Content-type: multipart/related; boundary="Boundary_(ID_BcXkQFfS2yz670tdAL4z+g)"; type="text/html" --Boundary_(ID_BcXkQFfS2yz670tdAL4z+g) Content-type: text/html; CHARSET=US-ASCII Content-transfer-encoding: quoted-printable

On Apr 09, 2013, at 01:46 PM, Joshua Stults <joshua.stults@gma=> wrote:

I am trying to speed up export of a regi= on to stl format, and I noticed in g-stl.c that there is a hard-coded ncpu= =3D1 [1].  Does changing this to 2 or 4 speed up the stl conversion (= if there are 2 or 4 cores available)?
If all it took to speed it up was changing that to= a bigger number, we would have done that.. ;)

It looks like = there is a command line option to set ncpu [2], but lower down in the code= it is hard-coded to 1 [3].

The exporter is not currently set up to perform a conver= sion in parallel.  Someone started working on infrastructure to suppo= rt that many years ago, but the work was never completed.

Depending on your use case, one faster = way to obtain an STL conversion is via the bot_dump command in mged.  = ;That will export a mesh without first validating that it's solid.  O= f course, you still need to convert to a BoT mesh first via the facetize c= ommand, and I suspect that's more likely where you're currently spending t= he majority of your time.

Ano= ther option would be to perform the conversion in parallel at a scripting = level, running multiple g-stl at the same time (on different objects).

Se= an

= --Boundary_(ID_BcXkQFfS2yz670tdAL4z+g)--