Experiment 19: The back story of Microsoft's work to bring NT core to ARM

A new Microsoft Research page acknowledges efforts dating back to 2008 to move the Windows NT core to ARM, which ended up impacting both Windows RT and Windows Phone 8.
Written by Mary Jo Foley, Senior Contributing Editor

Back in 2010, I wrote a bit about a Microsoft Research mobile project called Menlo. Part of that project involved a skunkworks team effort known as Experiment 19.


At that time, my sources told me that "Menlo" was a hush-hush research project which seems to be focused on replacing Windows CE/Windows Embedded Compact with Windows NT inside of mobile devices. "Experiment 19," my contacts said, was a graphics platform of some kind that researchers were building to complement Menlo.
As "Felix9" noted over in the Microsoft Channel 9 Coffee House forum on July 13, the Experiment 19 team has finally gone public. And members of that team are now admitting to have helped "prove... that Windows NT and the CLR (Common Language Runtime) could deliver better performance than Windows CE and the .NET Compact Framework on identical hardware."
"Within months of the completion of Experiment 19, Microsoft launched efforts to build what would become Windows Phone 8 and Windows RT for ARM tablets," according to the updated Experiment 19 profile on Microsoft Research site.
This was a bigger and more encompassing project than I had heard. More from the team:
"We started from a Windows core (called MinWin) and a port of the Windows NT kernel to the ARM processor. Working closely with MinWin pioneers—Adam Glass, Mark Russinovich, Richard Pletcher, Richard Neves and Bryce Cogswell—and with partners at NVIDIA (My aside: there's the graphics platform piece of the project), we created the device drivers and firmware necessary to boot and run MinWin on our prototype phones. We created an ARM JIT (just-in-time) compiler for the CLR and ported the CLR runtime to ARM (Another aside by me: Redhawk-related?)  To complete the system, we ported the phone implementation of Silverlight to run with our ARM implementation of the CLR."
The Experiment 19 team is echoing the message of the Windows Phone team, which claimed Microsoft decided to replace the Compact Embedded guts of Windows Phone with the NT core because the NT core would work better on the new multicore processors available today and in the future.
I've heard some Windows Embedded backers disagree with this contention, but that's what the official party line is here. And as @DrPizza, a k a Peter Bright with Ars Technica, notes, low-end Windows Phone hardware may not be Windows NT-kernel-compatible. There has been talk that Windows Phone OS 7.8, built on the existing Embedded Compact kernel, will continue to be the only OS available on lower-end Windows Phones. Microsoft officials are not commenting on that rumor. (I've asked.)

If you want to check out the Experiment 19 team's demo of Windows Embedd Compact vs. Windows NT core performance on phone prototypes, there's a video clip (dating back to 2009) that's also available for viewing as of this week. The new Menlo page on the Microsoft Research site also mentions Little Rock, a low-power sensing project inside MSR.  Blogger Manan Kakkar recently unearthed lots of  Little Rock details.

Update: As one of my readers privately reminded me, the timing on this Experiment 19 work also feeds in with reports of Microsoft's work to port Vista to ARM, in a project codenamed LongARM.

Editorial standards