I created some confusion in my special report on IBM's POWER5 processor. My story alluded to Windows for the POWER/PowerPC architecture, prompting many readers to send e-mail like this: I am a regular reader of your articles in InfoWorld and just finished reading the fascinating article on IBM's Power 5 processor. I saw mentioned a couple times in the article that Power 5 supports the Windows operating system. W I created some confusion in my special report on IBM’s POWER5 processor. My story alluded to Windows for the POWER/PowerPC architecture, prompting many readers to send e-mail like this: I am a regular reader of your articles in InfoWorld and just finished reading the fascinating article on IBM’s Power 5 processor. I saw mentioned a couple times in the article that Power 5 supports the Windows operating system. Wow, is this new? I knew Power 5 supported the IBM operating systems such as AIX and OS/400 (aka i5), as well as Linux, but the Windows support was news to me. I could not find the Power 5 Windows support mentioned on IBM’s nor Microsoft’s websites. Is this a pending announcement? The confusion here arose from my failure to make clear what I meant when I asserted that there is a Windows for PowerPC. Let me explain. Nearly all of the work to make applications “64-bit clean”–primarily, removing hard-coded assumptions about the size of an integer (32 bits on x86, 64 bits on Opteron and most everything else) or a memory pointer–is not done for a specific CPU architecture. Once the code compiles error-free and runs natively on any 64-bit architecture, it can be retargeted to another architecture with a change in compiler back end. This isn’t exactly a snap. I think that developing an efficient compiler and bulletproof debugger would be the hardest part of making Windows run on anything but x86. Retargeting the kernel, which has to wrestle with differences in CPU-level greasy bits like memory management and I/O, is no cakewalk, either. But as NetBSD, and later, Linux proved, if you do multiple platform support properly up front, porting to a new architecture later is much easier. Did Microsoft do multi-platform Windows right? Well, Windows is no NetBSD, but there are developers in Redmond paid to explore the worlds beyond Dell. I’m certain that Windows for PowerPC is running now within Microsoft, but not discussed. Microsoft has a rich reference library of working PowerPC code: Windows NT 4.0, the unreleased finished 64-bit editions of Windows 2000 and Microsoft’s own C++ compilers. When 64-bit support was pulled from Windows, developers could no longer target non-x86 platforms. Now the multi-targeted compiler back end has returned in Visual Studio 2005; one click (re)compiles your project for x86, x86-64 or Itanium 2. Old-timers may recall that Microsoft’s compilers once listed Macintosh among its targets. My certainty of the existence of a 64-bit Windows for PowerPC and POWER is by no means a prediction that such a release will ever see daylight. The volume isn’t there, and customers of the volume leader, Apple Computer, aren’t exactly clamoring to see the Windows flag raised over their iMacs and Xserves. But I do have some predictions about Microsoft and Apple in an upcoming column. ——– Technology Industry