I'm not sure about Linus's objections, but I've found that DMA accelerators for time sharing systems with general workloads haven't reaped benefits, as the over

I'm not sure about Linus's objections, but I've found that DMA accelerators for time sharing systems with general workloads haven't reaped benefits, as the overhead of multiplexing and synchronizing with them kills most of their benefits. At that point it's easier to blit memory yourself.