What's Your Favorite Processor on an FPGA?

John Larkin wrote:
> 
> On Mon, 22 Apr 2013 07:49:11 -0400, "Michael A. Terrell"
> <mike.terrell@earthlink.net> wrote:
> 
> >
> >John Larkin wrote:
> >>
> >> DEC wrote operating systems (TOPS10, VMS, RSTS) that ran for months between
> >> power failures, time-sharing multiple, sometimes hostile, users. We are now in
> >> the dark ages of computing, overwhelmed by bloat and slop and complexity. No
> >> wonder people are buying tablets. DEC understood things that Intel and Microsoft
> >> never really got, like: don't execute data.
> >
> >
> >   I've had Win2K run over nine months, between power failures that
> >lasted longer than the UPS batteries.  DEC had more control over the
> >computers, and a tiny fraction of the number running Windows or Linux.
> >
> >
> >   If DEC was so damned good, why were they unable to survive?  Their
> >IBM 'clone' (Rainbow 100) was very overpriced, not compatible, and did a
> >very quick death spiral.  Admit it.  It was a dinosaur company with a
> >very tiny customer base.
> 
> It was *the* minicomputer company and that changed the world.


   Really?  Could it have handled any modern application, let alone
dozens or hundreds of them at once.

John Larkin wrote:
> 
> On 22 Apr 2013 12:59:27 GMT, Allan Herriman <allanherriman@hotmail.com> wrote:
> 
> >On Sun, 21 Apr 2013 09:05:49 -0700, John Larkin wrote:
> >
> >> The annoying thing is the CPU-to-FPGA interface. It takes a lot of FPGA
> >> pins and it tends to be async and slow. It would be great to have an
> >> industry-standard LVDS-type fast serial interface, with hooks like
> >> shared memory, but transparent and easy to use.
> >
> >You've just described PCI Express.
> 
> No. PCIe is insanely complex and has horrible latency. It takes something like 2
> microseconds to do an 8-bit read over gen1 4-lane PCIe. It was designed for
> throughput, not latency.
> 
> We've done three PCIe projects so far, and it's the opposite of "transparent and
> easy to use." The PCIe spec reads like the tax code and Obamacare combined.
> 
> Next up is Thunderbolt, probably worse.


   Have you ever worked with PCI-X?

On Mon, 22 Apr 2013 16:02:14 +0000, Allan Herriman wrote:

> On Mon, 22 Apr 2013 08:16:04 -0700, John Larkin wrote:
> 
>> On 22 Apr 2013 14:57:24 GMT, Allan Herriman <allanherriman@hotmail.com>
>> wrote:
>> 
>>>On Mon, 22 Apr 2013 07:09:40 -0700, John Larkin wrote:
>>>
>>>> On 22 Apr 2013 12:59:27 GMT, Allan Herriman
>>>> <allanherriman@hotmail.com> wrote:
>>>> 
>>>>>On Sun, 21 Apr 2013 09:05:49 -0700, John Larkin wrote:
>>>>>
>>>>>> The annoying thing is the CPU-to-FPGA interface. It takes a lot of
>>>>>> FPGA pins and it tends to be async and slow. It would be great to
>>>>>> have an industry-standard LVDS-type fast serial interface, with
>>>>>> hooks like shared memory, but transparent and easy to use.
>>>>>
>>>>>You've just described PCI Express.
>>>> 
>>>> No. PCIe is insanely complex and has horrible latency. It takes
>>>> something like 2 microseconds to do an 8-bit read over gen1 4-lane
>>>> PCIe.
>>>> It was designed for throughput, not latency.
>>>
>>>I agree about it being designed for throughput, not latency.  However,
>>>with a fairly simple design, we can do 32 bit non-bursting reads or
>>>writes in about 350ns over a single lane of gen 1 through 1 layer of
>>>switching.  I suspect there's some problem with your implementation
>>>(unless your 2 microsecond figure was just hyperbole).
>> 
>> Writes are relatively fast, ballpark 350 ns gen1/4lane. Reads are slow,
>> around 2 us. That's from an x86 CPU into the PCIe hard core of an
>> Altera FPGA, cabled PCIe. A read requires two serial packets so is over
>> twice the time of a write.
> 
> 
> I thought it was faster than that.  If I remember, I'll measure some in
> the lab tomorrow.


I looked at a trace on a board at work.  I was surprised - the writes 
were fast(ish) - about 100 ns was the smallest gap I saw between writes.  
The reads were slower.  I didn't see reads closer together than about 2us.

This seems consistent with Larkin's measurements.

