Merge pull request #442 from xushiwei/q

README: chan

.github/workflows/go.yml

@@ -11,19 +11,38 @@ on:
 
 jobs:
 
-  test-macos:
-    runs-on: macos-latest
+  test:
     strategy:
       matrix:
+        os: [macos-latest, ubuntu-latest]
         llvm: [17]
+    runs-on: ${{ matrix.os }}
     steps:
       - uses: actions/checkout@v4
 
       - name: Update Homebrew
-        if: matrix.llvm == 17 # needed as long as LLVM 17 is still fresh
+        # needed as long as LLVM 17 is still fresh
+        if: matrix.llvm == 17 && startsWith(matrix.os, 'macos')
         run: brew update
-      - name: Install LLVM ${{ matrix.llvm }}
-        run: HOMEBREW_NO_AUTO_UPDATE=1 brew install llvm@${{ matrix.llvm }}
+      - name: Install LLVM ${{ matrix.llvm }} and bdw-gc
+        if: startsWith(matrix.os, 'macos')
+        run: |
+          HOMEBREW_NO_AUTO_UPDATE=1 brew install llvm@${{ matrix.llvm }} bdw-gc
+          echo `brew --prefix llvm@${{ matrix.llvm }}`/bin >> $GITHUB_PATH
+      - name: Install LLVM ${{ matrix.llvm }} and libgc-dev
+        if: startsWith(matrix.os, 'ubuntu')
+        run: |
+          echo "deb http://apt.llvm.org/$(lsb_release -cs)/ llvm-toolchain-$(lsb_release -cs)-${{ matrix.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 -y llvm-${{ matrix.llvm }}-dev clang-${{ matrix.llvm }} lld-${{ matrix.llvm }} pkg-config libgc-dev libcjson-dev libsqlite3-dev python3.11-dev
+          echo /usr/lib/llvm-${{ matrix.llvm }}/bin >> $GITHUB_PATH
+    
+      - name: Clang information
+        run: |
+          echo $PATH
+          which clang
+          clang --version
         
       - name: Set up Go
         uses: actions/setup-go@v5
@@ -34,36 +53,40 @@ jobs:
         run: go build -v ./...
 
       - name: Test
+        if: matrix.os != 'macos-latest'
         run: go test -v ./...
 
-  test-linux:
-    runs-on: ubuntu-20.04
-    strategy:
-      matrix:
-        llvm: [17]
-    steps:
-    - uses: actions/checkout@v4
+      - name: Test with coverage
+        if: matrix.os == 'macos-latest'
+        run: go test -v -coverprofile="coverage.txt" -covermode=atomic ./...
 
-    - name: Install LLVM ${{ matrix.llvm }}
-      run: |
-        echo 'deb http://apt.llvm.org/focal/ llvm-toolchain-focal-${{ matrix.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-${{ matrix.llvm }}-dev
+      - name: Install
+        run: go install ./...
+      
+      - name: LLGO tests
+        if: matrix.os != 'ubuntu-latest'
+        run: |
+          echo "Test result on ${{ matrix.os }} with LLVM ${{ matrix.llvm }}" > result.md
+          LLGOROOT=$PWD bash .github/workflows/test_llgo.sh
+      
+      - name: Test _demo and _pydemo
+        run: |
+          set +e
+          LLGOROOT=$PWD bash .github/workflows/test_demo.sh
+          exit 0
+      
+      - name: Show test result
+        run: cat result.md
 
-    - name: Set up Go
-      uses: actions/setup-go@v5
-      with:
-        go-version: '1.20'
+      - name: PR comment with test result
+        uses: thollander/actions-comment-pull-request@v2
+        if: false
+        with:
+          filePath: result.md
+          comment_tag: test-result-on-${{ matrix.os }}-with-llvm-${{ matrix.llvm }}
 
-    - name: Build
-      run: go build -v ./...
-
-    - name: Test
-      run: go test -v -coverprofile="coverage.txt" -covermode=atomic ./...
-
-    - name: Upload coverage reports to Codecov
-      uses: codecov/codecov-action@v4
-      with:
-        token: ${{ secrets.CODECOV_TOKEN }}
-        slug: goplus/llgo
+      - name: Upload coverage reports to Codecov
+        uses: codecov/codecov-action@v4
+        with:
+          token: ${{ secrets.CODECOV_TOKEN }}
+          slug: goplus/llgo

.github/workflows/test_demo.sh

@@ -0,0 +1,29 @@
+#!/bin/bash
+
+# llgo run subdirectories under _demo and _pydemo
+total=0
+failed=0
+failed_cases=""
+for d in ./_demo/* ./_pydemo/*; do
+  total=$((total+1))
+  if [ -d "$d" ]; then
+    echo "Testing $d"
+    if ! llgo run -v "$d"; then
+      echo "FAIL"
+      failed=$((failed+1))
+      failed_cases="$failed_cases\n* :x: $d"
+    else
+      echo "PASS"
+    fi
+  fi
+done
+echo "=== Done"
+echo "$((total-failed))/$total tests passed"
+
+if [ "$failed" -ne 0 ]; then
+  echo ":bangbang: Failed demo cases:" | tee -a result.md
+  echo -e "$failed_cases" | tee -a result.md
+  exit 1
+else
+  echo ":white_check_mark: All demo tests passed" | tee -a result.md
+fi

.github/workflows/test_llgo.sh

@@ -0,0 +1,38 @@
+#!/bin/bash
+set -e
+
+export LLGOROOT=$PWD
+
+testcmd=/tmp/test
+llgo build -o $testcmd ./_test
+cases=$($testcmd)
+total=$(echo "$cases" | wc -l | tr -d ' ')
+failed=0
+failed_cases=""
+
+for idx in $(seq 1 $((total))); do
+  case=$(echo "$cases" | sed -n "${idx}p")
+  case_name=$(echo "$case" | cut -d',' -f2)
+  echo "=== Test case: $case_name"
+  set +e
+  out=$("$testcmd" "$((idx-1))" 2>&1)
+  exit_code=$?
+  set -e
+  if [ "${exit_code:-0}" -ne 0 ]; then
+    echo "failed: $out"
+    failed=$((failed+1))
+    failed_cases="$failed_cases\n* :x: $case_name"
+  else
+    echo "passed"
+  fi
+done
+echo "=== Done"
+echo "$((total-failed))/$total tests passed"
+
+if [ "$failed" -ne 0 ]; then
+  echo ":bangbang: Failed llgo cases:" | tee -a result.md
+  echo -e "$failed_cases" | tee -a result.md
+  exit 1
+else
+  echo ":white_check_mark: All llgo tests passed" | tee -a result.md
+fi

.gitignore

@@ -9,7 +9,8 @@
 *.dylib
 
 test.db
-llgo_autogen.ll
+demo.ll
+llgo_autogen*.ll
 stories*.bin
 .DS_Store
 err.log
@@ -18,6 +19,7 @@ numpy.txt
 _go/
 _runtime/
 _tinygo/
+_output/
 build.dir/
 .vscode/
 

README.md

@@ -34,7 +34,6 @@ The `_demo` directory contains some C standard libary related demos (it start wi
 To run these demos (If you haven't installed `llgo` yet, please refer to [How to install](#how-to-install)):
 
 ```sh
-export LLGOROOT=`pwd`
 cd <demo-directory>  # eg. cd _demo/hello
 llgo run .
 ```
@@ -60,6 +59,8 @@ And you can import any Python library into `llgo` through a program called `llpy
 * [pytorch](https://pkg.go.dev/github.com/goplus/llgo/py/torch)
 * [matplotlib](https://pkg.go.dev/github.com/goplus/llgo/py/matplotlib)
 
+Note: For third-party libraries (such as pandas and pytorch), you still need to install the library files.
+
 Here is an example using the Python `math` library:
 
 ```go
@@ -129,10 +130,10 @@ You can also specify the path to tell `llgo` where the Python library is located
 export LLGO_LIB_PYTHON=/foo/bar/python3.12
 ```
 
-For example, `/opt/homebrew/Frameworks/Python.framework/Versions/3.12/libpython3.12.dylib` is a typical python library location under macOS. So we should set it like this:
+For example, `/opt/homebrew/Frameworks/Python.framework/Versions/3.12/lib/libpython3.12.dylib` is a typical python library location under macOS. So we should set it like this:
 
 ```sh
-export LLGO_LIB_PYTHON=/opt/homebrew/Frameworks/Python.framework/Versions/3.12/python3.12
+export LLGO_LIB_PYTHON=/opt/homebrew/Frameworks/Python.framework/Versions/3.12/lib/python3.12
 ```
 
 Note that the file name must be written in a platform-independent format, using `python3.12` instead of `libpython3.12.dylib`.
@@ -140,7 +141,6 @@ Note that the file name must be written in a platform-independent format, using
 Then you can run the demos:
 
 ```sh
-export LLGOROOT=`pwd`
 cd <demo-directory>  # eg. cd _pydemo/callpy
 llgo run .
 ```
@@ -167,22 +167,33 @@ Here are some examples related to them:
 
 ## Go syntax support
 
-Common Go syntax is already supported. Except for the following, which needs to be improved:
+Most of the Go syntax is already supported. Except for the following, which needs to be improved:
 
 * map (Very limited support)
-* panic (Limited support)
-* recover (Not supported yet)
-* defer (Not supported yet)
-* gc (Not supported yet)
-* chan (Not supported yet)
-* goroutine (Not supported yet)
-* generics (Not supported yet)
 
 Here are some examples related to Go syntax:
 
 * [concat](_demo/concat/concat.go): define a variadic function
 * [genints](_demo/genints/genints.go): various forms of closure usage (including C function, recv.method and anonymous function)
-* [errors](_demo/errors/errors.go): demo to implement error interface
+* [errors](_cmptest/errors/errors.go): demo to implement error interface
+* [defer](_cmptest/defer/defer.go): defer demo
+* [goroutine](_demo/goroutine/goroutine.go): goroutine demo
+
+
+## Defer
+
+LLGo `defer` does not support usage in loops. This is not a bug but a feature, because we think that using `defer` in a loop is a very unrecommended practice.
+
+
+### Garbage Collection (GC)
+
+By default, LLGo implements `gc` based on [bdwgc](https://www.hboehm.info/gc/) (also known as [libgc](https://www.hboehm.info/gc/)).
+
+However, you can disable gc by specifying the `nogc` tag. For example:
+
+```sh
+llgo run -tags nogc .
+```
 
 
 ## Go packages support
@@ -193,8 +204,35 @@ Here are the Go packages that can be imported correctly:
 * [unicode](https://pkg.go.dev/unicode)
 * [unicode/utf8](https://pkg.go.dev/unicode/utf8)
 * [unicode/utf16](https://pkg.go.dev/unicode/utf16)
+* [math](https://pkg.go.dev/math)
+* [math/bits](https://pkg.go.dev/math/bits)
+* [math/cmplx](https://pkg.go.dev/math/cmplx)
+* [sort](https://pkg.go.dev/sort)
+* [strconv](https://pkg.go.dev/strconv)
+* [sync/atomic](https://pkg.go.dev/sync/atomic)
+* [sync](https://pkg.go.dev/sync) (partially)
+* [syscall](https://pkg.go.dev/syscall) (partially)
+* [errors](https://pkg.go.dev/errors) (partially)
+* [io](https://pkg.go.dev/io) (partially)
+* [io/fs](https://pkg.go.dev/io/fs) (partially)
+* [os](https://pkg.go.dev/os) (partially)
+* [fmt](https://pkg.go.dev/fmt) (partially)
+* [reflect](https://pkg.go.dev/reflect) (partially)
+* [time](https://pkg.go.dev/time) (partially)
 
 
+## Dependencies
+
+- [Go 1.20+](https://go.dev) (build only)
+- [LLVM 17](https://llvm.org)
+- [LLD 17](https://lld.llvm.org)
+- [Clang 17](https://clang.llvm.org)
+- [pkg-config 0.29+](https://www.freedesktop.org/wiki/Software/pkg-config/)
+- [bdwgc/libgc 8.0+](https://www.hboehm.info/gc/)
+- [cJSON 1.7+](https://github.com/DaveGamble/cJSON) (optional, for [`github.com/goplus/llgo/c/cjson`](https://pkg.go.dev/github.com/goplus/llgo/c/cjson))
+- [SQLite 3](https://www.sqlite.org) (optional, for [`github.com/goplus/llgo/c/sqlite`](https://pkg.go.dev/github.com/goplus/llgo/c/sqlite))
+- [Python 3.11+](https://www.python.org) (optional, for [`github.com/goplus/llgo/py`](https://pkg.go.dev/github.com/goplus/llgo/py))
+
 ## How to install
 
 Follow these steps to generate the `llgo` command (its usage is the same as the `go` command):
@@ -202,18 +240,30 @@ Follow these steps to generate the `llgo` command (its usage is the same as the
 ### on macOS
 
 ```sh
-brew update  # execute if needed
-brew install llvm@17
+brew update # execute if needed
+brew install llvm@17 pkg-config libgc
+brew install cjson sqlite python@3.12 # optional
+export PATH=$(brew --prefix llvm@17)/bin:$PATH # you may want to add this to your shell RC file, e.g. ~/.zshrc
+export CC=clang CXX=clang++ # only for go build; optional if you have other compatible compilers
+git clone https://github.com/goplus/llgo.git
+cd llgo
+export LLGOROOT="/path/to/llgo" # Replace this with the root directory of the llgo project
 go install -v ./...
 ```
 
-### on Linux
+### on Linux (Debian/Ubuntu)
 
 ```sh
-echo 'deb http://apt.llvm.org/focal/ llvm-toolchain-focal-17 main' | sudo tee /etc/apt/sources.list.d/llvm.list
+echo "deb http://apt.llvm.org/$(lsb_release -cs)/ llvm-toolchain-$(lsb_release -cs)-17 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  # execute if needed
-sudo apt-get install --no-install-recommends llvm-17-dev
+sudo apt-get update # execute if needed
+sudo apt-get install -y llvm-17-dev clang-17 lld-17 pkg-config libgc-dev
+sudo apt-get install -y libcjson-dev libsqlite3-dev python3.12-dev # optional
+export PATH=/usr/lib/llvm-17/bin:$PATH # you may want to add this to your shell RC file, e.g. ~/.bashrc
+export CC=clang CXX=clang++ # only for go build; optional if you have other compatible compilers
+git clone https://github.com/goplus/llgo.git
+cd llgo
+export LLGOROOT="/path/to/llgo" # Replace this with the root directory of the llgo project
 go install -v ./...
 ```
 
