Got startup time down to 5 seconds by completely eliminating the use of lispreader and putting of loading the xml, set and zip namespaces. There's no clear bottlenecks left to fix.
On Fri, Feb 6, 2009 at 1:06 AM, Remco van 't Veer <rwvtv...@gmail.com> wrote: > I've managed to get to startup time down from 12 seconds to 7 on the > emulator and hope get it below 4 seconds tomorrow. A lot of time is > spend in LispReader when the core constants are initialized. I've > altered the compiler to emit type specific code for the basic cases, > thus bypassing LispReader. Need to learn more about GeneratorAdapter, > and need sleep, before I can speedup the loading of, for instance, seq > contants. > > My changes are available at github; http://github.com/remvee/clojure > > Remco > > > On Wed, Feb 4, 2009 at 5:55 PM, Remco van 't Veer <rwvtv...@gmail.com> wrote: >> Having a lisp to do android app development is definately worth some efford. >> Kawa already has some android support in svn trunk but it doesn't seem in a >> usable state; my experiments break because a lot of expressions do not seem >> to compile AOT. The parts that do compile are very fast on the other hand. >> >> But if it comes down to app development I would of course prefer clojure >> over kawa's schema and scheme over java.. >> >> I hope to do some tracing to find out what slowing down the startup later >> this week and will keep the list in formed. If anybody has any suggestions >> please yes pretty please. >> >> Remco >> >> On Feb 4, 2009 3:59 PM, "igorrumiha" <igorrum...@gmail.com> wrote: >> >> On Feb 4, 10:12 am, "Remco van 't Veer" <rwvtv...@gmail.com> wrote: > >> Caching Reflector.getMethods,... >> >> I would also be interested in running Clojure on the Android. I tried >> running a "Hello world" application on the emulator but the startup >> time od 15 seconds put me off. There is definitely room for >> improvement. I don't have much experience in Java (zero would be the >> best description) so I don't really know where to start with >> profiling. >> >> I've run the application through the Instrumentation mechanism and the >> trace shows me that much of the time is spent in a huge amount of >> calls to some character compare methods (something like 200000 calls) >> (I don't have the trace with me right now so I don't have the exact >> numbers and I don't know the exact method name). >> >> I find this curious when I know that all code is compiled ahead of >> time. When I get the chance I will recreate that call graph an upload >> it to the files section. >> >> -- >> IgorR >> >> >> > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en -~----------~----~----~----~------~----~------~--~---
