Hi Jason,
I tried running the VTR benchmarks on the k6_frac_N10_40nm architecture, but
some, like bgm, gave an error in read_blif: "Did not find matching model to
subckt multiply". This error did not make sense since there was clearly a
".model multiply" to correspond to the ".subckt multiply".
I thought this was a blif parsing issue and rearranged the blif to put the
.model before the .subckt, which gave a different error: "No corresponding
model multiply in architecture description." This error allowed me to diagnose
the problem: this architecture lacks the hard multipliers (and memories)
required by this circuit, and switching to the Comprehensive Architecture
solved the problem.
Would it be possible to always give the second error in this case? If users
synthesize their own blifs using the upstream tools, they might think that the
first error was caused by a corrupted blif, while the second error makes clear
that this is an architecture issue.
Thanks,
Michael
Original issue reported on code.google.com by JasonKai...@gmail.com on 27 Apr 2014 at 7:45
Original issue reported on code.google.com by
JasonKai...@gmail.com
on 27 Apr 2014 at 7:45