Services & ResourcesWolfram ForumsMathGroup Archive

[Date Index] [Thread Index] [Author Index]

Re: Re: Mathematica 7.0 slow on OS X

• To: mathgroup at smc.vnet.net
• Subject: [mg93806] Re: [mg93778] Re: Mathematica 7.0 slow on OS X
• From: Devendra Kapadia <dkapadia at wolfram.com>
• Date: Wed, 26 Nov 2008 05:14:00 -0500 (EST)
• References: <ggdqpu$k0$1@smc.vnet.net> <200811251219.HAA03170@smc.vnet.net>

On Tue, 25 Nov 2008, jsk618 at gmail.com wrote:

> On Nov 24, 4:06 am, "Matthias Gottschalk" <mg... at gfz-potsdam.de>
> wrote:
>> I just tried Mathematica 7.0
>>
>> In comparison to Mathematica 6.0.3, 7.0 seems to be by
>> factor of 2 slower under OS X.
>> When I run the "Mathematica Benchmark Test Mathematica 6.0" from
>> Karl UnterKofler I get for
>>
>> Mathematica 6.0.3: Benchmark = 0.97038
>>
>> and for
>>
>> Mathematica 7.0: Benchmark = 0.471266
>>
>> The numbers obtained by Mathematica#s own
>> BenchmarkReport[] are comparable, however: 1.88 vs. 1.91.
>>
>> If 7.0 is really so much slower on a Mac as indicated by
>> the UnterKofler test, then is hardly good news.
>>
>> So, what is your experience?
>>
>> Matt
>>
>> (numbers were generated on a MacBook Pro 2.2 GHz Intel
>> Core 2 Duo, OS X 10.5.5)
>
> I am having similar results with Mathematica 7 on Windows XP (HP
> laptop, AMD Athlon 64 2.2GHz, 1.25 GB RAM)
>
> Version 6.0.1: Benchmark = .649
> Version 7.0.0: Benchmark = .326
>
>   The main culprit is the next-to-last test (Binomial[m, k]) which
> jumps from 25.4s to 376.7s. Since the first 10 tests show an
> improvement and 4 of the last 5 are slower, I suspected a memory
> problem. However, free memory never drops below 300MB during the test,
> and there is essentially no disk activity, so it's not a lack of
> physical memory but probably an issue with Mathematica itself.
>
>   Also, the results with MathematicaMark seem dismal, lower than
> several computers which should be slower. I do not currently have
> Version 6 installed to compare it to, though.
>
> John Keith
>
Hello Matthias and John,

Thank you for reporting the problem with the Sum example used
in the benchmark test.

This example indeed evaluates much slower in Mathematica 7 than
it did in Mathematica 6. The slowdown appears to be caused by an 
autoevaluation of Gamma while simplifying the answer given by
the new algorithm for Sum.

It may be noted that the result given by Mathematica 7 is significantly
smaller in size (ByteCount=416) when compared with that given by 
Mathematica 6 (ByteCount=27264032).

Also, the test example uses an upper limit 10^7 for the summation.
The autoevaluation issue does not affect higher limits such as 10^8.
Thus, the timing for the test example is expected to be similar to that
in the following example (which uses an upper limit of 10^8).

===================

In[1]:= Sum[Binomial[m, k], {k, 0, 10^8}] // Timing // InputForm

Out[1]//InputForm=
{0.372024, 2^m -
   Binomial[m, 100000001]*Hypergeometric2F1[100000001 - m, 1,
     100000002, -1]}

===================

We are working on a fix for the problem with the time taken by
the test example.

Thank you once again for reporting the problem. I apologize
for the confusion caused by it.

Sincerely,

Devendra Kapadia.
Wolfram Research, Inc.

• References:
  • Re: Mathematica 7.0 slow on OS X
    • From: jsk618@gmail.com