Tuesday, November 8, 2011

On variadic templates

GCC developers, February 2008:
We have implemented the complete specification of variadic templates in the GNU C++ compiler [8], which are available in GCC 4.3. The implementation itself was relatively straightforward in GCC, implying that this feature can be implemented in other C++ compilers without architectural changes. Our basic implementation approach involved adding flags to each kind of template parameter and each function parameter stating whether these parameters are in fact parameter packs.
...
Microsoft guys,  September 2011:
We've developed a new scheme for simulating variadic templates.  Previously in VC9 SP1 and VC10, we repeatedly included subheaders with macros defined differently each time, in order to stamp out overloads for 0, 1, 2, 3, etc. arguments.  (For example, included the internal subheader repeatedly, in order to stamp out make_shared(args, args, args).)  In VC11, the subheaders are gone.  Now we define variadic templates themselves as macros (with lots of backslash-continuations), then expand them with master macros.  This internal implementation change has some user-visible effects.  First, the code is more maintainable, easier to use (adding subheaders was a fair amount of work), and slightly less hideously unreadable.  This is what allowed us to easily implement variadic emplacement, and should make it easier to squash bugs in the future.  Second, it's harder to step into with the debugger (sorry!).  Third, pair's pair(piecewise_construct_t, tuple, tuple) constructor had "interesting" effects.  This requires N^2 overloads (if we support up to 10-tuples, that means 121 overloads, since empty tuples count here too).  We initially observed that this (spamming out so many pair-tuple overloads, plus all of the emplacement overloads) consumed a massive amount of memory during compilation, so as a workaround we reduced infinity.  In VC9 SP1 and VC10, infinity was 10 (i.e. "variadic" templates supported 0 to 10 arguments inclusive).  In the VC11 Developer Preview, infinity is 5 by default.  This got our compiler memory consumption back to what it was in VC10.  If you need more arguments (e.g. you had code compiling with VC9 SP1 or VC10 that used 6-tuples), there's an escape hatch.  You can define _VARIADIC_MAX project-wide between 5 and 10 inclusive (it defaults to 5).  Increasing it will make the compiler consume more memory, and may require you to use the /Zm option to reserve more space for PCHes.