I'm still surprised though - 2 us is 20000 bit times on a 4 lane gen 1.
Ok, it's only 16000 bit times before the 8B10B coding.

Maybe the switch is configured for store-and-forward rather than cut 
through, or something equally silly.


Regards,
Allan

On Tue, 23 Apr 2013 02:48:53 -0400, "Michael A. Terrell"
<mike.terrell@earthlink.net> wrote:

>
>John Larkin wrote:
>> 
>> On Mon, 22 Apr 2013 07:49:11 -0400, "Michael A. Terrell"
>> <mike.terrell@earthlink.net> wrote:
>> 
>> >
>> >John Larkin wrote:
>> >>
>> >> DEC wrote operating systems (TOPS10, VMS, RSTS) that ran for months between
>> >> power failures, time-sharing multiple, sometimes hostile, users. We are now in
>> >> the dark ages of computing, overwhelmed by bloat and slop and complexity. No
>> >> wonder people are buying tablets. DEC understood things that Intel and Microsoft
>> >> never really got, like: don't execute data.
>> >
>> >
>> >   I've had Win2K run over nine months, between power failures that
>> >lasted longer than the UPS batteries.  DEC had more control over the
>> >computers, and a tiny fraction of the number running Windows or Linux.
>> >
>> >
>> >   If DEC was so damned good, why were they unable to survive?  Their
>> >IBM 'clone' (Rainbow 100) was very overpriced, not compatible, and did a
>> >very quick death spiral.  Admit it.  It was a dinosaur company with a
>> >very tiny customer base.
>> 
>> It was *the* minicomputer company and that changed the world.
>
>
>   Really?  Could it have handled any modern application, let alone
>dozens or hundreds of them at once.

As I recall, Unix and C were invented for the PDP11. As was Arpanet and the
Internet. The PDP8 was the first "personal" computer, a computer that one person
could buy and use all by himself, to automate a lab experiment or (in my case)
simulate a steamship power train. That changed everything.

DECs RT11 OS was cloned to become CPM and Microsoft DOS, and lives on in the
Windows command line.

What sort of computing system did you have in 1969? I had a PDP8 running Focal.
What did you compute on in 1975? I had a PDP11 timeshare system with around 20
users. 


-- 

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

On 23 Apr 2013 12:24:47 GMT, Allan Herriman <allanherriman@hotmail.com> wrote:

>On Mon, 22 Apr 2013 16:02:14 +0000, Allan Herriman wrote:
>
>> On Mon, 22 Apr 2013 08:16:04 -0700, John Larkin wrote:
>> 
>>> On 22 Apr 2013 14:57:24 GMT, Allan Herriman <allanherriman@hotmail.com>
>>> wrote:
>>> 
>>>>On Mon, 22 Apr 2013 07:09:40 -0700, John Larkin wrote:
>>>>
>>>>> On 22 Apr 2013 12:59:27 GMT, Allan Herriman
>>>>> <allanherriman@hotmail.com> wrote:
>>>>> 
>>>>>>On Sun, 21 Apr 2013 09:05:49 -0700, John Larkin wrote:
>>>>>>
>>>>>>> The annoying thing is the CPU-to-FPGA interface. It takes a lot of
>>>>>>> FPGA pins and it tends to be async and slow. It would be great to
>>>>>>> have an industry-standard LVDS-type fast serial interface, with
>>>>>>> hooks like shared memory, but transparent and easy to use.
>>>>>>
>>>>>>You've just described PCI Express.
>>>>> 
>>>>> No. PCIe is insanely complex and has horrible latency. It takes
>>>>> something like 2 microseconds to do an 8-bit read over gen1 4-lane
>>>>> PCIe.
>>>>> It was designed for throughput, not latency.
>>>>
>>>>I agree about it being designed for throughput, not latency.  However,
>>>>with a fairly simple design, we can do 32 bit non-bursting reads or
>>>>writes in about 350ns over a single lane of gen 1 through 1 layer of
>>>>switching.  I suspect there's some problem with your implementation
>>>>(unless your 2 microsecond figure was just hyperbole).
>>> 
>>> Writes are relatively fast, ballpark 350 ns gen1/4lane. Reads are slow,
>>> around 2 us. That's from an x86 CPU into the PCIe hard core of an
>>> Altera FPGA, cabled PCIe. A read requires two serial packets so is over
>>> twice the time of a write.
>> 
>> 
>> I thought it was faster than that.  If I remember, I'll measure some in
>> the lab tomorrow.
>
>
>I looked at a trace on a board at work.  I was surprised - the writes 
>were fast(ish) - about 100 ns was the smallest gap I saw between writes.  
>The reads were slower.  I didn't see reads closer together than about 2us.
>
>This seems consistent with Larkin's measurements.
>
>I'm still surprised though - 2 us is 20000 bit times on a 4 lane gen 1.
>Ok, it's only 16000 bit times before the 8B10B coding.
>
>Maybe the switch is configured for store-and-forward rather than cut 
>through, or something equally silly.
>
>
>Regards,
>Allan

Yup, PCIe is a pig! It's OK if you do DMA, but there's a heap of overhead
building descriptor chains and fielding interrupts and stuff.

