Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Can anybody fill me in on the state of play on FPGA's?

W

Wayne Morellini

Well-known member
Joined
Jan 3, 2008
Messages
6,157
Reaction score
0
Points
0
I usually avoid thinking about FPGA's for the type of small low powered cameras I was thinking to do. They have been slow, power hungry little beasts and complex to make a whole camera from (though there are open source cinema camera designs coming). I want something more universal circuitry programmable than a switched network shuttling between predefined processing units. I was waiting for magnetic FPGA, even with my own technology proposal.

However, I recently realised something.that was just gathering steam by the time I had turned off of fpga. There were processors with small sections of FPGA planned. But that seemed to go to FPGAs with an arm core. However, if there was a suitable low end low energy cheap processor with camera interface and FPGA, that could be useful. In this case the processor could ingress sensor data through its interface and the FPGA provide an avenue to accelerate some codec functionality. If the FPGA could define new instructions, then those instructions could be used with other cpu instructions to accelerate a new high efficiency codec I'be been hoping to do for years (a bit complex, wanted to do an os and do the codec as the first test application. The OS was more important than codec or cameras).

So, what type of fpga programmability options and technologies exist now for low energy or high speed?

Thanks.
 
This is a huge topic...

for a bit of background maybe check out https://www.ampltd.com/

They do a lot with embedded video systems for Defense applications and similar.

I used to live over the river from this guy in the U.K.... The company founder is one of the smartest people you will ever meet in your life...

If you subscribe to Military and Aerospace magazine and NASA Technical Briefs there is a LOT (traditionally) on embedded FPGAs...

______________________________________________________________________________________________________________

Since Direct Burst technology from 3d labs/ wild cat cards / Intergraph that off loads instructions to specific graphics hardware I have always been of the opinion that Graphics GPU type architecture to carry out extremely high bandwidth number crunching in parallel is extremely powerful indeed for many many diverse applications.

CUDA and Open CL have the ability to tap into that with the right programming mentality rather than having to write everything in Assembly language...

When I first used to visit my friend over the river and I'd explain what I'm doing and I mention C and C++ they just shake their heads like "Ohhh no, no, nothing as high level as that "... Their reaction being like I just dangled a bad smelling rotten rat in front of their noses.

@ Wayne over the next few days I'll dig around for things that are in your neck of the woods as to what you are asking here.

_____________________________________________________________________________________________________________

To be honest I never dug deep on Red's secret sauce nor am I inclined to spill any beans if I do...

So back side of the silicon and first stage processing and REDCODE... I have no idea as I figured it was none of my "Beeswax" :-)
 
@Wayne ,

Scratching around a bit and following a few bread crumbs to se what became of 3d labs... And also the rise of NVIDIA...

And how Intel suddenly enters the frame once more.

Stem cell processors... [Seems eminently applicable to a portable "Media machine" perhaps? ].

https://ziirocks.wordpress.com/2009/08/11/what-is-a-stemcell-computer/

https://en.wikipedia.org/wiki/ZiiLABS#StemCell_cores

https://en.wikipedia.org/wiki/Project_Denver

https://en.wikipedia.org/wiki/ARM_architecture

^^^This you know about.

______________________________________________________________________

a bit of history

https://en.wikipedia.org/wiki/3Dlabs

https://en.wikipedia.org/wiki/Creative_Technology

______________________________________________________________________

Interesting how x employees from SGI, 3D labs and Sun microsystems and AMD seem to re-group in different capacities to form various companies ...

Also interesting how certain patents and technology once again falls into the hands of Intel...

Small world.

______________________________________________________________________

Lessons learned from the past (let's hope so): https://www.cnet.com/news/intel-settles-intergraph-suit-for-225-million/

Intel paid Intergraph all together $675 million dollars due to various infringements and bad tactics that originally killed 3d labs. Had that not happened then 3d labs could have been what NVIDIA is today ... (maybe). Although Intergraph got some justice in the end after much hassle I think it still cost them way more in the long run.

