NEE with a single light source. There's still some bias since subsequent paths don't apply MIS weights if the indirect sample hits the light source... and probably other sources of error as well. )

Getting closer to feature parity with the linux port. It's going to be very fun.. to find and fix all those hacks/oversights.

WIP vulkan + linux backend for my path tracer. Fully bindless. HLSL to SPIR-V with DXC.