InternalLinkage)
-		global.SetUnnamedAddr(true)
-		global.SetGlobalConstant(true)
-	}
-
-	// Remove most objects created for interface and reflect lowering.
-	// They would normally be removed anyway in later passes, but not always.
-	// It also cleans up the IR for testing.
-	for _, typ := range types {
-		initializer := typ.typecode.Initializer()
-		references := state.builder.CreateExtractValue(initializer, 0, "")
-		typ.typecode.SetInitializer(llvm.ConstNull(initializer.Type()))
-		if strings.HasPrefix(typ.name, "reflect/types.type:struct:") {
-			// Structs have a 'references' field that is not a typecode but
-			// a pointer to a runtime.structField array and therefore a
-			// bitcast. This global should be erased separately, otherwise
-			// typecode objects cannot be erased.
-			structFields := references
-			if !structFields.IsAConstantExpr().IsNil() && structFields.Opcode() == llvm.BitCast {
-				structFields = structFields.Operand(0) // get global from bitcast, for LLVM 14 compatibility (non-opaque pointers)
-			}
-			structFields.EraseFromParentAsGlobal()
-		}
-	}
-}
-
-// getTypeCodeNum returns the typecode for a given type as expected by the
-// reflect package. Also see getTypeCodeName, which serializes types to a string
-// based on a types.Type value for this function.
-func (state *typeCodeAssignmentState) getTypeCodeNum(typecode llvm.Value) *big.Int {
-	// Note: see src/reflect/type.go for bit allocations.
-	class, value := getClassAndValueFromTypeCode(typecode)
-	name := ""
-	if class == "named" {
-		name = value
-		typecode = state.builder.CreateExtractValue(typecode.Initializer(), 0, "")
-		class, value = getClassAndValueFromTypeCode(typecode)
-	}
-	if class == "basic" {
-		// Basic types follow the following bit pattern:
-		//    ...xxxxx0
-		// where xxxxx is allocated for the 18 possible basic types and all the
-		// upper bits are used to indicate the named type.
-		num, ok := basicTypes[value]
-		if !ok {
-			panic("invalid basic type: " + value)
-		}
-		if name != "" {
-			// This type is named, set the upper bits to the name ID.
-			num |= int64(state.getBasicNamedTypeNum(name)) << 5
-		}
-		return big.NewInt(num << 1)
-	} else {
-		// Non-baisc types use the following bit pattern:
-		//    ...nxxx1
-		// where xxx indicates the non-basic type. The upper bits contain
-		// whatever the type contains. Types that wrap a single other type
-		// (channel, interface, pointer, slice) just contain the bits of the
-		// wrapped type. Other types (like struct) need more fields and thus
-		// cannot be encoded as a simple prefix.
-		var classNumber int64
-		if n, ok := nonBasicTypes[class]; ok {
-			classNumber = n
-		} else {
-			panic("unknown type kind: " + class)
-		}
-		var num *big.Int
-		lowBits := (classNumber << 1) + 1 // the 5 low bits of the typecode
-		if name == "" {
-			num = state.getNonBasicTypeCode(class, typecode)
-		} else {
-			// We must return a named type here. But first check whether it
-			// has already been defined.
-			if index, ok := state.namedNonBasicTypes[name]; ok {
-				num := big.NewInt(int64(index))
-				num.Lsh(num, 5).Or(num, big.NewInt((classNumber<<1)+1+(1<<4)))
-				return num
-			}
-			lowBits |= 1 << 4 // set the 'n' bit (see above)
-			if !state.needsNamedNonBasicTypesSidetable {
-				// Use simple small integers in this case, to make these numbers
-				// smaller.
-				index := len(state.namedNonBasicTypes) + 1
-				state.namedNonBasicTypes[name] = index
-				num = big.NewInt(int64(index))
-			} else {
-				// We need to store full type information.
-				// First allocate a number in the named non-basic type
-				// sidetable.
-				index := len(state.namedNonBasicTypesSidetable)
-				state.namedNonBasicTypesSidetable = append(state.namedNonBasicTypesSidetable, 0)
-				state.namedNonBasicTypes[name] = index
-				// Get the typecode of the underlying type (which could be the
-				// element type in the case of pointers, for example).
-				num = state.getNonBasicTypeCode(class, typecode)
-				if num.BitLen() > state.uintptrLen || !num.IsUint64() {
-					panic("cannot store value in sidetable")
-				}
-				// Now update the side table with the number we just
-				// determined. We need this multi-step approach to avoid stack
-				// overflow due to adding types recursively in the case of
-				// linked lists (a pointer which points to a struct that
