"-O" enables optimizations
Replaces movs with less movs??
X86 mov is turing complete: mov-only compiler
41–50 of 60 posts
Re: X86 mov is turing complete: mov-only compiler
#42"-O" enables optimizations Replaces movs with less movs??
Fewer.
Re: X86 mov is turing complete: mov-only compiler
#43When I read the title I thought "ah of course cmov is turing complete", but no, it's only mov apparently.
Which is quite amazing (and probably dog slow)
Re: X86 mov is turing complete: mov-only compiler
#44Extremely Reduced Instruction Set Computer.
Or is it a Ridiculous Instruction Set ?
Re: X86 mov is turing complete: mov-only compiler
#45Extremely Reduced Instruction Set Computer. Or is it a Ridiculous Instruction Set ?
One instruction set computer.
Re: X86 mov is turing complete: mov-only compiler
#46Earlier quoted context omitted.
> HN doesn't allow me to escape * with \, which the markdown spec says I should be able to do So? What does the markdown spec have to do with... anything? The XML spec says you have to begin your comment with a version number declaration, but you're not doing that.
I thought it was trying to implement markdown by italicizing half my sentence when I put * in. Markdown was what popped into my mind when I saw that behavior. It would be nice if I could find a way of escaping them without having to insert spaces where I don't want spaces.
There isn't one, HN doesn't use markdown. See https://news.ycombinator.com/formatdoc
Re: X86 mov is turing complete: mov-only compiler
#47This is almost surely a dumb question, but what would happen if someone made a processor that only executed mov instructions, which would be extremely simple, and presumably would be really goddamn fast. Would that be competitive with today's fastest CPUs? If it were, what would be the advantages and disadvantages of this design?
what would happen if someone made a processor that only executed mov instructions, which would be extremely simple, and presumably would be really goddamn fast.
The laws of physics get in the way. It's essentially the same reason why clock frequencies have stopped going up.
Re: X86 mov is turing complete: mov-only compiler
#48Earlier quoted context omitted.
Not exactly, most instructions are not "microcoded" for performance reasons. But there is a class of instructions that is.
Intel CPUs translate x86 to some internal RISC-like "uops" before doing optimizations and execution. This is a widely documented fact (though Intel doesn't really talk publicly about it as far as I know): see for example §2.1 in http://www.agner.org/optimize/microarchitecture.pdf
For a (hopefully!) more accessible introduction to the crazy world of what really happens inside the x86, I did a talk at GOTO: http://youtu.be/hgcNM-6wr34 - it's a really fascinating subject!
Re: X86 mov is turing complete: mov-only compiler
#49Earlier quoted context omitted.
Not exactly, most instructions are not "microcoded" for performance reasons. But there is a class of instructions that is.
Intel CPUs translate x86 to some internal RISC-like "uops" before doing optimizations and execution. This is a widely documented fact (though Intel doesn't really talk publicly about it as far as I know): see for example §2.1 in http://www.agner.org/optimize/microarchitecture.pdf
For a (hopefully!) more accessible introduction to the crazy world of what really happens inside the x86, I did a talk at GOTO: http://youtu.be/hgcNM-6wr34 - it's a really fascinating subject!
Re: X86 mov is turing complete: mov-only compiler
#50Earlier quoted context omitted.
I thought it was trying to implement markdown by italicizing half my sentence when I put * in. Markdown was what popped into my mind when I saw that behavior. It would be nice if I could find a way of escaping them without having to insert spaces where I don't want spaces.
There isn't one, HN doesn't use markdown. See https://news.ycombinator.com/formatdoc
It should use markdown, though.