Subject: RE: A VAX microcode optimized Java VM?
To: port-vax@netbsd.org, classiccmp <classiccmp@classiccmp.com>
From: Carlini, Antonio <Antonio.Carlini@riverstonenet.com>
List: port-vax
Date: 10/28/2001 12:24:41
> Gunther Schadow wrote:
> 
	>(c) the VAX 11/750 also has an easy access for microcode hacks
	>     as used by NetBSD.

	The original VAX-11/750 came with permament 
	microcode - an update required changing some
	hige number of roms (I heard in the three figures
	but I never checked myself). Shortly afterward
	an updated board was issued that allowed for
	patchable microcode, via a boottime load. 
	There is a limit to how much patching you can do
	without swapping ROMs. Enough for DEC to
	issue fixes (last I know of is 104 with VMS V5.0 but
	I stopped using a 750 shortly afterwards) but
	possibly not enough for you to implement
	Java primitives!
	 
	>Question remains: what about other platforms? Espcially the

	The VAX-11/730 was entirely microcoded: one of the
	DTJs describes how they used the 730 as the platform
	for testing out the MicroVAX II instruction subsetting
	before the uVAX II chip existed.

	The VAX-11/780 had microcoding tools available
	for customer use (and I assume the same was
	true of the VAX-11/785 since it was essentially
	the same but faster).

	I know of no such tools for the VAX 82xx/83xx
	machines.

	The Nautilus stuff (VAX 85xx/8700/8800) must
	have been microprogrammable because early
	Alpha development was done on Nautilus
	(or Polarstar ... much the same h/w) with ucode
	assists before the EV3 units were built.

	I don't know whether there was any microcode
	accessible to the user on CVAX or Rigel
	or Mariah. The NVAX chip spec mentions
	a bit that is set to indicate that the microcode
	has been patched (and I believe that at least
	one platform does this to disable the VIC) so
	presumably there is some level of hackability.

	>6000 series? I think the only way to update microcode would be

	Which ones? CVAX (62xx/63xx), Rigel (64xx),
	Mariah (65xx) or NVAX (66xx)? Anyway,
	I think the answer is *no* (with a possible
	exception for NVAX).

	>Next question is: would every generation of VAX require a
	>completely custom microcode hack or is there a level of
	>microcode compatibility?

	All the ones I've ever been even peripherally close
	to used MICRO as the assembler but beyond
	that I expect that they were all mostly
	different (although each might have been 
	heavily influenced by its predecessor).

	Antonio