Re: Mathematica v6: Slower in the following fields
- To: mathgroup at smc.vnet.net
- Subject: [mg75806] Re: Mathematica v6: Slower in the following fields
- From: "alexxx.magni at gmail.com" <alexxx.magni at gmail.com>
- Date: Fri, 11 May 2007 05:22:11 -0400 (EDT)
- References: <200705090828.EAA13415@smc.vnet.net><f1uojp$87t$1@smc.vnet.net>
Hi everybody,
thank you for the useful info.
To get a more correct comparison, I'm giving you a comparison 6.0 vs.
5.2 under Linux, using, for the Mathematica processes running:
vsz virtual memory size of
the process in KiB (1024-byte units).
rss resident set size, the non-swapped
physical memory that a task has used (in kiloBytes).
For 5.2:
---------------------------------------------------------------------
(VSZ) ( RSS)
8172 24240 Mathematica
12312 44264 MathKernel
(dopo apertura Help):
9304 25184 Mathematica
12312 44264 MathKernel
In[1]:=img=Table[Random[Integer],{400},{249}];
11704 27500 Mathematica
16756 48532 MathKernel
In[2]:=ListDensityPlot[img,Mesh->False,AspectRatio->Automatic]
11580 27304 Mathematica
16952 48624 MathKernel
... so we went from ~ (8/24 + 12/44) MB to (11/27 + 17/49) MB for
(Mathematica + MathKernel).
The new situation is instead:
For 6.0:
---------------------------------------------------------------------
36920 146980 Mathematica
14412 149840 MathKernel
29992 481012 java
(dopo apertura Help):
40388 150152 Mathematica
31216 166456 MathKernel
38948 509044 java
In[1]:= img = RandomInteger[1, {400, 249}];
42468 152148 Mathematica
31648 166848 MathKernel
39400 509044 java
In[2]:= ListDensityPlot[img, Mesh -> False, AspectRatio -> Automatic]
58144 169816 Mathematica
282284 418064 MathKernel
39400 509044 java
Therefore (in MB):
5.2
6.0
VSZ RSS VSZ
RSS
--------------------------------------------------------------------------------------------------------------
start 8 24
37 147 [Mathematica]
12 44
14 150 [MathKernel]
30 48 [java]
finish 11 27
58 170 [Mathematica]
17 49
280 418 [MathKernel]
39 509 [java]
--------------------------------------------------------------------------------------------------------------
Now I see why I had to shut it down after a few more calculations! The
frontend is relatively stable, while both the kernel and java (java?
Why?) exploded in memory usage.
I still have to try the additional option you suggested, I'll let you
know - but something is clearly wrong here, considering the small size
of the object I'm trying to plot...
thanks!
Alessandro
P.S.
Even worse problems I had using ListContourPlot, which I cannot
quantify since it slows my machine to a crawl forcing me to kill the
process (LCPlot using a more regular image, of course)
Chris Chiasson ha scritto:
> Are you manually setting PlotPoints in 6.0? In the new version,
> PlotPoints *seems* to mean something different than it did in 5.2
> (i.e. lots more subdivisions per plotpoint).
>
> On 5/9/07, alexxx.magni at gmail.com <alexxx.magni at gmail.com> wrote:
> > I installed v6, but had to return to 5.2 since - for some works I was
> > doing - the new version has become unacceptably slow. Now I changed my
> > mind and I keep both versions on, depending on what I have to do.
> >
> > I'd like to start this thread to see people post their observations
> > about what has become slower in v6 - and of course what is faster too!
> >
> > e.g. some posts ago I saw somebody having problems with eigenvalues
> > calculations.
> >
> > My problems, just to start, relate to graphics. Dealing with large
> > images with ListContourPlot, for example, is really slower (had to
> > kill the calculation). And ListDensityPlot is barely acceptable.
> >
> > So... please, add your experience here!
> >
> > Alessandro Magni
> >
> >
> >
>
>
> --
> http://chris.chiasson.name/
- References:
- Mathematica v6: Slower in the following fields
- From: "alexxx.magni@gmail.com" <alexxx.magni@gmail.com>
- Mathematica v6: Slower in the following fields