kicad-developers team mailing list archive

[Simon Richter <Simon.Richter@xxxxxxxxxx>]

Hi,

On 31.01.19 16:01, John Beard wrote:

> I'm not really sure. I suppose it depends on the benefits it brings (and
> in future with shared common libs, perhaps it won't matter as much?)

I doubt it makes much of a difference — the biggest hot spots in compile
time are:

 - waiting for pcbnew_wrap.cxx to be generated before we even start
compiling pcbnew objects
 - cmake's dependency handling, which uses a separate dependency scan
rather than generating them on the fly like automake does
 - compiling pcbnew_wrap.cxx (this is started fairly late, and takes the
longest)
 - linking pcbnew three times in parallel
 - linking common/pcbcommon statically because of incompatible base units

We could probably shave off a few seconds here or there by minimizing
includes or using abstract interfaces, but that is probably not worth
pursuing on its own, but might be interesting as part of unit tests.

TBH, I'm not even sure the kiface mechanism really buys us anything but
the option to start individual modules from the command line. I'd expect
a separate 3d-viewer to contain another copy of common/pcbcommon, so
it'd slow down linking and use more memory.

The biggest savings we could probably get from unified units. I'd think
64 bit nanometers would be the way to go, even if that will write a lot
of leading zeroes to memory, saving the extra code copies will
definitely offset that.

   Simon

Attachment: signature.asc
Description: OpenPGP digital signature