MobileRead Forums - View Single Post - call for testers from source: dramatic performance improvement

chaley · 12-14-2010, 02:01 PM

The trunk release contains changes that improve performance of getting information for the tag pane by a factor of 6. In addition, startup performance has been improved a fair amount. For example, on my win7 machine with my 20,000 book test library, the time to populate the tags pane has dropped from 3 seconds to .5 seconds. The time to start has been cut in half, from 19 seconds to 9 seconds. Kovid is seeing very similar numbers on his Linux machine.

The way the work is done is now totally different, and it is possible that I broke something. It would be extremely helpful if those of you who run from source could try this code. Feedback on performance improvement would be good, but at this point I am more interested in correctness.

Note: this source also contains code that doesn't update the tags pane at all if it isn't visible (the .5 seconds becomes 0 seconds). Don't confuse this change with the general improvement.

Caveat: tags containing commas will no longer display on the tags pane. There shouldn't be any such beasts, but there was a time many months ago when they could sneak in. To make them display, edit the book. Hit OK to split the tag at the comma, or change the comma to a semicolon. I am looking at doing the comma->semicolon change automatically at the next release.

12-14-2010, 02:01 PM	#1
chaley Grand Sorcerer Posts: 12,336 Karma: 8012652 Join Date: Jan 2010 Location: Notts, England Device: Kobo Libra 2	call for testers from source: dramatic performance improvement The trunk release contains changes that improve performance of getting information for the tag pane by a factor of 6. In addition, startup performance has been improved a fair amount. For example, on my win7 machine with my 20,000 book test library, the time to populate the tags pane has dropped from 3 seconds to .5 seconds. The time to start has been cut in half, from 19 seconds to 9 seconds. Kovid is seeing very similar numbers on his Linux machine. The way the work is done is now totally different, and it is possible that I broke something. It would be extremely helpful if those of you who run from source could try this code. Feedback on performance improvement would be good, but at this point I am more interested in correctness. Note: this source also contains code that doesn't update the tags pane at all if it isn't visible (the .5 seconds becomes 0 seconds). Don't confuse this change with the general improvement. Caveat: tags containing commas will no longer display on the tags pane. There shouldn't be any such beasts, but there was a time many months ago when they could sneak in. To make them display, edit the book. Hit OK to split the tag at the comma, or change the comma to a semicolon. I am looking at doing the comma->semicolon change automatically at the next release.