py: Add MICROPY_PERSISTENT_CODE so code can persist beyond the runtime.

Main changes when MICROPY_PERSISTENT_CODE is enabled are:

- qstrs are encoded as 2-byte fixed width in the bytecode
- all pointers are removed from bytecode and put in const_table (this
  includes const objects and raw code pointers)

Ultimately this option will enable persistence for not just bytecode but
also native code.

py/emitbc.c

@@ -56,7 +56,14 @@ struct _emit_t {
     mp_uint_t bytecode_offset;
     mp_uint_t bytecode_size;
     byte *code_base; // stores both byte code and code info
+
+    #if MICROPY_PERSISTENT_CODE
+    uint16_t ct_cur_obj;
+    uint16_t ct_num_obj;
+    uint16_t ct_cur_raw_code;
+    #endif
     mp_uint_t *const_table;
+
     // Accessed as mp_uint_t, so must be aligned as such
     byte dummy_data[DUMMY_DATA_SIZE];
 };
@@ -108,10 +115,6 @@ STATIC byte *emit_get_cur_to_write_code_info(emit_t *emit, int num_bytes_to_writ
     }
 }
 
-STATIC void emit_align_code_info_to_machine_word(emit_t *emit) {
-    emit->code_info_offset = (emit->code_info_offset + sizeof(mp_uint_t) - 1) & (~(sizeof(mp_uint_t) - 1));
-}
-
 STATIC void emit_write_code_info_byte(emit_t* emit, byte val) {
     *emit_get_cur_to_write_code_info(emit, 1) = val;
 }
@@ -121,7 +124,14 @@ STATIC void emit_write_code_info_uint(emit_t* emit, mp_uint_t val) {
 }
 
 STATIC void emit_write_code_info_qstr(emit_t *emit, qstr qst) {
+    #if MICROPY_PERSISTENT_CODE
+    assert((qst >> 16) == 0);
+    byte *c = emit_get_cur_to_write_code_info(emit, 2);
+    c[0] = qst;
+    c[1] = qst >> 8;
+    #else
     emit_write_uint(emit, emit_get_cur_to_write_code_info, qst);
+    #endif
 }
 
 #if MICROPY_ENABLE_SOURCE_LINE
@@ -163,10 +173,6 @@ STATIC byte *emit_get_cur_to_write_bytecode(emit_t *emit, int num_bytes_to_write
     }
 }
 
-STATIC void emit_align_bytecode_to_machine_word(emit_t *emit) {
-    emit->bytecode_offset = (emit->bytecode_offset + sizeof(mp_uint_t) - 1) & (~(sizeof(mp_uint_t) - 1));
-}
-
 STATIC void emit_write_bytecode_byte(emit_t *emit, byte b1) {
     byte *c = emit_get_cur_to_write_bytecode(emit, 1);
     c[0] = b1;
@@ -211,18 +217,55 @@ STATIC void emit_write_bytecode_byte_uint(emit_t *emit, byte b, mp_uint_t val) {
     emit_write_uint(emit, emit_get_cur_to_write_bytecode, val);
 }
 
-// aligns the pointer so it is friendly to GC
+#if MICROPY_PERSISTENT_CODE
+STATIC void emit_write_bytecode_byte_const(emit_t *emit, byte b, mp_uint_t n, mp_uint_t c) {
+    if (emit->pass == MP_PASS_EMIT) {
+        emit->const_table[n] = c;
+    }
+    emit_write_bytecode_byte_uint(emit, b, n);
+}
+#else
 STATIC void emit_write_bytecode_byte_ptr(emit_t *emit, byte b, void *ptr) {
+    // aligns the pointer so it is friendly to GC
     emit_write_bytecode_byte(emit, b);
-    emit_align_bytecode_to_machine_word(emit);
+    emit->bytecode_offset = (mp_uint_t)MP_ALIGN(emit->bytecode_offset, sizeof(mp_uint_t));
     mp_uint_t *c = (mp_uint_t*)emit_get_cur_to_write_bytecode(emit, sizeof(mp_uint_t));
     // Verify thar c is already uint-aligned
     assert(c == MP_ALIGN(c, sizeof(mp_uint_t)));
     *c = (mp_uint_t)ptr;
 }
+#endif
 