-				// contains that same pointer).
-				state.namedNonBasicTypesSidetable[index] = num.Uint64()
-				num = big.NewInt(int64(index))
-			}
-		}
-		// Concatenate the 'num' and 'lowBits' bitstrings.
-		num.Lsh(num, 5).Or(num, big.NewInt(lowBits))
-		return num
-	}
-}
-
-// getNonBasicTypeCode is used by getTypeCodeNum. It returns the upper bits of
-// the type code used there in the type code.
-func (state *typeCodeAssignmentState) getNonBasicTypeCode(class string, typecode llvm.Value) *big.Int {
-	switch class {
-	case "chan", "pointer", "slice":
-		// Prefix-style type kinds. The upper bits contain the element type.
-		sub := state.builder.CreateExtractValue(typecode.Initializer(), 0, "")
-		return state.getTypeCodeNum(sub)
-	case "array":
-		// An array is basically a pair of (typecode, length) stored in a
-		// sidetable.
-		return big.NewInt(int64(state.getArrayTypeNum(typecode)))
-	case "struct":
-		// More complicated type kind. The upper bits contain the index to the
-		// struct type in the struct types sidetable.
-		return big.NewInt(int64(state.getStructTypeNum(typecode)))
-	default:
-		// Type has not yet been implemented, so fall back by using a unique
-		// number.
-		num := big.NewInt(int64(state.fallbackIndex))
-		state.fallbackIndex++
-		return num
-	}
-}
-
-// getClassAndValueFromTypeCode takes a typecode (a llvm.Value of type
-// runtime.typecodeID), looks at the name, and extracts the typecode class and
-// value from it. For example, for a typecode with the following name:
-//
-//	reflect/types.type:pointer:named:reflect.ValueError
-//
-// It extracts:
-//
-//	class = "pointer"
-//	value = "named:reflect.ValueError"
-func getClassAndValueFromTypeCode(typecode llvm.Value) (class, value string) {
-	typecodeName := typecode.Name()
-	const prefix = "reflect/types.type:"
-	if !strings.HasPrefix(typecodeName, prefix) {
-		panic("unexpected typecode name: " + typecodeName)
-	}
-	id := typecodeName[len(prefix):]
-	class = id[:strings.IndexByte(id, ':')]
-	value = id[len(class)+1:]
-	return
-}
-
-// getBasicNamedTypeNum returns an appropriate (unique) number for the given
-// named type. If the name already has a number that number is returned, else a
-// new number is returned. The number is always non-zero.
-func (state *typeCodeAssignmentState) getBasicNamedTypeNum(name string) int {
-	if num, ok := state.namedBasicTypes[name]; ok {
-		return num
-	}
-	num := len(state.namedBasicTypes) + 1
-	state.namedBasicTypes[name] = num
-	return num
-}
-
-// getArrayTypeNum returns the array type number, which is an index into the
-// reflect.arrayTypesSidetable or a unique number for this type if this table is
-// not used.
-func (state *typeCodeAssignmentState) getArrayTypeNum(typecode llvm.Value) int {
-	name := typecode.Name()
-	if num, ok := state.arrayTypes[name]; ok {
-		// This array type already has an entry in the sidetable. Don't store
-		// it twice.
-		return num
-	}
-
-	if !state.needsArrayTypesSidetable {
-		// We don't need array sidetables, so we can just assign monotonically
-		// increasing numbers to each array type.
-		num := len(state.arrayTypes)
-		state.arrayTypes[name] = num
-		return num
-	}
-
-	elemTypeCode := state.builder.CreateExtractValue(typecode.Initializer(), 0, "")
-	elemTypeNum := state.getTypeCodeNum(elemTypeCode)
-	if elemTypeNum.BitLen() > state.uintptrLen || !elemTypeNum.IsUint64() {
-		// TODO: make this a regular error
-		panic("array element type has a type code that is too big")
-	}
-
-	// The array side table is a sequence of {element type, array length}.
-	arrayLength := state.builder.CreateExtractValue(typecode.Initializer(), 1, "").ZExtValue()
-	buf := makeVarint(elemTypeNum.Uint64())
-	buf = append(buf, makeVarint(arrayLength)...)
-
-	index := len(state.arrayTypesSidetable)
-	state.arrayTypes[name] = index
-	state.arrayTypesSidetable = append(state.arrayTypesSidetable, buf...)
-	return index
-}
-
-// getStructTypeNum returns the struct type number, which is an index into
-// reflect.structTypesSidetable or an unique number for every struct if this
-// sidetable is not needed in the to-be-compiled program.
-func (state *typeCodeAssignmentState) getStructTypeNum(typecode llvm.Value) int {
-	name := typecode.Name()
-	if num, ok := state.structTypes[name]; ok {
-		// This struct already has an assigned type code.
-		return num
-	}
-
