WIP alpine build

ci: simplify build-linux job
Split building the release and smoke-testing the release in two, and don't redo some tests that are already done by assert-test-linux. Some benefits: - Lower overall CI time because tests aren't done multiple times. - TinyHCI can run earlier because the build-linux job is finished as soon as the build artifact is ready. It does however have the downside of an extra job, which costs a few seconds to spin up and a few seconds to push and pull the workspace. But even with this, overall CI time is down by a few minutes per workflow run.
2026-07-31 09:07:46 +00:00 · 2021-11-16 21:57:51 +01:00 · 2021-11-16 18:43:32 +01:00 · 2021-11-16 16:23:59 +01:00 · 2021-11-16 11:08:30 +01:00 · 2021-11-16 11:08:30 +01:00
130 changed files with 2988 additions and 6868 deletions
@@ -6,26 +6,6 @@ commands:
      - run:
          name: "Pull submodules"
          command: git submodule update --init
-  apt-dependencies:
-    parameters:
-      llvm:
-        type: string
-    steps:
-      - run:
-          name: "Install apt dependencies"
-          command: |
-            echo 'deb https://apt.llvm.org/buster/ llvm-toolchain-buster-<<parameters.llvm>> main' | sudo tee /etc/apt/sources.list.d/llvm.list
-            wget -O - https://apt.llvm.org/llvm-snapshot.gpg.key|sudo apt-key add -
-            sudo apt-get update
-            sudo apt-get install --no-install-recommends \
-                llvm-<<parameters.llvm>>-dev \
-                clang-<<parameters.llvm>> \
-                libclang-<<parameters.llvm>>-dev \
-                lld-<<parameters.llvm>> \
-                qemu-system-arm \
-                qemu-user \
-                gcc-avr \
-                avr-libc
  install-node:
    steps:
      - run:
@@ -49,6 +29,13 @@ commands:
          command: |
            curl https://wasmtime.dev/install.sh -sSf | bash
            sudo ln -s ~/.wasmtime/bin/wasmtime /usr/local/bin/wasmtime
+  install-cmake:
+    steps:
+      - run:
+          name: "Install CMake"
+          command: |
+            wget https://github.com/Kitware/CMake/releases/download/v3.21.4/cmake-3.21.4-linux-x86_64.tar.gz
+            sudo tar --strip-components=1 -C /usr/local -xf cmake-3.21.4-linux-x86_64.tar.gz
  install-xtensa-toolchain:
    parameters:
      variant:
@@ -76,6 +63,39 @@ commands:
            - llvm-project/clang/include
            - llvm-project/lld/include
            - llvm-project/llvm/include
+  hack-ninja-jobs:
+    steps:
+      - run:
+          name: "Hack Ninja to use less jobs"
+          command: |
+            echo -e '#!/bin/sh\n/usr/bin/ninja -j3 "$@"' > /go/bin/ninja
+            chmod +x /go/bin/ninja
+  build-binaryen-linux:
+    steps:
+      - restore_cache:
+          keys:
+            - binaryen-linux-v1
+      - run:
+          name: "Build Binaryen"
+          command: |
+            make binaryen
+      - save_cache:
+          key: binaryen-linux-v1
+          paths:
+            - build/wasm-opt
+  build-binaryen-linux-stretch:
+    steps:
+      - restore_cache:
+          keys:
+            - binaryen-linux-stretch-v1
+      - run:
+          name: "Build Binaryen"
+          command: |
+            CC=$PWD/llvm-build/bin/clang make binaryen
+      - save_cache:
+          key: binaryen-linux-stretch-v1
+          paths:
+            - build/wasm-opt
  build-wasi-libc:
    steps:
      - restore_cache:
@@ -95,11 +115,23 @@ commands:
    steps:
      - checkout
      - submodules
-      - apt-dependencies:
-          llvm: "<<parameters.llvm>>"
-      - install-node
-      - install-chrome
-      - install-wasmtime
+      - run:
+          name: "Install apt dependencies"
+          command: |
+            echo 'deb https://apt.llvm.org/buster/ llvm-toolchain-buster-<<parameters.llvm>> main' | sudo tee /etc/apt/sources.list.d/llvm.list
+            wget -O - https://apt.llvm.org/llvm-snapshot.gpg.key | sudo apt-key add -
+            sudo apt-get update
+            sudo apt-get install --no-install-recommends \
+                llvm-<<parameters.llvm>>-dev \
+                clang-<<parameters.llvm>> \
+                libclang-<<parameters.llvm>>-dev \
+                lld-<<parameters.llvm>> \
+                gcc-avr \
+                avr-libc \
+                cmake \
+                ninja-build
+      - hack-ninja-jobs
+      - build-binaryen-linux
      - restore_cache:
          keys:
            - go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
@@ -114,14 +146,8 @@ commands:
          key: wasi-libc-sysroot-systemclang-v3
          paths:
            - lib/wasi-libc/sysroot
-      - run:
-          name: "Test TinyGo"
-          command: go test -v -timeout=20m -tags=llvm<<parameters.llvm>> ./builder ./cgo ./compileopts ./compiler ./interp ./transform .
-          no_output_timeout: 20m
      - run: make gen-device -j4
      - run: make smoketest XTENSA=0
-      - run: make tinygo-test
-      - run: make wasmtest
      - save_cache:
          key: go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
          paths:
@@ -141,9 +167,14 @@ commands:
                qemu-system-arm \
                qemu-user \
                gcc-avr \
-                avr-libc
+                avr-libc \
+                ninja-build \
+                python3
      - install-node
+      - install-chrome
      - install-wasmtime
+      - install-cmake
+      - hack-ninja-jobs
      - install-xtensa-toolchain:
          variant: "linux-amd64"
      - restore_cache:
@@ -159,14 +190,9 @@ commands:
          command: |
            if [ ! -f llvm-build/lib/liblldELF.a ]
            then
-              # fetch LLVM source
+              # fetch LLVM source (may only have headers right now)
              rm -rf llvm-project
              make llvm-source
-              # install dependencies
-              sudo apt-get install cmake ninja-build
-              # hack ninja to use less jobs
-              echo -e '#!/bin/sh\n/usr/bin/ninja -j3 "$@"' > /go/bin/ninja
-              chmod +x /go/bin/ninja
              # build!
              make ASSERT=1 llvm-build
              find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
@@ -175,7 +201,12 @@ commands:
          key: llvm-build-11-linux-v4-assert
          paths:
            llvm-build
-      - run: make ASSERT=1
+      - build-binaryen-linux-stretch
+      - run:
+          name: "Build TinyGo"
+          command: |
+            make ASSERT=1
+            echo 'export PATH=$(pwd)/build:$PATH' >> $BASH_ENV
      - build-wasi-libc
      - run:
          name: "Test TinyGo"
@@ -192,59 +223,67 @@ commands:
            - ~/.cache/go-build
            - /go/pkg/mod
      - run: make gen-device -j4
-      - run: make smoketest TINYGO=build/tinygo
+      - run: make smoketest
+      - run: make tinygo-test
+      - run: make wasmtest
  build-linux:
    steps:
-      - checkout
-      - submodules
      - run:
-          name: "Install apt dependencies"
+          name: "Install apk dependencies"
          command: |
-            sudo apt-get update
-            sudo apt-get install --no-install-recommends \
-                libgnutls30 libssl1.0.2 \
-                qemu-system-arm \
-                qemu-user \
-                gcc-avr \
-                avr-libc
-      - install-node
-      - install-wasmtime
-      - install-xtensa-toolchain:
-          variant: "linux-amd64"
+            apk add git openssh make g++ gcompat
+      - checkout
+      - run:
+          name: "Install Go"
+          command: |
+            wget https://dl.google.com/go/go1.17.3.linux-amd64.tar.gz
+            tar -C /usr/local -xzf go1.17.3.linux-amd64.tar.gz
+            ln -s /usr/local/go/bin/go /usr/local/bin/go
+      #- submodules
      - restore_cache:
+          name: "Restore Go cache"
          keys:
-            - go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
-            - go-cache-v2-{{ checksum "go.mod" }}
-      - llvm-source-linux
+            - go-cache-alpine-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_PREVIOUS_BUILD_NUM }}
+            - go-cache-alpine-v2-{{ checksum "go.mod" }}
      - restore_cache:
+          name: "Restore llvm-source cache"
          keys:
-            - llvm-build-11-linux-v4-noassert
+            - llvm-source-11-alpine-v2-x0
+      - run:
+          name: "Fetch LLVM source"
+          command: make llvm-source
+      - save_cache:
+          name: "Save llvm-source cache"
+          key: llvm-source-11-alpine-v2-x0
+          paths:
+            - llvm-project
+            #- llvm-project/clang/lib/Headers
+            #- llvm-project/clang/include
+            #- llvm-project/lld/include
+            #- llvm-project/llvm/include
+      - restore_cache:
+          name: "Restore llvm-build cache"
+          keys:
+            - llvm-build-11-alpine-v1-noassert-x0
      - run:
          name: "Build LLVM"
          command: |
            if [ ! -f llvm-build/lib/liblldELF.a ]
            then
-              # fetch LLVM source
-              rm -rf llvm-project
-              make llvm-source
+              # fetch LLVM source (may only have headers right now)
+              #rm -rf llvm-project
+              #make llvm-source
              # install dependencies
-              sudo apt-get install cmake ninja-build
-              # hack ninja to use less jobs
-              echo -e '#!/bin/sh\n/usr/bin/ninja -j3 "$@"' > /go/bin/ninja
-              chmod +x /go/bin/ninja
+              apk add cmake samurai python3
              # build!
-              make llvm-build
-              find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
+              SAMUFLAGS=-j3 make llvm-build
+              #find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
            fi
      - save_cache:
-          key: llvm-build-11-linux-v4-noassert
+          name: "Save llvm-build cache"
+          key: llvm-build-11-alpine-v1-noassert-x0
          paths:
            llvm-build
-      - build-wasi-libc
-      - run:
-          name: "Test TinyGo"
-          command: make test
-          no_output_timeout: 20m
      - run:
          name: "Install fpm"
          command: |
@@ -256,21 +295,41 @@ commands:
            make release deb -j3
            cp -p build/release.tar.gz /tmp/tinygo.linux-amd64.tar.gz
            cp -p build/release.deb    /tmp/tinygo_amd64.deb
+      - persist_to_workspace:
+          root: /tmp
+          paths:
+            - tinygo.linux-amd64.tar.gz
      - store_artifacts:
          path: /tmp/tinygo.linux-amd64.tar.gz
      - store_artifacts:
          path: /tmp/tinygo_amd64.deb
      - save_cache:
-          key: go-cache-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
+          name: "Save Go cache"
+          key: go-cache-alpine-v2-{{ checksum "go.mod" }}-{{ .Environment.CIRCLE_BUILD_NUM }}
          paths:
            - ~/.cache/go-build
            - /go/pkg/mod
+  test-linux-build:
+    # Now run the smoke tests for the generated binary.
+    steps:
+      - attach_workspace:
+          at: /tmp/workspace
+      - checkout
+      - run:
+          name: "Install apt dependencies"
+          command: |
+            sudo apt-get update
+            sudo apt-get install --no-install-recommends \
+                gcc-avr \
+                avr-libc
+      - install-xtensa-toolchain:
+          variant: "linux-amd64"
      - run:
          name: "Extract release tarball"
          command: |
            mkdir -p ~/lib
-            tar -C ~/lib -xf /tmp/tinygo.linux-amd64.tar.gz
-            ln -s ~/lib/tinygo/bin/tinygo /go/bin/tinygo
+            tar -C ~/lib -xf /tmp/workspace/tinygo.linux-amd64.tar.gz
+            ln -s ~/lib/tinygo/bin/tinygo ~/go/bin/tinygo
            tinygo version
      - run: make smoketest
  build-macos:
@@ -283,7 +342,7 @@ commands:
            curl https://dl.google.com/go/go1.17.darwin-amd64.tar.gz -o go1.17.darwin-amd64.tar.gz
            sudo tar -C /usr/local -xzf go1.17.darwin-amd64.tar.gz
            ln -s /usr/local/go/bin/go /usr/local/bin/go
-            HOMEBREW_NO_AUTO_UPDATE=1 brew install qemu
+            HOMEBREW_NO_AUTO_UPDATE=1 brew install qemu cmake ninja
      - install-xtensa-toolchain:
          variant: "macos"
      - restore_cache:
@@ -311,11 +370,9 @@ commands:
          command: |
            if [ ! -f llvm-build/lib/liblldELF.a ]
            then
-              # fetch LLVM source
+              # fetch LLVM source (may only have headers right now)
              rm -rf llvm-project
              make llvm-source
-              # install dependencies
-              HOMEBREW_NO_AUTO_UPDATE=1 brew install cmake ninja
              # build!
              make llvm-build
              find llvm-build -name CMakeFiles -prune -exec rm -r '{}' \;
@@ -324,6 +381,20 @@ commands:
          key: llvm-build-11-macos-v5
          paths:
            llvm-build
+      - restore_cache:
+          keys:
+            - binaryen-macos-v1
+      - run:
+          name: "Build Binaryen"
+          command: |
+            if [ ! -f build/wasm-opt ]
+            then
+              make binaryen
+            fi
+      - save_cache:
+          key: binaryen-macos-v1
+          paths:
+            - build/wasm-opt
      - restore_cache:
          keys:
            - wasi-libc-sysroot-macos-v4
@@ -379,9 +450,14 @@ jobs:
      - assert-test-linux
  build-linux:
    docker:
-      - image: circleci/golang:1.17-stretch
+      - image: alpine:3.14
    steps:
      - build-linux
+  test-linux-build:
+    docker:
+      - image: cimg/go:1.17
+    steps:
+      - test-linux-build
  build-macos:
    macos:
      xcode: "11.1.0" # macOS 10.14
@@ -393,8 +469,11 @@ jobs:
 workflows:
  test-all:
    jobs:
-      - test-llvm11-go115
-      - test-llvm11-go116
+      #- test-llvm11-go115
+      #- test-llvm11-go116
      - build-linux
-      - build-macos
-      - assert-test-linux
+      - test-linux-build:
+          requires:
+            - build-linux
+      #- build-macos
+      #- assert-test-linux
@@ -1,10 +1,15 @@
 name: Windows

-on: push
+on:
+  #pull_request:
+  push:
+    branches:
+      - dev
+      - release

 jobs:
  build-windows:
-    runs-on: windows-2016
+    runs-on: windows-2019
    steps:
      - name: Install Go
        uses: actions/setup-go@v2
@@ -15,6 +20,10 @@ jobs:
        run: |
          choco install qemu --version=2020.06.12
          echo "C:\Program Files\QEMU" >> $GITHUB_PATH
+      - name: Install Ninja
+        shell: bash
+        run: |
+          choco install ninja
      - name: Checkout
        uses: actions/checkout@v2
        with:
@@ -45,8 +54,6 @@ jobs:
          # fetch LLVM source
          rm -rf llvm-project
          make llvm-source
-          # install dependencies
-          choco install ninja
          # build!
          make llvm-build
          # Remove unnecessary object files (to reduce cache size).
@@ -60,6 +67,15 @@ jobs:
      - name: Build wasi-libc
        if: steps.cache-wasi-libc.outputs.cache-hit != 'true'
        run: make wasi-libc
+      - name: Cache Binaryen
+        uses: actions/cache@v2
+        id: cache-binaryen
+        with:
+          key: binaryen-v1
+          path: build/binaryen
+      - name: Build Binaryen
+        if: steps.cache-binaryen.outputs.cache-hit != 'true'
+        run: make binaryen
      - name: Test TinyGo
        shell: bash
        run: make test
@@ -83,3 +99,5 @@ jobs:
      - name: Smoke tests
        shell: bash
        run: make smoketest TINYGO=build/tinygo AVR=0 XTENSA=0
+      - name: Test stdlib packages
+        run: make tinygo-test
@@ -1,4 +1,3 @@
-build
 docs/_build
 src/device/avr/*.go
 src/device/avr/*.ld
@@ -26,3 +26,9 @@
 [submodule "lib/musl"]
 	path = lib/musl
 	url = git://git.musl-libc.org/musl
+[submodule "lib/binaryen"]
+	path = lib/binaryen
+	url = https://github.com/WebAssembly/binaryen.git
+[submodule "lib/mingw-w64"]
+	path = lib/mingw-w64
+	url = https://github.com/mingw-w64/mingw-w64.git
@@ -29,8 +29,9 @@ COPY --from=tinygo-base /tinygo/targets /tinygo/targets

 RUN cd /tinygo/ && \
    apt-get update && \
-    apt-get install -y make clang-11 libllvm11 lld-11 && \
-    make wasi-libc
+    apt-get install -y make clang-11 libllvm11 lld-11 cmake ninja-build && \
+    mkdir build && \
+    make wasi-libc binaryen

 # tinygo-avr stage installs the needed dependencies to compile TinyGo programs for AVR microcontrollers.
 FROM tinygo-base AS tinygo-avr
@@ -38,7 +38,7 @@ endif

 .PHONY: all tinygo test $(LLVM_BUILDDIR) llvm-source clean fmt gen-device gen-device-nrf gen-device-nxp gen-device-avr gen-device-rp

-LLVM_COMPONENTS = all-targets analysis asmparser asmprinter bitreader bitwriter codegen core coroutines coverage debuginfodwarf executionengine frontendopenmp instrumentation interpreter ipo irreader linker lto mc mcjit objcarcopts option profiledata scalaropts support target
+LLVM_COMPONENTS = all-targets analysis asmparser asmprinter bitreader bitwriter codegen core coroutines coverage debuginfodwarf debuginfopdb executionengine frontendopenmp instrumentation interpreter ipo irreader libdriver linker lto mc mcjit objcarcopts option profiledata scalaropts support target windowsmanifest

 ifeq ($(OS),Windows_NT)
    EXE = .exe
@@ -54,6 +54,8 @@ ifeq ($(OS),Windows_NT)
    CGO_LDFLAGS += -static -static-libgcc -static-libstdc++
    CGO_LDFLAGS_EXTRA += -lversion

+    BINARYEN_OPTION += -DCMAKE_EXE_LINKER_FLAGS='-static-libgcc -static-libstdc++'
+
    LIBCLANG_NAME = libclang

 else ifeq ($(shell uname -s),Darwin)
@@ -163,12 +165,18 @@ llvm-source: $(LLVM_PROJECTDIR)/llvm
 # Configure LLVM.
 TINYGO_SOURCE_DIR=$(shell pwd)
 $(LLVM_BUILDDIR)/build.ninja: llvm-source
-	mkdir -p $(LLVM_BUILDDIR); cd $(LLVM_BUILDDIR); cmake -G Ninja $(TINYGO_SOURCE_DIR)/$(LLVM_PROJECTDIR)/llvm "-DLLVM_TARGETS_TO_BUILD=X86;ARM;AArch64;RISCV;WebAssembly" "-DLLVM_EXPERIMENTAL_TARGETS_TO_BUILD=AVR;Xtensa" -DCMAKE_BUILD_TYPE=Release -DLIBCLANG_BUILD_STATIC=ON -DLLVM_ENABLE_TERMINFO=OFF -DLLVM_ENABLE_ZLIB=OFF -DLLVM_ENABLE_LIBEDIT=OFF -DLLVM_ENABLE_Z3_SOLVER=OFF -DLLVM_ENABLE_OCAMLDOC=OFF -DLLVM_ENABLE_LIBXML2=OFF -DLLVM_ENABLE_PROJECTS="clang;lld" -DLLVM_TOOL_CLANG_TOOLS_EXTRA_BUILD=OFF -DCLANG_ENABLE_STATIC_ANALYZER=OFF -DCLANG_ENABLE_ARCMT=OFF $(LLVM_OPTION)
+	mkdir -p $(LLVM_BUILDDIR) && cd $(LLVM_BUILDDIR) && cmake -G Ninja $(TINYGO_SOURCE_DIR)/$(LLVM_PROJECTDIR)/llvm "-DLLVM_TARGETS_TO_BUILD=X86;ARM;AArch64;RISCV;WebAssembly" "-DLLVM_EXPERIMENTAL_TARGETS_TO_BUILD=AVR;Xtensa" -DCMAKE_BUILD_TYPE=Release -DLIBCLANG_BUILD_STATIC=ON -DLLVM_ENABLE_TERMINFO=OFF -DLLVM_ENABLE_ZLIB=OFF -DLLVM_ENABLE_LIBEDIT=OFF -DLLVM_ENABLE_Z3_SOLVER=OFF -DLLVM_ENABLE_OCAMLDOC=OFF -DLLVM_ENABLE_LIBXML2=OFF -DLLVM_ENABLE_PROJECTS="clang;lld" -DLLVM_TOOL_CLANG_TOOLS_EXTRA_BUILD=OFF -DCLANG_ENABLE_STATIC_ANALYZER=OFF -DCLANG_ENABLE_ARCMT=OFF $(LLVM_OPTION)

 # Build LLVM.
-$(LLVM_BUILDDIR): $(LLVM_BUILDDIR)/build.ninja
-	cd $(LLVM_BUILDDIR); ninja $(NINJA_BUILD_TARGETS)
+$(LLVM_BUILDDIR): #$(LLVM_BUILDDIR)/build.ninja
+	cd $(LLVM_BUILDDIR) && ninja llvm-nm #$(NINJA_BUILD_TARGETS)

+# Build Binaryen
+.PHONY: binaryen
+binaryen: build/wasm-opt
+build/wasm-opt:
+	cd lib/binaryen && cmake -G Ninja . -DBUILD_STATIC_LIB=ON $(BINARYEN_OPTION) && ninja
+	cp lib/binaryen/bin/wasm-opt build/wasm-opt

 # Build wasi-libc sysroot
 .PHONY: wasi-libc
@@ -208,13 +216,16 @@ TEST_PACKAGES = \
 	hash/crc64 \
 	html \
 	index/suffixarray \
+	internal/itoa \
 	math \
 	math/cmplx \
+	net/mail \
+	reflect \
 	testing \
 	testing/iotest \
 	text/scanner \
-	text/scanner \
 	unicode \
+	unicode/utf16 \
 	unicode/utf8 \

 # Test known-working standard library packages.
@@ -466,7 +477,10 @@ endif
 	$(TINYGO) build -size short -o test.hex -target=pca10040 -opt=0     ./testdata/stdlib.go
 	@$(MD5SUM) test.hex
 	GOOS=linux GOARCH=arm $(TINYGO) build -size short -o test.elf       ./testdata/cgo
+	GOOS=windows GOARCH=amd64 $(TINYGO) build         -o test.exe       ./testdata/cgo
 ifneq ($(OS),Windows_NT)
+	# TODO: this does not yet work on Windows. Somehow, unused functions are
+	# not garbage collected.
 	$(TINYGO) build -o test.elf -gc=leaking -scheduler=none examples/serial
 endif

@@ -474,11 +488,13 @@ endif
 wasmtest:
 	$(GO) test ./tests/wasm

-build/release: tinygo gen-device wasi-libc
+build/release: tinygo gen-device wasi-libc binaryen
 	@mkdir -p build/release/tinygo/bin
 	@mkdir -p build/release/tinygo/lib/clang/include
 	@mkdir -p build/release/tinygo/lib/CMSIS/CMSIS
 	@mkdir -p build/release/tinygo/lib/compiler-rt/lib
+	@mkdir -p build/release/tinygo/lib/mingw-w64/mingw-w64-crt/lib-common
+	@mkdir -p build/release/tinygo/lib/mingw-w64/mingw-w64-headers/defaults
 	@mkdir -p build/release/tinygo/lib/musl/arch
 	@mkdir -p build/release/tinygo/lib/musl/crt
 	@mkdir -p build/release/tinygo/lib/musl/src
@@ -491,6 +507,7 @@ build/release: tinygo gen-device wasi-libc
 	@mkdir -p build/release/tinygo/pkg/armv7em-unknown-unknown-eabi
 	@echo copying source files
 	@cp -p  build/tinygo$(EXE)           build/release/tinygo/bin
+	@cp -p  build/wasm-opt$(EXE)         build/release/tinygo/bin
 	@cp -p $(abspath $(CLANG_SRC))/lib/Headers/*.h build/release/tinygo/lib/clang/include
 	@cp -rp lib/CMSIS/CMSIS/Include      build/release/tinygo/lib/CMSIS/CMSIS
 	@cp -rp lib/CMSIS/README.md          build/release/tinygo/lib/CMSIS
@@ -518,6 +535,11 @@ build/release: tinygo gen-device wasi-libc
 	@cp -rp lib/musl/src/thread          build/release/tinygo/lib/musl/src
 	@cp -rp lib/musl/src/time            build/release/tinygo/lib/musl/src
 	@cp -rp lib/musl/src/unistd          build/release/tinygo/lib/musl/src
+	@cp -rp lib/mingw-w64/mingw-w64-crt/def-include                 build/release/tinygo/lib/mingw-w64/mingw-w64-crt
+	@cp -rp lib/mingw-w64/mingw-w64-crt/lib-common/api-ms-win-crt-* build/release/tinygo/lib/mingw-w64/mingw-w64-crt/lib-common
+	@cp -rp lib/mingw-w64/mingw-w64-crt/lib-common/kernel32.def.in  build/release/tinygo/lib/mingw-w64/mingw-w64-crt/lib-common
+	@cp -rp lib/mingw-w64/mingw-w64-headers/crt/                    build/release/tinygo/lib/mingw-w64/mingw-w64-headers
+	@cp -rp lib/mingw-w64/mingw-w64-headers/defaults/include        build/release/tinygo/lib/mingw-w64/mingw-w64-headers/defaults
 	@cp -rp lib/nrfx/*                   build/release/tinygo/lib/nrfx
 	@cp -rp lib/picolibc/newlib/libc/ctype       build/release/tinygo/lib/picolibc/newlib/libc
 	@cp -rp lib/picolibc/newlib/libc/include     build/release/tinygo/lib/picolibc/newlib/libc
@@ -0,0 +1,2 @@
+*
+!.gitignore
@@ -3,6 +3,7 @@ package builder
 import (
 	"bytes"
 	"debug/elf"
+	"debug/pe"
 	"encoding/binary"
 	"errors"
 	"fmt"
@@ -39,29 +40,42 @@ func makeArchive(arfile *os.File, objs []string) error {
 		}

 		// Read the symbols and add them to the symbol table.
-		dbg, err := elf.NewFile(objfile)
-		if err != nil {
-			return fmt.Errorf("failed to open file %s: %w", objpath, err)
-		}
-		symbols, err := dbg.Symbols()
-		if err != nil {
-			return err
-		}
-		for _, symbol := range symbols {
-			bind := elf.ST_BIND(symbol.Info)
-			if bind != elf.STB_GLOBAL && bind != elf.STB_WEAK {
-				// Don't include local symbols (STB_LOCAL).
-				continue
+		if dbg, err := elf.NewFile(objfile); err == nil {
+			symbols, err := dbg.Symbols()
+			if err != nil {
+				return err
 			}
-			if elf.ST_TYPE(symbol.Info) != elf.STT_FUNC && elf.ST_TYPE(symbol.Info) != elf.STT_OBJECT {
-				// Not a function.
-				continue
+			for _, symbol := range symbols {
+				bind := elf.ST_BIND(symbol.Info)
+				if bind != elf.STB_GLOBAL && bind != elf.STB_WEAK {
+					// Don't include local symbols (STB_LOCAL).
+					continue
+				}
+				if elf.ST_TYPE(symbol.Info) != elf.STT_FUNC && elf.ST_TYPE(symbol.Info) != elf.STT_OBJECT {
+					// Not a function.
+					continue
+				}
+				// Include in archive.
+				symbolTable = append(symbolTable, struct {
+					name      string
+					fileIndex int
+				}{symbol.Name, i})
 			}
-			// Include in archive.
-			symbolTable = append(symbolTable, struct {
-				name      string
-				fileIndex int
-			}{symbol.Name, i})
+		} else if dbg, err := pe.NewFile(objfile); err == nil {
+			for _, symbol := range dbg.Symbols {
+				if symbol.StorageClass != 2 {
+					continue
+				}
+				if symbol.SectionNumber == 0 {
+					continue
+				}
+				symbolTable = append(symbolTable, struct {
+					name      string
+					fileIndex int
+				}{symbol.Name, i})
+			}
+		} else {
+			return fmt.Errorf("failed to open file %s as ELF or PE/COFF: %w", objpath, err)
 		}

 		// Close file, to avoid issues with too many open files (especially on
@@ -120,11 +134,13 @@ func makeArchive(arfile *os.File, objs []string) error {
 	}

 	// Add all object files to the archive.
+	var copyBuf bytes.Buffer
 	for i, objpath := range objs {
 		objfile, err := os.Open(objpath)
 		if err != nil {
 			return err
 		}
+		defer objfile.Close()

 		// Store the start index, for when we'll update the symbol table with
 		// the correct file start indices.
@@ -155,13 +171,21 @@ func makeArchive(arfile *os.File, objs []string) error {
 		}

 		// Copy the file contents into the archive.
-		n, err := io.Copy(arwriter, objfile)
+		// First load all contents into a buffer, then write it all in one go to
+		// the archive file. This is a bit complicated, but is necessary because
+		// io.Copy can't deal with files that are of an odd size.
+		copyBuf.Reset()
+		n, err := io.Copy(&copyBuf, objfile)
 		if err != nil {
-			return err
+			return fmt.Errorf("could not copy object file into ar file: %w", err)
 		}
 		if n != st.Size() {
 			return errors.New("file modified during ar creation: " + arfile.Name())
 		}
+		_, err = arwriter.Write(copyBuf.Bytes())
+		if err != nil {
+			return fmt.Errorf("could not copy object file into ar file: %w", err)
+		}

 		// File is not needed anymore.
 		objfile.Close()
@@ -16,9 +16,11 @@ import (
 	"io/ioutil"
 	"math/bits"
 	"os"
+	"os/exec"
 	"path/filepath"
 	"runtime"
 	"sort"
+	"strconv"
 	"strings"

 	"github.com/tinygo-org/tinygo/cgo"
@@ -113,6 +115,12 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(Buil
 			return errors.New("could not find wasi-libc, perhaps you need to run `make wasi-libc`?")
 		}
 		libcDependencies = append(libcDependencies, dummyCompileJob(path))
+	case "mingw-w64":
+		_, err := MinGW.load(config, dir)
+		if err != nil {
+			return err
+		}
+		libcDependencies = append(libcDependencies, makeMinGWExtraLibs(dir)...)
 	case "":
 		// no library specified, so nothing to do
 	default:
@@ -136,7 +144,6 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(Buil
 		DefaultStackSize:   config.Target.DefaultStackSize,
 		NeedsStackObjects:  config.NeedsStackObjects(),
 		Debug:              true,
-		LLVMFeatures:       config.LLVMFeatures(),
 	}

 	// Load the target machine, which is the LLVM object that contains all
@@ -517,6 +524,9 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(Buil
 	// Prepare link command.
 	linkerDependencies := []*compileJob{outputObjectFileJob}
 	executable := filepath.Join(dir, "main")
+	if config.GOOS() == "windows" {
+		executable += ".exe"
+	}
 	tmppath := executable // final file
 	ldflags := append(config.LDFlags(), "-o", executable)

@@ -659,6 +669,37 @@ func Build(pkgName, outpath string, config *compileopts.Config, action func(Buil
 				}
 			}

+			// Run wasm-opt if necessary.
+			if config.Scheduler() == "asyncify" {
+				var optLevel, shrinkLevel int
+				switch config.Options.Opt {
+				case "none", "0":
+				case "1":
+					optLevel = 1
+				case "2":
+					optLevel = 2
+				case "s":
+					optLevel = 2
+					shrinkLevel = 1
+				case "z":
+					optLevel = 2
+					shrinkLevel = 2
+				default:
+					return fmt.Errorf("unknown opt level: %q", config.Options.Opt)
+				}
+				cmd := exec.Command(goenv.Get("WASMOPT"), "--asyncify", "-g",
+					"--optimize-level", strconv.Itoa(optLevel),
+					"--shrink-level", strconv.Itoa(shrinkLevel),
+					executable, "--output", executable)
+				cmd.Stdout = os.Stdout
+				cmd.Stderr = os.Stderr
+
+				err := cmd.Run()
+				if err != nil {
+					return fmt.Errorf("wasm-opt failed: %w", err)
+				}
+			}
+
 			// Print code size if requested.
 			if config.Options.PrintSizes == "short" || config.Options.PrintSizes == "full" {
 				packagePathMap := make(map[string]string, len(lprogram.Packages))
@@ -12,8 +12,8 @@ import (
 	"tinygo.org/x/go-llvm"
 )

-// Test whether the Clang generated "target-cpu" attribute matches the CPU
-// property in TinyGo target files.
+// Test whether the Clang generated "target-cpu" and "target-features"
+// attributes match the CPU and Features property in TinyGo target files.
 func TestClangAttributes(t *testing.T) {
 	var targetNames = []string{
 		// Please keep this list sorted!
@@ -52,10 +52,13 @@ func TestClangAttributes(t *testing.T) {
 	for _, options := range []*compileopts.Options{
 		{GOOS: "linux", GOARCH: "386"},
 		{GOOS: "linux", GOARCH: "amd64"},
-		{GOOS: "linux", GOARCH: "arm"},
+		{GOOS: "linux", GOARCH: "arm", GOARM: "5"},
+		{GOOS: "linux", GOARCH: "arm", GOARM: "6"},
+		{GOOS: "linux", GOARCH: "arm", GOARM: "7"},
 		{GOOS: "linux", GOARCH: "arm64"},
 		{GOOS: "darwin", GOARCH: "amd64"},
 		{GOOS: "darwin", GOARCH: "arm64"},
+		{GOOS: "windows", GOARCH: "amd64"},
 	} {
 		t.Run("GOOS="+options.GOOS+",GOARCH="+options.GOARCH, func(t *testing.T) {
 			testClangAttributes(t, options)
@@ -112,16 +115,30 @@ func testClangAttributes(t *testing.T, options *compileopts.Options) {
 		t.Errorf("target has LLVM triple %#v but Clang makes it LLVM triple %#v", config.Triple(), mod.Target())
 	}

-	// Check the "target-cpu"  string attribute of the add function.
+	// Check the "target-cpu" and "target-features" string attribute of the add
+	// function.
 	add := mod.NamedFunction("add")
-	var cpu string
+	var cpu, features string
 	cpuAttr := add.GetStringAttributeAtIndex(-1, "target-cpu")
+	featuresAttr := add.GetStringAttributeAtIndex(-1, "target-features")
 	if !cpuAttr.IsNil() {
 		cpu = cpuAttr.GetStringValue()
 	}
+	if !featuresAttr.IsNil() {
+		features = featuresAttr.GetStringValue()
+	}
 	if cpu != config.CPU() {
 		t.Errorf("target has CPU %#v but Clang makes it CPU %#v", config.CPU(), cpu)
 	}
+	if features != config.Features() {
+		if llvm.Version != "11.0.0" {
+			// This needs to be removed once we switch to LLVM 12.
+			// LLVM 11.0.0 uses a different "target-features" string than LLVM
+			// 11.1.0 for Thumb targets. The Xtensa fork is still based on LLVM
+			// 11.0.0, so we need to skip this check on that version.
+			t.Errorf("target has LLVM features %#v but Clang makes it %#v", config.Features(), features)
+		}
+	}
 }

 // This TestMain is necessary because TinyGo may also be invoked to run certain
@@ -11,6 +11,11 @@ bool tinygo_link_elf(int argc, char **argv) {
 	return lld::elf::link(args, false, llvm::outs(), llvm::errs());
 }

