Test performance impact (was: The dreaded M-R)
Simon Marlow
simonmar at microsoft.com
Wed Feb 1 07:32:55 EST 2006
On 01 February 2006 11:42, Nils Anders Danielsson wrote:
> However, to stand on more solid ground I suggest that someone runs
> some performance tests, with and without
> -fno-monomorphism-restriction, to see whether the M-R has any great
> impact in practice. There are some performance test suites based on
> real code out there, right? Nofib?
Results below (ignore runtime differences, I was averaging over 5 runs
but most of these benchmarks run too quickly to get any reasonable
measurements).
Summary: 2 programs failed to compile due to type errors (anna, gg).
One program did 19% more allocation, a few other programs increased
allocation very slightly (<2%).
------------------------------------------------------------------------
--------
Program Size Allocs Runtime
------------------------------------------------------------------------
--------
anna ----- ----- -----
ansi +0.00% +0.00% 0.00
atom +0.00% +0.00% -2.92%
awards +0.00% +0.00% 0.00
banner +0.00% +0.00% 0.00
bernouilli -0.23% +0.00% -7.44%
boyer +0.00% +0.00% 0.00
boyer2 +0.00% +0.00% 0.01
bspt +0.00% +0.00% 0.02
cacheprof +0.14% +0.00% -9.86%
calendar +0.00% +0.00% 0.00
cichelli +0.00% +0.00% +0.99%
circsim +0.00% +0.00% -0.93%
clausify +0.00% +0.00% 0.01
comp_lab_zift +0.00% +0.00% +1.86%
compress +0.00% +0.00% -5.58%
compress2 +0.00% +0.00% -7.63%
constraints +0.00% +0.00% -9.03%
cryptarithm1 +0.00% +0.00% -6.27%
cryptarithm2 +0.00% +0.00% 0.03
cse +0.00% +0.00% 0.00
eliza +0.00% +0.00% 0.00
event +0.00% +0.00% -7.04%
exp3_8 +0.00% +0.00% -10.84%
expert +0.00% +0.00% 0.00
fem +0.35% +1.76% 0.06
fft +0.00% +0.00% 0.08
fft2 +0.00% +0.00% 0.15
fibheaps +0.00% +0.00% 0.08
fish +0.00% +0.00% 0.03
fluid +0.12% +0.04% 0.02
fulsom -0.14% +0.00% -2.10%
gamteb +2.47% +0.56% 0.15
gcd +0.00% +0.00% -5.88%
gen_regexps +0.00% +0.00% 0.00
genfft +0.00% +0.00% 0.04
gg ----- ----- -----
grep +0.00% +0.00% 0.00
hidden +0.00% +0.00% -12.48%
hpg +0.00% +0.00% -7.92%
ida +0.00% +0.00% -12.50%
infer +0.00% +0.00% 0.13
integer +0.00% +0.00% -9.81%
integrate +0.00% +0.00% -5.47%
knights +0.19% +0.00% 0.01
lcss +0.00% +0.00% -7.42%
life +0.75% +0.00% -16.22%
lift +0.00% +0.00% 0.00
listcompr +0.00% +0.00% -9.76%
listcopy +0.00% +0.00% -8.33%
maillist +0.00% +0.03% 0.06
mandel +0.00% +0.00% +0.00%
mandel2 +0.00% +0.00% 0.02
minimax +0.00% +0.00% 0.01
mkhprog +0.00% +0.00% 0.00
multiplier +0.00% +0.00% -10.43%
nucleic2 +0.61% (stdout) (stdout)
para +0.00% +0.00% -2.51%
paraffins +0.00% +0.00% -10.07%
parser +0.00% +0.00% 0.06
parstof +0.00% +0.00% 0.01
pic +0.28% +19.27% 0.02
power +0.29% +0.00% -11.62%
pretty +0.00% +0.00% 0.00
primes +0.00% +0.00% 0.13
primetest +0.00% +0.00% +0.77%
prolog +0.00% +0.00% 0.00
puzzle +0.00% +0.00% -5.03%
queens +0.00% +0.00% 0.06
reptile +0.00% +0.00% 0.03
rewrite +0.00% +0.00% 0.00
rfib +0.00% +0.00% 0.04
rsa +0.00% +0.00% 0.06
scc +0.00% +0.00% 0.00
sched +0.00% +0.00% 0.05
scs +0.54% +0.06% -11.14%
simple +1.70% +0.00% -4.68%
solid +0.00% +0.00% -2.03%
sorting +0.00% +0.00% 0.00
sphere +0.06% +0.00% -1.74%
symalg +0.00% +0.00% 0.13
tak +0.00% +0.00% 0.02
transform +0.00% +0.00% -7.73%
treejoin +0.00% +0.00% -2.70%
typecheck +0.00% +0.00% -1.29%
veritas +0.12% +0.00% 0.00
wang +0.00% +0.00% -18.99%
wave4main +0.00% +0.00% -9.04%
x2n1 +0.00% +0.00% -2.84%
------------------------------------------------------------------------
--------
Min -0.23% +0.00% -18.99%
Max +2.47% +19.27% +1.86%
Geometric Mean +0.08% +0.23% -6.71%
More information about the Haskell-prime
mailing list