Bruce Momjian wrote:
> Tom Lane wrote:
> > Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> > > Right, palloc shouldn't. I was thinking of having another version of
> > > palloc that _does_ clear out memory, and calling that from a newNode()
> > > macro. We already know palloc is going to call MemoryContextAlloc, so
> > > we could have a pallocC() that calls a new MemoryContextAllocC() that
> > > would call the underlying memory allocation function, then do the loop
> > > like MemSet to clear it.
> >
> > But if the MemSet is inside the called function then it cannot reduce
> > the if-tests to a compile-time decision to invoke the word-zeroing loop.
> > We want the MemSet to be expanded at the newNode call site, where the
> > size of the allocated memory is a compile-time constant.
>
> I can easily do the tests in the MemSet macro, but I can't do a loop in
> a macro that has to return a value; I need while(). Though a loop in a
> new fuction will not be as fast as a MemSet macro, I think it will be
> better than what we have now with newNode only because newNode will be a
> macro and not a function anymore, i.e. the MemSet will happen in the
> function called by pallocC and not in newNode anymore, and there will be
> zero code bloat. I wish I saw another way.
OK, here's a patch for testing. It needs cleanup because the final
version would remove the nodes/nodes.c file. The net effect of the
patch is to make newNode a macro with little code bloat.
--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073