The driver supports the Tundra semiconductors Universe PCI/VME-Bus bridge. It is included in a a VME-Bus CPU, the ELTEC E128, a Pentium-CPU with the usual PC Peripherals, a NCR53C810 SCSI controller and NE2000-Compatible Ethernet.
This CPU is the one on which i have tested the driver.
Another CPU with this PCI/VME Bridge is the Dynatem DCP1, which i have not yet seen or tested.
An important point to notice is that the Universe does not support hardware byte swapping. That means, that if the VME-Cpu is a little-endian machine, i.e. it has the 'wrong' byte order which is used on the i386, all the data read/written from/to the VME-Bus are in the wrong byte order! The VME-Bus is always big-endian, i.e. the more significant bytes have lower addresses, so that the most significant byte comes first. This byte order is also known as the network byte order, and that allows to use the ntohs()/htons() ntohl()/htonl() set of functions that are defined in netinet/in.h.
The driver does not do any byte-swapping on its own, because it does know nothing about the type of the data that are to be transferred, and doing a byte swap without knowing that would mess up the data, which is exactly the reason of the Universes inability to byte-swap in hardware, see http://www.tundra.com/Tundra/Products/Downloads/BIC/Universe/Articles/Intfacng.pdf, a reprint from VMEbus Systems Magazine, August 1996.