-	if !state.needsStructTypesSidetable {
-		// We don't need struct sidetables, so we can just assign monotonically
-		// increasing numbers to each struct type.
-		num := len(state.structTypes)
-		state.structTypes[name] = num
-		return num
-	}
-
-	// Get the fields this struct type contains.
-	// The struct number will be the start index of
-	structTypeGlobal := stripPointerCasts(state.builder.CreateExtractValue(typecode.Initializer(), 0, "")).Initializer()
-	numFields := structTypeGlobal.Type().ArrayLength()
-
-	// The first data that is stored in the struct sidetable is the number of
-	// fields this struct contains. This is usually just a single byte because
-	// most structs don't contain that many fields, but make it a varint just
-	// to be sure.
-	buf := makeVarint(uint64(numFields))
-
-	// Iterate over every field in the struct.
-	// Every field is stored sequentially in the struct sidetable. Fields can
-	// be retrieved from this list of fields at runtime by iterating over all
-	// of them until the right field has been found.
-	// Perhaps adding some index would speed things up, but it would also make
-	// the sidetable bigger.
-	for i := 0; i < numFields; i++ {
-		// Collect some information about this field.
-		field := state.builder.CreateExtractValue(structTypeGlobal, i, "")
-
-		nameGlobal := state.builder.CreateExtractValue(field, 1, "")
-		if nameGlobal == llvm.ConstPointerNull(nameGlobal.Type()) {
-			panic("compiler: no name for this struct field")
-		}
-		fieldNameBytes := getGlobalBytes(stripPointerCasts(nameGlobal), state.builder)
-		fieldNameNumber := state.getStructNameNumber(fieldNameBytes)
-
-		// See whether this struct field has an associated tag, and if so,
-		// store that tag in the tags sidetable.
-		tagGlobal := state.builder.CreateExtractValue(field, 2, "")
-		hasTag := false
-		tagNumber := 0
-		if tagGlobal != llvm.ConstPointerNull(tagGlobal.Type()) {
-			hasTag = true
-			tagBytes := getGlobalBytes(stripPointerCasts(tagGlobal), state.builder)
-			tagNumber = state.getStructNameNumber(tagBytes)
-		}
-
-		// The 'embedded' or 'anonymous' flag for this field.
-		embedded := state.builder.CreateExtractValue(field, 3, "").ZExtValue() != 0
-
-		// The first byte in the struct types sidetable is a flags byte with
-		// two bits in it.
-		flagsByte := byte(0)
-		if embedded {
-			flagsByte |= 1
-		}
-		if hasTag {
-			flagsByte |= 2
-		}
-		if ast.IsExported(string(fieldNameBytes)) {
-			flagsByte |= 4
-		}
-		buf = append(buf, flagsByte)
-
-		// Get the type number and add it to the buffer.
-		// All fields have a type, so include it directly here.
-		typeNum := state.getTypeCodeNum(state.builder.CreateExtractValue(field, 0, ""))
-		if typeNum.BitLen() > state.uintptrLen || !typeNum.IsUint64() {
-			// TODO: make this a regular error
-			panic("struct field has a type code that is too big")
-		}
-		buf = append(buf, makeVarint(typeNum.Uint64())...)
-
-		// Add the name.
-		buf = append(buf, makeVarint(uint64(fieldNameNumber))...)
-
-		// Add the tag, if there is one.
-		if hasTag {
-			buf = append(buf, makeVarint(uint64(tagNumber))...)
-		}
-	}
-
-	num := len(state.structTypesSidetable)
-	state.structTypes[name] = num
-	state.structTypesSidetable = append(state.structTypesSidetable, buf...)
-	return num
-}
-
-// getStructNameNumber stores this string (name or tag) onto the struct names
-// sidetable. The format is a varint of the length of the struct, followed by
-// the raw bytes of the name. Multiple identical strings are stored under the
-// same name for space efficiency.
-func (state *typeCodeAssignmentState) getStructNameNumber(nameBytes []byte) int {
-	name := string(nameBytes)
-	if n, ok := state.structNames[name]; ok {
-		// This name was used before, re-use it now (for space efficiency).
-		return n
-	}
-	// This name is not yet in the names sidetable. Add it now.
-	n := len(state.structNamesSidetable)
-	state.structNames[name] = n
-	state.structNamesSidetable = append(state.structNamesSidetable, makeVarint(uint64(len(nameBytes)))...)
-	state.structNamesSidetable = append(state.structNamesSidetable, nameBytes...)
-	return n
-}
-
-// makeVarint is a small helper function that returns the bytes of the number in
-// varint encoding.
-func makeVarint(n uint64) []byte {
-	buf := make([]byte, binary.MaxVarintLen64)
-	return buf[:binary.PutUvarint(buf, n)]
-}

