macOS Eclipse build takes 2x as long as Windows

[chelsel]

I switched to Mac from Windows about four months ago and except for Leopard it's been pretty smooth. However, I've recently setup Eclipse on OS X and find that all of the builds take a lot longer to complete... for example, one of my projects takes 20 seconds on the Mac vs. 7 seconds on Windows.

The Windows machine is actually not even spec'd as well as the Mac... the Windows machine is only a 2GB 2.16GHz Core Duo whereas the Mac is a MBP 4GB 2.4GHZ Core 2 Duo. Any ideas? I've already increased the memory for Eclipse (heap settings in eclipse.ini) to 768M which is 3x the heap given to the Windows implementation.

 

[wrldwzrd89]

Do you have the same version of the Java VM installed on both? If you're using Java 6 on Windows, and Java 5 on Mac (since Java 6 isn't available for Mac users yet), that could make a significant difference due to optimizations in the Java 6 compiler that aren't in the Java 5 compiler. Also, are you using the same version of Eclipse on both computers?

 

[chelsel]

Java 5 on both.

 

[wrldwzrd89]

Interesting. I've never built an Eclipse project on Windows, though I have it set up. In fact, I could run some tests for you using my Bootcamp partition running Windows XP, to take differing hardware out of the equation, and make the test results more meaningful, in an attempt to see if I experience the same thing.

 

[wrldwzrd89]

Okay, testing completed.

I'm using an application I made myself, called Maze Runner, as the test Java application. Here's the data:

Mac OS X 10.5:
Uncached launch: 10 seconds
Cached launch: 2 seconds

Windows:
Uncached launch: 15 seconds
Cached launch: 2 seconds

It appears that Windows is actually a little slower than the Mac OS, on the same hardware, with default Eclipse configurations on each platform.

By uncached launch I mean launching for the first time, after opening Eclipse. Cached launch means any subsequent launches after the initial rebuild completes.

 

[chelsel]

Thanks for the info. I'll have to do the test on the exact same hardware to be fair. Not sure what the problem could be though.