On Thu, 28 Feb 2013 06:23:45 -0500, deadalnix <deadal...@gmail.com> wrote:
On Thursday, 28 February 2013 at 10:57:29 UTC, Walter Bright wrote:
Notice the double read of *p. This is what SentinelInputRange is trying
to fix.
But it will be optimized away.
I have to say I agree with deadalnix.
you have essentially in lexer.c:
outer:
while(1)
{
switch(r.front)
{
case 0:
break outer;
...
}
}
whereas a range where empty is defined as r.front == 0:
while(!r.empty) // inlined to r.front != 0
{
switch(r.front) // why would another load occur here?
{
// no need to check for 0, already done
...
}
}
If this doesn't translate to the same code, I don't know why not. The
second implementation looks more elegant/straightforward. I think others
have pointed out that other compilers already do this. I hope we aren't
introducing library components to get around DMD optimizer shortcomings.
Note, r cannot be a D string in order to achieve the desired efficiency.
With a sentinel range, length is never checked, and should not have to be
updated.
I think it's not necessary to add extra isSentinelRange, etc. checks, just
define your sentinel range like any other input range.
-Steve
