Re: Using 128-bit integers for sum, avg and statistics aggregates

On 2015-03-22 22:00:13 +0100, Petr Jelinek wrote:
> On 22/03/15 13:59, Andreas Karlsson wrote:
> >Would this mean the bug is most likely somewhere in
> >int128_to_numericvar()? Maybe that version of gcc has a bug in some
> >__int128 operator or I messed up the code there somehow.

Yes, or a compiler bug. I looked through the code again and found and
fixed one minor bug, but that doesnt' explain the problem.

> Yeah that's what I was thinking also, and I went through the function and
> didn't find anything suspicious (besides it's same as the 64 bit version
> except for the int128 use).
>
> It really might be some combination of arithmetic + the conversion to 16bit
> uint bug in the compiler. Would be worth to try to produce test case and try
> it standalone maybe?

A compiler bug looks like a not unreasonable bet at this point. I've
asked Andrew to recompile without optimizations... We'll see whether
that makes a difference. Jacana is the only compiler with gcc 4.8.1 (or
is it 4.8.0? there's conflicting output). There've been a number of bugs
fixed that affect loop unrolling and such.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services