Tag Archives: DSM

Putting the DSM to use

So after the lengthy description of the DSM cards – how can we make use of them? As said in the previous chapter, they were shipped with an assembler and even an early version of GCC (1.3) so development is pretty straightforward.

Activation

First, you have to understand how the cards integrate themselves into an ISA/EISA system. While the three versions (8, 16, 32bit) differ in some areas, the integration is more or less similar:

Each version offers a latch for controlling the card. This means to activate the card by writing bits to that latch to define a memory-window inside the hosts RAM to blend-in the cards dual-ported RAM  and/or resetting it etc.. The latch is accessible through an IO-port set by jumpers on the card (default 0x300).

So for the ISA cards you have to for example write a 0xC2 at that port-adress to reset & activate the card and use the mem-window of 0xDC000-DC7FF. In Turbo-Pascal this would be something like:

port[$300] := $c2;

This gives you a 2K mem-window to exchange data between the DSM and the host (just 1K for the DSM860-8).

The EISA cards obviously use other ports depending on the slot-number, so this would be an example to do the same for am DSM860-32, this time in Turbo-C:

outportb(slot_no * 0x1000 + 0x800, 0xc2); // For slot #2 this would be 0x2800

This would also open a mem-window at 0xDC000, this time up to 0xDCFFF, i.e. 4K long.

Memory

As mentioned above, the Host and the DSM-card are communicating through a memory-window of diffenrent sizes, depending on the DSM used. Due to their nature, the memory is looking different though. That said, at least they’re both litte-endian, so no byte-swapping needed.

The 80×86 side

For the hosting PC, memory looks pretty straightforward. 1KB-4KB of RAM somewhere in ‘lower-RAM’, that’s it.
While we don’t use it, it’s worth mentioning that there’s a 2nd memory window called “Common“. This is fixed at a specific address and is shared between all possible cards plugged into one host. I guess you already got it: This enables easy multi-processor communication… and gives a lot of possibilities for f**k-ups.

The i860 side

The memory-mapping on the i860-side is the same for the 16 and 32bit cards, the dual-ported RAM is located at 0xd0040000 (0xC0000000 for the DSM-8).
In any case the i860 memory is linear, 64bit wide and always on a 64-bit boundary. This means you have to read the DP-RAM area differently depending on which card you run your code. Here’s an example of how the DP-RAM looks like on the Host- and i860 side:

Host DP-RAM in DOS ‘debug’
-d dc00:0000
DC00:0000 11 22 33 44 55 66 77 88 ...

which would look like this on the i860 side:

DSM/8
C0000000 - 11 xx xx xx xx xx xx xx 22 xx xx xx xx xx xx xx
C0000010 - 33 xx xx xx xx xx xx xx 44 xx xx xx xx xx xx xx
C0000020 - 55 xx xx xx xx xx xx xx 66 xx xx xx xx xx xx xx
C0000030 - 77 xx xx xx xx xx xx xx 88 xx xx xx xx xx xx xx

DSM/16
D0040000 - 11 22 xx xx xx xx xx xx 33 44 xx xx xx xx xx xx
D0040010 - 55 66 xx xx xx xx xx xx 77 88 xx xx xx xx xx xx

DSM/32
D0040000 - 11 22 33 44 xx xx xx xx 55 66 77 88 xx xx xx xx

So reading and writing from/to the DP-RAM involves some thinking to be done by the programmer. Here are two code-snippets showing the difference between reading the DP-RAM on a DSM860-8 and an DSM860-16. First the ‘8 bit version’:

mov 4*8,r4
readloop:
ld.b 0(r15),r14  // Load BYTE from DP-RAM
st.b r14,0(r29)  // store it destination
addu 8,r15,r15   // add 8 to read-mem-pointer
addu 1,r29,r29   // add 1 to desitination-mem-pointer
addu -1,r4,r4    // loop-counter
xor r0,r4,r0     // Test Zero
bnc readloop

And the same for the DSM860-16:

mov 2*8,r4
readloop:
ld.s 0(r15),r14  // Load SHORT (2 Bytes) from DP-RAM
st.s r14,0(r29)  // store it destination
addu 8,r15,r15   // add 8 to read-mem-pointer
addu 2,r29,r29   // add 2(!) to desitination-mem-pointer  (short <> byte)
addu -1,r4,r4    // loop-counter
xor r0,r4,r0     // Test Zero
bnc readloop

Because of reading SHORTs (ld.s) the DSM860-16 version has to loop just 16 times while the 8-bit version has to do that 32 times.
Same applies to writing. You will find an example in the Mandelbrot program (Commented source file).

