Made bake.cmd faster and added most thrd_* functions

bake.cmd

@@ -29,23 +29,16 @@ mkdir bin
 mkdir bin\%PLATFORM%
 
 del ciabatta.lib 2> nul
-for /R src\%PLATFORM% %%F in (*.c) do (
+call bake_cc.cmd
-    echo %%F
-    clang -Isrc/win -c -o bin\%PLATFORM%\%%~nF.obj %%F %CIABATTA_OPTIONS%
-)
 for /R src\%PLATFORM% %%F in (*.asm) do (
     echo %%F
     nasm %%F -f win64 -o bin\%PLATFORM%\%%~nF.obj
 )
-for /R src\code %%F in (*.c) do (
-    echo %%F
-    clang -c -o bin\%%~nF.obj %%F %CIABATTA_OPTIONS%
-)
 llvm-ar rc ciabatta.lib bin\*.obj bin\%PLATFORM%\*.obj
 
 :skip_crt_compilation
 
-if "%TEST%"=="" set TEST=assert
+if "%TEST%"=="" set TEST=threads
 
 echo Compiling test_%TEST%.c
 clang test\test_%TEST%.c ciabatta.lib -std=c11 -lDbghelp -lkernel32 -luser32 -lshell32 -nostdlib %CIABATTA_OPTIONS%

bake_cc.cmd

@@ -0,0 +1,8 @@
+for /R src\%PLATFORM% %%F in (*.c) do (
+    echo %%F
+    start /B clang -Isrc/win -c -o bin\%PLATFORM%\%%~nF.obj %%F %CIABATTA_OPTIONS%
+)
+for /R src\code %%F in (*.c) do (
+    echo %%F
+    start /B clang -c -o bin\%%~nF.obj %%F %CIABATTA_OPTIONS%
+)

src/win/win_threads.c

@@ -6,10 +6,80 @@
 //   https://preshing.com/20120305/implementing-a-recursive-mutex/
 //   https://preshing.com/20120226/roll-your-own-lightweight-mutex/
 #include <win.h>
+#include <stdlib.h>
 #include <threads.h>
 
 DWORD _tls_index = 0;
 
+typedef struct thrd__wrapper_info {
+    thrd_start_t func;
+    void* arg;
+} thrd__wrapper_info;
+
+static DWORD thrd__wrapper(void* arg) {
+    thrd__wrapper_info info = *((thrd__wrapper_info*) arg);
+    int result = info.func(info.arg);
+
+    // TODO(NeGate): setup TSS dtors here
+    return (DWORD) result;
+}
+
+thrd_t thrd_current(void) {
+    return (thrd_t){ GetCurrentThread() };
+}
+
+int thrd_create(thrd_t *thr, thrd_start_t func, void *arg) {
+    thrd__wrapper_info* info = malloc(sizeof(thrd__wrapper_info));
+    if (info == NULL) {
+        return thrd_nomem;
+    }
+
+    info->func = func;
+    info->arg = arg;
+
+    // technically thrd_start_t and LPTHREAD_START_ROUTINE aren't the same but are close
+    // enough to be ABI compatible, namely a difference in signedness of the return val.
+    thr->handle = CreateThread(NULL, 0, thrd__wrapper, info, 0, NULL);
+    return thr->handle != NULL ? thrd_success : thrd_error;
+}
+
+int thrd_detach(thrd_t thr) {
+    return CloseHandle(thr.handle) != 0 ? thrd_success : thrd_error;
+}
+
+int thrd_equal(thrd_t thr0, thrd_t thr1) {
+    return GetThreadId(thr0.handle) == GetThreadId(thr1.handle);
+}
+
+int thrd_join(thrd_t thr, int *res) {
+    if (WaitForSingleObject(thr.handle, INFINITE) == WAIT_FAILED) {
+        return thrd_error;
+    }
+
+    if (res != NULL) {
+        // snatch that exit code
+        DWORD ures;
+        if (GetExitCodeThread(thr.handle, &ures) == 0) {
+            CloseHandle(thr.handle);
+            return thrd_error;
+        }
+
+        *res = (int) ures;
+    }
+
+    CloseHandle(thr.handle);
+    return thrd_success;
+}
+
+void thrd_yield(void) {
+    Sleep(0);
+}
+
+_Noreturn void thrd_exit(int res) {
+    // TODO(NeGate): setup TSS dtors here
+    ExitThread((DWORD) res);
+}
+
 void mtx_destroy(mtx_t *mtx) {
     CloseHandle(mtx->semaphore);
 }

test/test_threads.c

@@ -0,0 +1,27 @@
+#include <stdio.h>
+#include <threads.h>
+#include <stdatomic.h>
+
+atomic_int acnt;
+int cnt;
+
+int f(void* thr_data)
+{
+    for(int n = 0; n < 1000; ++n) {
+        atomic_fetch_add_explicit(&acnt, 1, memory_order_relaxed); // atomic
+        ++cnt; // undefined behavior, in practice some updates missed
+    }
+    return 0;
+}
+
+int main(void)
+{
+    thrd_t thr[10];
+    for(int n = 0; n < 10; ++n)
+        thrd_create(&thr[n], f, NULL);
+    for(int n = 0; n < 10; ++n)
+        thrd_join(thr[n], NULL);
+
+    printf("The atomic counter is %d\n", acnt);
+    printf("The non-atomic counter is %d\n", cnt);
+}