@@ -233,6 +283,7 @@ TODO
 How do I generate these tools?
 
 ```sh
+export CC=clang CXX=clang++ # only for go build; optional if you have other compatible compilers
 go install -v ./...  # compile all tools except pydump
 cd chore/_xtool
 llgo install ./...   # compile pydump

_cmptest/chan/chan.go

@@ -0,0 +1,11 @@
+package main
+
+func main() {
+	ch := make(chan int, 10)
+	println(len(ch), cap(ch))
+	go func() {
+		ch <- 100
+	}()
+	n, ok := <-ch
+	println(n, ok)
+}

_cmptest/defer/defer.go

@@ -0,0 +1,17 @@
+package main
+
+func f(s string) bool {
+	return len(s) > 2
+}
+
+func main() {
+	defer func() {
+		println("hi")
+	}()
+	if s := "hello"; f(s) {
+		defer println(s)
+	} else {
+		defer println("world")
+	}
+	defer println("bye")
+}

_cmptest/errors/errors.go

@@ -17,6 +17,5 @@ func (e *errorString) Error() string {
 
 func main() {
 	err := New("an error")
-	println(err)
 	println(err.Error())
 }

_cmptest/fmtdemo/fmt.go

@@ -0,0 +1,7 @@
+package main
+
+import "fmt"
+
+func main() {
+	fmt.Println("Hello, world")
+}

_cmptest/interf/interf.go

@@ -1,7 +1,7 @@
 package main
 
 import (
-	"github.com/goplus/llgo/_demo/interf/foo"
+	"github.com/goplus/llgo/_cmptest/interf/foo"
 )
 
 func Foo() any {

_cmptest/iodemo/io.go

@@ -0,0 +1,14 @@
+package main
+
+import (
+	"io"
+	"os"
+)
+
+func f(w io.Writer) {
+	w.Write([]byte("Hello, world\n"))
+}
+
+func main() {
+	f(os.Stdout)
+}

_cmptest/osdemo/osd.go

@@ -0,0 +1,18 @@
+package main
+
+import (
+	"os"
+)
+
+func main() {
+	home, _ := os.UserHomeDir()
+	println("home:", home)
+
+	cfgdir, _ := os.UserConfigDir()
+	println("cfgdir:", cfgdir)
+
+	cache, _ := os.UserCacheDir()
+	println("cachedir:", cache)
+
+	os.Stdout.Write([]byte("Hello, World\n"))
+}

_cmptest/reflect/reflect.go

@@ -0,0 +1,13 @@
+package main
+
+import "reflect"
+
+func main() {
+	tyIntSlice := reflect.SliceOf(reflect.TypeOf(0))
+	v := reflect.Zero(tyIntSlice)
+	v = reflect.Append(v, reflect.ValueOf(1), reflect.ValueOf(2), reflect.ValueOf(3))
+	for i, n := 0, v.Len(); i < n; i++ {
+		item := v.Index(i)
+		println(item.Int())
+	}
+}

_cmptest/rtype/rtype.go

@@ -0,0 +1,14 @@
+package main
+
+import "reflect"
+
+func main() {
+	tyIntSlice := reflect.SliceOf(reflect.TypeOf(0))
+	println(tyIntSlice.String())
+
+	v := reflect.Zero(tyIntSlice)
+	println(v.Len())
+
+	v = reflect.ValueOf(100)
+	println(v.Int())
+}

_cmptest/sortdemo/sort.go

@@ -0,0 +1,23 @@
+package main
+
+import "sort"
+
+func main() {
+	vals := []int{32, 58, 25, 92, 45, 78}
+	sort.Ints(vals)
+	for _, v := range vals {
+		println(v)
+	}
+
+	texts := []string{"apple", "banana", "cherry", "date", "elderberry", "fig"}
+	sort.Slice(texts, func(i, j int) bool {
+		leni, lenj := len(texts[i]), len(texts[j])
+		if leni != lenj {
+			return leni < lenj
+		}
+		return texts[i] < texts[j]
+	})
+	for _, v := range texts {
+		println(v)
+	}
+}

_cmptest/strconv/strconv.go

@@ -0,0 +1,7 @@
+package main
+
+import "strconv"
+
+func main() {
+	println(strconv.Itoa(-123))
+}

_cmptest/syscall/syscall.go

@@ -0,0 +1,11 @@
+package main
+
+import "syscall"
+
+func main() {
+	wd, err := syscall.Getwd()
+	if err != nil {
+		panic(err)
+	}
+	println("cwd:", wd)
+}

_cmptest/timedemo/time.go

@@ -0,0 +1,8 @@
+package main
+
+import "time"
+
+func main() {
+	t := time.Date(2018, time.January, 1, 2, 3, 4, 5, time.UTC)
+	println(t.String())
+}

_demo/catomic/atomic.go

@@ -0,0 +1,24 @@
+package main
+
+import (
+	"github.com/goplus/llgo/c/sync/atomic"
+)
+
+func main() {
+	var v int64
+
+	atomic.Store(&v, 100)
+	println("store:", atomic.Load(&v))
+
+	ret := atomic.Add(&v, 1)
+	println("ret:", ret, "v:", v)
+
+	ret, _ = atomic.CompareAndExchange(&v, 100, 102)
+	println("ret:", ret, "vs 100, v:", v)
+
+	ret, _ = atomic.CompareAndExchange(&v, 101, 102)
+	println("ret:", ret, "vs 101, v:", v)
+
+	ret = atomic.Sub(&v, 1)
+	println("ret:", ret, "v:", v)
+}

_demo/cchan/cchan.go

@@ -0,0 +1,49 @@
+package main
+
+import (
+	"unsafe"
+
+	"github.com/goplus/llgo/internal/runtime"
+)
+
+const (
+	eltSize = int(unsafe.Sizeof(0))
+)
+
+func doChan(cap int) {
+	c := runtime.NewChan(eltSize, cap)
+	go func() {
+		v := 100
+		runtime.ChanSend(c, unsafe.Pointer(&v), eltSize)
+	}()
+	var ret int
+	runtime.ChanRecv(c, unsafe.Pointer(&ret), eltSize)
+	println(ret)
+}
+
+func main() {
+	doChan(10)
+	doChan(0)
+
+	c := runtime.NewChan(eltSize, 3)
+
+	v := 1
+	runtime.ChanSend(c, unsafe.Pointer(&v), eltSize)
+	v = 2
+	runtime.ChanSend(c, unsafe.Pointer(&v), eltSize)
+	v = 3
+	runtime.ChanSend(c, unsafe.Pointer(&v), eltSize)
+	runtime.ChanClose(c)
+
+	v = 10
+	if runtime.ChanTrySend(c, unsafe.Pointer(&v), eltSize) {
+		println("error: chan send to closed chan")
+	}
+
+	for {
+		if ok := runtime.ChanRecv(c, unsafe.Pointer(&v), eltSize); !ok {
+			break
+		}
+		println(v)
+	}
+}

_demo/complex/cmplx.go

@@ -0,0 +1,9 @@
+package main
+
+import (
+	"math/cmplx"
+)
+
+func main() {
+	println("abs(3+4i):", cmplx.Abs(3+4i))
+}

_demo/cppintf/cppintf.go

@@ -0,0 +1,38 @@
+package main
+
+import (
+	"github.com/goplus/llgo/_demo/cppintf/foo"
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/math"
+)
+
+type Bar struct {
+	foo.Callback
+	a c.Int
+}
+
+func NewBar(a c.Int) *Bar {
+	return &Bar{
+		Callback: foo.Callback{
+			Vptr: &foo.CallbackVtbl{
+				Val:  c.Func((*Bar).getA),
+				Calc: c.Func((*Bar).sqrt),
+			},
+		},
+		a: a,
+	}
+}
+
+func (p *Bar) getA() c.Int {
+	return p.a
+}
+
+func (p *Bar) sqrt(v float64) float64 {
+	return math.Sqrt(v)
+}
+
+func main() {
+	bar := NewBar(1)
+	foo.F(&bar.Callback)
+	foo.G(&bar.Callback)
+}

_demo/cppintf/foo/bar/bar.cpp

@@ -0,0 +1,15 @@
+#include <stdio.h>
+#define interface struct
+
+interface ICallback {
+	virtual int val() = 0;
+	virtual double calc(double v) = 0;
+};
+
+extern "C" void f(ICallback* cb) {
+	printf("val: %d\ncalc(2): %lf\n", cb->val(), cb->calc(2));
+}
+
+void g(ICallback* cb) {
+	f(cb);
+}

_demo/cppintf/foo/foo.go

@@ -0,0 +1,25 @@
+package foo
+
+import (
+	"unsafe"
+)
+
+const (
+	LLGoFiles   = "bar/bar.cpp"
+	LLGoPackage = "link"
+)
+
+type Callback struct {
+	Vptr *CallbackVtbl
+}
+
+type CallbackVtbl struct {
+	Val  unsafe.Pointer
+	Calc unsafe.Pointer
+}
+
+//go:linkname F C.f
+func F(cb *Callback)
+
+//go:linkname G C._Z1gP9ICallback
+func G(cb *Callback)

_demo/cppmintf/cpp_multi_intf.go

@@ -0,0 +1,50 @@
+package main
+
+import (
+	"unsafe"
+
+	"github.com/goplus/llgo/_demo/cppmintf/foo"
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/math"
+)
+
+type Bar struct {
+	foo.Callback
+	a c.Int
+}
+
+func NewBar(a c.Int) *Bar {
+	return &Bar{
+		Callback: foo.Callback{
+			ICalc: foo.ICalc{
+				Vptr: &foo.ICalcVtbl{
+					Calc: c.Func((*Bar).sqrt),
+				},
+			},
+			IVal: foo.IVal{
+				Vptr: &foo.IValVtbl{
+					Val: c.Func(bar_IVal_getA),
+				},
+			},
+		},
+		a: a,
+	}
+}
+
+func (p *Bar) getA() c.Int {
+	return p.a
+}
+
+func bar_IVal_getA(this c.Pointer) c.Int {
+	const delta = -int(unsafe.Offsetof(foo.Callback{}.IVal))
+	return (*Bar)(c.Advance(this, delta)).getA()
+}
+
+func (p *Bar) sqrt(v float64) float64 {
+	return math.Sqrt(v)
+}
+
+func main() {
+	bar := NewBar(1)
+	foo.F(&bar.Callback)
+}

_demo/cppmintf/foo/bar/bar.cpp

@@ -0,0 +1,17 @@
+#include <stdio.h>
+#define interface struct
+
+interface ICalc {
+	virtual double calc(double v) = 0;
+};
+
+interface IVal {
+	virtual int val() = 0;
+};
+
+class Callback : public ICalc, public IVal {
+};
+
+extern "C" void f(Callback* cb) {
+	printf("val: %d\ncalc(2): %lf\n", cb->val(), cb->calc(2));
+}

_demo/cppmintf/foo/foo.go

@@ -0,0 +1,42 @@
+package foo
+
+import (
+	"unsafe"
+)
+
+const (
+	LLGoFiles   = "bar/bar.cpp"
+	LLGoPackage = "link"
+)
+
+// -----------------------------------------------------------------------------
+
+type ICalc struct {
+	Vptr *ICalcVtbl
+}
+
+type ICalcVtbl struct {
+	Calc unsafe.Pointer
+}
+
+// -----------------------------------------------------------------------------
+
+type IVal struct {
+	Vptr *IValVtbl
+}
+
+type IValVtbl struct {
+	Val unsafe.Pointer
+}
+
+// -----------------------------------------------------------------------------
+
+type Callback struct {
+	ICalc
+	IVal
+}
+
+//go:linkname F C.f
+func F(cb *Callback)
+
+// -----------------------------------------------------------------------------

_demo/ctime/time.go

@@ -0,0 +1,12 @@
+package main
+
+import "github.com/goplus/llgo/c/time"
+
+func main() {
+	var tv time.Timespec
+	time.ClockGettime(time.CLOCK_REALTIME, &tv)
+	println("REALTIME sec:", tv.Sec, "nsec:", tv.Nsec)
+
+	time.ClockGettime(time.CLOCK_MONOTONIC, &tv)
+	println("MONOTONIC sec:", tv.Sec, "nsec:", tv.Nsec)
+}

_demo/getcwd/getcwd.go

@@ -0,0 +1,11 @@
+package main
+
+import (
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/os"
+)
+
+func main() {
+	wd := os.Getcwd(c.Alloca(os.PATH_MAX), os.PATH_MAX)
+	c.Printf(c.Str("cwd: %s\n"), wd)
+}

_demo/goroutine/goroutine.go

@@ -0,0 +1,12 @@
+package main
+
+func main() {
+	done := false
+	go func() {
+		println("Hello, goroutine")
+		done = true
+	}()
+	for !done {
+		print(".")
+	}
+}

_demo/gotime/time.go

@@ -0,0 +1,7 @@
+package main
+
+import "time"
+
+func main() {
+	println(time.Now().String())
+}

_demo/math/math.go

@@ -0,0 +1,11 @@
+package main
+
+import (
+	"math"
+)
+
+func main() {
+	println(math.Sqrt(2))
+	println(math.Abs(-1.2))
+	println(math.Ldexp(1.2, 3))
+}

_demo/rand/rand.go

@@ -0,0 +1,15 @@
+package main
+
+import (
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/math/rand"
+)
+
+func main() {
+	var s c.Uint = 6
+	rand.Srand(s)
+	rr := rand.RandR(&s)
+	r := rand.Rand()
+	println("r:", r)
+	println("rr:", rr)
+}

_demo/setjmp/setjmp.go

@@ -0,0 +1,16 @@
+package main
+
+import (
+	"github.com/goplus/llgo/c/setjmp"
+)
+
+func main() {
+	var jb setjmp.SigjmpBuf
+	switch ret := setjmp.Sigsetjmp(&jb, 0); ret {
+	case 0:
+		println("Hello, setjmp!")
+		setjmp.Siglongjmp(&jb, 1)
+	default:
+		println("exception:", ret)
+	}
+}

_demo/thread/thd.go

@@ -0,0 +1,25 @@
+package main
+
+import (
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/pthread"
+)
+
+var key pthread.Key
+
+func main() {
+	key.Create(nil)
+	key.Set(c.Pointer(c.Str("main value\n")))
+
+	var thd pthread.Thread
+	pthread.Create(&thd, nil, func(arg c.Pointer) c.Pointer {
+		key.Set(c.Pointer(c.Str("thread value\n")))
+		c.Printf(c.Str("Hello, thread\nTLS: %s"), key.Get())
+		return c.Pointer(c.Str("Back to main\n"))
+	}, nil)
+
+	var retval c.Pointer
+	pthread.Join(thd, &retval)
+
+	c.Printf(c.Str("%sTLS: %s"), retval, key.Get())
+}

_test/bdwgc.go

@@ -0,0 +1,101 @@
+package main
+
+import (
+	"unsafe"
+
+	"github.com/goplus/llgo/_test/testing"
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/bdwgc"
+)
+
+// ------ Test malloc ------
+
+func TestMalloc(t *testing.T) {
+	pn := (*int)(bdwgc.Malloc(unsafe.Sizeof(int(0))))
+	*pn = 1 << 30
+	c.Printf(c.Str("value: %d, %x, %p, %p\n"), *pn, *pn, pn, &pn)
+
+	pl := (*int64)(bdwgc.Realloc(c.Pointer(pn), unsafe.Sizeof(int64(0))))
+	*pl = 1 << 60
+	c.Printf(c.Str("value: %lld, %llx, %p, %p\n"), *pl, *pl, pl, &pl)
+
+	bdwgc.Free(c.Pointer(pl))
+}
+
+// ------ Test finalizer ------
+
+const (
+	RETURN_VALUE_FREED = 1 << 31
+)
+
+var called uint = 0
+
+func setReturnValueFreed(pobj c.Pointer, clientData c.Pointer) {
+	called |= RETURN_VALUE_FREED
+	c.Printf(c.Str("called: %x\n"), called)
+}
+
+func setLoopValueFreed(pobj c.Pointer, clientData c.Pointer) {
+	pmask := (*uint)(clientData)
+	called |= *pmask
+	c.Printf(c.Str("called: %x\n"), called)
+}
+
+func isCalled(mask uint) bool {
+	return called&mask != 0
+}
+
+func returnValue() *int {
+	pn := bdwgc.Malloc(unsafe.Sizeof(int(0)))
+	bdwgc.RegisterFinalizer(pn, setReturnValueFreed, nil, nil, nil)
+	return (*int)(pn)
+}
+
+func callFunc() {
+	pn := returnValue()
+	*pn = 1 << 30
+	c.Printf(c.Str("value: %d, %x, %p, %p\n"), *pn, *pn, pn, &pn)
+	bdwgc.Gcollect()
+	check(!isCalled(RETURN_VALUE_FREED), c.Str("finalizer should not be called"))
+}
+
+func loop() {
+	for i := 0; i < 5; i++ {
+		p := bdwgc.Malloc(unsafe.Sizeof(int(0)))
+		pn := (*int)(p)
+		*pn = i
+		c.Printf(c.Str("value: %d, %x, %p, %p\n"), *pn, *pn, pn, &pn)
+		pflag := (*uint)(c.Malloc(unsafe.Sizeof(uint(0))))
+		*pflag = 1 << i
+		bdwgc.RegisterFinalizer(p, setLoopValueFreed, c.Pointer(pflag), nil, nil)
+		bdwgc.Gcollect()
+		check(!isCalled(1<<i), c.Str("finalizer should not be called"))
+		for j := 0; j < i; j++ {
+			check(isCalled(1<<j), c.Str("finalizers of previous objects should be called"))
+		}
+	}
+}
+
+// Clear stack to avoid reference
+func clearStack() {
+	p := c.Alloca(128)
+	c.Memset(p, 0, 128)
+}
+
+func check(b bool, msg *c.Char) {
+	if !b {
+		c.Printf(c.Str("check failed: %s\n"), msg)
+		panic("check failed")
+	}
+}
+
+func TestFinalizer(t *testing.T) {
+	bdwgc.Init()
+
+	callFunc()
+	clearStack()
+	bdwgc.Gcollect()
+	check(isCalled(RETURN_VALUE_FREED), c.Str("finalizer should be called"))
+
+	loop()
+}

_test/main.go

@@ -0,0 +1,31 @@
+package main
+
+import (
+	"github.com/goplus/llgo/_test/testing"
+	"github.com/goplus/llgo/c"
+)
+
+type TestCase struct {
+	Name string
+	F    func(*testing.T)
+}
+
+func main() {
+	tests := []TestCase{
+		{"TestMalloc", TestMalloc},
+		{"TestFinalizer", TestFinalizer},
+	}
+	if c.Argc == 1 {
+		for _, test := range tests {
+			c.Printf(c.Str("%s\n"), c.AllocaCStr(test.Name))
+		}
+		return
+	}
+	c.Fprintf(c.Stderr, c.Str("arg: %s\n"), c.Index(c.Argv, 1))
+	idx := int(c.Atoi(c.Index(c.Argv, 1)))
+	if idx < 0 || idx >= len(tests) {
+		c.Printf(c.Str("invalid test index %d"), idx)
+		panic("invalid test index")
+	}
+	tests[idx].F(nil)
+}

_test/testing/testing.go

@@ -0,0 +1,4 @@
+package testing
+
+type T struct {
+}

c/bdwgc/bdwgc.go

@@ -0,0 +1,103 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package bdwgc
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "link: $(pkg-config --libs bdw-gc); -lgc"
+)
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Init C.GC_init
+func Init()
+
+//go:linkname Malloc C.GC_malloc
+func Malloc(size uintptr) c.Pointer
+
+//go:linkname Realloc C.GC_realloc
+func Realloc(ptr c.Pointer, size uintptr) c.Pointer
+
+//go:linkname Free C.GC_free
+func Free(ptr c.Pointer)
+
+// -----------------------------------------------------------------------------
+
+//go:linkname RegisterFinalizer C.GC_register_finalizer
+func RegisterFinalizer(
+	obj c.Pointer,
+	fn func(c.Pointer, c.Pointer), cd c.Pointer,
+	oldFn *func(c.Pointer, c.Pointer), oldCd *c.Pointer)
+
+//go:linkname RegisterFinalizerNoOrder C.GC_register_finalizer_no_order
+func RegisterFinalizerNoOrder(
+	obj c.Pointer,
+	fn func(c.Pointer, c.Pointer), cd c.Pointer,
+	oldFn *func(c.Pointer, c.Pointer), oldCd *c.Pointer)
+
+//go:linkname RegisterFinalizerIgnoreSelf C.GC_register_finalizer_ignore_self
+func RegisterFinalizerIgnoreSelf(
+	obj c.Pointer,
+	fn func(c.Pointer, c.Pointer), cd c.Pointer,
+	oldFn *func(c.Pointer, c.Pointer), oldCd *c.Pointer)
+
+//go:linkname RegisterFinalizerUnreachable C.GC_register_finalizer_unreachable
+func RegisterFinalizerUnreachable(
+	obj c.Pointer,
+	fn func(c.Pointer, c.Pointer), cd c.Pointer,
+	oldFn *func(c.Pointer, c.Pointer), oldCd *c.Pointer)
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Enable C.GC_enable
+func Enable()
+
+//go:linkname Disable C.GC_disable
+func Disable()
+
+//go:linkname IsDisabled C.GC_is_disabled
+func IsDisabled() c.Int
+
+//go:linkname Gcollect C.GC_gcollect
+func Gcollect()
+
+//go:linkname GetMemoryUse C.GC_get_memory_use
+func GetMemoryUse() uintptr
+
+// -----------------------------------------------------------------------------
+
+//go:linkname EnableIncremental C.GC_enable_incremental
+func EnableIncremental()
+
+//go:linkname IsIncrementalMode C.GC_is_incremental_mode
+func IsIncrementalMode() c.Int
+
+//go:linkname IncrementalProtectionNeeds C.GC_incremental_protection_needs
+func IncrementalProtectionNeeds() c.Int
+
+//go:linkname StartIncrementalCollection C.GC_start_incremental_collection
+func StartIncrementalCollection()
+
+//go:linkname CollectALittle C.GC_collect_a_little
+func CollectALittle()
+
+// -----------------------------------------------------------------------------

c/bitcast/_cast/cast.c

@@ -0,0 +1,17 @@
+typedef union {
+    double d;
+    float  f;
+    long   v;
+} castUnion;
+
+double llgoToFloat64(long v) {
+    castUnion k;
+    k.v = v;
+    return k.d;
+}
+
+float llgoToFloat32(long v) {
+    castUnion k;
+    k.v = v;
+    return k.f;
+}

c/bitcast/bitcast.go

@@ -0,0 +1,30 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package bitcast
+
+import _ "unsafe"
+
+const (
+	LLGoFiles   = "_cast/cast.c"
+	LLGoPackage = "link"
+)
+
+//go:linkname ToFloat64 C.llgoToFloat64
+func ToFloat64(v uintptr) float64
+
+//go:linkname ToFloat32 C.llgoToFloat32
+func ToFloat32(v uintptr) float32

c/c.go

@@ -17,6 +17,9 @@
 package c
 
 // typedef unsigned int uint;
+// typedef unsigned long ulong;
+// typedef unsigned long long ulonglong;
+// typedef long long longlong;
 import "C"
 import "unsafe"
 
@@ -25,17 +28,22 @@ const (
 )
 
 type (
-	Char      = int8
-	Int       = C.int
-	Uint      = C.uint
-	Long      = int32
-	Ulong     = uint32
-	LongLong  = int64
-	UlongLong = uint64
-	Float     = float32
-	Double    = float64
-	Pointer   = unsafe.Pointer
-	FilePtr   = unsafe.Pointer
+	Char    = int8
+	Float   = float32
+	Double  = float64
+	Pointer = unsafe.Pointer
+	FilePtr = unsafe.Pointer
+)
+
+type (
+	Int  C.int
+	Uint C.uint
+
+	Long  C.long
+	Ulong C.ulong
+
+	LongLong  C.longlong
+	UlongLong C.ulonglong
 )
 
 type integer interface {
@@ -45,8 +53,11 @@ type integer interface {
 //go:linkname Str llgo.cstr
 func Str(string) *Char
 
+//go:linkname Func llgo.funcAddr
+func Func(any) Pointer
+
 // llgo:link Advance llgo.advance
-func Advance[PtrT any](ptr PtrT, offset int) PtrT { return ptr }
+func Advance[PtrT any, I integer](ptr PtrT, offset I) PtrT { return ptr }
 
 // llgo:link Index llgo.index
 func Index[T any, I integer](ptr *T, offset I) T { return *ptr }
@@ -57,27 +68,107 @@ func Alloca(size uintptr) Pointer
 //go:linkname AllocaCStr llgo.allocaCStr
 func AllocaCStr(s string) *Char
 
-//go:linkname Unreachable llgo.unreachable
-func Unreachable()
-
 //go:linkname Malloc C.malloc
 func Malloc(size uintptr) Pointer
 
+//go:linkname Free C.free
+func Free(ptr Pointer)
+
 //go:linkname Memcpy C.memcpy
 func Memcpy(dst, src Pointer, n uintptr) Pointer
 
+//go:linkname Memmove C.memmove
+func Memmove(dst, src Pointer, n uintptr) Pointer
+
 //go:linkname Memset C.memset
 func Memset(s Pointer, c Int, n uintptr) Pointer
 
+//go:linkname Memchr C.memchr
+func Memchr(s Pointer, c Int, n uintptr) Pointer
+
+//go:linkname Memcmp C.memcmp
+func Memcmp(s1, s2 Pointer, n uintptr) Int
+
 // -----------------------------------------------------------------------------
 
+//go:linkname Strlen C.strlen
+func Strlen(s *Char) uintptr
+
+//go:linkname Strcpy C.strcpy
+func Strcpy(dst, src *Char) *Char
+
+//go:linkname Strncpy C.strncpy
+func Strncpy(dst, src *Char, n uintptr) *Char
+
+//go:linkname Strcat C.strcat
+func Strcat(dst, src *Char) *Char
+
+//go:linkname Strncat C.strncat
+func Strncat(dst, src *Char, n uintptr) *Char
+
+//go:linkname Strcmp C.strcmp
+func Strcmp(s1, s2 *Char) Int
+
+//go:linkname Strncmp C.strncmp
+func Strncmp(s1, s2 *Char, n uintptr) Int
+
+//go:linkname Strchr C.strchr
+func Strchr(s *Char, c Int) *Char
+
+//go:linkname Strrchr C.strrchr
+func Strrchr(s *Char, c Int) *Char
+
+//go:linkname Strstr C.strstr
+func Strstr(s1, s2 *Char) *Char
+
+//go:linkname Strdup C.strdup
+func Strdup(s *Char) *Char
+
+//go:linkname Strndup C.strndup
+func Strndup(s *Char, n uintptr) *Char
+
+//go:linkname Strtok C.strtok
+func Strtok(s, delim *Char) *Char
+
+//go:linkname Strerror C.strerror
+func Strerror(errnum Int) *Char
+
+//go:linkname Sprintf C.sprintf
+func Sprintf(s *Char, format *Char, __llgo_va_list ...any) Int
+
+//go:linkname Snprintf C.snprintf
+func Snprintf(s *Char, n uintptr, format *Char, __llgo_va_list ...any) Int
+
+//go:linkname Vsnprintf C.vsnprintf
+func Vsnprintf(s *Char, n uintptr, format *Char, ap Pointer) Int
+
+// -----------------------------------------------------------------------------
+
+// GoString converts a C string to a Go string.
+// TODO(xsw): any => int
+//
+//go:linkname GoString llgo.string
+func GoString(cstr *Char, __llgo_va_list /* n */ ...any) string
+
 //go:linkname GoStringData llgo.stringData
 func GoStringData(string) *Char
 
+//go:linkname GoDeferData llgo.deferData
+func GoDeferData() Pointer
+
 // -----------------------------------------------------------------------------
 
-//go:linkname Remove C.remove
-func Remove(path *Char) Int
+//go:linkname AllocaSigjmpBuf llgo.sigjmpbuf
+func AllocaSigjmpBuf() Pointer
+
+//go:linkname Sigsetjmp llgo.sigsetjmp
+func Sigsetjmp(jb Pointer, savemask Int) Int
+
+//go:linkname Siglongjmp llgo.siglongjmp
