The-OpenROAD-Project
Fanout violations during global routing
@tspyrou As I tried APU design with GF180 PDK, during global routing seeing rise in fanout count causing timing violations of WNS with -234ns.
Attached DEF and groute.tcl
Log snippet: Complete log available in attachment. gftest1_groute_packaged.zip
Startpoint: _7272_ (rising edge-triggered flip-flop clocked by clk)
Endpoint: Sample[9] (output port clocked by clk)
Path Group: clk
Path Type: max
Fanout Cap Slew Delay Time Description
-----------------------------------------------------------------------------
1.87 1.63 19.86 ^ _4788_/ZN (OAI21_X1_7T5P0)
2 0.01 _1461_ (net)
1.87 0.00 19.86 ^ _4789_/A3 (XOR3_X1_7T5P0)
349.42 224.19 244.05 ^ _4789_/Z (XOR3_X1_7T5P0)
2511 3.08 _1462_ (net)
349.42 0.21 244.26 ^ _4790_/I (CLKBUF_X1_7T5P0)
4.68 -3.95 240.31 ^ _4790_/Z (CLKBUF_X1_7T5P0)
1 0.01 net71 (net)
4.68 0.00 240.31 ^ output71/I (CLKBUF_X4_7T5P0)
1.18 2.39 242.70 ^ output71/Z (CLKBUF_X4_7T5P0)
1 0.07 Sample[9] (net)
1.18 0.00 242.71 ^ Sample[9] (out)
242.71 data arrival time
10.00 10.00 clock clk (rise edge)
0.00 10.00 clock network delay (propagated)
-0.25 9.75 clock uncertainty
0.00 9.75 clock reconvergence pessimism
-2.00 7.75 output external delay
7.75 data required time
-----------------------------------------------------------------------------
7.75 data required time
-242.71 data arrival time
-----------------------------------------------------------------------------
-234.96 slack (VIOLATED)
Please package a complete testcase with all of the data to run it. The only thing in this tarfile is a def file and a command file and references to undefined variables.
@jjcherry56 Its private PDK. I am not sure what are the files to attach here. I will send as Email. Thanks.
It is fine to reference the PDK in or1:/platforms etc but I have to be able to run it in openroad (not shell, not openlane) with all of the design files and variables.
There is no GF180.git repo in either of the 2 places I know to find it: or1:/platforms and /home/zf4_projects/OpenROAD-guest/platforms. So where is it hiding?
@tspyrou pointed me to the pdk. I had to make some changes to the run script to make it work, but got it to run.
The groute.tcl script seg faults while trying to do antenna repair in everybody's favorite opendb wire code. There is some chance those "fanouts" were diodes that were inserted during antenna repair that managed to complete because the memory corruption on your run didn't cause a seg fault. Use the following command to look at the connections on the net: report_net -connections -verbose 1462
report_net
It is indeed an antenna repair that is the culprit. Compiling with optimization on disables an assertion check that causes it to fail.
There are a number of bugs that this testcase exposes but your biggest problem appears to be a density so high that antenna repair has no place to put diodes so it is unable to fix the problem and keeps trying over and over. Try reducing the density earlier in the flow so it isn't packed so tight.
The GF180 LEF metal layers have constant values for ANTENNADIFFSIDEAREARATIO. This means that adding diffusion with diodes does not repair the antenna violations because the allowed ratio of metal to gate area or side area does not change as it would with a PWL model. I have a pending PR that detects this situation and reports it instead of adding tons of pointless diodes.
@jjcherry56
sorry for responding late.
My access to GF180 restored recently. Complete OpenLane flow got setup and ran APU design again with versionC.
The test case I shared with you is versionA PDK lib.
With latest lib i am not facing fanout violations.
There are slew violations, its related ANTEANNA. Post routing sta report:
===========================================================================
report_tns
============================================================================
tns -5274.90
tns_report_end
wns_report
===========================================================================
report_wns
============================================================================
wns -42.00
wns_report_end
worst_slack
===========================================================================
report_worst_slack -max (Setup)
============================================================================
worst slack -42.00
===========================================================================
report_worst_slack -min (Hold)
============================================================================
worst slack 0.35
worst_slack_end
clock_skew
===========================================================================
aes and aes_core design has fanout issue again with gf180.
aes_core post routing sta report follows.
