Well over a year ago Engadget posted a description of the (suspected) processor in the iPhone  - ARM1176JZF (ARM11 family, implementing the ARMv6 instruction set). It has a VFP11 vector coprocessor for SIMD operations. It does not support NEON as some have suggested.

In my attempt to figure out how to access the SIMD operations for matrix and vector math, I’ve found these reference PDFs that might be of use to others:

• The ARM Architecture Reference Manual
• The ARM1176JZF-S Technical Reference Manual
• The Vector Floating Point Instruction Set Quick Reference Card

There’s also a (pay-for) article on the IEEE Xplore website, called “Accelerating floating-point 3D graphics for vector microprocessors”, which talks specifically about ARM VFP processors.

It may also be worth keeping an eye on this math library, which forms part of the Oolong Engine.