Thread overview
[Issue 17736] bigint opunary should be better performing
August 10
https://issues.dlang.org/show_bug.cgi?id=17736

hsteoh@quickfur.ath.cx changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hsteoh@quickfur.ath.cx

--- Comment #1 from hsteoh@quickfur.ath.cx ---
I notice also that binary operators on BigInt (even the opOpAssign ones like +=) will create new instances of BigInt rather than update in-place.

One trouble with updating in-place is that it makes BigInt assignment either expensive (always copy) or exhibit reference semantics:

---
BigInt x = 1;
BigInt y = x;
++y;
writeln(x); // will this print 1 or 2?
---

If I understand the BigInt design correctly, reference semantics are *not* desirable because we want BigInt to be more-or-less a drop-in replacement of fixed-size ints, and lots of code will break in subtle ways if the by-value semantics was substituted with by-reference semantics.

One thought is that if BigInt uses some sort of reference-counting scheme, then if the refcount is 1 we can update in-place, otherwise allocate a new BigInt to hold the result as usual.

--
August 10
https://issues.dlang.org/show_bug.cgi?id=17736

Steven Schveighoffer <schveiguy@yahoo.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |INVALID

--- Comment #2 from Steven Schveighoffer <schveiguy@yahoo.com> ---
Oh, so assignment just rebinds to the existing data? Then this request is invalid.

One thing we could do is make a MutableBigInt, that is allowed to modify itself. But that's a much bigger project.

--
August 10
https://issues.dlang.org/show_bug.cgi?id=17736

--- Comment #3 from hsteoh@quickfur.ath.cx ---
IMO, the refcounting idea is still valid, if a bit more complicated to implement. It would be important for reducing GC load on BigInt-heavy code, I think.

--
August 10
https://issues.dlang.org/show_bug.cgi?id=17736

--- Comment #4 from Steven Schveighoffer <schveiguy@yahoo.com> ---
Either way, this is not a "simple" enhancement. But feel free to take over this enhancement request if you want to write it up.

--
August 11
https://issues.dlang.org/show_bug.cgi?id=17736

--- Comment #5 from hsteoh@quickfur.ath.cx ---
Wrote it up here:

https://issues.dlang.org/show_bug.cgi?id=17746

Doesn't necessarily mean I have the time to actually implement this, though. :-)

--
August 11
https://issues.dlang.org/show_bug.cgi?id=17736

hsteoh@quickfur.ath.cx changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://issues.dlang.org/sh
                   |                            |ow_bug.cgi?id=17746

--