[This is work-in-progress and will expanded over time]

Action!

So here we go, finally some program running showing all the power behind the i860. I took the Mandelbrot example from R.D.Klein and modified it a bit, well quite a bit as it was written for the DSM860-8 and provided CGA output (yuck!).

Like most “external accelerator” programs, there’s one part running on the accelerator (the i860 in this case) and one part running on the host doing useful things with the provided data. In this case we have an i860 assembler code doing the number-crunching on the Mandelbrot algorithm using the i860’s ability of ‘dual instruction-mode‘ and some code done in Trubo-Pascal handling the display and zooming.
The latter was extended to use SVGA (640x480x256) output and providing an interrupt driven timer. [sourcecode package cleanup is still work in progress]

Here are the two running full steam ahead:

Some things worth to mention:

  • The host being used here is a P1 133MHz, a bit unfair comparing that to a 40MHz i860 – OTOH they seem quite comparable when it comes down to Mandelbrot crunching speed.
  • To calculate the Mandelbrot the same speed as it took the Pentium (~15s) I needed five T800-20 according to my benchmarks.
  • To even achieve the 8.2s of the i860 I had to run 9(!) T800-20 in parallel.
  • A i486DX/33 took 66 sec to do the same (8.25 times slower!), while it still took 34s for a i486DX2/66!

So while all that moaning about the bad ‘programmability’ and slow context-changes of the i860 are completely correct, in certain tasks that CPU was indeed a real screamer!

The DSM-860 Series

Based on a public project from Rolf-Dieter Klein and Tobias Thiel (“PC-Karte mit i860”) in the German computer magazine “mc” (2/90 to 7/90), the Munich based company DSM built several i860 boards for the PC/AT which they called the DSM-860 series.
All DSM-860 have one thing in common: They offer a high level of hardware features – no costs were feared. So naturally, those cards were not cheap. But you really got something for your money. All versions featured

  • 4 Transputerlinks for networking multiple cards
  • Connection to the hosts system-bus (ISA/EISA) via dual-ported RAM
  • A 16-bit bus is also available via dual-ported RAM on special connectors giving a throughput speed of 8MByte/s for high-speed connections between several SPC boards.
  • RAM was put on an extra RAM board making the complete SPC-860 a double-card sandwich

DSM-860

The 1st incarnation was the SPC-860, very quickly renamed to DSM-860, an 8-bit ISA card with 4MB RAM (DIL) and 4 10mbps Transputerlinks via four C012’s (polled by the i860 resulting in 740kbps linkspeed).

Here’s a picture from a 1992 ad, with separate RAM card attached:

print-DSM860-8

It did cost 16450 DM including the (GNU) C compiler and assembler.

DSM-860/16

Next, they released the DSM-860/16 (renaming the DSM-860 into DSM-860/8) being a full fledged 16-bit ISA card. It has a real Transputer (16bit T222, having its own 32K SRAM) for handling the “multiprocessor communication” and is able to support up to 256MB on a sandwiched daughter-board, now using SIMM modules instead of DIL parts.
The Hitachi HD63310 dual-ported RAM, used in the 8-bit version to communicate with the host was replaced by faster Cypress IDT7130 types (“because of the high speed of the 16-bit ISA interface” ;-)), resulting in a peak-rate of 14MByte/s.

Here’s the schematic of the card and its components

SPC860schem

This is how it looks in reality… my DSM-860/16:

DSM860-16

As you can see, the card is not exceptionally high integrated – even built 1992 there is not a single SMD part used, everything is socketed, only some PALs could be called “custom parts”. But this does not necessarily mean it’s badly designed or build.
If you have a close look (click the picture for a bigger version), you’ll see that every part/socket/jumper on the board is nicely specified in the silk-screen printing. All GALs and the EPROM contents are available in the documentation… which has 426 pages by the way.

