another board P939

Looks like a cool board, the availability doesn't look so cool tho:

That's the solder sample, nominally free. PCB houses always make a few extra boards to cover defects, and we just got one that they would have thrown away.

One December, a PCB house called us and gave us a list of extra boards that they had around, extras left over from our various orders. They offered them half price, to fund their Christmas party.

We have a bunch in stock. Presumably this crunch will be over some day.

I've tended to rely on the microchip 8-bitters a lot in my work, and they've been really thin on the ground lately, particularly the parts with a luxurious 8k of Flash or more.

Parts with 1k are readily available though which encourages creativity, sometimes by offloading a few functions to a SPLC or using two instead of one I've got my relatively low-volume jobs done to my client's satisfaction.

We are considering a soft-core RiscV inside an efinix FPGA, sort of a poor-persons Zynq. Lots of applications need a cpu, but not much.

Isn't there a soft power architecture core meanwhile? Some time ago (a year or two) IBM made it "open". If available it might be a better option than risk-v or arm.

efinix has a risc-v block available.

A modest program could run in on-chip sram.

I suspect that risc-v has a big future. It's free.

torsdag den 23. juni 2022 kl. 21.35.30 UTC+2 skrev John Larkin:

yep, so people can invest time and effort in it knowing that it's theirs to use for free forever

With fpga's and uPs going end-of-life so often, one could imagine owning a block of Verilog or VHDL that could be recompiled for another FPGA when needed. That cuts the risk about in half.

Has anyone used risc-v? I wonder what the code density is like.

Methinks Power PC is to fat to do as a soft core.

I already had the pleasure with Power PC in a previous life. A Pipeline pig with 1024 ultrasonics channels and a PPC for each 16 or 32, don't remember. Luckily, I was responsible only for the Virtex FPGA nd the Coolrunner; the software guy nearly shot himself.

< >

Xilinx had a fixed core PPC in one of their Virtexes. I heard Peter Alfke complain that it was wrong to license it from Motorola instead of IBM. It seems, the price was lower, but support was nonexistent.

Cheers, also to Peter if he's still alive!

Gerhard

Am 23.06.22 um 22:53 schrieb John Larkin:

You get the VHDL source for the PicoBlaze, but you are allowed only to run it on Xilinx hardware. I once 'repaired' a picoblaze. It's registers ar really little windows into the configuration ram (called block rams). For a space application we had to scrubb the configuration ram along with triple module redundancy to make it more radiation resistant. That did overwrite the registers with the reset contents. The software people did not like that. I replaced the registers by banks of DFFs. That was not too hard to do.

After 50 years of research in this area I would exspect a saturation effect.

Cheers, Gerhard

On Thursday, 23 June 2022 at 15:24:53 UTC-7, Gerhard Hoffmann wrote: ...

Sorry to say he died over 10 years ago.

kw

Peter was cool. One day at the Foothill Flea Market I had my head down inside a big box of books and another head appeared. It was Peter.

He was a real FPGA missionary, and had fun doing it.

Perhaps so, though it would be nice to have a 603e. Can't see why risc-v would be much smaller, some of the small ARMs certainly would be smaller. But I am not (yet) much of an fpga person so I can't really judge what is a practical size.

Could not make out the PPC part no., was it that 404 (420?) or something like that from IBM? I considered it at some point some

15 or 20 years ago and got frightened by its errata sheet, was clearly half baked at best.

Which coolrunner did you use, the old Philips one or the Xilinx version (I have used both, even wrote my own logic compiler for the Philips one some 20-odd years ago).

I remember that, but I am pretty sure the PPC was exactly that half-baked IBM core I remember, not Motorola. Motorola did some really good 603e spinoffs, their failure was a 7500 or something core (could be hung by user level code so that only hard reset would have an effect....). Now they (well, NXP) have that 5500/6500 cores (64 bit beasts) waiting in my drawer to get around to them, hopefully they are as good as they look so far to me.

He was a really fine guy. Went out of his way to find some info for me at some point, a really knowledgeable guy. Someone else already mentioned that he died.

Dimiter

====================================================== Dimiter Popoff, TGI

I see, having the core on chip makes things different.

Perhaps so, but Power is also free now. I have only looked at risc-v docs - to begin with it is little-endian (a more serious drawback than people have been made to think for decades though not a show-stopper). I have been living with power for well over

20 years now, have had no buying issues (knock on wood) and have yet to identify any shortcoming in the architecture. Those who did it at IBM knew damn well what they were doing.

so is all the other major platforms

what serious drawbacks?

Do you have enough experience at low level programming to be able to judge that.

since you are not coming up with any examples I'll just assume there isn't any ....

Am 24.06.22 um 18:40 schrieb Dimiter_Popoff:

<

Arghh, the <> brackets ask the posting software not to insert CRLF, tabs or whatever that breaks things like URLs. This here would break if it was just a few chars longer.

The chip in the photo reads PPC405 in the small print.

The Coolrunner below/left of the PPC is XILINX. I still use Coolrunners 2c64 or 128 as garbage collectors.

Reminds me at the never ending instructions in the 68040/50? that could produce exceptions in a circular way, shortly b4 Moto gave up. Oh so wonderful stuff like double-memory- indirect deferred addressing. Who says x86 is complicated?

Gerhard.