ecl/src/gabriel/chart
2001-06-26 17:14:44 +00:00

62 lines
3.1 KiB
Text

AKCL version 1.20
LUCID version 2.1.1
KCL is the Jun 87 version unmodified.
Gabriel benchmarks run on a Sun 3-160 by Bill Schelter.
Colu
Thu Apr 7 10:47:59 CDT 1988
HARDWARE: Sun280 Sun280 Sun3-160 Sun3-160 Sun3-160
SOFTWARE: AKCL LUCID AKCL LUCID KCL
BOYER 6.667 6.533 9.200 12.200 15.160 15.950
BROWSE 12.900 13.433 18.500 24.183 29.560 41.283
CTAK 3.500 3.000 1.140 6.583 2.160 7.050
DDERIV 3.317 3.317 5.420 6.033 9.260 8.417
DERIV 2.633 2.700 5.240 4.850 8.200 6.450
DESTRU-MOD 1.000 1.090
DESTRU 1.350 1.667 1.080 2.883 1.720 3.717
DIV2 3.117 3.100 4.380 5.433 6.660 5.667
FFT-MOD 5.350 5.317 38.360 7.700 51.880 29.467
FFT 32.150 32.317 34.760 58.450 51.840 (error)
FPRINT 0.667 1.033 0.800 1.817 1.460 2.417
FREAD 1.212 1.333 1.670 2.500 3.220 2.783
FRPOLY 46.883 46.033 30.040 85.450 45.200 90.167
PUZZLE-MOD 4.383 4.350 4.380 7.117 6.960 10.183
PUZZLE 4.833 4.900 3.180 8.200 5.140 11.183
STAK 1.813 1.817 1.380 3.717 2.620 4.350
TAK-MOD 2.900 2.900 2.800 5.150 4.680 3.600
TAK 3.833 3.750 2.760 6.467 4.900 5.767
TAKL 1.350 1.583 1.680 2.600 2.640 3.017
TAKR 0.521 0.533 0.460 0.933 0.840 1.483
TPRINT 1.046 1.650 2.100 3.117 3.680 4.233
TRAVERSE 25.667 25.833 27.940 38.917 40.060 71.133
TRIANG-MOD 63.500 62.217 255.460 112.750 453.460 160.783
TRIANG 78.017 77.450 76.760 147.567 141.62 220.917
Additional notes: All files were proclaimed using the proclaim-file
facility in make-declare.lsp. The XX-MOD tests, differ from the
original XX tests only in the matter of additional proclamations and
perhaps changing an array type. This sometimes hurt lucid, mainly
because the original declarations were optimal for lucid. (Also in
several places we changed an eq of integers or characters to an eql,
to conform with correct Common Lisp, and ensure that the test would
actually run as intended.) In AKCL three of the four float types map
to 64 bit floats, with only short-float being a 32 bit float. On
lucid all four appear to be 32 bit (judging from the fact that 1.0e38
is the largest float in each case). Thus KCL and AKCL are using 64
bit floats in the fft examples, while lucid appears to be using 32 bit
floats. We did not know what switches to use to speed lucid floating
arithmetic.
To run the tests in a lisp called foo (eg akcl, lucid,..)
% cd gabriel
% make -e "LISP=foo" >> #errs &
The times will be recorded in a file called times. All files
will be proclaimed, and compiled automatically.