+func Siglongjmp(jb Pointer, retval Int)
+
+//go:linkname Unreachable llgo.unreachable
+func Unreachable()
 
 // -----------------------------------------------------------------------------
 
@@ -94,14 +185,10 @@ func Qsort(base Pointer, count, elem uintptr, compar func(a, b Pointer) Int)
 
 // -----------------------------------------------------------------------------
 
-//go:linkname Stdin __stdinp
-var Stdin FilePtr
+//go:linkname Atoi C.atoi
+func Atoi(s *Char) Int
 
-//go:linkname Stdout __stdoutp
-var Stdout FilePtr
-
-//go:linkname Stderr __stderrp
-var Stderr FilePtr
+// -----------------------------------------------------------------------------
 
 //go:linkname Printf C.printf
 func Printf(format *Char, __llgo_va_list ...any) Int
@@ -115,11 +202,20 @@ func Fwrite(data Pointer, size, count uintptr, fp FilePtr) uintptr
 //go:linkname Fputc C.fputc
 func Fputc(c Int, fp FilePtr) Int
 
+//go:linkname Fputs C.fputs
+func Fputs(s *Char, fp FilePtr) Int
+
+//go:linkname Fflush C.fflush
+func Fflush(fp FilePtr) Int
+
 // -----------------------------------------------------------------------------
 
 //go:linkname Time C.time
 func Time(*int32) int32
 
+//go:linkname Usleep C.usleep
+func Usleep(useconds Uint) Int
+
 // -----------------------------------------------------------------------------
 
 type Option struct {

c/c_default.go

@@ -0,0 +1,31 @@
+//go:build !linux
+// +build !linux
+
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package c
+
+import _ "unsafe"
+
+//go:linkname Stdin __stdinp
+var Stdin FilePtr
+
+//go:linkname Stdout __stdoutp
+var Stdout FilePtr
+
+//go:linkname Stderr __stderrp
+var Stderr FilePtr

c/c_linux.go

@@ -0,0 +1,31 @@
+//go:build linux
+// +build linux
+
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package c
+
+import _ "unsafe"
+
+//go:linkname Stdin stdin
+var Stdin FilePtr
+
+//go:linkname Stdout stdout
+var Stdout FilePtr
+
+//go:linkname Stderr stderr
+var Stderr FilePtr

c/cjson/README.md

@@ -1,22 +1,18 @@
 LLGo wrapper of DaveGamble/cJSON
 =====
-[![Build Status](https://github.com/goplus/cjson/actions/workflows/go.yml/badge.svg)](https://github.com/goplus/cjson/actions/workflows/go.yml)
-[![GitHub release](https://img.shields.io/github/v/tag/goplus/cjson.svg?label=release)](https://github.com/goplus/cjson/releases)
-[![GoDoc](https://pkg.go.dev/badge/github.com/goplus/cjson.svg)](https://pkg.go.dev/github.com/goplus/cjson)
-[![Compiler](https://img.shields.io/badge/compiler-llgo-darkgreen.svg)](https://github.com/goplus/llgo)
-[![Language](https://img.shields.io/badge/language-Go+-blue.svg)](https://github.com/goplus/gop)
 
 ## How to install
 
+### on macOS (Homebrew)
+
 ```sh
-git clone https://github.com/goplus/cjson.git
-cd cjson
-git submodule init
-git submodule update
-mkdir build.dir
-cd build.dir
-cmake ../cJSON
-sudo make install
+brew install cjson 
+```
+
+### on Linux (Debian/Ubuntu)
+
+```sh
+apt-get install -y libcjson-dev
 ```
 
 ## Demos

c/cjson/cjson.go

@@ -23,7 +23,7 @@ import (
 )
 
 const (
-	LLGoPackage = "link: cjson"
+	LLGoPackage = "link: $(pkg-config --libs libcjson); -lcjson"
 )
 
 // llgo:type C

c/clang/_demo/castdump/astdump.go

@@ -0,0 +1,49 @@
+package main
+
+import (
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/clang"
+)
+
+func visit(cursor, parent clang.Cursor, clientData c.Pointer) clang.ChildVisitResult {
+	depth := *(*c.Uint)(clientData)
+	printAST(cursor, depth+1)
+	return clang.ChildVisit_Continue
+}
+
+func printAST(cursor clang.Cursor, depth c.Uint) {
+	cursorKind := cursor.Kind.String()
+	cursorSpelling := cursor.String()
+
+	for i := c.Uint(0); i < depth; i++ {
+		c.Fputs(c.Str("  "), c.Stdout)
+	}
+
+	c.Printf(c.Str("%s: %s\n"), cursorKind.CStr(), cursorSpelling.CStr())
+
+	cursorKind.Dispose()
+	cursorSpelling.Dispose()
+
+	clang.VisitChildren(cursor, visit, c.Pointer(&depth))
+}
+
+func main() {
+	index := clang.CreateIndex(0, 0)
+	unit := index.ParseTranslationUnit(
+		c.Str("todo"),
+		nil, 0,
+		nil, 0,
+		clang.TranslationUnit_None,
+	)
+
+	if unit == nil {
+		println("Unable to parse translation unit. Quitting.")
+		c.Exit(1)
+	}
+
+	cursor := unit.Cursor()
+	printAST(cursor, 0)
+
+	unit.Dispose()
+	index.Dispose()
+}

c/clang/basic.go

@@ -0,0 +1,59 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package clang
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+/**
+ * A character string.
+ *
+ * The \c CXString type is used to return strings from the interface when
+ * the ownership of that string might differ from one call to the next.
+ * Use \c clang_getCString() to retrieve the string data and, once finished
+ * with the string data, call \c clang_disposeString() to free the string.
+ */
+type String struct {
+	Data         c.Pointer
+	PrivateFlags c.Uint
+}
+
+/**
+ * Retrieve the character data associated with the given string.
+ */
+// llgo:link C.clang_getCString
+func (String) CStr() *c.Char { return nil }
+
+/**
+ * Free the given string.
+ */
+// llgo:link C.clang_disposeString
+func (String) Dispose() {}
+
+type StringSet struct {
+	Strings *String
+	Count   c.Uint
+}
+
+/**
+ * Free the given string set.
+ */
+// llgo:link C.clang_disposeStringSet
+func (*StringSet) Dispose() {}

c/clang/clang.go

@@ -0,0 +1,266 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package clang
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "link: -L$(llvm-config --libdir) -lclang; -lclang"
+)
+
+/**
+ * Opaque pointer representing client data that will be passed through
+ * to various callbacks and visitors.
+ */
+type ClientData = c.Pointer
+
+/**
+ * Provides the contents of a file that has not yet been saved to disk.
+ *
+ * Each CXUnsavedFile instance provides the name of a file on the
+ * system along with the current contents of that file that have not
+ * yet been saved to disk.
+ */
+type UnsavedFile struct {
+	/**
+	 * The file whose contents have not yet been saved.
+	 *
+	 * This file must already exist in the file system.
+	 */
+	Filename *c.Char
+
+	/**
+	 * A buffer containing the unsaved contents of this file.
+	 */
+	Contents *c.Char
+
+	/**
+	 * The length of the unsaved contents of this buffer.
+	 */
+	Length c.Ulong
+}
+
+/**
+ * An "index" that consists of a set of translation units that would
+ * typically be linked together into an executable or library.
+ */
+type Index struct {
+	Unused [0]byte
+}
+
+/**
+ * Provides a shared context for creating translation units.
+ *
+ * It provides two options:
+ *
+ * - excludeDeclarationsFromPCH: When non-zero, allows enumeration of "local"
+ * declarations (when loading any new translation units). A "local" declaration
+ * is one that belongs in the translation unit itself and not in a precompiled
+ * header that was used by the translation unit. If zero, all declarations
+ * will be enumerated.
+ *
+ * Here is an example:
+ *
+ * \code
+ *   // excludeDeclsFromPCH = 1, displayDiagnostics=1
+ *   Idx = clang_createIndex(1, 1);
+ *
+ *   // IndexTest.pch was produced with the following command:
+ *   // "clang -x c IndexTest.h -emit-ast -o IndexTest.pch"
+ *   TU = clang_createTranslationUnit(Idx, "IndexTest.pch");
+ *
+ *   // This will load all the symbols from 'IndexTest.pch'
+ *   clang_visitChildren(clang_getTranslationUnitCursor(TU),
+ *                       TranslationUnitVisitor, 0);
+ *   clang_disposeTranslationUnit(TU);
+ *
+ *   // This will load all the symbols from 'IndexTest.c', excluding symbols
+ *   // from 'IndexTest.pch'.
+ *   char *args[] = { "-Xclang", "-include-pch=IndexTest.pch" };
+ *   TU = clang_createTranslationUnitFromSourceFile(Idx, "IndexTest.c", 2, args,
+ *                                                  0, 0);
+ *   clang_visitChildren(clang_getTranslationUnitCursor(TU),
+ *                       TranslationUnitVisitor, 0);
+ *   clang_disposeTranslationUnit(TU);
+ * \endcode
+ *
+ * This process of creating the 'pch', loading it separately, and using it (via
+ * -include-pch) allows 'excludeDeclsFromPCH' to remove redundant callbacks
+ * (which gives the indexer the same performance benefit as the compiler).
+ */
+//go:linkname CreateIndex C.clang_createIndex
+func CreateIndex(excludeDeclarationsFromPCH, displayDiagnostics c.Int) *Index
+
+/**
+ * Destroy the given index.
+ *
+ * The index must not be destroyed until all of the translation units created
+ * within that index have been destroyed.
+ */
+// llgo:link (*Index).Dispose C.clang_disposeIndex
+func (*Index) Dispose() {}
+
+/**
+ * Flags that control the creation of translation units.
+ *
+ * The enumerators in this enumeration type are meant to be bitwise
+ * ORed together to specify which options should be used when
+ * constructing the translation unit.
+ */
+const (
+	TranslationUnit_None = 0x0
+)
+
+/**
+ * Same as \c clang_parseTranslationUnit2, but returns
+ * the \c CXTranslationUnit instead of an error code.  In case of an error this
+ * routine returns a \c NULL \c CXTranslationUnit, without further detailed
+ * error codes.
+ */
+// llgo:link (*Index).ParseTranslationUnit C.clang_parseTranslationUnit
+func (*Index) ParseTranslationUnit(
+	sourceFilename *c.Char, commandLineArgs **c.Char, numCommandLineArgs c.Int,
+	unsavedFiles *UnsavedFile, numUnsavedFiles c.Uint, options c.Uint) *TranslationUnit {
+	return nil
+}
+
+/**
+ * A single translation unit, which resides in an index.
+ */
+type TranslationUnit struct {
+	Unused [0]byte
+}
+
+/**
+ * Destroy the specified CXTranslationUnit object.
+ */
+// llgo:linke (*TranslationUnit).Dispose C.clang_disposeTranslationUnit
+func (*TranslationUnit) Dispose() {}
+
+/**
+ * Retrieve the cursor that represents the given translation unit.
+ *
+ * The translation unit cursor can be used to start traversing the
+ * various declarations within the given translation unit.
+ */
+// llgo:link (*TranslationUnit).Cursor C.clang_getTranslationUnitCursor
+func (*TranslationUnit) Cursor() (ret Cursor) {
+	return
+}
+
+/**
+ * Describes the kind of entity that a cursor refers to.
+ */
+type CursorKind c.Int
+
+/* for debug/testing */
+// llgo:link (CursorKind).String C.clang_getCursorKindSpelling
+func (CursorKind) String() (ret String) {
+	return
+}
+
+/**
+ * A cursor representing some element in the abstract syntax tree for
+ * a translation unit.
+ *
+ * The cursor abstraction unifies the different kinds of entities in a
+ * program--declaration, statements, expressions, references to declarations,
+ * etc.--under a single "cursor" abstraction with a common set of operations.
+ * Common operation for a cursor include: getting the physical location in
+ * a source file where the cursor points, getting the name associated with a
+ * cursor, and retrieving cursors for any child nodes of a particular cursor.
+ *
+ * Cursors can be produced in two specific ways.
+ * clang_getTranslationUnitCursor() produces a cursor for a translation unit,
+ * from which one can use clang_visitChildren() to explore the rest of the
+ * translation unit. clang_getCursor() maps from a physical source location
+ * to the entity that resides at that location, allowing one to map from the
+ * source code into the AST.
+ */
+type Cursor struct {
+	Kind  CursorKind
+	Xdata c.Int
+	Data  [3]c.Pointer
+}
+
+/**
+ * Retrieve a name for the entity referenced by this cursor.
+ */
+// llgo:link C.clang_getCursorSpelling
+func (Cursor) String() (ret String) {
+	return
+}
+
+/**
+ * Describes how the traversal of the children of a particular
+ * cursor should proceed after visiting a particular child cursor.
+ *
+ * A value of this enumeration type should be returned by each
+ * \c CXCursorVisitor to indicate how clang_visitChildren() proceed.
+ */
+type ChildVisitResult c.Int
+
+const (
+	/**
+	 * Terminates the cursor traversal.
+	 */
+	ChildVisit_Break ChildVisitResult = iota
+	/**
+	 * Continues the cursor traversal with the next sibling of
+	 * the cursor just visited, without visiting its children.
+	 */
+	ChildVisit_Continue
+	/**
+	 * Recursively traverse the children of this cursor, using
+	 * the same visitor and client data.
+	 */
+	ChildVisit_Recurse
+)
+
+/**
+ * Visit the children of a particular cursor.
+ *
+ * This function visits all the direct children of the given cursor,
+ * invoking the given \p visitor function with the cursors of each
+ * visited child. The traversal may be recursive, if the visitor returns
+ * \c CXChildVisit_Recurse. The traversal may also be ended prematurely, if
+ * the visitor returns \c CXChildVisit_Break.
+ *
+ * \param parent the cursor whose child may be visited. All kinds of
+ * cursors can be visited, including invalid cursors (which, by
+ * definition, have no children).
+ *
+ * \param visitor the visitor function that will be invoked for each
+ * child of \p parent.
+ *
+ * \param client_data pointer data supplied by the client, which will
+ * be passed to the visitor each time it is invoked.
+ *
+ * \returns a non-zero value if the traversal was terminated
+ * prematurely by the visitor returning \c CXChildVisit_Break.
+ */
+//go:linkname VisitChildren C.clang_visitChildren
+func VisitChildren(
+	cusor Cursor,
+	visitor func(cursor, parent Cursor, clientData ClientData) ChildVisitResult,
+	clientData ClientData) c.Uint {
+	return 0
+}

c/llama2/README.md

@@ -0,0 +1,10 @@
+LLGo wrapper of karpathy/llama2.c
+=====
+
+## How to update source to llgo
+
+```
+git submodule init
+git submodule update
+cp llama2.c/run.c llama2/run.c
+```

c/llama2/llama2.go

@@ -23,7 +23,8 @@ import (
 )
 
 const (
-	LLGoPackage = "link"
+	LLGoFiles   = "llama2/llama2.c"
+	LLGoPackage = "link: -lm"
 )
 
 // -----------------------------------------------------------------------------

c/llama2/llama2/llama2.c

@@ -0,0 +1,2 @@
+#define TESTING
+#include "./run.c"

c/llama2/llama2/run.c

@@ -1,2 +1,973 @@
-#define TESTING
-#include "../llama2.c/run.c"
+/* Inference for Llama-2 Transformer model in pure C */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <ctype.h>
+#include <time.h>
+#include <math.h>
+#include <string.h>
+#include <fcntl.h>
+#if defined _WIN32
+    #include "win.h"
+#else
+    #include <unistd.h>
+    #include <sys/mman.h>
+#endif
+// ----------------------------------------------------------------------------
+// Transformer model
+
+typedef struct {
+    int dim; // transformer dimension
+    int hidden_dim; // for ffn layers
+    int n_layers; // number of layers
+    int n_heads; // number of query heads
+    int n_kv_heads; // number of key/value heads (can be < query heads because of multiquery)
+    int vocab_size; // vocabulary size, usually 256 (byte-level)
+    int seq_len; // max sequence length
+} Config;
+
+typedef struct {
+    // token embedding table
+    float* token_embedding_table;    // (vocab_size, dim)
+    // weights for rmsnorms
+    float* rms_att_weight; // (layer, dim) rmsnorm weights
+    float* rms_ffn_weight; // (layer, dim)
+    // weights for matmuls. note dim == n_heads * head_size
+    float* wq; // (layer, dim, n_heads * head_size)
+    float* wk; // (layer, dim, n_kv_heads * head_size)
+    float* wv; // (layer, dim, n_kv_heads * head_size)
+    float* wo; // (layer, n_heads * head_size, dim)
+    // weights for ffn
+    float* w1; // (layer, hidden_dim, dim)
+    float* w2; // (layer, dim, hidden_dim)
+    float* w3; // (layer, hidden_dim, dim)
+    // final rmsnorm
+    float* rms_final_weight; // (dim,)
+    // (optional) classifier weights for the logits, on the last layer
+    float* wcls;
+} TransformerWeights;
+
+typedef struct {
+    // current wave of activations
+    float *x; // activation at current time stamp (dim,)
+    float *xb; // same, but inside a residual branch (dim,)
+    float *xb2; // an additional buffer just for convenience (dim,)
+    float *hb; // buffer for hidden dimension in the ffn (hidden_dim,)
+    float *hb2; // buffer for hidden dimension in the ffn (hidden_dim,)
+    float *q; // query (dim,)
+    float *k; // key (dim,)
+    float *v; // value (dim,)
+    float *att; // buffer for scores/attention values (n_heads, seq_len)
+    float *logits; // output logits
+    // kv cache
+    float* key_cache;   // (layer, seq_len, dim)
+    float* value_cache; // (layer, seq_len, dim)
+} RunState;
+
+typedef struct {
+    Config config; // the hyperparameters of the architecture (the blueprint)
+    TransformerWeights weights; // the weights of the model
+    RunState state; // buffers for the "wave" of activations in the forward pass
+    // some more state needed to properly clean up the memory mapping (sigh)
+    int fd; // file descriptor for memory mapping
+    float* data; // memory mapped data pointer
+    ssize_t file_size; // size of the checkpoint file in bytes
+} Transformer;
+
+void malloc_run_state(RunState* s, Config* p) {
+    // we calloc instead of malloc to keep valgrind happy
+    int kv_dim = (p->dim * p->n_kv_heads) / p->n_heads;
+    s->x = calloc(p->dim, sizeof(float));
+    s->xb = calloc(p->dim, sizeof(float));
+    s->xb2 = calloc(p->dim, sizeof(float));
+    s->hb = calloc(p->hidden_dim, sizeof(float));
+    s->hb2 = calloc(p->hidden_dim, sizeof(float));
+    s->q = calloc(p->dim, sizeof(float));
+    s->key_cache = calloc(p->n_layers * p->seq_len * kv_dim, sizeof(float));
+    s->value_cache = calloc(p->n_layers * p->seq_len * kv_dim, sizeof(float));
+    s->att = calloc(p->n_heads * p->seq_len, sizeof(float));
+    s->logits = calloc(p->vocab_size, sizeof(float));
+    // ensure all mallocs went fine
+    if (!s->x || !s->xb || !s->xb2 || !s->hb || !s->hb2 || !s->q
+     || !s->key_cache || !s->value_cache || !s->att || !s->logits) {
+        fprintf(stderr, "malloc failed!\n");
+        exit(EXIT_FAILURE);
+    }
+}
+
+void free_run_state(RunState* s) {
+    free(s->x);
+    free(s->xb);
+    free(s->xb2);
+    free(s->hb);
+    free(s->hb2);
+    free(s->q);
+    free(s->att);
+    free(s->logits);
+    free(s->key_cache);
+    free(s->value_cache);
+}
+
+void memory_map_weights(TransformerWeights *w, Config* p, float* ptr, int shared_weights) {
+    int head_size = p->dim / p->n_heads;
+    // make sure the multiplications below are done in 64bit to fit the parameter counts of 13B+ models
+    unsigned long long n_layers = p->n_layers;
+    w->token_embedding_table = ptr;
+    ptr += p->vocab_size * p->dim;
+    w->rms_att_weight = ptr;
+    ptr += n_layers * p->dim;
+    w->wq = ptr;
+    ptr += n_layers * p->dim * (p->n_heads * head_size);
+    w->wk = ptr;
+    ptr += n_layers * p->dim * (p->n_kv_heads * head_size);
+    w->wv = ptr;
+    ptr += n_layers * p->dim * (p->n_kv_heads * head_size);
+    w->wo = ptr;
+    ptr += n_layers * (p->n_heads * head_size) * p->dim;
+    w->rms_ffn_weight = ptr;
+    ptr += n_layers * p->dim;
+    w->w1 = ptr;
+    ptr += n_layers * p->dim * p->hidden_dim;
+    w->w2 = ptr;
+    ptr += n_layers * p->hidden_dim * p->dim;
+    w->w3 = ptr;
+    ptr += n_layers * p->dim * p->hidden_dim;
+    w->rms_final_weight = ptr;
+    ptr += p->dim;
+    ptr += p->seq_len * head_size / 2; // skip what used to be freq_cis_real (for RoPE)
+    ptr += p->seq_len * head_size / 2; // skip what used to be freq_cis_imag (for RoPE)
+    w->wcls = shared_weights ? w->token_embedding_table : ptr;
+}
+
+void read_checkpoint(char* checkpoint, Config* config, TransformerWeights* weights,
+                     int* fd, float** data, ssize_t* file_size) {
+    FILE *file = fopen(checkpoint, "rb");
+    if (!file) { fprintf(stderr, "Couldn't open file %s\n", checkpoint); exit(EXIT_FAILURE); }
+    // read in the config header
+    if (fread(config, sizeof(Config), 1, file) != 1) { exit(EXIT_FAILURE); }
+    // negative vocab size is hacky way of signaling unshared weights. bit yikes.
+    int shared_weights = config->vocab_size > 0 ? 1 : 0;
+    config->vocab_size = abs(config->vocab_size);
+    // figure out the file size
+    fseek(file, 0, SEEK_END); // move file pointer to end of file
+    *file_size = ftell(file); // get the file size, in bytes
+    fclose(file);
+    // memory map the Transformer weights into the data pointer
+    *fd = open(checkpoint, O_RDONLY); // open in read only mode
+    if (*fd == -1) { fprintf(stderr, "open failed!\n"); exit(EXIT_FAILURE); }
+    *data = mmap(NULL, *file_size, PROT_READ, MAP_PRIVATE, *fd, 0);
+    if (*data == MAP_FAILED) { fprintf(stderr, "mmap failed!\n"); exit(EXIT_FAILURE); }
+    float* weights_ptr = *data + sizeof(Config)/sizeof(float);
+    memory_map_weights(weights, config, weights_ptr, shared_weights);
+}
+
+void build_transformer(Transformer *t, char* checkpoint_path) {
+    // read in the Config and the Weights from the checkpoint
+    read_checkpoint(checkpoint_path, &t->config, &t->weights, &t->fd, &t->data, &t->file_size);
+    // allocate the RunState buffers
+    malloc_run_state(&t->state, &t->config);
+}
+
+void free_transformer(Transformer* t) {
+    // close the memory mapping
