Feedback-ID: ie0a040ee:Fastmail
Message-ID: <84e2783f-8082-4fd9-bf5f-31982caf9aff@eisentraut.org>
Date: Sun, 26 May 2024 23:52:34 +0200
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: Improve conditional compilation for direct I/O alignment checks
To: Junwang Zhao <zhjwpku@gmail.com>,
 PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
References: <CAEG8a3J5w053U=O2+4_uxbXuMnmQFFVQef4GveV9QfW0wdeEaw@mail.gmail.com>
Content-Language: en-US
From: Peter Eisentraut <peter@eisentraut.org>
In-Reply-To: <CAEG8a3J5w053U=O2+4_uxbXuMnmQFFVQef4GveV9QfW0wdeEaw@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Archived-At: <https://www.postgresql.org/message-id/84e2783f-8082-4fd9-bf5f-31982caf9aff%40eisentraut.org>
Precedence: bulk

On 26.05.24 09:16, Junwang Zhao wrote:
> This patch refactors the alignment checks for direct I/O to preprocess phase,
> thereby reducing some CPU cycles.

This patch replaces for example

     if (PG_O_DIRECT != 0 && PG_IO_ALIGN_SIZE <= BLCKSZ)
         Assert((uintptr_t) buffer == TYPEALIGN(PG_IO_ALIGN_SIZE, buffer));

with

     #if PG_O_DIRECT != 0 && PG_IO_ALIGN_SIZE <= BLCKSZ
         Assert((uintptr_t) buffer == TYPEALIGN(PG_IO_ALIGN_SIZE, buffer));
     #endif

You appear to be assuming that this saves some CPU cycles.  But this is 
not the case.  The compiler will remove the unused code in the first 
case because all the constants in the if expression are known at 
compile-time.