[icedchai]

I messed around with that setting on one of my Macs. I wanted to load a large LLM model and it needed more than 75% of shared memory.