How difficult is it to "play nice" but sometimes things get complicated and I'm sure some very shrewd and mindful lawyers did pretty well out of that to. "Hired guns".

"We're now done with Integraph" Said Chuck Mulloy an Intel spokesman... (Good call).

http://techlawjournal.com/courts/intergraph/Default.htm

^^description of 1997 Alabama case and original case and basis thereof. "Nature of the Case. Intergraph, which makes workstations based on Intel microprocessors, sues its chip supplier, Intel, on a wide range of legal theories, including antitrust, anticompetitive behavior, and patent infringement, for withholding information and products.".
 
Wow thanks Eric, I was coming to quickly check on the way out.

Your friend across the river, at least I would not be alone, rather than hanging around with the future leaders here. I thought maybe Jeff might turn up and give a few summary paragraphs about where things are and are testing technically.

GPU's are still very wasteful, people I know can do a processing core in a few thousand transistors, but you get to a point of how to use hundreds of thousands low slower processors that get record low power consumption. They weren't on board and I didn't like the way they run things, and with policy on work carried out on campus by others. Moronic, and to the large disadvantage of owners. Americans just don't know how to do open culture. They're do putfullt greedy and power hungry for control and ownership that they can even be threatened by open source, a messily inately stupidedly run community. They can't see the forest for opening a few trees, instead of more trees. Starving themselves they go bankrupt on the futility of their obsessional deluded little minds. Untrustworthy bastads that want to encrouch take and even flatten the work of others. Like doing business with slick English con men. I'm sure many are not like that, but 8t is the culture, and a culture among those trying to claw their way to the top. So, I've insulted enough sub-groups today. It's interested, there is a web page that closely genetically links Europeans to a bkavk tribe in Pakistan, who give off some albino children. A business man I know was discussing with me about the issues with dealing with a certain nation, and I asked then why not dump them and deal with the Indians, and he told me bluntly that they were worse. The superior races seem to share this root, and as they are diluted, they maybe become more useful to the world before getting less useful again. Very interesting stuff. There was one race if Super Men, they used to call themselves, do you know who they were, the original Egyptians from a tribe in Mali. The depictions of them, before interbreeding, is very refined European looking, except black/dark brown, like this Pakistani tribe I mentioned. Also, there is a common base language family between Eygot and the cresent and trading roots with the Indus valley empire into Pakistan. It could be they were all originally related.

CPU parallelism.

Anyway, aside off. My proposal was to go higher functionality, higher speed, lots of local memory, but still low transistor CPU's using new schemes of parallel computing and technologies I have come up with. So, 10-1000 times increase in processing density on today's circuits, if you take away the memory space out of the metric. The memory space is to keep the chip cooler and stop it overheating, allowing more cores to run at higher speeds. The local low energy memory also allows a performance speedup from on chip fetches. But I don't have the resources for that.

So, a lot of useful things I could do with a little bit of programmable circuites to deliver a lot higher performance on a cheaper arm chip that otherwise couldn't, versus a bigger more power hungry chip with larger GPU. One of the things, is to do things that benefit from this size of fpga. But yeah, GPU use is easier.
 
Oh yes, I share a similar opinion about C/C##.
 
Ok. Just glanced. The zii architecture was an old one which I was interested when announced years back. I thought that had stalked, disappeared? I am interested in the riscv chip being designed by what's his name, for his next mass processor chip. That's funny named crowdfunded chip. Only as an easy to use more efficient open alternative to the norm. But it might take years to get that level of complexity we have on tap in phone chips for ease of use. They really getting there, but after these years, it's still not as great as I hoped.
 
Eric, none of those pages are about FPGA tech.
 
Which is not the point of the thread.
 
Wayne Morellini said:
<Snip>
So, what type of fpga programmability options and technologies exist now for low energy or high speed?

