Re: Incremental Compilers

Some poeple would say just buy faster machines.... But to be helpful,
some more data about the project is needed. For example, I'd imagine
your project must be a large project consisting of many many source
files. Some possible sources of slow down then are:


1) Too many source files, especially with many sub-directories may
make the overhead of "making" large. You can get this data just by
timing a no change make.


2) Lots of object files mean slower link time. If you are not using
dynmically linked libraries, then at least for development purpose,
use them.


3) Likewise, turn off optimizations for development compiles. Note
that this practice gets frowned upon, depeneding on who you talk to
(arguments against ranging from, you must debug what you ship to if
there are any bugs exposed/introduced by the optimizers, the
developers should see it first).


4) Are you using C++? Are you using templates? In the last big company
I worked for, switching from C to C++ and using templates added
substantially (10X?) to the build time. Needless to say we had a talk
to the C++ engineers next aisle. Template instantiations are slow, and
some methods are slower than others. Be careful.


5) Get a faster compiler. GCC is known to be slow. What host machine
are you using? You may be able to get a faster compiler. In
particular, LCC is known to be much faster but generates less
efficient code than GCC and is C only.


rinap@cisco.com wrote:
> I am an engineer at Cisco and every engineer spends about 10 minutes
> to compile an image even if only few of the files in the code base
> were modified. I am wondering why that time cannot (in theory) be
> reduced from 10 minutes to 10 seconds by using incremental compilation
> AND LINKING.
>
> ...


--
// richard
http://www.imagecraft.com

Return to the comp.compilers page.
Search the comp.compilers archives again.