I agree with your point, but the "benchmark" has always included
generating the matrix and that is typically a very small part of the
time and should be relatively stable (although I suppose inverting the
same "random" matrix over and over would remove some variation). Your
suggestion of using a left argument for 6!:2 is the best way to reduce
(or at lease smooth out) variability.
In repeating these expressions on my iPad, I discovered something I was
never aware of before... I accidentally tacked a & onto the end of the
expression - that has a very interesting (and actually useful) result.
But because it is a pain to email selections from the iPad, these were
done on an iMac i7 -
100 (6!:2) '%. 50 50 ?@$ 1000'
0.00080055
100 (6!:2) '%. 50 50 ?@$ 1000'&
0.000807989999999999967&
0.1
0.1
0.1&
0.100000000000000006&
0j40 ": 0.1
0.1000000000000000055511151231257827021182
I never realized that tacking on the & displays the number which might
be useful to see the actual floating point representation. May be about
as silly as inverting random matrices, but also amusing...
On 2012/09/12 14:51 , Devon McCormick wrote:
Part of the reason for the variability is that you're generating a new
matrix each time and including the generation in your timing.
Something like this should give a more stable result:
6!:2 '%.mat' [ mat=. 50 50?.@$1000
Also, this form better allows you to run multiple timings to get a
more stable number, e.g.:
(10) 6!:2 '%.mat'
On Mon, Sep 10, 2012 at 3:29 PM, Robert Cyr <robert....@gmail.com> wrote:
Results of the posted benchmark vary quie a bit: on this Nexus 7,
6!:2'%.50 50 ?.@$1000'
0.043118
6!:2'%.50 50 ?.@$1000'
0.033788
6!:2'%.50 50 ?.@$1000'
0.030006
6!:2'%.50 50 ?.@$1000'
0.034632
6!:2'%.50 50 ?.@$1000'
0.023798
And with a larger sample,
6!:2'%.500 500 ?.@$1000'
7.91767
6!:2'%.500 500 ?.@$1000'
8.07115
6!:2'%.500 500 ?.@$1000'
7.94837
About twice the time of the iPad
On Sep 9, 2012 6:53 AM, "bill lam" <bbill....@gmail.com> wrote:
From google, iphone 4 uses cortex-a8 and that cpu uses vfp-lite which
is 10 times slower than a regular vfp used in cortex-a9.
Сбт, 08 Сен 2012, Joey K Tuttle писал(а):
That's a nice speedup! It will be interesting to try the iPhone 5.
I'm curious if anyone has a timing from a generation 3 iPad, it may
well be faster than my iPad 2.
On 2012/09/08 06:30 , J. Patrick Harrington wrote:
The iPhone 4S is faster: 5.6 sec
On Fri, 7 Sep 2012, Joey K Tuttle wrote:
iPhone 4 - 18.2 seconds
iPad 2 - 4.6 seconds
On 2012/09/07 19:48 , bill lam wrote:
I am curious to know what are the timings for iphone and ipad.
It is around 6 to 11 seconds on android depending on CPU.
Срд, 05 Сен 2012, Paul Jackson писал(а):
Some time ago, you and Roger were talking about timings on
%. 500 500 ?@$ 1000
I believe Roger said he had timings from the IPSA days. He also
said the
matix was considerably smaller. Did you ever get those early machine
timings?
Paul
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