Thanks.
Click to expand...
^^^
I think you answered your own question there...
 
Hmm, so you don't know the technology.

Anyway, got to go. Watching American high culture. Some program called Hillbilly Squid after some program called Robot Chicken. You guys are so smart over there. Not like that backwards looking Ben Elton's Upstart Crow.
 
Why did Misha Engle immediately point you to OpenCl?

FPGA seem quite seductive to begin with and then when you look at the deeper implementation details they seem like a total pain in the arse... And there is a LOT of simulation and test software out there to help wrestle with them ...

Normally IMO FPGAs are SLOW and EXPENESIVE (for certain number crunching applications) compared to GPUs... not always so great for high bandwidth applications whereas things like CUDA and OpenCL using C and wrappered C++ and full on C++ in your normal programs (or C# if you must) to talk to high bandwidth massively parallel processing architectures based on GPUs is pretty much the fastest game in town (for many applications) and now there are SOCs (system on a chip)... And other super computing architectures on low power devices.

I wouldn't be surprised that in some cases a simulation of FPGA would run faster on a NVIDIA card than the real hardware... but for hard core (first stage) signal processing (outside of number crunching), that's a little out of my wheelhouse... I understand more basic digital cameras so from that point of view first stage signal processing needs to be handled by specific hardware and YES ASICs are very expensive to build and FPGAs provide some better options.

So as you probably know FPGA are not normally super- low power... Depends on the implementation (I have seen very power hungry FPGA / boards and also seem other implementations that are low power), fine for an embedded system on a aircraft/ military platform like a tank etc... not so much for something in your pocket; like a cell phone..?


Unless there is something wildly new that I don't know about FPGA's just seem very slow and difficult and clunky for floating point operations and low level when there are so many other beautiful an elegent newer solutions that are easier to implement nowadays... But maybe good for signal processing first stage (can be made less complex).

I'll google about to see if there is some amazing breakthrough about FPGAs that I DON"T know about :-) ...

Always good to learn something new.

If for some reason I have to learn new voodoo for FPGA and have strong reason to do so then I will... But these GPUs and more generalized memory "cells" like Stem Cell seem to be a better way to go for a lot of applications.


@ Wayne I know you know the basic arguments about FPGA versus ASIC (s) too... Pros and Cons...
_________________________________________________________________________________________________


Side note :

@Wayne as a result of mentioning Ben Elton and "Crow" I did watch a few episodes... Didn't know of their existence and the Third Black adder series (During the Prince Regent times) has Hugh Laurie (the guy that also played house) playing the Prince Regent... I think you mentioned that you like the second Black Adder... That one has Miranda Richardson playing "Queeny"/ Elizabeth the 1st.
 
Wayne Morellini said:
Hmm, so you don't know the technology.

Anyway, got to go. Watching American high culture. Some program called Hillbilly Squid after some program called Robot Chicken. You guys are so smart over there. Not like that backwards looking Ben Elton's Upstart Crow.
Click to expand...

The Robot chicken StarWars episodes are master pieces in their own right :-) Hillbilly squid lol!

I knew one of the writers from Blackadder , more minor writer... Collin Bostock Smith, but the Black-adders always had an amazing cast. Myriam Margoles is an institution in her won right as well as Stephen Fry... (and many others that went on to do amazing things (not least of course Rowan Atkinson)...

Didn't know that Ben Elton moved to Australia.
 
@Wayne maybe this is more up your alley

http://optomotive.com/

"High performance" FPGA + camera, small units...

_____________________________________________________

@Wayne maybe you like FPGAs for prototyping?

_____________________________________________________

http://mil-embedded.com/articles/fpga-gpu-evolution-continues/

FPGA vs GPU (military / defense perspective)... ^^^ Short summary.

IMO GPU type architectures / unified memory structures and things like SOC and Stem cell type approach IS The future...

