jasst3@cisunx.UUCP (Jeffrey A. Sullivan) (03/23/88)
Here is a benchmark examination of Coral Allegro CL vs Lucid CL that is part of a review article I did for our ITS Newsletter. Sorry the formatting is not great, but it's the best I could do on short notice. I also tried to run the gabriel suite on Lyric CL from Xerox, but the first one took so long to run (never stopped in 5 minutes) that I gave it up. If someone has the gabriel suite for Lyric CL, I'd like to see it. Also, there is a benchmark listing in a past issue of BYTE. I can't recall which. Benchmarks In order to determine the relative efficiency of Allegro CL, the Gabriel suite of benchmarks was run in Allegro CL and Lucid CL on a Vax 8860. The results appear below. Allegro Optimized Allegro Vax (C) Vax (I) Boyer 36.000 23.050 7.81/1.01 418.12/26.66 Browse 65.233 52.800 8.87/0.58 587.73/37.06 CTAK 4.317 3.500 0.74/0.01 35.81/2.99 Dderiv 31.933/4.267 31.017/4.283 3.83/0.2 48.02/2.68 Deriv 32.250/4.183 31.350/4.183 3.06/0.18 54.84/2.03 Destructive 9.017 7.383 0.99/0.09 113.85/1.44 Div-iter 5.333 3.117 1.99/0.11 98.85/6.22 DiOAv-rec 5.000 2.667 2.43/0.08 58.96/4.17 FFT 67.083/4.233 66.550/4.217 41.59/2.13 433.94/28.56 Fprint 9.650 9.567 0.6/0.06 0.63/0.03 Fread 3.700 3.333 1.26/0.13 1.23/0.04 Frpoly Power=2 r=x+y+z+1 0.033 0.050 0.0 0.3/0.02 Frpoly Power=2 r2=1000r 0.017 0.017 0.01 0.17/0.01 Frpoly Power=2 r3=r in flonums 0.017 0.017 0.02 0.18/0.01 Frpoly Power=5 r=x+y+z+1 0.167 0.117 0.02 1.61/0.13 Frpoly Power=5 r2=1000r 0.233 0.183 0.06/0.02 1.72/0.11 OAFrpoly Power=5 r3=r in flonums 0.183 0.150 0.04 1.6/0.16 Frpoly Power=10 r=x+y+z+1 1.517 1.167 0.22/0.02 18.2/1.27 Frpoly Power=10 r2=1000r 2.633 2.283 0.73/0.02 17.69/0.74 Frpoly Power=10 r3=r in flonums 1.900 1.550 0.36/0.01 18.07/0.82 Frpoly Power=15 r=x+y+z+1 9.767 7.567 1.54/0.13 115.77/1.78 Frpoly Power=15 r2=1000r 20.550 18.200 7.25/0.5 124.55/3.61 Frpoly Power=15 r3=r in flonums 12.133 9.867 3.83/0.24 118.95/3.26 Puzzle 69.283 68.267 20.29/0.87 823.23/46.0 STAK 15.200 15.667 0.77/0.03 45.38/0.86 TAK 1.250 0.700 0.26 29.48/0.26 TAKL 13.100 7.133 0.91 263.66/4.39 TAKR 1.433 0.917 0.38/0.03 29.75/0.56 Tprint 29.883 31.600 0.79/0.04 0.8/0.04 Traverse-init 36.533 35.517 5.57/0.29 913.73/85.41 Traverse 135.017 133.967 20.07/1.17 >2600 Note: The optimized Allegro column refers to compiled benchmarks including the form (declare (optimize (speed 3) (size 0) (safety 0) ... ) wrapped around them. The difference between the Lucid CL benchmarks when compiled and interpreted is quite marked. The (C) column refers to files loaded with (load "file.vbin") after (compile-file "file.cl") has been completed. The (I) column refers to files loaded with (load "file.cl"). -- .......................................................................... Jeffrey Sullivan | University of Pittsburgh jas@cadre.dsl.pittsburgh.edu | Intelligent Systems Studies Program jasper@PittVMS.BITNET, jasst3@cisunx.UUCP | Graduate Student