transform/reflect_test.go

@@ -1,77 +0,0 @@
-package transform_test
-
-import (
-	"testing"
-
-	"github.com/tinygo-org/tinygo/transform"
-	"tinygo.org/x/go-llvm"
-)
-
-type reflectAssert struct {
-	call           llvm.Value
-	name           string
-	expectedNumber uint64
-}
-
-// Test reflect lowering. This code looks at IR like this:
-//
-//	call void @main.assertType(i32 ptrtoint (%runtime.typecodeID* @"reflect/types.type:basic:int" to i32), i8* inttoptr (i32 3 to i8*), i32 4, i8* undef, i8* undef)
-//
-// and verifies that the ptrtoint constant (the first parameter of
-// @main.assertType) is replaced with the correct type code.  The expected
-// output is this:
-//
-//	call void @main.assertType(i32 4, i8* inttoptr (i32 3 to i8*), i32 4, i8* undef, i8* undef)
-//
-// The first and third parameter are compared and must match, the second
-// parameter is ignored.
-func TestReflect(t *testing.T) {
-	t.Parallel()
-
-	mod := compileGoFileForTesting(t, "./testdata/reflect.go")
-
-	// Run the instcombine pass, to clean up the IR a bit (especially
-	// insertvalue/extractvalue instructions).
-	pm := llvm.NewPassManager()
-	defer pm.Dispose()
-	pm.AddInstructionCombiningPass()
-	pm.Run(mod)
-
-	// Get a list of all the asserts in the source code.
-	assertType := mod.NamedFunction("main.assertType")
-	var asserts []reflectAssert
-	for user := assertType.FirstUse(); !user.IsNil(); user = user.NextUse() {
-		use := user.User()
-		if use.IsACallInst().IsNil() {
-			t.Fatal("expected call use of main.assertType")
-		}
-		global := use.Operand(0).Operand(0)
-		expectedNumber := use.Operand(2).ZExtValue()
-		asserts = append(asserts, reflectAssert{
-			call:           use,
-			name:           global.Name(),
-			expectedNumber: expectedNumber,
-		})
-	}
-
-	// Sanity check to show that the test is actually testing anything.
-	if len(asserts) < 3 {
-		t.Errorf("expected at least 3 test cases, got %d", len(asserts))
-	}
-
-	// Now lower the type codes.
-	transform.LowerReflect(mod)
-
-	// Check whether the values are as expected.
-	for _, assert := range asserts {
-		actualNumberValue := assert.call.Operand(0)
-		if actualNumberValue.IsAConstantInt().IsNil() {
-			t.Errorf("expected to see a constant for %s, got something else", assert.name)
-			continue
-		}
-		actualNumber := actualNumberValue.ZExtValue()
-		if actualNumber != assert.expectedNumber {
-			t.Errorf("%s: expected number 0b%b, got 0b%b", assert.name, assert.expectedNumber, actualNumber)
-		}
-	}
-}

