pmap, mmap, iomap... what unixes are these available on?

[Tony Anzelmo]

In article <7634 at ecsvax.uncecs.edu> uccjcm at ecsvax.uncecs.edu (John McLendon) writes:
>
>We're looking at un*x's to move our application to, but we need
>to be able to memory map a register based board we produce
>into a processes address space. HP has iomap. Sun has mmap
>which I presume can do the same thing. Regulus has a non-standard
>call that does this. Concurrent does it with mmap.
>Does anyone know if AT&T can do this with the latest SYSV/386?
>How about other un*x's?
>			Thanks in advance...
>			John
>-- 
>Signed: John McLendon                uunet\
>        (919) 846-7931 (home)		   >mcnc!ecsvax!uccjcm
>        (919) 941-5730 (play)	    gatech/


The Alliant version of UNIX (Concentrix) has recently released a rich
set of real-time extensions under the name FX/RT. It includes I/O primitives
including:

	o The ability to map process virtual addresses to external
	  bus address space (VME or multibus) via the call rt_map_vtoe.
	  This call allows direct access to device registers for control or
	  programmed I/O.

	o The ability to allocate and load external bus mapping registers
	  via rt_alloc_extba/rt_map_etov. These calls allow a dma device
	  to be controlled from a user process.

	o The ability to connect a device interrupt to a user written
	  interrupt service routine by allocating an interrupt vector
	  (rt_alloc_intvec), installing an interrupt service routine
	  (isr_install), and connecting the two (rt_conn_intr). In
	  addition, the interrupt service routine can wakeup waiting
	  processes through the use of semaphores and/or events.

The above facilities permit a complete user written driver to be developed.
In addition to these I/O related facilities, other system calls and library
routines in FX/RT include:

	o Place a process into a real-time scheduling class that uses
	  fixed-priority preemptive policies.

	o Asynchronous event and semaphore synchronization primitives.

	o A logical naming facility to allow unrelated processes to use
	  common events, semaphores, and I/O objects.

	o Shared memory support.

	o Finer control over processor scheduling (such as, binding to
	  a particular cpu) and memory residency (such as, locking a
	  process in memory to prevent paging/swapping).


Tony Anzelmo
Anzelmo Associates, Inc.
508-568-1880
...!linus!alliant!cairo!tony