Knowing FPGA is a useful skill for digital electronics and signal processing engineering development ... BUT...

The article is a really good summary so I'll just quote the whole thing!
 
Article by Charlotte Adams Abaco Systems* [Emphasis added].


FPGA or GPU? - The evolution continues

"Designers of high performance embedded computing (HPEC) systems for the military and aerospace market have some options when choosing the primary processor for signal- and image-processing applications. Designers can cast field programmable gate arrays (FPGAs) or graphics processing units (GPUs) in the starring role. In the past the military was wedded to FPGAs mostly because there was no middle ground between FPGAs and cost-prohibitive application-specific integrated circuits (ASICs). Program managers thought nothing of building a complete electronic warfare (EW) system with FPGAs.

Two developments, however, are changing this picture: First, GPUs have emerged that are nearing parity with FPGAs in both performance and power consumption. Second, the military itself has changed, with budgetary necessity driving officials to demand size, weight, and power (SWaP) tradeoffs. As a result, GPUs are becoming more popular and may eventually overshadow FPGAs, as the latter alternative takes on a more subordinate role.

FPGAs vs. GPUs

FPGAs have certain advantages. To begin with, these chips are hardware implementations of algorithms, and hardware is always faster than software. FPGAs are also more deterministic; their latencies are still an order of magnitude less than that of GPUs – hundreds of nanoseconds vs. single-digit microseconds. (GPU users compensate by accommodating the worst-possible timing case in their particular applications.)

GPUs historically have been power hogs, which is problematic in battery-dependent scenarios, but the latest GPU products have reduced that liability. NVIDIA’s Tegra K1 GPU/CPU board, for example, burns less than 10 W. GE Intelligent Platforms, taking notice of this improvement, has announced an agreement with NVIDIA to add Tegra K1-based products to its stable of GPU offerings (see Figure 1).


Figure 1: GE Intelligent Platforms is NVIDIA’s preferred provider of products based on the new Tegra K1 to serve users in the military/aerospace market.
(Click graphic to zoom by 1.9x)

Unlike FPGAs, GPUs run software, and executing an algorithm in software takes time. Instructions have to be fetched and cued up, math operations have to be performed, and results have to be sent to memory. GPUs also have their own advantages. On the hardware side, GPUs’ massively parallel construction enables them to run a software algorithm much faster than a conventional processor could. GPUs also run their software very close to the hardware, enhancing speed and controllability.

Unlike FPGAs, GPUs excel in floating-point operations. GPU cores are native hardware floating-point processors. A 384-core GPU can run 384 floating-point math operations every clock cycle. This capacity makes GPUs a natural fit for floating-point-intensive signal- and image-processing applications. In fact, many newer signal-processing algorithms are aimed at GPUs. Moreover, GPUs are designed with very fast memory, and new direct memory access (DMA) techniques allow high-volume sensor data to be streamed to the GPU without consuming GPU clock cycles.


GPUs also offer good backward compatibility. If an algorithm changes, the new software can run on older chips. FPGAs are more problematic on this count: It’s no small matter to upgrade the algorithm on an FPGA or to move an algorithm to a newer FPGA. GPUs, furthermore, are supported with a wide array of open development tools and free math function libraries.


GPUs are increasingly found in radar processing, for example, where flexibility is valuable. Radar has numerous modes, some of which pilots want to run simultaneously. GPUs are right for this application, as they can run multiple processing pipelines at the same time. While FPGA manufacturers offer the ability to synthesize a small number of algorithm “images” on the same chip, the algorithms can’t be run simultaneously. It takes a second or so – an eternity in EW – to switch between them.

Is collaboration the key?

The long-term trend in embedded military-signal and image-processing applications seems to be the adoption of the GPU as the primary processing engine, with the FPGA in a supporting role as the data pipe between the antenna and the GPU. Central processing units (CPUs) would play a management role, interpreting the results of the GPU and sending the “answer” to the user.

