Huge speed improvements possible:
1. by using other datastructures libraries
1.1. using Numpy datastructures (huge speed improvement)
1.2 using CyToolz, even faster than Numpy
1.3. using Databases instead of Datastructures would also improve speed of portage emerge hugely
2. by not using brut force to solve dependencies, but that may be nothing new for you.
3. by make use of many cores of cpu's
3.1. maybe by using gpu cores, but that is maybe more insane than rational
Steps to Reproduce:
emerge --search ....
emerge --searchdesc ....
to wait is even not neccessary
How have you actually benchmarked these changes? Show your work. Patches?
FWIW, Portage with the already existing sqlite backend doesn't show any appreciable speedup. How do you explain that?
We would welcome patches to improve performance, but they are not going to write themselves. I would suggest sending some patches to the gentoo-portage-dev mailing list, along with some benchmarks to demonstrate their effect.