John Cupitt has rewritten the original PTM fitter to use parallel tasks on multiple cores and other memory/speed optimisations. It also has a -cache option that uses more RAM to gain speed. WIth today’s 16GB machines that is useful.
Here are some speed tests on 76 x 24 MPixel images on a desktop with an i7-6700 CPU 3.40GHz, 32GB DDR4 RAM and SSD:
1 core: 142s
8 threads: with -cache 99s (showing we need to only use IM_CONCURRENCY=Ncores
4 cores: 51s
4 cores with -cache: 40s