Apache OpenOffice (AOO) Bugzilla – Issue 87022
Why OOBasic is so slow compared to MS VisualBasic (300x)
Last modified: 2017-05-20 11:35:19 UTC
[b]I tested on CPU Core 2 Quad Q6600 @ 3,2GHz[/b] [u] [code] Sub Main Dim x as Variant Dim y as Variant Dim i as Variant x = 1 y = 1 t! = Timer For i = 1 To 1000000000 x = x * y Next Msgbox (Timer - t!) End Sub [/code] [b]MS VB = 24,03 seconds OO Basic = 3424 seconds[/b] 140 times more slow [code]Sub Main Dim x as Long Dim y as Long Dim i as Long x = 1 y = 1 t! = Timer For i = 1 To 1000000000 x = x * y Next Msgbox (Timer - t!) End Sub[/code][/u] [b]MS VB = 11,42 seconds OO Basic = 3427 seconds[/b] 300 times more slow Who can explain this
This is a well known but difficult to fix problem as it's based on the general architecture of OOo Basic that obviously hasn't been designed for good performance. So to really fix this a complete redesign would be necessary. This is difficult because of limited resources and because of the high risk of breaking the compatibility somehow. Anyway this of course is no P1 at all according to the priority rules. -> P3, OOo later as this problem unfortunately has no real chance to be addressed in the near future.
I have an ideea At first to make the second interpretator SuperBasic - starting with little functionality but very optimized for speed. Results to recive througth functions from second command interpretator to the first. By time the second command interpretator will be more functional enought functional and at this time to use them as first
Andreas, as always we have a choice of slowly bleeding (300 times slower is actually dying) or having one-time surgery and then living happy life. Making surgery will be more and more difficult as the number of developed applications and number of users grows. I strongly suggest targeting this for 3.x just to have it in sight and in mind. WBR, KP.
ab->amdx4: Unfortunately I don't really understand your idea. You want to mix different interpreters at runtime?! If yes, do you really expect this could ever work? I don't. If no, I don't see which of the problems I've mentioned above would be solved by this. Of course if a new system was de- veloped this would be done besides the old first, but this is only develo- pement tactics and no solution for the real problems. ab->kpalagin: It's not as easy. Nobody can guarantee that the happy life status will ever be reached. It could easily end up in endlessly fixing compatibility problems. Believe me, it's hard enough to make minor chan- ges to the parser - as I'm just trying at the moment - without breaking anything. Rewriting the whole system including all the little bugs old programs now may depend on would be much worse. Besides this it's also a question of priority. The effort for rewriting the complete OOo Basic system would be huge. I guess, it could easily re- quire 12 months or more. First, I simply don't have this time. Second I doubt that such an effort is justified for one single issue. The OOoBasic performance is weak, but obviously it's not too weak for the most cases. The real life does not consist out of benchmarks and nobody uses OOo Ba- sic for number crunching. Usually it's used to handle dialog events and access the OOo API that often isn't too fast either. In most cases it does not matter if a Button handler needs 0.000001 or 0.001 seconds to be executed. In case a user really needs performance it will be better anyway to create an extension, e.g. using Java. So if we put it on 3.x it wouldn't change much as it surely would be set to 4.x some day. Anyone who sees different priorities here is free of course to start a SuperBasic project. Just don't expect me to do the work then... :-)
Reset assigne to the default "issues@openoffice.apache.org".