From: | Larry Rosenman <ler(at)lerctr(dot)org> |
---|---|
To: | pgsql-hackers(at)postgresql(dot)org |
Subject: | UnixWare/Compiler Stuff |
Date: | 2003-08-08 20:31:09 |
Message-ID: | 304900000.1060374669@lerlaptop-red.iadfw.net |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Here is Dave's reply re: -K no_host vs. removing -K inline. (I.E. ADDING
-K no_host vs. removing -K inline:
-------Dave Prosser:
> I had someone ask about performance impact. I gave him the alternate
> -Kno_host as well.
>
> Do you have a feel for what it would do to/for us?
There are only a few places where -Khost and -Kno_host makes a
difference. Obviously, it affects the value of the preprocessing
predefine __STDC_HOSTED__, but otherwise, the only effects I see
on a quick grepping are:
- whether "inline" on "main" is ignored
- whether the known library function rewrites/side effect checks
are enabled. Other than what you'll find in /usr/ccs/lib/intrinsics,
the only other rewrites are:
- strcpy(X, "string") -> BMOVE of "string" if X has no side effects
- strncpy(X, "string", sizeof("string")+1) -> as above
- strlen("string") -> the length
- some "redundant loop" unrolling, presumably since we don't know about
potential side effects
But, like with all optimizations, your mileage will vary. You can
try it with and without and see how much of a difference it makes.
------end Dave Prosser.
I can see doing either, but that's the team's decision.
The fix won't be out until Update Pack 3 at least (which is 3 months away).
It **MIGHT** make an update pack but I don't know.
LER
--
Larry Rosenman http://www.lerctr.org/~ler
Phone: +1 972-414-9812 E-Mail: ler(at)lerctr(dot)org
US Mail: 1905 Steamboat Springs Drive, Garland, TX 75044-6749
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2003-08-08 20:33:22 | Re: WITH HOLD and pooled connections |
Previous Message | Bruce Momjian | 2003-08-08 20:31:08 | Re: WITH HOLD and pooled connections |