Menu

#995 assertion in synth2 when attempting to process dpram

devel
closed-fixed
nobody
None
4
2016-01-30
2015-08-20
No
$ iverilog -V | grep "version 1"
Icarus Verilog version 11.0 (devel) (s20150603-96-g19009ec)
Icarus Verilog Preprocessor version 11.0 (devel) (s20150603-96-g19009ec)
Icarus Verilog Parser/Elaborator version 11.0 (devel) (s20150603-96-g19009ec)
$ iverilog -Wall dpram.v
$ iverilog -Wall dpram.v -tsizer
dpram.v:13: assert: synth2.cc:232: failed assertion nex_out.pin_count()==1
Aborted

This is from the most basic dpram model imaginable.

1 Attachments

Discussion

  • Larry Doolittle

    Larry Doolittle - 2015-08-20

    Here's a simple patch to turn the assertion error into a halfway-legible "unsupported" message. I don't see synthesis code for RAM in there now. After application, synthesis still fails, but the message becomes


    $ iverilog -Wall dpram.v -tsizer
    dpram.v:13: error: NetAssignBase::synth_async on unsupported lval {mem[word=<select=addra[7:0][(0)+:10] expr_type=logic>]}
    dpram.v:13: error: Unable to synthesize synchronous process.
    dpram.v:12: warning: Process not synthesized.
    2 error(s) in post-elaboration processing.

     
  • Martin Whitaker

    Martin Whitaker - 2016-01-30

    I've pushed an enhanced version of your patch to the master and v-10 branches. I needed to add a check for the memory index being constant, as that is supported (and used in the memsynth4 test in the test suite). I took the opportunity to add a more helpful error message.

    Do you want to close this bug, or keep it open as a feature request?

     
    • Larry Doolittle

      Larry Doolittle - 2016-01-30

      Martin -

      On Sat, Jan 30, 2016 at 05:15:24PM +0000, Martin Whitaker wrote:

      I've pushed an enhanced version of your patch to the master and v-10 branches. I needed to add a check for the memory index being constant, as that is supported (and used in the memsynth4 test in the test suite). I took the opportunity to add a more helpful error message.

      I see that. Thanks!

      Do you want to close this bug, or keep it open as a feature request?

      I say close it. A sorry message is not a crash.

      • Larry
       
  • Martin Whitaker

    Martin Whitaker - 2016-01-30
    • status: open --> closed-fixed
     
  • Martin Whitaker

    Martin Whitaker - 2016-01-30

    Done!

     

Log in to post a comment.