+bool tinygo_link_mingw(int argc, char **argv) {
+	std::vector<const char*> args(argv, argv + argc);
+	return lld::mingw::link(args, false, llvm::outs(), llvm::errs());
+}
+
 bool tinygo_link_wasm(int argc, char **argv) {
 	std::vector<const char*> args(argv, argv + argc);
 	return lld::wasm::link(args, false, llvm::outs(), llvm::errs());
@@ -0,0 +1,91 @@
+package builder
+
+import (
+	"io"
+	"os"
+	"path/filepath"
+	"strings"
+
+	"github.com/tinygo-org/tinygo/goenv"
+)
+
+var MinGW = Library{
+	name: "mingw-w64",
+	makeHeaders: func(target, includeDir string) error {
+		// copy _mingw.h
+		srcDir := filepath.Join(goenv.Get("TINYGOROOT"), "lib", "mingw-w64")
+		outf, err := os.Create(includeDir + "/_mingw.h")
+		if err != nil {
+			return err
+		}
+		defer outf.Close()
+		inf, err := os.Open(srcDir + "/mingw-w64-headers/crt/_mingw.h.in")
+		if err != nil {
+			return err
+		}
+		_, err = io.Copy(outf, inf)
+		return err
+	},
+	cflags: func(target, headerPath string) []string {
+		// No flags necessary because there are no files to compile.
+		return nil
+	},
+	librarySources: func(target string) []string {
+		// We only use the UCRT DLL file. No source files necessary.
+		return nil
+	},
+}
+
+// makeMinGWExtraLibs returns a slice of jobs to import the correct .dll
+// libraries. This is done by converting input .def files to .lib files which
+// can then be linked as usual.
+//
+// TODO: cache the result. At the moment, it costs a few hundred milliseconds to
+// compile these files.
+func makeMinGWExtraLibs(tmpdir string) []*compileJob {
+	var jobs []*compileJob
+	root := goenv.Get("TINYGOROOT")
+	// Normally all the api-ms-win-crt-*.def files are all compiled to a single
+	// .lib file. But to simplify things, we're going to leave them as separate
+	// files.
+	for _, name := range []string{
+		"kernel32.def.in",
+		"api-ms-win-crt-conio-l1-1-0.def",
+		"api-ms-win-crt-convert-l1-1-0.def",
+		"api-ms-win-crt-environment-l1-1-0.def",
+		"api-ms-win-crt-filesystem-l1-1-0.def",
+		"api-ms-win-crt-heap-l1-1-0.def",
+		"api-ms-win-crt-locale-l1-1-0.def",
+		"api-ms-win-crt-math-l1-1-0.def.in",
+		"api-ms-win-crt-multibyte-l1-1-0.def",
+		"api-ms-win-crt-private-l1-1-0.def.in",
+		"api-ms-win-crt-process-l1-1-0.def",
+		"api-ms-win-crt-runtime-l1-1-0.def.in",
+		"api-ms-win-crt-stdio-l1-1-0.def",
+		"api-ms-win-crt-string-l1-1-0.def",
+		"api-ms-win-crt-time-l1-1-0.def",
+		"api-ms-win-crt-utility-l1-1-0.def",
+	} {
+		outpath := filepath.Join(tmpdir, filepath.Base(name)+".lib")
+		inpath := filepath.Join(root, "lib/mingw-w64/mingw-w64-crt/lib-common/"+name)
+		job := &compileJob{
+			description: "create lib file " + inpath,
+			result:      outpath,
+			run: func(job *compileJob) error {
+				defpath := inpath
+				if strings.HasSuffix(inpath, ".in") {
+					// .in files need to be preprocessed by a preprocessor (-E)
+					// first.
+					defpath = outpath + ".def"
+					err := runCCompiler("-E", "-x", "c", "-Wp,-w", "-P", "-DDEF_X64", "-o", defpath, inpath, "-I"+goenv.Get("TINYGOROOT")+"/lib/mingw-w64/mingw-w64-crt/def-include/")
+					if err != nil {
+						return err
+					}
+				}
+				return link("ld.lld", "-m", "i386pep", "-o", outpath, defpath)
+			},
+		}
+		jobs = append(jobs, job)
+	}
+	return jobs
+}
@@ -288,6 +288,13 @@ func loadProgramSize(path string, packagePathMap map[string]string) (*programSiz
 			if symType != elf.STT_FUNC && symType != elf.STT_OBJECT && symType != elf.STT_NOTYPE {
 				continue
 			}
+			if symbol.Section >= elf.SHN_LORESERVE {
+				// Not a regular section, so skip it.
+				// One example is elf.SHN_ABS, which is used for symbols
+				// declared with an absolute value such as the memset function
+				// on the ESP32 which is defined in the mask ROM.
+				continue
+			}
 			section := file.Sections[symbol.Section]
 			if section.Flags&elf.SHF_ALLOC == 0 {
 				continue
@@ -13,6 +13,7 @@ import (
 #include <stdlib.h>
 bool tinygo_clang_driver(int argc, char **argv);
 bool tinygo_link_elf(int argc, char **argv);
+bool tinygo_link_mingw(int argc, char **argv);
 bool tinygo_link_wasm(int argc, char **argv);
 */
 import "C"
@@ -24,6 +25,10 @@ const hasBuiltinTools = true
 // This version actually runs the tools because TinyGo was compiled while
 // linking statically with LLVM (with the byollvm build tag).
 func RunTool(tool string, args ...string) error {
+	linker := "elf"
+	if tool == "ld.lld" && len(args) >= 2 && args[0] == "-m" && args[1] == "i386pep" {
+		linker = "mingw"
+	}
 	args = append([]string{"tinygo:" + tool}, args...)

 	var cflag *C.char
@@ -41,7 +46,14 @@ func RunTool(tool string, args ...string) error {
 	case "clang":
 		ok = C.tinygo_clang_driver(C.int(len(args)), (**C.char)(buf))
 	case "ld.lld":
-		ok = C.tinygo_link_elf(C.int(len(args)), (**C.char)(buf))
+		switch linker {
+		case "elf":
+			ok = C.tinygo_link_elf(C.int(len(args)), (**C.char)(buf))
+		case "mingw":
+			ok = C.tinygo_link_mingw(C.int(len(args)), (**C.char)(buf))
+		default:
+			return errors.New("unknown linker: " + linker)
+		}
 	case "wasm-ld":
 		ok = C.tinygo_link_wasm(C.int(len(args)), (**C.char)(buf))
 	default:
@@ -34,10 +34,16 @@ func (c *Config) CPU() string {
 }

 // Features returns a list of features this CPU supports. For example, for a
-// RISC-V processor, that could be ["+a", "+c", "+m"]. For many targets, an
-// empty list will be returned.
-func (c *Config) Features() []string {
-	return c.Target.Features
+// RISC-V processor, that could be "+a,+c,+m". For many targets, an empty list
+// will be returned.
+func (c *Config) Features() string {
+	if c.Target.Features == "" {
+		return c.Options.LLVMFeatures
+	}
+	if c.Options.LLVMFeatures == "" {
+		return c.Target.Features
+	}
+	return c.Target.Features + "," + c.Options.LLVMFeatures
 }

 // GOOS returns the GOOS of the target. This might not always be the actual OS:
@@ -54,6 +60,12 @@ func (c *Config) GOARCH() string {
 	return c.Target.GOARCH
 }

+// GOARM will return the GOARM environment variable given to the compiler when
+// building a program.
+func (c *Config) GOARM() string {
+	return c.Options.GOARM
+}
+
 // 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()}...)
@@ -151,7 +163,7 @@ func (c *Config) OptLevels() (optLevel, sizeLevel int, inlinerThreshold uint) {
 // target.
 func (c *Config) FuncImplementation() string {
 	switch c.Scheduler() {
-	case "tasks":
+	case "tasks", "asyncify":
 		// A func value is implemented as a pair of pointers:
 		//     {context, function pointer}
 		// where the context may be a pointer to a heap-allocated struct
@@ -260,6 +272,15 @@ func (c *Config) CFlags() []string {
 	case "wasi-libc":
 		root := goenv.Get("TINYGOROOT")
 		cflags = append(cflags, "--sysroot="+root+"/lib/wasi-libc/sysroot")
+	case "mingw-w64":
+		root := goenv.Get("TINYGOROOT")
+		path, _ := c.LibcPath("mingw-w64")
+		cflags = append(cflags,
+			"--sysroot="+path,
+			"-Xclang", "-internal-isystem", "-Xclang", filepath.Join(root, "lib", "mingw-w64", "mingw-w64-headers", "crt"),
+			"-Xclang", "-internal-isystem", "-Xclang", filepath.Join(root, "lib", "mingw-w64", "mingw-w64-headers", "defaults", "include"),
+			"-D_UCRT",
+		)
 	case "":
 		// No libc specified, nothing to add.
 	default:
@@ -441,10 +462,6 @@ func (c *Config) WasmAbi() string {
 	return c.Target.WasmAbi
 }

-func (c *Config) LLVMFeatures() string {
-	return c.Options.LLVMFeatures
-}
-
 type TestConfig struct {
 	CompileTestBinary bool
 	// TODO: Filter the test functions to run, include verbose flag, etc
@@ -8,7 +8,7 @@ import (

 var (
 	validGCOptions            = []string{"none", "leaking", "extalloc", "conservative"}
-	validSchedulerOptions     = []string{"none", "tasks", "coroutines"}
+	validSchedulerOptions     = []string{"none", "tasks", "coroutines", "asyncify"}
 	validSerialOptions        = []string{"none", "uart", "usb"}
 	validPrintSizeOptions     = []string{"none", "short", "full"}
 	validPanicStrategyOptions = []string{"print", "trap"}
@@ -21,6 +21,7 @@ var (
 type Options struct {
 	GOOS            string // environment variable
 	GOARCH          string // environment variable
+	GOARM           string // environment variable (only used with GOARCH=arm)
 	Target          string
 	Opt             string
 	GC              string
@@ -10,7 +10,7 @@ import (
 func TestVerifyOptions(t *testing.T) {

 	expectedGCError := errors.New(`invalid gc option 'incorrect': valid values are none, leaking, extalloc, conservative`)
-	expectedSchedulerError := errors.New(`invalid scheduler option 'incorrect': valid values are none, tasks, coroutines`)
+	expectedSchedulerError := errors.New(`invalid scheduler option 'incorrect': valid values are none, tasks, coroutines, asyncify`)
 	expectedPrintSizeError := errors.New(`invalid size option 'incorrect': valid values are none, short, full`)
 	expectedPanicStrategyError := errors.New(`invalid panic option 'incorrect': valid values are print, trap`)

@@ -5,6 +5,7 @@ package compileopts
 import (
 	"encoding/json"
 	"errors"
+	"fmt"
 	"io"
 	"os"
 	"os/exec"
@@ -25,7 +26,7 @@ type TargetSpec struct {
 	Inherits         []string `json:"inherits"`
 	Triple           string   `json:"llvm-target"`
 	CPU              string   `json:"cpu"`
-	Features         []string `json:"features"`
+	Features         string   `json:"features"`
 	GOOS             string   `json:"goos"`
 	GOARCH           string   `json:"goarch"`
 	BuildTags        []string `json:"build-tags"`
@@ -94,7 +95,7 @@ func (spec *TargetSpec) overrideProperties(child *TargetSpec) {
 					dst.Set(src)
 				case "append", "":
 					// or append the field of child to spec
-					dst.Set(reflect.AppendSlice(src, dst))
+					dst.Set(reflect.AppendSlice(dst, src))
 				default:
 					panic("override mode must be 'copy' or 'append' (default). I don't know how to '" + tag + "'.")
 				}
@@ -165,13 +166,26 @@ func LoadTarget(options *Options) (*TargetSpec, error) {
 	if options.Target == "" {
 		// Configure based on GOOS/GOARCH environment variables (falling back to
 		// runtime.GOOS/runtime.GOARCH), and generate a LLVM target based on it.
-		llvmarch := map[string]string{
-			"386":   "i386",
-			"amd64": "x86_64",
-			"arm64": "aarch64",
-			"arm":   "armv7",
-		}[options.GOARCH]
-		if llvmarch == "" {
+		var llvmarch string
+		switch options.GOARCH {
+		case "386":
+			llvmarch = "i386"
+		case "amd64":
+			llvmarch = "x86_64"
+		case "arm64":
+			llvmarch = "aarch64"
+		case "arm":
+			switch options.GOARM {
+			case "5":
+				llvmarch = "armv5"
+			case "6":
+				llvmarch = "armv6"
+			case "7":
+				llvmarch = "armv7"
+			default:
+				return nil, fmt.Errorf("invalid GOARM=%s, must be 5, 6, or 7", options.GOARM)
+			}
+		default:
 			llvmarch = options.GOARCH
 		}
 		llvmos := options.GOOS
@@ -192,6 +206,9 @@ func LoadTarget(options *Options) (*TargetSpec, error) {
 		if options.GOARCH == "arm" {
 			target += "-gnueabihf"
 		}
+		if options.GOOS == "windows" {
+			target += "-gnu"
+		}
 		return defaultTarget(options.GOOS, options.GOARCH, target)
 	}

@@ -208,16 +225,15 @@ func LoadTarget(options *Options) (*TargetSpec, error) {
 	if err != nil {
 		return nil, err
 	}
+
+	if spec.Scheduler == "asyncify" {
+		spec.ExtraFiles = append(spec.ExtraFiles, "src/internal/task/task_asyncify_wasm.S")
+	}
+
 	return spec, nil
 }

-// WindowsBuildNotSupportedErr is being thrown, when goos is windows and no target has been specified.
-var WindowsBuildNotSupportedErr = errors.New("Building Windows binaries is currently not supported. Try specifying a different target")
-
 func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) {
-	if goos == "windows" {
-		return nil, WindowsBuildNotSupportedErr
-	}
 	// No target spec available. Use the default one, useful on most systems
 	// with a regular OS.
 	spec := TargetSpec{
@@ -234,12 +250,23 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) {
 	switch goarch {
 	case "386":
 		spec.CPU = "pentium4"
+		spec.Features = "+cx8,+fxsr,+mmx,+sse,+sse2,+x87"
 	case "amd64":
 		spec.CPU = "x86-64"
+		spec.Features = "+cx8,+fxsr,+mmx,+sse,+sse2,+x87"
 	case "arm":
 		spec.CPU = "generic"
+		switch strings.Split(triple, "-")[0] {
+		case "armv5":
+			spec.Features = "+armv5t,+strict-align,-thumb-mode"
+		case "armv6":
+			spec.Features = "+armv6,+dsp,+fp64,+strict-align,+vfp2,+vfp2sp,-thumb-mode"
+		case "armv7":
+			spec.Features = "+armv7-a,+dsp,+fp64,+vfp2,+vfp2sp,+vfp3d16,+vfp3d16sp,-thumb-mode"
+		}
 	case "arm64":
 		spec.CPU = "generic"
+		spec.Features = "+neon"
 	}
 	if goos == "darwin" {
 		spec.CFlags = append(spec.CFlags, "-isysroot", "/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk")
@@ -249,12 +276,28 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) {
 		spec.RTLib = "compiler-rt"
 		spec.Libc = "musl"
 		spec.LDFlags = append(spec.LDFlags, "--gc-sections")
+	} else if goos == "windows" {
+		spec.Linker = "ld.lld"
+		spec.Libc = "mingw-w64"
+		spec.LDFlags = append(spec.LDFlags,
+			"-m", "i386pep",
+			"-Bdynamic",
+			"--image-base", "0x400000",
+			"--gc-sections",
+			"--no-insert-timestamp",
+		)
 	} else {
 		spec.LDFlags = append(spec.LDFlags, "-no-pie", "-Wl,--gc-sections") // WARNING: clang < 5.0 requires -nopie
 	}
 	if goarch != "wasm" {
-		spec.ExtraFiles = append(spec.ExtraFiles, "src/runtime/gc_"+goarch+".S")
-		spec.ExtraFiles = append(spec.ExtraFiles, "src/internal/task/task_stack_"+goarch+".S")
+		suffix := ""
+		if goos == "windows" {
+			// Windows uses a different calling convention from other operating
+			// systems so we need separate assembly files.
+			suffix = "_windows"
+		}
+		spec.ExtraFiles = append(spec.ExtraFiles, "src/runtime/gc_"+goarch+suffix+".S")
+		spec.ExtraFiles = append(spec.ExtraFiles, "src/internal/task/task_stack_"+goarch+suffix+".S")
 	}
 	if goarch != runtime.GOARCH {
 		// Some educated guesses as to how to invoke helper programs.
@@ -266,6 +309,11 @@ func defaultTarget(goos, goarch, triple string) (*TargetSpec, error) {
 			spec.Emulator = []string{"qemu-aarch64"}
 		}
 	}
+	if goos != runtime.GOOS {
+		if goos == "windows" {
+			spec.Emulator = []string{"wine"}
+		}
+	}
 	return &spec, nil
 }

@@ -27,7 +27,7 @@ func TestOverrideProperties(t *testing.T) {
 	base := &TargetSpec{
 		GOOS:             "baseGoos",
 		CPU:              "baseCpu",
-		Features:         []string{"bf1", "bf2"},
+		CFlags:           []string{"-base-foo", "-base-bar"},
 		BuildTags:        []string{"bt1", "bt2"},
 		Emulator:         []string{"be1", "be2"},
 		DefaultStackSize: 42,
@@ -37,7 +37,7 @@ func TestOverrideProperties(t *testing.T) {
 	child := &TargetSpec{
 		GOOS:             "",
 		CPU:              "chlidCpu",
-		Features:         []string{"cf1", "cf2"},
+		CFlags:           []string{"-child-foo", "-child-bar"},
 		Emulator:         []string{"ce1", "ce2"},
 		AutoStackSize:    &childAutoStackSize,
 		DefaultStackSize: 64,
@@ -51,8 +51,8 @@ func TestOverrideProperties(t *testing.T) {
 	if base.CPU != "chlidCpu" {
 		t.Errorf("Overriding failed : got %v", base.CPU)
 	}
-	if !reflect.DeepEqual(base.Features, []string{"cf1", "cf2", "bf1", "bf2"}) {
-		t.Errorf("Overriding failed : got %v", base.Features)
+	if !reflect.DeepEqual(base.CFlags, []string{"-base-foo", "-base-bar", "-child-foo", "-child-bar"}) {
+		t.Errorf("Overriding failed : got %v", base.CFlags)
 	}
 	if !reflect.DeepEqual(base.BuildTags, []string{"bt1", "bt2"}) {
 		t.Errorf("Overriding failed : got %v", base.BuildTags)
@@ -15,22 +15,16 @@ import (
 // createLookupBoundsCheck emits a bounds check before doing a lookup into a
 // slice. This is required by the Go language spec: an index out of bounds must
 // cause a panic.
-func (b *builder) createLookupBoundsCheck(arrayLen, index llvm.Value, indexType types.Type) {
+// The caller should make sure that index is at least as big as arrayLen.
+func (b *builder) createLookupBoundsCheck(arrayLen, index llvm.Value) {
 	if b.info.nobounds {
 		// The //go:nobounds pragma was added to the function to avoid bounds
 		// checking.
 		return
 	}

-	if index.Type().IntTypeWidth() < arrayLen.Type().IntTypeWidth() {
-		// Sometimes, the index can be e.g. an uint8 or int8, and we have to
-		// correctly extend that type.
-		if indexType.Underlying().(*types.Basic).Info()&types.IsUnsigned == 0 {
-			index = b.CreateZExt(index, arrayLen.Type(), "")
-		} else {
-			index = b.CreateSExt(index, arrayLen.Type(), "")
-		}
-	} else if index.Type().IntTypeWidth() > arrayLen.Type().IntTypeWidth() {
+	// Extend arrayLen if it's too small.
+	if index.Type().IntTypeWidth() > arrayLen.Type().IntTypeWidth() {
 		// The index is bigger than the array length type, so extend it.
 		arrayLen = b.CreateZExt(arrayLen, index.Type(), "")
 	}
@@ -70,27 +64,9 @@ func (b *builder) createSliceBoundsCheck(capacity, low, high, max llvm.Value, lo
 	}

 	// Extend low and high to be the same size as capacity.
-	if low.Type().IntTypeWidth() < capacityType.IntTypeWidth() {
-		if lowType.Info()&types.IsUnsigned != 0 {
-			low = b.CreateZExt(low, capacityType, "")
-		} else {
-			low = b.CreateSExt(low, capacityType, "")
-		}
-	}
-	if high.Type().IntTypeWidth() < capacityType.IntTypeWidth() {
-		if highType.Info()&types.IsUnsigned != 0 {
-			high = b.CreateZExt(high, capacityType, "")
-		} else {
-			high = b.CreateSExt(high, capacityType, "")
-		}
-	}
-	if max.Type().IntTypeWidth() < capacityType.IntTypeWidth() {
-		if maxType.Info()&types.IsUnsigned != 0 {
-			max = b.CreateZExt(max, capacityType, "")
-		} else {
-			max = b.CreateSExt(max, capacityType, "")
-		}
-	}
+	low = b.extendInteger(low, lowType, capacityType)
+	high = b.extendInteger(high, highType, capacityType)
+	max = b.extendInteger(max, maxType, capacityType)

 	// Now do the bounds check: low > high || high > capacity
 	outOfBounds1 := b.CreateICmp(llvm.IntUGT, low, high, "slice.lowhigh")
@@ -125,13 +101,7 @@ func (b *builder) createUnsafeSliceCheck(ptr, len llvm.Value, lenType *types.Bas
 	// using an unsiged greater than.

 	// Make sure the len value is at least as big as a uintptr.
-	if len.Type().IntTypeWidth() < b.uintptrType.IntTypeWidth() {
-		if lenType.Info()&types.IsUnsigned != 0 {
-			len = b.CreateZExt(len, b.uintptrType, "")
-		} else {
-			len = b.CreateSExt(len, b.uintptrType, "")
-		}
-	}
+	len = b.extendInteger(len, lenType, b.uintptrType)

 	// Determine the maximum slice size, and therefore the maximum value of the
 	// len parameter.
@@ -159,17 +129,8 @@ func (b *builder) createChanBoundsCheck(elementSize uint64, bufSize llvm.Value,
 		return
 	}

-	// Check whether the bufSize parameter must be cast to a wider integer for
-	// comparison.
-	if bufSize.Type().IntTypeWidth() < b.uintptrType.IntTypeWidth() {
-		if bufSizeType.Info()&types.IsUnsigned != 0 {
-			// Unsigned, so zero-extend to uint type.
-			bufSize = b.CreateZExt(bufSize, b.intType, "")
-		} else {
-			// Signed, so sign-extend to int type.
-			bufSize = b.CreateSExt(bufSize, b.intType, "")
-		}
-	}
+	// Make sure bufSize is at least as big as maxBufSize (an uintptr).
+	bufSize = b.extendInteger(bufSize, bufSizeType, b.uintptrType)

 	// Calculate (^uintptr(0)) >> 1, which is the max value that fits in an
 	// uintptr if uintptrs were signed.
@@ -294,3 +255,19 @@ func (b *builder) createRuntimeAssert(assert llvm.Value, blockPrefix, assertFunc
 	// Ok: assert didn't trigger so continue normally.
 	b.SetInsertPointAtEnd(nextBlock)
 }
+
+// extendInteger extends the value to at least targetType using a zero or sign
+// extend. The resulting value is not truncated: it may still be bigger than
+// targetType.
+func (b *builder) extendInteger(value llvm.Value, valueType types.Type, targetType llvm.Type) llvm.Value {
+	if value.Type().IntTypeWidth() < targetType.IntTypeWidth() {
+		if valueType.Underlying().(*types.Basic).Info()&types.IsUnsigned != 0 {
+			// Unsigned, so zero-extend to the target type.
+			value = b.CreateZExt(value, targetType, "")
+		} else {
+			// Signed, so sign-extend to the target type.
+			value = b.CreateSExt(value, targetType, "")
+		}
+	}
+	return value
+}
@@ -23,7 +23,7 @@ import (
 // Version of the compiler pacakge. Must be incremented each time the compiler
 // package changes in a way that affects the generated LLVM module.
 // This version is independent of the TinyGo version number.
-const Version = 24 // last change: add layout param to runtime.alloc calls
+const Version = 25 // last change: add "target-cpu" and "target-features" attributes

 func init() {
 	llvm.InitializeAllTargets()
@@ -43,7 +43,7 @@ type Config struct {
 	// Target and output information.
 	Triple          string
 	CPU             string
-	Features        []string
+	Features        string
 	GOOS            string
 	GOARCH          string
 	CodeModel       string
@@ -57,7 +57,6 @@ type Config struct {
 	DefaultStackSize   uint64
 	NeedsStackObjects  bool
 	Debug              bool // Whether to emit debug information in the LLVM module.
-	LLVMFeatures       string
 }

 // compilerContext contains function-independent data that should still be
@@ -189,12 +188,6 @@ func NewTargetMachine(config *Config) (llvm.TargetMachine, error) {
 		return llvm.TargetMachine{}, err
 	}

-	feat := config.Features
-	if len(config.LLVMFeatures) > 0 {
-		feat = append(feat, config.LLVMFeatures)
-	}
-	features := strings.Join(feat, ",")
-
 	var codeModel llvm.CodeModel
 	var relocationModel llvm.RelocMode

@@ -222,7 +215,7 @@ func NewTargetMachine(config *Config) (llvm.TargetMachine, error) {
 		relocationModel = llvm.RelocDynamicNoPic
 	}

-	machine := target.CreateTargetMachine(config.Triple, config.CPU, features, llvm.CodeGenLevelDefault, relocationModel, codeModel)
+	machine := target.CreateTargetMachine(config.Triple, config.CPU, config.Features, llvm.CodeGenLevelDefault, relocationModel, codeModel)
 	return machine, nil
 }

@@ -1629,10 +1622,14 @@ func (b *builder) createExpr(expr ssa.Value) (llvm.Value, error) {
 		array := b.getValue(expr.X)
 		index := b.getValue(expr.Index)

+		// Extend index to at least uintptr size, because getelementptr assumes
+		// index is a signed integer.
+		index = b.extendInteger(index, expr.Index.Type(), b.uintptrType)
+
 		// Check bounds.
 		arrayLen := expr.X.Type().Underlying().(*types.Array).Len()
 		arrayLenLLVM := llvm.ConstInt(b.uintptrType, uint64(arrayLen), false)
-		b.createLookupBoundsCheck(arrayLenLLVM, index, expr.Index.Type())
+		b.createLookupBoundsCheck(arrayLenLLVM, index)

 		// Can't load directly from array (as index is non-constant), so have to
 		// do it using an alloca+gep+load.
@@ -1673,8 +1670,12 @@ func (b *builder) createExpr(expr ssa.Value) (llvm.Value, error) {
 			return llvm.Value{}, b.makeError(expr.Pos(), "todo: indexaddr: "+ptrTyp.String())
 		}

+		// Make sure index is at least the size of uintptr becuase getelementptr
+		// assumes index is a signed integer.
+		index = b.extendInteger(index, expr.Index.Type(), b.uintptrType)
+
 		// Bounds check.
-		b.createLookupBoundsCheck(buflen, index, expr.Index.Type())
+		b.createLookupBoundsCheck(buflen, index)

 		switch expr.X.Type().Underlying().(type) {
 		case *types.Pointer:
@@ -1698,9 +1699,17 @@ func (b *builder) createExpr(expr ssa.Value) (llvm.Value, error) {
 				panic("lookup on non-string?")
 			}

+			// Sometimes, the index can be e.g. an uint8 or int8, and we have to
+			// correctly extend that type for two reasons:
+			//  1. The lookup bounds check expects an index of at least uintptr
+			//     size.
+			//  2. getelementptr has signed operands, and therefore s[uint8(x)]
+			//     can be lowered as s[int8(x)]. That would be a bug.
+			index = b.extendInteger(index, expr.Index.Type(), b.uintptrType)
+
 			// Bounds check.
 			length := b.CreateExtractValue(value, 1, "len")
-			b.createLookupBoundsCheck(length, index, expr.Index.Type())
+			b.createLookupBoundsCheck(length, index)

 			// Lookup byte
 			buf := b.CreateExtractValue(value, 0, "")
@@ -1826,13 +1835,7 @@ func (b *builder) createExpr(expr ssa.Value) (llvm.Value, error) {
 		if expr.Low != nil {
 			lowType = expr.Low.Type().Underlying().(*types.Basic)
 			low = b.getValue(expr.Low)
-			if low.Type().IntTypeWidth() < b.uintptrType.IntTypeWidth() {
-				if lowType.Info()&types.IsUnsigned != 0 {
-					low = b.CreateZExt(low, b.uintptrType, "")
-				} else {
-					low = b.CreateSExt(low, b.uintptrType, "")
-				}
-			}
+			low = b.extendInteger(low, lowType, b.uintptrType)
 		} else {
 			lowType = types.Typ[types.Uintptr]
 			low = llvm.ConstInt(b.uintptrType, 0, false)
@@ -1841,13 +1844,7 @@ func (b *builder) createExpr(expr ssa.Value) (llvm.Value, error) {
 		if expr.High != nil {
 			highType = expr.High.Type().Underlying().(*types.Basic)
 			high = b.getValue(expr.High)
-			if high.Type().IntTypeWidth() < b.uintptrType.IntTypeWidth() {
-				if highType.Info()&types.IsUnsigned != 0 {
-					high = b.CreateZExt(high, b.uintptrType, "")
-				} else {
-					high = b.CreateSExt(high, b.uintptrType, "")
-				}
-			}
+			high = b.extendInteger(high, highType, b.uintptrType)
 		} else {
 			highType = types.Typ[types.Uintptr]
 		}
@@ -1855,13 +1852,7 @@ func (b *builder) createExpr(expr ssa.Value) (llvm.Value, error) {
 		if expr.Max != nil {
 			maxType = expr.Max.Type().Underlying().(*types.Basic)
 			max = b.getValue(expr.Max)
-			if max.Type().IntTypeWidth() < b.uintptrType.IntTypeWidth() {
-				if maxType.Info()&types.IsUnsigned != 0 {
-					max = b.CreateZExt(max, b.uintptrType, "")
-				} else {
-					max = b.CreateSExt(max, b.uintptrType, "")
-				}
-			}
+			max = b.extendInteger(max, maxType, b.uintptrType)
 		} else {
 			maxType = types.Typ[types.Uintptr]
 		}
@@ -18,8 +18,9 @@ import (
 var flagUpdate = flag.Bool("update", false, "update tests based on test output")

 type testCase struct {
-	file   string
-	target string
+	file      string
+	target    string
+	scheduler string
 }

 // Basic tests for the compiler. Build some Go files and compare the output with
@@ -41,20 +42,21 @@ func TestCompiler(t *testing.T) {
 	}

 	tests := []testCase{
-		{"basic.go", ""},
-		{"pointer.go", ""},
-		{"slice.go", ""},
-		{"string.go", ""},
-		{"float.go", ""},
-		{"interface.go", ""},
-		{"func.go", ""},
-		{"pragma.go", ""},
-		{"goroutine.go", "wasm"},
-		{"goroutine.go", "cortex-m-qemu"},
-		{"channel.go", ""},
-		{"intrinsics.go", "cortex-m-qemu"},
-		{"intrinsics.go", "wasm"},
-		{"gc.go", ""},
+		{"basic.go", "", ""},
+		{"pointer.go", "", ""},
+		{"slice.go", "", ""},
+		{"string.go", "", ""},
+		{"float.go", "", ""},
+		{"interface.go", "", ""},
+		{"func.go", "", "coroutines"},
+		{"pragma.go", "", ""},
+		{"goroutine.go", "wasm", "asyncify"},
+		{"goroutine.go", "wasm", "coroutines"},
+		{"goroutine.go", "cortex-m-qemu", "tasks"},
+		{"channel.go", "", ""},
+		{"intrinsics.go", "cortex-m-qemu", ""},
+		{"intrinsics.go", "wasm", ""},
+		{"gc.go", "", ""},
 	}

 	_, minor, err := goenv.GetGorootVersion(goenv.Get("GOROOT"))
@@ -62,7 +64,7 @@ func TestCompiler(t *testing.T) {
 		t.Fatal("could not read Go version:", err)
 	}
 	if minor >= 17 {
-		tests = append(tests, testCase{"go1.17.go", ""})
+		tests = append(tests, testCase{"go1.17.go", "", ""})
 	}

 	for _, tc := range tests {
@@ -70,7 +72,10 @@ func TestCompiler(t *testing.T) {
 		targetString := "wasm"
 		if tc.target != "" {
 			targetString = tc.target
-			name = tc.file + "-" + tc.target
+			name += "-" + tc.target
+		}
+		if tc.scheduler != "" {
+			name += "-" + tc.scheduler
 		}

 		t.Run(name, func(t *testing.T) {
@@ -81,6 +86,9 @@ func TestCompiler(t *testing.T) {
 			if err != nil {
 				t.Fatal("failed to load target:", err)
 			}
+			if tc.scheduler != "" {
+				options.Scheduler = tc.scheduler
+			}
 			config := &compileopts.Config{
 				Options: options,
 				Target:  target,
@@ -94,6 +102,7 @@ func TestCompiler(t *testing.T) {
 				Scheduler:          config.Scheduler(),
 				FuncImplementation: config.FuncImplementation(),
 				AutomaticStackSize: config.AutomaticStackSize(),
+				DefaultStackSize:   config.Target.DefaultStackSize,
 			}
 			machine, err := NewTargetMachine(compilerConfig)
 			if err != nil {
@@ -142,6 +151,9 @@ func TestCompiler(t *testing.T) {
 			if tc.target != "" {
 				outFilePrefix += "-" + tc.target
 			}
+			if tc.scheduler != "" {
+				outFilePrefix += "-" + tc.scheduler
+			}
 			outPath := "./testdata/" + outFilePrefix + ".ll"

 			// Update test if needed. Do not check the result.
@@ -100,7 +100,7 @@ func (b *builder) createGo(instr *ssa.Go) {
 		switch b.Scheduler {
 		case "none", "coroutines":
 			// There are no additional parameters needed for the goroutine start operation.
-		case "tasks":
+		case "tasks", "asyncify":
 			// Add the function pointer as a parameter to start the goroutine.
 			params = append(params, funcPtr)
 		default:
@@ -112,7 +112,7 @@ func (b *builder) createGo(instr *ssa.Go) {
 	paramBundle := b.emitPointerPack(params)
 	var callee, stackSize llvm.Value
 	switch b.Scheduler {
-	case "none", "tasks":
+	case "none", "tasks", "asyncify":
 		callee = b.createGoroutineStartWrapper(funcPtr, prefix, hasContext, instr.Pos())
 		if b.AutomaticStackSize {
 			// The stack size is not known until after linking. Call a dummy
@@ -124,7 +124,7 @@ func (b *builder) createGo(instr *ssa.Go) {
 		} else {
 			// The stack size is fixed at compile time. By emitting it here as a
 			// constant, it can be optimized.
-			if b.Scheduler == "tasks" && b.DefaultStackSize == 0 {
+			if (b.Scheduler == "tasks" || b.Scheduler == "asyncify") && b.DefaultStackSize == 0 {
 				b.addError(instr.Pos(), "default stack size for goroutines is not set")
 			}
 			stackSize = llvm.ConstInt(b.uintptrType, b.DefaultStackSize, false)
@@ -170,6 +170,11 @@ func (c *compilerContext) createGoroutineStartWrapper(fn llvm.Value, prefix stri
 	builder := c.ctx.NewBuilder()
 	defer builder.Dispose()

+	var deadlock llvm.Value
+	if c.Scheduler == "asyncify" {
+		deadlock = c.getFunction(c.program.ImportedPackage("runtime").Members["deadlock"].(*ssa.Function))
+	}
+
 	if !fn.IsAFunction().IsNil() {
 		// See whether this wrapper has already been created. If so, return it.
 		name := fn.Name()
@@ -225,6 +230,12 @@ func (c *compilerContext) createGoroutineStartWrapper(fn llvm.Value, prefix stri
 		// Create the call.
 		builder.CreateCall(fn, params, "")

+		if c.Scheduler == "asyncify" {
+			builder.CreateCall(deadlock, []llvm.Value{
+				llvm.Undef(c.i8ptrType), llvm.Undef(c.i8ptrType),
+			}, "")
+		}
+
 	} else {
 		// For a function pointer like this:
 		//
@@ -292,11 +303,22 @@ func (c *compilerContext) createGoroutineStartWrapper(fn llvm.Value, prefix stri

 		// Create the call.
 		builder.CreateCall(fnPtr, params, "")
+
+		if c.Scheduler == "asyncify" {
+			builder.CreateCall(deadlock, []llvm.Value{
+				llvm.Undef(c.i8ptrType), llvm.Undef(c.i8ptrType),
+			}, "")
+		}
 	}

-	// Finish the function. Every basic block must end in a terminator, and
-	// because goroutines never return a value we can simply return void.
-	builder.CreateRetVoid()
+	if c.Scheduler == "asyncify" {
+		// The goroutine was terminated via deadlock.
+		builder.CreateUnreachable()
+	} else {
+		// Finish the function. Every basic block must end in a terminator, and
+		// because goroutines never return a value we can simply return void.
+		builder.CreateRetVoid()
+	}

 	// Return a ptrtoint of the wrapper, not the function itself.
 	return builder.CreatePtrToInt(wrapper, c.uintptrType, "")
@@ -346,6 +346,12 @@ func (c *compilerContext) addStandardDeclaredAttributes(llvmFn llvm.Value) {
 		attr := c.ctx.CreateEnumAttribute(kind, 0)
 		llvmFn.AddFunctionAttr(attr)
 	}
+	if c.CPU != "" {
+		llvmFn.AddFunctionAttr(c.ctx.CreateStringAttribute("target-cpu", c.CPU))
+	}
+	if c.Features != "" {
+		llvmFn.AddFunctionAttr(c.ctx.CreateStringAttribute("target-features", c.Features))
+	}
 }

