Improve allocation of stack for byte code.

py/bc.h

@@ -94,5 +94,5 @@
 #define PYBC_IMPORT_FROM (0xe1)
 #define PYBC_IMPORT_STAR (0xe2)
 
-py_obj_t py_execute_byte_code(const byte *code, const py_obj_t *args, uint n_args);
+py_obj_t py_execute_byte_code(const byte *code, const py_obj_t *args, uint n_args, uint n_state);
 bool py_execute_byte_code_2(const byte **ip_in_out, py_obj_t *fastn, py_obj_t **sp_in_out);

py/runtime.c

@@ -133,8 +133,8 @@ struct _py_obj_base_t {
         } u_fun;
         struct { // for O_FUN_BC
             int n_args;
+            uint n_state;
             byte *code;
-            uint len;
         } u_fun_bc;
         struct { // for O_FUN_ASM
             int n_args;
@@ -1474,8 +1474,8 @@ py_obj_t rt_make_function_from_id(int unique_code_id) {
         case PY_CODE_BYTE:
             o->kind = O_FUN_BC;
             o->u_fun_bc.n_args = c->n_args;
+            o->u_fun_bc.n_state = c->n_locals + c->n_stack;
             o->u_fun_bc.code = c->u_byte.code;
-            o->u_fun_bc.len = c->u_byte.len;
             break;
         case PY_CODE_NATIVE:
             switch (c->n_args) {
@@ -1660,7 +1660,7 @@ py_obj_t rt_call_function_n(py_obj_t fun, int n_args, const py_obj_t *args) {
             goto bad_n_args;
         }
         DEBUG_OP_printf("calling byte code %p(n_args=%d)\n", o->u_fun_bc.code, n_args);
-        return py_execute_byte_code(o->u_fun_bc.code, args, n_args);
+        return py_execute_byte_code(o->u_fun_bc.code, args, n_args, o->u_fun_bc.n_state);
 
     } else if (IS_O(fun, O_FUN_ASM)) {
         py_obj_base_t *o = fun;

py/vm.c

@@ -21,14 +21,19 @@
 #define POP() (*sp++)
 
 // args are in reverse order in array
-py_obj_t py_execute_byte_code(const byte *code, const py_obj_t *args, uint n_args) {
-    py_obj_t state[18]; // TODO allocate properly
+py_obj_t py_execute_byte_code(const byte *code, const py_obj_t *args, uint n_args, uint n_state) {
+    py_obj_t temp_state[10]; // TODO allocate properly
+    py_obj_t *state = &temp_state[0];
+    py_obj_t *sp = &state[10];
+    if (n_state > 10) {
+        state = m_new(py_obj_t, n_state);
+        sp = &state[n_state];
+    }
     // init args
     for (int i = 0; i < n_args; i++) {
         assert(i < 8);
         state[i] = args[n_args - 1 - i];
     }
-    py_obj_t *sp = &state[18];
     const byte *ip = code;
     if (py_execute_byte_code_2(&ip, &state[0], &sp)) {
         // it shouldn't yield