If you compile from source MyPaint can take advantage of some advanced cpu features, and multiple cores to some extent. I still think that single-thread performance is the best benchmark for MyPaint if you are looking at upgrading your computer.
Generally the bottlenecks depend on brush settings. Drawing too many dabs, using dabs that are not round, dabs that are too big, or too small, and especially using smudge settings which samples the canvas, too.
If you compile libmypaint you could try settings CFLAGS before compiling:
This is what I use for my cpu, this should automatically use your “native” cpu features:
export CFLAGS=’-fopenmp -O3 -ftree-vectorize -march=native -mtune=native -funsafe-math-optimizations -funsafe-loop-optimizations’
and when I run configure I enable openmp:
./configure --enable-openmp