 // addStandardDefinedAttributes adds the set of attributes that are added to
@@ -156,12 +156,12 @@ func (b *builder) createRawSyscall(call *ssa.CallCommon) (llvm.Value, error) {
 // createSyscall emits instructions for the syscall.Syscall* family of
 // functions, depending on the target OS/arch.
 func (b *builder) createSyscall(call *ssa.CallCommon) (llvm.Value, error) {
-	syscallResult, err := b.createRawSyscall(call)
-	if err != nil {
-		return syscallResult, err
-	}
 	switch b.GOOS {
 	case "linux", "freebsd":
+		syscallResult, err := b.createRawSyscall(call)
+		if err != nil {
+			return syscallResult, err
+		}
 		// Return values: r0, r1 uintptr, err Errno
 		// Pseudocode:
 		//     var err uintptr
@@ -180,6 +180,10 @@ func (b *builder) createSyscall(call *ssa.CallCommon) (llvm.Value, error) {
 		retval = b.CreateInsertValue(retval, errResult, 2, "")
 		return retval, nil
 	case "darwin":
+		syscallResult, err := b.createRawSyscall(call)
+		if err != nil {
+			return syscallResult, err
+		}
 		// Return values: r0, r1 uintptr, err Errno
 		// Pseudocode:
 		//     var err uintptr
@@ -195,6 +199,57 @@ func (b *builder) createSyscall(call *ssa.CallCommon) (llvm.Value, error) {
 		retval = b.CreateInsertValue(retval, zero, 1, "")
 		retval = b.CreateInsertValue(retval, errResult, 2, "")
 		return retval, nil
+	case "windows":
+		// On Windows, syscall.Syscall* is basically just a function pointer
+		// call. This is complicated in gc because of stack switching and the
+		// different ABI, but easy in TinyGo: just call the function pointer.
+		// The signature looks like this:
+		//   func Syscall(trap, nargs, a1, a2, a3 uintptr) (r1, r2 uintptr, err Errno)
+
+		// Prepare input values.
+		var paramTypes []llvm.Type
+		var params []llvm.Value
+		for _, val := range call.Args[2:] {
+			param := b.getValue(val)
+			params = append(params, param)
+			paramTypes = append(paramTypes, param.Type())
+		}
+		llvmType := llvm.FunctionType(b.uintptrType, paramTypes, false)
+		fn := b.getValue(call.Args[0])
+		fnPtr := b.CreateIntToPtr(fn, llvm.PointerType(llvmType, 0), "")
+
+		// Prepare some functions that will be called later.
+		setLastError := b.mod.NamedFunction("SetLastError")
+		if setLastError.IsNil() {
+			llvmType := llvm.FunctionType(b.ctx.VoidType(), []llvm.Type{b.ctx.Int32Type()}, false)
+			setLastError = llvm.AddFunction(b.mod, "SetLastError", llvmType)
+		}
+		getLastError := b.mod.NamedFunction("GetLastError")
+		if getLastError.IsNil() {
+			llvmType := llvm.FunctionType(b.ctx.Int32Type(), nil, false)
+			getLastError = llvm.AddFunction(b.mod, "GetLastError", llvmType)
+		}
+
+		// Now do the actual call. Pseudocode:
+		//     SetLastError(0)
+		//     r1 = trap(a1, a2, a3, ...)
+		//     err = uintptr(GetLastError())
+		//     return r1, 0, err
+		// Note that SetLastError/GetLastError could be replaced with direct
+		// access to the thread control block, which is probably smaller and
+		// faster. The Go runtime does this in assembly.
+		b.CreateCall(setLastError, []llvm.Value{llvm.ConstNull(b.ctx.Int32Type())}, "")
+		syscallResult := b.CreateCall(fnPtr, params, "")
+		errResult := b.CreateCall(getLastError, nil, "err")
+		if b.uintptrType != b.ctx.Int32Type() {
+			errResult = b.CreateZExt(errResult, b.uintptrType, "err.uintptr")
+		}
+
+		// Return r1, 0, err
+		retval := llvm.ConstNull(b.ctx.StructType([]llvm.Type{b.uintptrType, b.uintptrType, b.uintptrType}, false))
+		retval = b.CreateInsertValue(retval, syscallResult, 0, "")
+		retval = b.CreateInsertValue(retval, errResult, 2, "")
+		return retval, nil
 	default:
 		return llvm.Value{}, b.makeError(call.Pos(), "unknown GOOS/GOARCH for syscall: "+b.GOOS+"/"+b.GOARCH)
 	}
@@ -6,7 +6,8 @@ target triple = "wasm32-unknown-wasi"
 %runtime.channel = type { i32, i32, i8, %runtime.channelBlockedList*, i32, i32, i32, i8* }
 %runtime.channelBlockedList = type { %runtime.channelBlockedList*, %"internal/task.Task"*, %runtime.chanSelectState*, { %runtime.channelBlockedList*, i32, i32 } }
 %"internal/task.Task" = type { %"internal/task.Task"*, i8*, i64, %"internal/task.state" }
-%"internal/task.state" = type { i8* }
+%"internal/task.state" = type { i32, i8*, %"internal/task.stackState", i1 }
+%"internal/task.stackState" = type { i32, i32 }
 %runtime.chanSelectState = type { %runtime.channel*, i8* }

 declare noalias nonnull i8* @runtime.alloc(i32, i8*, i8*, i8*)
@@ -5,7 +5,6 @@ target triple = "wasm32-unknown-wasi"

 %runtime.typecodeID = type { %runtime.typecodeID*, i32, %runtime.interfaceMethodInfo*, %runtime.typecodeID*, i32 }
 %runtime.interfaceMethodInfo = type { i8*, i32 }
-%runtime.funcValue = type { i8*, i32 }
 %runtime._interface = type { i32, i8* }

@main.scalar1 = hidden global i8* null, align 4
@@ -72,11 +71,11 @@ entry:
 }

 ; Function Attrs: nounwind
-define hidden %runtime.funcValue* @main.newFuncValue(i8* %context, i8* %parentHandle) unnamed_addr #0 {
+define hidden { i8*, void ()* }* @main.newFuncValue(i8* %context, i8* %parentHandle) unnamed_addr #0 {
 entry:
  %new = call i8* @runtime.alloc(i32 8, i8* nonnull inttoptr (i32 197 to i8*), i8* undef, i8* null) #0
-  %0 = bitcast i8* %new to %runtime.funcValue*
-  ret %runtime.funcValue* %0
+  %0 = bitcast i8* %new to { i8*, void ()* }*
+  ret { i8*, void ()* }* %0
 }

 ; Function Attrs: nounwind
@@ -0,0 +1,210 @@
+; ModuleID = 'goroutine.go'
+source_filename = "goroutine.go"
+target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
+target triple = "wasm32-unknown-wasi"
+
+%runtime.channel = type { i32, i32, i8, %runtime.channelBlockedList*, i32, i32, i32, i8* }
+%runtime.channelBlockedList = type { %runtime.channelBlockedList*, %"internal/task.Task"*, %runtime.chanSelectState*, { %runtime.channelBlockedList*, i32, i32 } }
+%"internal/task.Task" = type { %"internal/task.Task"*, i8*, i64, %"internal/task.state" }
+%"internal/task.state" = type { i32, i8*, %"internal/task.stackState", i1 }
+%"internal/task.stackState" = type { i32, i32 }
+%runtime.chanSelectState = type { %runtime.channel*, i8* }
+
+@"main$string" = internal unnamed_addr constant [4 x i8] c"test", align 1
+
+declare noalias nonnull i8* @runtime.alloc(i32, i8*, i8*, i8*)
+
+; Function Attrs: nounwind
+define hidden void @main.init(i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  ret void
+}
+
+; Function Attrs: nounwind
+define hidden void @main.regularFunctionGoroutine(i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  call void @"internal/task.start"(i32 ptrtoint (void (i8*)* @"main.regularFunction$gowrapper" to i32), i8* nonnull inttoptr (i32 5 to i8*), i32 8192, i8* undef, i8* null) #0
+  ret void
+}
+
+declare void @main.regularFunction(i32, i8*, i8*)
+
+declare void @runtime.deadlock(i8*, i8*)
+
+; Function Attrs: nounwind
+define linkonce_odr void @"main.regularFunction$gowrapper"(i8* %0) unnamed_addr #1 {
+entry:
+  %unpack.int = ptrtoint i8* %0 to i32
+  call void @main.regularFunction(i32 %unpack.int, i8* undef, i8* undef) #0
+  call void @runtime.deadlock(i8* undef, i8* undef) #0
+  unreachable
+}
+
+declare void @"internal/task.start"(i32, i8*, i32, i8*, i8*)
+
+; Function Attrs: nounwind
+define hidden void @main.inlineFunctionGoroutine(i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  call void @"internal/task.start"(i32 ptrtoint (void (i8*)* @"main.inlineFunctionGoroutine$1$gowrapper" to i32), i8* nonnull inttoptr (i32 5 to i8*), i32 8192, i8* undef, i8* null) #0
+  ret void
+}
+
+; Function Attrs: nounwind
+define hidden void @"main.inlineFunctionGoroutine$1"(i32 %x, i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  ret void
+}
+
+; Function Attrs: nounwind
+define linkonce_odr void @"main.inlineFunctionGoroutine$1$gowrapper"(i8* %0) unnamed_addr #2 {
+entry:
+  %unpack.int = ptrtoint i8* %0 to i32
+  call void @"main.inlineFunctionGoroutine$1"(i32 %unpack.int, i8* undef, i8* undef)
+  call void @runtime.deadlock(i8* undef, i8* undef) #0
+  unreachable
+}
+
+; Function Attrs: nounwind
+define hidden void @main.closureFunctionGoroutine(i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  %n = call i8* @runtime.alloc(i32 4, i8* nonnull inttoptr (i32 3 to i8*), i8* undef, i8* null) #0
+  %0 = bitcast i8* %n to i32*
+  store i32 3, i32* %0, align 4
+  %1 = call i8* @runtime.alloc(i32 8, i8* null, i8* undef, i8* null) #0
+  %2 = bitcast i8* %1 to i32*
+  store i32 5, i32* %2, align 4
+  %3 = getelementptr inbounds i8, i8* %1, i32 4
+  %4 = bitcast i8* %3 to i8**
+  store i8* %n, i8** %4, align 4
+  call void @"internal/task.start"(i32 ptrtoint (void (i8*)* @"main.closureFunctionGoroutine$1$gowrapper" to i32), i8* nonnull %1, i32 8192, i8* undef, i8* null) #0
+  %5 = load i32, i32* %0, align 4
+  call void @runtime.printint32(i32 %5, i8* undef, i8* null) #0
+  ret void
+}
+
+; Function Attrs: nounwind
+define hidden void @"main.closureFunctionGoroutine$1"(i32 %x, i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  %unpack.ptr = bitcast i8* %context to i32*
+  store i32 7, i32* %unpack.ptr, align 4
+  ret void
+}
+
+; Function Attrs: nounwind
+define linkonce_odr void @"main.closureFunctionGoroutine$1$gowrapper"(i8* %0) unnamed_addr #3 {
+entry:
+  %1 = bitcast i8* %0 to i32*
+  %2 = load i32, i32* %1, align 4
+  %3 = getelementptr inbounds i8, i8* %0, i32 4
+  %4 = bitcast i8* %3 to i8**
+  %5 = load i8*, i8** %4, align 4
+  call void @"main.closureFunctionGoroutine$1"(i32 %2, i8* %5, i8* undef)
+  call void @runtime.deadlock(i8* undef, i8* undef) #0
+  unreachable
+}
+
+declare void @runtime.printint32(i32, i8*, i8*)
+
+; Function Attrs: nounwind
+define hidden void @main.funcGoroutine(i8* %fn.context, void ()* %fn.funcptr, i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  %0 = call i8* @runtime.alloc(i32 12, i8* null, i8* undef, i8* null) #0
+  %1 = bitcast i8* %0 to i32*
+  store i32 5, i32* %1, align 4
+  %2 = getelementptr inbounds i8, i8* %0, i32 4
+  %3 = bitcast i8* %2 to i8**
+  store i8* %fn.context, i8** %3, align 4
+  %4 = getelementptr inbounds i8, i8* %0, i32 8
+  %5 = bitcast i8* %4 to void ()**
+  store void ()* %fn.funcptr, void ()** %5, align 4
+  call void @"internal/task.start"(i32 ptrtoint (void (i8*)* @main.funcGoroutine.gowrapper to i32), i8* nonnull %0, i32 8192, i8* undef, i8* null) #0
+  ret void
+}
+
+; Function Attrs: nounwind
+define linkonce_odr void @main.funcGoroutine.gowrapper(i8* %0) unnamed_addr #4 {
+entry:
+  %1 = bitcast i8* %0 to i32*
+  %2 = load i32, i32* %1, align 4
+  %3 = getelementptr inbounds i8, i8* %0, i32 4
+  %4 = bitcast i8* %3 to i8**
+  %5 = load i8*, i8** %4, align 4
+  %6 = getelementptr inbounds i8, i8* %0, i32 8
+  %7 = bitcast i8* %6 to void (i32, i8*, i8*)**
+  %8 = load void (i32, i8*, i8*)*, void (i32, i8*, i8*)** %7, align 4
+  call void %8(i32 %2, i8* %5, i8* undef) #0
+  call void @runtime.deadlock(i8* undef, i8* undef) #0
+  unreachable
+}
+
+; Function Attrs: nounwind
+define hidden void @main.recoverBuiltinGoroutine(i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  ret void
+}
+
+; Function Attrs: nounwind
+define hidden void @main.copyBuiltinGoroutine(i8* %dst.data, i32 %dst.len, i32 %dst.cap, i8* %src.data, i32 %src.len, i32 %src.cap, i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  %copy.n = call i32 @runtime.sliceCopy(i8* %dst.data, i8* %src.data, i32 %dst.len, i32 %src.len, i32 1, i8* undef, i8* null) #0
+  ret void
+}
+
+declare i32 @runtime.sliceCopy(i8* nocapture writeonly, i8* nocapture readonly, i32, i32, i32, i8*, i8*)
+
+; Function Attrs: nounwind
+define hidden void @main.closeBuiltinGoroutine(%runtime.channel* dereferenceable_or_null(32) %ch, i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  call void @runtime.chanClose(%runtime.channel* %ch, i8* undef, i8* null) #0
+  ret void
+}
+
+declare void @runtime.chanClose(%runtime.channel* dereferenceable_or_null(32), i8*, i8*)
+
+; Function Attrs: nounwind
+define hidden void @main.startInterfaceMethod(i32 %itf.typecode, i8* %itf.value, i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  %0 = call i8* @runtime.alloc(i32 16, i8* null, i8* undef, i8* null) #0
+  %1 = bitcast i8* %0 to i8**
+  store i8* %itf.value, i8** %1, align 4
+  %2 = getelementptr inbounds i8, i8* %0, i32 4
+  %.repack = bitcast i8* %2 to i8**
+  store i8* getelementptr inbounds ([4 x i8], [4 x i8]* @"main$string", i32 0, i32 0), i8** %.repack, align 4
+  %.repack1 = getelementptr inbounds i8, i8* %0, i32 8
+  %3 = bitcast i8* %.repack1 to i32*
+  store i32 4, i32* %3, align 4
+  %4 = getelementptr inbounds i8, i8* %0, i32 12
+  %5 = bitcast i8* %4 to i32*
+  store i32 %itf.typecode, i32* %5, align 4
+  call void @"internal/task.start"(i32 ptrtoint (void (i8*)* @"interface:{Print:func:{basic:string}{}}.Print$invoke$gowrapper" to i32), i8* nonnull %0, i32 8192, i8* undef, i8* null) #0
+  ret void
+}
+
+declare void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(i8*, i8*, i32, i32, i8*, i8*) #5
+
+; Function Attrs: nounwind
+define linkonce_odr void @"interface:{Print:func:{basic:string}{}}.Print$invoke$gowrapper"(i8* %0) unnamed_addr #6 {
+entry:
+  %1 = bitcast i8* %0 to i8**
+  %2 = load i8*, i8** %1, align 4
+  %3 = getelementptr inbounds i8, i8* %0, i32 4
+  %4 = bitcast i8* %3 to i8**
+  %5 = load i8*, i8** %4, align 4
+  %6 = getelementptr inbounds i8, i8* %0, i32 8
+  %7 = bitcast i8* %6 to i32*
+  %8 = load i32, i32* %7, align 4
+  %9 = getelementptr inbounds i8, i8* %0, i32 12
+  %10 = bitcast i8* %9 to i32*
+  %11 = load i32, i32* %10, align 4
+  call void @"interface:{Print:func:{basic:string}{}}.Print$invoke"(i8* %2, i8* %5, i32 %8, i32 %11, i8* undef, i8* undef) #0
+  call void @runtime.deadlock(i8* undef, i8* undef) #0
+  unreachable
+}
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind "tinygo-gowrapper"="main.regularFunction" }
+attributes #2 = { nounwind "tinygo-gowrapper"="main.inlineFunctionGoroutine$1" }
+attributes #3 = { nounwind "tinygo-gowrapper"="main.closureFunctionGoroutine$1" }
+attributes #4 = { nounwind "tinygo-gowrapper" }
+attributes #5 = { "tinygo-invoke"="reflect/methods.Print(string)" "tinygo-methods"="reflect/methods.Print(string)" }
+attributes #6 = { nounwind "tinygo-gowrapper"="interface:{Print:func:{basic:string}{}}.Print$invoke" }
@@ -27,3 +27,8 @@ func stringCompareUnequal(s1, s2 string) bool {
 func stringCompareLarger(s1, s2 string) bool {
 	return s1 > s2
 }
+
+func stringLookup(s string, x uint8) byte {
+	// Test that x is correctly extended to an uint before comparison.
+	return s[x]
+}
@@ -78,4 +78,21 @@ entry:

 declare i1 @runtime.stringLess(i8*, i32, i8*, i32, i8*, i8*)

+; Function Attrs: nounwind
+define hidden i8 @main.stringLookup(i8* %s.data, i32 %s.len, i8 %x, i8* %context, i8* %parentHandle) unnamed_addr #0 {
+entry:
+  %0 = zext i8 %x to i32
+  %.not = icmp ult i32 %0, %s.len
+  br i1 %.not, label %lookup.next, label %lookup.throw
+
+lookup.throw:                                     ; preds = %entry
+  call void @runtime.lookupPanic(i8* undef, i8* null) #0
+  unreachable
+
+lookup.next:                                      ; preds = %entry
+  %1 = getelementptr inbounds i8, i8* %s.data, i32 %0
+  %2 = load i8, i8* %1, align 1
+  ret i8 %2
+}
+
 attributes #0 = { nounwind }
@@ -3,12 +3,15 @@
 package goenv

 import (
+	"bytes"
+	"errors"
 	"fmt"
 	"os"
 	"os/exec"
 	"os/user"
 	"path/filepath"
 	"runtime"
+	"strings"
 )

 // Keys is a slice of all available environment variable keys.
@@ -22,6 +25,12 @@ var Keys = []string{
 	"TINYGOROOT",
 }

+func init() {
+	if Get("GOARCH") == "arm" {
+		Keys = append(Keys, "GOARM")
+	}
+}
+
 // TINYGOROOT is the path to the final location for checking tinygo files. If
 // unset (by a -X ldflag), then sourceDir() will fallback to the original build
 // directory.
@@ -41,6 +50,20 @@ func Get(name string) string {
 			return dir
 		}
 		return runtime.GOARCH
+	case "GOARM":
+		if goarm := os.Getenv("GOARM"); goarm != "" {
+			return goarm
+		}
+		if goos := Get("GOOS"); goos == "windows" || goos == "android" {
+			// Assume Windows and Android are running on modern CPU cores.
+			// This matches upstream Go.
+			return "7"
+		}
+		// Default to ARMv6 on other devices.
+		// The difference between ARMv5 and ARMv6 is big, much bigger than the
+		// difference between ARMv6 and ARMv7. ARMv6 binaries are much smaller,
+		// especially when floating point instructions are involved.
+		return "6"
 	case "GOROOT":
 		return getGoroot()
 	case "GOPATH":
@@ -67,11 +90,98 @@ func Get(name string) string {
 		return "1"
 	case "TINYGOROOT":
 		return sourceDir()
+	case "WASMOPT":
+		if path := os.Getenv("WASMOPT"); path != "" {
+			err := wasmOptCheckVersion(path)
+			if err != nil {
+				fmt.Fprintf(os.Stderr, "cannot use %q as wasm-opt (from WASMOPT environment variable): %s", path, err.Error())
+				os.Exit(1)
+			}
+
+			return path
+		}
+
+		return findWasmOpt()
 	default:
 		return ""
 	}
 }

+// Find wasm-opt, or exit with an error.
+func findWasmOpt() string {
+	tinygoroot := sourceDir()
+	searchPaths := []string{
+		tinygoroot + "/bin/wasm-opt",
+		tinygoroot + "/build/wasm-opt",
+	}
+
+	var paths []string
+	for _, path := range searchPaths {
+		if runtime.GOOS == "windows" {
+			path += ".exe"
+		}
+
+		_, err := os.Stat(path)
+		if err != nil && os.IsNotExist(err) {
+			continue
+		}
+
+		paths = append(paths, path)
+	}
+
+	if path, err := exec.LookPath("wasm-opt"); err == nil {
+		paths = append(paths, path)
+	}
+
+	if len(paths) == 0 {
+		fmt.Fprintln(os.Stderr, "error: could not find wasm-opt, set the WASMOPT environment variable to override")
+		os.Exit(1)
+	}
+
+	errs := make([]error, len(paths))
+	for i, path := range paths {
+		err := wasmOptCheckVersion(path)
+		if err == nil {
+			return path
+		}
+
+		errs[i] = err
+	}
+	fmt.Fprintln(os.Stderr, "no usable wasm-opt found, update or run \"make binaryen\"")
+	for i, path := range paths {
+		fmt.Fprintf(os.Stderr, "\t%s: %s\n", path, errs[i].Error())
+	}
+	os.Exit(1)
+	panic("unreachable")
+}
+
+// wasmOptCheckVersion checks if a copy of wasm-opt is usable.
+func wasmOptCheckVersion(path string) error {
+	cmd := exec.Command(path, "--version")
+	var buf bytes.Buffer
+	cmd.Stdout = &buf
+	cmd.Stderr = os.Stderr
+	err := cmd.Run()
+	if err != nil {
+		return err
+	}
+
+	str := buf.String()
+	if strings.Contains(str, "(") {
+		// The git tag may be placed in parentheses after the main version string.
+		str = strings.Split(str, "(")[0]
+	}
+
+	str = strings.TrimSpace(str)
+	var ver uint
+	_, err = fmt.Sscanf(str, "wasm-opt version %d", &ver)
+	if err != nil || ver < 102 {
+		return errors.New("incompatible wasm-opt (need 102 or newer)")
+	}
+
+	return nil
+}
+
 // Return the TINYGOROOT, or exit with an error.
 func sourceDir() string {
 	// Use $TINYGOROOT as root, if available.
@@ -145,6 +145,9 @@ func Run(mod llvm.Module, debug bool) error {
 		if obj.buffer == nil {
 			continue
 		}
+		if obj.constant {
+			continue // constant buffers can't have been modified
+		}
 		initializer, err := obj.buffer.toLLVMValue(obj.llvmGlobal.Type().ElementType(), &mem)
 		if err == errInvalidPtrToIntSize {
 			// This can happen when a previous interp run did not have the
@@ -248,6 +251,9 @@ func RunFunc(fn llvm.Value, debug bool) error {
 		if obj.buffer == nil {
 			continue
 		}
+		if obj.constant {
+			continue // constant, so can't have been modified
+		}
 		initializer, err := obj.buffer.toLLVMValue(obj.llvmGlobal.Type().ElementType(), &mem)
 		if err != nil {
 			return err
@@ -941,6 +941,7 @@ func (r *runner) runAtRuntime(fn *function, inst instruction, locals []value, me
 		agg := operands[0]
 		for i := 0; i < len(indices)-1; i++ {
 			agg = r.builder.CreateExtractValue(agg, int(indices[i]), inst.name+".agg")
+			mem.instructions = append(mem.instructions, agg)
 		}
 		result = r.builder.CreateExtractValue(agg, int(indices[len(indices)-1]), inst.name)
 	case llvm.InsertValue:
@@ -953,11 +954,15 @@ func (r *runner) runAtRuntime(fn *function, inst instruction, locals []value, me
 		for i := 0; i < len(indices)-1; i++ {
 			agg = r.builder.CreateExtractValue(agg, int(indices[i]), inst.name+".agg"+strconv.Itoa(i))
 			aggregates = append(aggregates, agg)
+			mem.instructions = append(mem.instructions, agg)
 		}
 		result = operands[1]
 		for i := len(indices) - 1; i >= 0; i-- {
 			agg := aggregates[i]
 			result = r.builder.CreateInsertValue(agg, result, int(indices[i]), inst.name+".insertvalue"+strconv.Itoa(i))
+			if i != 0 { // don't add last result to mem.instructions as it will be done at the end already
+				mem.instructions = append(mem.instructions, result)
+			}
 		}

 	case llvm.Add:
@@ -41,6 +41,7 @@ type object struct {
 	globalName     string    // name, if not yet created (not guaranteed to be the final name)
 	buffer         value     // buffer with value as given by interp, nil if external
 	size           uint32    // must match buffer.len(), if available
+	constant       bool      // true if this is a constant global
 	marked         uint8     // 0 means unmarked, 1 means external read, 2 means external write
 }

@@ -223,7 +224,7 @@ func (mv *memoryView) hasExternalLoadOrStore(v pointerValue) bool {
 // possible for the interpreter to read from the object.
 func (mv *memoryView) hasExternalStore(v pointerValue) bool {
 	obj := mv.get(v.index())
-	return obj.marked >= 2
+	return obj.marked >= 2 && !obj.constant
 }

 // get returns an object that can only be read from, as it may return an object
@@ -263,6 +264,9 @@ func (mv *memoryView) put(index uint32, obj object) {
 	if checks && mv.get(index).buffer.len(mv.r) != obj.buffer.len(mv.r) {
 		panic("put() with a differently-sized object")
 	}
+	if checks && obj.constant {
+		panic("interp: store to a constant")
+	}
 	mv.objects[index] = obj
 }

@@ -1138,6 +1142,7 @@ func (r *runner) getValue(llvmValue llvm.Value) value {
 				obj.size = uint32(r.targetData.TypeAllocSize(llvmValue.Type().ElementType()))
 				if initializer := llvmValue.Initializer(); !initializer.IsNil() {
 					obj.buffer = r.getValue(initializer)
+					obj.constant = llvmValue.IsGlobalConstant()
 				}
 			} else if !llvmValue.IsAFunction().IsNil() {
 				// OK
@@ -6,6 +6,7 @@ target triple = "x86_64--linux"
@main.nonConst2 = global i64 0
@main.someArray = global [8 x {i16, i32}] zeroinitializer
@main.exportedValue = global [1 x i16*] [i16* @main.exposedValue1]
+@main.exportedConst = constant i64 42
@main.exposedValue1 = global i16 0
@main.exposedValue2 = global i16 0
@main.insertedValue = global {i8, i32, {float, {i64, i16}}} zeroinitializer
@@ -62,6 +63,11 @@ entry:
  call void @modifyExternal(i32* bitcast ([1 x i16*]* @main.exportedValue to i32*))
  store i16 5, i16* @main.exposedValue1

+  ; Test that marking a constant as external still allows loading from it.
+  call void @readExternal(i32* bitcast (i64* @main.exportedConst to i32*))
+  %constLoad = load i64, i64 * @main.exportedConst
+  call void @runtime.printint64(i64 %constLoad)
+
  ; Test that this even propagates through functions.
  call void @modifyExternal(i32* bitcast (void ()* @willModifyGlobal to i32*))
  store i16 7, i16* @main.exposedValue2
@@ -96,6 +102,8 @@ declare i64 @someValue()

 declare void @modifyExternal(i32*)

+declare void @readExternal(i32*)
+
 ; This function will modify an external value. By passing this function as a
 ; function pointer to an external function, @main.exposedValue2 should be
 ; marked as external.
@@ -5,6 +5,7 @@ target triple = "x86_64--linux"
@main.nonConst2 = local_unnamed_addr global i64 0
@main.someArray = global [8 x { i16, i32 }] zeroinitializer
@main.exportedValue = global [1 x i16*] [i16* @main.exposedValue1]
+@main.exportedConst = constant i64 42
@main.exposedValue1 = global i16 0
@main.exposedValue2 = local_unnamed_addr global i16 0
@main.insertedValue = local_unnamed_addr global { i8, i32, { float, { i64, i16 } } } zeroinitializer
@@ -24,6 +25,8 @@ entry:
  call void @modifyExternal(i32* bitcast (i8* getelementptr inbounds (i8, i8* bitcast ([8 x { i16, i32 }]* @main.someArray to i8*), i32 28) to i32*))
  call void @modifyExternal(i32* bitcast ([1 x i16*]* @main.exportedValue to i32*))
  store i16 5, i16* @main.exposedValue1, align 2
+  call void @readExternal(i32* bitcast (i64* @main.exportedConst to i32*))
+  call void @runtime.printint64(i64 42)
  call void @modifyExternal(i32* bitcast (void ()* @willModifyGlobal to i32*))
  store i16 7, i16* @main.exposedValue2, align 2
  call void @modifyExternal(i32* bitcast (void ()* @hasInlineAsm to i32*))
@@ -54,6 +57,8 @@ declare i64 @someValue() local_unnamed_addr

 declare void @modifyExternal(i32*) local_unnamed_addr

+declare void @readExternal(i32*) local_unnamed_addr
+
 define void @willModifyGlobal() {
 entry:
  store i16 8, i16* @main.exposedValue2, align 2
@@ -5,9 +5,12 @@ declare void @externalCall(i64)

@foo.knownAtRuntime = global i64 0
@bar.knownAtRuntime = global i64 0
+@baz.someGlobal = external global [3 x {i64, i32}]
+@baz.someInt = global i32 0

 define void @runtime.initAll() unnamed_addr {
 entry:
+  call void @baz.init(i8* undef, i8* undef)
  call void @foo.init(i8* undef, i8* undef)
  call void @bar.init(i8* undef, i8* undef)
  call void @main.init(i8* undef, i8* undef)
@@ -25,6 +28,14 @@ define internal void @bar.init(i8* %context, i8* %parentHandle) unnamed_addr {
  ret void
 }

+define internal void @baz.init(i8* %context, i8* %parentHandle) unnamed_addr {
+  ; Test extractvalue/insertvalue with more than one index.
+  %val = load [3 x {i64, i32}], [3 x {i64, i32}]* @baz.someGlobal
+  %part = extractvalue [3 x {i64, i32}] %val, 0, 1
+  %val2 = insertvalue [3 x {i64, i32}] %val, i32 5, 2, 1
+  unreachable ; trigger revert
+}
+
 define internal void @main.init(i8* %context, i8* %parentHandle) unnamed_addr {
 entry:
  call void @externalCall(i64 3)
@@ -3,11 +3,14 @@ target triple = "x86_64--linux"

@foo.knownAtRuntime = local_unnamed_addr global i64 0
@bar.knownAtRuntime = local_unnamed_addr global i64 0
+@baz.someGlobal = external local_unnamed_addr global [3 x { i64, i32 }]
+@baz.someInt = local_unnamed_addr global i32 0

 declare void @externalCall(i64) local_unnamed_addr

 define void @runtime.initAll() unnamed_addr {
 entry:
+  call fastcc void @baz.init(i8* undef, i8* undef)
  call fastcc void @foo.init(i8* undef, i8* undef)
  %val = load i64, i64* @foo.knownAtRuntime, align 8
  store i64 %val, i64* @bar.knownAtRuntime, align 8
@@ -19,3 +22,7 @@ define internal fastcc void @foo.init(i8* %context, i8* %parentHandle) unnamed_a
  store i64 5, i64* @foo.knownAtRuntime, align 8
  unreachable
 }
+
+define internal fastcc void @baz.init(i8* %context, i8* %parentHandle) unnamed_addr {
+  unreachable
+}
@@ -228,6 +228,7 @@ func pathsToOverride(needsSyscallPackage bool) map[string]bool {
 		"internal/bytealg/":     false,
 		"internal/reflectlite/": false,
 		"internal/task/":        false,
+		"internal/itoa/":        false, // TODO: Remove when we drop support for go 1.16
 		"machine/":              false,
 		"net/":                  true,
 		"os/":                   true,
@@ -244,8 +244,18 @@ func runPackageTest(config *compileopts.Config, result builder.BuildResult, test
 		cmd.Dir = result.MainDir
 	} else {
 		// Run in an emulator.
-		// TODO: pass the -test.v flag if needed.
 		args := append(config.Target.Emulator[1:], result.Binary)
+		if config.Target.Emulator[0] == "wasmtime" {
+			// allow reading from current directory: --dir=.
+			// mark end of wasmtime arguments and start of program ones: --
+			args = append(args, "--dir=.", "--")
+			if testVerbose {
+				args = append(args, "-test.v")
+			}
+			if testShort {
+				args = append(args, "-test.short")
+			}
+		}
 		cmd = executeCommand(config.Options, config.Target.Emulator[0], args...)
 	}
 	cmd.Stdout = os.Stdout
@@ -1182,6 +1192,7 @@ func main() {
 	options := &compileopts.Options{
 		GOOS:            goenv.Get("GOOS"),
 		GOARCH:          goenv.Get("GOARCH"),
+		GOARM:           goenv.Get("GOARM"),
 		Target:          *target,
 		Opt:             *opt,
 		GC:              *gc,
@@ -1391,6 +1402,7 @@ func main() {
 		fmt.Printf("LLVM triple:       %s\n", config.Triple())
 		fmt.Printf("GOOS:              %s\n", config.GOOS())
 		fmt.Printf("GOARCH:            %s\n", config.GOARCH())
+		fmt.Printf("GOARM:             %s\n", config.GOARM())
 		fmt.Printf("build tags:        %s\n", strings.Join(config.BuildTags(), " "))
 		fmt.Printf("garbage collector: %s\n", config.GC())
 		fmt.Printf("scheduler:         %s\n", config.Scheduler())
@@ -71,11 +71,9 @@ func TestCompiler(t *testing.T) {
 		return
 	}

-	if runtime.GOOS != "windows" {
-		t.Run("Host", func(t *testing.T) {
-			runPlatTests(optionsFromTarget(""), tests, t)
-		})
-	}
+	t.Run("Host", func(t *testing.T) {
+		runPlatTests(optionsFromTarget(""), tests, t)
+	})

 	if testing.Short() {
 		return
@@ -95,13 +93,13 @@ func TestCompiler(t *testing.T) {

 	if runtime.GOOS == "linux" {
 		t.Run("X86Linux", func(t *testing.T) {
-			runPlatTests(optionsFromOSARCH("linux", "386"), tests, t)
+			runPlatTests(optionsFromOSARCH("linux/386"), tests, t)
 		})
 		t.Run("ARMLinux", func(t *testing.T) {
-			runPlatTests(optionsFromOSARCH("linux", "arm"), tests, t)
+			runPlatTests(optionsFromOSARCH("linux/arm/6"), tests, t)
 		})
 		t.Run("ARM64Linux", func(t *testing.T) {
-			runPlatTests(optionsFromOSARCH("linux", "arm64"), tests, t)
+			runPlatTests(optionsFromOSARCH("linux/arm64"), tests, t)
 		})
 		t.Run("WebAssembly", func(t *testing.T) {
 			runPlatTests(optionsFromTarget("wasm"), tests, t)
@@ -113,10 +111,6 @@ func TestCompiler(t *testing.T) {

 	// Test a few build options.
 	t.Run("build-options", func(t *testing.T) {
-		if runtime.GOOS == "windows" {
-			// These tests assume a host that is supported by TinyGo.
-			t.Skip("can't test build options on Windows")
-		}
 		t.Parallel()

 		// Test with few optimizations enabled (no inlining, etc).
@@ -125,6 +119,7 @@ func TestCompiler(t *testing.T) {
 			runTestWithConfig("stdlib.go", t, compileopts.Options{
 				GOOS:   goenv.Get("GOOS"),
 				GOARCH: goenv.Get("GOARCH"),
+				GOARM:  goenv.Get("GOARM"),
 				Opt:    "1",
 			}, nil, nil)
 		})
@@ -136,6 +131,7 @@ func TestCompiler(t *testing.T) {
 			runTestWithConfig("print.go", t, compileopts.Options{
 				GOOS:   goenv.Get("GOOS"),
 				GOARCH: goenv.Get("GOARCH"),
+				GOARM:  goenv.Get("GOARM"),
 				Opt:    "0",
 			}, nil, nil)
 		})
@@ -145,6 +141,7 @@ func TestCompiler(t *testing.T) {
 			runTestWithConfig("ldflags.go", t, compileopts.Options{
 				GOOS:   goenv.Get("GOOS"),
 				GOARCH: goenv.Get("GOARCH"),
+				GOARM:  goenv.Get("GOARM"),
 				GlobalValues: map[string]map[string]string{
 					"main": {
 						"someGlobal": "foobar",
@@ -169,6 +166,14 @@ func runPlatTests(options compileopts.Options, tests []string, t *testing.T) {
 		t.Parallel()
 		runTest("env.go", options, t, []string{"first", "second"}, []string{"ENV1=VALUE1", "ENV2=VALUE2"})
 	})
+	if options.Target == "wasi" {
+		t.Run("alias.go-scheduler-none", func(t *testing.T) {
+			t.Parallel()
+			options := compileopts.Options(options)
+			options.Scheduler = "none"
+			runTest("alias.go", options, t, nil, nil)
+		})
+	}
 	if options.Target == "" || options.Target == "wasi" {
 		t.Run("filesystem.go", func(t *testing.T) {
 			t.Parallel()
@@ -200,15 +205,24 @@ func optionsFromTarget(target string) compileopts.Options {
 		// GOOS/GOARCH are only used if target == ""
 		GOOS:   goenv.Get("GOOS"),
 		GOARCH: goenv.Get("GOARCH"),
+		GOARM:  goenv.Get("GOARM"),
 		Target: target,
 	}
 }

-func optionsFromOSARCH(goos, goarch string) compileopts.Options {
-	return compileopts.Options{
-		GOOS:   goos,
-		GOARCH: goarch,
+// optionsFromOSARCH returns a set of options based on the "osarch" string. This
+// string is in the form of "os/arch/subarch", with the subarch only sometimes
+// being necessary. Examples are "darwin/amd64" or "linux/arm/7".
+func optionsFromOSARCH(osarch string) compileopts.Options {
+	parts := strings.Split(osarch, "/")
+	options := compileopts.Options{
+		GOOS:   parts[0],
+		GOARCH: parts[1],
 	}
+	if options.GOARCH == "arm" {
+		options.GOARM = parts[2]
+	}
+	return options
 }

 func runTest(name string, options compileopts.Options, t *testing.T, cmdArgs, environmentVars []string) {
@@ -278,6 +292,9 @@ func runTestWithConfig(name string, t *testing.T, options compileopts.Options, c

 	// Build the test binary.
 	binary := filepath.Join(tmpdir, "test")
+	if spec.GOOS == "windows" {
+		binary += ".exe"
+	}
 	err = runBuild("./"+path, binary, &options)
 	if err != nil {
 		printCompilerError(t.Log, err)
@@ -319,7 +336,13 @@ func runTestWithConfig(name string, t *testing.T, options compileopts.Options, c
 	}
 	go func() {
 		// Terminate the process if it runs too long.
-		timer := time.NewTimer(10 * time.Second)
+		maxDuration := 10 * time.Second
+		if runtime.GOOS == "windows" {
+			// For some reason, tests on Windows can take around
+			// 30s to complete. TODO: investigate why and fix this.
+			maxDuration = 40 * time.Second
+		}
+		timer := time.NewTimer(maxDuration)
 		select {
 		case <-runComplete:
 			timer.Stop()
@@ -349,7 +372,7 @@ func runTestWithConfig(name string, t *testing.T, options compileopts.Options, c
 		t.Log("failed to run:", err)
 		fail = true
 	} else if !bytes.Equal(expected, actual) {
-		t.Log("output did not match")
+		t.Logf("output did not match (expected %d bytes, got %d bytes):", len(expected), len(actual))
 		fail = true
 	}

@@ -0,0 +1,42 @@
+package rand
+
+import "errors"
+
+func init() {
+	Reader = &reader{}
+}
+
+type reader struct {
+}
+
+var errRandom = errors.New("failed to obtain random data from rand_s")
+
+func (r *reader) Read(b []byte) (n int, err error) {
+	if len(b) == 0 {
+		return
+	}
+
+	var randomByte uint32
+	for i := range b {
+		// Call rand_s every four bytes because it's a C int (always 32-bit in
+		// Windows).
+		if i%4 == 0 {
+			errCode := libc_rand_s(&randomByte)
+			if errCode != 0 {
+				// According to the documentation, it can return an error.
+				return n, errRandom
+			}
+		} else {
+			randomByte >>= 8
+		}
+		b[i] = byte(randomByte)
+	}
+
+	return len(b), nil
+}
+
+// Cryptographically secure random number generator.
+// https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/rand-s?view=msvc-170
+// errno_t rand_s(unsigned int* randomValue);
+//export rand_s
+func libc_rand_s(randomValue *uint32) int32
@@ -2,6 +2,7 @@
 // Hardfault aliases for definitions that have inconsistent naming (which are
 // auto-generated by gen-device-svd.go) among devices in package nxp.

