Hackaday readers (and writers) are an odd bunch. While the rest of the tech press falls over for the newest, shiniest CPU on the market, we’re the type who’s more interested the unexplored dark corners of metaphorical Silicon Alley. So when someone comes to us with a good writeup of a chip that we’d never heard about, we’re all ears.
[Remy]’s writeup of the CoolRISC 816 microcontroller CPU makes it obvious that he shares our taste for the esoteric. It has a 22-bit “RISC” instruction set. It has a dedicated 8-to-16 bit multiplier. Some of the instructions are so un-reduced that [Remy] calls bunk on its RISC claims. All of the operations, including the un-RISC ones, run in a single cycle. And the CoolRISC does this by cheating — the last stages of the pipeline run not on every clock tick, but on the rising and falling flanks of the clock respectively.
Why all these odd bits? They make the job of the assembly programmer, or compiler designer, a lot easier. With all single-cycle instructions, counting cycles is the same as counting lines of code. The not-really-RISC instructions are great for compiling C into. So what happened? [Remy] speculates that the MSP430, another not-really-RISC microcontroller that came out about the same time, ate the CoolRISC’s lunch. The MSP430 is a 16-bit machine, and chances are good that you’ve heard of TI. The same may not be true of Xemics, maker of the CoolRISC.
But still it’s nice to have someone saying the eulogy for this strange little chip. Or maybe the reports of the CR816’s death are premature — it seems to be inside TI’s bq20x80 chip that’s used in a number of battery power monitors. Oh, the irony! Indeed, watch [Charlie Miller] tear into a battery and find a CR816.
Have any of you used a CR816? What’s the strangest microcontroller architecture that you’ve ever seen?
Filed under: Microcontrollers
No comments:
Post a Comment