+    if (t->data != MAP_FAILED) { munmap(t->data, t->file_size); }
+    if (t->fd != -1) { close(t->fd); }
+    // free the RunState buffers
+    free_run_state(&t->state);
+}
+
+// ----------------------------------------------------------------------------
+// neural net blocks; the dynamics of the Transformer
+
+void rmsnorm(float* o, float* x, float* weight, int size) {
+    // calculate sum of squares
+    float ss = 0.0f;
+    for (int j = 0; j < size; j++) {
+        ss += x[j] * x[j];
+    }
+    ss /= size;
+    ss += 1e-5f;
+    ss = 1.0f / sqrtf(ss);
+    // normalize and scale
+    for (int j = 0; j < size; j++) {
+        o[j] = weight[j] * (ss * x[j]);
+    }
+}
+
+void softmax(float* x, int size) {
+    // find max value (for numerical stability)
+    float max_val = x[0];
+    for (int i = 1; i < size; i++) {
+        if (x[i] > max_val) {
+            max_val = x[i];
+        }
+    }
+    // exp and sum
+    float sum = 0.0f;
+    for (int i = 0; i < size; i++) {
+        x[i] = expf(x[i] - max_val);
+        sum += x[i];
+    }
+    // normalize
+    for (int i = 0; i < size; i++) {
+        x[i] /= sum;
+    }
+}
+
+void matmul(float* xout, float* x, float* w, int n, int d) {
+    // W (d,n) @ x (n,) -> xout (d,)
+    // by far the most amount of time is spent inside this little function
+    int i;
+    #pragma omp parallel for private(i)
+    for (i = 0; i < d; i++) {
+        float val = 0.0f;
+        for (int j = 0; j < n; j++) {
+            val += w[i * n + j] * x[j];
+        }
+        xout[i] = val;
+    }
+}
+
+float* forward(Transformer* transformer, int token, int pos) {
+
+    // a few convenience variables
+    Config* p = &transformer->config;
+    TransformerWeights* w = &transformer->weights;
+    RunState* s = &transformer->state;
+    float *x = s->x;
+    int dim = p->dim;
+    int kv_dim = (p->dim * p->n_kv_heads) / p->n_heads;
+    int kv_mul = p->n_heads / p->n_kv_heads; // integer multiplier of the kv sharing in multiquery
+    int hidden_dim =  p->hidden_dim;
+    int head_size = dim / p->n_heads;
+
+    // copy the token embedding into x
+    float* content_row = w->token_embedding_table + token * dim;
+    memcpy(x, content_row, dim*sizeof(*x));
+
+    // forward all the layers
+    for(unsigned long long l = 0; l < p->n_layers; l++) {
+
+        // attention rmsnorm
+        rmsnorm(s->xb, x, w->rms_att_weight + l*dim, dim);
+
+        // key and value point to the kv cache
+        int loff = l * p->seq_len * kv_dim; // kv cache layer offset for convenience
+        s->k = s->key_cache + loff + pos * kv_dim;
+        s->v = s->value_cache + loff + pos * kv_dim;
+
+        // qkv matmuls for this position
+        matmul(s->q, s->xb, w->wq + l*dim*dim, dim, dim);
+        matmul(s->k, s->xb, w->wk + l*dim*kv_dim, dim, kv_dim);
+        matmul(s->v, s->xb, w->wv + l*dim*kv_dim, dim, kv_dim);
+
+        // RoPE relative positional encoding: complex-valued rotate q and k in each head
+        for (int i = 0; i < dim; i+=2) {
+            int head_dim = i % head_size;
+            float freq = 1.0f / powf(10000.0f, head_dim / (float)head_size);
+            float val = pos * freq;
+            float fcr = cosf(val);
+            float fci = sinf(val);
+            int rotn = i < kv_dim ? 2 : 1; // how many vectors? 2 = q & k, 1 = q only
+            for (int v = 0; v < rotn; v++) {
+                float* vec = v == 0 ? s->q : s->k; // the vector to rotate (query or key)
+                float v0 = vec[i];
+                float v1 = vec[i+1];
+                vec[i]   = v0 * fcr - v1 * fci;
+                vec[i+1] = v0 * fci + v1 * fcr;
+            }
+        }
+
+        // multihead attention. iterate over all heads
+        int h;
+        #pragma omp parallel for private(h)
+        for (h = 0; h < p->n_heads; h++) {
+            // get the query vector for this head
+            float* q = s->q + h * head_size;
+            // attention scores for this head
+            float* att = s->att + h * p->seq_len;
+            // iterate over all timesteps, including the current one
+            for (int t = 0; t <= pos; t++) {
+                // get the key vector for this head and at this timestep
+                float* k = s->key_cache + loff + t * kv_dim + (h / kv_mul) * head_size;
+                // calculate the attention score as the dot product of q and k
+                float score = 0.0f;
+                for (int i = 0; i < head_size; i++) {
+                    score += q[i] * k[i];
+                }
+                score /= sqrtf(head_size);
+                // save the score to the attention buffer
+                att[t] = score;
+            }
+
+            // softmax the scores to get attention weights, from 0..pos inclusively
+            softmax(att, pos + 1);
+
+            // weighted sum of the values, store back into xb
+            float* xb = s->xb + h * head_size;
+            memset(xb, 0, head_size * sizeof(float));
+            for (int t = 0; t <= pos; t++) {
+                // get the value vector for this head and at this timestep
+                float* v = s->value_cache + loff + t * kv_dim + (h / kv_mul) * head_size;
+                // get the attention weight for this timestep
+                float a = att[t];
+                // accumulate the weighted value into xb
+                for (int i = 0; i < head_size; i++) {
+                    xb[i] += a * v[i];
+                }
+            }
+        }
+
+        // final matmul to get the output of the attention
+        matmul(s->xb2, s->xb, w->wo + l*dim*dim, dim, dim);
+
+        // residual connection back into x
+        for (int i = 0; i < dim; i++) {
+            x[i] += s->xb2[i];
+        }
+
+        // ffn rmsnorm
+        rmsnorm(s->xb, x, w->rms_ffn_weight + l*dim, dim);
+
+        // Now for FFN in PyTorch we have: self.w2(F.silu(self.w1(x)) * self.w3(x))
+        // first calculate self.w1(x) and self.w3(x)
+        matmul(s->hb, s->xb, w->w1 + l*dim*hidden_dim, dim, hidden_dim);
+        matmul(s->hb2, s->xb, w->w3 + l*dim*hidden_dim, dim, hidden_dim);
+
+        // SwiGLU non-linearity
+        for (int i = 0; i < hidden_dim; i++) {
+            float val = s->hb[i];
+            // silu(x)=x*σ(x), where σ(x) is the logistic sigmoid
+            val *= (1.0f / (1.0f + expf(-val)));
+            // elementwise multiply with w3(x)
+            val *= s->hb2[i];
+            s->hb[i] = val;
+        }
+
+        // final matmul to get the output of the ffn
+        matmul(s->xb, s->hb, w->w2 + l*dim*hidden_dim, hidden_dim, dim);
+
+        // residual connection
+        for (int i = 0; i < dim; i++) {
+            x[i] += s->xb[i];
+        }
+    }
+
+    // final rmsnorm
+    rmsnorm(x, x, w->rms_final_weight, dim);
+
+    // classifier into logits
+    matmul(s->logits, x, w->wcls, p->dim, p->vocab_size);
+    return s->logits;
+}
+
+// ----------------------------------------------------------------------------
+// The Byte Pair Encoding (BPE) Tokenizer that translates strings <-> tokens
+
+typedef struct {
+    char *str;
+    int id;
+} TokenIndex;
+
+typedef struct {
+    char** vocab;
+    float* vocab_scores;
+    TokenIndex *sorted_vocab;
+    int vocab_size;
+    unsigned int max_token_length;
+    unsigned char byte_pieces[512]; // stores all single-byte strings
+} Tokenizer;
+
+int compare_tokens(const void *a, const void *b) {
+    return strcmp(((TokenIndex*)a)->str, ((TokenIndex*)b)->str);
+}
+
+void build_tokenizer(Tokenizer* t, char* tokenizer_path, int vocab_size) {
+    // i should have written the vocab_size into the tokenizer file... sigh
+    t->vocab_size = vocab_size;
+    // malloc space to hold the scores and the strings
+    t->vocab = (char**)malloc(vocab_size * sizeof(char*));
+    t->vocab_scores = (float*)malloc(vocab_size * sizeof(float));
+    t->sorted_vocab = NULL; // initialized lazily
+    for (int i = 0; i < 256; i++) {
+        t->byte_pieces[i * 2] = (unsigned char)i;
+        t->byte_pieces[i * 2 + 1] = '\0';
+    }
+    // read in the file
+    FILE *file = fopen(tokenizer_path, "rb");
+    if (!file) { fprintf(stderr, "couldn't load %s\n", tokenizer_path); exit(EXIT_FAILURE); }
+    if (fread(&t->max_token_length, sizeof(int), 1, file) != 1) { fprintf(stderr, "failed read\n"); exit(EXIT_FAILURE); }
+    int len;
+    for (int i = 0; i < vocab_size; i++) {
+        if (fread(t->vocab_scores + i, sizeof(float), 1, file) != 1) { fprintf(stderr, "failed read\n"); exit(EXIT_FAILURE);}
+        if (fread(&len, sizeof(int), 1, file) != 1) { fprintf(stderr, "failed read\n"); exit(EXIT_FAILURE); }
+        t->vocab[i] = (char *)malloc(len + 1);
+        if (fread(t->vocab[i], len, 1, file) != 1) { fprintf(stderr, "failed read\n"); exit(EXIT_FAILURE); }
+        t->vocab[i][len] = '\0'; // add the string terminating token
+    }
+    fclose(file);
+}
+
+void free_tokenizer(Tokenizer* t) {
+    for (int i = 0; i < t->vocab_size; i++) { free(t->vocab[i]); }
+    free(t->vocab);
+    free(t->vocab_scores);
+    free(t->sorted_vocab);
+}
+
+char* decode(Tokenizer* t, int prev_token, int token) {
+    char *piece = t->vocab[token];
+    // following BOS (1) token, sentencepiece decoder strips any leading whitespace (see PR #89)
+    if (prev_token == 1 && piece[0] == ' ') { piece++; }
+    // careful, some tokens designate raw bytes, and look like e.g. '<0x01>'
+    // parse this and convert and return the actual byte
+    unsigned char byte_val;
+    if (sscanf(piece, "<0x%02hhX>", &byte_val) == 1) {
+        piece = (char*)t->byte_pieces + byte_val * 2;
+    }
+    return piece;
+}
+
+void safe_printf(char *piece) {
+    // piece might be a raw byte token, and we only want to print printable chars or whitespace
+    // because some of the other bytes can be various control codes, backspace, etc.
+    if (piece == NULL) { return; }
+    if (piece[0] == '\0') { return; }
+    if (piece[1] == '\0') {
+        unsigned char byte_val = piece[0];
+        if (!(isprint(byte_val) || isspace(byte_val))) {
+            return; // bad byte, don't print it
+        }
+    }
+    printf("%s", piece);
+}
+
+int str_lookup(char *str, TokenIndex *sorted_vocab, int vocab_size) {
+    // efficiently find the perfect match for str in vocab, return its index or -1 if not found
+    TokenIndex tok = { .str = str }; // acts as the key to search for
+    TokenIndex *res = bsearch(&tok, sorted_vocab, vocab_size, sizeof(TokenIndex), compare_tokens);
+    return res != NULL ? res->id : -1;
+}
+
+void encode(Tokenizer* t, char *text, int8_t bos, int8_t eos, int *tokens, int *n_tokens) {
+    // encode the string text (input) into an upper-bound preallocated tokens[] array
+    // bos != 0 means prepend the BOS token (=1), eos != 0 means append the EOS token (=2)
+    if (text == NULL) { fprintf(stderr, "cannot encode NULL text\n"); exit(EXIT_FAILURE); }
+
+    if (t->sorted_vocab == NULL) {
+        // lazily malloc and sort the vocabulary
+        t->sorted_vocab = malloc(t->vocab_size * sizeof(TokenIndex));
+        for (int i = 0; i < t->vocab_size; i++) {
+            t->sorted_vocab[i].str = t->vocab[i];
+            t->sorted_vocab[i].id = i;
+        }
+        qsort(t->sorted_vocab, t->vocab_size, sizeof(TokenIndex), compare_tokens);
+    }
+
+    // create a temporary buffer that will store merge candidates of always two consecutive tokens
+    // *2 for concat, +1 for null terminator +2 for UTF8 (in case max_token_length is 1)
+    char* str_buffer = malloc((t->max_token_length*2 +1 +2) * sizeof(char));
+    size_t str_len = 0;
+
+    // start at 0 tokens
+    *n_tokens = 0;
+
+    // add optional BOS (=1) token, if desired
+    if (bos) tokens[(*n_tokens)++] = 1;
+
+    // add_dummy_prefix is true by default
+    // so prepend a dummy prefix token to the input string, but only if text != ""
+    // TODO: pretty sure this isn't correct in the general case but I don't have the
+    // energy to read more of the sentencepiece code to figure out what it's doing
+    if (text[0] != '\0') {
+        int dummy_prefix = str_lookup(" ", t->sorted_vocab, t->vocab_size);
+        tokens[(*n_tokens)++] = dummy_prefix;
+    }
+
+    // Okay UTF-8 time. This will get messy. Here is the reference from Wikipedia:
+    // Code point ↔ UTF-8 conversion
+    // First code point	Last code point	Byte 1	Byte 2	Byte 3	Byte 4
+    // U+0000	U+007F	    0xxxxxxx
+    // U+0080	U+07FF	    110xxxxx	10xxxxxx
+    // U+0800	U+FFFF	    1110xxxx	10xxxxxx	10xxxxxx
+    // U+10000	U+10FFFF    11110xxx	10xxxxxx	10xxxxxx	10xxxxxx
+
+    // process the raw (UTF-8) byte sequence of the input string
+    for (char *c = text; *c != '\0'; c++) {
+
+        // reset buffer if the current byte is ASCII or a leading byte
+        // 0xC0 is 11000000, so (*c & 0xC0) keeps the first 2 bits and zeros the rest
+        // 0x80 is 10000000
+        // in UTF-8, all continuation bytes start with "10" in first two bits
+        // so in English this is: "if this byte is not a continuation byte"
+        if ((*c & 0xC0) != 0x80) {
+            // this byte must be either a leading byte (11...) or an ASCII char (0x...)
+            // => reset our location, as we're starting a new UTF-8 codepoint
+            str_len = 0;
+        }
+
+        // append the current byte to the buffer
+        str_buffer[str_len++] = *c; // ++ is post-increment, incremented after this line
+        str_buffer[str_len] = '\0';
+
+        // while the next character is a continuation byte, continue appending
+        // but if there are too many of them, just stop to avoid overruning str_buffer size.
+        if ((*(c+1) & 0xC0) == 0x80 && str_len < 4) {
+            continue;
+        }
+
+        // ok c+1 is not a continuation byte, so we've read in a full codepoint
+        int id = str_lookup(str_buffer, t->sorted_vocab, t->vocab_size);
+
+        if (id != -1) {
+            // we found this codepoint in vocab, add it as a token
+            tokens[(*n_tokens)++] = id;
+        } else {
+            // byte_fallback encoding: just encode each byte as a token
+            // +3 is here because the first 3 vocab elements are <unk>, <s>, </s>
+            // so the individual bytes only start at index 3
+            for (int i=0; i < str_len; i++) {
+                tokens[(*n_tokens)++] = (unsigned char)str_buffer[i] + 3;
+            }
+        }
+        str_len = 0; // protect against a sequence of stray UTF8 continuation bytes
+    }
+
+    // merge the best consecutive pair each iteration, according the scores in vocab_scores
+    while (1) {
+        float best_score = -1e10;
+        int best_id = -1;
+        int best_idx = -1;
+
+        for (int i=0; i < (*n_tokens-1); i++) {
+            // check if we can merge the pair (tokens[i], tokens[i+1])
+            sprintf(str_buffer, "%s%s", t->vocab[tokens[i]], t->vocab[tokens[i+1]]);
+            int id = str_lookup(str_buffer, t->sorted_vocab, t->vocab_size);
+            if (id != -1 && t->vocab_scores[id] > best_score) {
+                // this merge pair exists in vocab! record its score and position
+                best_score = t->vocab_scores[id];
+                best_id = id;
+                best_idx = i;
+            }
+        }
+
+        if (best_idx == -1) {
+            break; // we couldn't find any more pairs to merge, so we're done
+        }
+
+        // merge the consecutive pair (best_idx, best_idx+1) into new token best_id
+        tokens[best_idx] = best_id;
+        // delete token at position best_idx+1, shift the entire sequence back 1
+        for (int i = best_idx+1; i < (*n_tokens-1); i++) {
+            tokens[i] = tokens[i+1];
+        }
+        (*n_tokens)--; // token length decreased
+    }
+
+    // add optional EOS (=2) token, if desired
+    if (eos) tokens[(*n_tokens)++] = 2;
+
+    free(str_buffer);
+}
+
+// ----------------------------------------------------------------------------
+// The Sampler, which takes logits and returns a sampled token
+// sampling can be done in a few ways: greedy argmax, sampling, top-p sampling
+
+typedef struct {
+    float prob;
+    int index;
+} ProbIndex; // struct used when sorting probabilities during top-p sampling
+
+typedef struct {
+    int vocab_size;
+    ProbIndex* probindex; // buffer used in top-p sampling
+    float temperature;
+    float topp;
+    unsigned long long rng_state;
+} Sampler;
+
+int sample_argmax(float* probabilities, int n) {
+    // return the index that has the highest probability
+    int max_i = 0;
+    float max_p = probabilities[0];
+    for (int i = 1; i < n; i++) {
+        if (probabilities[i] > max_p) {
+            max_i = i;
+            max_p = probabilities[i];
+        }
+    }
+    return max_i;
+}
+
+int sample_mult(float* probabilities, int n, float coin) {
+    // sample index from probabilities (they must sum to 1!)
+    // coin is a random number in [0, 1), usually from random_f32()
+    float cdf = 0.0f;
+    for (int i = 0; i < n; i++) {
+        cdf += probabilities[i];
+        if (coin < cdf) {
+            return i;
+        }
+    }
+    return n - 1; // in case of rounding errors
+}
+
+int compare(const void* a, const void* b) {
+    ProbIndex* a_ = (ProbIndex*) a;
+    ProbIndex* b_ = (ProbIndex*) b;
+    if (a_->prob > b_->prob) return -1;
+    if (a_->prob < b_->prob) return 1;
+    return 0;
+}
+
+int sample_topp(float* probabilities, int n, float topp, ProbIndex* probindex, float coin) {
+    // top-p sampling (or "nucleus sampling") samples from the smallest set of
+    // tokens that exceed probability topp. This way we never sample tokens that
+    // have very low probabilities and are less likely to go "off the rails".
+    // coin is a random number in [0, 1), usually from random_f32()
+
+    int n0 = 0;
+    // quicksort indices in descending order of probabilities
+    // values smaller than (1 - topp) / (n - 1) cannot be part of the result
+    // so for efficiency we crop these out as candidates before sorting
+    const float cutoff = (1.0f - topp) / (n - 1);
+    for (int i = 0; i < n; i++) {
+        if (probabilities[i] >= cutoff) {
+            probindex[n0].index = i;
+            probindex[n0].prob = probabilities[i];
+            n0++;
+        }
+    }
+    qsort(probindex, n0, sizeof(ProbIndex), compare);
+
+    // truncate the list where cumulative probability exceeds topp
+    float cumulative_prob = 0.0f;
+    int last_idx = n0 - 1; // in case of rounding errors consider all elements
+    for (int i = 0; i < n0; i++) {
+        cumulative_prob += probindex[i].prob;
+        if (cumulative_prob > topp) {
+            last_idx = i;
+            break; // we've exceeded topp by including last_idx
+        }
+    }
+
+    // sample from the truncated list
+    float r = coin * cumulative_prob;
+    float cdf = 0.0f;
+    for (int i = 0; i <= last_idx; i++) {
+        cdf += probindex[i].prob;
+        if (r < cdf) {
+            return probindex[i].index;
+        }
+    }
+    return probindex[last_idx].index; // in case of rounding errors
+}
+
+void build_sampler(Sampler* sampler, int vocab_size, float temperature, float topp, unsigned long long rng_seed) {
+    sampler->vocab_size = vocab_size;
+    sampler->temperature = temperature;
+    sampler->topp = topp;
+    sampler->rng_state = rng_seed;
+    // buffer only used with nucleus sampling; may not need but it's ~small
+    sampler->probindex = malloc(sampler->vocab_size * sizeof(ProbIndex));
+}
+
+void free_sampler(Sampler* sampler) {
+    free(sampler->probindex);
+}
+
+unsigned int random_u32(unsigned long long *state) {
+    // xorshift rng: https://en.wikipedia.org/wiki/Xorshift#xorshift.2A
+    *state ^= *state >> 12;
+    *state ^= *state << 25;
+    *state ^= *state >> 27;
+    return (*state * 0x2545F4914F6CDD1Dull) >> 32;
+}
+float random_f32(unsigned long long *state) { // random float32 in [0,1)
+    return (random_u32(state) >> 8) / 16777216.0f;
+}
+
+int sample(Sampler* sampler, float* logits) {
+    // sample the token given the logits and some hyperparameters
+    int next;
+    if (sampler->temperature == 0.0f) {
+        // greedy argmax sampling: take the token with the highest probability
+        next = sample_argmax(logits, sampler->vocab_size);
+    } else {
+        // apply the temperature to the logits
+        for (int q=0; q<sampler->vocab_size; q++) { logits[q] /= sampler->temperature; }
+        // apply softmax to the logits to get the probabilities for next token
+        softmax(logits, sampler->vocab_size);
+        // flip a (float) coin (this is our source of entropy for sampling)
+        float coin = random_f32(&sampler->rng_state);
+        // we sample from this distribution to get the next token
+        if (sampler->topp <= 0 || sampler->topp >= 1) {
