I don't know if they realize this, but imo all they cause it having less people working with FPGA's. As a result it's not a tool people will reach to when it could be used.
However this has just changed. There is an excellent project at http://www.clifford.at/icestorm/ that has reversed the format for a series of the Lattice FPGA and provide an open source end-to-end solution... And AFAIK they haven't been told off for doing so.
I therefore have one of their USB stick evaluation board on my desk to have a play with as soon as I can. They also have 'hacker friendly' packages.
The most useful tools in fpga dev have little to do with the part of translating the Verilog/Vhdl into something the fpga understands.
Just going to add this:
I had to install VMware (on OS X) and load an Ubuntu ISO just to play with Xilinx's tooling. I dev on win, Ubuntu, and OS X. Primarily Ubuntu. But just a heads up to anyone reading this that can positively affect this: if even I am on a Mac, it's time to expand your offering.
I don't think that having a tool that can't produce working binaries(bitstream or native code) would be that useful. Can you imagine a GCC that can't produce native code but instead have to generate COBOL sources(with almost random feature set) to be compiled by vendors tools?
>[...]Intel about their microcode[...]
Intel have a widely known instruction set that can be used with their cpus, there are many compilers, interpreters(and more) that generate and use it without problems, so I don't see how this analogy is relevant
>There is so much more to the process of creating something useful with an fpga.
You need tools and potential creators of such tools need access to bitstream format used by fpgas to create useful tools
>The most useful tools in fpga dev have little to do with the part of translating the Verilog/Vhdl into something the fpga understands.
But without the ability to translate your code(any code not just vhdl/some variant of verilog) into bitstream, where you would run* it on? It's hard to consider tool useful for fpga development when it can't produce something that can be run* on fpga...
*oh come one, don't be pedantic
The tools available for that task are compact and I'd imagine quite portable based on the OSes they are already supported on. Huge vendor downloads aside, I don't think the problem in the ecosystem is with those tools not being OSS.
Intel's instruction set (ISA if you will, for those reading), is at a higher level than microcode. Microcode is specific to an architecture but not the instruction set. No programmer touches microcode and as far as I'm aware, it's secret and proprietary.
The useful tools to be created that would be open are not related to bitstream formats. (This is actually an assumption that I cannot fully verify at this time.)
I'm not not sure how to address the last point. I'm not trying to be pedantic at all. I'll come back to this as I wrap my head around it.
Edit: I'm continuing to make small edits to grammar and wording.
Also there are many usefull tools that could be created if bitstream format was known. Some of with could be usefull for other devs, not only fpga devs. And there are so many things you could potentially do with your bitstream, like generating it on the fly.
While floss tools would be nice,(gcc or llvm compiling a function to be accelarated by fpga, and then proper bitsream be genarated on startup by fpga driver) but just having alternatives from vendor suplied toolchain would be good.
Those tools generally are more "barely tolerable necessities" than "useful". Also, I think we're overdue some more languages for hardware development; both Verilog and VHDL are about 30 years old.
All this stuff, including the libraries, can be pretty compact. There is no justification whatsoever for the ISE and Vivado bloat. The essential tools should be very compact. There are multiple copies of a microblaze toolchain, for example. WTF? I need none exactly. Zero. Nil. Stop bundling all the crap, please! Put it in a separate package so everybody could happily ignore it.
That they were trying to include or support most of their devices or use cases in one distribution was my guess as to the reason for the bloat. Thanks for confirming it. I'll add that they'd do even better writing it in a LISP or ML-like language with macro's plus efficient compiler. The code would be readable, robust, efficient, and take no more space than necessary.
Cadence is using Lisp extensively, and their distributions are also not very small and are hard to maintain.
I just think they could be several times smaller if they only included necessary functionality and added extra's repo-style depending on one's device, use-case, and so on. Btw, is Lisp used for extensions over there or the whole application? And which LISP?
https://en.wikipedia.org/wiki/Cadence_SKILL
And, yes, it's about a time for the monolithic distros to die. People are already spoiled by package managers, there is no justification for 10gb downloads any more.