On 28/12/25 11:54, Paolo Bonzini wrote:
Il mer 24 dic 2025, 16:22 Philippe Mathieu-Daudé <[email protected] <mailto:[email protected]>> ha scritto:Replace the ldn_p/stn_p() calls by their explicit endianness variants. Duplicate the MemoryRegionOps, replacing the single DEVICE_NATIVE_ENDIAN entry by a pair of LITTLE and BIG ones. Select the proper MemoryRegionOps in subpage_init().This would need further adjustment when the target endianness is not fixed, but luckily the extra complexity you're introducing is not needed. The two ops add either 0 or 2 swaps to the underlying flatview_read/flatview_write so you only need one copy. As long as the ldn_*_p is consistent with the endianness of the MemoryRegionOps the result is the same.
Ack, patch dropped.
