Re: creating Graphics using ParallelTable[]
- To: mathgroup at smc.vnet.net
- Subject: [mg99871] Re: creating Graphics using ParallelTable[]
- From: Jens-Peer Kuska <kuska at informatik.uni-leipzig.de>
- Date: Mon, 18 May 2009 02:29:34 -0400 (EDT)
- References: <gum0ar$q33$1@smc.vnet.net>
Hi,
frame[t_?NumericQ] :=
Plot3D[Sin[x*y + t], {x, 0, Pi}, {y, 0, Pi}, PlotPoints -> 128]
DistributeDefinitions[frame]
movieframes = ParallelTable[frame[x], {x, 0, 16, 1}];
ListAnimate[movieframes]
gives a speedup of 3.26 in the Parallel Kernel Status window on
my Quad core
May be it help you post the full code and not only fragments.
Regards
Jens
Jan-Philip Gehrcke wrote:
> Hello all,
>
> I generate "video frames" from time-consuming 3DPlot[]s (to export
> them into a video file later on):
>
> frame[x_] := Plot3D[... something big using x ...];
> movieframes = Table[frame[x], {x, start, end, (end-start)/steps}];
>
> I have a double core CPU; so now I would like to create these frames
> in parallel with Mathematica 7, using ParallelTable[]. But I don't
> derive any advantage from doing this:
>
> My Windows taskmanager shows three "MathKernel.exe". When I use
> ParallelTable[] for the described problem, only one "MathKernel.exe"
> is working, causing 50% CPU load and using much RAM. The other two
> "MathKernel.exe"s don't cause any load and almost don't use any RAM.
>
> $ProcessorCount is 2, there is 1 "master" and 2 "local" in "Parallel
> Kernel Status". The "parallelizeabletest" ParallelTable[$KernelID,
> {10}] succeeds.
>
> I tried `DistributeDefinitions[frame];` before invoking ParallelTable
> [], but it did not change anything.
>
> One core needs about 20 seconds to create one single frame without
> displaying it: An extensive analytical function (among others there
> are nested Coth[]s) has to be calculated with Plotpoints->100 option.
>
> In my opinion - simply expressed - each core can take one `x` out of
> the queue and create the corresponding 3DPlot Graphics object, while
> the other core is doing the same. This should work, because the tasks
> are totally independent and my way to use the Table[] is the least
> complex one. I don't see the "data management overhead" that often
> reduces or even prevents advantages from parallelizing, because each
> core just needs to get the function definitions and a simple number:
> `x` - no more overhead.
>
> (At which point) do I think wrong or does Mathematica work weird (less
> likely..)?
>
> Is there a way for me to create these frames using all my CPU power?
>
> Sincereley,
>
> Jan-Philip Gehrcke
>