We're writing a Linux driver that lets us build descriptor chains and keep them,
and reuse them as needed to repeat data transfers.


-- 

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

On Tue, 23 Apr 2013 02:50:10 -0400, "Michael A. Terrell"
<mike.terrell@earthlink.net> wrote:

>
>John Larkin wrote:
>> 
>> On 22 Apr 2013 12:59:27 GMT, Allan Herriman <allanherriman@hotmail.com> wrote:
>> 
>> >On Sun, 21 Apr 2013 09:05:49 -0700, John Larkin wrote:
>> >
>> >> The annoying thing is the CPU-to-FPGA interface. It takes a lot of FPGA
>> >> pins and it tends to be async and slow. It would be great to have an
>> >> industry-standard LVDS-type fast serial interface, with hooks like
>> >> shared memory, but transparent and easy to use.
>> >
>> >You've just described PCI Express.
>> 
>> No. PCIe is insanely complex and has horrible latency. It takes something like 2
>> microseconds to do an 8-bit read over gen1 4-lane PCIe. It was designed for
>> throughput, not latency.
>> 
>> We've done three PCIe projects so far, and it's the opposite of "transparent and
>> easy to use." The PCIe spec reads like the tax code and Obamacare combined.
>> 
>> Next up is Thunderbolt, probably worse.
>
>
>   Have you ever worked with PCI-X?

No, but it's mostly dead, as PCI will soon be. Intel busses only last a few
years each. 




-- 

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

On 04/22/2013 07:03 AM, John Larkin wrote:

>>   Intel & Microsoft are the mainstream today, and DEC is in the
>> scrapyard of computing history.
>
> Tragically so. The thing that Intel and Microsoft had in common was brutal
> rapaciousness. That often overcomes quality.
>
> But Dec's PDP-11 bagat Unix, which begat Linux, then MacOS, then Android...
>
>
Bell Labs begat Unix, Ken Olson considered Unix "Snake Oil", didn't want 
anything to do with it.
-bill

On Tue, 23 Apr 2013 13:46:55 -0700, Bill Martin <wwm@wwmartin.net>
wrote:

>On 04/22/2013 07:03 AM, John Larkin wrote:
>
>>>   Intel & Microsoft are the mainstream today, and DEC is in the
>>> scrapyard of computing history.
>>
>> Tragically so. The thing that Intel and Microsoft had in common was brutal
>> rapaciousness. That often overcomes quality.
>>
>> But Dec's PDP-11 bagat Unix, which begat Linux, then MacOS, then Android...
>>
>>
>Bell Labs begat Unix, Ken Olson considered Unix "Snake Oil", didn't want 
>anything to do with it.

But C and Unix were developed on the PDP11. Unix was "born" in 1970,
on a PDP11/20. C is practically a PDP11 assembler.

>-bill

KO made a lot of mistakes. But DECs architectures - PDP11, VAX, Alpha
- were great. And their operating systems were excellent. It's
unfortunate that the winners were Microsoft and Intel.


-- 

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser drivers and controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

