**UPDATED BENCHMARK** Task: read 1.5 million price bars off disk and total the volume values
Platform: Win 10 Pro Workstation, 2x Xeon X5675 CPUs, 24 gigs memory, SSD @treeform approach - marshalling with flatty and supersnappy libs **Average: 0.14 secs** @cblake approach - using the memfiles lib **Average: 0.06 secs** As someone with a very sketchy understanding of OS internals I'm not clear why the memfiles approach is benchmarking 10x slower on Windoze than cblake was reporting for Unix. I'd appreciate any explanation.
