Re: Different string format options, benefits?

In article 91-10-061 coxs2@rpi.edu (Sean C. Cox) writes:
>I am curious about the benefits/costs related to the two general character
>string formats [null terminated vs. counted]


Certainly using the length string format is better on any computer
with any type of concurrency. A test for zero termination creates
a loop dependency, whereas most string operations can be vectorized/
concurrentized if the length is known beforehand. Even if the machine
only has a data prefetch mechanism, one must know that fetching the value
of the next byte will not cause a core dump. Even if the machine is
purely sequential, knowing the length beforehand allows one to
perform word-size data fetches, instead of byte-by-byte (you can do
this and check for zero-termination, too, but the code is a lot
uglier).


bb
--

Return to the comp.compilers page.
Search the comp.compilers archives again.