+//go:build nxp && !mimxrt1062
 // +build nxp,!mimxrt1062

 package nxp
@@ -2,12 +2,12 @@
 // Type definitions, fields, and constants associated with various clocks and
 // peripherals of the NXP MIMXRT1062.

+//go:build nxp && mimxrt1062
 // +build nxp,mimxrt1062

 package nxp

 import (
-	"device/arm"
 	"runtime/volatile"
 	"unsafe"
 )
@@ -378,6 +378,30 @@ func (clk Clock) setCcm(value uint32) {
 	}
 }

+func setSysPfd(value ...uint32) {
+	for i, val := range value {
+		pfd528 := CCM_ANALOG.PFD_528.Get() &
+			^((CCM_ANALOG_PFD_528_PFD0_CLKGATE_Msk | CCM_ANALOG_PFD_528_PFD0_FRAC_Msk) << (8 * uint32(i)))
+		frac := (val << CCM_ANALOG_PFD_528_PFD0_FRAC_Pos) & CCM_ANALOG_PFD_528_PFD0_FRAC_Msk
+		// disable the clock output first
+		CCM_ANALOG.PFD_528.Set(pfd528 | (CCM_ANALOG_PFD_528_PFD0_CLKGATE_Msk << (8 * uint32(i))))
+		// set the new value and enable output
+		CCM_ANALOG.PFD_528.Set(pfd528 | (frac << (8 * uint32(i))))
+	}
+}
+
+func setUsb1Pfd(value ...uint32) {
+	for i, val := range value {
+		pfd480 := CCM_ANALOG.PFD_480.Get() &
+			^((CCM_ANALOG_PFD_480_PFD0_CLKGATE_Msk | CCM_ANALOG_PFD_480_PFD0_FRAC_Msk) << (8 * uint32(i)))
+		frac := (val << CCM_ANALOG_PFD_480_PFD0_FRAC_Pos) & CCM_ANALOG_PFD_480_PFD0_FRAC_Msk
+		// disable the clock output first
+		CCM_ANALOG.PFD_480.Set(pfd480 | (CCM_ANALOG_PFD_480_PFD0_CLKGATE_Msk << (8 * uint32(i))))
+		// set the new value and enable output
+		CCM_ANALOG.PFD_480.Set(pfd480 | (frac << (8 * uint32(i))))
+	}
+}
+
 // PLL configuration for ARM
 type ClockConfigArmPll struct {
 	LoopDivider uint32 // PLL loop divider. Valid range for divider value: 54-108. Fout=Fin*LoopDivider/2.
@@ -448,178 +472,59 @@ func (cfg ClockConfigSysPll) Configure(pfd ...uint32) {
 	setSysPfd(pfd...)
 }

-func setSysPfd(value ...uint32) {
-	for i, val := range value {
-		pfd528 := CCM_ANALOG.PFD_528.Get() &
-			^((CCM_ANALOG_PFD_528_PFD0_CLKGATE_Msk | CCM_ANALOG_PFD_528_PFD0_FRAC_Msk) << (8 * uint32(i)))
-		frac := (val << CCM_ANALOG_PFD_528_PFD0_FRAC_Pos) & CCM_ANALOG_PFD_528_PFD0_FRAC_Msk
-		// disable the clock output first
-		CCM_ANALOG.PFD_528.Set(pfd528 | (CCM_ANALOG_PFD_528_PFD0_CLKGATE_Msk << (8 * uint32(i))))
-		// set the new value and enable output
-		CCM_ANALOG.PFD_528.Set(pfd528 | (frac << (8 * uint32(i))))
-	}
-}
-
-// PHY configuration for USB HS
-type ClockConfigUsbPhy struct {
-	Instance  uint8  // USB PHY number (1 or 2)
-	XtalFreq  uint32 // External reference clock frequency (Hz)
-	DCal      uint32 // Decode to trim nominal 17.78mA current source
-	TxCal45DP uint32 // Decode to trim nominal 45-Ohm series Rp on USB D+
-	TxCal45DM uint32 // Decode to trim nominal 45-Ohm series Rp on USB D-
-	PllConfig ClockConfigUsbPll
-}
-
-// Configure initializes the USB HS (480 Mbit/s) PHY and PLL clocks, including
-// the USB +3V regulator (PMU), for use as either USB host or device.
-func (cfg ClockConfigUsbPhy) Configure() {
-
-	var (
-		usb           *USB_Type
-		phy           *USBPHY_Type
-		chrgDetectReg *volatile.Register32
-		chrgDetectMsk uint32
-	)
-
-	// Select appropriate peripherals based on receiver Instance
-	switch cfg.Instance {
-	case 1:
-		usb = USB1    // Select USB1 HS PHY/PLL
-		phy = USBPHY1 //
-		chrgDetectReg = &USB_ANALOG.USB1_CHRG_DETECT_SET
-		chrgDetectMsk = USB_ANALOG_USB1_CHRG_DETECT_SET_CHK_CHRG_B |
-			USB_ANALOG_USB1_CHRG_DETECT_SET_EN_B
-	case 2:
-		usb = USB2    // Select USB2 HS PHY/PLL
-		phy = USBPHY2 //
-		chrgDetectReg = &USB_ANALOG.USB2_CHRG_DETECT_SET
-		chrgDetectMsk = USB_ANALOG_USB2_CHRG_DETECT_SET_CHK_CHRG_B |
-			USB_ANALOG_USB2_CHRG_DETECT_SET_EN_B
-	default:
-		panic("nxp: invalid USB PHY")
-	}
-
-	// Configure and enable USB PLL clocks
-	cfg.PllConfig.Configure()
-
-	// Release PHY from reset
-	phy.CTRL.ClearBits(USBPHY_CTRL_SFTRST)
-	phy.CTRL.ClearBits(USBPHY_CTRL_CLKGATE)
-
-	// Enable power to USB PHY
-	phy.PWD.Set(0)
-	phy.CTRL.SetBits(USBPHY_CTRL_ENAUTOCLR_PHY_PWD | USBPHY_CTRL_ENAUTOCLR_CLKGATE |
-		// enable support for low-speed device connection, direct and indirect (hub)
-		USBPHY_CTRL_ENUTMILEVEL2 | USBPHY_CTRL_ENUTMILEVEL3)
-
-	// Enable USB HS clocks gate
-	ClockIpUsbOh3.Enable(true)
-	// Reset USB peripheral
-	usb.USBCMD.SetBits(USB_USBCMD_RST)
-
-	// Add a delay after RST to ensure there is a USB D+ pullup sequence
-	nopDelay(400000)
-
-	// Enable USB LDO
-	PMU.REG_3P0.Set((PMU.REG_3P0.Get() & ^uint32(PMU_REG_3P0_OUTPUT_TRG_Msk)) |
-		(0x17 << PMU_REG_3P0_OUTPUT_TRG_Pos) | PMU_REG_3P0_ENABLE_LINREG)
-
-	// check whether we are connected to USB charger
-	chrgDetectReg.Set(chrgDetectMsk)
-
-	// Decode to trim nominal 17.78mA source for HS TX on USB D+/D-
-	phy.TX.Set((phy.TX.Get() &
-		^uint32(USBPHY_TX_D_CAL_Msk|USBPHY_TX_TXCAL45DN_Msk|USBPHY_TX_TXCAL45DP_Msk)) |
-		((cfg.DCal << USBPHY_TX_D_CAL_Pos) & USBPHY_TX_D_CAL_Msk) |
-		((cfg.TxCal45DM << USBPHY_TX_TXCAL45DN_Pos) & USBPHY_TX_TXCAL45DN_Msk) |
-		((cfg.TxCal45DP << USBPHY_TX_TXCAL45DP_Pos) & USBPHY_TX_TXCAL45DP_Msk))
-}
-
 // PLL configuration for USB
 type ClockConfigUsbPll struct {
-	Instance    uint8    // USB PLL number (1 or 2)
-	LoopDivider uint8    // PLL loop divider (0 [Fout=Fref*20] or 1 [Fout=Fref*22])
-	Src         uint8    // PLL bypass clock source (0 [OSC24M] or 1 [CLK1_P & CLK1_N])
-	Pfd         []uint32 // Phase fractional divisors (len=4, or nil for boot default)
+	Instance    uint8 // USB PLL number (1 or 2)
+	LoopDivider uint8 // PLL loop divider: 0 - Fout=Fref*20, 1 - Fout=Fref*22
+	Src         uint8 // Pll clock source, reference _clock_pll_clk_src
 }

-func (cfg ClockConfigUsbPll) Configure() {
-	// select USB peripheral registers based on receiver's Instance
-	switch cfg.Instance {
-	case 1: // USB1 PLL
-		if CCM_ANALOG.PLL_USB1.HasBits(CCM_ANALOG_PLL_USB1_ENABLE) {
-			// PLL already configured, enable USB clocks
-			CCM_ANALOG.PLL_USB1.SetBits(CCM_ANALOG_PLL_USB1_EN_USB_CLKS)
-		} else {
-			// bypass PLL first
-			src := (uint32(cfg.Src) << CCM_ANALOG_PLL_USB1_BYPASS_CLK_SRC_Pos) &
-				CCM_ANALOG_PLL_USB1_BYPASS_CLK_SRC_Msk
-			CCM_ANALOG.PLL_USB1.Set(
-				(CCM_ANALOG.PLL_USB1.Get() & ^uint32(CCM_ANALOG_PLL_USB1_BYPASS_CLK_SRC_Msk)) |
-					CCM_ANALOG_PLL_USB1_BYPASS | src)
-			// reconfigure PLL
-			sel := (uint32(cfg.LoopDivider) << CCM_ANALOG_PLL_USB1_DIV_SELECT_Pos) &
-				CCM_ANALOG_PLL_USB1_DIV_SELECT_Msk
-			CCM_ANALOG.PLL_USB1.Set(
-				(CCM_ANALOG.PLL_USB1.Get() & ^uint32(CCM_ANALOG_PLL_USB1_DIV_SELECT_Msk)) |
-					CCM_ANALOG_PLL_USB1_ENABLE | CCM_ANALOG_PLL_USB1_POWER |
-					CCM_ANALOG_PLL_USB1_EN_USB_CLKS | sel)
-			for !CCM_ANALOG.PLL_USB1.HasBits(CCM_ANALOG_PLL_USB1_LOCK) {
-			}
-			// disable bypass
-			CCM_ANALOG.PLL_USB1.ClearBits(CCM_ANALOG_PLL_USB1_BYPASS)
+func (cfg ClockConfigUsbPll) Configure(pfd ...uint32) {

-			// update PFDs (if provided)
-			if nil != cfg.Pfd {
-				setUsb1Pfd(cfg.Pfd...)
-			}
+	switch cfg.Instance {
+	case 1:
+
+		// bypass PLL first
+		src := (uint32(cfg.Src) << CCM_ANALOG_PLL_USB1_BYPASS_CLK_SRC_Pos) & CCM_ANALOG_PLL_USB1_BYPASS_CLK_SRC_Msk
+		CCM_ANALOG.PLL_USB1.Set(
+			(CCM_ANALOG.PLL_USB1.Get() & ^uint32(CCM_ANALOG_PLL_USB1_BYPASS_CLK_SRC_Msk)) |
+				CCM_ANALOG_PLL_USB1_BYPASS_Msk | src)
+
+		sel := uint32((cfg.LoopDivider << CCM_ANALOG_PLL_USB1_DIV_SELECT_Pos) & CCM_ANALOG_PLL_USB1_DIV_SELECT_Msk)
+		CCM_ANALOG.PLL_USB1_SET.Set(
+			(CCM_ANALOG.PLL_USB1.Get() & ^uint32(CCM_ANALOG_PLL_USB1_DIV_SELECT_Msk)) |
+				CCM_ANALOG_PLL_USB1_ENABLE_Msk | CCM_ANALOG_PLL_USB1_POWER_Msk |
+				CCM_ANALOG_PLL_USB1_EN_USB_CLKS_Msk | sel)
+
+		for !CCM_ANALOG.PLL_USB1.HasBits(CCM_ANALOG_PLL_USB1_LOCK_Msk) {
 		}
-	case 2: // USB2 PLL
-		if CCM_ANALOG.PLL_USB2.HasBits(CCM_ANALOG_PLL_USB2_ENABLE) {
-			// PLL already configured, enable USB clocks
-			CCM_ANALOG.PLL_USB2.SetBits(CCM_ANALOG_PLL_USB2_EN_USB_CLKS)
-		} else {
-			// bypass PLL first
-			src := (uint32(cfg.Src) << CCM_ANALOG_PLL_USB2_BYPASS_CLK_SRC_Pos) &
-				CCM_ANALOG_PLL_USB2_BYPASS_CLK_SRC_Msk
-			CCM_ANALOG.PLL_USB2.Set(
-				(CCM_ANALOG.PLL_USB2.Get() & ^uint32(CCM_ANALOG_PLL_USB2_BYPASS_CLK_SRC_Msk)) |
-					CCM_ANALOG_PLL_USB2_BYPASS | src)
-			// reconfigure PLL
-			sel := (uint32(cfg.LoopDivider) << CCM_ANALOG_PLL_USB2_DIV_SELECT_Pos) &
-				CCM_ANALOG_PLL_USB2_DIV_SELECT_Msk
-			CCM_ANALOG.PLL_USB2.Set(
-				(CCM_ANALOG.PLL_USB2.Get() & ^uint32(CCM_ANALOG_PLL_USB2_DIV_SELECT_Msk)) |
-					CCM_ANALOG_PLL_USB2_ENABLE | CCM_ANALOG_PLL_USB2_POWER |
-					CCM_ANALOG_PLL_USB2_EN_USB_CLKS | sel)
-			for !CCM_ANALOG.PLL_USB2.HasBits(CCM_ANALOG_PLL_USB2_LOCK) {
-			}
-			// disable bypass
-			CCM_ANALOG.PLL_USB2.ClearBits(CCM_ANALOG_PLL_USB2_BYPASS)
+
+		// disable bypass
+		CCM_ANALOG.PLL_USB1_CLR.Set(CCM_ANALOG_PLL_USB1_BYPASS_Msk)
+
+		// update PFDs after update
+		setUsb1Pfd(pfd...)
+
+	case 2:
+		// bypass PLL first
+		src := (uint32(cfg.Src) << CCM_ANALOG_PLL_USB2_BYPASS_CLK_SRC_Pos) & CCM_ANALOG_PLL_USB2_BYPASS_CLK_SRC_Msk
+		CCM_ANALOG.PLL_USB2.Set(
+			(CCM_ANALOG.PLL_USB2.Get() & ^uint32(CCM_ANALOG_PLL_USB2_BYPASS_CLK_SRC_Msk)) |
+				CCM_ANALOG_PLL_USB2_BYPASS_Msk | src)
+
+		sel := uint32((cfg.LoopDivider << CCM_ANALOG_PLL_USB2_DIV_SELECT_Pos) & CCM_ANALOG_PLL_USB2_DIV_SELECT_Msk)
+		CCM_ANALOG.PLL_USB2.Set(
+			(CCM_ANALOG.PLL_USB2.Get() & ^uint32(CCM_ANALOG_PLL_USB2_DIV_SELECT_Msk)) |
+				CCM_ANALOG_PLL_USB2_ENABLE_Msk | CCM_ANALOG_PLL_USB2_POWER_Msk |
+				CCM_ANALOG_PLL_USB2_EN_USB_CLKS_Msk | sel)
+
+		for !CCM_ANALOG.PLL_USB2.HasBits(CCM_ANALOG_PLL_USB2_LOCK_Msk) {
 		}

+		// disable bypass
+		CCM_ANALOG.PLL_USB2.ClearBits(CCM_ANALOG_PLL_USB2_BYPASS_Msk)
+
 	default:
 		panic("nxp: invalid USB PLL")
 	}
 }
-
-func setUsb1Pfd(value ...uint32) {
-	for i, val := range value {
-		pfd480 := CCM_ANALOG.PFD_480.Get() &
-			^((CCM_ANALOG_PFD_480_PFD0_CLKGATE_Msk | CCM_ANALOG_PFD_480_PFD0_FRAC_Msk) << (8 * uint32(i)))
-		frac := (val << CCM_ANALOG_PFD_480_PFD0_FRAC_Pos) & CCM_ANALOG_PFD_480_PFD0_FRAC_Msk
-		// disable the clock output first
-		CCM_ANALOG.PFD_480.Set(pfd480 | (CCM_ANALOG_PFD_480_PFD0_CLKGATE_Msk << (8 * uint32(i))))
-		// set the new value and enable output
-		CCM_ANALOG.PFD_480.Set(pfd480 | (frac << (8 * uint32(i))))
-	}
-}
-
-// We cannot use the sleep timer from this context (import cycle), but we need
-// an approximate method to spin CPU cycles for short periods of time.
-// go:inline
-func nopDelay(cycles uint32) {
-	for i := uint32(0); i < cycles; i++ {
-		arm.Asm(`nop`)
-	}
-}
@@ -2,6 +2,7 @@
 // Hardfault aliases for definitions that have inconsistent naming (which are
 // auto-generated by gen-device-svd.go) among devices in package nxp.

+//go:build nxp && mimxrt1062
 // +build nxp,mimxrt1062

 package nxp
@@ -278,83 +278,3 @@ func enableDcache(enable bool) {
 		}
 	}
 }
-
-// FlushDcache flushes data from cache to memory
-//
-// Normally FlushDcache is used when metadata written to memory will be used by
-// a DMA or a bus-controller peripheral. Any data in the cache is written to
-// memory. A copy remains in the cache, so this is typically used with special
-// fields you will want to quickly access in the future. For data transmission,
-// use FlushDeleteDcache.
-//go:inline
-func FlushDcache(addr, size uintptr) {
-	location := addr & 0xFFFFFFE0
-	endAddr := addr + size
-	arm.AsmFull(`
-		dsb 0xF
-	`, nil)
-	for {
-		SystemControl.DCCMVAC.Set(uint32(location))
-		location += 32
-		if location >= endAddr {
-			break
-		}
-	}
-	arm.AsmFull(`
-		dsb 0xF
-		isb 0xF
-	`, nil)
-}
-
-// DeleteDcache deletes data from the cache, without touching memory.
-//
-// Normally DeleteDcache is used before receiving data via DMA or from
-// bus-controller peripherals which write to memory. You want to delete anything
-// the cache may have stored, so your next read is certain to access the
-// physical memory.
-//go:inline
-func DeleteDcache(addr, size uintptr) {
-	location := addr & 0xFFFFFFE0
-	endAddr := addr + size
-	arm.AsmFull(`
-		dsb 0xF
-	`, nil)
-	for {
-		SystemControl.DCIMVAC.Set(uint32(location))
-		location += 32
-		if location >= endAddr {
-			break
-		}
-	}
-	arm.AsmFull(`
-		dsb 0xF
-		isb 0xF
-	`, nil)
-}
-
-// FlushDeleteDcache flushes data from cache to memory, and delete it from the
-// cache
-//
-// Normally FlushDeleteDcache is used when transmitting data via DMA or
-// bus-controller peripherals which read from memory. You want any cached data
-// written to memory, and then removed from the cache, because you no longer
-// need to access the data after transmission.
-//go:inline
-func FlushDeleteDcache(addr, size uintptr) {
-	location := addr & 0xFFFFFFE0
-	endAddr := addr + size
-	arm.AsmFull(`
-		dsb 0xF
-	`, nil)
-	for {
-		SystemControl.DCCIMVAC.Set(uint32(location))
-		location += 32
-		if location >= endAddr {
-			break
-		}
-	}
-	arm.AsmFull(`
-		dsb 0xF
-		isb 0xF
-	`, nil)
-}
@@ -1,6 +1,7 @@
 // Hand created file. DO NOT DELETE.
 // atsamd51x bitfield definitions that are not auto-generated by gen-device-svd.go

+//go:build sam && atsamd51
 // +build sam,atsamd51

 // These are the supported pchctrl function numberings on the atsamd51x
@@ -1,6 +1,7 @@
 // Hand created file. DO NOT DELETE.
 // atsamd51x bitfield definitions that are not auto-generated by gen-device-svd.go

+//go:build sam && atsame5x
 // +build sam,atsame5x

 // These are the supported pchctrl function numberings on the atsamd51x
@@ -1,251 +0,0 @@
-package main
-
-import (
-	"machine"
-	"machine/usb"
-	"time"
-)
-
-var keyboard = machine.HID0.Keyboard()
-
-func main() {
-
-	println("USB HID keyboard demo")
-
-	for {
-		time.Sleep(5 * time.Second)
-
-		// Open a new text editor
-		keyboard.Down(usb.KeyModifierAlt)
-		keyboard.Press(usb.KeySpace)
-		keyboard.Up(usb.KeyModifierAlt)
-		time.Sleep(2 * time.Second)
-		keyboard.Write([]byte("kate"))
-		time.Sleep(time.Second)
-		keyboard.Press(usb.KeyEnter)
-
-		time.Sleep(5 * time.Second)
-
-		// Use the io.Writer interface
-		keyboard.Write([]byte("TinyGo USB Keyboard Control Test\n"))
-		time.Sleep(2 * time.Second)
-
-		// Or manually specify keycodes and Unicode codepoints
-		testKeys([]Key{
-			// Print alphabet out-of-order
-			{Press: usb.KeyX},
-			{Press: usb.KeyY},
-			{Press: usb.KeyZ},
-			{Press: usb.KeyG},
-			{Press: usb.KeyH},
-			{Press: usb.KeyI},
-			{Press: usb.KeyJ},
-			{Press: usb.KeyK},
-			{Press: usb.KeyL},
-			{Press: usb.KeyM},
-			{Press: usb.KeyN},
-			{Press: usb.KeyO},
-			{Press: usb.KeyP},
-			{Press: usb.KeyQ},
-			{Press: usb.KeyR},
-			{Press: usb.KeyS},
-			{Press: usb.KeyT},
-			{Press: usb.KeyA},
-			{Press: usb.KeyB},
-			{Press: usb.KeyC},
-			{Press: usb.KeyD},
-			{Press: usb.KeyE},
-			{Press: usb.KeyF},
-			{Press: usb.KeyU},
-			{Press: usb.KeyV},
-			{Press: usb.KeyW},
-			// Pause 1 second
-			{Time: time.Second},
-			// Move cursor left x3
-			{Press: usb.KeyLeft},
-			{Press: usb.KeyLeft},
-			{Press: usb.KeyLeft},
-			// Pause 1 second
-			{Time: time.Second},
-			// Highlight 6 symbols to the left
-			{Down: usb.KeyModifierShift},
-			{Press: usb.KeyLeft},
-			{Press: usb.KeyLeft},
-			{Press: usb.KeyLeft},
-			{Press: usb.KeyLeft},
-			{Press: usb.KeyLeft},
-			{Press: usb.KeyLeft},
-			{Up: usb.KeyModifierShift},
-			// Pause 1 second
-			{Time: time.Second},
-			// Use Ctrl-X to cut
-			{Down: usb.KeyModifierCtrl, Press: usb.KeyX, Up: usb.KeyModifierCtrl},
-			// Pause 1 second
-			{Time: time.Second},
-			// Move to beginning of line
-			{Press: usb.KeyHome},
-			// Pause 1 second
-			{Time: time.Second},
-			// Use Ctrl-V to paste
-			{Down: usb.KeyModifierCtrl, Press: usb.KeyV, Up: usb.KeyModifierCtrl},
-			// Pause 1 second
-			{Time: time.Second},
-			// Highlight 3 symbols to the right
-			{Down: usb.KeyModifierShift},
-			{Press: usb.KeyRight},
-			{Press: usb.KeyRight},
-			{Press: usb.KeyRight},
-			{Up: usb.KeyModifierShift},
-			// Use Ctrl-X to cut
-			{Down: usb.KeyModifierCtrl, Press: usb.KeyX, Up: usb.KeyModifierCtrl},
-			// Pause 1 second
-			{Time: time.Second},
-			// Move to end of line
-			{Press: usb.KeyEnd},
-			// Pause 1 second
-			{Time: time.Second},
-			// Use Ctrl-V to paste
-			{Down: usb.KeyModifierCtrl, Press: usb.KeyV, Up: usb.KeyModifierCtrl},
-			// Pause 1 second
-			{Time: time.Second},
-			// Newline
-			{Press: usb.KeyEnter},
-			{Press: usb.KeyEnter},
-		}, 150*time.Millisecond)
-
-		// Highlight all text and delete
-		keyboard.Down(usb.KeyModifierCtrl)
-		keyboard.Press(usb.KeyA)
-		keyboard.Up(usb.KeyModifierCtrl)
-		time.Sleep(time.Second)
-		keyboard.Press(usb.KeyDelete)
-		time.Sleep(time.Second)
-
-		// Close window
-		keyboard.Down(usb.KeyModifierCtrl)
-		keyboard.Press(usb.KeyQ)
-		keyboard.Up(usb.KeyModifierCtrl)
-		time.Sleep(2 * time.Second)
-		// Confirm discard file
-		keyboard.Down(usb.KeyModifierAlt)
-		keyboard.Press(usb.KeyD)
-		keyboard.Up(usb.KeyModifierAlt)
-
-		time.Sleep(5 * time.Second)
-
-		// Open a new terminal
-		keyboard.Down(usb.KeyModifierAlt)
-		keyboard.Press(usb.KeySpace)
-		keyboard.Up(usb.KeyModifierAlt)
-		time.Sleep(2 * time.Second)
-		keyboard.Write([]byte("konsole"))
-		keyboard.Press(usb.KeyEnter)
-
-		time.Sleep(5 * time.Second)
-
-		// Open serial connection
-		keyboard.Write([]byte("screen /dev/ttyACM0 115200"))
-		time.Sleep(2 * time.Second)
-		keyboard.Press(usb.KeyEnter)
-		time.Sleep(4 * time.Second)
-
-		// Write to UART
-		keyboard.Write([]byte("hello!"))
-		time.Sleep(time.Second)
-		keyboard.Press(usb.KeyEnter)
-		time.Sleep(2 * time.Second)
-		keyboard.Write([]byte("NO U"))
-		time.Sleep(time.Second)
-		keyboard.Press(usb.KeyEnter)
-		time.Sleep(2 * time.Second)
-
-		// Close serial connection
-		keyboard.Down(usb.KeyModifierCtrl)
-		keyboard.Press(usb.KeyX)
-		keyboard.Up(usb.KeyModifierCtrl)
-		time.Sleep(time.Second)
-		keyboard.Press(usb.KeyBackslash)
-		time.Sleep(time.Second)
-		keyboard.Press(usb.KeyY)
-		keyboard.Press(usb.KeyEnter)
-		time.Sleep(2 * time.Second)
-
-		// Close terminal
-		keyboard.Down(usb.KeyModifierCtrl)
-		keyboard.Press(usb.KeyD)
-		keyboard.Up(usb.KeyModifierCtrl)
-
-		time.Sleep(25 * time.Second)
-	}
-}
-
-type Key struct {
-	Press usb.Keycode
-	Down  usb.Keycode
-	Up    usb.Keycode
-	Time  time.Duration
-}
-
-func testKeys(key []Key, delay time.Duration) {
-	for _, k := range key {
-		if 0 != k.Down {
-			keyboard.Down(k.Down)
-		}
-		if 0 != k.Press {
-			keyboard.Press(k.Press)
-		}
-		if 0 != k.Up {
-			keyboard.Up(k.Up)
-		}
-		if 0 != k.Time {
-			time.Sleep(k.Time)
-		} else {
-			time.Sleep(delay)
-		}
-	}
-}
-
-func testInternationalLayout() {
-	// International keyboard layouts also supported
-	keyboard.Write([]byte("TinyGo USB Keyboard Layout Test\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Lowercase:  abcdefghijklmnopqrstuvwxyz\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Uppercase:  ABCDEFGHIJKLMNOPQRSTUVWXYZ\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Numbers:    0123456789\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Symbols1:   !\"#$%&'()*+,-./\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Symbols2:   :;<=>?[\\]^_`{|}~\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Symbols3:   ¡¢£¤¥¦§¨©ª«¬®¯°±\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Symbols4:   ²³´µ¶·¸¹º»¼½¾¿×÷\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Grave:      ÀÈÌÒÙàèìòù\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Acute:      ÁÉÍÓÚÝáéíóúý\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Circumflex: ÂÊÎÔÛâêîôû\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Tilde:      ÃÑÕãñõ\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Diaeresis:  ÄËÏÖÜäëïöüÿ\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Cedilla:    Çç\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Ring Above: Åå\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("AE:         Ææ\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Thorn:      Þþ\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Sharp S:    ß\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("O-Stroke:   Øø\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Eth:        Ðð\n"))
-	time.Sleep(250 * time.Millisecond)
-	keyboard.Write([]byte("Euro:       €\n"))
-}
@@ -1,37 +0,0 @@
-// This is a echo console running on the device UART.
-// Connect using default baudrate for this hardware, 8-N-1 with your terminal program.
-package main
-
-import (
-	"machine"
-	"time"
-)
-
-func main() {
-	uart := machine.UART0
-	uart.Write([]byte("Echo console enabled. Type something then press enter:\r\n"))
-
-	input := make([]byte, 4096)
-	i := 0
-	for {
-		if uart.Buffered() > 0 {
-			data, _ := uart.ReadByte()
-
-			switch data {
-			case 13:
-				// return key
-				uart.Write([]byte("\r\n"))
-				uart.Write([]byte("You typed: "))
-				uart.Write(input[:i])
-				uart.Write([]byte("\r\n"))
-				i = 0
-			default:
-				// just echo the character
-				uart.WriteByte(data)
-				input[i] = data
-				i++
-			}
-		}
-		time.Sleep(10 * time.Millisecond)
-	}
-}
@@ -0,0 +1,2 @@
+internal/itoa is new to go as of 1.17.
+This directory should be removed when tinygo drops support for go 1.16.
@@ -0,0 +1,33 @@
+// Copyright 2021 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Simple conversions to avoid depending on strconv.
+
+package itoa
+
+// Itoa converts val to a decimal string.
+func Itoa(val int) string {
+	if val < 0 {
+		return "-" + Uitoa(uint(-val))
+	}
+	return Uitoa(uint(val))
+}
+
+// Uitoa converts val to a decimal string.
+func Uitoa(val uint) string {
+	if val == 0 { // avoid string allocation
+		return "0"
+	}
+	var buf [20]byte // big enough for 64bit value base 10
+	i := len(buf) - 1
+	for val >= 10 {
+		q := val / 10
+		buf[i] = byte('0' + val - q*10)
+		i--
+		val = q
+	}
+	// val < 10
+	buf[i] = byte('0' + val)
+	return string(buf[i:])
+}
@@ -0,0 +1,40 @@
+// Copyright 2021 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package itoa_test
+
+import (
+	"fmt"
+	"internal/itoa"
+	"math"
+	"testing"
+)
+
+var (
+	minInt64  int64  = math.MinInt64
+	maxInt64  int64  = math.MaxInt64
+	maxUint64 uint64 = math.MaxUint64
+)
+
+func TestItoa(t *testing.T) {
+	tests := []int{int(minInt64), math.MinInt32, -999, -100, -1, 0, 1, 100, 999, math.MaxInt32, int(maxInt64)}
+	for _, tt := range tests {
+		got := itoa.Itoa(tt)
+		want := fmt.Sprint(tt)
+		if want != got {
+			t.Fatalf("Itoa(%d) = %s, want %s", tt, got, want)
+		}
+	}
+}
+
+func TestUitoa(t *testing.T) {
+	tests := []uint{0, 1, 100, 999, math.MaxUint32, uint(maxUint64)}
+	for _, tt := range tests {
+		got := itoa.Uitoa(tt)
+		want := fmt.Sprint(tt)
+		if want != got {
+			t.Fatalf("Uitoa(%d) = %s, want %s", tt, got, want)
+		}
+	}
+}
@@ -0,0 +1,127 @@
+//go:build scheduler.asyncify
+// +build scheduler.asyncify
+
+package task
+
+import (
+	"unsafe"
+)
+
+// Stack canary, to detect a stack overflow. The number is a random number
+// generated by random.org. The bit fiddling dance is necessary because
+// otherwise Go wouldn't allow the cast to a smaller integer size.
+const stackCanary = uintptr(uint64(0x670c1333b83bf575) & uint64(^uintptr(0)))
+
+//go:linkname runtimePanic runtime.runtimePanic
+func runtimePanic(str string)
+
+// state is a structure which holds a reference to the state of the task.
+// When the task is suspended, the stack pointers are saved here.
+type state struct {
+	// entry is the entry function of the task.
+	// This is needed every time the function is invoked so that asyncify knows what to rewind.
+	entry uintptr
+
+	// args are a pointer to a struct holding the arguments of the function.
+	args unsafe.Pointer
+
+	// stackState is the state of the stack while unwound.
+	stackState
+
+	launched bool
+}
+
+// stackState is the saved state of a stack while unwound.
+// The stack is arranged with asyncify at the bottom, C stack at the top, and a gap of available stack space between the two.
+type stackState struct {
+	// asyncify is the stack pointer of the asyncify stack.
+	// This starts from the bottom and grows upwards.
+	asyncifysp uintptr
+
+	// asyncify is stack pointer of the C stack.
+	// This starts from the top and grows downwards.
+	csp uintptr
+}
+
+// start creates and starts a new goroutine with the given function and arguments.
+// The new goroutine is immediately started.
+func start(fn uintptr, args unsafe.Pointer, stackSize uintptr) {
+	t := &Task{}
+	t.state.initialize(fn, args, stackSize)
+	runqueuePushBack(t)
+}
+
+//export tinygo_launch
+func (*state) launch()
+
+//go:linkname align runtime.align
+func align(p uintptr) uintptr
+
+// initialize the state and prepare to call the specified function with the specified argument bundle.
+func (s *state) initialize(fn uintptr, args unsafe.Pointer, stackSize uintptr) {
+	// Save the entry call.
+	s.entry = fn
+	s.args = args
+
+	// Create a stack.
+	stack := make([]uintptr, stackSize/unsafe.Sizeof(uintptr(0)))
+
+	// Calculate stack base addresses.
+	s.asyncifysp = uintptr(unsafe.Pointer(&stack[0]))
+	s.csp = uintptr(unsafe.Pointer(&stack[0])) + uintptr(len(stack))*unsafe.Sizeof(uintptr(0))
+	stack[0] = stackCanary
+}
+
+//go:linkname runqueuePushBack runtime.runqueuePushBack
+func runqueuePushBack(*Task)
+
+// currentTask is the current running task, or nil if currently in the scheduler.
+var currentTask *Task
+
+// Current returns the current active task.
+func Current() *Task {
+	return currentTask
+}
+
+// Pause suspends the current task and returns to the scheduler.
+// This function may only be called when running on a goroutine stack, not when running on the system stack.
+func Pause() {
+	// This is mildly unsafe but this is also the only place we can do this.
+	if *(*uintptr)(unsafe.Pointer(currentTask.state.asyncifysp)) != stackCanary {
+		runtimePanic("stack overflow")
+	}
+
+	currentTask.state.unwind()
+
+	*(*uintptr)(unsafe.Pointer(currentTask.state.asyncifysp)) = stackCanary
+}
+
+//export tinygo_unwind
+func (*stackState) unwind()
+
+// Resume the task until it pauses or completes.
+// This may only be called from the scheduler.
+func (t *Task) Resume() {
+	// The current task must be saved and restored because this can nest on WASM with JS.
+	prevTask := currentTask
+	currentTask = t
+	if !t.state.launched {
+		t.state.launch()
+		t.state.launched = true
+	} else {
+		t.state.rewind()
+	}
+	currentTask = prevTask
+	if t.state.asyncifysp > t.state.csp {
+		runtimePanic("stack overflow")
+	}
+}
+
+//export tinygo_rewind
+func (*state) rewind()
+
+// OnSystemStack returns whether the caller is running on the system stack.
+func OnSystemStack() bool {
+	// If there is not an active goroutine, then this must be running on the system stack.
+	return Current() == nil
+}
@@ -0,0 +1,102 @@
+.globaltype __stack_pointer, i32
+
+.global  tinygo_unwind
+.hidden  tinygo_unwind
+.type    tinygo_unwind,@function
+tinygo_unwind: // func (state *stackState) unwind()
+    .functype tinygo_unwind (i32) -> ()
+    // Check if we are rewinding.
+    i32.const 0
+    i32.load8_u tinygo_rewinding
+    if // if tinygo_rewinding {
+    // Stop rewinding.
+    call stop_rewind
+    i32.const 0
+    i32.const 0
+    i32.store8 tinygo_rewinding // tinygo_rewinding = false;
+    else
+    // Save the C stack pointer (destination structure pointer is in local 0).
+    local.get 0
+    global.get __stack_pointer
+    i32.store 4 // state.csp = getCurrentStackPointer()
+    // Ask asyncify to unwind.
+    // When resuming, asyncify will return this function with tinygo_rewinding set to true.
+    local.get 0
+    call start_unwind // asyncify.start_unwind(state)
+    end_if
+    return
+    end_function
+
+.global tinygo_launch
+.hidden tinygo_launch
+.type tinygo_launch,@function
+tinygo_launch: // func (state *state) launch()
+    .functype tinygo_launch (i32) -> ()
+    // Switch to the goroutine's C stack.
+    global.get __stack_pointer // prev := getCurrentStackPointer()
+    local.get 0
+    i32.load 12
+    global.set __stack_pointer // setStackPointer(state.csp)
+    // Get the argument pack and entry pointer.
+    local.get 0
+    i32.load 4 // args := state.args
+    local.get 0
+    i32.load 0 // fn := state.entry
+    // Launch the entry function.
+    call_indirect (i32) -> () // fn(args)
+    // Stop unwinding.
+    call stop_unwind
+    // Restore the C stack.
+    global.set __stack_pointer // setStackPointer(prev)
+    return
+    end_function
+
+.global  tinygo_rewind
+.hidden  tinygo_rewind
+.type    tinygo_rewind,@function
+tinygo_rewind: // func (state *state) rewind()
+    .functype tinygo_rewind (i32) -> ()
+    // Switch to the goroutine's C stack.
+    global.get __stack_pointer // prev := getCurrentStackPointer()
+    local.get 0
+    i32.load 12
+    global.set __stack_pointer // setStackPointer(state.csp)
+    // Get the argument pack and entry pointer.
+    local.get 0
+    i32.load 4 // args := state.args
+    local.get 0
+    i32.load 0 // fn := state.entry
+    // Prepare to rewind.
+    i32.const 0
+    i32.const 1
+    i32.store8 tinygo_rewinding // tinygo_rewinding = true;
+    local.get 0
+    i32.const 8
+    i32.add
+    call start_rewind // asyncify.start_rewind(&state.stackState)
+    // Launch the entry function.
+    // This will actually rewind the call stack.
+    call_indirect (i32) -> () // fn(args)
+    // Stop unwinding.
+    call stop_unwind
+    // Restore the C stack.
+    global.set __stack_pointer // setStackPointer(prev)
+    return
+    end_function
+
+.functype start_unwind (i32) -> ()
+.import_module start_unwind, asyncify
+.functype stop_unwind () -> ()
+.import_module stop_unwind, asyncify
+.functype start_rewind (i32) -> ()
+.import_module start_rewind, asyncify
+.functype stop_rewind () -> ()
+.import_module stop_rewind, asyncify
+
+        .hidden tinygo_rewinding                # @tinygo_rewinding
+        .type   tinygo_rewinding,@object
+        .section        .bss.tinygo_rewinding,"",@
+        .globl  tinygo_rewinding
+tinygo_rewinding:
+        .int8   0                               # 0x0
+        .size   tinygo_rewinding, 1
@@ -1,4 +1,4 @@
-// +build scheduler.tasks,amd64
+// +build scheduler.tasks,amd64,!windows

