benchmarks against mlton-20011006

Stephen Weeks MLton@sourcelight.com
Sat, 30 Mar 2002 15:49:41 -0800


Here's how the new MLton stacks up against the previous release on all
our standard benchmarks.  Anoq, I thought you might be interested
since you noticed that MLton feels faster.  :-)

Anyways, the improvements are pretty good!


MLton0 -- mlton release
MLton1 -- mlton-20011006

run time ratio
benchmark         MLton1
barnes-hut           1.0
checksum             1.1
count-graphs         1.1
DLXSimulator         0.8
fft                  0.9
fib                  1.0
hamlet               1.1
imp-for              2.0
knuth-bendix         1.2
lexgen               1.1
life                 1.4
logic                1.3
mandelbrot           1.2
matrix-multiply      4.7
md5                  3.0
merge                0.9
mlyacc               1.1
mpuz                 1.0
nucleic              0.8
peek                 3.2
psdes-random         1.3
ratio-regions        1.0
ray                  1.0
raytrace             0.9
simple               1.0
smith-normal-form    0.8
tailfib              1.1
tak                  1.2
tensor               1.5
tsp                  1.0
tyan                 1.1
vector-concat        1.6
vector-rev           1.8
vliw                 1.2
wc-input1            1.5
wc-scanStream        1.7
zebra                1.2
zern                 1.1

compile time
benchmark         MLton0 MLton1
barnes-hut           1.6    1.6
checksum             0.4    0.5
count-graphs         1.1    1.2
DLXSimulator         2.9    2.8
fft                  0.8    0.8
fib                  0.4    0.4
hamlet              28.9   27.2
imp-for              0.4    0.4
knuth-bendix         1.3    1.5
lexgen               3.9    3.4
life                 0.8    0.9
logic                1.8    3.6
mandelbrot           0.4    0.5
matrix-multiply      0.4    0.5
md5                  0.8    1.1
merge                0.4    0.5
mlyacc              13.7   13.1
mpuz                 0.5    0.6
nucleic              1.5    2.0
peek                 0.7    0.7
psdes-random         0.4    0.5
ratio-regions        1.5    1.7
ray                  2.4    2.3
raytrace             6.5    5.7
simple               4.2    4.1
smith-normal-form    4.0    4.2
tailfib              0.3    0.4
tak                  0.4    0.4
tensor               1.9    1.9
tsp                  0.9    1.0
tyan                 2.4    2.3
vector-concat        0.4    0.5
vector-rev           0.4    0.5
vliw                 8.4    7.7
wc-input1            1.1    1.1
wc-scanStream        1.1    1.1
zebra                3.5    5.2
zern                 0.7    0.7

size
benchmark            MLton0  MLton1
barnes-hut           56,959  59,365
checksum             22,733  20,497
count-graphs         44,205  40,049
DLXSimulator         88,285  77,849
fft                  32,765  29,061
fib                  22,765  20,497
hamlet            1,105,596 946,276
imp-for              22,765  21,393
knuth-bendix         64,358  59,362
lexgen              152,165 122,745
life                 39,501  38,161
logic                79,917 147,137
mandelbrot           22,797  20,481
matrix-multiply      23,341  20,993
md5                  32,550  33,682
merge                24,013  21,521
mlyacc              471,365 416,329
mpuz                 27,309  26,257
nucleic              61,677  60,289
peek                 31,494  28,130
psdes-random         24,237  21,505
ratio-regions        42,605  41,521
ray                  85,692  66,708
raytrace            237,321 158,945
simple              181,661 148,109
smith-normal-form   137,519 140,689
tailfib              22,477  20,241
tak                  22,861  20,577
tensor               57,006  62,152
tsp                  37,894  33,378
tyan                 85,798  76,954
vector-concat        23,725  21,185
vector-rev           23,661  21,009
vliw                299,861 263,973
wc-input1            48,638  38,826
wc-scanStream        49,598  41,194
zebra               109,670 104,602
zern                 30,396  26,116