On Monday, 27 January 2014 at 18:30:43 UTC, Pierre Talbot wrote:
On Monday, 27 January 2014 at 04:07:04 UTC, Andrei Alexandrescu
wrote:
On 1/26/14 3:22 AM, Pierre Talbot wrote:
Hi,
I was wondering why CTFE is context sensitive, why don't we
check
every expressions and run the CTFE if it applies?
Compilation would get awfully slow (and sometimes won't
terminate).
Andrei
So it is theoretically possible? I mean if the compilation
doesn't terminate, the execution won't either for at least one
program input, so we can detect an infinite loop at
compile-time. Moreover, isn't the same problem with
context-sensitive CTFE?
Pierre
Just because code runs for a long time doesn't mean that it will
run forever. This is known as the Halting problem [1]. We could
have configurable switches to halt CTFE if it runs for more than
N seconds or allocates more than N MiB.
However, this still doesn't take away from the point that it
would make compilation awfully slow with presumably little
benefit. There are further issues in that CTFE isn't exactly
stable yet (I've had it produce incorrect results in a few
cases), so relying on it to produce correct, optimized code is a
bad idea.
I would be in favor of some "lite-CTFE" for the sake of
optimization. Simply halting if the interpreter encounters a loop
or function recursion should make it pretty snappy. But I think
many compilers / optimizers already do this to some extent.
[1] https://en.wikipedia.org/wiki/Halting_problem
