Sorry for bumping a month-old thread, but I was searching for this:
more time is spent building and precompiling the system image (which
cannot be sped up with -j N) than actually compiling and linking its C
sources.
Can you explain why building the system image can't be parallelized? Could
Theoretically, yes. The tedious-but-possible theoretical is to have the
following things, plus independent pre-compilation of packages; then the
minimal base can be compiled first and all the internal packages
compiled independently and linked. Big wins would come from precompiling
linalg and
On Sunday, February 2, 2014 9:29:09 PM UTC-6, Jeff Pickhardt wrote:
Finally, when I compiled it, there were a ton of warnings. Stuff like:
unused variables, incompatible pointer types, etc.
You'll find that, as with the warnings you pasted here, these almost always
come from Julia's