+            // simply sample from the predicted probability distribution
+            next = sample_mult(logits, sampler->vocab_size, coin);
+        } else {
+            // top-p (nucleus) sampling, clamping the least likely tokens to zero
+            next = sample_topp(logits, sampler->vocab_size, sampler->topp, sampler->probindex, coin);
+        }
+    }
+    return next;
+}
+
+// ----------------------------------------------------------------------------
+// utilities: time
+
+long time_in_ms() {
+    // return time in milliseconds, for benchmarking the model speed
+    struct timespec time;
+    clock_gettime(CLOCK_REALTIME, &time);
+    return time.tv_sec * 1000 + time.tv_nsec / 1000000;
+}
+
+// ----------------------------------------------------------------------------
+// generation loop
+
+void generate(Transformer *transformer, Tokenizer *tokenizer, Sampler *sampler, char *prompt, int steps) {
+    char *empty_prompt = "";
+    if (prompt == NULL) { prompt = empty_prompt; }
+
+    // encode the (string) prompt into tokens sequence
+    int num_prompt_tokens = 0;
+    int* prompt_tokens = (int*)malloc((strlen(prompt)+3) * sizeof(int)); // +3 for '\0', ?BOS, ?EOS
+    encode(tokenizer, prompt, 1, 0, prompt_tokens, &num_prompt_tokens);
+    if (num_prompt_tokens < 1) {
+        fprintf(stderr, "something is wrong, expected at least 1 prompt token\n");
+        exit(EXIT_FAILURE);
+    }
+
+    // start the main loop
+    long start = 0;  // used to time our code, only initialized after first iteration
+    int next;        // will store the next token in the sequence
+    int token = prompt_tokens[0]; // kick off with the first token in the prompt
+    int pos = 0;     // position in the sequence
+    while (pos < steps) {
+
+        // forward the transformer to get logits for the next token
+        float* logits = forward(transformer, token, pos);
+
+        // advance the state machine
+        if (pos < num_prompt_tokens - 1) {
+            // if we are still processing the input prompt, force the next prompt token
+            next = prompt_tokens[pos + 1];
+        } else {
+            // otherwise sample the next token from the logits
+            next = sample(sampler, logits);
+        }
+        pos++;
+
+        // data-dependent terminating condition: the BOS (=1) token delimits sequences
+        if (next == 1) { break; }
+
+        // print the token as string, decode it with the Tokenizer object
+        char* piece = decode(tokenizer, token, next);
+        safe_printf(piece); // same as printf("%s", piece), but skips "unsafe" bytes
+        fflush(stdout);
+        token = next;
+
+        // init the timer here because the first iteration can be slower
+        if (start == 0) { start = time_in_ms(); }
+    }
+    printf("\n");
+
+    // report achieved tok/s (pos-1 because the timer starts after first iteration)
+    if (pos > 1) {
+        long end = time_in_ms();
+        fprintf(stderr, "achieved tok/s: %f\n", (pos-1) / (double)(end-start)*1000);
+    }
+
+    free(prompt_tokens);
+}
+
+void read_stdin(const char* guide, char* buffer, size_t bufsize) {
+    // read a line from stdin, up to but not including \n
+    printf("%s", guide);
+    if (fgets(buffer, bufsize, stdin) != NULL) {
+        size_t len = strlen(buffer);
+        if (len > 0 && buffer[len - 1] == '\n') {
+            buffer[len - 1] = '\0'; // strip newline
+        }
+    }
+}
+
+// ----------------------------------------------------------------------------
+// chat loop
+// I manually inspected the tokens for a few chat conversations compared to
+// python reference and that seemed ok, but this was not thoroughly tested and
+// is not safely implemented, it's more a proof of concept atm.
+
+void chat(Transformer *transformer, Tokenizer *tokenizer, Sampler *sampler,
+          char *cli_user_prompt, char *cli_system_prompt, int steps) {
+
+    // buffers for reading the system prompt and user prompt from stdin
+    // you'll notice they are soomewhat haphazardly and unsafely set atm
+    char system_prompt[512];
+    char user_prompt[512];
+    char rendered_prompt[1152];
+    int num_prompt_tokens = 0;
+    int* prompt_tokens = (int*)malloc(1152 * sizeof(int));
+    int user_idx;
+
+    // start the main loop
+    int8_t user_turn = 1; // user starts
+    int next;        // will store the next token in the sequence
+    int token;       // stores the current token to feed into the transformer
+    int prev_token;
+    int pos = 0;     // position in the sequence
+    while (pos < steps) {
+
+        // when it is the user's turn to contribute tokens to the dialog...
+        if (user_turn) {
+            // get the (optional) system prompt at position 0
+            if (pos == 0) {
+                // at position 0, the user can also contribute a system prompt
+                if (cli_system_prompt == NULL) {
+                    // system prompt was not passed in, attempt to get it from stdin
+                    read_stdin("Enter system prompt (optional): ", system_prompt, sizeof(system_prompt));
+                } else {
+                    // system prompt was passed in, use it
+                    strcpy(system_prompt, cli_system_prompt);
+                }
+            }
+            // get the user prompt
+            if (pos == 0 && cli_user_prompt != NULL) {
+                // user prompt for position 0 was passed in, use it
+                strcpy(user_prompt, cli_user_prompt);
+            } else {
+                // otherwise get user prompt from stdin
+                read_stdin("User: ", user_prompt, sizeof(user_prompt));
+            }
+            // render user/system prompts into the Llama 2 Chat schema
+            if (pos == 0 && system_prompt[0] != '\0') {
+                char system_template[] = "[INST] <<SYS>>\n%s\n<</SYS>>\n\n%s [/INST]";
+                sprintf(rendered_prompt, system_template, system_prompt, user_prompt);
+            } else {
+                char user_template[] = "[INST] %s [/INST]";
+                sprintf(rendered_prompt, user_template, user_prompt);
+            }
+            // encode the rendered prompt into tokens
+            encode(tokenizer, rendered_prompt, 1, 0, prompt_tokens, &num_prompt_tokens);
+            user_idx = 0; // reset the user index
+            user_turn = 0;
+            printf("Assistant: ");
+        }
+
+        // determine the token to pass into the transformer next
+        if (user_idx < num_prompt_tokens) {
+            // if we are still processing the input prompt, force the next prompt token
+            token = prompt_tokens[user_idx++];
+        } else {
+            // otherwise use the next token sampled from previous turn
+            token = next;
+        }
+        // EOS (=2) token ends the Assistant turn
+        if (token == 2) { user_turn = 1; }
+
+        // forward the transformer to get logits for the next token
+        float* logits = forward(transformer, token, pos);
+        next = sample(sampler, logits);
+        pos++;
+
+        if (user_idx >= num_prompt_tokens && next != 2) {
+            // the Assistant is responding, so print its output
+            char* piece = decode(tokenizer, token, next);
+            safe_printf(piece); // same as printf("%s", piece), but skips "unsafe" bytes
+            fflush(stdout);
+        }
+        if (next == 2) { printf("\n"); }
+    }
+    printf("\n");
+    free(prompt_tokens);
+}
+
+
+// ----------------------------------------------------------------------------
+// CLI, include only if not testing
+#ifndef TESTING
+
+void error_usage() {
+    fprintf(stderr, "Usage:   run <checkpoint> [options]\n");
+    fprintf(stderr, "Example: run model.bin -n 256 -i \"Once upon a time\"\n");
+    fprintf(stderr, "Options:\n");
+    fprintf(stderr, "  -t <float>  temperature in [0,inf], default 1.0\n");
+    fprintf(stderr, "  -p <float>  p value in top-p (nucleus) sampling in [0,1] default 0.9\n");
+    fprintf(stderr, "  -s <int>    random seed, default time(NULL)\n");
+    fprintf(stderr, "  -n <int>    number of steps to run for, default 256. 0 = max_seq_len\n");
+    fprintf(stderr, "  -i <string> input prompt\n");
+    fprintf(stderr, "  -z <string> optional path to custom tokenizer\n");
+    fprintf(stderr, "  -m <string> mode: generate|chat, default: generate\n");
+    fprintf(stderr, "  -y <string> (optional) system prompt in chat mode\n");
+    exit(EXIT_FAILURE);
+}
+
+int main(int argc, char *argv[]) {
+
+    // default parameters
+    char *checkpoint_path = NULL;  // e.g. out/model.bin
+    char *tokenizer_path = "tokenizer.bin";
+    float temperature = 1.0f;   // 0.0 = greedy deterministic. 1.0 = original. don't set higher
+    float topp = 0.9f;          // top-p in nucleus sampling. 1.0 = off. 0.9 works well, but slower
+    int steps = 256;            // number of steps to run for
+    char *prompt = NULL;        // prompt string
+    unsigned long long rng_seed = 0; // seed rng with time by default
+    char *mode = "generate";    // generate|chat
+    char *system_prompt = NULL; // the (optional) system prompt to use in chat mode
+
+    // poor man's C argparse so we can override the defaults above from the command line
+    if (argc >= 2) { checkpoint_path = argv[1]; } else { error_usage(); }
+    for (int i = 2; i < argc; i+=2) {
+        // do some basic validation
+        if (i + 1 >= argc) { error_usage(); } // must have arg after flag
+        if (argv[i][0] != '-') { error_usage(); } // must start with dash
+        if (strlen(argv[i]) != 2) { error_usage(); } // must be -x (one dash, one letter)
+        // read in the args
+        if (argv[i][1] == 't') { temperature = atof(argv[i + 1]); }
+        else if (argv[i][1] == 'p') { topp = atof(argv[i + 1]); }
+        else if (argv[i][1] == 's') { rng_seed = atoi(argv[i + 1]); }
+        else if (argv[i][1] == 'n') { steps = atoi(argv[i + 1]); }
+        else if (argv[i][1] == 'i') { prompt = argv[i + 1]; }
+        else if (argv[i][1] == 'z') { tokenizer_path = argv[i + 1]; }
+        else if (argv[i][1] == 'm') { mode = argv[i + 1]; }
+        else if (argv[i][1] == 'y') { system_prompt = argv[i + 1]; }
+        else { error_usage(); }
+    }
+
+    // parameter validation/overrides
+    if (rng_seed <= 0) rng_seed = (unsigned int)time(NULL);
+    if (temperature < 0.0) temperature = 0.0;
+    if (topp < 0.0 || 1.0 < topp) topp = 0.9;
+    if (steps < 0) steps = 0;
+
+    // build the Transformer via the model .bin file
+    Transformer transformer;
+    build_transformer(&transformer, checkpoint_path);
+    if (steps == 0 || steps > transformer.config.seq_len) steps = transformer.config.seq_len; // override to ~max length
+
+    // build the Tokenizer via the tokenizer .bin file
+    Tokenizer tokenizer;
+    build_tokenizer(&tokenizer, tokenizer_path, transformer.config.vocab_size);
+
+    // build the Sampler
+    Sampler sampler;
+    build_sampler(&sampler, transformer.config.vocab_size, temperature, topp, rng_seed);
+
+    // run!
+    if (strcmp(mode, "generate") == 0) {
+        generate(&transformer, &tokenizer, &sampler, prompt, steps);
+    } else if (strcmp(mode, "chat") == 0) {
+        chat(&transformer, &tokenizer, &sampler, prompt, system_prompt, steps);
+    } else {
+        fprintf(stderr, "unknown mode: %s\n", mode);
+        error_usage();
+    }
+
+    // memory and file handles cleanup
+    free_sampler(&sampler);
+    free_tokenizer(&tokenizer);
+    free_transformer(&transformer);
+    return 0;
+}
+#endif

c/llama2/llgo.cfg

@@ -1,6 +0,0 @@
-{
-    "cl": [
-        "clang -emit-llvm -S -o llgo_autogen.ll -c llama2/run.c",
-        "rm llgo_autogen.lla; zip llgo_autogen.lla llgo_autogen.ll"
-    ]
-}

c/math/cmplx/complex.go

@@ -0,0 +1,151 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package cmplx
+
+import (
+	_ "unsafe"
+)
+
+const (
+	LLGoPackage = "decl"
+)
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Abs C.cabs
+func Abs(z complex128) float64
+
+//go:linkname Acos C.cacos
+func Acos(z complex128) complex128
+
+//go:linkname Acosh C.cacosh
+func Acosh(z complex128) complex128
+
+//go:linkname Asin C.casin
+func Asin(z complex128) complex128
+
+//go:linkname Asinh C.casinh
+func Asinh(z complex128) complex128
+
+//go:linkname Atan C.catan
+func Atan(z complex128) complex128
+
+//go:linkname Atanh C.catanh
+func Atanh(z complex128) complex128
+
+//go:linkname Cos C.ccos
+func Cos(z complex128) complex128
+
+//go:linkname Cosh C.ccosh
+func Cosh(z complex128) complex128
+
+//go:linkname Exp C.cexp
+func Exp(z complex128) complex128
+
+//go:linkname Log C.clog
+func Log(z complex128) complex128
+
+//go:linkname Log10 C.clog10
+func Log10(z complex128) complex128
+
+//go:linkname Arg C.carg
+func Arg(z complex128) float64
+
+//go:linkname Phase C.carg
+func Phase(z complex128) float64
+
+//go:linkname Pow C.cpow
+func Pow(x, y complex128) complex128
+
+//go:linkname Sin C.csin
+func Sin(z complex128) complex128
+
+//go:linkname Sinh C.csinh
+func Sinh(z complex128) complex128
+
+//go:linkname Sqrt C.csqrt
+func Sqrt(z complex128) complex128
+
+//go:linkname Tan C.ctan
+func Tan(z complex128) complex128
+
+//go:linkname Tanh C.ctanh
+func Tanh(z complex128) complex128
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Absf C.cabsf
+func Absf(z complex64) float32
+
+//go:linkname Acosf C.cacosf
+func Acosf(z complex64) complex64
+
+//go:linkname Acoshf C.cacoshf
+func Acoshf(z complex64) complex64
+
+//go:linkname Asinf C.casinf
+func Asinf(z complex64) complex64
+
+//go:linkname Asinhf C.casinhf
+func Asinhf(z complex64) complex64
+
+//go:linkname Atanf C.catanf
+func Atanf(z complex64) complex64
+
+//go:linkname Atanhf C.catanhf
+func Atanhf(z complex64) complex64
+
+//go:linkname Cosf C.ccosf
+func Cosf(z complex64) complex64
+
+//go:linkname Coshf C.ccoshf
+func Coshf(z complex64) complex64
+
+//go:linkname Expf C.cexpf
+func Expf(z complex64) complex64
+
+//go:linkname Logf C.clogf
+func Logf(z complex64) complex64
+
+//go:linkname Log10f C.clog10f
+func Log10f(z complex64) complex64
+
+//go:linkname Argf C.cargf
+func Argf(z complex64) float32
+
+//go:linkname Phasef C.cargf
+func Phasef(z complex64) float32
+
+//go:linkname Powf C.cpowf
+func Powf(x, y complex64) complex64
+
+//go:linkname Sinf C.csinf
+func Sinf(z complex64) complex64
+
+//go:linkname Sinhf C.csinhf
+func Sinhf(z complex64) complex64
+
+//go:linkname Sqrtf C.csqrtf
+func Sqrtf(z complex64) complex64
+
+//go:linkname Tanf C.ctanf
+func Tanf(z complex64) complex64
+
+//go:linkname Tanhf C.ctanhf
+func Tanhf(z complex64) complex64
+
+// -----------------------------------------------------------------------------

c/math/math.go

@@ -0,0 +1,333 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package math
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "decl"
+)
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Acos C.acos
+func Acos(x float64) float64
+
+//go:linkname Acosh C.acosh
+func Acosh(x float64) float64
+
+//go:linkname Asin C.asin
+func Asin(x float64) float64
+
+//go:linkname Asinh C.asinh
+func Asinh(x float64) float64
+
+//go:linkname Atan C.atan
+func Atan(x float64) float64
+
+//go:linkname Atan2 C.atan2
+func Atan2(y, x float64) float64
+
+//go:linkname Atanh C.atanh
+func Atanh(x float64) float64
+
+//go:linkname Cbrt C.cbrt
+func Cbrt(x float64) float64
+
+//go:linkname Ceil C.ceil
+func Ceil(x float64) float64
+
+//go:linkname Cos C.cos
+func Cos(x float64) float64
+
+//go:linkname Cosh C.cosh
+func Cosh(x float64) float64
+
+//go:linkname Copysign C.copysign
+func Copysign(x, y float64) float64
+
+//go:linkname Erf C.erf
+func Erf(x float64) float64
+
+//go:linkname Erfc C.erfc
+func Erfc(x float64) float64
+
+//go:linkname Exp C.exp
+func Exp(x float64) float64
+
+//go:linkname Exp2 C.exp2
+func Exp2(x float64) float64
+
+//go:linkname Expm1 C.expm1
+func Expm1(x float64) float64
+
+//go:linkname Fdim C.fdim
+func Fdim(x, y float64) float64
+
+//go:linkname Floor C.floor
+func Floor(x float64) float64
+
+//go:linkname Fma C.fma
+func Fma(x, y, z float64) float64
+
+//go:linkname Fmax C.fmax
+func Fmax(x, y float64) float64
+
+//go:linkname Fmin C.fmin
+func Fmin(x, y float64) float64
+
+//go:linkname Fmod C.fmod
+func Fmod(x, y float64) float64
+
+//go:linkname Frexp C.frexp
+func Frexp(x float64, exp *c.Int) float64
+
+//go:linkname Gamma C.gamma
+func Gamma(x float64) float64
+
+//go:linkname Hypot C.hypot
+func Hypot(x, y float64) float64
+
+//go:linkname Ilogb C.ilogb
+func Ilogb(x float64) c.Int
+
+//go:linkname J0 C.j0
+func J0(x float64) float64
+
+//go:linkname J1 C.j1
+func J1(x float64) float64
+
+//go:linkname Jn C.jn
+func Jn(n c.Int, x float64) float64
+
+//go:linkname Ldexp C.ldexp
+func Ldexp(x float64, exp c.Int) float64
+
+//go:linkname Lgamma C.lgamma
+func Lgamma(x float64) float64
+
+//go:linkname Log C.log
+func Log(x float64) float64
+
+//go:linkname Log10 C.log10
+func Log10(x float64) float64
+
+//go:linkname Log1p C.log1p
+func Log1p(x float64) float64
+
+//go:linkname Log2 C.log2
+func Log2(x float64) float64
+
+//go:linkname Logb C.logb
+func Logb(x float64) float64
+
+//go:linkname Modf C.modf
+func Modf(x float64, ipart *float64) float64
+
+//go:linkname Nan C.nan
+func Nan(tag *c.Char) float64
+
+//go:linkname Nextafter C.nextafter
+func Nextafter(x, y float64) float64
+
+//go:linkname Pow C.pow
+func Pow(x, y float64) float64
+
+//go:linkname Remainder C.remainder
+func Remainder(x, y float64) float64
+
+//go:linkname Round C.round
+func Round(x float64) float64
+
+//go:linkname Sin C.sin
+func Sin(x float64) float64
+
+//go:linkname Sinh C.sinh
+func Sinh(x float64) float64
+
+//go:linkname Sqrt C.sqrt
+func Sqrt(x float64) float64
+
+//go:linkname Tan C.tan
+func Tan(x float64) float64
+
+//go:linkname Tanh C.tanh
+func Tanh(x float64) float64
+
+//go:linkname Tgamma C.tgamma
+func Tgamma(x float64) float64
+
+//go:linkname Trunc C.trunc
+func Trunc(x float64) float64
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Acosf C.acosf
+func Acosf(x float32) float32
+
+//go:linkname Acoshf C.acoshf
+func Acoshf(x float32) float32
+
+//go:linkname Asinf C.asinf
+func Asinf(x float32) float32
+
+//go:linkname Asinhf C.asinhf
+func Asinhf(x float32) float32
+
+//go:linkname Atanf C.atanf
+func Atanf(x float32) float32
+
+//go:linkname Atan2f C.atan2f
+func Atan2f(y, x float32) float32
+
+//go:linkname Atanhf C.atanhf
+func Atanhf(x float32) float32
+
+//go:linkname Cbrtf C.cbrtf
+func Cbrtf(x float32) float32
+
+//go:linkname Ceilf C.ceilf
+func Ceilf(x float32) float32
+
+//go:linkname Cosf C.cosf
+func Cosf(x float32) float32
+
+//go:linkname Coshf C.coshf
+func Coshf(x float32) float32
+
+//go:linkname Copysignf C.copysignf
+func Copysignf(x, y float32) float32
+
+//go:linkname Erff C.erff
+func Erff(x float32) float32
+
+//go:linkname Erfcf C.erfcf
+func Erfcf(x float32) float32
+
+//go:linkname Expf C.expf
+func Expf(x float32) float32
+
+//go:linkname Exp2f C.exp2f
+func Exp2f(x float32) float32
+
+//go:linkname Expm1f C.expm1f
+func Expm1f(x float32) float32
+
+//go:linkname Fdimf C.fdimf
+func Fdimf(x, y float32) float32
+
+//go:linkname Floorf C.floorf
+func Floorf(x float32) float32
+
+//go:linkname Fmaf C.fmaf
+func Fmaf(x, y, z float32) float32
+
+//go:linkname Fmaxf C.fmaxf
+func Fmaxf(x, y float32) float32
+
+//go:linkname Fminf C.fminf
+func Fminf(x, y float32) float32
+
+//go:linkname Fmodf C.fmodf
+func Fmodf(x, y float32) float32
+
+//go:linkname Frexpf C.frexpf
+func Frexpf(x float32, exp *c.Int) float32
+
+//go:linkname Gammaf C.gammaf
+func Gammaf(x float32) float32
+
+//go:linkname Hypotf C.hypotf
+func Hypotf(x, y float32) float32
+
+//go:linkname Ilogbf C.ilogbf
+func Ilogbf(x float32) c.Int
+
+//go:linkname J0f C.j0f
+func J0f(x float32) float32
+
+//go:linkname J1f C.j1f
+func J1f(x float32) float32
+
+//go:linkname Jnf C.jnf
+func Jnf(n c.Int, x float32) float32
+
+//go:linkname Ldexpf C.ldexpf
+func Ldexpf(x float32, exp c.Int) float32
+
+//go:linkname Lgammaf C.lgammaf
+func Lgammaf(x float32) float32
+
+//go:linkname Logf C.logf
+func Logf(x float32) float32
+
+//go:linkname Log10f C.log10f
+func Log10f(x float32) float32
+
+//go:linkname Log1pf C.log1pf
+func Log1pf(x float32) float32
+
+//go:linkname Log2f C.log2f
+func Log2f(x float32) float32
+
+//go:linkname Logbf C.logbf
+func Logbf(x float32) float32
+
+//go:linkname Modff C.modff
+func Modff(x float32, ipart *float32) float32
+
+//go:linkname Nanf C.nanf
+func Nanf(tag *c.Char) float32
+
+//go:linkname Nextafterf C.nextafterf
+func Nextafterf(x, y float32) float32
+
+//go:linkname Powf C.powf
+func Powf(x, y float32) float32
+
+//go:linkname Remainderf C.remainderf
+func Remainderf(x, y float32) float32
+
+//go:linkname Roundf C.roundf
+func Roundf(x float32) float32
+
+//go:linkname Sinf C.sinf
+func Sinf(x float32) float32