 STATIC void emit_write_bytecode_byte_qstr(emit_t* emit, byte b, qstr qst) {
+    #if MICROPY_PERSISTENT_CODE
+    assert((qst >> 16) == 0);
+    byte *c = emit_get_cur_to_write_bytecode(emit, 3);
+    c[0] = b;
+    c[1] = qst;
+    c[2] = qst >> 8;
+    #else
     emit_write_bytecode_byte_uint(emit, b, qst);
+    #endif
+}
+
+STATIC void emit_write_bytecode_byte_obj(emit_t *emit, byte b, void *ptr) {
+    #if MICROPY_PERSISTENT_CODE
+    emit_write_bytecode_byte_const(emit, b,
+        emit->scope->num_pos_args + emit->scope->num_kwonly_args
+        + emit->ct_cur_obj++, (mp_uint_t)ptr);
+    #else
+    emit_write_bytecode_byte_ptr(emit, b, ptr);
+    #endif
+}
+
+STATIC void emit_write_bytecode_byte_raw_code(emit_t *emit, byte b, mp_raw_code_t *rc) {
+    #if MICROPY_PERSISTENT_CODE
+    emit_write_bytecode_byte_const(emit, b,
+        emit->scope->num_pos_args + emit->scope->num_kwonly_args
+        + emit->ct_num_obj + emit->ct_cur_raw_code++, (mp_uint_t)rc);
+    #else
+    emit_write_bytecode_byte_ptr(emit, b, rc);
+    #endif
 }
 
 // unsigned labels are relative to ip following this instruction, stored as 16 bits
@@ -318,6 +361,11 @@ void mp_emit_bc_start_pass(emit_t *emit, pass_kind_t pass, scope_t *scope) {
     }
     emit_write_bytecode_byte(emit, 255); // end of list sentinel
 
+    #if MICROPY_PERSISTENT_CODE
+    emit->ct_cur_obj = 0;
+    emit->ct_cur_raw_code = 0;
+    #endif
+
     if (pass == MP_PASS_EMIT) {
         // Write argument names (needed to resolve positional args passed as
         // keywords).  We store them as full word-sized objects for efficient access
@@ -360,16 +408,30 @@ void mp_emit_bc_end_pass(emit_t *emit) {
 
     emit_write_code_info_byte(emit, 0); // end of line number info
 
+    #if MICROPY_PERSISTENT_CODE
+    assert(emit->pass <= MP_PASS_STACK_SIZE || (emit->ct_num_obj == emit->ct_cur_obj));
+    emit->ct_num_obj = emit->ct_cur_obj;
+    #endif
+
     if (emit->pass == MP_PASS_CODE_SIZE) {
+        #if !MICROPY_PERSISTENT_CODE
         // so bytecode is aligned
-        emit_align_code_info_to_machine_word(emit);
+        emit->code_info_offset = (mp_uint_t)MP_ALIGN(emit->code_info_offset, sizeof(mp_uint_t));
+        #endif
 
         // calculate size of total code-info + bytecode, in bytes
         emit->code_info_size = emit->code_info_offset;
         emit->bytecode_size = emit->bytecode_offset;
         emit->code_base = m_new0(byte, emit->code_info_size + emit->bytecode_size);
 
-        emit->const_table = m_new0(mp_uint_t, emit->scope->num_pos_args + emit->scope->num_kwonly_args);
+        #if MICROPY_PERSISTENT_CODE
+        emit->const_table = m_new0(mp_uint_t,
+            emit->scope->num_pos_args + emit->scope->num_kwonly_args
+            + emit->ct_cur_obj + emit->ct_cur_raw_code);
+        #else
+        emit->const_table = m_new0(mp_uint_t,
+            emit->scope->num_pos_args + emit->scope->num_kwonly_args);
+        #endif
 
     } else if (emit->pass == MP_PASS_EMIT) {
         mp_emit_glue_assign_bytecode(emit->scope->raw_code, emit->code_base,
@@ -457,7 +519,7 @@ void mp_emit_bc_load_const_tok(emit_t *emit, mp_token_kind_t tok) {
         case MP_TOKEN_KW_NONE: emit_write_bytecode_byte(emit, MP_BC_LOAD_CONST_NONE); break;
         case MP_TOKEN_KW_TRUE: emit_write_bytecode_byte(emit, MP_BC_LOAD_CONST_TRUE); break;
         no_other_choice:
-        case MP_TOKEN_ELLIPSIS: emit_write_bytecode_byte_ptr(emit, MP_BC_LOAD_CONST_OBJ, (void*)&mp_const_ellipsis_obj); break;
+        case MP_TOKEN_ELLIPSIS: emit_write_bytecode_byte_obj(emit, MP_BC_LOAD_CONST_OBJ, (void*)&mp_const_ellipsis_obj); break;
         default: assert(0); goto no_other_choice; // to help flow control analysis
     }
 }
@@ -478,7 +540,7 @@ void mp_emit_bc_load_const_str(emit_t *emit, qstr qst) {
 
