RSSA backend
Matthew Fluet
fluet@CS.Cornell.EDU
Fri, 18 Jan 2002 13:19:33 -0500 (EST)
> I'm currently running benchmarks against the 2001-10-06 release. IIRC,
> that release has the broken limit-check coalescing, so we're likely to see
> some slowdowns.
Looks pretty good. A couple of slowdowns, but nothing compared to some of
the speedups. We're still hurting a bit in codesize because of the extra
limit checks. And, in general, compile times are down a little.
MLton0 -- mlton-stable
MLton1 -- mlton
compile time
benchmark MLton0 MLton1
barnes-hut 2.14 2.13
checksum 0.58 0.58
count-graphs 1.62 1.52
DLXSimulator 3.77 3.76
fft 1.11 1.17
fib 0.56 0.55
hamlet 44.17 45.33
imp-for 0.62 0.54
knuth-bendix 2.13 1.94
lexgen 4.84 4.99
life 1.29 1.17
logic 5.70 2.49
mandelbrot 0.60 0.60
matrix-multiply 0.67 0.65
md5 1.52 1.14
merge 0.61 0.60
mlyacc 19.98 18.64
mpuz 0.87 0.78
nucleic 3.00 2.42
peek 0.99 0.94
psdes-random 0.64 0.62
ratio-regions 2.47 2.23
ray 3.24 3.08
raytrace 8.62 9.03
simple 6.10 6.48
smith-normal-form 7.94 7.38
tailfib 0.59 0.58
tak 0.59 0.58
tensor 2.79 2.77
tsp 1.46 1.34
tyan 3.64 3.55
vector-concat 0.65 0.65
vector-rev 0.64 0.61
vliw 11.84 10.98
wc-input1 1.53 1.49
wc-scanStream 1.62 1.56
zebra 9.22 4.95
zern 1.02 1.01
run time
benchmark MLton0 MLton1
barnes-hut 3.93 4.57
checksum 3.22 3.18
count-graphs 4.97 4.23
DLXSimulator 14.95 14.65
fft 7.85 8.73
fib 3.41 3.45
hamlet 8.15 7.14
imp-for 12.95 7.16
knuth-bendix 6.59 5.91
lexgen 10.63 9.26
life 7.83 6.22
logic 26.52 18.30
mandelbrot 6.69 5.98
matrix-multiply 5.25 2.83
md5 3.32 1.90
merge 51.55 51.55
mlyacc 9.79 9.39
mpuz 4.64 4.22
nucleic 7.04 8.31
peek 3.44 0.82
psdes-random 3.41 3.17
ratio-regions 8.76 9.11
ray 3.92 3.99
raytrace 4.64 5.01
simple 6.34 6.37
smith-normal-form 0.98 0.98
tailfib 16.33 10.95
tak 7.88 7.75
tensor 7.13 3.67
tsp 9.18 7.67
tyan 20.91 17.19
vector-concat 5.56 2.26
vector-rev 4.50 4.55
vliw 6.58 6.23
wc-input1 2.30 2.01
wc-scanStream 3.70 2.24
zebra 2.10 1.77
zern 37.35 37.38
run time ratio
benchmark MLton1
barnes-hut 1.16
checksum 0.99
count-graphs 0.85
DLXSimulator 0.98
fft 1.11
fib 1.01
hamlet 0.88
imp-for 0.55
knuth-bendix 0.90
lexgen 0.87
life 0.79
logic 0.69
mandelbrot 0.89
matrix-multiply 0.54
md5 0.57
merge 1.00
mlyacc 0.96
mpuz 0.91
nucleic 1.18
peek 0.24
psdes-random 0.93
ratio-regions 1.04
ray 1.02
raytrace 1.08
simple 1.01
smith-normal-form 1.00
tailfib 0.67
tak 0.98
tensor 0.51
tsp 0.83
tyan 0.82
vector-concat 0.41
vector-rev 1.01
vliw 0.95
wc-input1 0.87
wc-scanStream 0.60
zebra 0.85
zern 1.00
size
benchmark MLton0 MLton1
barnes-hut 59,793 66,177
checksum 20,917 20,429
count-graphs 40,461 42,869
DLXSimulator 78,237 85,221
fft 29,441 30,897
fib 20,909 20,493
hamlet 945,328 1,333,240
imp-for 21,805 20,445
knuth-bendix 59,710 62,750
lexgen 122,061 147,789
life 38,565 38,693
logic 147,501 79,037
mandelbrot 20,901 20,549
matrix-multiply 21,405 21,117
md5 34,038 30,238
merge 21,885 21,709
mlyacc 409,501 518,909
mpuz 26,645 25,005
nucleic 60,653 60,421
peek 28,542 28,510
psdes-random 21,901 21,341
ratio-regions 41,893 42,653
ray 66,688 79,376
raytrace 159,381 240,061
simple 146,913 186,977
smith-normal-form 141,053 145,925
tailfib 20,637 20,157
tak 20,957 20,613
tensor 62,516 64,684
tsp 33,774 35,334
tyan 77,054 83,502
vector-concat 21,557 21,149
vector-rev 21,389 21,101
vliw 261,417 299,913
wc-input1 39,222 42,294
wc-scanStream 41,614 43,406
zebra 103,502 105,014
zern 26,504 26,592