Here’s the left side in more detail:

DSM860-16_left

This end of the card is the “external comms” side. Beside the all-mighty i860 you can easily spot the golden Transputer being the communication controller.
To its left, there’s the first dual-ported RAM (1k x 8) connected to the socket for the external bus (Ring-A, located on the edge above). Below that are the two 16k SRAMs -marked MHS- for the Transputer. Then to the right are the two dual-ported RAMs (a 7C131 and a so-called slave 7C141) making the 16-bit connection to the i860’s bus. The rest of the parts are quartz oscillators (5MHz & 40MHz) and drivers/buffers for the buses. On the top edge next to the Ring-connectors you can spot the 4 Transputer links (JP9-12).

The right side is comparably boring:

DSM860-16_right

The boot EPROM, 6 GALs, again two dual-ported RAMs (this time for the ISA-bus connection) and some buffers… well, and 5 LEDs. LEDs are good. 😉
Also, you can see the pin-row connectors at the lower edge and on the left of the photo. That’s the expansion-bus. The lower connector is more or less 1:1 the ISA bus, the left one is a 16-bit bus to the i860. AFAIK, they never offered an expansion for sale.
The last (but not least) interesting thing on this picture is the copyright. Yes, it’s a DSM860/16 from 1992, RDK made it (Rolf-Dieter Klein), DSM in Munich distributed it – but it’s obviously also a rev. 1.6, which means there could be others before or after that one. If you have more informations I would be happy to learn more.

DSM860-OEM/16

I’m pretty sure there was one more version after this rev.1.6, this b/w picture from a DSM press release about shipping the 250.000th slot-CPU card shows a very changed design. The silkscreen print says “DSM860-OEM/16”, so it’s obviously nothing for the normal market.
Mind the onboard-RAM, the missing comms-section and the high integration (SMD parts all-over) and even an early form of an FPGA from Lattice – my assumption is that this version could be the answer to the Kontron SBC860 showing nearly the same layout:

OEM_250ste
At least this ‘riddle’ is finally solved. I was able to buy an DSM860-OEM/16..Yay! As assumed, it is a modern (for 1990 standards) version of the DSM860-16 now consisting of just one board, so no more RAM card as described below- and without the Transputer and Ring-A/B stuff.

Here are my pictures of it. First of all, the card in full view:

OEM16_full

The right half shows a very high degree of integration compared to its predecessor. All DIL ICs were replaced by SMD parts and lot of logic went into PALs and even an FPGA (The contents of the EPROM is the same as with the other card, minus Transputer handling):

OEM16_right

The left half contains the CPU and the RAM. This time only 8 SIMM slots:

OEM16_left

The RAM card

DSM860-RAMcard

This is the 2nd part of every DSM-860 – the RAM card (except the OEM-16). Same dimensions as the SPC itself… i.e. full length. The biggest part of it is consumed by the 16 angled SIMM sockets, obviously interleaved, thus named altering slot-a & b. Only the 8 a-slots are populated on mine.
The rest of the board is used for lots of buffers and drivers, some GALs (doing the mem-mapping) and there’s quite a big copyright… again.

DSM860-RAMcard_C

One typical detail of those days is the fact that manufacturers were not very keen on having users doing upgrades themselves. Even this RAM board has standard SIMM sockets and one could simply plug in more SIMMs to expand the RAM you had to change a GAL (the one in the lower left corner with a while label saying “UXM24Wxx” on it). Obviously these GALs are the only ones notdocumented. All I know is that there were 3 GALs available differing in the last two letters of the label:

  • 8B = only a-slots can be used by either 1M or 4M SIMMs giving 8 or 32MB total
  • 16 = both a- and b-slots used with 1M SIMMs = 16MB total
  • 64 = a- and b-slots filled with 1M and/or 4M SIMMs giving 16, 40 or 64MB total.

Because the card as well as the documentation says the maximum supported amount of RAM is 256MB there might be the chance of supporting 16MB SIMMs – I did not try this yet.