 void mp_emit_bc_load_const_obj(emit_t *emit, void *obj) {
     emit_bc_pre(emit, 1);
-    emit_write_bytecode_byte_ptr(emit, MP_BC_LOAD_CONST_OBJ, obj);
+    emit_write_bytecode_byte_obj(emit, MP_BC_LOAD_CONST_OBJ, obj);
 }
 
 void mp_emit_bc_load_null(emit_t *emit) {
@@ -821,22 +883,22 @@ void mp_emit_bc_unpack_ex(emit_t *emit, mp_uint_t n_left, mp_uint_t n_right) {
 void mp_emit_bc_make_function(emit_t *emit, scope_t *scope, mp_uint_t n_pos_defaults, mp_uint_t n_kw_defaults) {
     if (n_pos_defaults == 0 && n_kw_defaults == 0) {
         emit_bc_pre(emit, 1);
-        emit_write_bytecode_byte_ptr(emit, MP_BC_MAKE_FUNCTION, scope->raw_code);
+        emit_write_bytecode_byte_raw_code(emit, MP_BC_MAKE_FUNCTION, scope->raw_code);
     } else {
         emit_bc_pre(emit, -1);
-        emit_write_bytecode_byte_ptr(emit, MP_BC_MAKE_FUNCTION_DEFARGS, scope->raw_code);
+        emit_write_bytecode_byte_raw_code(emit, MP_BC_MAKE_FUNCTION_DEFARGS, scope->raw_code);
     }
 }
 
 void mp_emit_bc_make_closure(emit_t *emit, scope_t *scope, mp_uint_t n_closed_over, mp_uint_t n_pos_defaults, mp_uint_t n_kw_defaults) {
     if (n_pos_defaults == 0 && n_kw_defaults == 0) {
         emit_bc_pre(emit, -n_closed_over + 1);
-        emit_write_bytecode_byte_ptr(emit, MP_BC_MAKE_CLOSURE, scope->raw_code);
+        emit_write_bytecode_byte_raw_code(emit, MP_BC_MAKE_CLOSURE, scope->raw_code);
         emit_write_bytecode_byte(emit, n_closed_over);
     } else {
         assert(n_closed_over <= 255);
         emit_bc_pre(emit, -2 - n_closed_over + 1);
-        emit_write_bytecode_byte_ptr(emit, MP_BC_MAKE_CLOSURE_DEFARGS, scope->raw_code);
+        emit_write_bytecode_byte_raw_code(emit, MP_BC_MAKE_CLOSURE_DEFARGS, scope->raw_code);
         emit_write_bytecode_byte(emit, n_closed_over);
     }
 }

py/emitnative.c

@@ -830,10 +830,16 @@ STATIC void emit_native_end_pass(emit_t *emit) {
         ASM_DATA(emit->as, 1, emit->scope->num_kwonly_args);
         ASM_DATA(emit->as, 1, emit->scope->num_def_pos_args);
 
-        // write code info (just contains block name and source file)
+        // write code info
+        #if MICROPY_PERSISTENT_CODE
         ASM_DATA(emit->as, 1, 5);
-        ASM_DATA(emit->as, 2, emit->scope->simple_name);
-        ASM_DATA(emit->as, 2, emit->scope->source_file);
+        ASM_DATA(emit->as, 1, emit->scope->simple_name);
+        ASM_DATA(emit->as, 1, emit->scope->simple_name >> 8);
+        ASM_DATA(emit->as, 1, emit->scope->source_file);
+        ASM_DATA(emit->as, 1, emit->scope->source_file >> 8);
+        #else
+        ASM_DATA(emit->as, 1, 1);
+        #endif
 
         // bytecode prelude: initialise closed over variables
         for (int i = 0; i < emit->scope->id_info_len; i++) {

py/mpconfig.h

@@ -192,6 +192,11 @@
 /*****************************************************************************/
 /* Micro Python emitters                                                     */
 
+// Whether generated code can persist independently of the VM/runtime instance
+#ifndef MICROPY_PERSISTENT_CODE
+#define MICROPY_PERSISTENT_CODE (0)
+#endif
+
 // Whether to emit x64 native code
 #ifndef MICROPY_EMIT_X64
 #define MICROPY_EMIT_X64 (0)

py/objfun.c

@@ -106,7 +106,11 @@ const mp_obj_type_t mp_type_fun_builtin = {
 
 qstr mp_obj_code_get_name(const byte *code_info) {
     mp_decode_uint(&code_info); // skip code_info_size entry
+    #if MICROPY_PERSISTENT_CODE
+    return code_info[0] | (code_info[1] << 8);
+    #else
     return mp_decode_uint(&code_info);
+    #endif
 }
 
