[XSL-LIST Mailing List Archive Home] [By Thread] [By Date]

Re: [xsl] Is xsl:for-each "syntactic sugar"?

---

---

> the width of your CPU limiting the number of iterations.

I think, the number of iterations is bound by the max-bound (on a
given hardware & compiler implementation) of the type used for the
iterating variable.

> Also, xsl:for-each is much easier to understand than a potentially
> complex recursive function (you SEE that it is a loop, while you have
> NOTICE that it is recursive).

very true :)

> Also, xsl:for-each can (in theory) be implemented using parallel execution (multiple cores)
> while recursion necessarily must be a single core as it is up to the coder where / when in
> his particular algorithm the recursive call takes place.

your statement, that recursion cannot be parallelized, is true to a
large extent. But IMO, this cannot be generalized to every possible
situation.

I agree, that many "for loops" are potential candidates of
parallelization. I think parallelization support for "for loops" could
be available at following levels:
a) a given end-user program could try to do this (if the host language
provides support for parallel execution, say something like language
threads).
b) a language compiler could provide some support. IMO, since all "for
loops" cannot be parallelized, compilers could provide parallelization
support for specific scenarios. I think, the extent to which this
benefit could be provided by a compiler, is dependent on a
corresponding optimizer in a compiler.
c) there could be some support for parallelization at hardware level.
IMO, it's generally quite difficult (or impossible?) for compilers, to
exploit hardware-level parallelization benefits, if the compilers are
written in platform independent languages.

> Totally wrong, as it is much more complex to read and understand as a simple loop.

Some algorithms can be expressed naturally (and sometimes can only
be), when written recursively.


-- 
Regards,
Mukul Gandhi

---