"Eye-Test" Calyx Sharing numbers for Lenet and Squeezenet

[calebmkim]

The number of resources used in a fully inlined (and then compiled) Calyx file, with different sharing bounds. Since it was fully inlined, I copy and pasted just the main component into a separate file and ran ctrl+f for different resources to get the estimates.

As a refresher for what the bounds mean, the first number is the number of times a combinational component will be shared, second number is the number of times a register will be shared, and third number is number of times all other components will be shared. I can check the numbers for sharing for combinational components too , but I think @rachitnigam mentioned that we probably don't want any sharing of these types of cells. Another disclaimer to add is that I haven't verified the correctness of these large Calyx files either.

Lenet:

(ctrl + f)	no sharing	bounds = 1,2,2	bounds=1,4,4	always share
"std_reg"	601	517	478	454
"std_fp_smult_pipe"	38	34	32	31
"std_mult_pipe"	8	8	8	8
"std_div_pipe"	3	2	2	1

Squeezenet:

(ctrl + f)	no sharing	bounds = 1,2,2	bounds=1,4,4	always share
"std_reg"	2197	1774	1565	1428
"std_fp_smult_pipe"	27	14	7	1
"std_mult_pipe"	60	60	60	60
"std_div_pipe"	1	1	1	1

As you can see, there wasn't much sharing of multipliers (but the register sharing looks good, and you can see that the bounded sharing stuff is working as expected). I think if you guys agree with my thinking on #1181 and its corresponding issue (#1180) then I can retry the sharing numbers for multipliers.

[rachitnigam]

Wowowow! These numbers are already very cool. I am particularly excited to see std_fp_smult_pipe getting shared down to exactly 1 for Squeezenet!

@sampsyo even without synthesis results, this table is looking pretty compelling for a short paper. I think even if we can't figure out a way to get resource estimates, this seems a plausible direction for a short paper