So, there is a strong chance that waking up from hibernation and/or (hybrid) sleep may be slower without the prefetch data. I also read the prefetch data is used during restore from hibernation/sleep (at least in Windows Vista+), and that if the prefetch data is missing, our executables will be paged in from disk instead of from the optimized (already defragmented, and designed to eliminate seeking) hibernate area of the disk. The purpose is to improve performance in situations where running an anti-virus scan or back-up utility would result in otherwise recently-used information being paged out to disk, or disposed from in-memory caches, resulting in lengthy delays when a user comes back to their computer after a period of non-use.' It is based on the 'prefetcher' function in Windows XP. 'SuperFetch is a technology that pre-loads commonly used applications into memory to reduce their load times.

If this is done, Windows will need to re-create all the prefetch files again, thereby slowing down Windows during boot and program starts until the prefetch files are created-unless the prefetcher is disabled.' 'A second myth is that the user should delete the prefetch folder contents to speed up the computer. We should make sure we are not slowing down common cases for other common cases. Once we have a service that nukes prefetch, this code should probably be gated on (firefoxUpdateService || !ReadOperationCount) Is currently gated on an unreliably prefetch-detection(ie it's overly conservative), make sure that codepath is live when prefetch is off in your case. * Prefetch does not use windows readahead.įor a test system I used an acer i7 laptop with a crappy 5400rpm hd, there the differences was 1.6seconds(no prefetch) vs 2.6+ with prefetch. problem is that all this happens before our code runs, so it delays UI showing(.firstPaint) hugely * Crap can be all of the fonts on the system, all of the plugins, etc.all of those often get read during session restore if websites like facebook are in there. * It reads in a bunch of crap before app even runs I recommend using xperf and seeing how much io is attributed to firefox, and how many IOPs it constitutes. This is pretty annoying to test since amount of IO varies.

Old prefetch data