Hi Richard, I hoped I could argue with you on that, but unfortunately I am not an expert on this matter....
I am just reporting what I have experienced (and it is aligned with what was reported by other fellows). Before I used the patch, I got < 10 inserts/sec on some brand new computers. After the pacth (which is only adding FILE_FLAG_WRITE_THROUGH flag to OpenReadWrite and OpenExclusive, and bypassing the winSync), I saw a 10x improvement. Using the base version 3.3.2 with synchronous=OFF was a LOT more faster as I got close to 300 inserts/sec. So I am pretty sure this patch is not equivalent to setting synchronous=OFF, and I am also convinced it really helped in my case. Now, I am not saying that using FILE_FLAG_WRITE_THROUGH is as safe as using the FlushFileBuffers as I do have enough knowledge on the subject, but I have a good gut feeling that, from what Microsoft is advertising (and from reading Tegs comment), we should be very close to it and it does improve performance quite a bit. Nono. --- [EMAIL PROTECTED] wrote: > Nono BEZMan <[EMAIL PROTECTED]> wrote: > > > > My point was that there is light for windows > users, if > > this > > FILE_FLAG_WRITE_THROUGH flag is added to the code. > > Otherwise, it seems that > > FlushFileBuffers really hinders SLQite to a point > it > > becomes unusable. '7 > > Inserts per seconds' on a brand new AMD64 3500+ > with > > SATA drives is extremely > > low, and this could be fixed by using this option, > > IMO. > > > > A quick google of FILE_FLAG_WRITE_THROUGH shows me > that > this option is a no-op for SATA drives. It only > makes > a difference on SCSI disks. So it seems your patch > is really just the equivalent of setting > synchronous=OFF, > at least on your brand new AMD64 with the SATA disk > drives. > > -- > D. Richard Hipp <[EMAIL PROTECTED]> > > __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
