Fix upsampling of multichannel textures

drjit/tensor.py

@@ -110,11 +110,11 @@ def upsample(t, shape=None, scale_factor=None):
 
         # Concatenate output values to a flatten buffer
         channels = len(values)
-        width = _dr.width(values[0]) * channels
-        index = _dr.arange(_dr.uint32_array_t(value_type), width) // channels
-        data = _dr.zeros(value_type, width)
+        width = _dr.width(values[0])
+        index = _dr.arange(_dr.uint32_array_t(value_type), width)
+        data = _dr.zeros(value_type, width * channels)
         for c in range(channels):
-            _dr.scatter(data, values[c], index + c)
+            _dr.scatter(data, values[c], channels * index + c)
 
         # Create the up-sampled texture
         texture = type(t)(shape[:-1], channels,

tests/python/test_tensor.py

@@ -407,3 +407,9 @@ def test14_upsampling_texture(pkg):
     assert dr.allclose(b.tensor().array, [1.0, 1.5, 2.0,
                                           2.0, 2.5, 3.0,
                                           3.0, 3.5, 4.0])
+
+    a = tex_t(t([1, 1, 5, 2, 2, 6, 3, 3, 7, 4, 4, 8], shape=(2, 2, 3)), filter_mode=dr.FilterMode.Linear)
+    b = dr.upsample(a, shape=[3, 3])
+    assert dr.allclose(b.tensor().array, [1.0, 1.0, 5.0, 1.5, 1.5, 5.5, 2.0, 2.0, 6.0,
+                                          2.0, 2.0, 6.0, 2.5, 2.5, 6.5, 3.0, 3.0, 7.0,
+                                          3.0, 3.0, 7.0, 3.5, 3.5, 7.5, 4.0, 4.0, 8.0])