what is /g/'s opinion of field programmable gate arrays?
>>54049835
No one here knows what those are. But I'm sure they could advise you on some fine Linux distros.
>>54049835
They are cool, I worked with them. Too cool for /g/ tier neckbeards though.
>>54049835
they are the future. bitcoin mining has brought down their cost considerably
>>54049835
Pretty nice concept, good for prototyping.
Only real thing I hated about it is that singe you're programming at an hardware level, two thing that are logically the same would work only if written in one specific way.
I used it ages ago so I don't remember it in detail, but it pissed me off.
>>54049835
Had to use a DE2 by Altera in one class. Had to use multiplexers and binary adders and had to display the result of various sums depending on which button is pushed
I failed.
can't wait to implement x86 on one of those things
>>54050912
is this realistically possible?
>>54049835
good for exactly 2 things:
> prototypes or otherwise just fucking around
> very low volume shipping products
they run much hotter and slower than a custom ASIC for like a thousands of the tool-up cost.
also, a lot of the things they were traditionally used for have been eroded away by commodity ASICs with huge amounts of multipliers/adders that don't need cycle-perfect deterministic throughput.
>>54050157
idiot
>>54050183
another one
they are just shit and only fun for the first 5 minutes
after that you can only use them to cut your wrists because you realize what a fedora faggot hipster you are
>>54049835
Never had the chance to study or work with one.
Going to in the next 2-3 years so we will see.
>>54049835
I'm in my first few weeks of college and we're working with them.
I was pretty surprised to find out we're using them and not some microcontrollers or SPS. Pretty cool, but also a lot of work, I think.
>>54050912
>implement x86
>>54050924
>is this realistically possible?
of course. 80386 only had ~275k transistors and the instruction set wasn't -that- ridiculous.
it would still be a pain in the ass and probably not worth anyone's time though.
>>54050924
theoretically yes
realistically maybe. i guess it would be possible with older x86 that aren't as complex as the new ones.
anyways, it would be useless.
>>54051243
>a lot of work, I think
There's no "I think", it IS a lot of work
I used Verilog for two semesters to program FPGAs, and these guys are right
>>54051198
>>54050694
I wasn't even shit at it, got A's in both courses
But god fucking damn did I hate it
>>54049835
What you can do with it is pretty cool, but fuck do I fucking hate working with VHDL or anything that Xilinx produces.
Plus they have shit support for Linux, which makes me very sad.
They are really fun. Currently working on making some cool VGA things. Too bad the tools for development are proprietary and can weight 6-10 gb. Picrelated was my first project. I got it up and running within 8 hours of finding an old FPGA and starting the software download.
>>54053456
This ain't true. Xilinx ISE is a breeze to get working on linux. Just check out the archwiki info and you-ll see that its no more difficult than double clicking an exe
Love the concept, I just really fucking hate HDL. I cannot stand Verilog. And I have no reasoning why; I just irrationally hate it.
>>54050157
>there is no technology other than electronics
KYS
>>54053916
54053916 := (others=> 'Ur Dumb');
>>54049835
Gennerally unnecessary for most things, and a ASIC would suffice to most end users, but I really like them and it gives a feeling of "wow, this is cool" when I find out something has a FPGA in it.
Mostly good for prototyping and tweaking things.
They're too damn expensive.
>>54050825
352 wasn't that hard, pussy. The exams sucked though
>>54049835
isn't that just a fancy name for eeprom? or does it work differently?
Made an image decoder on one. Shit is hard as balls to learn but really cool. Would highly recommend.
>>54054151
No it's an array of logic cells.
>>54054151
uuuhhh... Very different... FPGAs allow you to "program" arrays of logic cells to perform operations on signals. Essentially, the HDL you use for an FPGA could ideally be used to assemble a physical digital circuit to perform the same logic operations.
>>54053910
Xilinx ISE isn't the problem.
Getting drivers to program the damn things is.
>>54053927
holy shit is all you do quote posts and write "kys" under some shitty greentext?
im not good enough to use things like that (yet(i hope))
I really wanted to get into those things, but poorfag plus I just begun getting into electronics and discrete logic and have not much time to learn. Support- and documentation-wise, would you advise to go with xilinx or altera?
