Re: [HACKERS] [PATCH] relocation truncated to fit: citus build failure on s390x

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> On Mon, May 29, 2017 at 3:45 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> I wonder what the overhead is of using -fPIC when -fpic would be
>> sufficient.

> Do we have an idea how to measure the increased overhead? Just from
> reading the description, I'm guessing that the increased cost would
> happen when the extension calls back into core, but maybe that doesn't
> happen often enough to worry about?

My gut feeling is that it'd be a pretty distributed cost, because every
internal cross-reference in the .so (for instance, loading the address of
a string literal) would involve a bit more overhead to support a wider
offset field. An easy thing to look at would be how much the code expands
by. That might or might not be a good proxy for the runtime slowdown
percentage, but it seems like it ought to serve as a zero-order
approximation.

regards, tom lane