diff --git a/py/builtin.h b/py/builtin.h
index a5e0f5f2d0fbb..863956e2b9a38 100644
--- a/py/builtin.h
+++ b/py/builtin.h
@@ -75,7 +75,8 @@ MP_DECLARE_CONST_FUN_OBJ_KW(mp_builtin_print_obj);
 MP_DECLARE_CONST_FUN_OBJ_1(mp_builtin_repr_obj);
 MP_DECLARE_CONST_FUN_OBJ_VAR_BETWEEN(mp_builtin_round_obj);
 MP_DECLARE_CONST_FUN_OBJ_KW(mp_builtin_sorted_obj);
-MP_DECLARE_CONST_FUN_OBJ_VAR_BETWEEN(mp_builtin_sum_obj);
+//MP_DECLARE_CONST_FUN_OBJ_VAR_BETWEEN(mp_builtin_sum_obj);
+extern const struct _mp_obj_fun_bc_t mp_builtin_sum_obj;
 // Defined by a port, but declared here for simplicity
 MP_DECLARE_CONST_FUN_OBJ_VAR_BETWEEN(mp_builtin_input_obj);
 MP_DECLARE_CONST_FUN_OBJ_KW(mp_builtin_open_obj);
diff --git a/py/modbuiltins.c b/py/modbuiltins.c
index a65f3beecfa8b..67e8605a3e870 100644
--- a/py/modbuiltins.c
+++ b/py/modbuiltins.c
@@ -521,6 +521,106 @@ STATIC mp_obj_t mp_builtin_round(size_t n_args, const mp_obj_t *args) {
 }
 MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(mp_builtin_round_obj, 1, 2, mp_builtin_round);
 
+#if 1
+
+// Implement sum() in Python!
+//
+// def sum(iterable, start=0):
+//     for x in iterable:
+//         start += x
+//     return start
+
+#include "py/bc0.h"
+#include "py/runtime0.h"
+#include "py/objfun.h"
+
+extern const mp_obj_type_t mp_type_fun_native;
+
+#if 1
+// Implement sum in Python bytecode
+
+STATIC const uint8_t mp_builtin_sum_bytecode[] = {
+    0x09, // n_state
+    0x00, // n_exc_stack
+    0x80, // scope_flags (special top-bit set for internal bytecode function)
+    0x02, // n_pos_args
+    0x00, // n_kwonly_args
+    0x01, // n_def_pos_args
+    0x03, // code_info_size
+    MP_QSTR_sum & 0xff, MP_QSTR_sum >> 8, // simple_name
+    0xff, // end-of-list-sentinel for cells
+
+    MP_BC_LOAD_FAST_MULTI,
+    MP_BC_GET_ITER_STACK,
+    MP_BC_FOR_ITER, 0x08, 0x00, // +8
+    MP_BC_STORE_FAST_MULTI + 2,
+    MP_BC_LOAD_FAST_MULTI + 1,
+    MP_BC_LOAD_FAST_MULTI + 2,
+    MP_BC_BINARY_OP_MULTI + MP_BINARY_OP_INPLACE_ADD,
+    MP_BC_STORE_FAST_MULTI + 1,
+    MP_BC_JUMP, 0xf5, 0x7f, // -11
+    MP_BC_LOAD_FAST_MULTI + 1,
+    MP_BC_RETURN_VALUE,
+};
+const mp_obj_fun_bc_t mp_builtin_sum_obj = {
+    .base = { &mp_type_fun_bc },
+    .globals = NULL,
+    .bytecode = mp_builtin_sum_bytecode,
+    .const_table = NULL,
+    { MP_OBJ_NEW_SMALL_INT(0) }, // default start value
+};
+
+#else
+// Implement sum in Python native code
+
+#if defined(__x86_64__)
+STATIC const byte mp_builtin_sum_fun_data[159] __attribute__((section(".text,\"ax\",@progbits # "))) = {
+    0x55, 0x53, 0x41, 0x54, 0x41, 0x55, 0x48, 0x83, 0xec, 0x78, 0x4c, 0x8b, 0x6f, 0x18, 0x49, 0x8b,
+    0x6d, 0x10, 0x48, 0x89, 0x3c, 0x24, 0xbf, 0x92, 0x00, 0x00, 0x00, 0x48, 0x89, 0x7c, 0x24, 0x08,
+    0x48, 0x89, 0xe7, 0x48, 0x8b, 0x85, 0x70, 0x01, 0x00, 0x00, 0xff, 0xd0, 0x48, 0x8b, 0x5c, 0x24,
+    0x68, 0x4c, 0x8b, 0x64, 0x24, 0x60, 0x4c, 0x8b, 0x6c, 0x24, 0x58, 0x48, 0x89, 0xdf, 0x48, 0x8d,
+    0x74, 0x24, 0x28, 0x48, 0x8b, 0x85, 0xf0, 0x00, 0x00, 0x00, 0xff, 0xd0, 0x48, 0x8d, 0x7c, 0x24,
+    0x28, 0x48, 0x8b, 0x85, 0xf8, 0x00, 0x00, 0x00, 0xff, 0xd0, 0x48, 0x85, 0xc0, 0x0f, 0x84, 0x1c,
+    0x00, 0x00, 0x00, 0x49, 0x89, 0xc5, 0x4c, 0x89, 0xea, 0x4c, 0x89, 0xe6, 0xbf, 0x0e, 0x00, 0x00,
+    0x00, 0x48, 0x8b, 0x85, 0x90, 0x00, 0x00, 0x00, 0xff, 0xd0, 0x49, 0x89, 0xc4, 0xeb, 0xcd, 0x4c,
+    0x89, 0xe0, 0xe9, 0x00, 0x00, 0x00, 0x00, 0x48, 0x83, 0xec, 0x88, 0x41, 0x5d, 0x41, 0x5c, 0x5b,
+    0x5d, 0xc3,
+    0x80, 0x09, 0x00, 0x00, 0x02, 0x00, 0x01, 0x05,
+    MP_QSTR_sum & 0xff, MP_QSTR_sum >> 8,
+    MP_QSTR_sum_dot_py & 0xff, MP_QSTR_sum_dot_py >> 8,
+    0xff,
+};
+#elif defined(__thumb2__)
+STATIC const byte mp_builtin_sum_fun_data[87] __attribute__((section(".text,\"ax\",%progbits @ "))) __attribute__ ((aligned (2))) = {
+    0xfe, 0xb5, 0x8c, 0xb0, 0xc6, 0x68, 0xb7, 0x68, 0x00, 0x90, 0x4a, 0x20, 0x01, 0x90, 0x00, 0xa8,
+    0xd7, 0xf8, 0xb8, 0x40, 0xa0, 0x47, 0x0d, 0x9c, 0x0c, 0x9d, 0x0b, 0x9e, 0x20, 0x46, 0x05, 0xa9,
+    0xbb, 0x6f, 0x98, 0x47, 0x05, 0xa8, 0xfb, 0x6f, 0x98, 0x47, 0x00, 0x28, 0x00, 0xf0, 0x08, 0x80,
+    0x06, 0x46, 0x32, 0x46, 0x29, 0x46, 0x0e, 0x20, 0xbb, 0x6c, 0x98, 0x47, 0x05, 0x46, 0xf1, 0xe7,
+    0x28, 0x46, 0x00, 0xf0, 0x00, 0xb8, 0x0c, 0xb0, 0xfe, 0xbd,
+    0x80, 0x09, 0x00, 0x00, 0x02, 0x00, 0x01, 0x05,
+    MP_QSTR_sum & 0xff, MP_QSTR_sum >> 8,
+    MP_QSTR_sum_dot_py & 0xff, MP_QSTR_sum_dot_py >> 8,
+    0xff,
+};
+#endif
+STATIC const mp_rom_obj_t mp_builtin_sum_const_table[3] = {
+    MP_ROM_QSTR(MP_QSTR_iterable),
+    MP_ROM_QSTR(MP_QSTR_start),
+    mp_fun_table,
+};
+const mp_obj_fun_bc_t mp_builtin_sum_obj = {
+    .base = { &mp_type_fun_native },
+    .globals = NULL,
+    .bytecode = mp_builtin_sum_fun_data,
+    .const_table = (void*)mp_builtin_sum_const_table,
+    { MP_OBJ_NEW_SMALL_INT(0) }, // default start value
+};
+
+#endif
+
+#else
+// Implement sum in C
+
 STATIC mp_obj_t mp_builtin_sum(size_t n_args, const mp_obj_t *args) {
     mp_obj_t value;
     switch (n_args) {
@@ -537,6 +637,8 @@ STATIC mp_obj_t mp_builtin_sum(size_t n_args, const mp_obj_t *args) {
 }
 MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(mp_builtin_sum_obj, 1, 2, mp_builtin_sum);
 
