RFC: Enhancing VHDL for OO, Randomization, Functional Coverage

[Jim Lewis]

Hi,
The VHDL standards community needs feedback from VHDL users.

Currently the Accellera VHDL TSC is working on enhancements
to add classes/OO, Randomization constructs, and Functional
Coverage with a goal of giving VHDL the same verification
capability as SystemVerilog or E.

One of the VHDL simulation vendors has indicated that they
only want to implement new features if the user community
wants these features.

The questions come down to:
Do you want these features added to VHDL?
Do you want to VHDL to be capable of handling all of your
testbench needs?

This work is work in progress and below is the current status.
Keep in mind too that your interest/support of this work will
help raise the focus and inspiration of those doing the work.

Classes / OO:
Peter Ashenden submitted a class proposal last year and
provided updates to it this year at DVCon. Currently
he plans on finishing an updated draft soon.

Randomization:
I just submitted the first draft of the randomization proposal.

Functional Coverage:
I have started working on this - anyone else who is interested
is welcome to contribute as much as they would like.


With a focused effort, like the one to finish the Accellera 3.0
draft of the standard, I think we can be done with these by
September.

Although some have expressed doubt, it is clear that vendors
will do what their user community asks them to do - otherwise,
someone else will and, as a result, will earn your business.

You can post here, send your reply to me (let me know if I can use
either your name and/or company name when I tally the results
for the Accellera VHDL TSC), or join the Accellera VHDL TSC
(which you can do as a non-Accellera member by registering)
and post your reply there.

Thanks,
Jim Lewis
VHDL Evangelist
SynthWorks VHDL Training

 

[Amal]

Hi,
The VHDL standards community needs feedback from VHDL users.

Currently the Accellera VHDL TSC is working on enhancements
to add classes/OO, Randomization constructs, and Functional
Coverage with a goal of giving VHDL the same verification
capability as SystemVerilog or E.

One of the VHDL simulation vendors has indicated that they
only want to implement new features if the user community
wants these features.

The questions come down to:
Do you want these features added to VHDL?
Do you want to VHDL to be capable of handling all of your
testbench needs?

This work is work in progress and below is the current status.
Keep in mind too that your interest/support of this work will
help raise the focus and inspiration of those doing the work.

Classes / OO:
Peter Ashenden submitted a class proposal last year and
provided updates to it this year at DVCon. Currently
he plans on finishing an updated draft soon.

Randomization:
I just submitted the first draft of the randomization proposal.

Functional Coverage:
I have started working on this - anyone else who is interested
is welcome to contribute as much as they would like.

With a focused effort, like the one to finish the Accellera 3.0
draft of the standard, I think we can be done with these by
September.

Although some have expressed doubt, it is clear that vendors
will do what their user community asks them to do - otherwise,
someone else will and, as a result, will earn your business.

You can post here, send your reply to me (let me know if I can use
either your name and/or company name when I tally the results
for the Accellera VHDL TSC), or join the Accellera VHDL TSC
(which you can do as a non-Accellera member by registering)
and post your reply there.

Thanks,
Jim Lewis
VHDL Evangelist
SynthWorks VHDL Training

I have been a long time for Classes/OO, Randomization and Functional
coverage that is already part of SystemVerilog. VHDL is my favorite
HDL and I think it has always had very high-level constructs ahead of
its time. The VHDL-2006 is a much needed update, but still comes
short of the features that SystemVerilog boasts for verification. I
think Classes/OO is great for synthesis as well and brings the
language to higher abstraction level that is already enjoyed by
software developers.

It is a shame that (big) vendors are pushing SystemVerilog and think
that there is not enough customer base for VHDL. I beg to differ and
have be anxiously waiting for VHDL update and OO, verification
extensions.

I do not want to start language wars, but take a good look at table on
page () of the following article and you would agree that VHDL has
been at higher-level of abstraction than Verilog.
http://www.edn.com/contents/images/376625t1.pdf

Verilog is still good for gate-level descriptions and with Verilog
2001 extensions they brought it up to par with VHDL, and then
SystemVerilog extensions added more features to get market share from
Verisity and E supporters by pushing a supposedly standard driven
language. Although we all know the big vendor influence.

I hope these proposed features and the good works of Peter, Jim and
other drivers of this great language is not taken for granted.

-- Amal

 

[Thomas Stanka]

Hi Jim,

The VHDL standards community needs feedback from VHDL users.

Currently the Accellera VHDL TSC is working on enhancements
to add classes/OO, Randomization constructs, and Functional
Coverage with a goal of giving VHDL the same verification
capability as SystemVerilog or E.

One of the VHDL simulation vendors has indicated that they
only want to implement new features if the user community
wants these features.

And I guess mti won't support these features with Modelsim, to force
Questa sells, like it is with SV. This might be a major issue when it
comes to market usage.

The questions come down to:
Do you want these features added to VHDL?