+
+//go:linkname Sinhf C.sinhf
+func Sinhf(x float32) float32
+
+//go:linkname Sqrtf C.sqrtf
+func Sqrtf(x float32) float32
+
+//go:linkname Tanf C.tanf
+func Tanf(x float32) float32
+
+//go:linkname Tanhf C.tanhf
+func Tanhf(x float32) float32
+
+//go:linkname Tgammaf C.tgammaf
+func Tgammaf(x float32) float32
+
+//go:linkname Truncf C.truncf
+func Truncf(x float32) float32
+
+// -----------------------------------------------------------------------------

c/math/rand/rand.go

@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package rand
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "decl"
+)
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Rand C.rand
+func Rand() c.Int
+
+//go:linkname RandR C.rand_r
+func RandR(*c.Uint) c.Int
+
+//go:linkname Srand C.srand
+func Srand(c.Uint)
+
+//go:linkname Sranddev C.sranddev
+func Sranddev()
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Random C.random
+func Random() c.Long
+
+//go:linkname Srandom C.srandom
+func Srandom(c.Uint)
+
+//go:linkname Srandomdev C.srandomdev
+func Srandomdev()
+
+// -----------------------------------------------------------------------------

c/os/os.go

@@ -0,0 +1,231 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package os
+
+// #include <sys/stat.h>
+// #include <limits.h>
+import "C"
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "decl"
+)
+
+const (
+	PATH_MAX = C.PATH_MAX
+)
+
+type (
+	ModeT C.mode_t
+	UidT  C.uid_t
+	GidT  C.gid_t
+	OffT  C.off_t
+	DevT  C.dev_t
+	StatT C.struct_stat
+)
+
+//go:linkname Errno errno
+var Errno c.Int
+
+//go:linkname Umask C.umask
+func Umask(cmask ModeT) ModeT
+
+//go:linkname Mkdir C.mkdir
+func Mkdir(path *c.Char, mode ModeT) c.Int
+
+//go:linkname Rmdir C.rmdir
+func Rmdir(path *c.Char) c.Int
+
+//go:linkname Link C.link
+func Link(oldpath *c.Char, newpath *c.Char) c.Int
+
+//go:linkname Symlink C.symlink
+func Symlink(target *c.Char, linkpath *c.Char) c.Int
+
+//go:linkname Readlink C.readlink
+func Readlink(path *c.Char, buf c.Pointer, bufsize uintptr) int
+
+//go:linkname Unlink C.unlink
+func Unlink(path *c.Char) c.Int
+
+//go:linkname Remove C.remove
+func Remove(path *c.Char) c.Int
+
+//go:linkname Rename C.rename
+func Rename(oldpath *c.Char, newpath *c.Char) c.Int
+
+//go:linkname Stat C.stat
+func Stat(path *c.Char, buf *StatT) c.Int
+
+//go:linkname Lstat C.lstat
+func Lstat(path *c.Char, buf *StatT) c.Int
+
+//go:linkname Truncate C.truncate
+func Truncate(path *c.Char, length OffT) c.Int
+
+//go:linkname Chmod C.chmod
+func Chmod(path *c.Char, mode ModeT) c.Int
+
+//go:linkname Chown C.chown
+func Chown(path *c.Char, owner UidT, group GidT) c.Int
+
+//go:linkname Lchown C.lchown
+func Lchown(path *c.Char, owner UidT, group GidT) c.Int
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Getcwd C.getcwd
+func Getcwd(buffer c.Pointer, size uintptr) *c.Char
+
+//go:linkname Chdir C.chdir
+func Chdir(path *c.Char) c.Int
+
+//go:linkname Chroot C.chroot
+func Chroot(path *c.Char) c.Int
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Environ C.environ
+func Environ() **c.Char
+
+//go:linkname Getenv C.getenv
+func Getenv(name *c.Char) *c.Char
+
+//go:linkname Setenv C.setenv
+func Setenv(name *c.Char, value *c.Char, overwrite c.Int) c.Int
+
+//go:linkname Putenv C.putenv
+func Putenv(env *c.Char) c.Int
+
+//go:linkname Unsetenv C.unsetenv
+func Unsetenv(name *c.Char) c.Int
+
+//go:linkname Clearenv C.clearenv
+func Clearenv()
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Fchdir C.fchdir
+func Fchdir(dirfd c.Int) c.Int
+
+//go:linkname Faccessat C.faccessat
+func Faccessat(dirfd c.Int, path *c.Char, mode c.Int, flags c.Int) c.Int
+
+//go:linkname Fchmodat C.fchmodat
+func Fchmodat(dirfd c.Int, path *c.Char, mode ModeT, flags c.Int) c.Int
+
+//go:linkname Fchownat C.fchownat
+func Fchownat(dirfd c.Int, path *c.Char, owner UidT, group GidT, flags c.Int) c.Int
+
+//go:linkname Fstatat C.fstatat
+func Fstatat(dirfd c.Int, path *c.Char, buf *StatT, flags c.Int) c.Int
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Open C.open
+func Open(path *c.Char, flags c.Int, mode ModeT) c.Int
+
+//go:linkname Openat C.openat
+func Openat(dirfd c.Int, path *c.Char, flags c.Int, mode ModeT) c.Int
+
+//go:linkname Creat C.creat
+func Creat(path *c.Char, mode ModeT) c.Int
+
+//go:linkname Dup C.dup
+func Dup(fd c.Int) c.Int
+
+//go:linkname Dup2 C.dup2
+func Dup2(oldfd c.Int, newfd c.Int) c.Int
+
+/* TODO(xsw):
+On Alpha, IA-64, MIPS, SuperH, and SPARC/SPARC64, pipe() has the following prototype:
+struct fd_pair {
+	long fd[2];
+};
+struct fd_pair pipe(void);
+*/
+//go:linkname Pipe C.pipe
+func Pipe(fds *[2]c.Int) c.Int
+
+//go:linkname Mkfifo C.mkfifo
+func Mkfifo(path *c.Char, mode ModeT) c.Int
+
+//go:linkname Mknod C.mknod
+func Mknod(path *c.Char, mode ModeT, dev DevT) c.Int
+
+//go:linkname Close C.close
+func Close(fd c.Int) c.Int
+
+//go:linkname Read C.read
+func Read(fd c.Int, buf c.Pointer, count uintptr) int
+
+//go:linkname Write C.write
+func Write(fd c.Int, buf c.Pointer, count uintptr) int
+
+//go:linkname Lseek C.lseek
+func Lseek(fd c.Int, offset OffT, whence c.Int) OffT
+
+//go:linkname Fsync C.fsync
+func Fsync(fd c.Int) c.Int
+
+//go:linkname Ftruncate C.ftruncate
+func Ftruncate(fd c.Int, length OffT) c.Int
+
+//go:linkname Fchmod C.fchmod
+func Fchmod(fd c.Int, mode ModeT) c.Int
+
+//go:linkname Fchown C.fchown
+func Fchown(fd c.Int, owner UidT, group GidT) c.Int
+
+//go:linkname Fstat C.fstat
+func Fstat(fd c.Int, buf *StatT) c.Int
+
+//go:linkname Isatty C.isatty
+func Isatty(fd c.Int) c.Int
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Kill C.kill
+func Kill(pid c.Int, sig c.Int) c.Int
+
+//go:linkname Exit C.exit
+func Exit(c.Int)
+
+//go:linkname Getpid C.getpid
+func Getpid() c.Int
+
+//go:linkname Getppid C.getppid
+func Getppid() c.Int
+
+//go:linkname Getuid C.getuid
+func Getuid() UidT
+
+//go:linkname Geteuid C.geteuid
+func Geteuid() UidT
+
+//go:linkname Getgid C.getgid
+func Getgid() GidT
+
+//go:linkname Getegid C.getegid
+func Getegid() GidT
+
+// -----------------------------------------------------------------------------

c/pthread/pthread.go

@@ -0,0 +1,160 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package pthread
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "decl"
+)
+
+func __noop__() c.Int {
+	return 0
+}
+
+// -----------------------------------------------------------------------------
+
+type aThread struct {
+	Unused [8]byte
+}
+
+// Thread represents a POSIX thread.
+type Thread = *aThread
+
+// The pthread_create() function starts a new thread in the calling
+// process.  The new thread starts execution by invoking
+// start_routine(); arg is passed as the sole argument of
+// start_routine().
+//
+// The new thread terminates in one of the following ways:
+//
+//   - It calls pthread_exit(3), specifying an exit status value that
+//     is available to another thread in the same process that calls
+//     pthread_join(3).
+//
+//   - It returns from start_routine().  This is equivalent to
+//     calling pthread_exit(3) with the value supplied in the return
+//     statement.
+//
+//   - It is canceled (see pthread_cancel(3)).
+//
+//   - Any of the threads in the process calls exit(3), or the main
+//     thread performs a return from main().  This causes the
+//     termination of all threads in the process.
+//
+// On success, pthread_create() returns 0; on error, it returns an
+// error number, and the contents of *thread are undefined.
+//
+// See https://man7.org/linux/man-pages/man3/pthread_create.3.html
+//
+//go:linkname Create C.pthread_create
+func Create(pthread *Thread, attr *Attr, routine func(c.Pointer) c.Pointer, arg c.Pointer) c.Int
+
+// The pthread_join() function waits for the thread specified by
+// thread to terminate.  If that thread has already terminated, then
+// pthread_join() returns immediately.  The thread specified by
+// thread must be joinable.
+//
+// If retval is not NULL, then pthread_join() copies the exit status
+// of the target thread (i.e., the value that the target thread
+// supplied to pthread_exit(3)) into the location pointed to by
+// retval.  If the target thread was canceled, then PTHREAD_CANCELED
+// is placed in the location pointed to by retval.
+//
+// If multiple threads simultaneously try to join with the same
+// thread, the results are undefined.  If the thread calling
+// pthread_join() is canceled, then the target thread will remain
+// joinable (i.e., it will not be detached).
+//
+// See https://man7.org/linux/man-pages/man3/pthread_join.3.html
+//
+//go:linkname Join C.pthread_join
+func Join(thread Thread, retval *c.Pointer) c.Int
+
+// The pthread_exit() function terminates the calling thread and
+// returns a value via retval that (if the thread is joinable) is
+// available to another thread in the same process that calls
+// pthread_join(3).
+//
+// See https://man7.org/linux/man-pages/man3/pthread_exit.3.html
+//
+//go:linkname Exit C.pthread_exit
+func Exit(retval c.Pointer)
+
+// The pthread_cancel() function sends a cancelation request to the
+// thread thread.
+//
+// See https://man7.org/linux/man-pages/man3/pthread_cancel.3.html
+//
+//go:linkname Cancel C.pthread_cancel
+func Cancel(thread Thread) c.Int
+
+// -----------------------------------------------------------------------------
+
+// Attr represents a POSIX thread attributes.
+type Attr struct {
+	Detached byte
+	SsSp     *c.Char
+	SsSize   uintptr
+}
+
+// llgo:link (*Attr).Init C.pthread_attr_init
+func (attr *Attr) Init() c.Int { return 0 }
+
+// llgo:link (*Attr).Destroy C.pthread_attr_destroy
+func (attr *Attr) Destroy() c.Int { return 0 }
+
+// llgo:link (*Attr).GetDetached C.pthread_attr_getdetachstate
+func (attr *Attr) GetDetached(detached *c.Int) c.Int { return 0 }
+
+// llgo:link (*Attr).SetDetached C.pthread_attr_setdetachstate
+func (attr *Attr) SetDetached(detached c.Int) c.Int { return 0 }
+
+// llgo:link (*Attr).GetStackSize C.pthread_attr_getstacksize
+func (attr *Attr) GetStackSize(stackSize *uintptr) c.Int { return 0 }
+
+// llgo:link (*Attr).SetStackSize C.pthread_attr_setstacksize
+func (attr *Attr) SetStackSize(stackSize uintptr) c.Int { return 0 }
+
+// llgo:link (*Attr).GetStackAddr C.pthread_attr_getstackaddr
+func (attr *Attr) GetStackAddr(stackAddr *c.Pointer) c.Int { return 0 }
+
+// llgo:link (*Attr).SetStackAddr C.pthread_attr_setstackaddr
+func (attr *Attr) SetStackAddr(stackAddr c.Pointer) c.Int { return 0 }
+
+// -----------------------------------------------------------------------------
+// Thread Local Storage
+
+type Key c.Uint
+
+// llgo:link (*Key).Create C.pthread_key_create
+func (key *Key) Create(destructor func(c.Pointer)) c.Int { return 0 }
+
+// llgo:link Key.Delete C.pthread_key_delete
+func (key Key) Delete() c.Int { return 0 }
+
+// llgo:link Key.Get C.pthread_getspecific
+func (key Key) Get() c.Pointer { return nil }
+
+// llgo:link Key.Set C.pthread_setspecific
+func (key Key) Set(value c.Pointer) c.Int { return __noop__() }
+
+// -----------------------------------------------------------------------------

c/pthread/sync/_pthd/pthd.c

@@ -0,0 +1,7 @@
+#include <pthread.h>
+
+// -----------------------------------------------------------------------------
+
+pthread_once_t llgoSyncOnceInitVal = PTHREAD_ONCE_INIT;
+
+// -----------------------------------------------------------------------------

c/pthread/sync/sync.go

@@ -0,0 +1,174 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package sync
+
+// #include <pthread.h>
+import "C"
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/time"
+)
+
+const (
+	LLGoFiles   = "_pthd/pthd.c"
+	LLGoPackage = "link"
+)
+
+// -----------------------------------------------------------------------------
+
+// Once is an object that will perform exactly one action.
+type Once C.pthread_once_t
+
+//go:linkname OnceInit llgoSyncOnceInitVal
+var OnceInit Once
+
+// llgo:link (*Once).Do C.pthread_once
+func (o *Once) Do(f func()) c.Int { return 0 }
+
+// -----------------------------------------------------------------------------
+
+type MutexType c.Int
+
+const (
+	MUTEX_NORMAL     MutexType = C.PTHREAD_MUTEX_NORMAL
+	MUTEX_ERRORCHECK MutexType = C.PTHREAD_MUTEX_ERRORCHECK
+	MUTEX_RECURSIVE  MutexType = C.PTHREAD_MUTEX_RECURSIVE
+	MUTEX_DEFAULT    MutexType = C.PTHREAD_MUTEX_DEFAULT
+)
+
+// MutexAttr is a mutex attribute object.
+type MutexAttr C.pthread_mutexattr_t
+
+// llgo:link (*MutexAttr).Init C.pthread_mutexattr_init
+func (a *MutexAttr) Init(attr *MutexAttr) c.Int { return 0 }
+
+// llgo:link (*MutexAttr).Destroy C.pthread_mutexattr_destroy
+func (a *MutexAttr) Destroy() {}
+
+// llgo:link (*MutexAttr).SetType C.pthread_mutexattr_settype
+func (a *MutexAttr) SetType(typ MutexType) c.Int { return 0 }
+
+// -----------------------------------------------------------------------------
+
+// Mutex is a mutual exclusion lock.
+type Mutex C.pthread_mutex_t
+
+// llgo:link (*Mutex).Init C.pthread_mutex_init
+func (m *Mutex) Init(attr *MutexAttr) c.Int { return 0 }
+
+// llgo:link (*Mutex).Destroy C.pthread_mutex_destroy
+func (m *Mutex) Destroy() {}
+
+// llgo:link (*Mutex).Lock C.pthread_mutex_lock
+func (m *Mutex) Lock() {}
+
+// llgo:link (*Mutex).TryLock C.pthread_mutex_trylock
+func (m *Mutex) TryLock() c.Int { return 0 }
+
+// llgo:link (*Mutex).Unlock C.pthread_mutex_unlock
+func (m *Mutex) Unlock() {}
+
+// -----------------------------------------------------------------------------
+
+// RWLockAttr is a read-write lock attribute object.
+type RWLockAttr C.pthread_rwlockattr_t
+
+// llgo:link (*RWLockAttr).Init C.pthread_rwlockattr_init
+func (a *RWLockAttr) Init(attr *RWLockAttr) c.Int { return 0 }
+
+// llgo:link (*RWLockAttr).Destroy C.pthread_rwlockattr_destroy
+func (a *RWLockAttr) Destroy() {}
+
+// llgo:link (*RWLockAttr).SetPShared C.pthread_rwlockattr_setpshared
+func (a *RWLockAttr) SetPShared(pshared c.Int) c.Int { return 0 }
+
+// llgo:link (*RWLockAttr).GetPShared C.pthread_rwlockattr_getpshared
+func (a *RWLockAttr) GetPShared(pshared *c.Int) c.Int { return 0 }
+
+// -----------------------------------------------------------------------------
+
+// RWLock is a read-write lock.
+type RWLock C.pthread_rwlock_t
+
+// llgo:link (*RWLock).Init C.pthread_rwlock_init
+func (rw *RWLock) Init(attr *RWLockAttr) c.Int { return 0 }
+
+// llgo:link (*RWLock).Destroy C.pthread_rwlock_destroy
+func (rw *RWLock) Destroy() {}
+
+// llgo:link (*RWLock).RLock C.pthread_rwlock_rdlock
+func (rw *RWLock) RLock() {}
+
+// llgo:link (*RWLock).TryRLock C.pthread_rwlock_tryrdlock
+func (rw *RWLock) TryRLock() c.Int { return 0 }
+
+// llgo:link (*RWLock).RUnlock C.pthread_rwlock_unlock
+func (rw *RWLock) RUnlock() {}
+
+// llgo:link (*RWLock).Lock C.pthread_rwlock_wrlock
+func (rw *RWLock) Lock() {}
+
+// llgo:link (*RWLock).TryLock C.pthread_rwlock_trywrlock
+func (rw *RWLock) TryLock() c.Int { return 0 }
+
+// llgo:link (*RWLock).Unlock C.pthread_rwlock_unlock
+func (rw *RWLock) Unlock() {}
+
+// -----------------------------------------------------------------------------
+
+// CondAttr is a condition variable attribute object.
+type CondAttr C.pthread_condattr_t
+
+// llgo:link (*CondAttr).Init C.pthread_condattr_init
+func (a *CondAttr) Init(attr *CondAttr) c.Int { return 0 }
+
+// llgo:link (*CondAttr).Destroy C.pthread_condattr_destroy
+func (a *CondAttr) Destroy() {}
+
+// llgo:link (*CondAttr).SetClock C.pthread_condattr_setclock
+func (a *CondAttr) SetClock(clock time.ClockidT) c.Int { return 0 }
+
+// llgo:link (*CondAttr).GetClock C.pthread_condattr_getclock
+func (a *CondAttr) GetClock(clock *time.ClockidT) c.Int { return 0 }
+
+// -----------------------------------------------------------------------------
+
+// Cond is a condition variable.
+type Cond C.pthread_cond_t
+
+// llgo:link (*Cond).Init C.pthread_cond_init
+func (c *Cond) Init(attr *CondAttr) c.Int { return 0 }
+
+// llgo:link (*Cond).Destroy C.pthread_cond_destroy
+func (c *Cond) Destroy() {}
+
+// llgo:link (*Cond).Signal C.pthread_cond_signal
+func (c *Cond) Signal() c.Int { return 0 }
+
+// llgo:link (*Cond).Broadcast C.pthread_cond_broadcast
+func (c *Cond) Broadcast() c.Int { return 0 }
+
+// llgo:link (*Cond).Wait C.pthread_cond_wait
+func (c *Cond) Wait(m *Mutex) c.Int { return 0 }
+
+// llgo:link (*Cond).TimedWait C.pthread_cond_timedwait
+func (c *Cond) TimedWait(m *Mutex, abstime *time.Timespec) c.Int { return 0 }
+
+// -----------------------------------------------------------------------------

c/raylib/_demo/raylibdemo/raydemo.go

@@ -0,0 +1,18 @@
+package main
+
+import (
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/raylib"
+)
+
+func main() {
+	const screenWidth = 800
+	const screenHeight = 450
+	raylib.InitWindow(screenWidth, screenHeight, c.Str("Raylib DEMO"))
+	for !raylib.WindowShouldClose() {
+		raylib.BeginDrawing()
+		raylib.ClearBackground(raylib.RAYWHITE)
+		raylib.DrawRectangle(screenWidth/2-50, screenHeight/2-50, 100, 100, raylib.BLUE)
+		raylib.EndDrawing()
+	}
+}

c/raylib/_demo/tetris/tetris.go

@@ -0,0 +1,181 @@
+package main
+
+import (
+	"unsafe"
+
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/raylib"
+)
+
+const (
+	BOARD_WIDTH  = 10
+	BOARD_HEIGHT = 20
+	BLOCK_SIZE   = 30
+
+	SCREENWIDTH  = 300
+	SCREENHEIGHT = 600
+)
+
+const MAX_BLOCKS = 4
+
+type Shape struct {
+	Blocks [MAX_BLOCKS]raylib.Vector2
+	Color  raylib.Color
+}
+
+var SHAPES = []Shape{
+	{Blocks: [MAX_BLOCKS]raylib.Vector2{{X: 0, Y: 0}, {X: 1, Y: 0}, {X: 2, Y: 0}, {X: 3, Y: 0}}, Color: raylib.SKYBLUE},
+	{Blocks: [MAX_BLOCKS]raylib.Vector2{{X: 0, Y: 0}, {X: 1, Y: 0}, {X: 0, Y: 1}, {X: 1, Y: 1}}, Color: raylib.YELLOW},
+	{Blocks: [MAX_BLOCKS]raylib.Vector2{{X: 1, Y: 0}, {X: 0, Y: 1}, {X: 1, Y: 1}, {X: 2, Y: 1}}, Color: raylib.PURPLE},
+	{Blocks: [MAX_BLOCKS]raylib.Vector2{{X: 1, Y: 0}, {X: 2, Y: 0}, {X: 0, Y: 1}, {X: 1, Y: 1}}, Color: raylib.GREEN},
+	{Blocks: [MAX_BLOCKS]raylib.Vector2{{X: 0, Y: 0}, {X: 1, Y: 0}, {X: 1, Y: 1}, {X: 2, Y: 1}}, Color: raylib.RED},
+	{Blocks: [MAX_BLOCKS]raylib.Vector2{{X: 0, Y: 0}, {X: 0, Y: 1}, {X: 1, Y: 1}, {X: 2, Y: 1}}, Color: raylib.BLUE},
+	{Blocks: [MAX_BLOCKS]raylib.Vector2{{X: 2, Y: 0}, {X: 0, Y: 1}, {X: 1, Y: 1}, {X: 2, Y: 1}}, Color: raylib.ORANGE},
+}
+
+var board [BOARD_HEIGHT][BOARD_WIDTH]raylib.Color
+
+var curShape Shape
+var curPos raylib.Vector2
+
+var fallTime = c.Float(0)
+var fallSpeed = c.Float(0.2)
+var score = 0
+var scoreText = make([]c.Char, 20)
+var gameOver = false
+
+func genShape() {
+	curShape = SHAPES[raylib.GetRandomValue(c.Int(0), c.Int(6))]
+	curPos = raylib.Vector2{BOARD_WIDTH/2 - 1, 0}
+}
+
+func checkCollision() bool {
+	for i := 0; i < MAX_BLOCKS; i++ {
+		x := int(curPos.X + curShape.Blocks[i].X)
+		y := int(curPos.Y + curShape.Blocks[i].Y)
+		if x < 0 || x >= BOARD_WIDTH || y >= BOARD_HEIGHT || (y >= 0 && board[y][x] != raylib.BLANK) {
+			return true
+		}
+	}
+	return false
+}
+
+func lockShape() {
+	for i := 0; i < MAX_BLOCKS; i++ {
+		x := int(curPos.X + curShape.Blocks[i].X)
+		y := int(curPos.Y + curShape.Blocks[i].Y)
+		if y >= 0 {
+			board[y][x] = curShape.Color
+		}
+	}
+}
+
+func rotateShape() {
+	rotated := curShape
+	for i := 0; i < MAX_BLOCKS; i++ {
+		x := rotated.Blocks[i].X
+		rotated.Blocks[i].X = -rotated.Blocks[i].Y
+		rotated.Blocks[i].Y = x
+	}
+
+	temp := curShape
+	curShape = rotated
+	if checkCollision() {
+		curShape = temp
+	}
+}
+
+func clearLines() int {
+	linesCleared := 0
+	for y := BOARD_HEIGHT - 1; y >= 0; y-- {
+		lineFull := true
+		for x := 0; x < BOARD_WIDTH; x++ {
+			if board[y][x] == raylib.BLANK {
+				lineFull = false
+				break
+			}
+		}
+		if lineFull {
+			for yy := y; yy > 0; yy-- {
+				for x := 0; x < BOARD_WIDTH; x++ {
+					board[yy][x] = board[yy-1][x]
+				}
+			}
+			for x := 0; x < BOARD_WIDTH; x++ {
+				board[0][x] = raylib.BLANK
+			}
+			y += 1
+			linesCleared += 1
+		}
+	}
+	return linesCleared
+}
+
+func keyPressed(key c.Int) bool {
+	return raylib.IsKeyPressed(key) || raylib.IsKeyPressedRepeat(key)
+}
+
+func main() {
+	raylib.InitWindow(SCREENWIDTH, SCREENHEIGHT, c.Str("tetris (powered by raylib + llgo)"))
+	raylib.SetTargetFPS(c.Int(60))
+	genShape()
+	for !raylib.WindowShouldClose() && !gameOver {
+		fallTime += raylib.GetFrameTime()
+		if fallTime >= fallSpeed {
+			fallTime = 0
+			curPos.Y += 1
+			if checkCollision() {
+				curPos.Y -= 1
+				lockShape()
+				linesCleared := clearLines()
+				score += linesCleared * 100
+				genShape()
+				if checkCollision() {
+					gameOver = true
+				}
+			}
+		}
+
+		if keyPressed(raylib.KEY_LEFT) {
+			curPos.X -= 1
+			if checkCollision() {
+				curPos.X += 1
+			}
+		}
+		if keyPressed(raylib.KEY_RIGHT) {
+			curPos.X += 1
+			if checkCollision() {
+				curPos.X -= 1
+			}
+		}
+		if keyPressed(raylib.KEY_SPACE) || keyPressed(raylib.KEY_UP) || keyPressed(raylib.KEY_DOWN) {
