diff --git a/python/jittor/__init__.py b/python/jittor/__init__.py
index 068cc7b2..645a65d0 100644
--- a/python/jittor/__init__.py
+++ b/python/jittor/__init__.py
@@ -8,7 +8,7 @@
 # This file is subject to the terms and conditions defined in
 # file 'LICENSE.txt', which is part of this source code package.
 # ***************************************************************
-__version__ = '1.2.2.43'
+__version__ = '1.2.2.44'
 from . import lock
 with lock.lock_scope():
     ori_int = int
diff --git a/python/jittor/compile_extern.py b/python/jittor/compile_extern.py
index 68f2ed8f..d1f3b62d 100644
--- a/python/jittor/compile_extern.py
+++ b/python/jittor/compile_extern.py
@@ -271,6 +271,7 @@ def setup_cutt():
 
 def install_nccl(root_folder):
     url = "https://github.com/NVIDIA/nccl/archive/v2.8.4-1.tar.gz"
+    url = "https://codeload.github.com/NVIDIA/nccl/tar.gz/v2.8.4-1"
 
     filename = "nccl.tgz"
     fullname = os.path.join(root_folder, filename)
diff --git a/python/jittor/misc.py b/python/jittor/misc.py
index 671f9f20..13e5aad8 100644
--- a/python/jittor/misc.py
+++ b/python/jittor/misc.py
@@ -897,7 +897,7 @@ def auto_parallel(n, src, **kw):
         tid_def += f"\nauto tnum{i} = 1<<tn{i};"
         tid_def += f"\ntid = tid>>tn{i};"
     for i in range(n):
-        tid_loop += f"\nfor (int i{i}=tid{i}; i{i}<{pnargs2[i]}; i{i}+=tn{i})"
+        tid_loop += f"\nfor (int i{i}=tid{i}; i{i}<{pnargs2[i]}; i{i}+=tnum{i})"
         call_args.append(pnargs2[i])
         call_args.append(f"i{i}")
     call_args += oargs2
diff --git a/python/jittor/nn.py b/python/jittor/nn.py
index c766bed9..e299743f 100644
--- a/python/jittor/nn.py
+++ b/python/jittor/nn.py
@@ -1053,8 +1053,10 @@ def upsample(img, size, mode="nearest", align_corners=False):
         x = (hid + 0.5) * (h / H) - 0.5
         y = (wid + 0.5) * (w / W) - 0.5
     else:
-        x = hid * (h / H)
-        y = wid * (w / W)
+        x = hid * (h / H) + (h / H * 0.5 - 0.5)
+        if H > h: x = x.clamp(0, h - 1)
+        y = wid * (w / W) + (w / W * 0.5 - 0.5)
+        if W > w: y = y.clamp(0, w - 1)
     return _interpolate(img, x, y, (nid, cid), mode)