 package task

@@ -0,0 +1,57 @@
+// Windows on amd64 has a slightly different ABI than other (*nix) systems.
+// Therefore, assembly functions need to be tweaked slightly.
+
+.section .text.tinygo_startTask,"ax"
+.global  tinygo_startTask
+tinygo_startTask:
+    // Small assembly stub for starting a goroutine. This is already run on the
+    // new stack, with the callee-saved registers already loaded.
+    // Most importantly, r12 contain the pc of the to-be-started function and
+    // r13 contain the only argument it is given. Multiple arguments are packed
+    // into one by storing them in a new allocation.
+
+    // Set the first argument of the goroutine start wrapper, which contains all
+    // the arguments.
+    movq %r13, %rcx
+
+    // Branch to the "goroutine start" function.
+    callq *%r12
+
+    // After return, exit this goroutine. This is a tail call.
+    jmp tinygo_pause
+
+.global tinygo_swapTask
+.section .text.tinygo_swapTask,"ax"
+tinygo_swapTask:
+    // This function gets the following parameters:
+    // %rcx = newStack uintptr
+    // %rdx = oldStack *uintptr
+
+    // Save all callee-saved registers:
+    pushq %r15
+    pushq %r14
+    pushq %r13
+    pushq %r12
+    pushq %rsi
+    pushq %rdi
+    pushq %rbp
+    pushq %rbx
+
+    // Save the current stack pointer in oldStack.
+    movq %rsp, (%rdx)
+
+    // Switch to the new stack pointer.
+    movq %rcx, %rsp
+
+    // Load saved register from the new stack.
+    popq %rbx
+    popq %rbp
+    popq %rdi
+    popq %rsi
+    popq %r12
+    popq %r13
+    popq %r14
+    popq %r15
+
+    // Return into the new task, as if tinygo_swapTask was a regular call.
+    ret
@@ -0,0 +1,66 @@
+// +build scheduler.tasks,amd64,windows
+
+package task
+
+// This is almost the same as task_stack_amd64.go, but with the extra rdi and
+// rsi registers saved: Windows has a slightly different calling convention.
+
+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_amd64_windows.S that relies on
+// the exact layout of this struct.
+type calleeSavedRegs struct {
+	rbx uintptr
+	rbp uintptr
+	rdi uintptr
+	rsi uintptr
+	r12 uintptr
+	r13 uintptr
+	r14 uintptr
+	r15 uintptr
+
+	pc 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_amd64_windows.S). This assembly code calls a
+	// function (passed in r12) with a single argument (passed in r13). After
+	// the function returns, it calls Pause().
+	r.pc = uintptr(unsafe.Pointer(&startTask))
+
+	// Pass the function to call in r12.
+	// 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.r12 = fn
+
+	// Pass the pointer to the arguments struct in r13.
+	r.r13 = 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
+}
@@ -1,53 +0,0 @@
-// +build sam,atsamd51,matrixportal_m4
-
-package machine
-
-import (
-	"device/sam"
-	"runtime/interrupt"
-)
-
-// UART on the MatrixPortal M4
-var (
-	Serial = UART1
-	UART1  = &_UART1
-	_UART1 = UART{
-		Buffer: NewRingBuffer(),
-		Bus:    sam.SERCOM1_USART_INT,
-		SERCOM: 1,
-	}
-
-	UART2  = &_UART2
-	_UART2 = UART{
-		Buffer: NewRingBuffer(),
-		Bus:    sam.SERCOM4_USART_INT,
-		SERCOM: 4,
-	}
-)
-
-func init() {
-	UART1.Interrupt = interrupt.New(sam.IRQ_SERCOM1_1, _UART1.handleInterrupt)
-	UART2.Interrupt = interrupt.New(sam.IRQ_SERCOM4_1, _UART2.handleInterrupt)
-}
-
-// I2C on the MatrixPortal M4
-var (
-	I2C0 = &I2C{
-		Bus:    sam.SERCOM5_I2CM,
-		SERCOM: 5,
-	}
-)
-
-// SPI on the MatrixPortal M4
-var (
-	SPI0 = SPI{
-		Bus:    sam.SERCOM3_SPIM,
-		SERCOM: 3, // BUG: SDO on SERCOM1!
-	}
-	NINA_SPI = SPI0
-
-	SPI1 = SPI{
-		Bus:    sam.SERCOM0_SPIM,
-		SERCOM: 0,
-	}
-)
@@ -1,9 +1,14 @@
+//go:build nano_33_ble
 // +build nano_33_ble

 // This contains the pin mappings for the Arduino Nano 33 BLE [Sense] boards.
+// - https://store.arduino.cc/arduino-nano-33-ble
+// - https://store.arduino.cc/arduino-nano-33-ble-sense
 //
-// Flashing the board requires special version of bossac.
+// ----------------------------------------------------------------------------
+// Flashing
 //
+// Special version of bossac is required.
 // This executable can be obtained two ways:
 // 1) In Arduino IDE, install support for the board ("Arduino Mbed OS Nano Boards")
 //    Search for "tools/bossac/1.9.1-arduino2/bossac" in Arduino IDEs directory
@@ -16,9 +21,14 @@
 // It is possible to replace original bossac with this new one (this only adds support for nrf chip).
 // In that case make "bossac_arduino2" symlink on it, for the board target to be able to find it.
 //
-// For more information, see:
-// - https://store.arduino.cc/arduino-nano-33-ble
-// - https://store.arduino.cc/arduino-nano-33-ble-sense
+// ----------------------------------------------------------------------------
+// Bluetooth
+//
+// SoftDevice (s140v7) must be flashed first to enable use of bluetooth on this board.
+// See https://github.com/tinygo-org/bluetooth
+//
+// SoftDevice overwrites original bootloader and flashing method described above is not avalable anymore.
+// Instead, please use debug probe and flash your code with "nano-33-ble-s140v7" target.
 //
 package machine

@@ -62,6 +72,7 @@ const (
 	LED_RED     = P0_24
 	LED_GREEN   = P0_16
 	LED_BLUE    = P0_06
+	LED_PWR     = P1_09
 )

 // UART0 pins
@@ -72,8 +83,19 @@ const (

 // I2C pins
 const (
-	SDA_PIN = P0_31
-	SCL_PIN = P0_02
+	// Defaults to internal
+	SDA_PIN = SDA1_PIN
+	SCL_PIN = SCL1_PIN
+
+	// I2C0 (external) pins
+	SDA0_PIN = P0_31
+	SCL0_PIN = P0_02
+
+	// I2C1 (internal) pins
+	SDA1_PIN = P0_14
+	SCL1_PIN = P0_15
+
+	I2C_PULLUP = P1_00 // Set high for I2C to work
 )

 // SPI pins
@@ -83,6 +105,19 @@ const (
 	SPI0_SDI_PIN = P1_08
 )

+// Peripherals
+const (
+	APDS_INT = P0_19 // Proximity (APDS9960) interrupt pin
+
+	LSM_PWR = P0_22 // IMU (LSM9DS1) power
+	LSP_PWR = P0_22 // Pressure (LSP22) power
+	HTS_PWR = P0_22 // Humidity (HTS221) power
+
+	MIC_PWR = P0_17 // Microphone (MP34DT06JTR) power
+	MIC_CLK = P0_26
+	MIC_DIN = P0_25
+)
+
 // USB CDC identifiers
 const (
 	usb_STRING_PRODUCT      = "Nano 33 BLE"
@@ -4,7 +4,6 @@ package machine

 import (
 	"device/nxp"
-	"machine/usb"
 	"runtime/interrupt"
 )

@@ -56,21 +55,21 @@ const (

 // Analog pins
 const (
-	//  = Pin  // Dig [Pad]      {ADC1/ADC2}
-	A0  = PA18 // D14 [AD_B1_02] {  7 / 7  }
-	A1  = PA19 // D15 [AD_B1_03] {  8 / 8  }
-	A2  = PA23 // D16 [AD_B1_07] { 12 / 12 }
-	A3  = PA22 // D17 [AD_B1_06] { 11 / 11 }
-	A4  = PA17 // D18 [AD_B1_01] {  6 / 6  }
-	A5  = PA16 // D19 [AD_B1_00] {  5 / 5  }
-	A6  = PA26 // D20 [AD_B1_10] { 15 / 15 }
-	A7  = PA27 // D21 [AD_B1_11] {  0 / 0  }
-	A8  = PA24 // D22 [AD_B1_08] { 13 / 13 }
-	A9  = PA25 // D23 [AD_B1_09] { 14 / 14 }
-	A10 = PA12 // D24 [AD_B0_12] {  1 / -  }
-	A11 = PA13 // D25 [AD_B0_13] {  2 / -  }
-	A12 = PA30 // D26 [AD_B1_14] {  - / 3  }
-	A13 = PA31 // D27 [AD_B1_15] {  - / 4  }
+	//  = Pin  // Dig | [Pad]      {ADC1/ADC2}
+	A0  = PA18 // D14 | [AD_B1_02] {  7 / 7  }
+	A1  = PA19 // D15 | [AD_B1_03] {  8 / 8  }
+	A2  = PA23 // D16 | [AD_B1_07] { 12 / 12 }
+	A3  = PA22 // D17 | [AD_B1_06] { 11 / 11 }
+	A4  = PA17 // D18 | [AD_B1_01] {  6 / 6  }
+	A5  = PA16 // D19 | [AD_B1_00] {  5 / 5  }
+	A6  = PA26 // D20 | [AD_B1_10] { 15 / 15 }
+	A7  = PA27 // D21 | [AD_B1_11] {  0 / 0  }
+	A8  = PA24 // D22 | [AD_B1_08] { 13 / 13 }
+	A9  = PA25 // D23 | [AD_B1_09] { 14 / 14 }
+	A10 = PA12 // D24 | [AD_B0_12] {  1 / -  }
+	A11 = PA13 // D25 | [AD_B0_13] {  2 / -  }
+	A12 = PA30 // D26 | [AD_B1_14] {  - / 3  }
+	A13 = PA31 // D27 | [AD_B1_15] {  - / 4  }
 )

 // Default peripheral pins
@@ -105,15 +104,6 @@ func init() {
 	_UART7.Interrupt = interrupt.New(nxp.IRQ_LPUART7, _UART7.handleInterrupt)
 }

-// #=====================================================#
-// |                        USB                          |
-// #=====================================================#
-var (
-	// UART0 = usb.UART{Port: 0}
-	HID0  = usb.HID{Port: 0}
-	UART0 = &UART1
-)
-
 // #=====================================================#
 // |                        UART                         |
 // #===========#===========#=============#===============#
@@ -11,7 +11,6 @@ import (

 // Peripheral abstraction layer for the MIMXRT1062

-//go:inline
 func CPUFrequency() uint32 {
 	return 600000000
 }
@@ -1,3 +1,4 @@
+//go:build rp2040
 // +build rp2040

 package machine
@@ -114,3 +115,11 @@ func init() {
 	UART0.Interrupt = interrupt.New(rp.IRQ_UART0_IRQ, _UART0.handleInterrupt)
 	UART1.Interrupt = interrupt.New(rp.IRQ_UART1_IRQ, _UART1.handleInterrupt)
 }
+
+// CurrentCore returns the core number the call was made from.
+func CurrentCore() int {
+	return int(rp.SIO.CPUID.Get())
+}
+
+// NumCores returns number of cores available on the device.
+func NumCores() int { return 2 }
@@ -1,9 +1,11 @@
+//go:build rp2040
 // +build rp2040

