优化: KGM解密并行化，多线程处理block，提升服务器卡顿时转码速度

vipsong.py

@@ -9,6 +9,7 @@ from mutagen.flac import FLAC
 from mutagen.oggvorbis import OggVorbis
 from mutagen.id3 import ID3, TIT2, TPE1, TALB, APIC
 from pydub import AudioSegment
+from concurrent.futures import ThreadPoolExecutor
 
 NCM_MAGIC = b"CTENFDAM"
 CORE_KEY = b"hzHRAmso5kInbaxW"
@@ -262,16 +263,26 @@ class KgmDecoder:
         pub_len = len(self.pub_key)
         out = bytearray(len(audio))
         block_count = (len(audio) + BLOCK - 1) // BLOCK
-        for block_idx in range(block_count):
-            if block_idx >= pub_len:
-                raise ValueError("Key file mismatch")
+        if block_count > pub_len:
+            raise ValueError("Key file mismatch")
+        # Pre-compute all 17 phase tables; workers only read from cache
+        for phase in range(17):
+            self._phase_tables(phase, own_key, self.pub_key)
+        # Parallel block decryption: each block is independent
+        def _decrypt_block(block_idx):
             phase = block_idx % 17
             tables = self._phase_tables(phase, own_key, self.pub_key)
             pub_value = self.pub_key[block_idx]
             start = block_idx * BLOCK
             end = min(start + BLOCK, len(audio))
-            for col, offset in enumerate(range(start, end)):
-                out[offset] = tables[col][pub_value][audio[offset]]
+            blk = bytearray(end - start)
+            for col in range(end - start):
+                blk[col] = tables[col][pub_value][audio[start + col]]
+            return start, bytes(blk)
+        max_workers = min(16, max(1, block_count))
+        with ThreadPoolExecutor(max_workers=max_workers) as executor:
+            for start, blk in executor.map(_decrypt_block, range(block_count)):
+                out[start:start + len(blk)] = blk
         payload = bytes(out)
         container = _sniff_container(payload[:64])
         return payload, container, {}, None