Understanding MATLAB file access
1 view (last 30 days)
I'm working on a script where I have to read hundreds of images and do some calculations on them. I'm using a simple for loop with imread and I noticed something quite strange. The first time that I run the script, the imread function takes a time longer than the image analysis itself, but if I clear the memory and I run it again, the imread function is incredibly fast!
How is it possible that the second time is so much faster? I clear the memory before running the script and the images must be loaded again from the hard disk, so the bottleneck should be the disk read speed in both the cases. Is there any "hidden" memory that matlab uses to store files that I have already loaded? Or is it a matter of windows prefetch stuff?
It would be useful to know what is happening to speed up the code in advance.
Jan on 2 Mar 2016
Edited: Jan on 2 Mar 2016
Hard disks store last accessed data in a fast cache, which has a size of about 64 or 128 MB. So reading the file a 2nd time is usually much faster.
The operating system stores data temporarily in the RAM also, which depends on the available RAM. Under Windows 7 the size can be influenced in the registry:
Session Manager\Memory Management -> "LargeSystemCache"
You can disable the caching of the OS in the device manager also in the settings for the hard disk. But even then the caching of the hard disk is working.