This PR fixes an issue that came about with the recent release of NumPy 2.0. There were a few places where np.array() was used with copy=False, and that started causing exceptions because NumPy could not avoid making a copy.
To solve this, I simply removed copy=False. The performance impact should be low, because the copy=False option only came into play when writing NumPy integers (not arrays) as a UMI or AXI/AXI-Lite payload. Worst case, a few extra bytes will have to be copied for write() and atomic() operations, which should not be a bottleneck. In fact, it's possible that those bytes were already being copied and not throwing an exception under NumPy 1.0 behavior.
This PR fixes an issue that came about with the recent release of NumPy 2.0. There were a few places where
np.array()
was used withcopy=False
, and that started causing exceptions because NumPy could not avoid making a copy.To solve this, I simply removed
copy=False
. The performance impact should be low, because thecopy=False
option only came into play when writing NumPy integers (not arrays) as a UMI or AXI/AXI-Lite payload. Worst case, a few extra bytes will have to be copied forwrite()
andatomic()
operations, which should not be a bottleneck. In fact, it's possible that those bytes were already being copied and not throwing an exception under NumPy 1.0 behavior.