transform/rtcalls.go

@@ -113,11 +113,6 @@ func OptimizeReflectImplements(mod llvm.Module) {
 	builder := mod.Context().NewBuilder()
 	defer builder.Dispose()
 
-	// Get a few useful object for use later.
-	targetData := llvm.NewTargetData(mod.DataLayout())
-	defer targetData.Dispose()
-	uintptrType := mod.Context().IntType(targetData.PointerSize() * 8)
-
 	// Look up the (reflect.Value).Implements() method.
 	var implementsFunc llvm.Value
 	for fn := mod.FirstFunction(); !fn.IsNil(); fn = llvm.NextFunction(fn) {
@@ -141,14 +136,13 @@ func OptimizeReflectImplements(mod llvm.Module) {
 		}
 		interfaceType := stripPointerCasts(call.Operand(2))
 		if interfaceType.IsAGlobalVariable().IsNil() {
-			// The asserted interface is not constant, so can't optimize this
-			// code.
+			// Interface is unknown at compile time. This can't be optimized.
 			continue
 		}
 
 		if strings.HasPrefix(interfaceType.Name(), "reflect/types.type:named:") {
 			// Get the underlying type.
-			interfaceType = builder.CreateExtractValue(interfaceType.Initializer(), 0, "")
+			interfaceType = stripPointerCasts(builder.CreateExtractValue(interfaceType.Initializer(), 2, ""))
 		}
 		if !strings.HasPrefix(interfaceType.Name(), "reflect/types.type:interface:") {
 			// This is an error. The Type passed to Implements should be of
@@ -156,16 +150,15 @@ func OptimizeReflectImplements(mod llvm.Module) {
 			// reported at runtime.
 			continue
 		}
-		if interfaceType.IsAGlobalVariable().IsNil() {
-			// Interface is unknown at compile time. This can't be optimized.
+		typeAssertFunction := mod.NamedFunction(strings.TrimPrefix(interfaceType.Name(), "reflect/types.type:") + ".$typeassert")
+		if typeAssertFunction.IsNil() {
 			continue
 		}
-		typeAssertFunction := builder.CreateExtractValue(interfaceType.Initializer(), 4, "").Operand(0)
 
 		// Replace Implements call with the type assert call.
 		builder.SetInsertPointBefore(call)
 		implements := builder.CreateCall(typeAssertFunction.GlobalValueType(), typeAssertFunction, []llvm.Value{
-			builder.CreatePtrToInt(call.Operand(0), uintptrType, ""), // typecode to check
+			call.Operand(0), // typecode to check
 		}, "")
 		call.ReplaceAllUsesWith(implements)
 		call.EraseFromParentAsInstruction()

transform/testdata/globals-function-sections.ll

@@ -1,8 +0,0 @@
-target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
-target triple = "armv7em-none-eabi"
-
-declare void @foo()
-
-define void @bar() {
-  ret void
-}

transform/testdata/globals-function-sections.out.ll

@@ -1,8 +0,0 @@
-target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
-target triple = "armv7em-none-eabi"
-
-declare void @foo()
-
-define void @bar() section ".text.bar" {
-  ret void
-}