Confused! CLD Failed the second time

[crossrulz]

Disclaimer: I do not work for NI and this knowledge is based on a conversation I had with one of the NI Certification heads ~4 years ago.

Most of the CLD grading is automatic now and each requirement has a different weight.  So if you didn't hit one of the main requirements, that would hurt you more than if you missed one of the minor requirements.  And there is also the cascading effect: requirement B is tested with results from requirement A, but requirement A was not done properly (failed) and so requirement B will therefore also fail.  Manual reviews are done, but pretty much only when the automatic grading results in a close to passing grade.

[Billy_Firestone]

Thanks for your insight.

That makes sense to me. For example, if the initialized data were not read properly from the file provided, you are very likely to lose most of functionality points since it is usually at the very beginning of the program.

But for the exam I just took, I tested all functionality over and over to make sure it works. I still don't understand why I failed and this is what I am frustrated about. 

[Fisel]

Hi all

We have received Billy's request for feedback, and I'm preparing that for him. Because of privacy issues I won't be posting that in this forum, but Billy will be able to forward basic information consistent with the NDA (keep it at the category/general LabVIEW programming level, no details about the given exam scenario).

In the meantime, a couple of details may be helpful for this audience.

• For Piotr and others who do not believe their score reflects the code they turned in: If you believe you coded correctly functionality for which you were not credited, please do appeal by emailing certification@ni.com . We can't regrade based on simply "I think I should score higher." But if your exam has demonstrated functionality that we didn't see, do let us know. 
• The functionality grading is not linear.
  
  • As Greg pointed out, some functionality, such as reading and using the IO data and making selections, can be "gateway" functionalities. If your system can't read the data, or you can't choose which subset of data to use for the specified functionality, your score will be very low for any functionality that depends on these gateway functionalities.
  • More complex functionality categories have a larger allotment of points than simpler categories.
• The phrase " ... does not demonstrate full specified functionality". Key words there are
  • "demonstrate": We only verify functionality that is demonstrated on the front panel of the main VI. This is to help standardize grading over a set of graders whose ability and interest in troubleshooting may vary. Knowing this ahead of time helps the test-taker prioritize their coding, which is a significant skill for a developer.
  • "full": Each category is divided into specific sub-categories, and the points are pro-rated over that category. So if, for example, a category called "RedLight" requires a timer to count down a given amount, then turn on a red LED and display a string message,  might be split into three categories. An exam that counted down correctly and turned on the red LED but did not display the string message would receive 2/3 of the points allotted for the RedLight functionality, and the feedback sheet would contain a message "RedLight did not demonstrate full specified functionality."
• On the automated grading, we hand-verify every failing grade. The automated grading is more strict, so we don't want to fail without checking. For example, occasionally we get exams that reverse the functionality of the Power button. As you can imagine, an automated system grades this as 0 functionality. But when we see this, we dock for the functionality of the Power button, then grade the remaining functionality without further penalty. So if you pass but scored a lot lower than expected, that might be the automated grading system, but a failed exam has been hand-graded.
• I'd refer everyone with questions about grading (pre- or post-exam) to our CLD exam preparation webcast.

Hope this clears up some confusion.

Best Regards,

[Billy_Firestone]

Hi,

Thanks for the message.

I would be better prepared if I know what I did wrong for the exam.

It is kind of a learning curve that will benefit my career. That is my purpose of taking the CLD exam, which is to evaluate how good I am in programming.

[piterx]

Hi,

is anyone entitled to get more detailed feedback? Although I don't want to dispute the results, it would be really helpful if I can see the feedback in more detail.

Best wishes,

Piotr Golacki

[Fisel]

I'm sorry, Piotr, but due to the volume of grading we can't provide more specific feedback for all exams. The appeal process is in place to ensure that the exam evaluation is consistent with our grading guidelines and policies, to make sure we didn't miss crediting important functionality.

Best Regards,

Elwood

[Billy_Firestone]

Hi Piotr,

Per discussion with the NI certification team, one portion of my code somehow locked out the front panel which prevented the program from further executing. But after the exam re-evaluation, I am able to pass the exam.

My suggestion is that always be aware of the response time of your program. Using a counter may be a good way to avoid this issue.

In addition, be careful when you use the event structure in a state machine, it may freeze your front panel if you do not manage it correctly.

[piterx]

Hi Billy,

Thanks for your response. I need to admit that when I recreated my exam at home I realized that I had a bug in my code which would lockout my code. I used Queued Message Handler and I actually en-queued idle state within the idle state which was unnecessary and accumulated in the queue causing it to gradually slow down the response and eventually making it unresponsive.

Best wishes,

Piotr Golacki

[Billy_Firestone]

Hi Poitr,

I normally use a state machine to write the code.

As for producer consumer loop and QMH, you have to make sure that the name of sent/received message matches with the corresponding one in the case structure.

Just use the structure you feel most comfortable for the exam.

[crossrulz]

---

@piterx wrote:

I used Queued Message Handler and I actually en-queued idle state within the idle state which was unnecessary and accumulated in the queue causing it to gradually slow down the response and eventually making it unresponsive.

---

This has been discussed many times at various places on the forum (too lazy to search at the moment).  Basically, a QMH should almost never enqueue onto itself.  As you have discovered, it can cause really weird behavior and/or horrible race conditions.  If you have a repeated task, use the queue/event timeout.