Such a combined system would play to the strengths of each type of processor while maximizing system efficiency. The FPGA would forward incoming sensor data at high speeds, while the GPU would handle the heavy algorithmic work. Then the CPU would step in to winnow out false positives from the GPU’s output. Since the FPGA would have fewer responsibilities, it could be smaller and less difficult to design and therefore cheaper and faster to field.".

defense.ge-ip.com

This article was published on September 16th, 2014.

____________________________________________________________________


Things have moved on considerably on the GPU/SOC front since 2014...

____________________________________________________________________

* no affiliation.
 
I've stopped watching high brow American comedy. Your philosopher "Homer", was right about Frazer.

So, you do know the classics. The Upstart Crow is in the same flavour as Black Adder the second, except for Black Adder if course (but why are they so racist against accountants) but a bit more subtle and refined with age. Even "Bob" turned up.

Anyway, Upstart Crow, it almost seems like the Blackadder cast covering lines and characters, very good. Not like imitating. I think I'm almost in love with his land Lord's daughter already. Good actress. I really hope an old Black Adder and Baldric and the lanky one, Percy was it, turns up. They are broaderly just before this time period Ifhink. But unfortunately Queenie, would be out of place, with Queen Elizabeth's relation to these matters. Unfortunately, they keep kli g black adders off at the end of the series. But if they can save the team.from Red dwarf from the seemingly impossible to restart the series then I guess they can do it here (that was humour).

Mr Ben was over here. Even had his own show etc, but didn't work out. Don't know when.be went back. Comedy over here had dins through the classics like Paul Hogan and Norman Gunston. I don't think they were ready for drama comedy (or whatever that series was about) or the low-key intellectual comedy routine. About the closest they got to that was Michael Parkinson and Andrew Denton, like his Guns the musical, with let McFeast who went into.heriwn show (A Lawyer and comedian). Most of the comedy was American and English. The goodies, Are you being served, Faulty towers, all the witty ones. Brilliant stuff. I guess he moved back. We get guys come over here and marry Australian girls, like John Denver, who died here from memory, in some sort of ultralight crash, and Sasha Coen? married a Sydney girl? Incidentally there is a guy called by that name in the open source Axiom project, who has an address in Sydney which I had copied down, as Facebook gives a map which they might have got from his location on the phone. If they ever do get through their camera and hold an event, hopefully I get to meet him (he denies he is the actor by the way, but wouldn't it be funny). But since his married that women, his comedy has improved. Wild practicality us one of the old hallmarks of previous Australian culture that people in some of those areas hold. I remember, I was filming train stations through one of those areas, and one of the conductors at one of the stations warned me it wasn't permitted (after the towers). He wasn't going to have me arrested, and no other station had told me. But some robbers had bombed or burnt a car they were using at the station the previous day. Anyway, I complied straight off. It used to sort of flow like that here. The minor lawd were enforced practically, if needed. More lsid back. I enjoyed Sasha's Dictator film. Not do much, Borat, though I did get the idea to buy those red swimming tights and stick on the back of my friends medical delivery van with a sign that said hink for Borat (but I'm too nice and didn't do it).
 
Eric, I covered the issues with FPGA's in the original posts. I'm also a long-term advocate of GPU use instead of FPGA'S in camera development. Indeed, when they wanted to do FPGA, I said it was too complex. I think it took them a millionaire to finish it, Here. It's useful, as they then had the money to covert to ASIC's and the Redrocket experience. They had a huge challenge to edit originally, and people were negative on the usefulness of GPU's back then due to issues I thought could be solved with prudent programming but I deferred to their seeming superior knowledge (which I now think we're a bit bunkim). I advocated a coprocessing solution I had heard of (like you, I used to note energy efficient high performance processing solutions, of which there were around three on market with excellent properties). Redrocket seemed to be from that.
 
 
You must log in or register to reply here.
Back
Top