Don't know if OO-update is a good idea to a language, but we will see.
C++ is not the best idea for OO IMHO because it relays to much on
procedural statements. VHDL has no choice, when it comes to procedural
statements.

But I would realy appreciate, if vdhl would support built-in "modern"
assertions like PSL or e.

Do you want to VHDL to be capable of handling all of your
testbench needs?

Yes I like VHDL to be enhanced for tomorrows tb needs.
I am currently working on a design, with included ADC. I have the
netlist with ADC as blackbox for some purpose, the netlist with
behavioral model for other purpose. In some cases I need to use
another library element for a digital gate, than normal. Netlist and
and one testbench are verilog and it is a pain to select the
corresponding tb-gtl-library settings without configurations(is no
porblem using configurations).

bye Thomas

 

[Ben Jones]

Hi Jim,

Hi,
The VHDL standards community needs feedback from VHDL users.

Currently the Accellera VHDL TSC is working on enhancements
to add classes/OO, Randomization constructs, and Functional
Coverage with a goal of giving VHDL the same verification
capability as SystemVerilog or E.

One of the VHDL simulation vendors has indicated that they
only want to implement new features if the user community
wants these features.

The questions come down to:
Do you want these features added to VHDL?
Do you want to VHDL to be capable of handling all of your
testbench needs?

I have been looking at SystemVerilog recently, although I am a die-hard VHDL
guy at heart.

My wish-list for VHDL, in a prioritized summary, looks something like this.
There are several language features I think I would rate as more important
than the verification-related ones you mention:

(1) Synthesizable fixed/floating-point types for arithmetic datapaths. ISTR
this being pretty much done now, modulo certain tools supporting it (ahem).

(2) Types as generics, a.k.a. templates - so one can write e.g. FIFO logic
which is completely type-agnostic. This flexibility would make code reuse in
VHDL approximately 10 times easier than it is today.

(3) Interfaces. Lots of people are coming around to using record types to
bundle related signals together when interfacing components together. This
is currently a biy kludgey as it requires separate bundles for each
direction. I remember seeing a pretty acceptable proposal for properly
adding this functionality to VHDL.

(4) Object-orientation. I must look at this Ashenden stuff. The ability to
create class hierarchies of VHDL components would be another great boost for
code re-use - particularly in a language where abstraction by
interface/implementation separation is so prominent.

(5) Make std_logic_unsigned, std_logic_signed and std_logic_arith into
reserved words so they are invalid as library names ;-)

(6) Improved assertions, like SystemVerilog. VHDL assertions are pretty
limited and hard to use as they stand.

(7) Binary file I/O that is portable.

(8) Constrained randomization, like SystemVerilog. Although personally I
have found procedural randomization to be perfectly adequate in most
contexts, that's probably just because of the projects I've worked on. I can
see how this could be a very useful feature.

(9) Native functional coverage support, like SystemVerilog. This is
progressively more important as the system being verified gets bigger, but
it's certainly possible to manage without it.

There are a few other things that bug me, which I might as well add while
I'm on a roll:

(10) Fix the operator precedence rules so they make sense.

(11) Extend the mandatory range of integers to 64-bit representation, and
force implementors to support the full range of 2's complement values rather
than just [-2**N-1 .. 2**N-1]. Not being able to treat VHDL's 32-bit
"integer" as being the same as C's 32-bit "int" is a royal pain.

(12) Allow bitwise operations (masks, shifts, rotates, xors etc) on
integers.

(13) Include a pre-processor for conditional compilation.

That's probably enough for now!

Basically I think verification is important, but having a language which
facilitates good design is more important still. I would prefer to be able
to do the verification in the same language that I do the design. And I
would prefer that language to be VHDL!

Cheers,

-Ben-

 

[Jim Lewis]

All,
If you want to continue to have good vendor support for VHDL,
you need to step up and voice your opinion, even if it is just
to say that you want these features in VHDL.

Ben,

My wish-list for VHDL, in a prioritized summary, looks something like this.
There are several language features I think I would rate as more important
than the verification-related ones you mention:

(1) Synthesizable fixed/floating-point types for arithmetic datapaths. ISTR
this being pretty much done now, modulo certain tools supporting it (ahem).

In Accellera VHDL standard 3.0 (standardized July 2006 by Accellera)
available through Accellera.org. Ask your favorite vendors to implement it.

(2) Types as generics, a.k.a. templates - so one can write e.g. FIFO logic
which is completely type-agnostic. This flexibility would make code reuse in
VHDL approximately 10 times easier than it is today.

In Accellera VHDL standard 3.0. Ask your favorite vendors to implement it.

(3) Interfaces. Lots of people are coming around to using record types to
bundle related signals together when interfacing components together. This
is currently a biy kludgey as it requires separate bundles for each
direction. I remember seeing a pretty acceptable proposal for properly
adding this functionality to VHDL.

