What is a 'unexpandable_clocks' in PrimeTime's check_timing report?

[gomlife330]

Hello everyone.

I have a warning in my PrimeTime check_timing report as follows.

Warning: There is 1 clock set for which periods were not expandable.

Failed clocks set Related clocks
----------------------------------------------------
1 CLK0
1 CLK1
1 CLK2
1 CLK3


Actually, I found a description about 'unexpandable_clocks' using 'man check_timing' in my pt_shell promt.
But, I can not understand exact meaning of 'unexpandable_clocks'.

Someone, could you please explain to me about this warning?

 

[jrwebsterco]

Are you already familiar with how PT will "expand" the periods of two clocks with different periods in order to find e.g. the least-common-multiple time where both clocks have a coincident edge?
(e.g. the effective period gets expended in the timing-report in order to find the tightest slack path between the two clocks)

I suspect (even though nothing shows up in SolvNet) that "unexpandable" means that PT could not eventually find an interval where "beating" of the two clocks occurs in a repeating pattern.
The assumption here is that PT gives up after iterating over larger and larger multiples of the period to find a common repeating pattern.

The bottom line is that you are either mis-defining your clock periods, or else you are overlooking e.g. false-path or set_max_delay declarations between totally unrelated clocks.

BTW, coincidentally there is an article that talks about this exact topic which you should read.
It is in the latest Xilinx Xcell Journal # 85, available on Xilinx's website, starting on page 30: "Demystifying Unexpanded Clocks".
While this is talking about Xilinx's Vivado timing analyzer, I feel the content is applicable to PT as well (noting Vivado supports the Synopsys SDC format for timing setups).