Closed jimbo1990 closed 2 years ago
There is no attached test case so this is unactionable. Including the tool output is not a method of reproduction. Please see https://github.com/The-OpenROAD-Project/OpenLane/blob/master/docs/source/using_or_issue.md
There is no attached test case so this is unactionable. Including the tool output is not a method of reproduction. Please see https://github.com/The-OpenROAD-Project/OpenLane/blob/master/docs/source/using_or_issue.md
Thanks. The only problem is the tool never fails so I don't think that or_issue.py ever executes.
You can run it yourself
@jimbo1990
If synthesis issue you need provide your config.tcl
and source .v/.sv files to check it our end.
@jimbo1990 If synthesis issue you need provide your
config.tcl
and source .v/.sv files to check it our end.
Thanks, I will leave the files here if someone would be kind enough to try and run it to see if it passes the clock_skew_report on their side.
@jimbo1990
Yes getting same issue at my end.
we had same issue with previous shuttle and project called sta_hangs
. I'll try to get that project and its solutions and let you know.
@jimbo1990 Yes getting same issue at my end. we had same issue with previous shuttle and project called
sta_hangs
. I'll try to get that project and its solutions and let you know.
Thanks for your help. This has been driving me mad.
@mattvenn I think for one of the design, you are raised such STA hangs issue. I don't remember the issue tag. Do you have any solution/workaround for this issue? Thanks...
Is it possible to turn off the clock reporting? It's not really a solution but possibly to see if synthesis continues
This is still not a well package test case. Please provide a single self-contained invocation of the tool that fails rather than a pile of files.
@maliberty
For PnR stage failing through OpenROAD App auto generated as openroad_issue_reproducible
.
For standalone run like synthesis,gds_magic,gds_drc_klayout etc., I'm not found any steps to be followed.
Will check with @donn maybe I've missed the documentation, how to attach test case if flow fail other than OR Flow stage.
I fixed temporarily solved this problem as suggested in a previous post by commenting out the report_clock_skew from the sta.tcl and also from the multiple_cornes.tcl. obviously not ideal however at least the flow finished.
Probably IMPORTANT is that in the SDA.tcl file there is a comment which states that the function will hang if a clock less design is detected. What I don't get is why it doesn't detect the clock at that point or why it is detecting the design as clock less at that point in the synthesis while before it was calculating the slack and other parameters fine.
I think that is the wrong solution as the tool should be fixed if it hangs. However without a reproducer this whole discussion is going nowhere. Please provide one or close the issue.
Hi @jimbo1990 and @vijayank88,
There are instructions in the issue template to create the items required by the developers to fix this problem. I have included these instructions below for your convenience too;
Environment
Please run the following command in the OpenLane folder: python3 ./env.py issue-survey And copy and paste the ENTIRE output between the triple-backticks. Please do not gzip and upload the output. If there's no env.py, you are using an out of date version of OpenLane and should probably update.
Reproduction Material
- Upload a tarball containing the relevant design.
- List the commands used to run the design.
If you see a message like
Reproducible packaged: Please tarball and upload <PATH> if you're going to submit an issue
in your logs, please also tarball and include that path. This will greatly speed up the fixing process.
If one of you can provide a tarball containing the relevant design, I'm sure someone like @maliberty or @donn will be able to fix the report_clock_skew
to not hang on your design. This will be a great help to future users of these tools!
Looking forward to seeing you upload the needed information and seeing the problem getting fixed.
Tim '@mithro' Ansell
@jimbo1990 If synthesis issue you need provide your
config.tcl
and source .v/.sv files to check it our end.Thanks, I will leave the files here if someone would be kind enough to try and run it to see if it passes the clock_skew_report on their side.
@mithro @maliberty or_issue.py used to reproduce issue tar file only for OpenRoad flow stage fails only(Floorplan to routing). For synthesis stage if flow fails we normally upload source RTL files with config.tcl. Steps to reproduce the issue:
@maliberty @donn
Here's standalone testcase for sta hangs:
run3_sta_packaged.zip
Steps to reproduce the issue:
-Goto OpenLane directory
-unzip the attached file inside OpenLane/_build
-Do make mount
from OpenLane
directory
-cd _build/run3_sta_packaged
-sta
%source opensta.tcl
This will hang during report_clock_skew
Issue confirmed... @maliberty Be aware that the issue is packaged a bit differently and so the usual run scripts are not functional here.
I packaged it myself and have reported it. @donn please address this hole in packaging test cases.
@Manarabdelaty when you did this in sta.tcl you should have filed a bug report:
# report clock skew if the clock port is defined
# OR hangs if this command is run on clockless designs
How did you package the case? Thanks
How did you package the case? Thanks
By hand, just looked at the sta.tcl and pulled out the relevant parts. I'll upload my test case for comparison.
@Manarabdelaty this comment is wrong. report_clock_skew works just find if there are no clocks defined.
# OR hangs if this command is run on clockless designs
The issue has nothing to do with whether or not clocks are defined. It is a bug dealing with combinational feedback loops. OpenSTA has already been fixed but has not been integrated into openroad yet because of all the required verification steps.
@jjcherry56 @maliberty @Manarabdelaty is no longer with us aside from informal consultations. If you have any comments or observations, you tell me.
OpenROAD e3315ba41 fixes it on our end
Description
Synthesis hangs at report_clock_skew and never continues passed that point. No errors are given it simply doesnt print and then doesnt proceed forward.
Environment
Reproduction Material
Expected behavior
Should print clock skew report and continue synthesis