Mac OS 10.5 - Multicore usage

cellprofiler

#1

Hello,

I have been testing CellProfiler on my MacBook with OS 10.5 installed. It seems that only one core is being used for the analysis (i.e. thread = 1 during initialization). I was wondering if it was possible to set CellProfiler to use both cores and perhaps have more than 1 independent thread.

Thanks very much,
Ulric


#2

Hello,
Matlab doesn’t automatically do so. If you would like to have it run faster, you could try running two instances of CellProfiler at once(for the compiled version).

~martha


#3

Thanks for the answer Martha.

I have been thinking about purchasing a new computer to perform image analysis. I have a series of TIF files (usually over 300 images) approximately 2.5MB in size and I have 2 channels for each image. What would configuration would you recommend I purchase?

I know that a cluster would be better, but this computer set up is for my small apartment, not my lab.

Thanks so much for your help,
Ulric


#4

Hello,
We have successfully run the SBS image set on our Mac Pro (with 8GB of RAM). Do you have 300 image sets (so 600 in total) or are you have 150 image sets (300 images total)? This would assume one image per channel and both channels make up one image set. It seems that the more RAM the better, but of course, there are cost-limiting issues. 1GB of RAM (minimum) is required to run CellProfiler properly with a very limited number of image sets. However, for larger sets, you can set up the pipeline using a small test set (usually a mix of image sets from the beginning, middle, and end of your acquisition.). Once you have the pipeline adjusted, you can run it on all of your images. Just be sure that you change your preferences to ‘Do not display windows’, so you will receive a MEMORY error.

Please let me know if there are any other questions that I can answer for you!


#5

Thanks so much Martha, I really appreciate the help :smile:

I am sorry I was so vague, your first guess was correct, 300 sets of images, with 2 channels per set and one image per channel is my typical image set. I was considering a Mac Pro with 8 GB of RAM, but was unsure if CellProfiler could take advantage of the Mac Pro’s 4 or 8 core set up. Also, I was not sure if a Mac or a PC tower would be better for CellProfiler analysis. Have you had more success with one set up or the other?

Thanks again for all the help Martha.

Regards,
Ulric


#6

Hello,
Currently, we have both PC and Macs available in the Imaging Platform. I am usually a PC-person, but found that since Macs can handle more memory, it tends to run better on very high resolution images. Really, though, if you are saving images as a lossless format (PNG or TIF, NOT JPG) at a reasonable size, you shouldn’t have any issues, nor should you need a full 8GB of RAM. Since you will be running these cycles locally and not on a cluster, you may want to have the 4GB of RAM. There is a chance that even if you have ~96 or more image sets with several figure windows open, CellProfiler will crash. Once you have your pipeline completely setup, I would run the entire cycle without any figure windows displayed. You can modify this setting in your Preferences (File >>Preferences).
~martha