Re: Crazy cross-MIPS-boards stunts possible?

On Wed, 2 Dec 2020, Swift Griggs wrote:

> > sgimips is big-endian, so "mipseb".  There are a couple of ABIs supported
> > ... "o32" (traditional 32-bit) and "n32" (32-bit ABI for 64-bit processors).
> 
> Ah, I remember those terms (n32/o32) from IRIX and SGI-land, I guess that
> situation is the same for NetBSD.

 Actually anywhere in the MIPS world, including Linux and bare metal 
embedded applications.  A newer ABI (NUBI) was outlined to phase out the 
old o32/n32/n64 ones along with their deficiencies, but never implemented: 
<https://www.linux-mips.org/wiki/NUBI> (yes, the cert is outdated and 
there's currently no way to fix it).  Then there were further attempts 
with the nanoMIPS architecture, including a functional toolchain, but it 
all has gone once the latest incarnation of the owner of the architecture 
went bust earlier this year.

> > With a few exceptions (for very system-specific programs), mipseb-o32
> > binaries built for sgimips will run just fine on e.g. evbmips (like MIPS
> > Malta) running in big-endian mode.
> 
> Oh, neat. So, I guess that means the Malta boards are/were bi-endian. It
> appears the only way to get a Malta board is to be patient and watch ebay.
> Nobody makes anything like that now, right?

 The Malta was designed by MIPS Copenhagen (a former branch of LSI Logic) 
back in ~2000 and produced in rather limited quantities (e.g. mine has S/N 
534; I don't think I've ever seen one numbered 2000 or higher).  Once the 
Copenhagen site shut down in 2002 production stopped and all the remaining 
board material, including in particular blank PCBs, was moved to the US.  
It was then used to assemble more boards sometime in 2000s.

 Given the persistent demand around 2008 a new, software-compatible 
variant was designed, called Malta-R, based on documentation from MIPS 
Copenhagen, as some parts went out of production and consequently the PCB 
had to be relaid to accomodate their replacements.  The immediately 
visible difference is that unlike the original it has 3 rather than 4 PCI 
slots.  It was short-lived however, because production was limited by 
whatever remaining quantity of Intel PIIX4 South Bridge stock the company 
had, as the part has long gone out of production by then, and no 
compatible replacement was available.

 So no, nobody makes them anymore and neither there is a feasible way to.  
You may have a hard time finding one that someone wants to get rid off, as 
usually with this kind of stuff.  Most everyone seems to be using QEMU 
rather than actual Malta boards.

 NB a completely new design using contemporary components and interfaces 
was of course also made, around 2009, which was called SEAD-3.  Unlike the 
older SEAD and SEAD-2 devices it was actually usable out of the box for 
running an OS, as it had storage (microSD), USB and Ethernet interfaces.  
I gather NetBSD may not have support for that board (Linux does, as it was 
contributed by Imagination Technologies, the owner of the architecture at 
the time), but that is probably irrelevant as I doubt getting a SEAD-3 
would be any easier than a Malta.

 The SEAD-3 has an FPGA rather than a core card, so (unlike the Malta) it 
always has a soft CPU core.  It has been notable for microMIPS CPU 
support.

 Yes, the Malta, as all development boards designed by MIPS Technologies 
and the company's descendents, can boot the CPU in either endianness, 
either by flipping a DIP switch, or, with the SEAD-3, via a software 
configuration setting (also adjustable with a button press sequence).  
They are development boards after all, so they had to suit whatever the 
customer's requirements were.

 FWIW,

  Maciej