Apologies if it's already been discussed here, but...
# openlanet
Tobias Strauch
10/09/2022, 1:26 PMApologies if it's already been discussed here, but SDF timing arcs of each FF seem to be unconditional (having the same value fo bc, nc, wc conditions), which looks odd to me. Random example:
(CELL
(CELLTYPE "sky130_fd_sc_hd__dfrtp_1")
(INSTANCE 29240)
(DELAY
(ABSOLUTE
(IOPATH CLK Q (0.3900.3900.390) (0.4060.4060.406))
(IOPATH RESET_B Q () (0.0000.0000.000))
)
)
(TIMINGCHECK
(REMOVAL (posedge RESET_B) (posedge CLK) (0.3370.3370.337))
(RECOVERY (posedge RESET_B) (posedge CLK) (-0.211 0.211-0.211))
(HOLD (posedge D) (posedge CLK) (-0.058 0.058-0.058))
(HOLD (negedge D) (posedge CLK) (-0.074 0.074-0.074))
(SETUP (posedge D) (posedge CLK) (0.0860.0860.086))
(SETUP (negedge D) (posedge CLK) (0.1440.1440.144))
)
)
Any idea why ?
m
Matt Liberty
10/09/2022, 8:51 PMmost likely those values come straight from the .lib. Do they differ there?
t
Tobias Strauch
10/10/2022, 6:45 AMWhich lib file should I check ?
Has this been reported so far ? I don't want to submit an issue of a problem, which has already been reported.
m
Matt Venn
10/10/2022, 3:22 PM@Tim Edwards might know
t
Tim Edwards
10/10/2022, 3:29 PM@Matt Venn: I have no idea how the SDF is generated, so I don't know how those numbers are derived. More likely @mshalan would know the answer.
m
Matt Venn
10/10/2022, 3:29 PMthanks
m
Matt Liberty
10/10/2022, 5:35 PMyou would look at the liberty containing sky130_fd_sc_hd__dfrtp_1. Are you running with multiple corners or a single corner?
t
Tobias Strauch
10/10/2022, 6:06 PM@Matt Liberty Thanks for your reply.
I'm still having a hard time following your advice. I was searching below
https://github.com/google/skywater-pdk-libs-sky130_fd_sc_hd
and on my local PDK installation, but couldn't find the explicit sky130_fd_sc_hd__dfrtp_1 file which contains the relevant information and which is the source for the SDF file generation. A file name ending would greatly simplify the search š
The "multiple corners or single corner" question is confusing to me. This is the default Caravel\Openlane design flow, where the SDF is generated by default and which is then used for multiple corners or single corner STA checks. I didn't modify the flow in this regard. I checked and it looks as if this issue has been around ever since.
Cheers !
m
Matt Liberty
10/10/2022, 6:20 PMTiming libraries end in .lib. I'd have to dig through how OL is setup to really understand this. If it reads two libraries and they have different values then I would expect to see that. If it reads one or both libraries have the same value then this result is correct
t
Tobias Strauch
10/10/2022, 6:37 PMThis result cannot be correct. Occasionally timing arcs can be the same, especially for lower values. But having always all bc, nc, wc values with the same value is clearly a bug. I think I'll submit an OL isssue tomorrow.
Tobias Strauch
10/11/2022, 10:45 AMOkay, so we have one SDF per process corner. Now your comment is getting more clear to me, Matt.
The sdf file in the sdf/ directory is the nominal one, as min and max get overwritten, right?
Then I guess it is okay to have the same value for each bc, nc, wc timing arc.
But for all non-FF related timing arcs (non-FF cells, interconnects), the values differ. What do they stand for ?
It is stated, that āThe values only differ if there are conditional timing arcs or slew mergingā. But Iām having trouble understanding it. Any help on this one would be awesome.
https://github.com/The-OpenROAD-Project/OpenLane/issues/1419
3 Views