Experimental Results
BOOM Benchmarks with 20% of DCs were solved by BOOM 1.2 and ESPRESSO 2.3.
Comparison 1. Comparing the result quality
First, the benchmarks were solved by ESPRESSO, then BOOM was run iteratively using the same runtime.
The quality of results was compared.
As the quality criterion was selected the sum of the number of literals and the output cost, which approximates the number of 2-input gates (NAND, NOR)
needed to implement the function.
All the tables contain the average values of the ten benchmarks (of the same size)
The dark shaded cell represents the only case, where the result obtained by BOOM was worse.
The results of the benchmarks with more than 200 input variables and/or terms are not available yet, because
the ESPRESSO runtimes are too high.
Computer with AMD Athlon 900 MHz and 256 MB RAM was used
iv - number of input variables
p - number of care terms
Table entry format:
First row: BOOM results: #of literals / output cost / #of implicants (# of iterations)
Second row: ESPRESSO results: #of literals / output cost / #of implicants / time in seconds
|
p / iv |
50 |
100 |
150 |
200 |
|
50 |
110/41/25 (58)
122/54/27/3.89 |
96/35/23 (90)
104/45/23/10.29 |
90/32/21 (147)
92/41/21/24.87 |
84/29/20 (199)
89/39/20/41.99 |
|
100 |
284/86/52 (46)
289/104/51/19.31 |
229/68/42 (94)
231/84/42/77.07 |
217/61/40 (140)
213/80/39/199.17 |
207/57/38 (140)
201/74/37/246.21 |
|
150 |
474/132/76 (43)
481/158/76/54.76 |
389/101/63 (101)
384/125/62/282.80 |
362/92/61 (116)
345/113/56/646.20 |
381/90/64 (64)
322/107/52/1066.14 |
|
200 |
678/177/101 (51)
686/209/101/162.62 |
553/137/83 (116)
539/165/81/730.91 |
492/125/75 (207)
480/149/72/1913.65 |
469/110/71 (277)
450/136/68/3372.66 |
Comparison 2. Comparing the computational time
First, the benchmarks were solved by ESPRESSO, then BOOM was run iteratively, until the result of the same or better quality was reached.
The runtimes were compared.
As the quality criterion was selected the sum of the number of literals and the output cost.
All the tables contain the average values of the ten benchmarks (of the same size)
Note that the BOOM results may vary from the results in the first table, as many random events are present in BOOM.
In all cases BOOM was faster.
The results of the benchmarks with more than 200 input variables and/or terms are not available yet, because
the ESPRESSO runtimes are too high.
Computer with AMD Athlon 900 MHz and 256 MB RAM was used
iv - number of input variables
p - number of care terms
Table entry format:
First row: BOOM results: # of literals+output cost / time in seconds (# of iterations)
Second row: ESPRESSO results: # of literals+output cost / time in seconds
|
p / iv |
50 |
100 |
150 |
200 |
|
50 |
170/0,64 (12)
176/3,89 |
145/1,89 (21)
149/10,29 |
131/14,52 (73)
133/24,87 |
126/3,26 (25)
128/41,99 |
|
|
100 |
388/7,15 (23)
393/19,31 |
313/25,5 (48)
315/77,07 |
291/38,91 (56)
293/199,17 |
273/86,51 (83)
275/246,21 |
|
|
150 |
631/20,38 (25)
639/54,76 |
506/153,84 (70)
509/282,8 |
456/374,68 (105)
458/646,20 |
427/974,40 (161)
429/1066,14 |
|
|
200 |
890/71,97 (31)
895/162,62 |
697/467,63 (86)
704/730,91 |
625/1026,28 (149)
629/1913,65 |
582/1759,27 (220)
586/3372,66 |
|
|