Re:Why iPhone Xs performance on JavaScript is so good
by olliej on Monday October 08, 2018 @04:03AM
(#18165188)
The traditional (eg mips) definition of risc was that the cpu should not hide how it operated - in essence, the compiler was required to schedule operations efficiently itself for both performance and correctness. That’s why MIPS and others have things like branch delay slots. It turns out that that is a terrible idea - it means binaries are tied to a specific micro architecture, let alone a separate implementation of the same ISA. Traditional risc also did not support floating point or integer division (both in software), except again software is necessarily slower than a hardware pipeline.
So risc has slowly become analogous to “more registers, orthogonal instruction set, not x86”
Re:Why iPhone Xs performance on JavaScript is so good
by floatboth on Monday October 08, 2018 @04:17AM
(#18165290)
> Did someone say ARM is RISC?
Load/store are still always separate instructions, and instruction length is fixed — that's RISC enough. (But the distinction is not super meaningful these days anyway.)
Re:Why iPhone Xs performance on JavaScript is so good
by userbinator on Sunday October 07, 2018 @11:35PM
(#18163961)
Indeed, it could be said that ARM not being pure RISC is what keeps it competitive with x86. I wonder how long it'll be before x86 also gets a similar set of instructions...
Re:Why iPhone Xs performance on JavaScript is so good
by Dylan16807 on Sunday October 07, 2018 @11:52PM
(#18164016)
The instruction is just specifying particular semantics, that doesn't really disqualify it from being RISC. The instruction doesn't do more than the baseline variant.
[1]: http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc....
Did someone say ARM is RISC?