[MLton-devel] gcc -O1 vs -O2 on Sparc
Stephen Weeks
MLton@mlton.org
Mon, 28 Apr 2003 22:38:52 -0700
I ran all the benchmarks -cc-opt -O{1,2} on my Sparc machine. The
numbers are below. There are some pretty decent speedups with -O2,
usually on the floating point benchmarks. But the compile times are
quite a bit worse. So, I'm going to leave -O1 as the default and put
a note in the user guide.
MLton0 -- /usr/local/bin/mlton -cc-opt -O1
MLton1 -- /usr/local/bin/mlton -cc-opt -O2
WARNING: /usr/local/bin/mlton -cc-opt -O1 failed on: vliw
run time ratio
benchmark MLton1
barnes-hut 0.86
boyer 0.99
checksum 0.90
count-graphs 0.99
DLXSimulator 1.00
fft 1.00
fib 1.00
hamlet 0.99
imp-for 1.00
knuth-bendix 0.98
lexgen 0.93
life 1.06
logic 0.98
mandelbrot 1.05
matrix-multiply 0.99
md5 0.87
merge 0.95
mlyacc 0.98
model-elimination 0.98
mpuz 0.99
nucleic 0.82
peek 0.86
psdes-random 0.93
ratio-regions 0.95
ray 0.95
raytrace 0.94
simple 0.92
smith-normal-form 1.00
tailfib 1.00
tak 0.93
tensor 0.73
tsp 0.84
tyan 0.97
vector-concat 1.00
vector-rev 0.92
wc-input1 1.37
wc-scanStream 1.04
zebra 1.01
zern 0.99
compile time
benchmark MLton0 MLton1
barnes-hut 15.11 24.84
boyer 66.55 187.54
checksum 3.51 4.02
count-graphs 9.94 18.22
DLXSimulator 24.07 52.95
fft 6.73 9.37
fib 3.07 3.59
hamlet 429.21 1405.30
imp-for 3.25 3.65
knuth-bendix 17.51 53.29
lexgen 39.99 134.22
life 9.27 19.73
logic 21.33 66.49
mandelbrot 3.39 3.93
matrix-multiply 3.64 4.25
md5 6.48 12.99
merge 3.40 4.21
mlyacc 185.26 740.39
model-elimination 177.08 578.89
mpuz 4.78 6.65
nucleic 62.38 155.11
peek 5.38 7.74
psdes-random 3.36 3.98
ratio-regions 14.04 28.74
ray 24.20 60.86
raytrace 61.10 159.22
simple 55.86 204.98
smith-normal-form 256.64 512.96
tailfib 3.00 3.48
tak 3.10 3.59
tensor 15.49 28.78
tsp 8.02 12.44
tyan 25.10 65.87
vector-concat 3.52 4.23
vector-rev 3.19 3.79
vliw 97.35 313.17
wc-input1 9.02 18.59
wc-scanStream 9.23 19.27
zebra 26.72 77.02
zern 5.57 7.29
run time
benchmark MLton0 MLton1
barnes-hut 149.23 129.01
boyer 185.88 184.18
checksum 446.99 401.86
count-graphs 207.67 205.69
DLXSimulator 248.62 248.29
fft 103.26 103.74
fib 209.93 209.15
hamlet 252.87 250.48
imp-for 108.88 109.04
knuth-bendix 188.24 183.55
lexgen 224.83 208.90
life 290.11 307.43
logic 203.47 199.57
mandelbrot 179.51 187.70
matrix-multiply 191.27 189.41
md5 974.39 852.11
merge 216.53 206.40
mlyacc 190.94 186.90
model-elimination 343.72 337.62
mpuz 187.83 186.69
nucleic 168.68 139.09
peek 140.23 120.20
psdes-random 136.84 127.13
ratio-regions 133.91 127.65
ray 113.40 107.77
raytrace 149.14 140.17
simple 255.28 233.79
smith-normal-form 545.45 544.83
tailfib 188.90 188.34
tak 333.30 308.85
tensor 143.29 104.18
tsp 149.42 124.89
tyan 222.22 216.21
vector-concat 244.02 243.79
vector-rev 211.68 195.79
wc-input1 131.46 180.30
wc-scanStream 151.40 157.84
zebra 287.21 289.52
zern 204.90 201.93
size
benchmark MLton0 MLton1
barnes-hut 163,597 162,893
boyer 186,425 162,393
checksum 72,590 72,366
count-graphs 92,926 91,614
DLXSimulator 138,101 134,965
fft 81,930 81,290
fib 72,542 72,382
hamlet 1,377,259 1,329,771
imp-for 72,358 72,198
knuth-bendix 120,917 117,397
lexgen 200,274 194,546
life 95,142 92,486
logic 126,929 123,825
mandelbrot 72,430 72,238
matrix-multiply 72,982 72,758
md5 83,154 82,226
merge 74,158 73,870
mlyacc 529,494 518,998
model-elimination 756,789 733,045
mpuz 78,918 78,374
nucleic 183,675 175,291
peek 81,202 80,658
psdes-random 73,206 72,886
ratio-regions 98,846 96,894
ray 142,104 139,064
raytrace 255,177 244,265
simple 245,009 236,049
smith-normal-form 299,841 232,993
tailfib 72,270 72,046
tak 72,598 72,406
tensor 162,728 161,544
tsp 88,771 89,131
tyan 146,109 142,461
vector-concat 73,894 73,574
vector-rev 72,990 72,734
vliw 407,354 396,538
wc-input1 95,690 94,346
wc-scanStream 96,650 95,178
zebra 162,736 156,400
zern 79,580 79,004
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
MLton-devel mailing list
MLton-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlton-devel