Please stop using the Flex Compiler Shell. Compilation speed could be increased a lot when using a different technique. Let me explain this in a little bit more detail. Say, we have modules A, B, C and D.

A depends on C
B depends on C
C depends on D

So this means we have a possible topsort of {D, C, B, A} but we can also have a two-dimensional sort like D}, {C}, {B, A. In that case we know which modules can be compiled in parallel. So we can compile D, then C, and then A and B in parallel. Basically the number of compiler processes should be limited of course since those "parallel projects" can grow large and you can have 50 of them.

Now that we can compile projects in parallel it would be an ideal option to stop using FCSH and spawn only a process of compc or mxmlc for each individual module. The cool thing is that FCSH is absolutely not needed when using the "-incremental" flag which will store a *.cache file for each compilation target next to the output file.

Why is this a better approach? If you have a really big project with 100+ modules the FCSH will run out of memory every time which makes compiling the project in our case a very long process. First FCSH will grow to some ridicolous amount of memory like 1GB and then crash, grow again to 1GB and crash ... The problem is that FCSH becomes slower the more memory it consumes. The time is ridicolous. It takes at the moment approx 10min to compile the whole project. With a simple application making use of this technique it takes about 1 minute – and it scales very well with the amount of cores.


Best,

Joa