>>54053893
Cool!
>>54054585
Oh god the cable drivers
>>54055042
Yes, the cable drivers.
Shit literally gave me a message about how I had to use a very specific version of red hat or else it wouldn't work. This was for a college project and installing that specific version of red hat just for some cable drivers wasn't exactly an option.
Mind sharing your experiences with it?
Cool as shit for drafting hardware, and great learning tools. Here's a 16-bit RISC CPU I made for a class project. It's the thing that connected the dots between transistor properties and computer science for my brain.
>>54053893
ugh Xilinx ISE is 20GB, what a shithole that software is.
So impressive that it starts up in less than a second though!
>>54054791
Altera's website is a goldmine of tutorials
The future
http://www.theregister.co.uk/2016/03/14/intel_xeon_fpga/
>>54055427
we didn't have time to fuck around I just somehow managed to make it work on Windows (stupid 32bit usb driver libs where even more retarded) then just transfered over the Bitstream from the actual workstation. That was at school and I don't really need it at work because fpga farms but I might soon get the play with it again. I found some guys that make some neat pcbs to play around with, looking for link...
>>54057827
http://www.ztex.de/usb-fpga-2/usb-fpga-2.01.e.html
so these things are just chips you can program yourself rather than having it specified at manufacture?
>>54057986
Yes, you can create digital hardware at your desk
I've spent month of free time on a FPGA project. It's trap next to trap. Pay attention to time analysis reports.
>>54058023
that's pretty cool
not got any use for it myself
wouldn't they need some kind of board?
I just stick to stm32's for now
>>54053910
Also, Altera's (Quartis) Linux support is superb. You can automate pretty much everything via command line tools. It also uses Tcl for scripting and description of almost everything.
>>54058052
Note that any pitfalls are associated with hardware design in general, not specific to FPGAs
Remember to unit test with a simulator like Modelsim and pay attention to the compiled RTL design to see if it is what you expected from your HDL
>>54058065
FPGA boards essentially emulate real digital hardware using a bunch of logic elements as shown in >>54054170
>>54053963
FPGAs are better than ASICs in that you can update your circuits in production to fix issues or add features.
>>54058137
In the same way play-dough is better than steel
>>54058107
I lost the possiblity of using Modelsim by having part of my circuit in Qsys. It spews out both Verilog and VHDL and free version of ModelSim doesn't allow mixed languages. Also, the design is already so complicated (avalon st/mm, dc fifo) I wonder If I would be able to debug it in it.
>>54058193
you can still test individually the parts that you wrote yourself, correct? Just check the inputs/outputs and see if those correspond to what they should be to communicate with the rest of your SOPC
>>54049835
1) Gets fucked over by CPU in terms of development time and maintainability. A FPGA bitstream only runs on a specific FPGA but an software program runs on every CPU. If a calculation is not simple and embarrassingly parallel the mult-core CPU of today is always faster.
2) Gets fucked over by GPU in development time and maintainability. A FPGA bitstream only runs on a specific FPGA but GPGPU program usually work on cards of the same architecture/generation without tweaking. Graphics cards are faster.
3) Gets fucked over by ASIC in terms of speed and power consumption. Get fucked over in the price department if mass production.
What is left for the FPGA? Prototyping of a small number of units of real-time signal processing systems. Because ASIC are more costly in small units and general purpose computers have high latencies.
>>54049835
OP what the fuck are you even asking? What opinion is there to have on FPGA's?
You probably think they serve the same purpose as an Arduino or a Raspberry Pi.
>>54058320
>but an software program runs on every CPU
Stopped reading there. This is simply not true.
>>54058421
You know what I mean. Stop being autistic.
>>54058435
You're the one being autistic by drafting up a lengthy comparison of apples to oranges.
>>54058456
You still haven't told me why it isn't true.
>>54058485
(I'm someone else)
> A FPGA bitstream only runs on a specific FPGA but an software program runs on every CPU.
Nonsense.
Binary image (bitstream) compiled for specific board with a CPU will only work with that board with that CPU. You can switch CPU for compatible one in the same way you can swap FPGA. You'll have to recompile your sources in both cases.
