Linting tool setup

[RCIngham]

Greetings all,

Further to previous thread(s), has anyone here experience in setting up a
linting tool such as Spyglass or LEDA? How long did it take?

My thoughts are that a suite of test-case files is needed to test for
detection of each code breach, and that preparing that would take a week or
2.

Thanks in advance,
Robert




---------------------------------------
Posted through http://www.FPGARelated.com

[jt_eaton]

Setup can be very simple. Most tools use the same command line syntax as
verilog so all you have to do is configure your code for synthesizable
only
and remove all the bus functional models from your testbench.

You only need a suite if you have multiple configurations to test. One
test
checks for all violations in the same pass.


John Eaton

[RCIngham]

>>Greetings all,
>>
>>Further to previous thread(s), has anyone here experience in setting up
a
>>linting tool such as Spyglass or LEDA? How long did it take?
>>
>>My thoughts are that a suite of test-case files is needed to test for
>>detection of each code breach, and that preparing that would take a week
>or
>>2.
>>
>>Thanks in advance,
>>Robert
>>

>Setup can be very simple. Most tools use the same command line syntax as
>verilog so all you have to do is configure your code for synthesizable
>only
>and remove all the bus functional models from your testbench.
>
>You only need a suite if you have multiple configurations to test. One
>test
>checks for all violations in the same pass.
>
>
>John Eaton
>

A clarification. My intention is to test the tool(s), to check whether they
find the rule breaches correctly...

[jone...@comcast.net]

How long it will take to test the linting tool depends entirely on the number and complexity of the rules you are trying to enforce.

If you have half a dozen simple rules, a couple of weeks is probably plenty of time. If you have a hundred rules, many of which are rather complex, two months might not be enough time.

What code language are you wanting to lint, VHDL or Verilog?

What kind of policies are you trying to enforce? Hazardous, frequently misused or mistake-prone usage? Maintainability guidelines (e.g. unused declarations)? Local style guide?

Andy

[RCIngham]

VHDL.

All of the above for preference. We have many rules.
Precise number subject to next week's document review.

[HT-Lab]

On 22/05/2013 14:20, RCIngham wrote:
>> How long it will take to test the linting tool depends entirely on the
> number and complexity of the rules you are trying to enforce.
>>
>> If you have half a dozen simple rules, a couple of weeks is probably
> plenty of time. If you have a hundred rules, many of which are rather
> complex, two months might not be enough time.
>>
>> What code language are you wanting to lint, VHDL or Verilog?
>>
>> What kind of policies are you trying to enforce? Hazardous, frequently
> misused or mistake-prone usage? Maintainability guidelines (e.g. unused
> declarations)? Local style guide?
>>
>> Andy
>
> VHDL.
>
> All of the above for preference. We have many rules.
> Precise number subject to next week's document review.
>

Why do you want to write tests for a linting tool, is this a DO-254
requirement?

There is no question that a static/dynamic linting tool will have a
positive impact on your development, it all comes down to your EDA
budget and not on the tools usefulness.

I would get some evaluation licenses and simply run them on all your
previous designs. All linting tools are push-button with a standard set
of rules (DO-254, RMM, Best Design Practise, Xilinx, Altera etc) so it
shouldn't take too long to get some results. I am sure that analysing
these results will highlight some interesting coding issues (and the
usual bunch of false positives). It is a bit like running Code Coverage
for the first time on your design. Then during the project work on
creating/improving your own ruleset.

If you are doing an FPGA design then I would probably go for Mentor's
DesignChecker or Aldec's Alint rather than Spyglass which is mainly an
ASIC tool and hence might be limited in their VHDL support.

Good luck,

Hans
www.ht-lab.com