Date: Thu, 14 Feb 2019 21:22:08 -0300
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Thomas Munro <thomas.munro@enterprisedb.com>,
	Andrew Gierth <andrew@tao11.riddles.org.uk>,
	David Rowley <david.rowley@2ndquadrant.com>,
	PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: Using POPCNT and other advanced bit manipulation instructions
Message-ID: <20190215002208.GA25042@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <4022.1550185074@sss.pgh.pa.us>
User-Agent: Mutt/1.9.4 (2018-02-28)
Precedence: bulk

On 2019-Feb-14, Tom Lane wrote:

> I think we need a clean test for __builtin_popcount(), and to be willing
> to use it if available, independently of -mpopcnt.  Then separately we
> should test to see if -mpopcnt works, probably with the same
> infrastructure we use for checking for other compiler flags, viz
> 
>    # Optimization flags for specific files that benefit from vectorization
>    PGAC_PROG_CC_VAR_OPT(CFLAGS_VECTOR, [-funroll-loops])
>    PGAC_PROG_CC_VAR_OPT(CFLAGS_VECTOR, [-ftree-vectorize])
> +  # Optimization flags for bit-twiddling
> +  PGAC_PROG_CC_VAR_OPT(CFLAGS_POPCNT, [-mpopcnt])
>    # We want to suppress clang's unhelpful unused-command-line-argument warnings
> 
> Then the correct test to see if we want to build pg_popcount.c (BTW,
> please pick a less generic name for that) and the choose function
> is whether we have *both* HAVE__BUILTIN_POPCOUNT and nonempty
> CFLAGS_POPCNT.

Yeah, this works.  I'll post the patch tomorrow.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services