shallow-comparison is longer running on xilinx than lattice

[bkushigian]

Xilinx times:

OP	Xilinx (Sec)	Lattice (Sec)
eq8	8.13897705078125s	1.3178119659423828s
eq16	34.87053894996643s	1.3003449440002441s
eq32	61.21352696418762s	1.3902618885040283s
eq64	88.46406102180481s	2.665834903717041s
neq8	16.96060276031494s	1.1544711589813232s
neq16	17.297454118728638s	1.5855050086975098s
neq32	65.9597601890564s	1.7859752178192139s
neq64	268.5559937953949s	2.378412961959839s

[bkushigian]

This is almost 100% because of the extra unused inputs into the LUT6s. These currently require symbolic bits which probably blow up the search space

[gussmith23]

one easy solution is to plug constants into the unused inputs!

On Mon, Feb 20, 2023, 6:54 PM Ben A Kushigian @.***> wrote:

This is almost 100% because of the extra unused inputs into the LUT6s. These currently require symbolic bits which probably blow up the search space

— Reply to this email directly, view it on GitHub https://github.com/uwsampl/lakeroad/issues/192#issuecomment-1437788756, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJFZALWZI4N2SRSOQHIWYLWYQU6ZANCNFSM6AAAAAAVCPYNMI . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[gussmith23]

I think this is adding a non-negligible amount of time to CI -- should make this a priority just for our own sanity