Add some more builtins to chibicc

https://justine.lol/cosmopolitan/documentation.html should now contain a
lot of functions that had been missing previously due to not having them

third_party/chibicc/test/msabi_test.c

@@ -0,0 +1,6 @@
+#include "third_party/chibicc/test/test.h"
+
+int (*__attribute__((__ms_abi__)) NtFoo)(int x, int y);
+
+int main() {
+}

third_party/chibicc/test/spinlock_test.c

@@ -0,0 +1,26 @@
+#include "third_party/chibicc/test/test.h"
+
+#define SPINLOCK(lock)                                \
+  do {                                                \
+    for (;;) {                                        \
+      typeof(*(lock)) x;                              \
+      __atomic_load(lock, &x, __ATOMIC_RELAXED);      \
+      if (!x && !__sync_lock_test_and_set(lock, 1)) { \
+        break;                                        \
+      } else {                                        \
+        __builtin_ia32_pause();                       \
+      }                                               \
+    }                                                 \
+  } while (0)
+
+#define SPUNLOCK(lock) __sync_lock_release(lock)
+
+_Alignas(64) char lock;
+
+main() {
+  ASSERT(0, lock);
+  SPINLOCK(&lock);
+  ASSERT(1, lock);
+  SPUNLOCK(&lock);
+  ASSERT(0, lock);
+}

third_party/chibicc/test/vector_test.c

@@ -18,6 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "third_party/chibicc/test/test.h"
 
+typedef char byte16 __attribute__((__vector_size__(16)));
 typedef float float4 __attribute__((__vector_size__(16)));
 typedef float float4a1 __attribute__((__vector_size__(16), __aligned__(1)));
 typedef float float4a16 __attribute__((__vector_size__(16), __aligned__(16)));
@@ -39,18 +40,33 @@ int main(void) {
   ASSERT(1, _Alignof(double2a1));
   ASSERT(16, _Alignof(double2a16));
 
-  float4 v1;
-  float4 v2;
-  float x[4] = {1, 2, 3, 4};
-  float y[4] = {1, 1, 1, 1};
-  memcpy(&v1, x, 16);
-  memcpy(&v2, y, 16);
-  v1 = v1 + v2;
-  memcpy(x, &v1, 16);
-  ASSERT(2, x[0]);
-  /* ASSERT(3, x[1]); */
-  /* ASSERT(4, x[2]); */
-  /* ASSERT(5, x[3]); */
+  {
+    float4 v1;
+    float4 v2;
+    float x[4] = {1, 2, 3, 4};
+    float y[4] = {1, 1, 1, 1};
+    memcpy(&v1, x, 16);
+    memcpy(&v2, y, 16);
+    v1 = v1 + v2;
+    memcpy(x, &v1, 16);
+    ASSERT(2, x[0]);
+    // TODO(jart): fix me
+    /* ASSERT(3, x[1]); */
+    /* ASSERT(4, x[2]); */
+    /* ASSERT(5, x[3]); */
+  }
+
+  {
+    byte16 v;
+    float x1[4] = {1, 2, 3, 4};
+    float x2[4];
+    memcpy(&v, x1, 16);
+    __builtin_ia32_movntdq(x1, &v);
+    memcpy(x2, &v, 16);
+    ASSERT(1, x2[0]);
+    // TODO(jart): fix me
+    /* ASSERT(2, x[1]); */
+  }
 
   return 0;
 }