Carsten Koch wrote:
I tried your patch today and watched the available memory (using top) when replaying. The number stayed the same. Excellent!
Thanks a lot.
However, when I did FF or FR, the number decreased. Ralf, can you reproduce the problem?
With FF there are no decreases here - but with FR. I didn't notice that before (why ever). But I'm happy you found that. When jumping backward I missed to clear the read ahead buffer. Interestingly it doesn't help to clear that buffer - it may have to do with the man page description for POSIX_FADV_WILLNEED: "start a non blocking read". It seems this read may or may not be finished when I try to say that I don't need the previously asked read ahead buffer anymore. So as a work around I dropped the POSIX_FADV_WILLNEED call. This also helped for the noad patch.
I will try to release a new patch today in the afternoon/evening.
Regards Ralf