pynuttx: optimize memory leak check by skipping previously checked po…

…inters

Reduce leak analysis time from 12s to 7.8s, achieving a 35% performance improvement that tested on an internal project.

1. Use memoryview to avoid creating int objects, reducing overhead
2. Store checked pointers in a dictionary for efficient lookups, improving performance.
3. While dictionary operations have some cost, the overall speedup is significant.

Signed-off-by: xuxingliang <xuxingliang@xiaomi.com>

tools/pynuttx/nxgdb/memleak.py

@@ -131,19 +131,34 @@ def collect_leaks(
 
         longsize = utils.get_long_type().sizeof
 
+        checked_ptr = {}
+
         def pointers(node: memdump.MMNodeDump) -> Generator[int, None, None]:
             # Return all possible pointers stored in this node
             size = node.nodesize - node.overhead
             memory = node.read_memory()
+            prev_mem = None
             while size > 0:
                 size -= longsize
-                ptr = int.from_bytes(memory[size : size + longsize], "little")
+                mem = memory[size : size + longsize]
+                if mem == prev_mem:
+                    continue
+
+                prev_mem = mem
+                ptr = int.from_bytes(mem, "little")
+                if ptr in checked_ptr:
+                    continue
+
+                checked_ptr[ptr] = True
+
                 if any(region["start"] <= ptr < region["end"] for region in regions):
                     yield ptr
 
         print("Leak analyzing...", flush=True, end="")
         t = time.time()
         for good in good_nodes:
+            if not sorted_addr:  # All nodes are checked
+                break
             for ptr in pointers(good):
                 if not (idx := bisect.bisect_right(sorted_addr, ptr)):
                     continue