Re: CISC vs RISC concepts -- from an assembly view

CISC CPU speed is relied on microprocessor optimizations (like 3DNow!
and MMX, SSE2, SSE3 extensions/optimizations) which are enabled by the
compiler.

SIMD has nothing to do with CISC - vector operations can also be used
in a RISC design

So if this is true, shouldn't everyone be using RISC processors and
just use really smart compilers to create their executable.(...)
Am I thinking clearly and realistically?

RISC is good if you can compile and profile your code on a certain
machine directly - at best, using real work data for profiling.
However, if you want to support a large variety of hardware with
different internal architectures as it is the case for most x86
machines in use, the abstraction offered by CISC is better as it
allows the CPU to optimize the processing in a way which suits its
internals better than a simple fixed and highly architecture-dependant
RISC machine.

Anyway, many processors are using a mixture of both: If you take a
look at todays CISC CPUs, you´ll notice that the fastest and most
common instructions are pretty much the same as could be found in RISC
machines. On the other hand, typical RISC style architectures offer
more and more instructions, the more complex one often not implemented
in hardware but as some sort of microcode macros.

.

Relevant Pages

  • RISC binary fragility (was Re: Code density and performance?)
    ... > My belief is that RISC, with a compiler optimized for the specific ... It is the requirement that the compiler optimize ... I suspect that typical CISC instructions only contain the ... performance for binary performance compatibility. ...
    (comp.arch)
  • Re: New itaniums out at 2.5x perform gain
    ... EPIC vs RISC vs CISC does nothing to the end user. ... compilers have to be written and how chip engineers have to work on the ... "pure" EPIC relies on the compiler to predict the outcome of the ...
    (comp.os.vms)
  • Re: RISC binary fragility (was Re: Code density and performance?)
    ... is faster than a comparable CISC. ... This might have been strengthened by the fact that RISCs were successful against CISCs only where applications could be relatively easily ported (i.e., absolute binary compatibility was not necessary). ... I have the impression that there was significant optimism about the practicality of application portability (e.g., there seems to have been something like an idea that any reasonably large company could produce and successfully market its own RISC architecture). ... but they tend to assume that the compiler will do it. ...
    (comp.arch)
  • Re: Code density and performance?
    ... "RISC" architectures have incorporated more "CISC" functions over the years as needed to do useful work. ... , not having a barrel shifter might have made the difference between a buildable design and an unbuildable design. ...
    (comp.arch)
  • Re: Code density and performance?
    ... "RISC" architectures have incorporated more "CISC" functions over the years as needed to do useful work. ... When you have only two or three metal layers, not having a barrel shifter might have made the difference between a buildable design and an unbuildable design. ...
    (comp.arch)