Add selective collect to memory allocations

By selectively collecting an allocation, we can skip scanning many
allocations for pointers because we know up front they won't have
them. This helps a ton when large buffers are being used and memory is
slow (PSRAM). In one Fruit Jam example GC times drop from 80+ms to
~25ms. The example uses a number of bitmaps that are now no longer
scanned.

Author: tannewt

extmod/modasyncio.c

@@ -179,7 +179,8 @@ mp_obj_t mp_asyncio_context = MP_OBJ_NULL;
 
 static mp_obj_t task_make_new(const mp_obj_type_t *type, size_t n_args, size_t n_kw, const mp_obj_t *args) {
     mp_arg_check_num(n_args, n_kw, 1, 2, false);
-    mp_obj_task_t *self = m_new_obj(mp_obj_task_t);
+    // CIRCUITPY-CHANGE: Task holds onto core and data so collect it.
+    mp_obj_task_t *self = m_malloc_with_collect(sizeof(mp_obj_task_t));
     self->pairheap.base.type = type;
     mp_pairheap_init_node(task_lt, &self->pairheap);
     self->coro = args[0];

extmod/vfs.c

@@ -237,7 +237,8 @@ mp_obj_t mp_vfs_mount(size_t n_args, const mp_obj_t *pos_args, mp_map_t *kw_args
     }
 
     // create new object
-    mp_vfs_mount_t *vfs = m_new_obj(mp_vfs_mount_t);
+    // CIRCUITPY-CHANGE: Collect the mount object because it references others
+    mp_vfs_mount_t *vfs = m_malloc_with_collect(sizeof(mp_vfs_mount_t));
     vfs->str = mnt_str;
     vfs->len = mnt_len;
     vfs->obj = vfs_obj;

ports/raspberrypi/boards/adafruit_fruit_jam/board.c

@@ -16,8 +16,8 @@
 
 #define I2S_RESET_PIN_NUMBER 22
 
-#if defined(DEFAULT_USB_HOST_5V_POWER)
 bool board_reset_pin_number(uint8_t pin_number) {
+    #if defined(DEFAULT_USB_HOST_5V_POWER)
     if (pin_number == DEFAULT_USB_HOST_5V_POWER->number) {
         // doing this (rather than gpio_init) in this specific order ensures no
         // glitch if pin was already configured as a high output. gpio_init() temporarily
@@ -29,6 +29,7 @@ bool board_reset_pin_number(uint8_t pin_number) {
 
         return true;
     }
+    #endif
     // Set I2S out of reset.
     if (pin_number == I2S_RESET_PIN_NUMBER) {
         gpio_put(pin_number, 1);
@@ -39,7 +40,6 @@ bool board_reset_pin_number(uint8_t pin_number) {
     }
     return false;
 }
-#endif
 
 void board_init(void) {
     // Reset the DAC to put it in a known state.

ports/unix/alloc.c

@@ -58,7 +58,8 @@ void mp_unix_alloc_exec(size_t min_size, void **ptr, size_t *size) {
     }
 
     // add new link to the list of mmap'd regions
-    mmap_region_t *rg = m_new_obj(mmap_region_t);
+    // CIRCUITPY-CHANGE: Collect the mmap region because it points to others.
+    mmap_region_t *rg = m_malloc_with_collect(sizeof(mmap_region_t));
     rg->ptr = *ptr;
     rg->len = min_size;
     rg->next = MP_STATE_VM(mmap_region_head);

ports/unix/mpconfigport.h

@@ -112,6 +112,7 @@ typedef long mp_off_t;
 
 // Always enable GC.
 #define MICROPY_ENABLE_GC           (1)
+#define MICROPY_ENABLE_SELECTIVE_COLLECT (1)
 
 #if !(defined(MICROPY_GCREGS_SETJMP) || defined(__x86_64__) || defined(__i386__) || defined(__thumb2__) || defined(__thumb__) || defined(__arm__))
 // Fall back to setjmp() implementation for discovery of GC pointers in registers.