+			rotateShape()
+		}
+
+		raylib.BeginDrawing()
+		raylib.ClearBackground(raylib.RAYWHITE)
+		for y := 0; y < BOARD_HEIGHT; y++ {
+			for x := 0; x < BOARD_WIDTH; x++ {
+				raylib.DrawRectangle(c.Int(x*BLOCK_SIZE), c.Int(y*BLOCK_SIZE), c.Int(BLOCK_SIZE-1), c.Int(BLOCK_SIZE-1), board[y][x])
+			}
+		}
+
+		for i := 0; i < MAX_BLOCKS; i++ {
+			raylib.DrawRectangle(c.Int((curPos.X+curShape.Blocks[i].X)*BLOCK_SIZE), c.Int((curPos.Y+curShape.Blocks[i].Y)*BLOCK_SIZE),
+				BLOCK_SIZE-1, BLOCK_SIZE-1, curShape.Color)
+		}
+
+		c.Sprintf(unsafe.SliceData(scoreText), c.Str("Score:%d"), score)
+		raylib.DrawText(unsafe.SliceData(scoreText), 10, 10, 20, raylib.BLACK)
+
+		raylib.EndDrawing()
+	}
+
+	for !raylib.WindowShouldClose() {
+		raylib.BeginDrawing()
+		raylib.ClearBackground(raylib.RAYWHITE)
+		raylib.DrawText(c.Str("Game Over"), SCREENWIDTH/2-50, SCREENHEIGHT/2-10, 20, raylib.RED)
+		raylib.DrawText(unsafe.SliceData(scoreText), SCREENWIDTH/2-50, SCREENHEIGHT/2+10, 20, raylib.BLACK)
+		raylib.EndDrawing()
+	}
+}

c/raylib/raylib.go

@@ -0,0 +1,485 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package raylib
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "link: $(pkg-config --libs raylib); -lraylib"
+)
+
+// -----------------------------------------------------------------------------
+
+// Vector2, 2 components
+type Vector2 struct {
+	X float32 // Vector x component
+	Y float32 // Vector y component
+}
+
+// Vector3, 3 components
+type Vector3 struct {
+	X float32 // Vector x component
+	Y float32 // Vector y component
+	Z float32 // Vector z component
+}
+
+// Vector4, 4 components
+type Vector4 struct {
+	X float32 // Vector x component
+	Y float32 // Vector y component
+	Z float32 // Vector z component
+	W float32 // Vector w component
+}
+
+// Quaternion, 4 components (Vector4 alias)
+type Quaternion = Vector4
+
+//----------------------------------------------------------------------------------
+// Enumerators Definition
+
+// System/Window config flags
+// By default all flags are set to 0
+
+// Trace log level
+
+// Keyboard keys (US keyboard layout)
+// required keys for alternative layouts
+const (
+	KEY_NULL = 0 // Key: NULL, used for no key pressed
+
+	// Alphanumeric keys
+	KEY_APOSTROPHE    = 39 // Key: '
+	KEY_COMMA         = 44 // Key: ,
+	KEY_MINUS         = 45 // Key: -
+	KEY_PERIOD        = 46 // Key: .
+	KEY_SLASH         = 47 // Key: /
+	KEY_ZERO          = 48 // Key: 0
+	KEY_ONE           = 49 // Key: 1
+	KEY_TWO           = 50 // Key: 2
+	KEY_THREE         = 51 // Key: 3
+	KEY_FOUR          = 52 // Key: 4
+	KEY_FIVE          = 53 // Key: 5
+	KEY_SIX           = 54 // Key: 6
+	KEY_SEVEN         = 55 // Key: 7
+	KEY_EIGHT         = 56 // Key: 8
+	KEY_NINE          = 57 // Key: 9
+	KEY_SEMICOLON     = 59 // Key: ;
+	KEY_EQUAL         = 61 // Key: =
+	KEY_A             = 65 // Key: A | a
+	KEY_B             = 66 // Key: B | b
+	KEY_C             = 67 // Key: C | c
+	KEY_D             = 68 // Key: D | d
+	KEY_E             = 69 // Key: E | e
+	KEY_F             = 70 // Key: F | f
+	KEY_G             = 71 // Key: G | g
+	KEY_H             = 72 // Key: H | h
+	KEY_I             = 73 // Key: I | i
+	KEY_J             = 74 // Key: J | j
+	KEY_K             = 75 // Key: K | k
+	KEY_L             = 76 // Key: L | l
+	KEY_M             = 77 // Key: M | m
+	KEY_N             = 78 // Key: N | n
+	KEY_O             = 79 // Key: O | o
+	KEY_P             = 80 // Key: P | p
+	KEY_Q             = 81 // Key: Q | q
+	KEY_R             = 82 // Key: R | r
+	KEY_S             = 83 // Key: S | s
+	KEY_T             = 84 // Key: T | t
+	KEY_U             = 85 // Key: U | u
+	KEY_V             = 86 // Key: V | v
+	KEY_W             = 87 // Key: W | w
+	KEY_X             = 88 // Key: X | x
+	KEY_Y             = 89 // Key: Y | y
+	KEY_Z             = 90 // Key: Z | z
+	KEY_LEFT_BRACKET  = 91 // Key: [
+	KEY_BACKSLASH     = 92 // Key: '\'
+	KEY_RIGHT_BRACKET = 93 // Key: ]
+	KEY_GRAVE         = 96 // Key: `
+
+	// Function keys
+	KEY_SPACE         = 32  // Key: Space
+	KEY_ESCAPE        = 256 // Key: Esc
+	KEY_ENTER         = 257 // Key: Enter
+	KEY_TAB           = 258 // Key: Tab
+	KEY_BACKSPACE     = 259 // Key: Backspace
+	KEY_INSERT        = 260 // Key: Ins
+	KEY_DELETE        = 261 // Key: Del
+	KEY_RIGHT         = 262 // Key: Cursor right
+	KEY_LEFT          = 263 // Key: Cursor left
+	KEY_DOWN          = 264 // Key: Cursor down
+	KEY_UP            = 265 // Key: Cursor up
+	KEY_PAGE_UP       = 266 // Key: Page up
+	KEY_PAGE_DOWN     = 267 // Key: Page down
+	KEY_HOME          = 268 // Key: Home
+	KEY_END           = 269 // Key: End
+	KEY_CAPS_LOCK     = 280 // Key: Caps lock
+	KEY_SCROLL_LOCK   = 281 // Key: Scroll down
+	KEY_NUM_LOCK      = 282 // Key: Num lock
+	KEY_PRINT_SCREEN  = 283 // Key: Print screen
+	KEY_PAUSE         = 284 // Key: Pause
+	KEY_F1            = 290 // Key: F1
+	KEY_F2            = 291 // Key: F2
+	KEY_F3            = 292 // Key: F3
+	KEY_F4            = 293 // Key: F4
+	KEY_F5            = 294 // Key: F5
+	KEY_F6            = 295 // Key: F6
+	KEY_F7            = 296 // Key: F7
+	KEY_F8            = 297 // Key: F8
+	KEY_F9            = 298 // Key: F9
+	KEY_F10           = 299 // Key: F10
+	KEY_F11           = 300 // Key: F11
+	KEY_F12           = 301 // Key: F12
+	KEY_LEFT_SHIFT    = 340 // Key: Shift left
+	KEY_LEFT_CONTROL  = 341 // Key: Control left
+	KEY_LEFT_ALT      = 342 // Key: Alt left
+	KEY_LEFT_SUPER    = 343 // Key: Super left
+	KEY_RIGHT_SHIFT   = 344 // Key: Shift right
+	KEY_RIGHT_CONTROL = 345 // Key: Control right
+	KEY_RIGHT_ALT     = 346 // Key: Alt right
+	KEY_RIGHT_SUPER   = 347 // Key: Super right
+	KEY_KB_MENU       = 348 // Key: KB menu
+
+	// Keypad keys
+	KEY_KP_0        = 320 // Key: Keypad 0
+	KEY_KP_1        = 321 // Key: Keypad 1
+	KEY_KP_2        = 322 // Key: Keypad 2
+	KEY_KP_3        = 323 // Key: Keypad 3
+	KEY_KP_4        = 324 // Key: Keypad 4
+	KEY_KP_5        = 325 // Key: Keypad 5
+	KEY_KP_6        = 326 // Key: Keypad 6
+	KEY_KP_7        = 327 // Key: Keypad 7
+	KEY_KP_8        = 328 // Key: Keypad 8
+	KEY_KP_9        = 329 // Key: Keypad 9
+	KEY_KP_DECIMAL  = 330 // Key: Keypad .
+	KEY_KP_DIVIDE   = 331 // Key: Keypad /
+	KEY_KP_MULTIPLY = 332 // Key: Keypad *
+	KEY_KP_SUBTRACT = 333 // Key: Keypad -
+	KEY_KP_ADD      = 334 // Key: Keypad +
+	KEY_KP_ENTER    = 335 // Key: Keypad Enter
+	KEY_KP_EQUAL    = 336 // Key: Keypad =
+
+	// Android key buttons
+	KEY_BACK        = 4  // Key: Android back button
+	KEY_MENU        = 82 // Key: Android menu button
+	KEY_VOLUME_UP   = 24 // Key: Android volume up button
+	KEY_VOLUME_DOWN = 25 // Key: Android volume down button
+)
+
+// Mouse buttons
+
+// Mouse cursor
+
+// Gamepad buttons
+
+// Gamepad axis
+
+// Material map index
+
+// Shader location index
+
+// Shader uniform data type
+
+// Shader attribute data types
+
+// Pixel formats
+
+// Texture parameters: filter mode
+
+// Texture parameters: wrap mode
+
+// Cubemap layouts
+
+// Font type, defines generation method
+
+// Color blending modes (pre-defined)
+
+// Gesture
+
+// Camera system modes
+
+// Camera projection
+
+// N-patch layout
+
+// -----------------------------------------------------------------------------
+// Window-related functions
+
+//go:linkname InitWindow C.InitWindow
+func InitWindow(width, height c.Int, title *c.Char)
+
+//go:linkname CloseWindow C.CloseWindow
+func CloseWindow()
+
+//go:linkname WindowShouldClose C.WindowShouldClose
+func WindowShouldClose() bool
+
+//go:linkname IsWindowReady C.IsWindowReady
+func IsWindowReady() bool
+
+//go:linkname IsWindowFullscreen C.IsWindowFullscreen
+func IsWindowFullscreen() bool
+
+//go:linkname IsWindowHidden C.IsWindowHidden
+func IsWindowHidden() bool
+
+//go:linkname IsWindowMinimized C.IsWindowMinimized
+func IsWindowMinimized() bool
+
+//go:linkname IsWindowMaximized C.IsWindowMaximized
+func IsWindowMaximized() bool
+
+//go:linkname IsWindowFocused C.IsWindowFocused
+func IsWindowFocused() bool
+
+//go:linkname IsWindowResized C.IsWindowResized
+func IsWindowResized() bool
+
+//go:linkname IsWindowState C.IsWindowState
+func IsWindowState(flag c.Int) bool
+
+//go:linkname SetWindowState C.SetWindowState
+func SetWindowState(flags c.Int)
+
+//go:linkname ClearWindowState C.ClearWindowState
+func ClearWindowState(flags c.Int)
+
+//go:linkname ToggleFullscreen C.ToggleFullscreen
+func ToggleFullscreen()
+
+//go:linkname MaximizeWindow C.MaximizeWindow
+func MaximizeWindow()
+
+//go:linkname MinimizeWindow C.MinimizeWindow
+func MinimizeWindow()
+
+//go:linkname RestoreWindow C.RestoreWindow
+func RestoreWindow()
+
+//go:linkname SetWindowIcon C.SetWindowIcon
+func SetWindowIcon(icon Image)
+
+//go:linkname SetWindowIcons C.SetWindowIcons
+func SetWindowIcons(icons *Image, count c.Int)
+
+//go:linkname SetWindowTitle C.SetWindowTitle
+func SetWindowTitle(title *c.Char)
+
+//go:linkname SetWindowPosition C.SetWindowPosition
+func SetWindowPosition(x, y c.Int)
+
+//go:linkname SetWindowMonitor C.SetWindowMonitor
+func SetWindowMonitor(monitor c.Int)
+
+//go:linkname SetWindowMinSize C.SetWindowMinSize
+func SetWindowMinSize(width, height c.Int)
+
+//go:linkname SetWindowSize C.SetWindowSize
+func SetWindowSize(width, height c.Int)
+
+//go:linkname SetWindowOpacity C.SetWindowOpacity
+func SetWindowOpacity(opacity c.Float)
+
+//go:linkname SetWindowFocused C.SetWindowFocused
+func SetWindowFocused()
+
+//go:linkname GetWindowHandle C.GetWindowHandle
+func GetWindowHandle() c.Pointer
+
+//go:linkname GetScreenWidth C.GetScreenWidth
+func GetScreenWidth() c.Int
+
+//go:linkname GetScreenHeight C.GetScreenHeight
+func GetScreenHeight() c.Int
+
+// Get current render width (it considers HiDPI)
+//
+//go:linkname GetRenderWidth C.GetRenderWidth
+func GetRenderWidth() c.Int
+
+// Get current render height (it considers HiDPI)
+//
+//go:linkname GetRenderHeight C.GetRenderHeight
+func GetRenderHeight() c.Int
+
+// Get number of connected monitors
+//
+//go:linkname GetMonitorCount C.GetMonitorCount
+func GetMonitorCount() c.Int
+
+// Get current connected monitor
+//
+//go:linkname GetCurrentMonitor C.GetCurrentMonitor
+func GetCurrentMonitor() c.Int
+
+// Get specified monitor position
+//
+//go:linkname GetMonitorPosition C.GetMonitorPosition
+func GetMonitorPosition(monitor c.Int) Vector2
+
+//go:linkname GetMonitorWidth C.GetMonitorWidth
+func GetMonitorWidth(monitor c.Int) c.Int
+
+//go:linkname GetMonitorHeight C.GetMonitorHeight
+func GetMonitorHeight(monitor c.Int) c.Int
+
+//go:linkname GetMonitorPhysicalWidth C.GetMonitorPhysicalWidth
+func GetMonitorPhysicalWidth(monitor c.Int) c.Int
+
+//go:linkname GetMonitorPhysicalHeight C.GetMonitorPhysicalHeight
+func GetMonitorPhysicalHeight(monitor c.Int) c.Int
+
+//go:linkname GetMonitorRefreshRate C.GetMonitorRefreshRate
+func GetMonitorRefreshRate(monitor c.Int) c.Int
+
+//go:linkname GetWindowPosition C.GetWindowPosition
+func GetWindowPosition() Vector2
+
+//go:linkname GetWindowScaleDPI C.GetWindowScaleDPI
+func GetWindowScaleDPI() Vector2
+
+//go:linkname GetMonitorName C.GetMonitorName
+func GetMonitorName(monitor c.Int) *c.Char
+
+//go:linkname SetClipboardText C.SetClipboardText
+func SetClipboardText(text *c.Char)
+
+//go:linkname GetClipboardText C.GetClipboardText
+func GetClipboardText() *c.Char
+
+// Enable waiting for events on EndDrawing(), no automatic event polling
+//
+//go:linkname EnableEventWaiting C.EnableEventWaiting
+func EnableEventWaiting()
+
+// Disable waiting for events on EndDrawing(), automatic events polling
+//
+//go:linkname DisableEventWaiting C.DisableEventWaiting
+func DisableEventWaiting()
+
+// -----------------------------------------------------------------------------
+// Cursor-related functions
+
+//go:linkname ShowCursor C.ShowCursor
+func ShowCursor()
+
+//go:linkname HideCursor C.HideCursor
+func HideCursor()
+
+//go:linkname IsCursorHidden C.IsCursorHidden
+func IsCursorHidden() bool
+
+//go:linkname EnableCursor C.EnableCursor
+func EnableCursor()
+
+//go:linkname DisableCursor C.DisableCursor
+func DisableCursor()
+
+//go:linkname IsCursorOnScreen C.IsCursorOnScreen
+func IsCursorOnScreen() bool
+
+// -----------------------------------------------------------------------------
+// Shader management functions
+
+// -----------------------------------------------------------------------------
+// Screen-space-related functions
+
+// -----------------------------------------------------------------------------
+// Timing-related functions
+
+// Set target FPS (maximum)
+//
+//go:linkname SetTargetFPS C.SetTargetFPS
+func SetTargetFPS(fps c.Int)
+
+// Returns current FPS
+//
+//go:linkname GetFPS C.GetFPS
+func GetFPS() c.Int
+
+// Returns time in seconds for last frame drawn (delta time)
+//
+//go:linkname GetFrameTime C.GetFrameTime
+func GetFrameTime() c.Float
+
+// Returns elapsed time in seconds since InitWindow()
+//
+//go:linkname GetTime C.GetTime
+func GetTime() c.Double
+
+// -----------------------------------------------------------------------------
+// Custom frame control functions
+
+// NOTE: Those functions are intended for advance users that want full control over the frame processing
+// By default EndDrawing() does this job: draws everything + SwapScreenBuffer() + manage frame timing + PollInputEvents()
+// To avoid that behaviour and control frame processes manually, enable in config.h: SUPPORT_CUSTOM_FRAME_CONTROL
+
+// -----------------------------------------------------------------------------
+// Random values generation functions
+
+//go:linkname SetRandomSeed C.SetRandomSeed
+func SetRandomSeed(seed c.Uint)
+
+//go:linkname GetRandomValue C.GetRandomValue
+func GetRandomValue(min c.Int, max c.Int) c.Int
+
+// -----------------------------------------------------------------------------
+// Misc. functions
+
+// -----------------------------------------------------------------------------
+// Input-related functions: keyboard
+
+//go:linkname IsKeyPressed C.IsKeyPressed
+func IsKeyPressed(key c.Int) bool
+
+//go:linkname IsKeyPressedRepeat C.IsKeyPressedRepeat
+func IsKeyPressedRepeat(key c.Int) bool
+
+//go:linkname IsKeyDown C.IsKeyDown
+func IsKeyDown(key c.Int) bool
+
+//go:linkname IsKeyReleased C.IsKeyReleased
+func IsKeyReleased(key c.Int) bool
+
+//go:linkname IsKeyUp C.IsKeyUp
+func IsKeyUp(key c.Int) bool
+
+//go:linkname GetKeyPressed C.GetKeyPressed
+func GetKeyPressed() c.Int
+
+//go:linkname GetCharPressed C.GetCharPressed
+func GetCharPressed() c.Int
+
+//go:linkname SetExitKey C.SetExitKey
+func SetExitKey(key c.Int)
+
+// -----------------------------------------------------------------------------
+// Input-related functions: gamepads
+
+// -----------------------------------------------------------------------------
+// Input-related functions: mouse
+
+// -----------------------------------------------------------------------------
+// Input-related functions: touch
+
+// -----------------------------------------------------------------------------
+// Gestures and Touch Handling Functions (Module: rgestures)
+
+// -----------------------------------------------------------------------------

c/raylib/shape.go

@@ -0,0 +1,215 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package raylib
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+// -----------------------------------------------------------------------------
+
+// Color, 4 components, R8G8B8A8 (32bit)
+// R, G, B, A uint8
+type Color uint32
+
+const (
+	LIGHTGRAY = Color(200 | 200<<8 | 200<<16 | 255<<24) // Light Gray
+	GRAY      = Color(130 | 130<<8 | 130<<16 | 255<<24) // Gray
+	DARKGRAY  = Color(80 | 80<<8 | 80<<16 | 255<<24)    // Dark Gray
+	YELLOW    = Color(253 | 249<<8 | 0<<16 | 255<<24)   // Yellow
+	GOLD      = Color(255 | 203<<8 | 0<<16 | 255<<24)   // Gold
+	ORANGE    = Color(255 | 161<<8 | 0<<16 | 255<<24)   // Orange
+	PINK      = Color(255 | 109<<8 | 194<<16 | 255<<24) // Pink
+	RED       = Color(230 | 41<<8 | 55<<16 | 255<<24)   // Red
+	MAROON    = Color(190 | 33<<8 | 55<<16 | 255<<24)   // Maroon
+	GREEN     = Color(0 | 228<<8 | 48<<16 | 255<<24)    // Green
+	LIME      = Color(0 | 158<<8 | 47<<16 | 255<<24)    // Lime
+	DARKGREEN = Color(0 | 117<<8 | 44<<16 | 255<<24)    // Dark Green
+	SKYBLUE   = Color(102 | 191<<8 | 255<<16 | 255<<24) // Sky Blue
+	BLUE      = Color(0 | 121<<8 | 241<<16 | 255<<24)   // Blue
+
+	DARKBLUE   = Color(0 | 82<<8 | 172<<16 | 255<<24)    // Dark Blue
+	PURPLE     = Color(200 | 122<<8 | 255<<16 | 255<<24) // Purple
+	VIOLET     = Color(135 | 60<<8 | 190<<16 | 255<<24)  // Violet
+	DARKPURPLE = Color(112 | 31<<8 | 126<<16 | 255<<24)  // Dark Purple
+	BEIGE      = Color(211 | 176<<8 | 131<<16 | 255<<24) // Beige
+	BROWN      = Color(127 | 106<<8 | 79<<16 | 255<<24)  // Brown
+	DARKBROWN  = Color(76 | 63<<8 | 47<<16 | 255<<24)    // Dark Brown
+
+	WHITE    = Color(255 | 255<<8 | 255<<16 | 255<<24) // White
+	BLACK    = Color(0 | 0<<8 | 0<<16 | 255<<24)       // Black
+	BLANK    = Color(0 | 0<<8 | 0<<16 | 0<<24)         // Blank (Transparent)
+	MAGENTA  = Color(255 | 0<<8 | 255<<16 | 255<<24)   // Magenta
+	RAYWHITE = Color(245 | 245<<8 | 245<<16 | 255<<24) // My own White (raylib logo)
+)
+
+// Image, pixel data stored in CPU memory (RAM)
+type Image struct {
+	Data    c.Pointer // Image raw data
+	Width   c.Int     // Image base width
+	Height  c.Int     // Image base height
+	Mipmaps c.Int     // Mipmap levels, 1 by default
+	Format  c.Int     // Data format (PixelFormat type)
+}
+
+// Camera, defines position/orientation in 3d space
+type Camera3D struct {
+	Position   Vector3 // Camera position
+	Target     Vector3 // Camera target it looks-at
+	Up         Vector3 // Camera up vector (rotation over its axis)
+	Fovy       float32 // Camera field-of-view aperture in Y (degrees) in perspective, used as near plane width in orthographic
+	Projection c.Int   // Camera projection: CAMERA_PERSPECTIVE or CAMERA_ORTHOGRAPHIC
+}
+
+type Camera = Camera3D // Camera type fallback, defaults to Camera3D
+
+// Camera2D, defines position/orientation in 2d space
+type Camera2D struct {
+	Offset   Vector2 // Camera offset (displacement from target)
+	Target   Vector2 // Camera target (rotation and zoom origin)
+	Rotation float32 // Camera rotation in degrees
+	Zoom     float32 // Camera zoom (scaling), should be 1.0f by default
+}
+
+// Shader
+type Shader struct {
+	Id   c.Uint // Shader program id
+	Locs *c.Int // Shader locations array (RL_MAX_SHADER_LOCATIONS)
+}
+
+// -----------------------------------------------------------------------------
+// Drawing-related functions
+
+// Set background color (framebuffer clear color)
+//
+//go:linkname ClearBackground C.ClearBackground
+func ClearBackground(color Color)
+
+// Begin drawing (call before drawing anything)
+//
+//go:linkname BeginDrawing C.BeginDrawing
+func BeginDrawing()
+
+// End drawing and swap buffers (call after drawing)
+//
+//go:linkname EndDrawing C.EndDrawing
+func EndDrawing()
+
+// Begin 2D mode with custom camera (2D)
+//
+//go:linkname BeginMode2D C.BeginMode2D
+func BeginMode2D(camera Camera2D)
+
+// End mode (2D)
+//
+//go:linkname EndMode2D C.EndMode2D
+func EndMode2D()
+
+// Begin 3D mode with custom camera (3D)
+//
+//go:linkname BeginMode3D C.BeginMode3D
+func BeginMode3D(camera Camera3D)
+
+// End mode (3D)
+//
+//go:linkname EndMode3D C.EndMode3D
+func EndMode3D()
+
+// Begin drawing to render texture
+//-go:linkname BeginTextureMode C.BeginTextureMode
+//func BeginTextureMode(target RenderTexture2D)
+
+// End drawing to render texture
+//
+//go:linkname EndTextureMode C.EndTextureMode
+func EndTextureMode()
+
+// Begin custom shader drawing
+//
+//go:linkname BeginShaderMode C.BeginShaderMode
+func BeginShaderMode(shader Shader)
+
+// End custom shader drawing (use default shader)
+//
+//go:linkname EndShaderMode C.EndShaderMode
+func EndShaderMode()
+
+// Color blending modes (pre-defined)
+type BlendMode c.Int
+
+const (
+	BLEND_ALPHA             BlendMode = iota // Blend textures considering alpha (default)
+	BLEND_ADDITIVE                           // Blend textures adding colors
+	BLEND_MULTIPLIED                         // Blend textures multiplying colors
+	BLEND_ADD_COLORS                         // Blend textures adding colors (alternative)
+	BLEND_SUBTRACT_COLORS                    // Blend textures subtracting colors (alternative)
+	BLEND_ALPHA_PREMULTIPLY                  // Blend premultiplied textures considering alpha
+	BLEND_CUSTOM                             // Blend textures using custom src/dst factors (use rlSetBlendFactors())
+	BLEND_CUSTOM_SEPARATE                    // Blend textures using custom rgb/alpha separate src/dst factors (use rlSetBlendFactorsSeparate())
+)
+
+// Begin blending mode (alpha, additive, multiplied, subtract, custom)
+//
+//go:linkname BeginBlendMode C.BeginBlendMode
