Benchmarks for Verilog Modules

[kb33]

I have a Verilog design (with several modules) and I need to benchmark
the same. Is there any industry standard for verilog design
benchmarking? A Google search doesn't bring up much authentic stuff.
Else, is it possible to use software bench marks such as SPEC2000,
MiBench and so on? Please advise.

Thanks
Kanchan

 

[Mike Treseler]

kb33 wrote:

I have a Verilog design (with several modules) and I need to benchmark
the same. Is there any industry standard for verilog design
benchmarking?

Since you already have a design, I would
write a simulation testbench for a functional
test then check fit and fmax using using
fpga vendor tools.

Else, is it possible to use software bench marks such as SPEC2000,
MiBench and so on?

No, but I would be more concerned about fitting
my own design than designs by others.

-- Mike Treseler

 

[kb33]

Hi Mike,
This design is not a commercial design (as yet!) but rather, targeted
towards a technical publication, hence the issue of benchmarking. I
have the test benches in place, and I will try your suggestion about
using fit and fmax, but that would not be sufficient to write a
technical paper. In essence, I need to present my design in a way that
even software guys (or for that matter, anyone else) can relate to the
design concept. Therefore I need more suggestions.

Thanks
Kanchan

 

[Mike Treseler]

kb33 wrote:

This design is not a commercial design (as yet!) but rather, targeted
towards a technical publication, hence the issue of benchmarking.

You've lost me.
What is it you are trying to measure?

In essence, I need to present my design in a way that
even software guys (or for that matter, anyone else) can relate to the
design concept. Therefore I need more suggestions.

If I wanted software guys to understand a
verilog module, I would write it as a single
named synchronous block with local reg
declarations.

-- Mike Treseler

 

[John_H]

"kb33" <kanchan.devarakonda@gmail.com> wrote in message
news:1184345833.730995.81900@o61g2000hsh.googlegroups.com...

I have a Verilog design (with several modules) and I need to benchmark
the same. Is there any industry standard for verilog design
benchmarking? A Google search doesn't bring up much authentic stuff.
Else, is it possible to use software bench marks such as SPEC2000,
MiBench and so on? Please advise.

Thanks
Kanchan

If you had a software program, would you use SPEC2000 or MiBench to
benchmark the *software*? The benchmarking I'm familiar with are all for
hardware, not for the software that goes in them. If your Verilog module is
a soft processor, you could use the industry standard benchmarks that are
for processors to test the processor performance. I wouldn't know how to
benchmark a simple process like "drink a beer" but I could benchmark the
time it takes me to get measurably drunk; I don't think there are standards
for either - it's all comparative with the same benchmark for other heavy
drinkers.

 

[Ssa]

This design is not a commercial design (as yet!) but rather, targeted
towards a technical publication, hence the issue of benchmarking.

You've lost me.
What is it you are trying to measure?

At first, I thought he wanted to profile his RTL-code. You know, to
see which parts of his design took the most CPU-time (in simulation.)