Finding skew in report_checks -format full_clock

[oharboe]

I would like to understand how the new skew column relates to to report_checks -format full_clock.

Presumably, I should be able to sum + subtract numbers from either report_checks -format full_clock or report_checks (output below) and get the skew?

>>> report_checks -from {ces_6_0/io_lsbOuts_7} -to {ces_6_4/io_lsbIns_4} -format full_clock
Startpoint: ces_6_0 (rising edge-triggered flip-flop clocked by clock)
Endpoint: ces_6_4 (rising edge-triggered flip-flop clocked by clock)
Path Group: clock
Path Type: max

  Delay    Time   Description
---------------------------------------------------------
   0.00    0.00   clock clock (rise edge)
   0.00    0.00   clock source latency
   0.00    0.00 ^ clock (in)
  16.91   16.91 ^ wire2/Y (BUFx16f_ASAP7_75t_R)
  73.09   89.99 ^ wire1/Y (BUFx16f_ASAP7_75t_R)
  96.82  186.81 ^ clkbuf_0_clock/Y (BUFx24_ASAP7_75t_R)
  43.83  230.64 ^ clkbuf_1_1_0_clock/Y (BUFx24_ASAP7_75t_R)
  25.71  256.35 ^ clkbuf_2_2_0_clock/Y (BUFx24_ASAP7_75t_R)
  46.96  303.31 ^ clkbuf_3_5_0_clock/Y (BUFx24_ASAP7_75t_R)
  24.92  328.23 ^ ces_6_0/clock (Element)
 153.56  481.79 ^ ces_6_0/io_lsbOuts_7 (Element)
  41.14  522.94 ^ ces_6_1/io_lsbOuts_6 (Element)
  40.00  562.93 ^ ces_6_2/io_lsbOuts_5 (Element)
  36.94  599.87 ^ ces_6_3/io_lsbOuts_4 (Element)
   0.00  599.87 ^ ces_6_4/io_lsbIns_4 (Element)
         599.87   data arrival time

 250.00  250.00   clock clock (rise edge)
   0.00  250.00   clock source latency
   0.00  250.00 ^ clock (in)
  16.36  266.36 ^ wire2/Y (BUFx16f_ASAP7_75t_R)
  70.62  336.98 ^ wire1/Y (BUFx16f_ASAP7_75t_R)
  92.17  429.15 ^ clkbuf_0_clock/Y (BUFx24_ASAP7_75t_R)
  42.17  471.32 ^ clkbuf_1_1_0_clock/Y (BUFx24_ASAP7_75t_R)
  24.31  495.63 ^ clkbuf_2_3_0_clock/Y (BUFx24_ASAP7_75t_R)
  43.62  539.25 ^ clkbuf_3_7_0_clock/Y (BUFx24_ASAP7_75t_R)
  14.56  553.81 ^ ces_6_4/clock (Element)
 -10.00  543.81   clock uncertainty
   9.32  553.13   clock reconvergence pessimism
  44.57  597.70   library setup time
         597.70   data required time
---------------------------------------------------------
         597.70   data required time
        -599.87   data arrival time
---------------------------------------------------------
          -2.17   slack (VIOLATED)

>>> report_checks -from {ces_6_0/io_lsbOuts_7} -to {ces_6_4/io_lsbIns_4}
Startpoint: ces_6_0 (rising edge-triggered flip-flop clocked by clock)
Endpoint: ces_6_4 (rising edge-triggered flip-flop clocked by clock)
Path Group: clock
Path Type: max

  Delay    Time   Description
---------------------------------------------------------
   0.00    0.00   clock clock (rise edge)
 328.23  328.23   clock network delay (propagated)
   0.00  328.23 ^ ces_6_0/clock (Element)
 153.56  481.79 ^ ces_6_0/io_lsbOuts_7 (Element)
  41.14  522.94 ^ ces_6_1/io_lsbOuts_6 (Element)
  40.00  562.93 ^ ces_6_2/io_lsbOuts_5 (Element)
  36.94  599.87 ^ ces_6_3/io_lsbOuts_4 (Element)
   0.00  599.87 ^ ces_6_4/io_lsbIns_4 (Element)
         599.87   data arrival time

 250.00  250.00   clock clock (rise edge)
 303.81  553.81   clock network delay (propagated)
 -10.00  543.81   clock uncertainty
   9.32  553.13   clock reconvergence pessimism
         553.13 ^ ces_6_4/clock (Element)
  44.57  597.70   library setup time
         597.70   data required time
---------------------------------------------------------
         597.70   data required time
        -599.87   data arrival time
---------------------------------------------------------
          -2.17   slack (VIOLATED)


>>> report_checks -from {ces_6_0/io_lsbOuts_7} -to {ces_6_4/io_lsbIns_4}
Startpoint: ces_6_0 (rising edge-triggered flip-flop clocked by clock)
Endpoint: ces_6_4 (rising edge-triggered flip-flop clocked by clock)
Path Group: clock
Path Type: max

  Delay    Time   Description
---------------------------------------------------------
   0.00    0.00   clock clock (rise edge)
 328.23  328.23   clock network delay (propagated)
   0.00  328.23 ^ ces_6_0/clock (Element)
 153.56  481.79 ^ ces_6_0/io_lsbOuts_7 (Element)
  41.14  522.94 ^ ces_6_1/io_lsbOuts_6 (Element)
  40.00  562.93 ^ ces_6_2/io_lsbOuts_5 (Element)
  36.94  599.87 ^ ces_6_3/io_lsbOuts_4 (Element)
   0.00  599.87 ^ ces_6_4/io_lsbIns_4 (Element)
         599.87   data arrival time

 250.00  250.00   clock clock (rise edge)
 303.81  553.81   clock network delay (propagated)
 -10.00  543.81   clock uncertainty
   9.32  553.13   clock reconvergence pessimism
         553.13 ^ ces_6_4/clock (Element)
  44.57  597.70   library setup time
         597.70   data required time
---------------------------------------------------------
         597.70   data required time
        -599.87   data arrival time
---------------------------------------------------------
          -2.17   slack (VIOLATED)

[maliberty]

Are you accounting for CRPR?

[maliberty]

https://www.design-reuse.com/articles/40155/clock-path-pessimism-statistical-vs-logical.html

[oharboe]

CPPR is beyond my level of sophistication when it comes to understanding of clocks.

Silly question is "clock reconvergence pessimism" and CPPR the same? If so, why use two different nomenclatures? For unsophisticated users, like me, it helps enormously when things are named consistently.

[oharboe]

report_clock_skew uses, CRPR. Is that an anagram or something else?

Spelling it out in the report_clock_skew report would reduce confusion and disambiguation for unsophisticated users...

>>> report_clock_skew
Clock core_clock
   0.00 source latency ctrl.state.out[1]$_DFF_P_/CK ^
   0.00 target latency ctrl.state.out[1]$_DFF_P_/CK ^
   0.00 CRPR
--------------
   0.00 setup skew

[maliberty]

There is crpr and cppr (don't blame me) which are different. See the paragraph "CRPR vs. CPPR".

The upshot is that skew is not just source clock arrival - destination clock arrival but involved some additional calculation

[oharboe]

Summary from explanation I got: CRPR and CPPR are the same thing; the common part of the clock tree for the start and end point. This common path is then subtracted from the setup path time, thus reducing the effective skew, which report_clock_skew reports.

CRPR/CPPR is not listed in the timing report or in report_checks(which I suppose could be useful in some cases?), only in report_clock_skew.