I have four targets for the linux version of my application:
linux32_debug
linux32_release
linux64_debug
linux64_release
I build all four, then run linux32_debug and watch output being printed onto the console, which is the average number of machine cycles per vertex required to transform the vertex from local-to-world coordinates.
Then I stop execution, switch "build target" to linux32_release, run again, and watch similar output.
Then I stop execution, switch "build target" to linux64_debug, run again, and watch similar output.
Then I stop execution, switch "build target" to linux64_release, run again, and watch similar output.
This is wrong. The 64-bit versions of these routines are much faster. That can easily be made evident by then selecting "build clean" and "build build", then running again.
Presto-chango... the output shows 35% improvement in speed.
The problem may be one of my not understanding how codeblocks is supposed to work. I was thinking that switching "build target" would change which executes when I click the red or green "run" arrows on the GUI. But I don't see any other obvious way to switch which build I execute.
Please explain. Is this a codeblocks bug, or am I the bug?