py/bc.h

@@ -302,14 +302,14 @@ static inline void mp_module_context_alloc_tables(mp_module_context_t *context,
     #if MICROPY_EMIT_BYTECODE_USES_QSTR_TABLE
     size_t nq = (n_qstr * sizeof(qstr_short_t) + sizeof(mp_uint_t) - 1) / sizeof(mp_uint_t);
     size_t no = n_obj;
-    mp_uint_t *mem = m_new(mp_uint_t, nq + no);
+    mp_uint_t *mem = m_malloc_items(nq + no);
     context->constants.qstr_table = (qstr_short_t *)mem;
     context->constants.obj_table = (mp_obj_t *)(mem + nq);
     #else
     if (n_obj == 0) {
         context->constants.obj_table = NULL;
     } else {
-        context->constants.obj_table = m_new(mp_obj_t, n_obj);
+        context->constants.obj_table = m_malloc_items(n_obj);
     }
     #endif
 }

py/circuitpy_mpconfig.h

@@ -62,6 +62,7 @@ extern void common_hal_mcu_enable_interrupts(void);
 #define MICROPY_EMIT_X64                 (0)
 #define MICROPY_ENABLE_DOC_STRING        (0)
 #define MICROPY_ENABLE_FINALISER         (1)
+#define MICROPY_ENABLE_SELECTIVE_COLLECT (1)
 #define MICROPY_ENABLE_GC                (1)
 #define MICROPY_ENABLE_PYSTACK           (1)
 #define MICROPY_TRACKED_ALLOC            (CIRCUITPY_SSL_MBEDTLS)

py/compile.c

@@ -221,7 +221,8 @@ static void mp_emit_common_start_pass(mp_emit_common_t *emit, pass_kind_t pass)
         if (emit->ct_cur_child == 0) {
             emit->children = NULL;
         } else {
-            emit->children = m_new0(mp_raw_code_t *, emit->ct_cur_child);
+            // CIRCUITPY-CHANGE: Use m_malloc_helper with collect flag to support selective collection
+            emit->children = m_malloc_helper(sizeof(mp_raw_code_t *) * (emit->ct_cur_child), M_MALLOC_ENSURE_ZEROED | M_MALLOC_RAISE_ERROR | M_MALLOC_COLLECT);
         }
     }
     emit->ct_cur_child = 0;
@@ -3688,7 +3689,7 @@ void mp_compile_to_raw_code(mp_parse_tree_t *parse_tree, qstr source_file, bool
 
 mp_obj_t mp_compile(mp_parse_tree_t *parse_tree, qstr source_file, bool is_repl) {
     mp_compiled_module_t cm;
-    cm.context = m_new_obj(mp_module_context_t);
+    cm.context = m_malloc_with_collect(sizeof(mp_module_context_t));
     cm.context->module.globals = mp_globals_get();
     mp_compile_to_raw_code(parse_tree, source_file, is_repl, &cm);
     // return function that executes the outer module

py/emitbc.c

@@ -76,7 +76,7 @@ struct _emit_t {
 };
 
 emit_t *emit_bc_new(mp_emit_common_t *emit_common) {
-    emit_t *emit = m_new0(emit_t, 1);
+    emit_t *emit = m_new_struct_with_collect(emit_t, 1);
     emit->emit_common = emit_common;
     return emit;
 }

py/emitglue.c

@@ -54,7 +54,8 @@ mp_uint_t mp_verbose_flag = 0;
 #endif
 
 mp_raw_code_t *mp_emit_glue_new_raw_code(void) {
-    mp_raw_code_t *rc = m_new0(mp_raw_code_t, 1);
+    // CIRCUITPY-CHANGE: Use m_malloc_helper with collect flag because raw code children are allocations too.
+    mp_raw_code_t *rc = m_malloc_helper(sizeof(mp_raw_code_t), M_MALLOC_ENSURE_ZEROED | M_MALLOC_RAISE_ERROR | M_MALLOC_COLLECT);
     rc->kind = MP_CODE_RESERVED;
     #if MICROPY_PY_SYS_SETTRACE
     rc->line_of_definition = 0;