 package machine

 import (
 	"device/rp"
+	"runtime/interrupt"
 	"runtime/volatile"
 	"unsafe"
 )
@@ -208,3 +210,108 @@ func (p Pin) Set(value bool) {
 func (p Pin) Get() bool {
 	return p.get()
 }
+
+// PinChange represents one or more trigger events that can happen on a given GPIO pin
+// on the RP2040. ORed PinChanges are valid input to most IRQ functions.
+type PinChange uint8
+
+// Pin change interrupt constants for SetInterrupt.
+const (
+	// PinLevelLow triggers whenever pin is at a low (around 0V) logic level.
+	PinLevelLow PinChange = 1 << iota
+	// PinLevelLow triggers whenever pin is at a high (around 3V) logic level.
+	PinLevelHigh
+	// Edge falling
+	PinFalling
+	// Edge rising
+	PinRising
+)
+
+// Callbacks to be called for pins configured with SetInterrupt.
+var (
+	pinCallbacks [2]func(Pin)
+	setInt       [2]bool
+)
+
+// SetInterrupt sets an interrupt to be executed when a particular pin changes
+// state. The pin should already be configured as an input, including a pull up
+// or down if no external pull is provided.
+//
+// This call will replace a previously set callback on this pin. You can pass a
+// nil func to unset the pin change interrupt. If you do so, the change
+// parameter is ignored and can be set to any value (such as 0).
+func (p Pin) SetInterrupt(change PinChange, callback func(Pin)) error {
+	if p > 31 || p < 0 {
+		return ErrInvalidInputPin
+	}
+	core := CurrentCore()
+	if callback == nil {
+		// disable current interrupt
+		p.setInterrupt(change, false)
+		pinCallbacks[core] = nil
+		return nil
+	}
+
+	if pinCallbacks[core] != nil {
+		// Callback already configured. Should disable callback by passing a nil callback first.
+		return ErrNoPinChangeChannel
+	}
+	p.setInterrupt(change, true)
+	pinCallbacks[core] = callback
+
+	if setInt[core] {
+		// interrupt has already been set. Exit.
+		println("core set")
+		return nil
+	}
+	interrupt.New(rp.IRQ_IO_IRQ_BANK0, gpioHandleInterrupt).Enable()
+	irqSet(rp.IRQ_IO_IRQ_BANK0, true)
+	return nil
+}
+
+// gpioHandleInterrupt finds the corresponding pin for the interrupt.
+// C SDK equivalent of gpio_irq_handler
+func gpioHandleInterrupt(intr interrupt.Interrupt) {
+	// panic("END") // if program is not ended here rp2040 will call interrupt again when finished, a vicious spin cycle.
+	core := CurrentCore()
+	callback := pinCallbacks[core]
+	if callback != nil {
+		// TODO fix gpio acquisition (see below)
+		// For now all callbacks get pin 255 (nonexistent).
+		callback(0xff)
+	}
+	var gpio Pin
+	for gpio = 0; gpio < _NUMBANK0_GPIOS; gpio++ {
+		// Acknowledge all GPIO interrupts for now
+		// since we are yet unable to acquire interrupt status
+		gpio.acknowledgeInterrupt(0xff) // TODO fix status get. For now we acknowledge all pending interrupts.
+		// Commented code below from C SDK not working.
+		// statreg := base.intS[gpio>>3]
+		// change := getIntChange(gpio, statreg.Get())
+		// if change != 0 {
+		// 	gpio.acknowledgeInterrupt(change)
+		// 	if callback != nil {
+		// 		callback(gpio)
+		// 		return
+		// 	} else {
+		// 		panic("unset callback in handler")
+		// 	}
+		// }
+	}
+}
+
+// events returns the bit representation of the pin change for the rp2040.
+func (change PinChange) events() uint32 {
+	return uint32(change)
+}
+
+// intBit is the bit storage form of a PinChange for a given Pin
+// in the IO_BANK0 interrupt registers (page 269 RP2040 Datasheet).
+func (p Pin) ioIntBit(change PinChange) uint32 {
+	return change.events() << (4 * (p % 8))
+}
+
+// Acquire interrupt data from a INT status register.
+func getIntChange(p Pin, status uint32) PinChange {
+	return PinChange(status>>(4*(p%8))) & 0xf
+}
@@ -0,0 +1,73 @@
+//go:build rp2040
+// +build rp2040
+
+package machine
+
+import (
+	"device/rp"
+)
+
+// machine_rp2040_sync.go contains interrupt and
+// lock primitives similar to those found in Pico SDK's
+// irq.c
+
+const (
+	// Number of spin locks available
+	_NUMSPINLOCKS = 32
+	// Number of interrupt handlers available
+	_NUMIRQ               = 32
+	_PICO_SPINLOCK_ID_IRQ = 9
+	_NUMBANK0_GPIOS       = 30
+)
+
+// Clears interrupt flag on a pin
+func (p Pin) acknowledgeInterrupt(change PinChange) {
+	ioBank0.intR[p>>3].Set(p.ioIntBit(change))
+}
+
+// Basic interrupt setting via ioBANK0 for GPIO interrupts.
+func (p Pin) setInterrupt(change PinChange, enabled bool) {
+	// Separate mask/force/status per-core, so check which core called, and
+	// set the relevant IRQ controls.
+	switch CurrentCore() {
+	case 0:
+		p.ctrlSetInterrupt(change, enabled, &ioBank0.proc0IRQctrl)
+	case 1:
+		p.ctrlSetInterrupt(change, enabled, &ioBank0.proc1IRQctrl)
+	}
+}
+
+// ctrlSetInterrupt acknowledges any pending interrupt and enables or disables
+// the interrupt for a given IRQ control bank (IOBANK, DormantIRQ, QSPI).
+//
+// pico-sdk calls this the _gpio_set_irq_enabled, not to be confused with
+// gpio_set_irq_enabled (no leading underscore).
+func (p Pin) ctrlSetInterrupt(change PinChange, enabled bool, base *irqCtrl) {
+	p.acknowledgeInterrupt(change)
+	enReg := &base.intE[p>>3]
+	if enabled {
+		enReg.SetBits(p.ioIntBit(change))
+	} else {
+		enReg.ClearBits(p.ioIntBit(change))
+	}
+}
+
+// Enable or disable a specific interrupt on the executing core.
+// num is the interrupt number which must be in [0,31].
+func irqSet(num uint32, enabled bool) {
+	if num >= _NUMIRQ {
+		return
+	}
+	irqSetMask(1<<num, enabled)
+}
+
+func irqSetMask(mask uint32, enabled bool) {
+	if enabled {
+		// Clear pending before enable
+		// (if IRQ is actually asserted, it will immediately re-pend)
+		rp.PPB.NVIC_ICPR.Set(mask)
+		rp.PPB.NVIC_ISER.Set(mask)
+	} else {
+		rp.PPB.NVIC_ICER.Set(mask)
+	}
+}
@@ -1,935 +0,0 @@
-package usb
-
-// Implementation of 32-bit target-agnostic USB device controller driver (dcd).
-
-import "unsafe"
-
-func init() {
-	if unsafe.Sizeof(uintptr(0)) > 4 {
-		panic("USB device controller is only supported on 32-bit systems")
-	}
-}
-
-// dcdCount defines the number of USB cores to configure for device mode. It is
-// computed as the sum of all declared device configuration descriptors.
-const dcdCount = descCDCACMCount + descHIDCount
-
-// dcdInstance provides statically-allocated instances of each USB device
-// controller configured on this platform.
-var dcdInstance [dcdCount]dcd
-
-// dhwInstance provides statically-allocated instances of each USB hardware
-// abstraction for ports configured as device on this platform.
-var dhwInstance [dcdCount]dhw
-
-// dcd implements a generic USB device controller driver (dcd) for 32-bit ARM
-// targets.
-type dcd struct {
-	*dhw // USB hardware abstraction layer
-
-	core *core // Parent USB core this instance is attached to
-	port int   // USB port index
-	cc   class // USB device class
-	id   int   // USB device controller index
-}
-
-// initDCD initializes and assigns a free device controller instance to the
-// given USB port. Returns the initialized device controller or nil if no free
-// device controller instances remain.
-func initDCD(port int, class class) (*dcd, status) {
-	if 0 == dcdCount {
-		return nil, statusInvalid // Must have defined device controllers
-	}
-	switch class.id {
-	case classDeviceCDCACM:
-		if 0 == class.config || class.config > descCDCACMCount {
-			return nil, statusInvalid // Must have defined descriptors
-		}
-	default:
-	}
-	// Return the first instance whose assigned core is currently nil.
-	for i := range dcdInstance {
-		if nil == dcdInstance[i].core {
-			// Initialize device controller.
-			dcdInstance[i].dhw = allocDHW(port, i, &dcdInstance[i])
-			dcdInstance[i].core = &coreInstance[port]
-			dcdInstance[i].port = port
-			dcdInstance[i].cc = class
-			dcdInstance[i].id = i
-			return &dcdInstance[i], statusOK
-		}
-	}
-	return nil, statusBusy // No free device controller instances available.
-}
-
-// class returns the receiver's current device class configuration.
-func (d *dcd) class() class { return d.cc }
-
-// dcdSetupSize defines the size (bytes) of a USB standard setup packet.
-const dcdSetupSize = 8 // bytes
-
-// dcdSetup contains the USB standard setup packet used to configure a device.
-type dcdSetup struct {
-	bmRequestType uint8
-	bRequest      uint8
-	wValue        uint16
-	wIndex        uint16
-	wLength       uint16
-}
-
-// pack returns the receiver setup packet encoded as uint64.
-func (s dcdSetup) pack() uint64 {
-	return ((uint64(s.bmRequestType) & 0xFF) << 0) |
-		((uint64(s.bRequest) & 0xFF) << 8) |
-		((uint64(s.wValue) & 0xFFFF) << 16) |
-		((uint64(s.wIndex) & 0xFFFF) << 32) |
-		((uint64(s.wLength) & 0xFFFF) << 48)
-}
-
-// dcdEvent is used to describe virtual interrupts on the USB bus to a device
-// controller.
-//
-// Since the device controller software is intended for use with multiple TinyGo
-// targets, all of which may not have exactly the same USB bus interrupts, a
-// "virtual interrupt" is defined that is common to all targets. The target's
-// hardware implementation (type dhw) is responsible for translating real system
-// interrupts it receives into the appropriate virtual interrupt code, defined
-// below, and notifying the device controller via method (*dcd).event(dcdEvent).
-type dcdEvent struct {
-	id    uint8
-	setup dcdSetup
-	mask  uint32
-}
-
-// Enumerated constants for all possible USB device controller interrupt codes.
-const (
-	dcdEventInvalid          uint8 = iota // Invalid interrupt
-	dcdEventStatusReset                   // USB reset received
-	dcdEventStatusRun                     // USB controller entered run state
-	dcdEventStatusSuspend                 // USB suspend received
-	dcdEventStatusError                   // USB error condition detected on bus
-	dcdEventControlSetup                  // USB setup received
-	dcdEventPeripheralReady               // USB PHY powered and ready to _go_
-	dcdEventTransactComplete              // USB transaction complete
-	dcdEventTimer                         // USB (system) timer
-)
-
-func (d *dcd) event(ev dcdEvent) {
-
-	switch ev.id {
-	case dcdEventInvalid:
-	case dcdEventStatusReset:
-		d.endpointMask = 0
-
-	case dcdEventPeripheralReady:
-		// Configure and enable control endpoint 0
-		d.endpointEnable(0, true, 0)
-
-	case dcdEventStatusRun:
-	case dcdEventStatusSuspend:
-	case dcdEventStatusError:
-	case dcdEventControlSetup:
-		// On control endpoint 0 setup events, the ev.setup field will be defined
-		switch d.controlSetup(ev.setup) {
-		case dcdStageSetup:
-		case dcdStageData:
-		case dcdStageStatus:
-		case dcdStageStall:
-			d.controlStall()
-		}
-
-	case dcdEventTransactComplete:
-	case dcdEventTimer:
-	default:
-	}
-}
-
-// dcdStage represents the USB transaction stage of a control request.
-type dcdStage uint8
-
-// Enumerated constants for all possible USB transaction stages.
-const (
-	dcdStageSetup  dcdStage = iota // Indicates no stage transition required
-	dcdStageData                   // IN/OUT data transfer
-	dcdStageStatus                 // Setup request complete
-	dcdStageStall                  // Unhandled or invalid request
-)
-
-// controlSetup handles setup messages on control endpoint 0.
-func (d *dcd) controlSetup(sup dcdSetup) dcdStage {
-
-	// Reset endpoint 0 notify mask
-	d.controlMask = 0
-
-	// First, switch on the type of request (standard, class, or vendor)
-	switch sup.bmRequestType & descRequestTypeTypeMsk {
-
-	// === STANDARD REQUEST ===
-	case descRequestTypeTypeStandard:
-
-		// Switch on the recepient and direction of the request
-		switch sup.bmRequestType &
-			(descRequestTypeRecipientMsk | descRequestTypeDirMsk) {
-
-		// --- DEVICE Rx (OUT) ---
-		case descRequestTypeRecipientDevice | descRequestTypeDirOut:
-
-			// Identify which request was received
-			switch sup.bRequest {
-
-			// SET ADDRESS (0x05):
-			case descRequestStandardSetAddress:
-				d.setDeviceAddress(sup.wValue)
-				d.controlReceive(uintptr(0), 0, false)
-				return dcdStageSetup
-
-			// SET CONFIGURATION (0x09):
-			case descRequestStandardSetConfiguration:
-				d.cc.config = int(sup.wValue)
-				if 0 == d.cc.config || d.cc.config > dcdCount {
-					// Use default if invalid index received
-					d.cc.config = 1
-				}
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// CDC-ACM (single)
-				case classDeviceCDCACM:
-					d.uartConfigure()
-					d.controlReceive(uintptr(0), 0, false)
-
-				// HID
-				case classDeviceHID:
-					d.serialConfigure()
-					d.keyboardConfigure()
-					d.mouseConfigure()
-					d.joystickConfigure()
-					d.controlReceive(uintptr(0), 0, false)
-
-				default:
-					// Unhandled device class
-				}
-				return dcdStageSetup
-
-			default:
-				// Unhandled request
-			}
-
-		// --- DEVICE Tx (IN) ---
-		case descRequestTypeRecipientDevice | descRequestTypeDirIn:
-
-			// Identify which request was received
-			switch sup.bRequest {
-
-			// GET STATUS (0x00):
-			case descRequestStandardGetStatus:
-				d.controlReply[0] = 0
-				d.controlReply[1] = 0
-				d.controlTransmit(
-					uintptr(unsafe.Pointer(&d.controlReply[0])), 2, false)
-				return dcdStageSetup
-
-			// GET DESCRIPTOR (0x06):
-			case descRequestStandardGetDescriptor:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// CDC-ACM (single)
-				case classDeviceCDCACM:
-					d.controlDescriptorCDCACM(sup)
-					return dcdStageSetup
-
-				// HID
-				case classDeviceHID:
-					d.controlDescriptorHID(sup)
-					return dcdStageSetup
-
-				default:
-					// Unhandled device class
-				}
-
-			// GET CONFIGURATION (0x08):
-			case descRequestStandardGetConfiguration:
-				d.controlReply[0] = uint8(d.cc.config)
-				d.controlTransmit(
-					uintptr(unsafe.Pointer(&d.controlReply[0])), 1, false)
-				return dcdStageSetup
-
-			default:
-				// Unhandled request
-			}
-
-		// --- INTERFACE Tx (IN) ---
-		case descRequestTypeRecipientInterface | descRequestTypeDirIn:
-
-			// Identify which request was received
-			switch sup.bRequest {
-
-			// GET DESCRIPTOR (0x06):
-			case descRequestStandardGetDescriptor:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// CDC-ACM (single)
-				case classDeviceCDCACM:
-					d.controlDescriptorCDCACM(sup)
-					return dcdStageSetup
-
-				// HID
-				case classDeviceHID:
-					d.controlDescriptorHID(sup)
-					return dcdStageSetup
-
-				default:
-					// Unhandled device class
-				}
-
-				// GET DESCRIPTOR (0x06):
-			case descHIDRequestGetReport:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// HID
-				case classDeviceHID:
-					d.controlDescriptorHID(sup)
-					return dcdStageSetup
-
-				default:
-					// Unhandled device class
-				}
-
-			default:
-				// Unhandled request
-			}
-
-		// --- ENDPOINT Rx (OUT) ---
-		case descRequestTypeRecipientEndpoint | descRequestTypeDirOut:
-
-			// Identify which request was received
-			switch sup.bRequest {
-
-			// CLEAR FEATURE (0x01):
-			case descRequestStandardClearFeature:
-				d.endpointClearFeature(uint8(sup.wIndex))
-				d.controlReceive(uintptr(0), 0, false)
-				return dcdStageSetup
-
-			// SET FEATURE (0x03):
-			case descRequestStandardSetFeature:
-				d.endpointSetFeature(uint8(sup.wIndex))
-				d.controlReceive(uintptr(0), 0, false)
-				return dcdStageSetup
-
-			default:
-				// Unhandled request
-			}
-
-		// --- ENDPOINT Tx (IN) ---
-		case descRequestTypeRecipientEndpoint | descRequestTypeDirIn:
-
-			// Identify which request was received
-			switch sup.bRequest {
-
-			// GET STATUS (0x00):
-			case descRequestStandardGetStatus:
-				status := d.endpointStatus(uint8(sup.wIndex))
-				d.controlReply[0] = uint8(status)
-				d.controlReply[1] = uint8(status >> 8)
-				d.controlTransmit(
-					uintptr(unsafe.Pointer(&d.controlReply[0])), 2, false)
-				return dcdStageSetup
-
-			default:
-				// Unhandled request
-			}
-
-		default:
-			// Unhandled request recepient or direction
-		}
-
-	// === CLASS REQUEST ===
-	case descRequestTypeTypeClass:
-
-		// Switch on the recepient and direction of the request
-		switch sup.bmRequestType &
-			(descRequestTypeRecipientMsk | descRequestTypeDirMsk) {
-
-		// --- INTERFACE Rx (OUT) ---
-		case descRequestTypeRecipientInterface | descRequestTypeDirOut:
-
-			// Identify which request was received
-			switch sup.bRequest {
-
-			// CDC | SET LINE CODING (0x20):
-			case descCDCRequestSetLineCoding:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// CDC-ACM (single)
-				case classDeviceCDCACM:
-					// line coding must contain exactly 7 bytes
-					if descCDCACMCodingSize == sup.wLength {
-						d.setup = sup
-						d.controlReceive(
-							uintptr(unsafe.Pointer(&descCDCACM[d.cc.config-1].cx[0])),
-							descCDCACMCodingSize, true)
-						// CDC Line Coding packet receipt handling occurs in method
-						// controlComplete().
-						return dcdStageSetup
-					}
-
-				default:
-					// Unhandled device class
-				}
-
-			// CDC | SET CONTROL LINE STATE (0x22):
-			case descCDCRequestSetControlLineState:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// CDC-ACM (single)
-				case classDeviceCDCACM:
-
-					// Determine interface destination of the request
-					switch sup.wIndex {
-
-					// Control/status interface:
-					case descCDCACMInterfaceCtrl:
-						// DTR is bit 0 (mask 0x01), RTS is bit 1 (mask 0x02)
-						d.uartSetLineState(0 != sup.wValue&0x01, 0 != sup.wValue&0x02)
-						d.controlReceive(uintptr(0), 0, false)
-						return dcdStageSetup
-
-					default:
-						// Unhandled device interface
-					}
-
-				default:
-					// Unhandled device class
-				}
-
-			// CDC | SEND BREAK (0x23):
-			case descCDCRequestSendBreak:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// CDC-ACM (single)
-				case classDeviceCDCACM:
-					d.controlReceive(uintptr(0), 0, false)
-					return dcdStageSetup
-
-				default:
-					// Unhandled device class
-				}
-
-			// HID | SET REPORT (0x09)
-			case descHIDRequestSetReport:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// HID
-				case classDeviceHID:
-					if sup.wLength <= descHIDCxCount {
-						d.setup = sup
-						descHID[d.cc.config-1].cx[0] = 0xE9
-						d.controlReceive(
-							uintptr(unsafe.Pointer(&descHID[d.cc.config-1].cx[0])),
-							uint32(sup.wLength), true)
-						return dcdStageSetup
-					}
-
-				default:
-					// Unhandled device class
-				}
-
-			// HID | SET IDLE (0x0A)
-			case descHIDRequestSetIdle:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// HID
-				case classDeviceHID:
-					idleRate := sup.wValue >> 8
-					// TBD: do we need to handle this request? wIndex contains the target
-					// interface of the request.
-					_ = idleRate
-					d.controlReceive(uintptr(0), 0, false)
-					return dcdStageSetup
-
-				default:
-					// Unhandled device class
-				}
-
-			default:
-				// Unhandled request
-			}
-
-		// --- INTERFACE Tx (IN) ---
-		case descRequestTypeRecipientInterface | descRequestTypeDirIn:
-
-			// Identify which request was received
-			switch sup.bRequest {
-
-			// HID | GET REPORT (0x01)
-			case descHIDRequestGetReport:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// HID
-				case classDeviceHID:
-					reportType := uint8(sup.wValue >> 8)
-					reportID := uint8(sup.wValue)
-					// TBD: do we need to handle this request? wIndex contains the target
-					// interface of the request.
-					_, _ = reportType, reportID
-					d.controlReply[0] = 0
-					d.controlReply[1] = 0
-					d.controlTransmit(
-						uintptr(unsafe.Pointer(&d.controlReply[0])), 2, false)
-					return dcdStageSetup
-
-				default:
-					// Unhandled device class
-				}
-
-			default:
-				// Unhandled request
-			}
-
-		default:
-			// Unhandled request recepient or direction
-		}
-
-	case descRequestTypeTypeVendor:
-	default:
-		// Unhandled request type
-	}
-
-	// All successful requests return early. If we reach this point, the request
-	// was invalid or unhandled. Stall the endpoint.
-	return dcdStageStall
-}
-
-// controlComplete handles the setup completion of control endpoint 0.
-func (d *dcd) controlComplete(status uint32) {
-
-	// Reset endpoint 0 notify mask
-	d.controlMask = 0
-
-	// First, switch on the type of request (standard, class, or vendor)
-	switch d.setup.bmRequestType & descRequestTypeTypeMsk {
-
-	// === CLASS REQUEST ===
-	case descRequestTypeTypeClass:
-
-		// Switch on the recepient and direction of the request
-		switch d.setup.bmRequestType &
-			(descRequestTypeRecipientMsk | descRequestTypeDirMsk) {
-
-		// --- INTERFACE Rx (OUT) ---
-		case descRequestTypeRecipientInterface | descRequestTypeDirOut:
-
-			// Identify which request was received
-			switch d.setup.bRequest {
-
-			// CDC | SET LINE CODING (0x20):
-			case descCDCRequestSetLineCoding:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// CDC-ACM (single)
-				case classDeviceCDCACM:
-					acm := &descCDCACM[d.cc.config-1]
-
-					// Determine interface destination of the request
-					switch d.setup.wIndex {
-
-					// CDC-ACM Control Interface:
-					case descCDCACMInterfaceCtrl:
-						// Notify PHY to handle triggers like special baud rates, which
-						// signal to reboot into bootloader or begin receiving OTA updates
-						d.uartSetLineCoding(descCDCACMLineCoding{
-							baud:     packU32(acm.cx[:]),
-							stopBits: acm.cx[4],
-							parity:   acm.cx[5],
-							numBits:  acm.cx[6],
-						})
-
-					default:
-						// Unhandled device interface
-					}
-
-				default:
-					// Unhandled device class
-				}
-
-			// HID | SET REPORT (0x09)
-			case descHIDRequestSetReport:
-
-				// Respond based on our device class configuration
-				switch d.cc.id {
-
-				// HID
-				case classDeviceHID:
-					hid := &descHID[d.cc.config-1]
-
-					// Determine interface destination of the request
-					switch d.setup.wIndex {
-
-					// HID Keyboard Interface
-					case descHIDInterfaceKeyboard:
-
-						// Determine the type of descriptor being requested
-						switch d.setup.wValue >> 8 {
-
-						// Configuration descriptor
-						case descTypeConfigure:
-							if 1 == d.setup.wLength {
-								hid.keyboard.led = hid.cx[0]
-								d.controlTransmit(uintptr(0), 0, false)
-							}
-
-						default:
-							// Unhandled descriptor type
-						}
-
-					// HID Serial Interface
-					case descHIDInterfaceSerial:
-
-						// Determine the type of descriptor being requested
-						switch d.setup.wValue >> 8 {
-
-						// String descriptor
-						case descTypeString:
-							if d.setup.wLength >= 4 && 0x68C245A9 == packU32(hid.cx[0:4]) {
-								d.enableSOF(true, descHIDInterfaceCount)
-							}
-
-						default:
-							// Unhandled descriptor type
-						}
-
-					default:
-						// Unhandled device interface
-					}
-
-				default:
-					// Unhandled device class
-				}
-
-			default:
-				// Unhandled request
-			}
-
-		default:
-			// Unhandled recepient or direction
-		}
-
-	default:
-		// Unhandled request type
-	}
-}
-
-func (d *dcd) controlDescriptorCDCACM(sup dcdSetup) {
-
-	acm := &descCDCACM[d.cc.config-1]
-	dxn := uint8(0)
-
-	// Determine the type of descriptor being requested
-	switch sup.wValue >> 8 {
-
-	// Device descriptor
-	case descTypeDevice:
-		dxn = descLengthDevice
-		_ = copy(acm.dx[:], acm.device[:dxn])
-
-	// Configuration descriptor
-	case descTypeConfigure:
-		dxn = uint8(descCDCACMConfigSize)
-		_ = copy(acm.dx[:], acm.config[:dxn])
-
-	// String descriptor
-	case descTypeString:
-		if 0 == len(acm.locale) {
-			break // No string descriptors defined!
-		}
-		var sd []uint8
-		if 0 == uint8(sup.wValue) {
-
-			// setup.wIndex contains an arbitrary index referring to a collection of
-			// strings in some given language. This case (setup.wValue = [0x03]00)
-			// is a string request from the host to determine what that language is.
-			//
-			// In subsequent string requests, the host will populate setup.wIndex
-			// with the language code we return here in this string descriptor.
-			//
-			// This way all strings returned to the host are in the same language,
-			// whatever language that may be.
-			code := int(sup.wIndex)
-			if code >= len(acm.locale) {
-				code = 0
-			}
-			sd = acm.locale[code].descriptor[sup.wValue&0xFF][:]
-
-		} else {
-
-			// setup.wIndex now contains a language code, which we specified in a
-			// previous request (above: setup.wValue = [0x03]00). We need to locate
-			// the set of strings whose language matches the language code given in
-			// this new setup.wIndex.
-			for code := range acm.locale {
-				if sup.wIndex == acm.locale[code].language {
-					// Found language, check if string descriptor at given index exists
-					if int(sup.wValue&0xFF) < len(acm.locale[code].descriptor) {
-
-						// Found language with a string defined at the requested index.
-						//
-						// TODO: Add API methods to device controller that allows the user
-						//       to provide these strings at/before driver initialization.
-						//
-						// For now, we just always use the descCommon* strings.
-						var s string
-						switch uint8(sup.wValue) {
-						case 1:
-							s = descCommonManufacturer
-						case 2:
-							s = descCommonProduct + " CDC-ACM"
-						case 3:
-							s = descCommonSerialNumber
-						}
-
-						// Construct a string descriptor dynamically to be transmitted on
-						// the serial bus.
-						sd = acm.locale[code].descriptor[int(sup.wValue&0xFF)][:]
-						// String descriptor format is 2-byte header + 2-bytes per rune
-						sd[0] = uint8(2 + 2*len(s)) // header[0] = descriptor length
-						sd[1] = descTypeString      // header[1] = descriptor type
-						// Copy UTF-8 string into string descriptor as UTF-16
-						for n, c := range s {
-							if 2+2*n >= len(sd) {
-								break
-							}
-							sd[2+2*n] = uint8(c)
-							sd[3+2*n] = 0
-						}
-						break // end search for matching language code
-					}
-				}
-			}
-		}
-		// Copy string descriptor into descriptor transmit buffer
-		if nil != sd && len(sd) >= 0 {
-			dxn = sd[0]
-			_ = copy(acm.dx[:], sd[:dxn])
-		}
-
-	// Device qualification descriptor
-	case descTypeQualification:
-		dxn = descLengthQualification
-		_ = copy(acm.dx[:], acm.qualif[:dxn])
-
-	// Alternate configuration descriptor
-	case descTypeOtherSpeedConfiguration:
-		// TODO
-
-	default:
-		// Unhandled descriptor type
-	}
-
-	if dxn > 0 {
-		if dxn > uint8(sup.wLength) {
-			dxn = uint8(sup.wLength)
-		}
-		flushCache(
-			uintptr(unsafe.Pointer(&acm.dx[0])), uintptr(dxn))
-		d.controlTransmit(
-			uintptr(unsafe.Pointer(&acm.dx[0])), uint32(dxn), false)
-	}
-}
-
-func (d *dcd) controlDescriptorHID(sup dcdSetup) {
-
-	hid := &descHID[d.cc.config-1]
-	dxn := uint8(0)
-	pos := uint8(0)
-
-	// Determine the type of descriptor being requested
-	switch sup.wValue >> 8 {
-
-	// Device descriptor
-	case descTypeDevice:
-		dxn = descLengthDevice
-		_ = copy(hid.dx[:], hid.device[:dxn])
-
-	// Configuration descriptor
-	case descTypeConfigure:
-		dxn = uint8(descHIDConfigSize)
-		_ = copy(hid.dx[:], hid.config[:dxn])
-
-	// String descriptor
-	case descTypeString:
-		if 0 == len(hid.locale) {
-			break // No string descriptors defined!
-		}
-		var sd []uint8
-		if 0 == uint8(sup.wValue) {
-
-			// setup.wIndex contains an arbitrary index referring to a collection of
-			// strings in some given language. This case (setup.wValue = [0x03]00)
-			// is a string request from the host to determine what that language is.
-			//
-			// In subsequent string requests, the host will populate setup.wIndex
-			// with the language code we return here in this string descriptor.
-			//
-			// This way all strings returned to the host are in the same language,
-			// whatever language that may be.
-			code := int(sup.wIndex)
-			if code >= len(hid.locale) {
-				code = 0
-			}
-			sd = hid.locale[code].descriptor[sup.wValue&0xFF][:]
-
-		} else {
-
-			// setup.wIndex now contains a language code, which we specified in a
-			// previous request (above: setup.wValue = [0x03]00). We need to locate
-			// the set of strings whose language matches the language code given in
-			// this new setup.wIndex.
-			for code := range hid.locale {
-				if sup.wIndex == hid.locale[code].language {
-					// Found language, check if string descriptor at given index exists
-					if int(sup.wValue&0xFF) < len(hid.locale[code].descriptor) {
-
-						// Found language with a string defined at the requested index.
-						//
-						// TODO: Add API methods to device controller that allows the user
-						//       to provide these strings at/before driver initialization.
-						//
-						// For now, we just always use the descCommon* strings.
-						var s string
-						switch uint8(sup.wValue) {
-						case 1:
-							s = descCommonManufacturer
-						case 2:
-							s = descCommonProduct + " HID"
-						case 3:
-							s = descCommonSerialNumber
-						}
-
-						// Construct a string descriptor dynamically to be transmitted on
-						// the serial bus.
-						sd = hid.locale[code].descriptor[int(sup.wValue&0xFF)][:]
-						// String descriptor format is 2-byte header + 2-bytes per rune
-						sd[0] = uint8(2 + 2*len(s)) // header[0] = descriptor length
-						sd[1] = descTypeString      // header[1] = descriptor type
-						// Copy UTF-8 string into string descriptor as UTF-16
-						for n, c := range s {
-							if 2+2*n >= len(sd) {
-								break
-							}
-							sd[2+2*n] = uint8(c)
-							sd[3+2*n] = 0
-						}
-						break // end search for matching language code
-					}
-				}
-			}
-		}
-		// Copy string descriptor into descriptor transmit buffer
-		if nil != sd && len(sd) >= 0 {
-			dxn = sd[0]
-			_ = copy(hid.dx[:], sd[:dxn])
-		}
-
-	// Device qualification descriptor
-	case descTypeQualification:
-		dxn = descLengthQualification
-		_ = copy(hid.dx[:], hid.qualif[:dxn])
-
-	// Alternate configuration descriptor
-	case descTypeOtherSpeedConfiguration:
-		// TODO
-
-	// HID descriptor
-	case descTypeHID:
-
-		// Determine interface destination of the request
-		switch sup.wIndex {
-		case descHIDInterfaceKeyboard:
-			pos = descHIDConfigKeyboardPos
-
-		case descHIDInterfaceMouse:
-			pos = descHIDConfigMousePos
-
-		case descHIDInterfaceSerial:
-			pos = descHIDConfigSerialPos
-
-		case descHIDInterfaceJoystick:
-			pos = descHIDConfigJoystickPos
-
-		case descHIDInterfaceMediaKey:
-			pos = descHIDConfigMediaKeyPos
-
-		default:
-			// Unhandled HID interface
-		}
-
-		if 0 != pos {
-			dxn = descLengthInterface
-			_ = copy(hid.dx[:], hid.config[pos:pos+dxn])
-		}
-
-	// HID report descriptor
-	case descTypeHIDReport:
-
-		// Determine interface destination of the request
-		switch sup.wIndex {
-		case descHIDInterfaceKeyboard:
-			dxn = uint8(len(descHIDReportKeyboard))
-			_ = copy(hid.dx[:], descHIDReportKeyboard[:])
-
-		case descHIDInterfaceMouse:
-			dxn = uint8(len(descHIDReportMouse))
-			_ = copy(hid.dx[:], descHIDReportMouse[:])
-
-		case descHIDInterfaceSerial:
-			dxn = uint8(len(descHIDReportSerial))
-			_ = copy(hid.dx[:], descHIDReportSerial[:])
-
-		case descHIDInterfaceJoystick:
-			dxn = uint8(len(descHIDReportJoystick))
-			_ = copy(hid.dx[:], descHIDReportJoystick[:])
-
-		case descHIDInterfaceMediaKey:
-			dxn = uint8(len(descHIDReportMediaKey))
-			_ = copy(hid.dx[:], descHIDReportMediaKey[:])
-
-		default:
-			// Unhandled HID interface
-		}
-
-	default:
-		// Unhandled descriptor type
-	}
-
-	if dxn > 0 {
-		if dxn > uint8(sup.wLength) {
-			dxn = uint8(sup.wLength)
-		}
-		flushCache(
-			uintptr(unsafe.Pointer(&hid.dx[0])), uintptr(dxn))
-		d.controlTransmit(
-			uintptr(unsafe.Pointer(&hid.dx[0])), uint32(dxn), false)
-	}
-}
@@ -1,604 +0,0 @@
-// +build mimxrt1062
-
-package usb
-
-// descCPUFrequencyHz defines the target CPU frequency (Hz).
-const descCPUFrequencyHz = 600000000
-
-// descCDCACMCount defines the number of USB cores that may be configured as
-// CDC-ACM (single) devices.
-const descCDCACMCount = 0
-
-// descHIDCount defines the number of USB cores that may be configured as a
-// composite (keyboard + mouse + joystick) human interface device (HID).
-const descHIDCount = 1
-
-// General USB device identification constants.
-const (
-	descCommonVendorID  = 0x16C0
-	descCommonProductID = 0x0483
-	descCommonReleaseID = 0x0101 // BCD (1.1)
-
-	descCommonLanguage     = descLanguageEnglish
-	descCommonManufacturer = "TinyGo"
-	descCommonProduct      = "USB"
-	descCommonSerialNumber = "00000"
-)
-
-// Constants for USB CDC-ACM device classes.
-const (
-	descCDCACMMaxPower = 50 // 100 mA
-
-	// CDC-ACM Control Buffers
-
-	descCDCACMQHCount = 2 * (descCDCACMEndpointCount + 1)
-	descCDCACMCxCount = 8
-
-	// CDC-ACM Data Buffers
-
-	descCDCACMRDCount = 2 * descCDCACMEndpointCount
-	descCDCACMRxSize  = descCDCACMDataRxPacketSize
-	descCDCACMRxCount = descCDCACMRxSize * descCDCACMRDCount
-
-	descCDCACMTDCount = descCDCACMEndpointCount
-	descCDCACMTxSize  = 4 * descCDCACMDataTxPacketSize
-	descCDCACMTxCount = descCDCACMTxSize * descCDCACMTDCount
-
-	descCDCACMTxTimeoutMs = 120 // millisec
-	descCDCACMTxSyncUs    = 75  // microsec
-
-	// Default CDC-ACM Endpoint Configurations (High-Speed)
-
-	descCDCACMStatusInterval   = descCDCACMStatusHSInterval   // Status
-	descCDCACMStatusPacketSize = descCDCACMStatusHSPacketSize //
-
-	descCDCACMDataRxPacketSize = descCDCACMDataRxHSPacketSize // Data Rx
-
-	descCDCACMDataTxPacketSize = descCDCACMDataTxHSPacketSize // Data Tx
-
-	// CDC-ACM Endpoint Configurations for Full-Speed Device
-
-	descCDCACMStatusFSInterval   = 5  // Status
-	descCDCACMStatusFSPacketSize = 16 //  (full-speed)
-
-	descCDCACMDataRxFSPacketSize = 64 // Data Rx (full-speed)
-
-	descCDCACMDataTxFSPacketSize = 64 // Data Tx (full-speed)
-
-	// CDC-ACM Endpoint Configurations for High-Speed Device
-
-	descCDCACMStatusHSInterval   = 5  // Status
-	descCDCACMStatusHSPacketSize = 16 //  (high-speed)
-
-	descCDCACMDataRxHSPacketSize = 512 // Data Rx (high-speed)
-
-	descCDCACMDataTxHSPacketSize = 512 // Data Tx (high-speed)
-)
-
-// Constants for USB HID (keyboard, mouse, joystick) device classes.
-const (
-	descHIDMaxPower = 50 // 100 mA
-
-	// HID Control Buffers
-
-	descHIDQHCount = 2 * (descHIDEndpointCount + 1)
-	descHIDCxCount = 8
-
-	// HID Serial Buffers
-
-	descHIDSerialRDCount = 8
-	descHIDSerialRxSize  = descHIDSerialRxPacketSize
-	descHIDSerialRxCount = descHIDSerialRxSize * descHIDSerialRDCount
-
-	descHIDSerialTDCount = 12
-	descHIDSerialTxSize  = descHIDSerialTxPacketSize
-	descHIDSerialTxCount = descHIDSerialTxSize * descHIDSerialTDCount
-
-	descHIDSerialTxTimeoutMs = 50 // millisec
-	descHIDSerialTxSyncUs    = 75 // microsec
-
-	// HID Keyboard Buffers
-
-	descHIDKeyboardTDCount = 12
-	descHIDKeyboardTxSize  = 4 * descHIDKeyboardTxPacketSize
-	descHIDKeyboardTxCount = descHIDKeyboardTxSize * descHIDKeyboardTDCount
-
-	descHIDKeyboardTxTimeoutMs = 50 // millisec
-
-	// HID Mouse Buffers
-
-	descHIDMouseTDCount = 4
-	descHIDMouseTxSize  = 4 * descHIDMouseTxPacketSize
-	descHIDMouseTxCount = descHIDMouseTxSize * descHIDMouseTDCount
-
-	descHIDMouseTxTimeoutMs = 30 // millisec
-
-	// HID Joystick Buffers
-
-	descHIDJoystickTDCount = 4
-	descHIDJoystickTxSize  = 4 * descHIDJoystickTxPacketSize
-	descHIDJoystickTxCount = descHIDJoystickTxSize * descHIDJoystickTDCount
-
-	descHIDJoystickTxTimeoutMs = 30 // millisec
-
-	// Default HID Endpoint Configurations (High-Speed)
-
-	descHIDSerialRxInterval   = descHIDSerialRxHSInterval   // Serial Rx
-	descHIDSerialRxPacketSize = descHIDSerialRxHSPacketSize //
-
-	descHIDSerialTxInterval   = descHIDSerialTxHSInterval   // Serial Tx
-	descHIDSerialTxPacketSize = descHIDSerialTxHSPacketSize //
-
-	descHIDKeyboardTxInterval   = descHIDKeyboardTxHSInterval   // Keyboard
-	descHIDKeyboardTxPacketSize = descHIDKeyboardTxHSPacketSize //
-
-	descHIDMediaKeyTxInterval   = descHIDMediaKeyTxHSInterval   // Keyboard Media Keys
-	descHIDMediaKeyTxPacketSize = descHIDMediaKeyTxHSPacketSize //
-
-	descHIDMouseTxInterval   = descHIDMouseTxHSInterval   // Mouse
-	descHIDMouseTxPacketSize = descHIDMouseTxHSPacketSize //
-
-	descHIDJoystickTxInterval   = descHIDJoystickTxHSInterval   // Joystick
-	descHIDJoystickTxPacketSize = descHIDJoystickTxHSPacketSize //
-
-	// HID Endpoint Configurations for Full-Speed Device
-
-	descHIDSerialRxFSInterval   = 2 // Serial Rx
-	descHIDSerialRxFSPacketSize = 8 //  (full-speed)
-
-	descHIDSerialTxFSInterval   = 1  // Serial Tx
-	descHIDSerialTxFSPacketSize = 16 //  (full-speed)
-
-	descHIDKeyboardTxFSInterval   = 4 // Keyboard
-	descHIDKeyboardTxFSPacketSize = 8 //  (full-speed)
-
-	descHIDMediaKeyTxFSInterval   = 4 // Keyboard Media Keys
-	descHIDMediaKeyTxFSPacketSize = 8 //  (full-speed)
-
-	descHIDMouseTxFSInterval   = 4 // Mouse
-	descHIDMouseTxFSPacketSize = 8 //  (full-speed)
-
-	descHIDJoystickTxFSInterval   = 4  // Joystick
-	descHIDJoystickTxFSPacketSize = 12 //  (full-speed)
-
-	// HID Endpoint Configurations for High-Speed Device
-
-	descHIDSerialRxHSInterval   = 2  // Serial
-	descHIDSerialRxHSPacketSize = 32 //  (high-speed)
-
-	descHIDSerialTxHSInterval   = 1  // Serial Tx
-	descHIDSerialTxHSPacketSize = 64 //  (high-speed)
-
-	descHIDKeyboardTxHSInterval   = 1 // Keyboard
-	descHIDKeyboardTxHSPacketSize = 8 //  (high-speed)
-
-	descHIDMediaKeyTxHSInterval   = 4 // Keyboard Media Keys
-	descHIDMediaKeyTxHSPacketSize = 8 //  (high-speed)
-
-	descHIDMouseTxHSInterval   = 1 // Mouse
-	descHIDMouseTxHSPacketSize = 8 //  (high-speed)
-
-	descHIDJoystickTxHSInterval   = 2  // Joystick
-	descHIDJoystickTxHSPacketSize = 12 //  (high-speed)
-)
-
-// descCDCACM0QH is an array of endpoint queue heads, which is where all
-// transfers for a given endpoint are managed, for the default CDC-ACM (single)
-// device class configuration (index 1).
-//
-// From the iMXRT1062 Reference Manual:
-//
-//   Software must ensure that no interface data structure reachable
-//   by the Device Controller spans a 4K-page boundary.
-//
-//   The [queue head] is a 48-byte data structure, but must be aligned on
-//   64-byte boundaries.
-//
-//   Endpoint queue heads are arranged in an array in a continuous area of
-//   memory pointed to by the USB.ENDPOINTLISTADDR pointer. The even-numbered
-//   device queue heads in the list support receive endpoints (OUT/SETUP) and
-//   the odd-numbered queue heads in the list are used for transmit endpoints
-//   (IN/INTERRUPT). The device controller will index into this array based upon
-//   the endpoint number received from the USB bus. All information necessary to
-//   respond to transactions for all primed transfers is contained in this list
-//   so the Device Controller can readily respond to incoming requests without
-//   having to traverse a linked list.
-//go:align 4096
-var descCDCACM0QH [descCDCACMQHCount]dhwEndpoint
-
-// descCDCACM0CD is the transfer descriptor for messages transmitted or received
-// on the status/control endpoint 0 for the default CDC-ACM (single) device
-// class configuration (index 1).
-//go:align 32
-var descCDCACM0CD dhwTransfer
-
-// descCDCACM0Cx is the buffer for control/status data received on endpoint 0 of
-// the default CDC-ACM (single) device class configuration (index 1).
-//go:align 32
-var descCDCACM0Cx [descCDCACMCxCount]uint8
-
-// descCDCACM0AD is the transfer descriptor for ackowledgement (ACK) messages
-// transmitted or received on the status/control endpoint 0 for the default
-// CDC-ACM (single) device class configuration (index 1).
-//go:align 32
-var descCDCACM0AD dhwTransfer
-
-// descCDCACM0Dx is the transmit (Tx) buffer of descriptor data on endpoint 0
-// for the default CDC-ACM (single) device class configuration (index 1).
-//go:align 32
-var descCDCACM0Dx [descCDCACMConfigSize]uint8
-
-// descCDCACM0RD is an array of transfer descriptors for Rx (OUT) transfers,
-// which describe to the device controller the location and quantity of data
-// being received for a given transfer, for the default CDC-ACM (single) device
-// class configuration (index 1).
-//go:align 32
-var descCDCACM0RD [descCDCACMRDCount]dhwTransfer
-
-// descCDCACM0Rx is the receive (Rx) transfer buffer for the default CDC-ACM
-// (single) device class configuration (index 1).
-//go:align 32
-var descCDCACM0Rx [descCDCACMRxCount]uint8
-
-// descCDCACM0TD is an array of transfer descriptors for Tx (IN) transfers,
-// which describe to the device controller the location and quantity of data
-// being transmitted for a given transfer, for the default CDC-ACM (single)
-// device class configuration (index 1).
-//go:align 32
-var descCDCACM0TD [descCDCACMTDCount]dhwTransfer
-
-// descCDCACM0Tx is the transmit (Tx) transfer buffer for the default CDC-ACM
-// (single) device class configuration (index 1).
-//go:align 32
-var descCDCACM0Tx [descCDCACMTxCount]uint8
-
-var descCDCACM0RDNum [descCDCACMRDCount]uint16
-var descCDCACM0RDIdx [descCDCACMRDCount]uint16
-var descCDCACM0RDQue [(descCDCACMRDCount + 1)]uint16
-
-// descCDCACMClassData holds the buffers and control states for all CDC-ACM
-// (single) device class configurations, ordered by index (offset by -1), for
-// iMXRT1062 targets only.
-//
-// Instances of this type (elements of descCDCACMData) are embedded in elements
-// of the common/target-agnostic CDC-ACM class configurations (descCDCACM).
-// Methods defined on this type implement target-specific functionality, and
-// some of these methods are required by the common device controller driver.
-// Thus, this type functions as a hardware abstraction layer (HAL).
-type descCDCACMClassData struct {
-
-	// CDC-ACM Control Buffers
-
-	qh *[descCDCACMQHCount]dhwEndpoint // endpoint queue heads
-
-	cd *dhwTransfer                 // control endpoint 0 Rx/Tx transfer descriptor
-	cx *[descCDCACMCxCount]uint8    // control endpoint 0 Rx/Tx transfer buffer
-	ad *dhwTransfer                 // control endpoint 0 Rx/Tx ACK transfer descriptor
-	dx *[descCDCACMConfigSize]uint8 // control endpoint 0 Tx (IN) descriptor transfer buffer
-
-	// CDC-ACM Data Buffers
-
-	rd *[descCDCACMRDCount]dhwTransfer // bulk data endpoint Rx (OUT) transfer descriptors
-	rx *[descCDCACMRxCount]uint8       // bulk data endpoint Rx (OUT) transfer buffer
-	td *[descCDCACMTDCount]dhwTransfer // bulk data endpoint Tx (IN) transfer descriptors
-	tx *[descCDCACMTxCount]uint8       // bulk data endpoint Tx (IN) transfer buffer
-
-	rxCount *[descCDCACMRDCount]uint16
-	rxIndex *[descCDCACMRDCount]uint16
-	rxQueue *[(descCDCACMRDCount + 1)]uint16
-
-	sxSize uint16
-	rxSize uint16
-	txSize uint16
-
-	txHead uint8
-	txFree uint16
-	txPrev bool
-
-	rxHead uint8
-	rxTail uint8
-	rxFree uint16
-}
-
-// descCDCACMData holds statically-allocated instances for each of the target-
-// specific (iMXRT1062) CDC-ACM (single) device class configurations' control
-// and data structures, ordered by configuration index (offset by -1). Each
-// element is embedded in a corresponding element of descCDCACM.
-//go:align 64
-var descCDCACMData = [dcdCount]descCDCACMClassData{
-
-	{ // -- CDC-ACM (single) Class Configuration Index 1 --
-
-		// CDC-ACM Control Buffers
-
-		qh: &descCDCACM0QH,
-
-		cd: &descCDCACM0CD,
-		cx: &descCDCACM0Cx,
-		ad: &descCDCACM0AD,
-		dx: &descCDCACM0Dx,
-
-		// CDC-ACM Data Buffers
-
-		rd: &descCDCACM0RD,
-		rx: &descCDCACM0Rx,
-		td: &descCDCACM0TD,
-		tx: &descCDCACM0Tx,
-
-		rxCount: &descCDCACM0RDNum,
-		rxIndex: &descCDCACM0RDIdx,
-		rxQueue: &descCDCACM0RDQue,
-
-		sxSize: descCDCACMStatusPacketSize,
-		rxSize: descCDCACMDataRxPacketSize,
-		txSize: descCDCACMDataTxPacketSize,
-	},
-}
-
-// descHID0QH is an array of endpoint queue heads, which is where all transfers
-// for a given endpoint are managed, for the default HID device class
-// configuration (index 1).
-//
-// From the iMXRT1062 Reference Manual:
-//
-//   Software must ensure that no interface data structure reachable
-//   by the Device Controller spans a 4K-page boundary.
-//
-//   The [queue head] is a 48-byte data structure, but must be aligned on
-//   64-byte boundaries.
-//
-//   Endpoint queue heads are arranged in an array in a continuous area of
-//   memory pointed to by the USB.ENDPOINTLISTADDR pointer. The even-numbered
-//   device queue heads in the list support receive endpoints (OUT/SETUP) and
-//   the odd-numbered queue heads in the list are used for transmit endpoints
-//   (IN/INTERRUPT). The device controller will index into this array based upon
-//   the endpoint number received from the USB bus. All information necessary to
-//   respond to transactions for all primed transfers is contained in this list
-//   so the Device Controller can readily respond to incoming requests without
-//   having to traverse a linked list.
-//go:align 4096
-var descHID0QH [descHIDQHCount]dhwEndpoint
-
-// descHID0CD is the transfer descriptor for messages transmitted or received on
-// the status/control endpoint 0 for the default HID device class configuration
-// (index 1).
-//go:align 32
-var descHID0CD dhwTransfer
-
-// descHID0Cx is the buffer for control/status data received on endpoint 0 of
-// the default HID device class configuration (index 1).
-//go:align 32
-var descHID0Cx [descHIDCxCount]uint8
-
-// descHID0AD is the transfer descriptor for ackowledgement (ACK) messages
-// transmitted or received on the status/control endpoint 0 for the default HID
-// device class configuration (index 1).
-//go:align 32
-var descHID0AD dhwTransfer
-
-// descHID0Dx is the transmit (Tx) buffer of descriptor data on endpoint 0 for
-// the default HID device class configuration (index 1).
-//go:align 32
-var descHID0Dx [descHIDConfigSize]uint8
-
-// descHID0SerialRD is an array of transfer descriptors for serial Rx (OUT)
-// transfers, which describe to the device controller the location and quantity
-// of data being received for a given transfer, for the default HID device class
-// configuration (index 1).
-//go:align 32
-var descHID0SerialRD [descHIDSerialRDCount]dhwTransfer
-
-// descHID0SerialRx is the serial receive (Rx) transfer buffer for the default
-// HID device class configuration (index 1).
-//go:align 32
-var descHID0SerialRx [descHIDSerialRxCount]uint8
-
-// descHID0SerialTD is an array of transfer descriptors for serial Tx (IN)
-// transfers, which describe to the device controller the location and quantity
-// of data being transmitted for a given transfer, for the default HID device
-// class configuration (index 1).
-//go:align 32
-var descHID0SerialTD [descHIDSerialTDCount]dhwTransfer
-
-// descHID0SerialTx is the serial transmit (Tx) transfer buffer for the default
-// HID device class configuration (index 1).
-//go:align 32
-var descHID0SerialTx [descHIDSerialTxCount]uint8
-
-var descHID0SerialRDIdx [descHIDSerialRDCount]uint16
-var descHID0SerialRDQue [(descHIDSerialRDCount + 1)]uint16
-
-// descHID0KeyboardTD is an array of transfer descriptors for keyboard Tx (IN)
-// transfers, which describe to the device controller the location and quantity
-// of data being transmitted for a given transfer, for the default HID device
-// class configuration (index 1).
-//go:align 32
-var descHID0KeyboardTD [descHIDKeyboardTDCount]dhwTransfer
-
-// descHID0KeyboardTx is the keyboard transmit (Tx) transfer buffer for the
-// default HID device class configuration (index 1).
-//go:align 32
-var descHID0KeyboardTx [descHIDKeyboardTxCount]uint8
-
-// descHID0KeyboardTp is the keyboard HID report transmit (Tx) transfer buffer
-// for the default HID device class configuration (index 1).
-//go:align 32
-var descHID0KeyboardTp [descHIDKeyboardTxPacketSize]uint8
-
-//go:align 32
-var descHID0KeyboardTxKey [hidKeyboardKeyCount]uint8
-
-//go:align 32
-var descHID0KeyboardTxCon [hidKeyboardConCount]uint16
-
-//go:align 32
-var descHID0KeyboardTxSys [hidKeyboardSysCount]uint8
-
-// descHID0MouseTD is an array of transfer descriptors for mouse Tx (IN)
-// transfers, which describe to the device controller the location and quantity
-// of data being transmitted for a given transfer, for the default HID device
-// class configuration (index 1).
-//go:align 32
-var descHID0MouseTD [descHIDMouseTDCount]dhwTransfer
-
-// descHID0MouseTx is the mouse transmit (Tx) transfer buffer for the default
-// HID device class configuration (index 1).
-//go:align 32
-var descHID0MouseTx [descHIDMouseTxCount]uint8
-
-// descHID0JoystickTD is an array of transfer descriptors for joystick Tx (IN)
-// transfers, which describe to the device controller the location and quantity
-// of data being transmitted for a given transfer, for the default HID device
-// class configuration (index 1).
-//go:align 32
-var descHID0JoystickTD [descHIDJoystickTDCount]dhwTransfer
-
-// descHID0JoystickTx is the joystick transmit (Tx) transfer buffer for the
-// default HID device class configuration (index 1).
-//go:align 32
-var descHID0JoystickTx [descHIDJoystickTxCount]uint8
-
-// descHID0Keyboard is the Keyboard instance with which the user may interact
-// when using the default HID device class configuration (index 1).
-//go:align 64
-var descHID0Keyboard = Keyboard{
-	key: &descHID0KeyboardTxKey,
-	con: &descHID0KeyboardTxCon,
-	sys: &descHID0KeyboardTxSys,
-}
-
-// descHIDClassData holds the buffers and control states for all of the HID
-// device class configurations, ordered by index (offset by -1), for iMXRT1062
-// targets only.
-//
-// Instances of this type (elements of descHIDData) are embedded in elements
-// of the common/target-agnostic HID class configurations (descHID).
-// Methods defined on this type implement target-specific functionality, and
-// some of these methods are required by the common device controller driver.
-// Thus, this type functions as a hardware abstraction layer (HAL).
-type descHIDClassData struct {
-
-	// HID Control Buffers
-
-	qh *[descHIDQHCount]dhwEndpoint // endpoint queue heads
-
-	cd *dhwTransfer              // control endpoint 0 Rx/Tx transfer descriptor
-	cx *[descHIDCxCount]uint8    // control endpoint 0 Rx/Tx transfer buffer
-	ad *dhwTransfer              // control endpoint 0 Rx/Tx ACK transfer descriptor
-	dx *[descHIDConfigSize]uint8 // control endpoint 0 Tx (IN) descriptor transfer buffer
-
-	// HID Serial Buffers
-
-	rdSerial *[descHIDSerialRDCount]dhwTransfer // interrupt endpoint serial Rx (OUT) transfer descriptors
-	rxSerial *[descHIDSerialRxCount]uint8       // interrupt endpoint serial Rx (OUT) transfer buffer
-	tdSerial *[descHIDSerialTDCount]dhwTransfer // interrupt endpoint serial Tx (IN) transfer descriptors
-	txSerial *[descHIDSerialTxCount]uint8       // interrupt endpoint serial Tx (IN) transfer buffer
-
-	rxSerialIndex *[descHIDSerialRDCount]uint16
-	rxSerialQueue *[(descHIDSerialRDCount + 1)]uint16
-
-	rxSerialSize uint16
-	txSerialSize uint16
-
-	txSerialHead uint8
-	txSerialFree uint16
-	txSerialPrev bool
-
-	rxSerialHead uint8
-	rxSerialTail uint8
-	rxSerialFree uint16
-
-	// HID Keyboard Buffers
-
-	tdKeyboard *[descHIDKeyboardTDCount]dhwTransfer // interrupt endpoint keyboard Tx (IN) transfer descriptors
-	txKeyboard *[descHIDKeyboardTxCount]uint8       // interrupt endpoint keyboard Tx (IN) transfer buffer
-	tpKeyboard *[descHIDKeyboardTxPacketSize]uint8  // interrupt endpoint keyboard Tx (IN) HID report bbuffer
-
-	txKeyboardSize uint16
-
-	txKeyboardHead uint8
-	txKeyboardPrev bool
-
-	// HID Mouse Buffers
-
-	tdMouse *[descHIDMouseTDCount]dhwTransfer // interrupt endpoint mouse Tx (IN) transfer descriptors
-	txMouse *[descHIDMouseTxCount]uint8       // interrupt endpoint mouse Tx (IN) transfer buffer
-
-	txMouseSize uint16
-
-	txMouseHead uint8
-	txMousePrev bool
-
-	// HID Joystick Buffers
-
-	tdJoystick *[descHIDJoystickTDCount]dhwTransfer // interrupt endpoint joystick Tx (IN) transfer descriptors
-	txJoystick *[descHIDJoystickTxCount]uint8       // interrupt endpoint joystick Tx (IN) transfer buffer
-
-	txJoystickSize uint16
-
-	txJoystickHead uint8
-	txJoystickPrev bool
-
-	// HID Device Instances
-
-	keyboard *Keyboard
-}
-
-// descHIDData holds statically-allocated instances for each of the target-
-// specific (iMXRT1062) HID device class configurations' control and data
-// structures, ordered by configuration index (offset by -1). Each element is
-// embedded in a corresponding element of descHID.
-//go:align 64
-var descHIDData = [dcdCount]descHIDClassData{
-
-	{ // -- HID Class Configuration Index 1 --
-
-		// HID Control Buffers
-
-		qh: &descHID0QH,
-
-		cd: &descHID0CD,
-		cx: &descHID0Cx,
-		ad: &descHID0AD,
-		dx: &descHID0Dx,
-
-		// HID Serial Buffers
-
-		rdSerial: &descHID0SerialRD,
-		rxSerial: &descHID0SerialRx,
-		tdSerial: &descHID0SerialTD,
-		txSerial: &descHID0SerialTx,
-
-		rxSerialIndex: &descHID0SerialRDIdx,
-		rxSerialQueue: &descHID0SerialRDQue,
-
-		rxSerialSize: descHIDSerialRxPacketSize,
-		txSerialSize: descHIDSerialTxPacketSize,
-
-		// HID Keyboard Buffers
-
-		tdKeyboard: &descHID0KeyboardTD,
-		txKeyboard: &descHID0KeyboardTx,
-		tpKeyboard: &descHID0KeyboardTp,
-
-		txKeyboardSize: descHIDKeyboardTxPacketSize,
-
-		// HID Mouse Buffers
-
-		tdMouse: &descHID0MouseTD,
-		txMouse: &descHID0MouseTx,
-
-		txMouseSize: descHIDMouseTxPacketSize,
-
-		// HID Joystick Buffers
-
-		tdJoystick: &descHID0JoystickTD,
-		txJoystick: &descHID0JoystickTx,
-
-		txJoystickSize: descHIDJoystickTxPacketSize,
-
-		// HID Device Instances
-
-		keyboard: &descHID0Keyboard,
-	},
-}
@@ -1,53 +0,0 @@
-package usb
-
-// Implementation of 32-bit target-agnostic USB host controller driver (hcd).
-
-// hcdCount defines the number of USB cores to configure for host mode. It is
-// computed as the sum of all declared host configuration descriptors.
-const hcdCount = 0 // + ...
-
-// hcdInstance provides statically-allocated instances of each USB host
-// controller configured on this platform.
-var hcdInstance [hcdCount]hcd
-
-// hhwInstance provides statically-allocated instances of each USB hardware
-// abstraction for ports configured as host on this platform.
-var hhwInstance [hcdCount]hhw
-
-// hcd implements USB host controller driver (hcd) interface.
-type hcd struct {
-	*hhw // USB hardware abstraction layer
-
-	core *core // Parent USB core this instance is attached to
-	port int   // USB port index
-	cc   class // USB host class
-	id   int   // USB host controller index
-}
-
-// initHCD initializes and assigns a free host controller instance to the given
-// USB port. Returns the initialized host controller or nil if no free host
-// controller instances remain.
-func initHCD(port int, class class) (*hcd, status) {
-	if 0 == hcdCount {
-		return nil, statusInvalid // Must have defined host controllers
-	}
-	switch class.id {
-	default:
-	}
-	// Return the first instance whose assigned core is currently nil.
-	for i := range hcdInstance {
-		if nil == hcdInstance[i].core {
-			// Initialize host controller.
-			hcdInstance[i].hhw = allocHHW(port, i, &hcdInstance[i])
-			hcdInstance[i].core = &coreInstance[port]
-			hcdInstance[i].port = port
-			hcdInstance[i].cc = class
-			hcdInstance[i].id = i
-			return &hcdInstance[i], statusOK
-		}
-	}
-	return nil, statusBusy // No free host controller instances available.
-}
-
-// class returns the receiver's current host class configuration.
-func (h *hcd) class() class { return h.cc }
@@ -1,59 +0,0 @@
-// +build mimxrt1062
-
-package usb
-
-// Implementation of USB host controller driver (hcd) for NXP iMXRT1062.
-
-import (
-	"device/nxp"
-	"runtime/interrupt"
-)
-
-// hcdInterruptPriority defines the priority for all USB host interrupts.
-const hcdInterruptPriority = 3
-
-// hcd implements USB host controller driver (hcd) interface.
-type hhw struct {
-	*hcd // USB host controller driver
-
-	bus *nxp.USB_Type       // USB core register
-	phy *nxp.USBPHY_Type    // USB PHY register
-	irq interrupt.Interrupt // USB IRQ, only a single interrupt on iMXRT1062
-}
-
-// allocHHW returns a reference to the USB hardware abstraction for the given
-// host controller driver. Should be called only one time and during host
-// controller initialization.
-func allocHHW(port, instance int, hc *hcd) *hhw {
-	switch port {
-	case 0:
-		hhwInstance[instance].hcd = hc
-		hhwInstance[instance].bus = nxp.USB1
-		hhwInstance[instance].phy = nxp.USBPHY1
-
-	case 1:
-		hhwInstance[instance].hcd = hc
-		hhwInstance[instance].bus = nxp.USB2
-		hhwInstance[instance].phy = nxp.USBPHY2
-	}
-
-	return &hhwInstance[instance]
-}
-
-// init configures the USB port for host mode operation by initializing all
-// endpoint and transfer descriptor data structures, initializing core registers
-// and interrupts, resetting the USB PHY, and enabling power on the bust.
-func (h *hhw) init() status {
-
-	return statusOK
-}
-
-// enable causes the USB core to enter (or exit) the normal run state and
-// enables/disables all interrupts on the receiver's USB port.
-func (h *hhw) enable(enable bool) {
-	if enable {
-		h.irq.Enable() // Enable USB interrupts
-	} else {
-		h.irq.Disable() // Disable USB interrupts
-	}
-}
@@ -1,42 +0,0 @@
-package usb
-
-import (
-	"errors"
-)
-
-var (
-	ErrHIDInvalidPort    = errors.New("invalid USB port")
-	ErrHIDInvalidCore    = errors.New("invalid USB core")
-	ErrHIDReportTransfer = errors.New("failed to transfer HID report")
-)
-
-type HID struct {
-	Port int
-	core *core
-}
-
-type HIDConfig struct {
-	// Port is the MCU's native USB core number. If in doubt, leave it
-	// uninitialized for default (0).
-	Port int
-}
-
-func (hid *HID) Configure(config HIDConfig) error {
-
-	if config.Port >= CoreCount || config.Port >= dcdCount {
-		return ErrHIDInvalidPort
-	}
-	hid.Port = config.Port
-
-	// verify we have a free USB port and take ownership of it
-	var st status
-	hid.core, st = initCore(hid.Port, class{id: classDeviceHID, config: 1})
-	if !st.ok() {
-		return ErrHIDInvalidPort
-	}
-	return nil
-}
-
-func (hid *HID) Keyboard() *Keyboard {
-	return hid.core.dc.keyboard()
-}
@@ -1,74 +0,0 @@
-package usb
-
-import (
-	"errors"
-)
-
-var (
-	ErrUARTInvalidPort = errors.New("invalid USB port")
-	ErrUARTInvalidCore = errors.New("invalid USB core")
-	ErrUARTEmptyBuffer = errors.New("USB receive buffer empty")
-	ErrUARTWriteFailed = errors.New("USB write failure")
-)
-
-// UART represents a virtual serial (UART) device emulation using the USB
-// CDC-ACM device class driver.
-type UART struct {
-	Port int
-	core *core
-}
-
-type UARTConfig struct {
-	// Port is the MCU's native USB core number. If in doubt, leave it
-	// uninitialized for default (0).
-	Port int
-}
-
-func (uart *UART) Configure(config UARTConfig) error {
-
-	if config.Port >= CoreCount || config.Port >= dcdCount {
-		return ErrUARTInvalidPort
-	}
-	uart.Port = config.Port
-
-	// verify we have a free USB port and take ownership of it
-	var st status
-	uart.core, st = initCore(uart.Port, class{id: classDeviceCDCACM, config: 1})
-	if !st.ok() {
-		return ErrUARTInvalidPort
-	}
-	return nil
-}
-
-// Buffered returns the number of bytes currently stored in the RX buffer.
-func (uart UART) Buffered() int {
-	return uart.core.dc.uartAvailable()
-}
-
-// ReadByte reads a single byte from the RX buffer.
-// If there is no data in the buffer, returns an error.
-func (uart UART) ReadByte() (byte, error) {
-	n, ok := uart.core.dc.uartReadByte()
-	if !ok {
-		return 0, ErrUARTEmptyBuffer
-	}
-	return n, nil
-}
-
-// Read from the RX buffer.
-func (uart UART) Read(data []byte) (n int, err error) {
-	return uart.core.dc.uartRead(data), nil
-}
-
-// WriteByte writes a single byte of data to the UART interface.
-func (uart UART) WriteByte(c byte) error {
-	if !uart.core.dc.uartWriteByte(c) {
-		return ErrUARTWriteFailed
-	}
-	return nil
-}
-
-// Write data to the UART.
-func (uart UART) Write(data []byte) (n int, err error) {
-	return uart.core.dc.uartWrite(data), nil
-}
@@ -1,143 +0,0 @@
-package usb
-
-// Hardware abstraction for USB ports configured as either host or device.
-
-// CoreCount defines the total number of USB cores to configure in device or
-// host mode.
-const CoreCount = dcdCount + hcdCount
-
-// coreInstance provides statically-allocated instances of each USB core
-// configured on this platform.
-var coreInstance [CoreCount]core
-
-// core represents the core of a USB port configured as either host or device.
-type core struct {
-	port int
-	mode int
-	dc   *dcd
-	hc   *hcd
-}
-
-// Constant definitions for USB core operating modes.
-const (
-	modeIdle   = 0 // USB port has not been configured
-	modeDevice = 1
-	modeHost   = 2
-)
-
-// initCore initializes a free USB core with given operating mode on the USB
-// port at given index, if available. Returns a reference to the initialized
-// core or nil if the core is unavailable.
-func initCore(port int, class class) (*core, status) {
-
-	iv := disableInterrupts()
-	defer enableInterrupts(iv)
-
-	if port < 0 || port >= CoreCount || 0 == class.config {
-		return nil, statusInvalid
-	}
-
-	if modeIdle != coreInstance[port].mode {
-		// Check if requested port is already configured as requested class. If so,
-		// just return a reference to the existing core instead of an error.
-		//
-		// This will allow, for instance, TinyGo examples that try to reconfigure
-		// the USB (CDC-ACM) UART port (which is already configured by the runtime)
-		// to continue without error.
-		if coreInstance[port].mode == class.mode() {
-			switch class.mode() {
-			case modeDevice:
-				if coreInstance[port].dc.class().equals(class) {
-					return &coreInstance[port], statusOK
-				}
-			case modeHost:
-				if coreInstance[port].hc.class().equals(class) {
-					return &coreInstance[port], statusOK
-				}
-			}
-		}
-		return nil, statusBusy
-	}
-
-	switch class.mode() {
-	case modeDevice:
-		// Allocate a free device controller and install interrupts
-		dc, st := initDCD(port, class)
-		if !st.ok() {
-			return nil, st
-		}
-		// Initialize buffers and device descriptors
-		if st = dc.init(); !st.ok() {
-			return nil, st
-		}
-		coreInstance[port].port = port
-		coreInstance[port].mode = modeDevice
-		coreInstance[port].dc = dc
-		dc.enable(true) // Enable interrupts and enter runtime
-
-	case modeHost:
-		// Allocate a free host controller and install interrupts
-		hc, st := initHCD(port, class)
-		if !st.ok() {
-			return nil, st
-		}
-		// Initialize buffers and device descriptors
-		if st = hc.init(); !st.ok() {
-			return nil, st
-		}
-		coreInstance[port].port = port
-		coreInstance[port].mode = modeHost
-		coreInstance[port].hc = hc
-		hc.enable(true) // Enable interrupts and enter runtime
-
-	default:
-		return nil, statusInvalid
-	}
-
-	return &coreInstance[port], statusOK
-}
-
-// class represents the type of a host/device and its class configuration index.
-// The first valid configuration index is 1. Index 0 is reserved and invalid.
-type class struct {
-	id     int
-	config int
-}
-
-// Enumerated constants for all host/device class configurations.
-const (
-	classDeviceCDCACM = 0
-	classDeviceHID    = 1
-)
-
-// mode returns the USB core operating mode of the receiver class c.
-//go:inline
-func (c class) mode() int {
-	switch c.id {
-	case classDeviceCDCACM, classDeviceHID:
-		return modeDevice
-	default:
-		return modeIdle
-	}
-}
-
-// equals returns true if and only if all fields of the given class are equal to
-// those of the receiver c.
-//go:inline
-func (c class) equals(class class) bool {
-	return c.id == class.id && c.config == class.config
-}
-
-// status represents the return code of a subroutine.
-type status uint8
-
-// Constant definitions for all status codes used within the package.
-const (
-	statusOK      status = iota // Success
-	statusBusy                  // Busy
-	statusInvalid               // Invalid argument
-)
-
-// ok returns true if and only if the receiver st equals statusOK.
-//go:inline
-func (s status) ok() bool { return statusOK == s }
@@ -1,271 +0,0 @@
-package usb
-
-//go:linkname ticks runtime.ticks
-func ticks() int64
-
-// leU64 returns a slice containing 8 bytes from the given uint64 u.
-//
-// The returned bytes have little-endian ordering; that is, the first element
-// at index 0 is the least-significant byte in u and index 7 is the most-
-// significant byte.
-//go:inline
-func leU64(u uint64) []uint8 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return []uint8{0, 0, 0, 0, 0, 0, 0, 0}
-	}
-	return []uint8{
-		uint8(u), uint8(u >> 8), uint8(u >> 16), uint8(u >> 24),
-		uint8(u >> 32), uint8(u >> 40), uint8(u >> 48), uint8(u >> 56),
-	}
-}
-
-// leU32 returns a slice containing 4 bytes from the given uint32 u.
-//
-// The returned bytes have little-endian ordering; that is, the first element
-// at index 0 is the least-significant byte in u and index 3 is the most-
-// significant byte.
-//go:inline
-func leU32(u uint32) []uint8 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return []uint8{0, 0, 0, 0}
-	}
-	return []uint8{
-		uint8(u), uint8(u >> 8), uint8(u >> 16), uint8(u >> 24),
-	}
-}
-
-// leU16 returns a slice containing 2 bytes from the given uint16 u.
-//
-// The returned bytes have little-endian ordering; that is, the first element
-// at index 0 is the least-significant byte in u and index 1 is the most-
-// significant byte.
-//go:inline
-func leU16(u uint16) []uint8 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return []uint8{0, 0}
-	}
-	return []uint8{
-		uint8(u), uint8(u >> 8),
-	}
-}
-
-// beU64 returns a slice containing 8 bytes from the given uint64 u.
-//
-// The returned bytes have big-endian ordering; that is, the first element at
-// index 0 is the most-significant byte in u and index 7 is the least-
-// significant byte.
-//go:inline
-func beU64(u uint64) []uint8 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return []uint8{0, 0, 0, 0, 0, 0, 0, 0}
-	}
-	return []uint8{
-		uint8(u >> 56), uint8(u >> 48), uint8(u >> 40), uint8(u >> 32),
-		uint8(u >> 24), uint8(u >> 16), uint8(u >> 8), uint8(u),
-	}
-}
-
-// beU32 returns a slice containing 4 bytes from the given uint32 u.
-//
-// The returned bytes have big-endian ordering; that is, the first element at
-// index 0 is the most-significant byte in u and index 3 is the least-
-// significant byte.
-//go:inline
-func beU32(u uint32) []uint8 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return []uint8{0, 0, 0, 0}
-	}
-	return []uint8{
-		uint8(u >> 24), uint8(u >> 16), uint8(u >> 8), uint8(u),
-	}
-}
-
-// beU16 returns a slice containing 2 bytes from the given uint16 u.
-//
-// The returned bytes have big-endian ordering; that is, the first element at
-// index 0 is the most-significant byte in u and index 1 is the least-
-// significant byte.
-//go:inline
-func beU16(u uint16) []uint8 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return []uint8{0, 0}
-	}
-	return []uint8{
-		uint8(u >> 8), uint8(u),
-	}
-}
-
-// revU64 returns the given uint64 u with bytes in the reverse order.
-//go:inline
-func revU64(u uint64) uint64 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return 0
-	}
-	return ((u & 0x00000000000000FF) << 56) |
-		((u & 0x000000000000FF00) << 40) |
-		((u & 0x0000000000FF0000) << 24) |
-		((u & 0x00000000FF000000) << 8) |
-		((u & 0x000000FF00000000) >> 8) |
-		((u & 0x0000FF0000000000) >> 24) |
-		((u & 0x00FF000000000000) >> 40) |
-		((u & 0xFF00000000000000) >> 56)
-}
-
-// revU32 returns the given uint32 u with bytes in the reverse order.
-//go:inline
-func revU32(u uint32) uint32 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return 0
-	}
-	return ((u & 0x000000FF) << 24) | ((u & 0x0000FF00) << 8) |
-		((u & 0x00FF0000) >> 8) | ((u & 0xFF000000) >> 24)
-}
-
-// revU16 returns the given uint16 u with bytes in the reverse order.
-//go:inline
-func revU16(u uint16) uint16 {
-	if u == 0 {
-		// skip all processing for the common case (u = 0)
-		return 0
-	}
-	return ((u & 0x00FF) << 8) | ((u & 0xFF00) >> 8)
-}
-
-// packU64 returns a uint64 constructed by concatenating the bytes in slice b.
-//
-// The least-significant byte in the returned value is the first element at
-// index 0 in b and the most significant byte is index 7, if given. If fewer
-// than 8 elements are given in b, the corresponding bytes in the returned value
-// are all 0.
-//go:inline
-func packU64(b []uint8) (u uint64) {
-	for i := 0; i < 8 && i < len(b); i++ {
-		u |= uint64(b[i]) << (i * 8)
-	}
-	return
-}
-
-// packU32 returns a uint32 constructed by concatenating the bytes in slice b.
-//
-// The least-significant byte in the returned value is the first element at
-// index 0 in b and the most significant byte is index 3, if given. If fewer
-// than 4 elements are given in b, the corresponding bytes in the returned value
-// are all 0.
-//go:inline
-func packU32(b []uint8) (u uint32) {
-	for i := 0; i < 4 && i < len(b); i++ {
-		u |= uint32(b[i]) << (i * 8)
-	}
-	return
-}
-
-// packU16 returns a uint16 constructed by concatenating the bytes in slice b.
-//
-// The least-significant byte in the returned value is the first element at
-// index 0 in b and the most significant byte is index 1, if given. If fewer
-// than 2 elements are given in b, the corresponding bytes in the returned value
-// are all 0.
-//go:inline
-func packU16(b []uint8) (u uint16) {
-	for i := 0; i < 2 && i < len(b); i++ {
-		u |= uint16(b[i]) << (i * 8)
-	}
-	return
-}
-
-// msU8 returns the most-significant byte of u.
-//go:inline
-func msU8(u uint16) uint8 { return uint8(u >> 8) }
-
-// lsU8 returns the least-significant byte of u.
-//go:inline
-func lsU8(u uint16) uint8 { return uint8(u) }
-
-// cycles converts the given number of microseconds to CPU cycles for a CPU with
-// given frequency.
-//go:inline
-func cycles(microsec, cpuFreqHz uint32) uint32 {
-	return uint32((uint64(microsec) * uint64(cpuFreqHz)) / 1000000)
-}
-
-//go:inline
-func unpackEndpoint(address uint8) (number, direction uint8) {
-	return (address & descEndptAddrNumberMsk) >> descEndptAddrNumberPos,
-		(address & descEndptAddrDirectionMsk) >> descEndptAddrDirectionPos
-}
-
-//go:inline
-func rxEndpoint(number uint8) uint8 {
-	return (number & descEndptAddrNumberMsk) | descEndptAddrDirectionOut
-}
-
-//go:inline
-func txEndpoint(number uint8) uint8 {
-	return (number & descEndptAddrNumberMsk) | descEndptAddrDirectionIn
-}
-
-//go:inline
-func endpointIndex(address uint8) uint8 {
-	return ((address & descEndptAddrNumberMsk) << 1) |
-		((address & descEndptAddrDirectionMsk) >> descEndptAddrDirectionPos)
-}
-
-// The following buffLo and buffHi are helper methods for slice definitions from
-// potentially zero-length arrays (depending on compile-time constants).
-//
-// For example, if we have an array containing a 5-element buffer for three
-// instances of some device class (15 total elements), partitioned as follows,
-// then we compute the indices for instance 2 as usual:
-//
-//    Index:   01234 56789 ABCDE
-//    Array:  [  1  |  2  |  3  ]
-//
-//       Lo:  (n-1) * size   =>   (2-1) * 5   =>   5
-//       Hi:    (n) * size   =>     (2) * 5   =>   10 (0xA)
-//
-// However, if we have specified (via const definition) that 0 instances of some
-// device class be allocated, then the associated device class buffer arrays
-// will all be zero-length arrays, and the arithmetic to compute the slice
-// indices used above will result in out-of-bounds indices:
-//
-//    Index:
-//    Array:  []
-//
-//       Lo:  (n-1) * size   =>   (2-1) * 5   =>   5           [Error!]
-//       Hi:    (n) * size   =>     (2) * 5   =>   10 (0xA)    [Error!]
-//
-//
-// I couldn't figure out a straight-forward way to resolve these slice indices
-// using only arithmetic, so I've resorted to simple conditionals. If the number
-// of instances for some given class is zero (count=0), defined via compile-time
-// constant, then just use the empty slice range [0:0].
-
-// buffLo returns the starting array slice index for the n'th region of size
-// elements from an array containing count regions of size elements.
-// Regions are specified using a 1-based index (n > 0). Returns 0 if any given
-// argument equals 0.
-func buffLo(n, count, size uint16) uint16 {
-	if 0 == n || 0 == count || 0 == size {
-		return 0
-	}
-	return (n - 1) * size
-}
-
-// buffHi returns the ending array slice index for the n'th region of size
-// elements from an array containing count regions of size elements.
-// Regions are specified using a 1-based index (n > 0). Returns 0 if any given
-// argument equals 0.
-func buffHi(n, count, size uint16) uint16 {
-	if 0 == n || 0 == count || 0 == size {
-		return 0
-	}
-	return n * size
-}
@@ -1,24 +0,0 @@
-// +build arm
-
-package usb
-
-import "device/arm"
-
-// udelay waits for the given number of microseconds before returning.
-// We cannot use the sleep timer from this context (import cycle), but we need
-// an approximate method to spin CPU cycles for short periods of time.
-//go:inline
-func udelay(microsec uint32) {
-	n := cycles(microsec, descCPUFrequencyHz)
-	for i := uint32(0); i < n; i++ {
-		arm.Asm(`nop`)
-	}
-}
-
-func disableInterrupts() uintptr {
-	return arm.DisableInterrupts()
-}
-
-func enableInterrupts(mask uintptr) {
-	arm.EnableInterrupts(mask)
-}
@@ -14,7 +14,10 @@