Startpoint: reset_n (input port clocked by clk)
Endpoint: _41039_ (recovery check against rising-edge clock clk)
Path Group: **async_default**
Path Type: max
Fanout Cap Slew Delay Time Description
-----------------------------------------------------------------------------
0.00 0.00 clock clk (rise edge)
0.00 0.00 clock network delay (propagated)
4.40 4.40 ^ input external delay
2.63 1.51 5.91 ^ reset_n (in)
2 0.04 reset_n (net)
2.63 0.00 5.91 ^ input389/I (BUF_X20_7T5P0)
10.62 6.99 12.90 ^ input389/Z (BUF_X20_7T5P0)
598 3.85 net389 (net)
10.63 0.18 13.08 ^ repeater523/I (BUF_X20_7T5P0)
14.34 7.65 20.72 ^ repeater523/Z (BUF_X20_7T5P0)
820 5.64 net523 (net)
14.34 0.14 20.87 ^ repeater522/I (BUF_X20_7T5P0)
16.74 9.66 30.53 ^ repeater522/Z (BUF_X20_7T5P0)
890 6.34 net522 (net)
17.83 2.38 32.91 ^ repeater521/I (BUF_X20_7T5P0)
11.98 7.37 40.28 ^ repeater521/Z (BUF_X20_7T5P0)
662 4.56 net521 (net)
14.32 3.16 43.44 ^ _41039_/RN (DFFRNQ_X2_7T5P0)
43.44 data arrival time
22.00 22.00 clock clk (rise edge)
0.00 22.00 clock source latency
3.73 2.10 24.10 ^ clk (in)
2 0.26 clk (net)
3.73 0.00 24.10 ^ clkbuf_0_clk/I (CLKBUF_X16_7T5P0)
0.80 1.66 25.75 ^ clkbuf_0_clk/Z (CLKBUF_X16_7T5P0)
8 0.18 clknet_0_clk (net)
0.80 0.01 25.76 ^ clkbuf_2_0_0_clk/I (CLKBUF_X8_7T5P0)
0.29 0.81 26.57 ^ clkbuf_2_0_0_clk/Z (CLKBUF_X8_7T5P0)
1 0.02 clknet_2_0_0_clk (net)
0.29 0.00 26.57 ^ clkbuf_2_0_1_clk/I (CLKBUF_X8_7T5P0)
0.29 0.65 27.21 ^ clkbuf_2_0_1_clk/Z (CLKBUF_X8_7T5P0)
1 0.02 clknet_2_0_1_clk (net)
0.29 0.00 27.22 ^ clkbuf_2_0_2_clk/I (CLKBUF_X8_7T5P0)
0.55 0.83 28.04 ^ clkbuf_2_0_2_clk/Z (CLKBUF_X8_7T5P0)
4 0.06 clknet_2_0_2_clk (net)
0.55 0.00 28.05 ^ clkbuf_3_1_0_clk/I (CLKBUF_X8_7T5P0)
0.27 0.71 28.76 ^ clkbuf_3_1_0_clk/Z (CLKBUF_X8_7T5P0)
1 0.01 clknet_3_1_0_clk (net)
0.27 0.00 28.76 ^ clkbuf_3_1_1_clk/I (CLKBUF_X8_7T5P0)
0.53 0.81 29.56 ^ clkbuf_3_1_1_clk/Z (CLKBUF_X8_7T5P0)
4 0.05 clknet_3_1_1_clk (net)
0.53 0.00 29.57 ^ clkbuf_4_2_0_clk/I (CLKBUF_X8_7T5P0)
0.66 0.97 30.53 ^ clkbuf_4_2_0_clk/Z (CLKBUF_X8_7T5P0)
4 0.07 clknet_4_2_0_clk (net)
0.66 0.00 30.54 ^ clkbuf_5_4_0_clk/I (CLKBUF_X8_7T5P0)
0.49 0.91 31.45 ^ clkbuf_5_4_0_clk/Z (CLKBUF_X8_7T5P0)
4 0.05 clknet_5_4_0_clk (net)
0.49 0.00 31.45 ^ clkbuf_6_9_0_clk/I (CLKBUF_X8_7T5P0)
2.77 2.17 33.62 ^ clkbuf_6_9_0_clk/Z (CLKBUF_X8_7T5P0)
28 0.38 clknet_6_9_0_clk (net)
2.77 0.02 33.64 ^ clkbuf_leaf_858_clk/I (CLKBUF_X16_7T5P0)
0.30 1.15 34.79 ^ clkbuf_leaf_858_clk/Z (CLKBUF_X16_7T5P0)
2 0.01 clknet_leaf_858_clk (net)
0.30 0.00 34.79 ^ _41039_/CLK (DFFRNQ_X2_7T5P0)
-0.25 34.54 clock uncertainty
0.00 34.54 clock reconvergence pessimism
-5.89 28.65 library recovery time
28.65 data required time
-----------------------------------------------------------------------------
28.65 data required time
-43.44 data arrival time
-----------------------------------------------------------------------------
-14.79 slack (VIOLATED)
Did you update the version of openroad you are using to include commit 062004eff that includes the change I described in a previous post? The testcase you originally submitted works as intended with the latest version of openroad when I run it.
Inserted antennas will always show up as "fanout" because they are connected to the net.
I can't tell if you have changed the pdk or not. If you are going change the design/testcase you must attach it to the issue or post it someplace to make it available.
@jjcherry56
Shared another test case to openroad.tools mail ID with test case for aes design for GF180 PDK.
I'm seeing fanout violation, with latest commit again.
Please keep this issue open, until GF180 test completes. Thanks
As gf180 PDK become public and OpenLane/open-pdks update going on. Once PDK enabled with OpenLane, I'll test again with latest OpenROAD and open a new issue if I'm facing anything again. Thanks!!!