+#endif
+
 STATIC mp_obj_t mp_builtin_sorted(size_t n_args, const mp_obj_t *args, mp_map_t *kwargs) {
     if (n_args > 1) {
         mp_raise_TypeError("must use keyword argument for key function");
diff --git a/py/objfun.c b/py/objfun.c
index e0c6fb9271276..0df33fb63463f 100644
--- a/py/objfun.c
+++ b/py/objfun.c
@@ -148,7 +148,7 @@ qstr mp_obj_code_get_name(const byte *code_info) {
 }
 
 #if MICROPY_EMIT_NATIVE
-STATIC const mp_obj_type_t mp_type_fun_native;
+const mp_obj_type_t mp_type_fun_native;
 #endif
 
 qstr mp_obj_fun_get_name(mp_const_obj_t fun_in) {
@@ -406,7 +406,7 @@ STATIC mp_obj_t fun_native_call(mp_obj_t self_in, size_t n_args, size_t n_kw, co
     return fun(self_in, n_args, n_kw, args);
 }
 
-STATIC const mp_obj_type_t mp_type_fun_native = {
+const mp_obj_type_t mp_type_fun_native = {
     { &mp_type_type },
     .name = MP_QSTR_function,
     .call = fun_native_call,
diff --git a/py/vm.c b/py/vm.c
index 260a7f38baf9d..4c861e014499f 100644
--- a/py/vm.c
+++ b/py/vm.c
@@ -1365,6 +1365,10 @@ unwind_jump:;
                 const byte *ip = code_state->fun_bc->bytecode;
                 ip = mp_decode_uint_skip(ip); // skip n_state
                 ip = mp_decode_uint_skip(ip); // skip n_exc_stack
+                if (*ip & 0x80) {
+                    // builtin bytecode, don't add traceback info
+                    goto skip_traceback;
+                }
                 ip++; // skip scope_params
                 ip++; // skip n_pos_args
                 ip++; // skip n_kwonly_args
@@ -1407,6 +1411,7 @@ unwind_jump:;
                     }
                 }
                 mp_obj_exception_add_traceback(MP_OBJ_FROM_PTR(nlr.ret_val), source_file, source_line, block_name);
+            skip_traceback:;
             }
 
             while (exc_sp >= exc_stack && exc_sp->handler <= code_state->ip) {