Me too. I made the initial proposal. I will revisit it after we finish
classes, randomization, and functional coverage - however to do this
we need people to voice their opinion.

(4) Object-orientation. I must look at this Ashenden stuff. The ability to
create class hierarchies of VHDL components would be another great boost for
code re-use - particularly in a language where abstraction by
interface/implementation separation is so prominent.

(5) Make std_logic_unsigned, std_logic_signed and std_logic_arith into
reserved words so they are invalid as library names ;-)

Don't we all wish, but even if it was in the standard,
vendors would still have to provide a way to allow old designs.

(6) Improved assertions, like SystemVerilog. VHDL assertions are pretty
limited and hard to use as they stand.

PSL is integrated into VHDL in Accellera VHDL standard 3.0.
Ask your favorite vendors to implement it.

(7) Binary file I/O that is portable.

(8) Constrained randomization, like SystemVerilog. Although personally I
have found procedural randomization to be perfectly adequate in most
contexts, that's probably just because of the projects I've worked on. I can
see how this could be a very useful feature.

Keep in mind there is a new version due soon.

(9) Native functional coverage support, like SystemVerilog. This is
progressively more important as the system being verified gets bigger, but
it's certainly possible to manage without it.

There are a few other things that bug me, which I might as well add while
I'm on a roll:

(10) Fix the operator precedence rules so they make sense.

(11) Extend the mandatory range of integers to 64-bit representation, and
force implementors to support the full range of 2's complement values rather
than just [-2**N-1 .. 2**N-1]. Not being able to treat VHDL's 32-bit
"integer" as being the same as C's 32-bit "int" is a royal pain.

(12) Allow bitwise operations (masks, shifts, rotates, xors etc) on
integers.

(13) Include a pre-processor for conditional compilation.

That's probably enough for now!

Basically I think verification is important, but having a language which
facilitates good design is more important still. I would prefer to be able
to do the verification in the same language that I do the design. And I
would prefer that language to be VHDL!

Cheers,

-Ben-

 

[Torsten Landschoff]

Hi Jim,

Currently the Accellera VHDL TSC is working on enhancements
to add classes/OO, Randomization constructs, and Functional
Coverage with a goal of giving VHDL the same verification
capability as SystemVerilog or E.

One of the VHDL simulation vendors has indicated that they
only want to implement new features if the user community
wants these features.

I just started with VHDL two months ago and I have to say that I find
it frustrating that I keep running into unimplemented parts of the
VHDL standard. That's with the Xilinx toolchain and I read a lot of
that already that I can confirm. A few days before I tried to model a
ADC wrt. the timing constraints and had to notice that the Xilinx ISE
Simulator (non ModelSim ) does not support the stable attribute.

Okay, enough of the rant, what I wanted to say is that I'd expect the
implementations to conform to the standard.

Classes / OO:
Peter Ashenden submitted a class proposal last year and
provided updates to it this year at DVCon. Currently
he plans on finishing an updated draft soon.

I'd really like to see that proposal. Coming from software development
it really seems logical to apply OO to HDLs.

Randomization:
I just submitted the first draft of the randomization proposal.

Also very interesting, we will need pseudo random data for our project
as well. Some kind of LFSR will be good enough but having it directly
supported would be nice, also for synthesis. I can dream, right

You can post here, send your reply to me (let me know if I can use
either your name and/or company name when I tally the results
for the Accellera VHDL TSC), or join the Accellera VHDL TSC
(which you can do as a non-Accellera member by registering)
and post your reply there.

While I am just starting with VHDL, I'd like to contribute. In fact, I
became member of IEEE-SA just for that. I figured following that
discussion would also be the best way to learn VHDL, just as my
involvement with Debian GNU/Linux as a developer was a real boost for
my Linux know how. However, I don't think my company is interested in
spending money and developer time on it so I could only use my spare
time for it.

Greetings, Torsten

 

[Torsten Landschoff]

Hi Jim,

I just noticed that I did not really answer your questions, therefore
I'll try again

The questions come down to:
Do you want these features added to VHDL?
Do you want to VHDL to be capable of handling all of your
testbench needs?
[...]

Classes / OO:
Peter Ashenden submitted a class proposal last year and
provided updates to it this year at DVCon. Currently
he plans on finishing an updated draft soon.

Yes, classes/OO would be very welcome.

Randomization:
I just submitted the first draft of the randomization proposal.

Same for randomization.

Functional Coverage:
I have started working on this - anyone else who is interested
is welcome to contribute as much as they would like.

Everybody should want coverage tests as it is important for
correctness. I am not sure though how this will require language
changes?

And of course I'd like to use only VHDL for our test benches. Tcl and
VHDL kind of works but is not as portable and standardized.

Greetings, Torsten

PS: I am working for nAmbition GmbH but any opinions given above are
my own. Feel free to use my name.