[hpsdr] "FPGA Floor Planning"

[Scott Traurig]

Admittedly Barry's assertion was somewhat strong, but that does not make it
wholly incorrect.

As I understand it, both Phil and Joe are self-taught firmware developers
as an avocation, not a profession (please correct me if I am wrong). Not
that this makes them any less professional in fact, indeed they are both
brilliant individuals, but it also does not make anyone else less
professional than they are, either.

People may not realize that there are at least a few, and probably more
than a few, highly experienced and highly professional engineers on this
list. People who have lead many large teams of similarly professional
engineers on complex and difficult projects. People who have seen the exact
problems, many times, currently challenging the design of HPSDR firmware,
and who have seen and understood the solutions. Even if they were not the
firmware developers themselves, they were nevertheless responsible for the
entire job and therefore paid attention to tools, techniques, skills, etc.
In that way these people have been exposed to many facets of FPGA
development that perhaps Phil and Joe have not.

So please don't be so quick to discount the opinions of others.

With respect to the issues associated with developing in Quartus Prime
Standard vs. Quartus Prime Lite: the code should be the same. The compiler
directives and such certainly won't be the same. This is little different
than developing software in two different versions of Visual Studio.
Certainly less involved than porting from Windows to Linux. There is one
feature of Standard that might be worth avoiding, which is High Level
Synthesis (HLS). HLS takes C code and compiles it to VHDL or Verilog. That
level of abstraction might make it difficult for others to interpret any
synthesized code. Although it is a very powerful feature and not to be
completely discounted.

Another consideration is that the code may become literally too complex and
capable to reliably and easily manage using Quartus Prime Lite. I submit
that this has already occurred. If that is truly the case, then a point has
been reached whereby it may be impossible to resist moving to a more
capable development environment, at least if forward progress is going to
continue at it's traditional, rapid pace.

In the meantime, firmware development is clearly suffering from challenges
more easily addressed using a more capable development environment. Note
that I did not write "more properly addressed", although I believe that to
be true as well, as expressed in the above paragraph.

There are a number of concerns:

1. People will not be able to develop in an open source fashion with
Quartus Prime Lite. While I may not have fully debunked that idea, and
leaving aside the code complexity issue, I will make it a point to double
check my assertions from a purely mechanical perspective with some
"professional" ;-) firmware engineers this week and report back on my
findings.
2. Developers have to learn a new toolset and new techniques. It appears
that Joe has made it clear that he has no desire to learn a new toolset.
I'm not sure about Phil.
3. Someone has to buy the tools. This is no different than a loaner radio.
If Apache loaned a license it would be the same value as loaning an 8000DLE.

These concerns do not seem insurmountable. There are also many parallels in
the hardware sense. Very, very few people hand solder their boards anymore,
much less send out for their own board fabrication. Many boards are not
available as open source, nor even available for hand assembly. As the
complexity and capability of the HPSDR architecture continues to increase,
there may be some things that are better served by diverging from the path
of pure open source.

73,

Scott/w-u-2-o
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/attachments/20170212/f5d1793e/attachment.htm>