On Tue, Mar 24, 2009 at 4:59 PM, John Tamplin <[email protected]> wrote:
> On Tue, Mar 24, 2009 at 3:59 PM, Vitali Lovich <[email protected]> wrote: > >> I think it might be a good idea to also reference the -localWorkers flag. >> I find that it cuts down on my startup time significantly (especially on my >> desktop which is quad-core, but even on my dual-core laptop). With OOPHM I >> find this to be particularly more annoying because the browser freezes on >> startup while the server is compiling the code. >> > > I don't know that the OOPHM document is the right place to discuss other > compiler flags. Also, I don't think -localWorkers will have any impact on > hosted mode at all, since it is used to compile different permutations in > parallel. it will speedup web-mode compilation if you have more than one > permutation, but nothing for hosted mode. > Sorry my mistake. It was probably just the placebo effect :D. However, it does take about 11 seconds between when the OOPHM connection is established (request for module in log) & when the onModuleLoad starts executing (8 seconds between Request for module & Refreshing module from source). My IDE takes about 1 second to compile my project cleanly (same with ant). On my laptop which takes 4 seconds to build using ant, takes 28 seconds to go from Request for module to Refreshing module from source & then takes 7 seconds to compile. When I run HostedMode using the release configuration, then the times drop to 1/7 for the desktop, & 1/25 for the laptop (compilation time/module request time). Also, this happens on every launch regardless of whether or not code was actually changed. > > > >> Also, perhaps a mention that every refresh of the page launches a fresh >> compilation of the code & thus the browser locks up for a while there as >> well. >> > > It is doing a bytecode compile (including of generated code), not a > web-mode compile, so it should be very long unless you have a really large > app. There is a project we call instant hosted mode which will allow reuse > of your IDE's bytecode compilation rather than having to recompile, but it > isn't ready yet (soon though). > Sorry, my mistake. Refreshing isn't actually slow. But if it's doing a bytecode compile, why is it even taking any time on the initial startup when there was no source code change & the compiled classes haven't changed? Shouldn't it notice this & not do anything? > > > >> The other thing I noticed is that after changing client-side code & >> refreshing, although the debugger claims the source is out of sync, it works >> perfectly with the changed code. > > > What IDE are you using? I know there are issues with hot-swap between JDT > and Javac, so if you aren't using Eclipse you may have some hot-swap issues > (Eclipse and GWT both use JDT), which allows you to change some classes > without doing a refresh. That is independent of a refresh, which is > basically tossing the running classes and reloading them. Using Eclipse. The IDE warns me that hot swapping failed on save. The debugger claims the source is out of sync, but like I said, it actually hits the breakpoints correctly & executes the changed code. I can't reproduce it right now (I can't remember if it was a persistent problem and it was fixed by me updating the trunk today or if it was an intermittent problem). If it happens again, I'll let you know. > > > -- > John A. Tamplin > Software Engineer (GWT), Google > > > > --~--~---------~--~----~------------~-------~--~----~ http://groups.google.com/group/Google-Web-Toolkit-Contributors -~----------~----~----~----~------~----~------~--~---
