modeld: 2x faster chestnut build

SConstruct

@@ -55,6 +55,16 @@ assert arch in [
   "Darwin",       # macOS arm64 (x86 not supported)
 ]
 
+if arch == "comma_arm64":
+  from openpilot.common.utils import sudo_write
+  for cpu in range(4, 8):
+    sudo_write('1', f'/sys/devices/system/cpu/cpu{cpu}/online')
+  for policy in ('0', '4'):
+    governor_path = f'/sys/devices/system/cpu/cpufreq/policy{policy}'
+    sudo_write('performance', f'{governor_path}/scaling_governor')
+    with open(f'{governor_path}/cpuinfo_max_freq') as f:
+      sudo_write(f.read().strip(), f'{governor_path}/scaling_max_freq')
+
 pkg_names = ['acados', 'capnproto', 'ffmpeg', 'json11', 'ncurses', 'zeromq', 'zstd']
 pkgs = [importlib.import_module(name) for name in pkg_names]
 acados = pkgs[pkg_names.index('acados')]

openpilot/selfdrive/modeld/SConscript

@@ -45,7 +45,7 @@ tg_devices = { # which device to put jit inputs to at runtime
 
 USBGPU = usbgpu_present()
 if USBGPU:
-  usbgpu_tg_flags = f'DEBUG=2 DEV=USB+AMD:LLVM WARP_DEV={tg_backend} FLOAT16=1 JIT_BATCH_SIZE=0 GMMU=0 TC_OPT=2'
+  usbgpu_tg_flags = f'DEBUG=1 DEV=USB+AMD:LLVM WARP_DEV={tg_backend} FLOAT16=1 JIT_BATCH_SIZE=0 GMMU=0 TC_OPT=2'
   # the USB+AMD GPU takes an exclusive flock; serialize all targets that touch it
   usbgpu_lock = File("models/.usb_gpu.lock").abspath
 
@@ -79,8 +79,8 @@ for usbgpu in [False, True] if USBGPU else [False]:
   file_prefix, cmd_flags = ('big_', usbgpu_tg_flags) if usbgpu else ('big_' if os.getenv('BIG_INTO_SMALL') else '', tg_flags)
   driving_onnx_deps = get_existing_chunks(File(f"models/{file_prefix}driving_supercombo.onnx").abspath)
   camera_res_args = ' '.join(f'{cw}x{ch}' for cw, ch in CAMERA_CONFIGS)
-  # CPU 7 is isolated with isolcpus on AGNOS, so explicitly pin the compiler to it.
-  taskset = 'taskset -c 7 ' if arch == 'comma_arm64' else ''
+  # CPU 7 is isolated with isolcpus on AGNOS, so explicitly include it along with the other cores.
+  taskset = 'taskset -c 0-7 ' if arch == 'comma_arm64' else ''
   cmd = (f'{cmd_flags} {mac_brew_string} {taskset}python3 {modeld_dir}/compile_modeld.py '
         f'--model-size {model_w}x{model_h} '
         f'--camera-resolutions {camera_res_args} '

openpilot/selfdrive/modeld/compile_modeld.py

@@ -3,15 +3,14 @@
 import atexit
 import math
 import os
-import tempfile
 import time
 import shutil
 from functools import partial
 from collections import namedtuple
 
 import numpy as np
 
-from openpilot.selfdrive.modeld.helpers import dump_oob, load_oob
+from openpilot.selfdrive.modeld.helpers import dump_oob
 
 def _patch_tinygrad_fetch_fw():
   import hashlib
@@ -35,7 +34,6 @@ def fetch_fw(path, name, sha256):
 from tinygrad.device import Device
 from tinygrad.engine.jit import TinyJit
 
-
 NV12Frame = namedtuple("NV12Frame", ['width', 'height', 'stride', 'y_height', 'uv_height', 'size'])
 WARP_INPUTS = ['tfm', 'big_tfm']
 POLICY_INPUTS = ['img_q', 'big_img_q', 'feat_q', 'desire_q', 'packed_npy_inputs']
@@ -257,11 +255,7 @@ def random_inputs_run(fn, seed, test_val=None, test_buffers=None, expect_match=T
 
   print('capture + replay')
   test_val, test_buffers = random_inputs_run(jit, SEED)
-  print('pickle round trip')
-  with tempfile.TemporaryFile(dir=".") as f:
-    dump_oob(jit, f)
-    f.seek(0)
-    jit = load_oob(f)
+  print('validate replay')
   random_inputs_run(jit, SEED, test_val, test_buffers, expect_match=True)
   random_inputs_run(jit, SEED+1, test_val, test_buffers, expect_match=False)
   return jit
@@ -274,6 +268,8 @@ def _parse_size(s):
 
 def read_file_chunked_to_disk(path):
   from openpilot.common.file_chunker import open_file_chunked
+  if os.path.isfile(path):
+    return path
   tmp_path = f'{path}.unchunked'
   with open(tmp_path, 'wb') as f, open_file_chunked(path) as src:
     shutil.copyfileobj(src, f)