In my application, I need to search the entire data file many number of times. I am testing my program with fread and mmap functions of unix separately for reading or accessing data from the file. When I am accessing the data file more than two times in a program , mmap is taking less time than fread but when I am accessing the data file two times or only one time, mmap is taking more time than fread. what are the reasons for this
I think there are two separate parts.
(1) Why mmap becomes faster than fread after accessing data files for more than one time?
Because the fread has system call overhead each time you issue the request.
Please see my reply in https://www.researchgate.net/post/What_is_the_actual_difference_between_fread_and_mmap_functions_in_linux_unix?ev=tp_feed_post_xview
(2) Why mmap is slower than fread during the first-time accessing?
(I am not pretty sure about it, but following two possible reasons are my GUESSES).
(a) From system call interface, fread triggers file system component which internally might have more pre-fetching mechanism than mmap.
Note that the mmap does not map or fetch the disk content when you issue the mmap() call. So when you first access the memory content, page fault happens and the disk content will be fetched and populated to page cache via page fault handler.
(b) The page fault handler is more expensive than regular fread operation
Bests,
Jui-Hao
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Software
- System Software
- Operating Systems