Project

General

Profile

[logo] 
 
Home
News
Activity
About/Contact
Major Tools
  Dinotrace
  Verilator
  Verilog-mode
  Verilog-Perl
Other Tools
  BugVise
  CovVise
  Force-Gate-Sim
  Gspice
  IPC::Locker
  Rsvn
  SVN::S4
  Voneline
  WFH
General Info
  Papers

Issue #1456

sameHash error on type parameters after 4.014

Added by Stefan Wallentowitz 4 months ago. Updated 3 months ago.

Status:
Closed
Priority:
Normal
Assignee:
-
Category:
TranslationError
% Done:

0%


Description

Hi,

when building the ariane core (https://github.com/pulp-platform/ariane) with the current master, I get the following internal error:

%Error: Internal Error: /data/ariane/src/load_store_unit.sv:211: ../V3Hashed.cpp:70: sameHash function undefined (returns 0) for node under CFunc. ... See the manual and http://www.veripool.org/verilator for more assistance.

I can try to bisect or look into candidates later this month. It works with 4.014

History

#1 Updated by Todd Strader 4 months ago

Ugh, smells like this guy: https://www.veripool.org/issues/1427-Verilator-Incorrect-real-parameter-assignment

I probably still have this swapped in. I'll pull the Ariane repo and see what I can find.

#2 Updated by Todd Strader 4 months ago

Looks like I broke something with parameter types: https://github.com/toddstrader/verilator-dev/tree/samehash-fix

I'll work on a fix.

#3 Updated by Todd Strader 4 months ago

This fixes the Ariane build, but my new test still has issues: https://github.com/toddstrader/verilator-dev/tree/bug_escapes_to_cpp

It's possible that those issues predate my #1427 fix. I'm still investigating.

#4 Updated by Todd Strader 4 months ago

I re-ran the test suite with this fix and everything passes. I also rewound to just before the #1427 fix and tried out my new test. It's breaks in the same way that it does with my latest fix. So there are two things going on here:
  1. I broke type parameters (under certain circumstances) with my #1427 fix. That should be resolved now.
  2. Dotted references to type parameters don't seem to work correctly. See the commented out portions of my new test. I'm always getting $bits() == 32. Which is interesting because that's not the default type size. I guess it just returns the size of an int or something.

Wilson, what do you want to do here? Do you want to land this in two parts? If so, #1 should be ready here: https://github.com/toddstrader/verilator-dev/tree/samehash-fix And if we're landing in two parts I'll open another issue. I'd prefer this, because fixing #2 isn't at the top of my list right now.

#5 Updated by Wilson Snyder 4 months ago

  • Subject changed from Internal error after 4.014 to sameHash error on type parameters after 4.014
  • Category set to TranslationError
  • Status changed from New to Resolved

Todd, great patch, thanks. Merged that, please file another bug if you can't look at #2, as I can't look immediately either.

Pushed to git towards 4.015.

#6 Updated by Todd Strader 4 months ago

Sorry, one more update to the same branch: https://github.com/toddstrader/verilator-dev/tree/samehash-fix

It dawned on me that this test should have a --debug-collision variant.

#7 Updated by Wilson Snyder 4 months ago

Suspect you forgot to push samehash-fix.

#8 Updated by Todd Strader 4 months ago

Sorry, what do you think is missing? This commit hash the additional --debug-collision test: https://github.com/toddstrader/verilator-dev/commit/1c99e6f908670d699c264c145f89e0d4391a73b7

#9 Updated by Wilson Snyder 4 months ago

Pushed, thanks.

#10 Updated by Wilson Snyder 3 months ago

  • Status changed from Resolved to Closed

In 4.016.

Also available in: Atom