 package net

-import "internal/bytealg"
+import (
+	"internal/bytealg"
+	"internal/itoa"
+)

 // IP address lengths (bytes).
 const (
@@ -533,7 +536,7 @@ func (n *IPNet) String() string {
 	if l == -1 {
 		return nn.String() + "/" + m.String()
 	}
-	return nn.String() + "/" + uitoa(uint(l))
+	return nn.String() + "/" + itoa.Uitoa(uint(l))
 }

 // Parse IPv4 address (d.d.d.d).
@@ -64,24 +64,6 @@ func xtoi2(s string, e byte) (byte, bool) {
 	return byte(n), ok && ei == 2
 }

-// Convert unsigned integer to decimal string.
-func uitoa(val uint) string {
-	if val == 0 { // avoid string allocation
-		return "0"
-	}
-	var buf [20]byte // big enough for 64bit value base 10
-	i := len(buf) - 1
-	for val >= 10 {
-		q := val / 10
-		buf[i] = byte('0' + val - q*10)
-		i--
-		val = q
-	}
-	// val < 10
-	buf[i] = byte('0' + val)
-	return string(buf[i:])
-}
-
 // Convert i to a hexadecimal string. Leading zeros are not printed.
 func appendHex(dst []byte, i uint32) []byte {
 	if i == 0 {
@@ -0,0 +1,114 @@
+// +build !baremetal,!js
+
+package os
+
+import (
+	"io"
+	"syscall"
+)
+
+func init() {
+	// Mount the host filesystem at the root directory. This is what most
+	// programs will be expecting.
+	Mount("/", unixFilesystem{})
+}
+
+// Stdin, Stdout, and Stderr are open Files pointing to the standard input,
+// standard output, and standard error file descriptors.
+var (
+	Stdin  = &File{unixFileHandle(syscall.Stdin), "/dev/stdin"}
+	Stdout = &File{unixFileHandle(syscall.Stdout), "/dev/stdout"}
+	Stderr = &File{unixFileHandle(syscall.Stderr), "/dev/stderr"}
+)
+
+// isOS indicates whether we're running on a real operating system with
+// filesystem support.
+const isOS = true
+
+// unixFilesystem is an empty handle for a Unix/Linux filesystem. All operations
+// are relative to the current working directory.
+type unixFilesystem struct {
+}
+
+func (fs unixFilesystem) Mkdir(path string, perm FileMode) error {
+	return handleSyscallError(syscall.Mkdir(path, uint32(perm)))
+}
+
+func (fs unixFilesystem) Remove(path string) error {
+	return handleSyscallError(syscall.Unlink(path))
+}
+
+func (fs unixFilesystem) OpenFile(path string, flag int, perm FileMode) (FileHandle, error) {
+	// Map os package flags to syscall flags.
+	syscallFlag := 0
+	if flag&O_RDONLY != 0 {
+		syscallFlag |= syscall.O_RDONLY
+	}
+	if flag&O_WRONLY != 0 {
+		syscallFlag |= syscall.O_WRONLY
+	}
+	if flag&O_RDWR != 0 {
+		syscallFlag |= syscall.O_RDWR
+	}
+	if flag&O_APPEND != 0 {
+		syscallFlag |= syscall.O_APPEND
+	}
+	if flag&O_CREATE != 0 {
+		syscallFlag |= syscall.O_CREAT
+	}
+	if flag&O_EXCL != 0 {
+		syscallFlag |= syscall.O_EXCL
+	}
+	if flag&O_SYNC != 0 {
+		syscallFlag |= syscall.O_SYNC
+	}
+	if flag&O_TRUNC != 0 {
+		syscallFlag |= syscall.O_TRUNC
+	}
+	fp, err := syscall.Open(path, syscallFlag, uint32(perm))
+	return unixFileHandle(fp), handleSyscallError(err)
+}
+
+// unixFileHandle is a Unix file pointer with associated methods that implement
+// the FileHandle interface.
+type unixFileHandle uintptr
+
+// Read reads up to len(b) bytes from the File. It returns the number of bytes
+// read and any error encountered. At end of file, Read returns 0, io.EOF.
+func (f unixFileHandle) Read(b []byte) (n int, err error) {
+	n, err = syscall.Read(syscallFd(f), b)
+	err = handleSyscallError(err)
+	if n == 0 && err == nil {
+		err = io.EOF
+	}
+	return
+}
+
+// Write writes len(b) bytes to the File. It returns the number of bytes written
+// and an error, if any. Write returns a non-nil error when n != len(b).
+func (f unixFileHandle) Write(b []byte) (n int, err error) {
+	n, err = syscall.Write(syscallFd(f), b)
+	err = handleSyscallError(err)
+	return
+}
+
+// Close closes the File, rendering it unusable for I/O.
+func (f unixFileHandle) Close() error {
+	return handleSyscallError(syscall.Close(syscallFd(f)))
+}
+
+// handleSyscallError converts syscall errors into regular os package errors.
+// The err parameter must be either nil or of type syscall.Errno.
+func handleSyscallError(err error) error {
+	if err == nil {
+		return nil
+	}
+	switch err.(syscall.Errno) {
+	case syscall.EEXIST:
+		return ErrExist
+	case syscall.ENOENT:
+		return ErrNotExist
+	default:
+		return err
+	}
+}
@@ -2,113 +2,4 @@

 package os

-import (
-	"io"
-	"syscall"
-)
-
-func init() {
-	// Mount the host filesystem at the root directory. This is what most
-	// programs will be expecting.
-	Mount("/", unixFilesystem{})
-}
-
-// Stdin, Stdout, and Stderr are open Files pointing to the standard input,
-// standard output, and standard error file descriptors.
-var (
-	Stdin  = &File{unixFileHandle(0), "/dev/stdin"}
-	Stdout = &File{unixFileHandle(1), "/dev/stdout"}
-	Stderr = &File{unixFileHandle(2), "/dev/stderr"}
-)
-
-// isOS indicates whether we're running on a real operating system with
-// filesystem support.
-const isOS = true
-
-// unixFilesystem is an empty handle for a Unix/Linux filesystem. All operations
-// are relative to the current working directory.
-type unixFilesystem struct {
-}
-
-func (fs unixFilesystem) Mkdir(path string, perm FileMode) error {
-	return handleSyscallError(syscall.Mkdir(path, uint32(perm)))
-}
-
-func (fs unixFilesystem) Remove(path string) error {
-	return handleSyscallError(syscall.Unlink(path))
-}
-
-func (fs unixFilesystem) OpenFile(path string, flag int, perm FileMode) (FileHandle, error) {
-	// Map os package flags to syscall flags.
-	syscallFlag := 0
-	if flag&O_RDONLY != 0 {
-		syscallFlag |= syscall.O_RDONLY
-	}
-	if flag&O_WRONLY != 0 {
-		syscallFlag |= syscall.O_WRONLY
-	}
-	if flag&O_RDWR != 0 {
-		syscallFlag |= syscall.O_RDWR
-	}
-	if flag&O_APPEND != 0 {
-		syscallFlag |= syscall.O_APPEND
-	}
-	if flag&O_CREATE != 0 {
-		syscallFlag |= syscall.O_CREAT
-	}
-	if flag&O_EXCL != 0 {
-		syscallFlag |= syscall.O_EXCL
-	}
-	if flag&O_SYNC != 0 {
-		syscallFlag |= syscall.O_SYNC
-	}
-	if flag&O_TRUNC != 0 {
-		syscallFlag |= syscall.O_TRUNC
-	}
-	fp, err := syscall.Open(path, syscallFlag, uint32(perm))
-	return unixFileHandle(fp), handleSyscallError(err)
-}
-
-// unixFileHandle is a Unix file pointer with associated methods that implement
-// the FileHandle interface.
-type unixFileHandle uintptr
-
-// Read reads up to len(b) bytes from the File. It returns the number of bytes
-// read and any error encountered. At end of file, Read returns 0, io.EOF.
-func (f unixFileHandle) Read(b []byte) (n int, err error) {
-	n, err = syscall.Read(int(f), b)
-	err = handleSyscallError(err)
-	if n == 0 && err == nil {
-		err = io.EOF
-	}
-	return
-}
-
-// Write writes len(b) bytes to the File. It returns the number of bytes written
-// and an error, if any. Write returns a non-nil error when n != len(b).
-func (f unixFileHandle) Write(b []byte) (n int, err error) {
-	n, err = syscall.Write(int(f), b)
-	err = handleSyscallError(err)
-	return
-}
-
-// Close closes the File, rendering it unusable for I/O.
-func (f unixFileHandle) Close() error {
-	return handleSyscallError(syscall.Close(int(f)))
-}
-
-// handleSyscallError converts syscall errors into regular os package errors.
-// The err parameter must be either nil or of type syscall.Errno.
-func handleSyscallError(err error) error {
-	if err == nil {
-		return nil
-	}
-	switch err.(syscall.Errno) {
-	case syscall.EEXIST:
-		return ErrExist
-	case syscall.ENOENT:
-		return ErrNotExist
-	default:
-		return err
-	}
-}
+type syscallFd = int
@@ -0,0 +1,7 @@
+// +build windows
+
+package os
+
+import "syscall"
+
+type syscallFd = syscall.Handle
@@ -0,0 +1,188 @@
+// Copyright 2009 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package reflect_test
+
+import (
+	"math"
+	. "reflect"
+	"testing"
+)
+
+type Basic struct {
+	x int
+	y float32
+}
+
+type NotBasic Basic
+
+type DeepEqualTest struct {
+	a, b interface{}
+	eq   bool
+}
+
+// Simple functions for DeepEqual tests.
+var (
+	fn1 func()             // nil.
+	fn2 func()             // nil.
+	fn3 = func() { fn1() } // Not nil.
+)
+
+type self struct{}
+
+type Loopy interface{}
+
+var loopy1, loopy2 Loopy
+var cycleMap1, cycleMap2, cycleMap3 map[string]interface{}
+
+type structWithSelfPtr struct {
+	p *structWithSelfPtr
+	s string
+}
+
+func init() {
+	loopy1 = &loopy2
+	loopy2 = &loopy1
+
+	cycleMap1 = map[string]interface{}{}
+	cycleMap1["cycle"] = cycleMap1
+	cycleMap2 = map[string]interface{}{}
+	cycleMap2["cycle"] = cycleMap2
+	cycleMap3 = map[string]interface{}{}
+	cycleMap3["different"] = cycleMap3
+}
+
+// Note: all tests involving maps have been commented out because they aren't
+// supported yet.
+var deepEqualTests = []DeepEqualTest{
+	// Equalities
+	{nil, nil, true},
+	{1, 1, true},
+	{int32(1), int32(1), true},
+	{0.5, 0.5, true},
+	{float32(0.5), float32(0.5), true},
+	{"hello", "hello", true},
+	{make([]int, 10), make([]int, 10), true},
+	{&[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},
+	{fn1, fn2, true},
+	{[]byte{1, 2, 3}, []byte{1, 2, 3}, true},
+	{[]MyByte{1, 2, 3}, []MyByte{1, 2, 3}, true},
+	{MyBytes{1, 2, 3}, MyBytes{1, 2, 3}, true},
+
+	// Inequalities
+	{1, 2, false},
+	{int32(1), int32(2), false},
+	{0.5, 0.6, false},
+	{float32(0.5), float32(0.6), false},
+	{"hello", "hey", false},
+	{make([]int, 10), make([]int, 11), false},
+	{&[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},
+	{nil, 1, false},
+	{1, nil, false},
+	{fn1, fn3, false},
+	{fn3, fn3, false},
+	{[][]int{{1}}, [][]int{{2}}, false},
+	{&structWithSelfPtr{p: &structWithSelfPtr{s: "a"}}, &structWithSelfPtr{p: &structWithSelfPtr{s: "b"}}, false},
+
+	// Fun with floating point.
+	{math.NaN(), math.NaN(), false},
+	{&[1]float64{math.NaN()}, &[1]float64{math.NaN()}, false},
+	{&[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},
+
+	// 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},
+
+	// Mismatched types
+	{1, 1.0, false},
+	{int32(1), int64(1), false},
+	{0.5, "hello", false},
+	{[]int{1, 2, 3}, [3]int{1, 2, 3}, false},
+	{&[3]interface{}{1, 2, 4}, &[3]interface{}{1, 2, "s"}, false},
+	{Basic{1, 0.5}, NotBasic{1, 0.5}, false},
+	{map[uint]string{1: "one", 2: "two"}, map[int]string{2: "two", 1: "one"}, false},
+	{[]byte{1, 2, 3}, []MyByte{1, 2, 3}, false},
+	{[]MyByte{1, 2, 3}, MyBytes{1, 2, 3}, false},
+	{[]byte{1, 2, 3}, MyBytes{1, 2, 3}, false},
+
+	// Possible loops.
+	{&loopy1, &loopy1, true},
+	{&loopy1, &loopy2, true},
+	//{&cycleMap1, &cycleMap2, true},
+	{&cycleMap1, &cycleMap3, false},
+}
+
+func TestDeepEqual(t *testing.T) {
+	for _, test := range deepEqualTests {
+		if test.b == (self{}) {
+			test.b = test.a
+		}
+		if r := DeepEqual(test.a, test.b); r != test.eq {
+			t.Errorf("DeepEqual(%#v, %#v) = %v, want %v", test.a, test.b, r, test.eq)
+		}
+	}
+}
+
+type Recursive struct {
+	x int
+	r *Recursive
+}
+
+func TestDeepEqualRecursiveStruct(t *testing.T) {
+	a, b := new(Recursive), new(Recursive)
+	*a = Recursive{12, a}
+	*b = Recursive{12, b}
+	if !DeepEqual(a, b) {
+		t.Error("DeepEqual(recursive same) = false, want true")
+	}
+}
+
+type _Complex struct {
+	a int
+	b [3]*_Complex
+	c *string
+	d map[float64]float64
+}
+
+func TestDeepEqualComplexStruct(t *testing.T) {
+	m := make(map[float64]float64)
+	stra, strb := "hello", "hello"
+	a, b := new(_Complex), new(_Complex)
+	*a = _Complex{5, [3]*_Complex{a, b, a}, &stra, m}
+	*b = _Complex{5, [3]*_Complex{b, a, a}, &strb, m}
+	if !DeepEqual(a, b) {
+		t.Error("DeepEqual(complex same) = false, want true")
+	}
+}
+
+func TestDeepEqualComplexStructInequality(t *testing.T) {
+	m := make(map[float64]float64)
+	stra, strb := "hello", "helloo" // Difference is here
+	a, b := new(_Complex), new(_Complex)
+	*a = _Complex{5, [3]*_Complex{a, b, a}, &stra, m}
+	*b = _Complex{5, [3]*_Complex{b, a, a}, &strb, m}
+	if DeepEqual(a, b) {
+		t.Error("DeepEqual(complex different) = true, want false")
+	}
+}
+
+type MyBytes []byte
+type MyByte byte
@@ -1,9 +1,196 @@
+// Copyright 2009 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Deep equality test via reflection
+
 package reflect

