3 ms·
For awhile, CISC architectures (basically, just "architectures" before RISC came around) kept accumulating features. I took a VAX assembly language ("VAX Macro"
by stcredzero 10y ago
For awhile, CISC architectures (basically, just "architectures" before RISC came around) kept accumulating features. I took a VAX assembly language ("VAX Macro") class, and I remember that there was a string compare command in VAX Macro that was about as full featured as what you'd expect in some kind of standard library.
So I'm not surprised to learn that there's something in an Intel architecture like this: Intel is just showing its CISC roots. Since it's associated with SSE, there's probably something advantageously parallel about it.
- pcwalton 10y agoIt's really more about SIMD than about CISC or RISC. The pcmpxstrx family of instructions aren't like classical CISC string-compare instructions (though note that those are there too—"repe cmps"). Rather, they take 16-byte registers and perform operations (find first byte matching X, etc) on the bytes in them. I imagine that they can take advantage of effectively map-reduce style techniques for this (which is what you would do in software too if you were using simpler SIMD instructions). Note that pcmpxstrx isn't really that fast in many cases compared to just doing ordinary SIMD operations.
- stcredzero 10y agoThe string compare operations built into old CISC architectures also weren't that fast. It's more like they were designed as bullet points for the salesmen. (Who were talking to execs who remember managing people who wrote things in assembly.)