How about the one instruction "RunInst a, b, c, d", whose behavior is defined as "Do the instruction described by parameter a in [your favorite architecture], upon parameters b, c, d"?
One instruction set computer
11–20 of 27 posts
Re: One instruction set computer
#12I always thought MADD (Multiply + Add) made a great candidate for a single instruction set computer, if the program counter was usable as an operand to the instruction.
You will want a way of decreasing registers, hence the subtract parts of the single instruction sets.
Re: One instruction set computer
#13I'm curious, is there any real-world application for this? Or is it mostly interesting from a research perspective? It seems like this wouldn't perform very well in the real world for most applications, but I could see it being useful for very simple architectures.
My final year project was a hardware implementation of this. My group made an 8 core coprocessor and an OpenCL-like implementation complete with an LLVM IR to SUBLEQ translator. We actually got performance better than the proprietary NIOS II e soft processor on our image processing demo (Y colorspace conversion).
Re: One instruction set computer
#14How about the one instruction "RunInst a, b, c, d", whose behavior is defined as "Do the instruction described by parameter a in [your favorite architecture], upon parameters b, c, d"?
I think you just reinvented assemblerprogramming.
Re: One instruction set computer
#15How about the one instruction "RunInst a, b, c, d", whose behavior is defined as "Do the instruction described by parameter a in [your favorite architecture], upon parameters b, c, d"?
Yeah, the notion of "a single instruction" really doesn't mean a whole lot, just like how (in many languages) "a single line of code" doesn't mean much.
Re: One instruction set computer
#16How about the one instruction "RunInst a, b, c, d", whose behavior is defined as "Do the instruction described by parameter a in [your favorite architecture], upon parameters b, c, d"?
Yeah, the notion of "a single instruction" really doesn't mean a whole lot, just like how (in many languages) "a single line of code" doesn't mean much.
Well, a long line of code could hold a lot of meaning.
Re: One instruction set computer
#17How about the simplest Turing machine?
http://en.wikipedia.org/wiki/Wolfram%27s_2-state_3-symbol_Tu...
Re: One instruction set computer
#18I'm curious, is there any real-world application for this? Or is it mostly interesting from a research perspective? It seems like this wouldn't perform very well in the real world for most applications, but I could see it being useful for very simple architectures.
Well it could be interesting as a test architecture for a new logic family as long as you have wide memory.
The Manchester Baby was the first programmable electronic computer and was close to a single instruction computer. It can emulate the single instruction computer with 5 of its 7 instructions: load, subtract, store, skip if negative and jump. The two extra instructions were halt and relative jump.
http://en.wikipedia.org/wiki/Manchester_Small-Scale_Experime...
Re: One instruction set computer
#19Here's a single-instruction computer built out of the page translation behavior of X86 processors: http://events.ccc.de/congress/2012/Fahrplan/events/5265.en.h...
If programming with just one instruction seems too low-level, I have just the solution for you! Now you can compile the user-friendly, high-level language of Brainfuck to this single-instruction machine!
http://gergo.erdi.hu/blog/2013-10-05-yo_dawg_i_heard_you_lik...
Re: One instruction set computer
#20I always thought MADD (Multiply + Add) made a great candidate for a single instruction set computer, if the program counter was usable as an operand to the instruction.
You will want a way of decreasing registers, hence the subtract parts of the single instruction sets.
If you've got fixed length registers you can subtract via addition or multiplication.