+func BeginBlendMode(mode BlendMode)
+
+// End blending mode (reset to default: alpha blending)
+//
+//go:linkname EndBlendMode C.EndBlendMode
+func EndBlendMode()
+
+// Begin scissor mode (define screen area for following drawing)
+//
+//go:linkname BeginScissorMode C.BeginScissorMode
+func BeginScissorMode(x, y, width, height c.Int)
+
+// End scissor mode
+//
+//go:linkname EndScissorMode C.EndScissorMode
+func EndScissorMode()
+
+// -----------------------------------------------------------------------------
+// VR stereo config functions for VR simulator
+
+// -----------------------------------------------------------------------------
+// Camera System Functions (Module: rcamera)
+
+// Update camera position for selected mode
+//
+//go:linkname UpdateCamera C.UpdateCamera
+func UpdateCamera(camera *Camera, mode c.Int)
+
+// Update camera movement/rotation
+//
+//go:linkname UpdateCameraPro C.UpdateCameraPro
+func UpdateCameraPro(camera *Camera, movement, rotation Vector3, zoom float32)
+
+// -----------------------------------------------------------------------------
+
+// Draw a color-filled rectangle
+//
+//go:linkname DrawRectangle C.DrawRectangle
+func DrawRectangle(posX, posY, width, height c.Int, color Color)
+
+// Draw text (using default font)
+//
+//go:linkname DrawText C.DrawText
+func DrawText(text *c.Char, posX, posY, fontSize c.Int, color Color)
+
+// -----------------------------------------------------------------------------

c/raylib/utils.go

@@ -0,0 +1,67 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package raylib
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+// -----------------------------------------------------------------------------
+
+// Show trace log messages (LOG_DEBUG, LOG_INFO, LOG_WARNING, LOG_ERROR...)
+//
+//go:linkname TraceLog C.TraceLog
+func TraceLog(logLevel int, text *c.Char, __llgo_va_list ...any)
+
+// Set the current threshold (minimum) log level
+//
+//go:linkname SetTraceLogLevel C.SetTraceLogLevel
+func SetTraceLogLevel(logLevel int)
+
+// -----------------------------------------------------------------------------
+// Set custom callbacks
+
+// -----------------------------------------------------------------------------
+// Files management functions
+
+// -----------------------------------------------------------------------------
+// File system functions
+
+// Check if file exists
+//
+//go:linkname FileExists C.FileExists
+func FileExists(fileName *c.Char) bool
+
+// Check if a directory path exists
+//
+//go:linkname DirectoryExists C.DirectoryExists
+func DirectoryExists(dirPath *c.Char) bool
+
+// Check file extension (including point: .png, .wav)
+//
+//go:linkname IsFileExtension C.IsFileExtension
+func IsFileExtension(fileName *c.Char, ext *c.Char) bool
+
+// -----------------------------------------------------------------------------
+// Compression/Encoding functionality
+
+// -----------------------------------------------------------------------------
+// Automation events functionality
+
+// -----------------------------------------------------------------------------

c/setjmp/setjmp.go

@@ -0,0 +1,53 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package setjmp
+
+// #include <setjmp.h>
+import "C"
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "decl"
+)
+
+type (
+	JmpBuf    = C.jmp_buf
+	SigjmpBuf = C.sigjmp_buf
+)
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Setjmp C.setjmp
+func Setjmp(env *JmpBuf) c.Int
+
+//go:linkname Longjmp C.longjmp
+func Longjmp(env *JmpBuf, val c.Int)
+
+// -----------------------------------------------------------------------------
+
+//go:linkname Sigsetjmp C.sigsetjmp
+func Sigsetjmp(env *SigjmpBuf, savemask c.Int) c.Int
+
+//go:linkname Siglongjmp C.siglongjmp
+func Siglongjmp(env *SigjmpBuf, val c.Int)
+
+// -----------------------------------------------------------------------------

c/setjmp/trycatch/_code/demo.cpp

@@ -0,0 +1,5 @@
+#include <stdexcept>
+
+extern "C" void throwCppException() {
+    throw std::runtime_error("C++ exception");
+}

c/setjmp/trycatch/_code/try_catch.cpp

@@ -0,0 +1,13 @@
+#include <exception>
+#include <stdio.h>
+
+extern "C" void throwCppException();
+
+int main() {
+    try {
+        throwCppException();
+    } catch (std::exception& e) {
+        printf("Hi, %s\n", e.what());
+    }
+    return 0;
+}

c/setjmp/trycatch/demo.go

@@ -0,0 +1,28 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package trycatch
+
+import (
+	_ "unsafe"
+)
+
+const (
+	LLGoPackage = "link: c++"
+)
+
+//go:linkname ThrowCppException C.throwCppException
+func ThrowCppException()

c/setjmp/trycatch/llgo.cfg

@@ -0,0 +1,8 @@
+{
+    "cl": [
+        "clang -emit-llvm -S -o demo.ll -c _code/demo.cpp",
+        "clang -emit-llvm -S -o _code/llgo_autogen.ll -c _code/try_catch.cpp",
+        "llgen .",
+        "rm llgo_autogen.lla; zip llgo_autogen.lla llgo_autogen.ll demo.ll",
+    ]
+}

c/sqlite/README.md

@@ -1,22 +1,18 @@
 LLGo wrapper of sqlite
 =====
-[![Build Status](https://github.com/goplus/sqlite/actions/workflows/go.yml/badge.svg)](https://github.com/goplus/sqlite/actions/workflows/go.yml)
-[![GitHub release](https://img.shields.io/github/v/tag/goplus/sqlite.svg?label=release)](https://github.com/goplus/sqlite/releases)
-[![GoDoc](https://pkg.go.dev/badge/github.com/goplus/sqlite.svg)](https://pkg.go.dev/github.com/goplus/sqlite)
-[![Compiler](https://img.shields.io/badge/compiler-llgo-darkgreen.svg)](https://github.com/goplus/llgo)
-[![Language](https://img.shields.io/badge/language-Go+-blue.svg)](https://github.com/goplus/gop)
 
 ## How to install
 
+### on macOS (Homebrew)
+
 ```sh
-git clone https://github.com/goplus/sqlite.git
-cd sqlite
-git submodule init
-git submodule update
-mkdir build.dir
-cd build.dir
-../sqlite/configure --enable-shared
-sudo make install
+brew install sqlite3
+```
+
+### on Linux (Debian/Ubuntu)
+
+```sh
+apt-get install -y libsqlite3-dev
 ```
 
 ## Demos

c/sqlite/_demo/sqlitedemo/demo.go

@@ -2,11 +2,12 @@ package main
 
 import (
 	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/os"
 	"github.com/goplus/llgo/c/sqlite"
 )
 
 func main() {
-	c.Remove(c.Str("test.db"))
+	os.Remove(c.Str("test.db"))
 
 	db, err := sqlite.Open(c.Str("test.db"))
 	check(err, db, "sqlite: Open")

c/sqlite/sqlite.go

@@ -23,7 +23,7 @@ import (
 )
 
 const (
-	LLGoPackage = "link: sqlite3"
+	LLGoPackage = "link: $(pkg-config --libs sqlite3); -lsqlite3"
 )
 
 // llgo:type C

c/sync/atomic/atomic.go

@@ -0,0 +1,72 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package atomic
+
+import (
+	"unsafe"
+	_ "unsafe"
+)
+
+const (
+	LLGoPackage = "decl"
+)
+
+type valtype interface {
+	~int | ~uint | ~uintptr | ~int32 | ~uint32 | ~int64 | ~uint64 | ~unsafe.Pointer
+}
+
+// llgo:link Add llgo.atomicAdd
+func Add[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link Sub llgo.atomicSub
+func Sub[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link And llgo.atomicAnd
+func And[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link NotAnd llgo.atomicNand
+func NotAnd[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link Or llgo.atomicOr
+func Or[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link Xor llgo.atomicXor
+func Xor[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link Max llgo.atomicMax
+func Max[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link Min llgo.atomicMin
+func Min[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link UMax llgo.atomicUMax
+func UMax[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link UMin llgo.atomicUMin
+func UMin[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link Load llgo.atomicLoad
+func Load[T valtype](ptr *T) T { return *ptr }
+
+// llgo:link Store llgo.atomicStore
+func Store[T valtype](ptr *T, v T) {}
+
+// llgo:link Exchange llgo.atomicXchg
+func Exchange[T valtype](ptr *T, v T) T { return v }
+
+// llgo:link CompareAndExchange llgo.atomicCmpXchg
+func CompareAndExchange[T valtype](ptr *T, old, new T) (T, bool) { return old, false }

c/time/time.go

@@ -0,0 +1,132 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package time
+
+// #include <time.h>
+import "C"
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "decl"
+)
+
+// -----------------------------------------------------------------------------
+
+type TimeT C.time_t
+
+//go:linkname Time C.time
+func Time(timer *TimeT) TimeT
+
+//go:linkname Mktime C.mktime
+func Mktime(timer *Tm) TimeT
+
+//go:linkname Ctime C.ctime
+func Ctime(timer *TimeT) string
+
+//go:linkname Difftime C.difftime
+func Difftime(end, start TimeT) float64
+
+// -----------------------------------------------------------------------------
+
+type Tm struct {
+	Sec    c.Int
+	Min    c.Int
+	Hour   c.Int
+	Mday   c.Int
+	Mon    c.Int
+	Year   c.Int
+	Wday   c.Int
+	Yday   c.Int
+	Isdst  c.Int
+	Gmtoff c.Long
+	Zone   *c.Char
+}
+
+//go:linkname Gmtime C.gmtime
+func Gmtime(timer *TimeT) *Tm
+
+//go:linkname Localtime C.localtime
+func Localtime(timer *TimeT) *Tm
+
+//go:linkname Strftime C.strftime
+func Strftime(buf *c.Char, bufSize uintptr, format *c.Char, timeptr *Tm) uintptr
+
+// -----------------------------------------------------------------------------
+
+type ClockT C.clock_t
+
+//go:linkname Clock C.clock
+func Clock() ClockT
+
+// -----------------------------------------------------------------------------
+
+type ClockidT C.clockid_t
+
+const (
+	// the system's real time (i.e. wall time) clock, expressed as the amount of time since the Epoch.
+	// This is the same as the value returned by gettimeofday
+	CLOCK_REALTIME = ClockidT(C.CLOCK_REALTIME)
+
+	// clock that increments monotonically, tracking the time since an arbitrary point, and will continue
+	// to increment while the system is asleep.
+	CLOCK_MONOTONIC = ClockidT(C.CLOCK_MONOTONIC)
+
+	// clock that increments monotonically, tracking the time since an arbitrary point like CLOCK_MONOTONIC.
+	// However, this clock is unaffected by frequency or time adjustments.  It should not be compared to
+	// other system time sources.
+	CLOCK_MONOTONIC_RAW = ClockidT(C.CLOCK_MONOTONIC_RAW)
+
+	// like CLOCK_MONOTONIC_RAW, but reads a value cached by the system at context switch. This can be
+	// read faster, but at a loss of accuracy as it may return values that are milliseconds old.
+	// CLOCK_MONOTONIC_RAW_APPROX = ClockidT(C.CLOCK_MONOTONIC_RAW_APPROX)
+
+	// clock that increments monotonically, in the same manner as CLOCK_MONOTONIC_RAW, but that does
+	// not increment while the system is asleep. The returned value is identical to the result of
+	// mach_absolute_time() after the appropriate mach_timebase conversion is applied.
+	// CLOCK_UPTIME_RAW = ClockidT(C.CLOCK_UPTIME_RAW)
+
+	// like CLOCK_UPTIME_RAW, but reads a value cached by the system at context switch. This can be read
+	// faster, but at a loss of accuracy as it may return values that are milliseconds old.
+	// CLOCK_UPTIME_RAW_APPROX = ClockidT(C.CLOCK_UPTIME_RAW_APPROX)
+
+	// clock that tracks the amount of CPU (in user- or kernel-mode) used by the calling process.
+	CLOCK_PROCESS_CPUTIME_ID = ClockidT(C.CLOCK_PROCESS_CPUTIME_ID)
+
+	// clock that tracks the amount of CPU (in user- or kernel-mode) used by the calling thread.
+	CLOCK_THREAD_CPUTIME_ID = ClockidT(C.CLOCK_THREAD_CPUTIME_ID)
+)
+
+type Timespec struct {
+	Sec  TimeT  // seconds
+	Nsec c.Long // and nanoseconds
+}
+
+//go:linkname ClockGettime C.clock_gettime
+func ClockGettime(clkId ClockidT, tp *Timespec) c.Int
+
+//go:linkname ClockSettime C.clock_settime
+func ClockSettime(clkId ClockidT, tp *Timespec) c.Int
+
+//go:linkname ClockGetres C.clock_getres
+func ClockGetres(clkId ClockidT, res *Timespec) c.Int
+
+// -----------------------------------------------------------------------------

c/zlib/_demo/efficiency/compress.go

@@ -0,0 +1,35 @@
+package main
+
+import (
+	"unsafe"
+
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/zlib"
+)
+
+func main() {
+	txt := []byte("zlib is a software library used for data compression. It was created by Jean-loup Gailly and Mark Adler and first released in 1995. zlib is designed to be a free, legally unencumbered—that is, not covered by any patents—alternative to the proprietary DEFLATE compression algorithm, which is often used in software applications for data compression.The library provides functions to compress and decompress data using the DEFLATE algorithm, which is a combination of the LZ77 algorithm and Huffman coding. zlib is notable for its versatility; it can be used in a wide range of applications, from web servers and web clients compressing HTTP data, to the compression of data for storage or transmission in various file formats, such as PNG, ZIP, and GZIP.")
+	txtLen := c.Ulong(len(txt))
+
+	for level := 0; level <= 9; level++ {
+		cmpSize := zlib.CompressBound(txtLen)
+		cmpData := make([]byte, int(cmpSize))
+
+		res := zlib.Compress2(unsafe.SliceData(cmpData), &cmpSize, unsafe.SliceData(txt), txtLen, c.Int(level))
+		if res != zlib.OK {
+			c.Printf(c.Str("\nCompression failed at level %d: %d\n"), level, res)
+			continue
+		}
+
+		c.Printf(c.Str("Compression level %d: Text length = %d, Compressed size = %d\n"), level, txtLen, cmpSize)
+
+		ucmpSize := txtLen
+		ucmpData := make([]byte, int(ucmpSize))
+
+		unRes := zlib.Uncompress(unsafe.SliceData(ucmpData), &ucmpSize, unsafe.SliceData(cmpData), cmpSize)
+		if unRes != zlib.OK {
+			c.Printf(c.Str("\nDecompression failed at level %d: %d\n"), level, unRes)
+			continue
+		}
+	}
+}

c/zlib/_demo/normal/compress.go

@@ -0,0 +1,46 @@
+package main
+
+import (
+	"unsafe"
+
+	"github.com/goplus/llgo/c"
+	"github.com/goplus/llgo/c/zlib"
+)
+
+func main() {
+	txt := []byte("zlib is a software library used for data compression. It was created by Jean-loup Gailly and Mark Adler and first released in 1995. zlib is designed to be a free, legally unencumbered—that is, not covered by any patents—alternative to the proprietary DEFLATE compression algorithm, which is often used in software applications for data compression.The library provides functions to compress and decompress data using the DEFLATE algorithm, which is a combination of the LZ77 algorithm and Huffman coding. zlib is notable for its versatility; it can be used in a wide range of applications, from web servers and web clients compressing HTTP data, to the compression of data for storage or transmission in various file formats, such as PNG, ZIP, and GZIP.")
+	txtLen := c.Ulong(len(txt))
+
+	cmpSize := zlib.CompressBound(txtLen)
+	cmpData := make([]byte, int(cmpSize))
+
+	res := zlib.Compress(unsafe.SliceData(cmpData), &cmpSize, unsafe.SliceData(txt), txtLen)
+	if res != zlib.OK {
+		c.Printf(c.Str("\nCompression failed: %d\n"), res)
+		return
+	}
+
+	c.Printf(c.Str("Text length = %d, Compressed size = %d\n"), txtLen, cmpSize)
+
+	ucmpSize := txtLen
+	ucmpData := make([]byte, int(ucmpSize))
+
+	unRes := zlib.Uncompress(unsafe.SliceData(ucmpData), &ucmpSize, unsafe.SliceData(cmpData), cmpSize)
+	c.Printf(c.Str("Decompression result = %d, Decompressed size %d\n"), unRes, ucmpSize)
+
+	if unRes != zlib.OK {
+		c.Printf(c.Str("\nDecompression failed: %d\n"), unRes)
+		return
+	}
+
+	c.Printf(c.Str("Decompressed data: \n"))
+	for i := 0; i < int(ucmpSize); i++ {
+		c.Printf(c.Str("%c"), ucmpData[i])
+	}
+}
+
+/* Expected output:
+origin textLen = 73 compressed_size = 36
+uncompress result = 0 uncompress result size 73
+after uncompressed data: Hello, zlib compression!Hello, zlib compression!Hello, zlib compression!
+*/

c/zlib/zlib.go

@@ -0,0 +1,92 @@
+/*
+ * Copyright (c) 2024 The GoPlus Authors (goplus.org). All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package zlib
+
+import (
+	_ "unsafe"
+
+	"github.com/goplus/llgo/c"
+)
+
+const (
+	LLGoPackage = "link: $(pkg-config --libs zlib); -lz"
+)
+
+/* errno */
+const (
+	OK            = 0
+	STREAM_END    = 1
+	NEED_DICT     = 2
+	ERRNO         = -1
+	STREAM_ERROR  = -2
+	DATA_ERROR    = -3
+	MEM_ERROR     = -4
+	BUF_ERROR     = -5
+	VERSION_ERROR = -6
+)
+
+/* compression levels */
+const (
+	NO_COMPRESSION      = 0
+	BEST_SPEED          = 1
+	BEST_COMPRESSION    = 9
+	DEFAULT_COMPRESSION = -1
+)
+
+const (
+	NO_FLUSH      = 0
+	PARTIAL_FLUSH = 1
+	SYNC_FLUSH    = 2
+	FULL_FLUSH    = 3
+	FINISH        = 4
+	BLOCK         = 5
+	TREES         = 6
+)
+
+const (
+	FILTERED         = 1
+	HUFFMAN_ONLY     = 2
+	RLE              = 3
+	FIXED            = 4
+	DEFAULT_STRATEGY = 0
+)
+
+const (
+	BINARY  = 0
+	TEXT    = 1
+	ASCII   = TEXT
+	UNKNOWN = 2
+)
+
+const (
+	DEFLATED = 8
+)
+
+//go:linkname CompressBound C.compressBound
+func CompressBound(sourceLen c.Ulong) c.Ulong
+
+//go:linkname Compress C.compress
+func Compress(dest *byte, destLen *c.Ulong, source *byte, sourceLen c.Ulong) c.Int
+
+//go:linkname Compress2 C.compress2
+func Compress2(dest *byte, destLen *c.Ulong, source *byte, sourceLen c.Ulong, level c.Int) c.Int
+
+//go:linkname Uncompress C.uncompress
+func Uncompress(dest *byte, destLen *c.Ulong, source *byte, sourceLen c.Ulong) c.Int
+
+//go:linkname Uncompress2 C.uncompress2
+func Uncompress2(dest *byte, destLen *c.Ulong, source *byte, sourceLen *c.Ulong) c.Int

chore/_xtool/astdump/astdump.cpp

@@ -0,0 +1,63 @@
+#include <clang-c/Index.h>
+#include <iostream>
+#include <string>
+
+CXChildVisitResult visit(CXCursor c, CXCursor parent, CXClientData client_data);
+
+void printAST(CXCursor cursor, unsigned int depth = 0) {
+    CXString cursorKind = clang_getCursorKindSpelling(clang_getCursorKind(cursor));
+    CXString cursorSpelling = clang_getCursorSpelling(cursor);
+
+    for (unsigned int i = 0; i < depth; ++i) {
+        std::cout << "  ";
+    }
+
+    std::cout << clang_getCString(cursorKind) << ": " 
+              << clang_getCString(cursorSpelling) << std::endl;
+
+    clang_disposeString(cursorKind);
+    clang_disposeString(cursorSpelling);
+
+    CXCursor child;
+    clang_visitChildren(
+        cursor,
+        visit,
+        &depth
+    );
+}
+
+CXChildVisitResult visit(CXCursor c, CXCursor parent, CXClientData client_data) {
+    unsigned int* depth = (unsigned int*)client_data;
+    printAST(c, *depth + 1);
+    return CXChildVisit_Continue;
+}
+
+int main(int argc, char* argv[]) {
+    if (argc < 2) {
+        std::cerr << "Usage: " << argv[0] << " <header_file>" << std::endl;
+        return 1;
+    }
+
+    CXIndex index = clang_createIndex(0, 0);
+    CXTranslationUnit unit = clang_parseTranslationUnit(
+        index,
+        argv[1],
+        nullptr, 0,
+        nullptr, 0,
+        CXTranslationUnit_None
+    );
+
+    if (unit == nullptr) {
+        std::cerr << "Unable to parse translation unit. Quitting." << std::endl;
+        return 1;
+    }
+
+    CXCursor cursor = clang_getTranslationUnitCursor(unit);
+    std::cout << "AST for " << argv[1] << ":" << std::endl;
+    printAST(cursor);
+
+    clang_disposeTranslationUnit(unit);
+    clang_disposeIndex(index);
+
+    return 0;
+}

chore/_xtool/astdump/build.sh

@@ -0,0 +1,2 @@
+export LLVM_DIR=/opt/homebrew/Cellar/llvm@17/17.0.6
+clang -L$LLVM_DIR/lib -lclang -lc++ -I$LLVM_DIR/include astdump.cpp

chore/gentests/gentests.go

@@ -34,6 +34,7 @@ func main() {
 	llgen.Verbose = false
 
 	llgenDir(dir + "/cl/_testlibc")
+	llgenDir(dir + "/cl/_testlibgo")
 	llgenDir(dir + "/cl/_testrt")
 	llgenDir(dir + "/cl/_testgo")
 	llgenDir(dir+"/cl/_testpy", "")

chore/nmdump/nmdump.go

@@ -30,7 +30,7 @@ func main() {
 		return
 	}
 
-	nm := llvm.New().Nm()
+	nm := llvm.New("").Nm()
 	items, err := nm.List(os.Args[1])
 	for _, item := range items {
 		if item.File != "" {

chore/nmindex/nmindex.go

@@ -54,7 +54,7 @@ The commands are:
 }
 
 func makeIndex() {
-	env := llvm.New()
+	env := llvm.New("")
 	idxDir := indexDir()
 	os.MkdirAll(idxDir, 0755)
 

cl/_testdata/apkg/out.ll

@@ -1,7 +1,7 @@
 ; ModuleID = 'apkg'
 source_filename = "apkg"
 
-@"apkg.init$guard" = global ptr null
+@"apkg.init$guard" = global i1 false, align 1
 
 define double @apkg.Max(double %0, double %1) {
 _llgo_0:

cl/_testdata/fncall/out.ll

@@ -1,9 +1,9 @@
 ; ModuleID = 'main'
 source_filename = "main"
 
-@"main.init$guard" = global ptr null
-@__llgo_argc = global ptr null
-@__llgo_argv = global ptr null
+@"main.init$guard" = global i1 false, align 1
+@__llgo_argc = global i32 0, align 4
+@__llgo_argv = global ptr null, align 8
 
 define void @main.init() {
 _llgo_0:

cl/_testdata/importpkg/out.ll

@@ -1,10 +1,10 @@
 ; ModuleID = 'main'
 source_filename = "main"
 
-@main.hello = global [7 x i8] undef
-@"main.init$guard" = global ptr null
-@__llgo_argc = global ptr null
-@__llgo_argv = global ptr null
+@main.hello = global [7 x i8] zeroinitializer, align 1
+@"main.init$guard" = global i1 false, align 1
+@__llgo_argc = global i32 0, align 4
+@__llgo_argv = global ptr null, align 8
 
 define void @main.init() {
 _llgo_0: