Community
Participate
Working Groups
Bug 42196 dealt with underlying JDT performance. According to that bug, now closed, the issue has been resolved for the general JDT case. Since this still lags on a Ti/667 with 3.0M4, it seems likely that the remaining speed issues are SWT issues. 3.0M4 feels faster than previous versions at this task, but it still brings up the Spinning Wait Cursor on a Ti/667. I believe that sub-second response times are a reasonable target for a machine of this class, at least in the long term. The java gaming initiative has made JOGL a front runner for code development. Because this is a direct port of OpenGL, the GL class has a great number of methods and static final constants in it. These cause the method popup in the editor to show up very, very slowly. import net.java.games.jogl.*; import net.java.games.jogl.util.*; public class Test { public void display(GLDrawable gLDrawable) { final GL gl = gLDrawable.getGL(); gl.glClear(GL.GL_COLOR_BUFFER_BIT | GL.GL_DEPTH_BUFFER_BIT); } } Start typing a duplicate of the second line of display into the function. On my TiBook 667 with 10.2.6 running Eclipse 3.0M3, I see the following: Typing gl. pegs the cpu for 17 seconds when typed the first time, 9, 8, 9, 7, 8, and 9 seconds for later invocations. (Timing by eye, so these all seem to be about 8 seconds.) Typing GL. pegs the cpu for 5 seconds reliably. There does not seem to be a cache effect going on. A build of JOGL for MacOS X 10.2 can be found at <http://homepage.mac.com/fuz>. This tgz file contains the needed jar file to reproduce the problem.
Changed priority to critical, as it makes this feature almost unusable when working with jogl. It is not a blocker because the automatic method popup can be turned off.
Mac performance issue, Silenio to investigate. Question for Scott in the meantime: have you observed the performance of this on other platforms?
Unfortunately, I do not have easy access to other platforms.
A friend with a 1G windows machine, so about half again faster by raw CPU, and faster/slower depending on your MHz position, says that any lag is under a half second, as he does not see it. Scott
This might be related to bug #42264 and #42368
Table/Tree widgets performance were improved quite a bit since this bug was entered. The code assist popup comes up in less than a second for me (accept for the first time, probably because of some caching in JDT). Please the latest.