+import "unsafe"
+
+// During deepValueEqual, must keep track of checks that are
+// in progress. The comparison algorithm assumes that all
+// checks in progress are true when it reencounters them.
+// Visited comparisons are stored in a map indexed by visit.
+type visit struct {
+	a1  unsafe.Pointer
+	a2  unsafe.Pointer
+	typ rawType
+}
+
+// Tests for deep equality using reflected types. The map argument tracks
+// comparisons that have already been seen, which allows short circuiting on
+// recursive types.
+func deepValueEqual(v1, v2 Value, visited map[visit]struct{}) bool {
+	if !v1.IsValid() || !v2.IsValid() {
+		return v1.IsValid() == v2.IsValid()
+	}
+	if v1.typecode != v2.typecode {
+		return false
+	}
+
+	// We want to avoid putting more in the visited map than we need to.
+	// For any possible reference cycle that might be encountered,
+	// hard(v1, v2) needs to return true for at least one of the types in the cycle,
+	// and it's safe and valid to get Value's internal pointer.
+	hard := func(v1, v2 Value) bool {
+		switch v1.Kind() {
+		case Map, Slice, Ptr, Interface:
+			// Nil pointers cannot be cyclic. Avoid putting them in the visited map.
+			return !v1.IsNil() && !v2.IsNil()
+		}
+		return false
+	}
+
+	if hard(v1, v2) {
+		addr1 := v1.pointer()
+		addr2 := v2.pointer()
+		if uintptr(addr1) > uintptr(addr2) {
+			// Canonicalize order to reduce number of entries in visited.
+			// Assumes non-moving garbage collector.
+			addr1, addr2 = addr2, addr1
+		}
+
+		// Short circuit if references are already seen.
+		v := visit{addr1, addr2, v1.typecode}
+		if _, ok := visited[v]; ok {
+			return true
+		}
+
+		// Remember for later.
+		visited[v] = struct{}{}
+	}
+
+	switch v1.Kind() {
+	case Array:
+		for i := 0; i < v1.Len(); i++ {
+			if !deepValueEqual(v1.Index(i), v2.Index(i), visited) {
+				return false
+			}
+		}
+		return true
+	case Slice:
+		if v1.IsNil() != v2.IsNil() {
+			return false
+		}
+		if v1.Len() != v2.Len() {
+			return false
+		}
+		if v1.Pointer() == v2.Pointer() {
+			return true
+		}
+		for i := 0; i < v1.Len(); i++ {
+			if !deepValueEqual(v1.Index(i), v2.Index(i), visited) {
+				return false
+			}
+		}
+		return true
+	case Interface:
+		if v1.IsNil() || v2.IsNil() {
+			return v1.IsNil() == v2.IsNil()
+		}
+		return deepValueEqual(v1.Elem(), v2.Elem(), visited)
+	case Ptr:
+		if v1.Pointer() == v2.Pointer() {
+			return true
+		}
+		return deepValueEqual(v1.Elem(), v2.Elem(), visited)
+	case Struct:
+		for i, n := 0, v1.NumField(); i < n; i++ {
+			if !deepValueEqual(v1.Field(i), v2.Field(i), visited) {
+				return false
+			}
+		}
+		return true
+	case Map:
+		if v1.IsNil() != v2.IsNil() {
+			return false
+		}
+		if v1.Len() != v2.Len() {
+			return false
+		}
+		if v1.Pointer() == v2.Pointer() {
+			return true
+		}
+		for _, k := range v1.MapKeys() {
+			val1 := v1.MapIndex(k)
+			val2 := v2.MapIndex(k)
+			if !val1.IsValid() || !val2.IsValid() || !deepValueEqual(val1, val2, visited) {
+				return false
+			}
+		}
+		return true
+	case Func:
+		if v1.IsNil() && v2.IsNil() {
+			return true
+		}
+		// Can't do better than this:
+		return false
+	default:
+		// Normal equality suffices
+		return valueInterfaceUnsafe(v1) == valueInterfaceUnsafe(v2)
+	}
+}
+
+// DeepEqual reports whether x and y are ``deeply equal,'' defined as follows.
+// Two values of identical type are deeply equal if one of the following cases applies.
+// Values of distinct types are never deeply equal.
+//
+// Array values are deeply equal when their corresponding elements are deeply equal.
+//
+// Struct values are deeply equal if their corresponding fields,
+// both exported and unexported, are deeply equal.
+//
+// Func values are deeply equal if both are nil; otherwise they are not deeply equal.
+//
+// Interface values are deeply equal if they hold deeply equal concrete values.
+//
+// Map values are deeply equal when all of the following are true:
+// they are both nil or both non-nil, they have the same length,
+// and either they are the same map object or their corresponding keys
+// (matched using Go equality) map to deeply equal values.
+//
+// Pointer values are deeply equal if they are equal using Go's == operator
+// or if they point to deeply equal values.
+//
+// Slice values are deeply equal when all of the following are true:
+// they are both nil or both non-nil, they have the same length,
+// and either they point to the same initial entry of the same underlying array
+// (that is, &x[0] == &y[0]) or their corresponding elements (up to length) are deeply equal.
+// Note that a non-nil empty slice and a nil slice (for example, []byte{} and []byte(nil))
+// are not deeply equal.
+//
+// Other values - numbers, bools, strings, and channels - are deeply equal
+// if they are equal using Go's == operator.
+//
+// In general DeepEqual is a recursive relaxation of Go's == operator.
+// However, this idea is impossible to implement without some inconsistency.
+// Specifically, it is possible for a value to be unequal to itself,
+// either because it is of func type (uncomparable in general)
+// or because it is a floating-point NaN value (not equal to itself in floating-point comparison),
+// or because it is an array, struct, or interface containing
+// such a value.
+// On the other hand, pointer values are always equal to themselves,
+// even if they point at or contain such problematic values,
+// because they compare equal using Go's == operator, and that
+// is a sufficient condition to be deeply equal, regardless of content.
+// DeepEqual has been defined so that the same short-cut applies
+// to slices and maps: if x and y are the same slice or the same map,
+// they are deeply equal regardless of content.
+//
+// As DeepEqual traverses the data values it may find a cycle. The
+// second and subsequent times that DeepEqual compares two pointer
+// values that have been compared before, it treats the values as
+// equal rather than examining the values to which they point.
+// This ensures that DeepEqual terminates.
 func DeepEqual(x, y interface{}) bool {
 	if x == nil || y == nil {
 		return x == y
 	}
-
-	panic("unimplemented: reflect.DeepEqual()")
+	v1 := ValueOf(x)
+	v2 := ValueOf(y)
+	if v1.typecode != v2.typecode {
+		return false
+	}
+	return deepValueEqual(v1, v2, make(map[visit]struct{}))
 }
@@ -67,6 +67,14 @@ func (v Value) Interface() interface{} {
 // valueInterfaceUnsafe is used by the runtime to hash map keys. It should not
 // be subject to the isExported check.
 func valueInterfaceUnsafe(v Value) interface{} {
+	if v.typecode.Kind() == Interface {
+		// The value itself is an interface. This can happen when getting the
+		// value of a struct field of interface type, like this:
+		//     type T struct {
+		//         X interface{}
+		//     }
+		return *(*interface{})(v.value)
+	}
 	if v.isIndirect() && v.typecode.Size() <= unsafe.Sizeof(uintptr(0)) {
 		// Value was indirect but must be put back directly in the interface
 		// value.
@@ -132,10 +140,7 @@ func (v Value) IsNil() bool {
 func (v Value) Pointer() uintptr {
 	switch v.Kind() {
 	case Chan, Map, Ptr, UnsafePointer:
-		if v.isIndirect() {
-			return *(*uintptr)(v.value)
-		}
-		return uintptr(v.value)
+		return uintptr(v.pointer())
 	case Slice:
 		slice := (*sliceHeader)(v.value)
 		return uintptr(slice.data)
@@ -146,6 +151,15 @@ func (v Value) Pointer() uintptr {
 	}
 }

+// pointer returns the underlying pointer represented by v.
+// v.Kind() must be Ptr, Map, Chan, or UnsafePointer
+func (v Value) pointer() unsafe.Pointer {
+	if v.isIndirect() {
+		return *(*unsafe.Pointer)(v.value)
+	}
+	return v.value
+}
+
 func (v Value) IsValid() bool {
 	return v.typecode != 0
 }
@@ -392,7 +406,14 @@ func (v Value) Elem() Value {
 			value:    ptr,
 			flags:    v.flags | valueFlagIndirect,
 		}
-	default: // not implemented: Interface
+	case Interface:
+		typecode, value := decomposeInterface(*(*interface{})(v.value))
+		return Value{
+			typecode: typecode,
+			value:    value,
+			flags:    v.flags &^ valueFlagIndirect,
+		}
+	default:
 		panic(&ValueError{"Elem"})
 	}
 }
@@ -513,11 +534,17 @@ func (v Value) Index(i int) Value {
 		if size > unsafe.Sizeof(uintptr(0)) {
 			// The element fits in a pointer, but the array does not.
 			// Load the value from the pointer.
-			addr := uintptr(v.value) + elemSize*uintptr(i) // pointer to new value
+			addr := unsafe.Pointer(uintptr(v.value) + elemSize*uintptr(i)) // pointer to new value
+			value := addr
+			if !v.isIndirect() {
+				// Use a pointer to the value (don't load the value) if the
+				// 'indirect' flag is set.
+				value = unsafe.Pointer(loadValue(addr, elemSize))
+			}
 			return Value{
 				typecode: v.typecode.elem(),
 				flags:    v.flags,
-				value:    unsafe.Pointer(loadValue(unsafe.Pointer(addr), elemSize)),
+				value:    value,
 			}
 		}

@@ -0,0 +1,22 @@
+.section .text.tinygo_scanCurrentStack,"ax"
+.global tinygo_scanCurrentStack
+tinygo_scanCurrentStack:
+    // Save callee-saved registers.
+    pushq %rbx
+    pushq %rbp
+    pushq %rdi
+    pushq %rsi
+    pushq %r12
+    pushq %r13
+    pushq %r14
+    pushq %r15
+
+    // Scan the stack.
+    subq $8, %rsp // adjust the stack before the call to maintain 16-byte alignment
+    movq %rsp, %rdi
+    callq tinygo_scanstack
+
+    // Restore the stack pointer. Registers do not need to be restored as they
+    // were only pushed to be discoverable by the GC.
+    addq $72, %rsp
+    retq
@@ -1,3 +1,4 @@
+//go:build gc.conservative
 // +build gc.conservative

 package runtime
@@ -0,0 +1,3 @@
+package runtime
+
+const GOOS = "windows"
@@ -273,7 +273,9 @@ func printspace() {
 }

 func printnl() {
-	putchar('\r')
+	if baremetal {
+		putchar('\r')
+	}
 	putchar('\n')
 }

@@ -35,7 +35,7 @@ func handleHardFault(sp *interruptStack) {
 	if fault.Mem().WhileUnstackingException() {
 		print(" while unstacking exception")
 	}
-	if fault.Mem().WhileStackingException() {
+	if fault.Mem().WileStackingException() {
 		print(" while stacking exception")
 	}
 	if fault.Mem().DuringFPLazyStatePres() {
@@ -161,13 +161,13 @@ func (fs MemFaultStatus) WhileUnstackingException() bool {
 	return fs&arm.SCB_CFSR_MUNSTKERR != 0
 }

-// WhileStackingException: stacking for an exception entry has caused one or more
+// WileStackingException: stacking for an exception entry has caused one or more
 // access violations
 //
 // "When this bit is 1, the SP is still adjusted but the values in the context
 // area on the stack might be incorrect. The processor has not written a fault
 // address to the MMAR."
-func (fs MemFaultStatus) WhileStackingException() bool {
+func (fs MemFaultStatus) WileStackingException() bool {
 	return fs&arm.SCB_CFSR_MSTKERR != 0
 }

@@ -6,7 +6,6 @@ import (
 	"device/arm"
 	"device/nxp"
 	"machine"
-	"machine/usb"
 	"math/bits"
 	"unsafe"
 )
@@ -37,7 +36,7 @@ func main() {
 	// initialize cache and MPU
 	initCache()

-	// enable SysTick, GPIO, USB, and other peripherals
+	// enable SysTick, GPIO, and peripherals
 	initPeripherals()

 	// reenable interrupts
@@ -109,8 +108,7 @@ func initPeripherals() {
 	initPins()        // configure GPIO

 	enablePeripheralClocks() // activate peripheral clock gates
-	initUSB()                // configure USB CDC-ACM (UART0)
-	initUART()               // configure hardware UART (UART1)
+	initUART()               // configure UART (initialized first for debugging)
 }

 func initPins() {
@@ -125,14 +123,8 @@ func initUART() {
 	machine.UART1.Configure(machine.UARTConfig{})
 }

-func initUSB() {
-	machine.HID0.Configure(usb.HIDConfig{})
-	// machine.UART0.Configure(usb.UARTConfig{})
-}
-
 func putchar(c byte) {
-	// machine.UART0.WriteByte(c) // print to USB UART
-	machine.UART1.WriteByte(c) // print to hardware UART
+	machine.UART1.WriteByte(c)
 }

 func exit(code int) {
@@ -35,31 +35,15 @@ var (
 		Denominator: 1, // 30-bit DENOM of fractional loop divider
 		Src:         0, // bypass clock source, 0=OSC24M, 1=CLK1_P & CLK1_N
 	}
-	Usb1PhyConfig = nxp.ClockConfigUsbPhy{
-		Instance:  1,        // USB PHY number (1 or 2)
-		XtalFreq:  OSC_FREQ, // External reference clock frequency (Hz)
-		DCal:      0xC,      // Decode to trim nominal 17.78mA current source
-		TxCal45DP: 0x6,      // Decode to trim nominal 45-Ohm series Rp on USB D+
-		TxCal45DM: 0x6,      // Decode to trim nominal 45-Ohm series Rp on USB D-
-		PllConfig: nxp.ClockConfigUsbPll{
-			Instance:    1,   // USB PLL number (1 or 2)
-			LoopDivider: 0,   // PLL loop divider (0 [Fout=Fref*20] or 1 [Fout=Fref*22])
-			Src:         0,   // PLL bypass clock source (0 [OSC24M] or 1 [CLK1_P & CLK1_N])
-			Pfd:         nil, // Phase fractional divisors (len=4, or nil for boot default)
-		},
+	Usb1PllConfig = nxp.ClockConfigUsbPll{
+		Instance:    1, // USB PLL instance
+		LoopDivider: 0, // PLL loop divider, Fout=Fin*20
+		Src:         0, // bypass clock source, 0=OSC24M, 1=CLK1_P & CLK1_N
 	}
-	Usb2PhyConfig = nxp.ClockConfigUsbPhy{
-		Instance:  2,        // USB PHY number (1 or 2)
-		XtalFreq:  OSC_FREQ, // External reference clock frequency (Hz)
-		DCal:      0xC,      // Decode to trim the nominal 17.78mA current source
-		TxCal45DP: 0x6,      // Decode to trim the nominal 45-Ohm series Rp on USB D+
-		TxCal45DM: 0x6,      // Decode to trim the nominal 45-Ohm series Rp on USB D-
-		PllConfig: nxp.ClockConfigUsbPll{
-			Instance:    2,   // USB PLL number (1 or 2)
-			LoopDivider: 0,   // PLL loop divider (0 [Fout=Fref*20] or 1 [Fout=Fref*22])
-			Src:         0,   // PLL bypass clock source (0 [OSC24M] or 1 [CLK1_P & CLK1_N])
-			Pfd:         nil, // Phase fractional divisors (len=4, or nil for boot default)
-		},
+	Usb2PllConfig = nxp.ClockConfigUsbPll{
+		Instance:    2, // USB PLL instance
+		LoopDivider: 0, // PLL loop divider, Fout=Fin*20
+		Src:         0, // bypass clock source, 0=OSC24M, 1=CLK1_P & CLK1_N
 	}
 )

@@ -101,7 +85,7 @@ func initClocks() {

 	// set VDD_SOC to 1.275V, necessary to config AHB to 600 MHz
 	nxp.DCDC.REG3.Set((nxp.DCDC.REG3.Get() & ^uint32(nxp.DCDC_REG3_TRG_Msk)) |
-		((0x13 << nxp.DCDC_REG3_TRG_Pos) & nxp.DCDC_REG3_TRG_Msk))
+		((13 << nxp.DCDC_REG3_TRG_Pos) & nxp.DCDC_REG3_TRG_Msk))

 	// wait until DCDC_STS_DC_OK bit is asserted
 	for !nxp.DCDC.REG0.HasBits(nxp.DCDC_REG0_STS_DC_OK_Msk) {
@@ -120,8 +104,6 @@ func initClocks() {
 	nxp.DivIpArm.Div(1)        // divide ARM_PODF (DIV2)
 	nxp.DivIpPeriphClk2.Div(0) // divide PERIPH_CLK2_PODF (DIV1)

-	nxp.ClockIpUsbOh3.Enable(false) // disable USB
-
 	nxp.ClockIpGpt1.Enable(false)  // disable GPT/PIT
 	nxp.ClockIpGpt1S.Enable(false) //
 	nxp.ClockIpGpt2.Enable(false)  //
@@ -130,11 +112,6 @@ func initClocks() {

 	nxp.DivIpPerclk.Div(0) // divide PERCLK_PODF (DIV1)

-	nxp.ClockIpGpio1.Enable(false) // disable GPIO
-	nxp.ClockIpGpio2.Enable(false) //
-	nxp.ClockIpGpio3.Enable(false) //
-	nxp.ClockIpGpio4.Enable(false) //
-
 	nxp.ClockIpUsdhc1.Enable(false) // disable USDHC1
 	nxp.DivIpUsdhc1.Div(1)          // divide USDHC1_PODF (DIV2)
 	nxp.MuxIpUsdhc1.Mux(1)          // USDHC1 select PLL2_PFD0
@@ -143,9 +120,9 @@ func initClocks() {
 	nxp.MuxIpUsdhc2.Mux(1)          // USDHC2 select PLL2_PFD0

 	nxp.ClockIpSemc.Enable(false) // disable SEMC
-	nxp.DivIpSemc.Div(7)          // divide SEMC_PODF (DIV8)
+	nxp.DivIpSemc.Div(1)          // divide SEMC_PODF (DIV2)
 	nxp.MuxIpSemcAlt.Mux(0)       // SEMC_ALT select PLL2_PFD2
-	nxp.MuxIpSemc.Mux(0)          // SEMC select PERIPH_CLK
+	nxp.MuxIpSemc.Mux(1)          // SEMC select SEMC_ALT

 	if false {
 		// TODO: external flash is on this bus, configured via DCD block
@@ -214,7 +191,7 @@ func initClocks() {
 	nxp.ClockIpLcdPixel.Enable(false) // disable LCDIF
 	nxp.DivIpLcdifPre.Div(1)          // divide LCDIF_PRED (DIV2)
 	nxp.DivIpLcdif.Div(3)             // divide LCDIF_CLK_PODF (DIV4)
-	nxp.MuxIpLcdifPre.Mux(4)          // LCDIF_PRE select PLL2_PFD1
+	nxp.MuxIpLcdifPre.Mux(5)          // LCDIF_PRE select PLL3_PFD1

 	nxp.ClockIpSpdif.Enable(false) // disable SPDIF
 	nxp.DivIpSpdif0Pre.Div(1)      // divide SPDIF0_CLK_PRED (DIV2)
@@ -232,43 +209,43 @@ func initClocks() {

 	nxp.MuxIpPll3Sw.Mux(0) // PLL3_SW select PLL3_MAIN

-	// Disable Audio/Video/Ethernet PLLs
-	nxp.CCM_ANALOG.PLL_AUDIO.Set(nxp.CCM_ANALOG_PLL_AUDIO_POWERDOWN_Msk)
-	nxp.CCM_ANALOG.PLL_VIDEO.Set(nxp.CCM_ANALOG_PLL_VIDEO_POWERDOWN_Msk)
-	nxp.CCM_ANALOG.PLL_ENET.Set(nxp.CCM_ANALOG_PLL_ENET_POWERDOWN_Msk)
-
 	ArmPllConfig.Configure() // init ARM PLL
 	// SYS PLL (PLL2) @ 528 MHz
 	//   PFD0 = 396    MHz -> USDHC1/USDHC2(DIV2)=198 MHz
 	//   PFD1 = 594    MHz -> (currently unused)
-	//   PFD2 = 327.72 MHz -> FlexSPI/FlexSPI2=327.72 MHz
-	//   PFD3 = 594    MHz -> (currently unused)
+	//   PFD2 = 327.72 MHz -> SEMC(DIV2)=163.86 MHz, FlexSPI/FlexSPI2=327.72 MHz
+	//   PFD3 = 454.73 MHz -> (currently unused)
 	SysPllConfig.Configure(24, 16, 29, 16) // init SYS PLL and PFDs

-	Usb1PhyConfig.Configure() // init USB1 HS PHY/PLL
-	Usb2PhyConfig.Configure() // init USB2 HS PHY/PLL
+	// USB1 PLL (PLL3) @ 480 MHz
+	//   PFD0 -> (currently unused)
+	//   PFD1 -> (currently unused)
+	//   PFD2 -> (currently unused)
+	//   PFD3 -> (currently unused)
+	Usb1PllConfig.Configure() // init USB1 PLL and PFDs
+	Usb2PllConfig.Configure() // init USB2 PLL

 	nxp.MuxIpPrePeriph.Mux(3)  // PRE_PERIPH select ARM_PLL
 	nxp.MuxIpPeriph.Mux(0)     // PERIPH select PRE_PERIPH
 	nxp.MuxIpPeriphClk2.Mux(1) // PERIPH_CLK2 select OSC
 	nxp.MuxIpPerclk.Mux(1)     // PERCLK select OSC

-	// set LVDS1 clock source (ARM_PLL)
+	// set LVDS1 clock source
 	nxp.CCM_ANALOG.MISC1.Set((nxp.CCM_ANALOG.MISC1.Get() & ^uint32(nxp.CCM_ANALOG_MISC1_LVDS1_CLK_SEL_Msk)) |
 		((0 << nxp.CCM_ANALOG_MISC1_LVDS1_CLK_SEL_Pos) & nxp.CCM_ANALOG_MISC1_LVDS1_CLK_SEL_Msk))

-	// set CLOCK_OUT1 divider (DIV1)
+	// set CLOCK_OUT1 divider
 	nxp.CCM.CCOSR.Set((nxp.CCM.CCOSR.Get() & ^uint32(nxp.CCM_CCOSR_CLKO1_DIV_Msk)) |
 		((0 << nxp.CCM_CCOSR_CLKO1_DIV_Pos) & nxp.CCM_CCOSR_CLKO1_DIV_Msk))
-	// set CLOCK_OUT1 source (PLL2/DIV2)
+	// set CLOCK_OUT1 source
 	nxp.CCM.CCOSR.Set((nxp.CCM.CCOSR.Get() & ^uint32(nxp.CCM_CCOSR_CLKO1_SEL_Msk)) |
 		((1 << nxp.CCM_CCOSR_CLKO1_SEL_Pos) & nxp.CCM_CCOSR_CLKO1_SEL_Msk))
-	// set CLOCK_OUT2 divider (DIV1)
+	// set CLOCK_OUT2 divider
 	nxp.CCM.CCOSR.Set((nxp.CCM.CCOSR.Get() & ^uint32(nxp.CCM_CCOSR_CLKO2_DIV_Msk)) |
 		((0 << nxp.CCM_CCOSR_CLKO2_DIV_Pos) & nxp.CCM_CCOSR_CLKO2_DIV_Msk))
-	// set CLOCK_OUT2 source (SPDIF0)
+	// set CLOCK_OUT2 source
 	nxp.CCM.CCOSR.Set((nxp.CCM.CCOSR.Get() & ^uint32(nxp.CCM_CCOSR_CLKO2_SEL_Msk)) |
-		((29 << nxp.CCM_CCOSR_CLKO2_SEL_Pos) & nxp.CCM_CCOSR_CLKO2_SEL_Msk))
+		((18 << nxp.CCM_CCOSR_CLKO2_SEL_Pos) & nxp.CCM_CCOSR_CLKO2_SEL_Msk))

 	nxp.CCM.CCOSR.ClearBits(nxp.CCM_CCOSR_CLK_OUT_SEL_Msk) // set CLK_OUT1 drives CLK_OUT
 	nxp.CCM.CCOSR.SetBits(nxp.CCM_CCOSR_CLKO1_EN_Msk)      // enable CLK_OUT1
@@ -276,10 +253,15 @@ func initClocks() {

 	nxp.ClockIpIomuxcGpr.Enable(false) // disable IOMUXC_GPR
 	nxp.ClockIpIomuxc.Enable(false)    // disable IOMUXC
-	// set GPT1 High frequency reference clock source (PERCLK)
+	// set GPT1 High frequency reference clock source
 	nxp.IOMUXC_GPR.GPR5.ClearBits(nxp.IOMUXC_GPR_GPR5_VREF_1M_CLK_GPT1_Msk)
-	// set GPT2 High frequency reference clock source (PERCLK)
+	// set GPT2 High frequency reference clock source
 	nxp.IOMUXC_GPR.GPR5.ClearBits(nxp.IOMUXC_GPR_GPR5_VREF_1M_CLK_GPT2_Msk)
+
+	nxp.ClockIpGpio1.Enable(false) // disable GPIO
+	nxp.ClockIpGpio2.Enable(false) //
+	nxp.ClockIpGpio3.Enable(false) //
+	nxp.ClockIpGpio4.Enable(false) //
 }

 func enableTimerClocks() {
@@ -141,7 +141,7 @@ func timerSleep(cycles uint32) bool {
 	nxp.PIT.TIMER[pitSleepTimer].TCTRL.Set(nxp.PIT_TIMER_TCTRL_TIE)     // enable interrupts
 	nxp.PIT.TIMER[pitSleepTimer].TCTRL.SetBits(nxp.PIT_TIMER_TCTRL_TEN) // start timer
 	for {
-		waitForEvents()
+		//arm.Asm("wfi") // TODO: causes hardfault! why?
 		if pitActive.Get() == 0 {
 			return true
 		}
@@ -1,3 +1,4 @@
+//go:build wasm && !wasi
 // +build wasm,!wasi

 package runtime
@@ -6,13 +7,19 @@ import "unsafe"

 type timeUnit float64 // time in milliseconds, just like Date.now() in JavaScript

+// wasmNested is used to detect scheduler nesting (WASM calls into JS calls back into WASM).
+// When this happens, we need to use a reduced version of the scheduler.
+var wasmNested bool
+
 //export _start
 func _start() {
 	// These need to be initialized early so that the heap can be initialized.
 	heapStart = uintptr(unsafe.Pointer(&heapStartSymbol))
 	heapEnd = uintptr(wasm_memory_size(0) * wasmPageSize)

+	wasmNested = true
 	run()
+	wasmNested = false
 }

 var handleEvent func()
@@ -27,12 +34,27 @@ func resume() {
 	go func() {
 		handleEvent()
 	}()
+
+	if wasmNested {
+		minSched()
+		return
+	}
+
+	wasmNested = true
 	scheduler()
+	wasmNested = false
 }

 //export go_scheduler
 func go_scheduler() {
+	if wasmNested {
+		minSched()
+		return
+	}
+
+	wasmNested = true
 	scheduler()
+	wasmNested = false
 }

 func ticksToNanoseconds(ticks timeUnit) int64 {
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Ayke van Laethem	22b5b69cf0	WIP alpine build	2021-11-16 21:57:51 +01:00
Ayke van Laethem	2081380b5c	ci: simplify build-linux job Split building the release and smoke-testing the release in two, and don't redo some tests that are already done by assert-test-linux. Some benefits: - Lower overall CI time because tests aren't done multiple times. - TinyHCI can run earlier because the build-linux job is finished as soon as the build artifact is ready. It does however have the downside of an extra job, which costs a few seconds to spin up and a few seconds to push and pull the workspace. But even with this, overall CI time is down by a few minutes per workflow run.	2021-11-16 18:43:32 +01:00
Ayke van Laethem	3d0c6dd02d	ci: simplify test-linux jobs Instead of doing lots of repetitive tests in test-llvm11-go115 and test-llvm11-go116, do those tests only once in assert-test-linux and only run smoke tests for older Go versions. Benefits: - This should reduce total CI time, because these jobs don't do tests that are done elsewere anyway. They only do the minimal work necessary to prove that the given Go/LLVM version works. - Doing all tests in assert-test-linux hopefully catches bugs that might not be found in regular LLVM builds.	2021-11-16 16:23:59 +01:00
Ayke van Laethem	869e917dc6	all: add support for windows/amd64 This uses Mingw-w64, which seems to be the de facto standard for porting Unixy programs to Windows.	2021-11-16 11:08:30 +01:00
Ayke van Laethem	41bcad9c19	runtime: only use CRLF on baremetal systems We should only do this on baremetal systems, not on Linux, macOS, and Windows. In fact, Windows will convert LF into CRLF in its putchar function.	2021-11-16 11:08:30 +01:00
Ayke van Laethem	73ab44c178	builder: support -size= flag on the esp32 This fixes a small mistake when calculating binary size for an Xtensa file. Previously it would exit with the following error: $ tinygo build -o test.elf -size=short -target=esp32-mini32 examples/serial panic: runtime error: index out of range [65521] with length 18 Now it runs as expected: $ tinygo build -o test.elf -size=short -target=esp32-mini32 examples/serial code data bss \| flash ram 2897 0 4136 \| 2897 4136	2021-11-16 07:42:58 +01:00
Dan Kegel	b70b6076e3	net/ip, syscall/errno: Reduce code duplication by switching to internal/itoa. internal/itoa wasn't around back in go 1.12 days when tinygo's syscall/errno.go was written. It was only added as of go 1.17 ( https://github.com/golang/go/commit/061a6903a232cb868780b ) so we have to have an internal copy for now. The internal copy should be deleted when tinygo drops support for go 1.16. FWIW, the new version seems nicer. It uses no allocations when converting 0, and although the optimizer might make this moot, uses a multiplication x 10 instead of a mod operation.	2021-11-16 02:13:52 +01:00
sago35	7b41d92198	device: add build tag for go1.17	2021-11-16 02:10:03 +01:00
Ayke van Laethem	2f4f3bd1ba	wasi: restore support for -scheduler=none The assembly symbols were not marked as hidden and so were exported, leading to unreferenced symbols. Example error message: Error: failed to run main module `/tmp/tinygo3961039405/main` Caused by: 0: failed to instantiate "/tmp/tinygo3961039405/main" 1: unknown import: `asyncify::stop_rewind` has not been defined This commit fixes this issue.	2021-11-15 22:17:40 +01:00
Ayke van Laethem	7cb44fb373	all: add support for GOARM This environment variable can be set to 5, 6, or 7 and controls which ARM version (ARMv5, ARMv6, ARMv7) is used when compiling for GOARCH=arm. I have picked the default value ARMv6, which I believe is supported on most common single board computers including all Raspberry Pis. The difference in code size is pretty big. We could even go further and support ARMv4 if anybody is interested. It should be pretty simple to add this if needed.	2021-11-15 11:53:44 +01:00
Ayke van Laethem	73ad825b67	ci: update Windows runner to windows-2019 The windows-2016 runner will soon be removed: https://github.blog/changelog/2021-10-19-github-actions-the-windows-2016-runner-image-will-be-removed-from-github-hosted-runners-on-march-15-2022/ Therefore, switch to the next version.	2021-11-14 13:18:54 +01:00
Nia Waldvogel	539e1324c8	fix binaryen build in docker Oops I forgot to install cmake and ninja.	2021-11-14 10:49:28 +01:00
Nia Waldvogel	641dcd7c16	internal/task: use asyncify on webassembly This change implements a new "scheduler" for WebAssembly using binaryen's asyncify transform. This is more reliable than the current "coroutines" transform, and works with non-Go code in the call stack. runtime (js/wasm): handle scheduler nesting If WASM calls into JS which calls back into WASM, it is possible for the scheduler to nest. The event from the callback must be handled immediately, so the task cannot simply be deferred to the outer scheduler. This creates a minimal scheduler loop which is used to handle such nesting.	2021-11-14 10:49:28 +01:00
Yurii Soldak	523d1f28cf	nano-33-ble: internal i2c config, power led and sensors switch	2021-11-13 12:46:54 +01:00
Damian Gryski	782c57cc11	pass testing arguments to wasmtime	2021-11-13 12:01:20 +01:00
Ayke van Laethem	c1d697f868	compiler: fix indices into strings and arrays This PR fixes two bugs at once: 1. Indices were incorrectly extended to a bigger type. Specifically, unsigned integers were sign extended and signed integers were zero extended. This commit swaps them around. 2. The getelementptr instruction was given the raw index, even if it was a uint8 for example. However, getelementptr assumes the indices are signed, and therefore an index of uint8(200) was interpreted as an index of int8(-56).	2021-11-13 11:04:24 +01:00
Ayke van Laethem	335fb71d2f	reflect: add support for DeepEqual The implementation has been mostly copied from the Go reference implementation with some small changes to fit TinyGo. Source: https://github.com/golang/go/blob/77a11c05d6a6f766c75f804ea9b8796f9a9f85a3/src/reflect/deepequal.go In addition, this commit also contains the following: - A set of tests copied from the Go reflect package. - An increased stack size for the riscv-qemu and hifive1-qemu targets (because they otherwise fail to run the tests). Because these targets are only used for testing, this seems fine to me.	2021-11-12 21:27:27 +01:00
Ayke van Laethem	5866a47e77	reflect: fix Value.Index() in a specific case In the case where: - Value.Index() was called on an array - that array was bigger than a pointer - the element type fits in a pointer - the 'indirect' flag isn't set the Value.Index() method would still (incorrectly) load the value. This commit fixes that. The next commit adds a test which would have triggered this bug so works as a regression test.	2021-11-12 21:27:27 +01:00
Ayke van Laethem	823c9c25cf	reflect: implement Value.Elem() for interface values	2021-11-12 21:27:27 +01:00
Ayke van Laethem	d15e32fb89	reflect: don't construct an interface-in-interface value v.Interaface() could construct an interface in interface value if v was of type interface. This is not correct, and doesn't follow upstream Go behavior. Instead, it should return the interface value itself.	2021-11-12 21:27:27 +01:00
soypat	b534dd67e0	machine/rp2040: add interrupt API	2021-11-12 10:38:02 +01:00
Ayke van Laethem	6c02b4956c	interp: fix reverting of extractvalue/insertvalue with multiple indices	2021-11-11 10:36:22 +01:00
Ayke van Laethem	1681ed02d3	interp: take care of constant globals Constant globals can't have been modified, even if a pointer is passed externally. Therefore, don't treat it as such in hasExternalStore. In addition, it doesn't make sense to update values of constant globals after the interp pass is finished. So don't do this. TODO: track whether objects are actually modified and only update the globals if this is the case.	2021-11-11 08:59:32 +01:00
Ayke van Laethem	7e68980c39	ci: improve caching for GitHub Actions Previously the cache would be stale for every new branch. With this change, PRs use the cache from the base branch and therefore don't need to rebuild LLVM from scratch.	2021-11-10 18:55:17 +01:00
Ayke van Laethem	cf640290a3	compiler: add "target-cpu" and "target-features" attributes This matches Clang, and with that, it adds support for inlining between Go and C because LLVM only allows inlining if the "target-cpu" and "target-features" string attributes match. For example, take a look at the following code: // int add(int a, int b) { // return a + b; // } import "C" func main() { println(C.add(3, 5)) } The 'add' function is not inlined into the main function before this commit, but after it, it can be inlined and trivially be optimized to `println(8)`.	2021-11-10 11:16:13 +01:00
Ayke van Laethem	78fec3719f	all: add target-features string to all targets This makes sure that the LLVM target features match the one generated by Clang: - This fixes a bug introduced when setting the target CPU for all targets: Cortex-M4 would now start using floating point operations while they were disabled in C. - This will make it possible in the future to inline C functions in Go and vice versa. This will need some more work though. There is a code size impact. Cortex-M4 targets are increased slightly in binary size while Cortex-M0 targets tend to be reduced a little bit. Other than that, there is little impact.	2021-11-07 09:26:46 +01:00
Ayke van Laethem	af4d0fe191	compileopts: fix reversed append in the target file With this fix, `cflags` in the target JSON files is correctly ordered. Previously, the cflags of a parent JSON file would come after the ones in the child JSON file, which makes it hard to override properties in the child JSON file. Specifically, this fixes the case where targets/riscv32.json sets `-march=rv32imac` and targets/esp32c3.json wants to override this using `-march=rv32imc` but can't do this because its `-march` comes before the riscv32.json one.	2021-11-07 09:26:46 +01:00