On Tuesday, 23 April 2013 23:59:05 UTC+10, John Larkin  wrote:
> On Tue, 23 Apr 2013 02:48:53 -0400, "Michael A. Terrell"
> <mike.terrell@earthlink.net> wrote:=20
> >John Larkin wrote:
> >> On Mon, 22 Apr 2013 07:49:11 -0400, "Michael A. Terrell"
> >> <mike.terrell@earthlink.net> wrote:
> >> >John Larkin wrote:
> >> >>
> >> >> DEC wrote operating systems (TOPS10, VMS, RSTS) that ran for months=
 =20
> >> >> between power failures, time-sharing multiple, sometimes hostile, =
=20
> >> >> users. We are now in the dark ages of computing, overwhelmed by blo=
at=20
> >> >> and slop and complexity. No wonder people are buying tablets. DEC=
=20
> >> >> understood things that Intel and Microsoftnever really got, like: d=
on't > >> >> execute data.=20
> >> > =20
> >> >   I've had Win2K run over nine months, between power failures that
> >> >lasted longer than the UPS batteries.  DEC had more control over the=
=20
> >> >computers, and a tiny fraction of the number running Windows or Linux=
.
> >> >=20
> >> >   If DEC was so damned good, why were they unable to survive?  Their=
=20
> >> >IBM 'clone' (Rainbow 100) was very overpriced, not compatible, and di=
d a
> >> >very quick death spiral.  Admit it.  It was a dinosaur company with a
> >> >very tiny customer base.=20
> >> =20
> >> It was *the* minicomputer company and that changed the world.=20
> >=20
> >   Really?  Could it have handled any modern application, let alone=20
> >dozens or hundreds of them at once.
> =20
> As I recall, Unix and C were invented for the PDP11. As was Arpanet and t=
he
>  Internet. The PDP8 was the first "personal" computer, a computer that on=
e =20
> person could buy and use all by himself, to automate a lab experiment or =
(in =20
> my case) to simulate a steamship power train. That changed everything.
> =20
> DECs RT11 OS was cloned to become CPM and Microsoft DOS, and lives on in =
the
> Windows command line.
> =20
> What sort of computing system did you have in 1969? I had a PDP8 running=
=20
> Focal.
>=20
> What did you compute on in 1975? I had a PDP11 timeshare system with arou=
nd=20
> 20 users.=20

I had access to a PDP-8 in 1968. It ran Focal, but I programmed it in it's =
assembly language - Macro-8 - so I could set up and interrupt-driven progra=
m to monitor and document the chemical reaction I was researching.

I crunched the - more or less - raw numbers I got from the PDP-8 (it did se=
lective data averaging, but no more) on an IBM 7040/44. The PDP-8 numbers c=
ame out on paper tape, and went through a tape-to-Hollerith card converter =
before they got onto the IBM machine, which I programmed in Fortran, and go=
t to run as a single user machine for hours in the middle of the night - us=
ually 2:00am to 6:00am.

In my first industrial job, I had remote access to a Control Data machine r=
unning Fortran from an ASR-33 teletype terminal, but after that it was most=
ly DEC.=20

IRRR DEC failed because it didn't take personal computers seriously, rather=
 like the railway companies who thought that they were in the railway busin=
ess when they were actually in the transportation business.

--=20
Bill Sloman, Sydney

On 4/23/2013 5:58 PM, John Larkin wrote:
> On Tue, 23 Apr 2013 13:46:55 -0700, Bill Martin<wwm@wwmartin.net>
> wrote:
>
>> On 04/22/2013 07:03 AM, John Larkin wrote:
>>
>>>>    Intel&  Microsoft are the mainstream today, and DEC is in the
>>>> scrapyard of computing history.
>>>
>>> Tragically so. The thing that Intel and Microsoft had in common was brutal
>>> rapaciousness. That often overcomes quality.
>>>
>>> But Dec's PDP-11 bagat Unix, which begat Linux, then MacOS, then Android...
>>>
>>>
>> Bell Labs begat Unix, Ken Olson considered Unix "Snake Oil", didn't want
>> anything to do with it.
>
> But C and Unix were developed on the PDP11. Unix was "born" in 1970,
> on a PDP11/20. C is practically a PDP11 assembler.
>
>> -bill
>
> KO made a lot of mistakes. But DECs architectures - PDP11, VAX, Alpha
> - were great. And their operating systems were excellent. It's
> unfortunate that the winners were Microsoft and Intel.

"Unfortunate"...  KO made the humongously large mega-mistake of thinking 
there was no need for personal computers.  But to his credit at that 
time computers were a bit larger than they are now.  I think there was 
an ad of a beautiful woman in late 60's attire, perhaps rather elegant 
attire, but in a kitchen standing by a decwriter printing something. 
Not quite the ticket for storing recipes really.

Bill, on the other hand only made two mistakes.  The first was thinking 
there would never be a need for more than 640 kB of computer memory.  He 
got over that one pretty quickly since he wasn't designing hardware. 
But then he seems to have felt the Internet was not something important 
to a company writing computer OS.  lol  They saw the winds blowing a 
different direction and had to change their rigging all the way into 
court to justify why they were trying to kick other browsers off "their" 
computers.

Who else made fatal mistakes were made in the computer industry?  Who at 
Osborn decided to promote the next generation before they were ready to 
ship and killed the current sales?

Why did the Alpha die?  Was that more an issue of DEC going away?  I 
don't recall who ended up with it.  Was it Intel who let it die a 
lingering death?

-- 

Rick