An embedded DSP produced by Texas Instruments. It comes with many peripherals (SDRAM controllers, multichannel serial port controllers, etc.) on board, as well as a limited amount of memory that can be configured as either L2 cache or fast program memory.
The most remarkable thing about this DSP to an assembly programmer is that all eight logic units are programmed independently (this is referred to as a VLIW architecture), allowing a programmer to tune code to near 100% cpu utilization, without breaking his or her head; thus, despite the fact that my sample TMS320C6211 ran at only 150Mhz, it easily outperformed its slow-as-molasses external memory bus.
and that is why holistic system design is important.
(to be fair, the 'C6211 was probably meant to be used on algorithms with smaller working sets and higher computational complexity than the ones I was toying with.)