Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
Beginner
382 Views

Matching timing constraints regularly fails when Full Moon

For Quartus 17.1 and 18.1 (and its Design Space Explorer) and our design using Cyclone V, timing constraints are never met when full moon time. It sounds crazy, but we regularly observe that behavior. As the fitting process simulates annealing process when liquid metal cools down, the only explanation can be that the influence of moon gravity was (not) taken into account. Have anybody else experienced that or what is the explanation ?

Please, note that we usually have to wait for up to 3 days before full moon is gone and we match the constraints (not modifying anything).

0 Kudos
9 Replies
Highlighted
New Contributor I
10 Views

Hi, from your crazy scene, I cannot observe same behavior but strange thing happen also at my side.

I am working on how to isolate and avoid they can plague me...

I observed strange mangling of files, clipboard content appear somewhere.

Loss of files lines.

Loss of file file link as if they where stored in m$ mode but I am using Linux.

Non deterministic Behavior as VHDL has to be.

Sometimes loses lot of setting...

Refuse programming .pof files and old 15.0 programmed without problem...

Odd Havoc, BUT not moon influence ;) I am missing this great feature.

 

I am not currently working on Cyclone, I don't have this moment board at hand, tomorrow I try some core to that board to see what happen.

Please add more details about what happen, OS version.

Did you try'd clean project?

0 Kudos
Highlighted
Beginner
10 Views

Yes, I regularly clean project files and I even remove rubbish left somewhere after removal. Normally I run Quartus with seed set e.g. to 1, then use Design Space Explorer to try 40 or 60 seeds. Meeting timing constraints never works when full moon time. Our design takes about 75% of logic resources, I use Windows 7. We have 2 iterations of the design, one in Cyclone III using Quartus 12.1, the other in Cyclone V using Quartus 17.1 or 18.1. Our design takes time stamp from a server, so a design varies every day as far as date (in separate vhdl file) concerns.

May be that time stamp causes that chaos ?

0 Kudos
Highlighted
New Contributor I
10 Views

>> May be that time stamp causes that chaos ?

Hi Paul, sorry my magic bowl still at repair service so I have to struggle over and look somewhere for a cure.

Don't remember what was version when I started using Cyclone V but 12.1 seems have cyclone support, did you try'd compile on that version to see if meet timing constraint?

 

https://www.intel.com/content/www/us/en/programmable/support/support-resources/knowledge-base/tools/...

Some issue remain open for a long time, I seen intermixing Verilog/VHDL sometimes has oddity like wrong signal propagation or signal name not assigned to, but you as me appear as using VHDL.

In the past I experienced trouble on constant conversion, sometimes where ok sometimes if then elsif else failed.

Actually I discovered VHDL 2018 sometimes fail convert , 3x"5" compare ok with "101", no warning no error but seldom havoc on code.

This forum is also bad and doesn't work searching old thread hiding knowledge.

try this link:

https://forums.intel.com/s/question/0D70P000005Peb5/is-if-elsif-else-in-this-case-errant?s1oid=00DU0...

Actually changing constants from format 3x"hex digit" 3x"0", 3x"1", 3x"2" .. to binary "000", "001", "010"... seems mitigated, more work in progress but primary task is to finish my work on time, then i can beta test Quartus.

Last question by KTAN9, yes I am using VHDL 2008, otherwise 3x"n" is invalid syntax.

No idea when and why sometimes fail nor found how to repeat in non stocastic mode.

0 Kudos
Highlighted
Moderator
10 Views

Is there any update?
0 Kudos
Highlighted
Beginner
10 Views

I am expecting some feedback from Altera, sorry, Intel. The behaviour persists.

0 Kudos
Highlighted
Moderator
10 Views

Can you help to clarify few things below?

 

1) Does your design need to run 3 days in order to finish the compilation? Or only this happened when you used DSE?

 

2) Did you try analyze your design on the timing violation before using the DSE to close timing? Using DSE is one of the method to close timing, but you will need to analyze the path is that make sense to use DSE to close timing, if you have a slack that is very large, it does not make sense to use DSE to close the timing.

 

3) Is that fine for you to send the timing report to have a look on the failure path?

 

4) I am not sure what it means by full moon, may be you can make some explanation on it?

0 Kudos
Highlighted
New Contributor I
10 Views

Hi IPaul, both version of quartus fail or just 17.1 18.1?

I am looking around on how to isolate my ghosts and skeleton...

If applicable, how long is in term of code line the file causing failure?

Yesterday bug appeared again on a large file then dissolved asap I trimmed commented code, I am investigating other troubled design...

I am thinking really odd influence from Van Hallen Belt and sun wind too ;)

0 Kudos
Highlighted
Moderator
10 Views

Is there any update with the above? wiill need more infomration for it.
0 Kudos
Highlighted
Beginner
10 Views

  1. No, it is just when using DSE and only sometimes.
  2. Yes, I am doing that when a slack is large, e.g. 2ns.
  3. Yes, I will, see below
  4. It occurs every month, I have been observing this for about a year. Always when full Moon (i.e. lunar phase). I know it sounds crazy but we do have that issue regularly. We do have a date stamp in our design, it comes from a server using tcl script which creates vhdl file with a date.

I mean this is not preventing us from our work except of 2-3 days a month. I will send you timing report as soon as it occurs again, next time is 18.May.

0 Kudos