 #if MICROPY_EMIT_NATIVE

py/showbc.c

@@ -40,6 +40,18 @@
 }
 #define DECODE_ULABEL do { unum = (ip[0] | (ip[1] << 8)); ip += 2; } while (0)
 #define DECODE_SLABEL do { unum = (ip[0] | (ip[1] << 8)) - 0x8000; ip += 2; } while (0)
+
+#if MICROPY_PERSISTENT_CODE
+
+#define DECODE_QSTR \
+    qst = ip[0] | ip[1] << 8; \
+    ip += 2;
+#define DECODE_PTR \
+    DECODE_UINT; \
+    unum = mp_showbc_const_table[unum]
+
+#else
+
 #define DECODE_QSTR { \
     qst = 0; \
     do { \
@@ -52,10 +64,14 @@
     ip += sizeof(mp_uint_t); \
 } while (0)
 
+#endif
+
 const byte *mp_showbc_code_start;
+const mp_uint_t *mp_showbc_const_table;
 
 void mp_bytecode_print(const void *descr, const byte *ip, mp_uint_t len, const mp_uint_t *const_table) {
     mp_showbc_code_start = ip;
+    mp_showbc_const_table = const_table;
 
     // get bytecode parameters
     mp_uint_t n_state = mp_decode_uint(&ip);
@@ -69,8 +85,13 @@ void mp_bytecode_print(const void *descr, const byte *ip, mp_uint_t len, const m
     mp_uint_t code_info_size = mp_decode_uint(&code_info);
     ip += code_info_size;
 
+    #if MICROPY_PERSISTENT_CODE
+    qstr block_name = code_info[0] | (code_info[1] << 8);
+    qstr source_file = code_info[2] | (code_info[3] << 8);
+    #else
     qstr block_name = mp_decode_uint(&code_info);
     qstr source_file = mp_decode_uint(&code_info);
+    #endif
     printf("File %s, code block '%s' (descriptor: %p, bytecode @%p " UINT_FMT " bytes)\n",
         qstr_str(source_file), qstr_str(block_name), descr, mp_showbc_code_start, len);
 

py/vm.c

@@ -65,6 +65,18 @@ typedef enum {
     } while ((*ip++ & 0x80) != 0)
 #define DECODE_ULABEL mp_uint_t ulab = (ip[0] | (ip[1] << 8)); ip += 2
 #define DECODE_SLABEL mp_uint_t slab = (ip[0] | (ip[1] << 8)) - 0x8000; ip += 2
+
+#if MICROPY_PERSISTENT_CODE
+
+#define DECODE_QSTR \
+    qstr qst = ip[0] | ip[1] << 8; \
+    ip += 2;
+#define DECODE_PTR \
+    DECODE_UINT; \
+    void *ptr = (void*)code_state->const_table[unum]
+
+#else
+
 #define DECODE_QSTR qstr qst = 0; \
     do { \
         qst = (qst << 7) + (*ip & 0x7f); \
@@ -73,6 +85,9 @@ typedef enum {
     ip = (byte*)(((mp_uint_t)ip + sizeof(mp_uint_t) - 1) & (~(sizeof(mp_uint_t) - 1))); /* align ip */ \
     void *ptr = (void*)*(mp_uint_t*)ip; \
     ip += sizeof(mp_uint_t)
+
+#endif
+
 #define PUSH(val) *++sp = (val)
 #define POP() (*sp--)
 #define TOP() (*sp)
@@ -1280,8 +1295,14 @@ unwind_loop:
             if (mp_obj_is_exception_instance(nlr.ret_val) && nlr.ret_val != &mp_const_GeneratorExit_obj && nlr.ret_val != &mp_const_MemoryError_obj) {
                 const byte *ip = code_state->code_info;
                 mp_uint_t code_info_size = mp_decode_uint(&ip);
+                #if MICROPY_PERSISTENT_CODE
+                qstr block_name = ip[0] | (ip[1] << 8);
+                qstr source_file = ip[2] | (ip[3] << 8);
+                ip += 4;
+                #else
                 qstr block_name = mp_decode_uint(&ip);
                 qstr source_file = mp_decode_uint(&ip);
+                #endif
                 mp_uint_t bc = code_state->ip - code_state->code_info - code_info_size;
                 mp_uint_t source_line = 1;
                 mp_uint_t c;