Both cards attached together give quite a big and heavy sandwich

DSM860-16_sandwich

It’s obvious that you not only need 2 full-size slots but also guide-rails inside the case to hold the weight of this beast.

DSM-860/32

This seems to be the king of the SPC hill. Technically it’s pretty much the same as the DSM-860/16 but this time featuring a 32-bit system bus – namely EISA. The EISA bus was a dead-end like IBMs Microchannel but comparably easy to implement and free of license fees.

So the main difference to the DSM-860/16 is the 32-bit wide connection to the hosts bus, visible by the 4 dual-ported RAMs used for a 32-bit wide connection to the EISA slot.

Again, here’s a 1992 magazine ad showing a probably early version of the card as the marking says “DSMß860-32” (mind the beta) and the year “1991”:

print-DSM860-32

Compared to that, my version (1.2) does not look that much cluttered – also the Transputer comms part were moved to the left side of the i860 and two more LEDs were added:

DSM860-32

So the left side of the card looks pretty identical to the DSM860-16, while the right side has a bit more logic to satisfy the EISA standard, the already mentioned 2 more DP-RAMs, a bigger expansion-bus due to the EISA slot and most important: 2 more LEDs! Did I mention that LEDs are good? 😉

DSM860-32_C

This more detailed picture shows that the DSM860/32 was released the same year as the ISA version. This card is a rev.1.2 – again, if you know more about revisions, I’d be happy to hear from you.
You might have spotted that this card looks a bit shabby. That’s because it was pulled from some universities dumpster. It was missing some components and had some “scars”. The good thing was that none of the GALs were missing and due to the fact that every piece is documented on the card it was easy to replace the missing parts.
On the above picture you can clearly see e.g. the 100nF capacitor C40 below the i860 or the 40MHz OSC. I wish everything would be that well documented.

The Infinity card

This is a rare and mysterious beast. The documentation only touches it very briefly. It’s definitely nothing been built for the average DSM860 user – if something like that existed. For sure it was extremely expensive… and it has LEDs 😀

DSM-EINF860M

So at the first look you see 2×3 connectors for 40pin cables – the same used on the DSM860 cards (Ring-A and -B). Then there are a lot of drivers and buffers and a big Lattice pLSI 1032-50 which is a 6000 gates PLD (Programmable Logic Device). A closer look to the board gives more hints – thanks to the DSM (or RDK) habit to print as much info as possible onto the board:

DSM-EINF860M_C

Ok, first information we get is that this is a EINF860M or INFINITY 32Bit Extender. It’s like all other boards (c)1992 by DSM Munich and -as one would expect- designed by Rolf-Dieter Klein (RDK).
The three connectors are labeled ADDR(ess), DATA MSB (Most Significant Byte) and DATA LSB (Least Significant Byte).
The the right of the connectors is an Intel 85C098-20. I think that’s a One-Time-PLD, not 100% sure.

My educated guess is that this card is what the print on it says: A bus extender. Using the 16-Bit bus on the DSM860 cards one can build quite a big parallel computer. But the max. length of the flat-cable to connect each card with the next one is limited. So this card would be connected to other DSM860 cards in the same case (i.e. a 19″ case in a rack) and the extender would then “amplify” the bus-signals to be send over into the next rack full of DSM860 cards.
That said, my fear is, you’ll need two of those cards as the seperation of Adresses and Data (MSB and LSB) is nothing being used on the DSM860 cards. So my next guess is, that the INFINITY communicates over the EISA bus with the other cards and has its own external bus. Again – I’m happy if you contact me if you know more/better!

Nostalgia

And here’s a full page ad I found in my old (German) computer magazines – it shows all 3 versions:

DSM Software

I’m in the lucky position to own the last offically released software package for the DSM860 cards.
That said, it’s a mishmash of different tools, code and compilers not always compatible with the different boards (860/8, 860/16 and 860/32). Still the sources and specifically the assembler and C compiler may be helpful for a kickstart.
I’ve adjusted the installer to create a single-install instead of the original 3 floppies installation. See the readme.txt file for more details.

Download it here.

And if you’re interested, here is the the source & binaries for my Mandelbrot example.