This morning at Embedded World ARM announced the new Cortex M4 core:
The key feature of the M4 is significantly improved DSP capabilities. The M4 can execute a 32x32+64=64 MAC instruction in a single cycle (This takes 7 cycles on a M3, and 3 cycles in the closest competitor (AVR32)), and also implements SIMD instructions including a (16x16)+(16x16)+64=64, also in a single cycle.
This will greatly improve the already power signal processing capabilities if the M4 over the M3 in demanding applications like motor control and pro audio.
The M4 can also be implemented with an optional FPU to ad floating point capability. The FPU is single precision. It is my expectation that most usage of the M4 will be without the FPU since it will add a significant gate count to the core, but no doubt there will be lines of high end devices with floating point.
Publicly announced licensees include Texas Instruments, ST, and NXP so stay tuned to your suppliers for specific device plans.
I have a short presentation on M4 for interested parties, contact me if interested. There is also an article on M4 coming out on the March issue of Technology Review
I expect to have a “Microcontroller Technology Spotlight” available on the Cotrex-M4 core shortly after 3/2 (waiting on some more public announcements) which you can forward to target customers as you see fit.