Re: 64-bit Compile Failure on Solaris 10 with OpenSSL

Поиск
Список
Период
Сортировка
От Zdenek Kotala
Тема Re: 64-bit Compile Failure on Solaris 10 with OpenSSL
Дата
Msg-id 48C6411F.3060607@sun.com
обсуждение исходный текст
Ответ на 64-bit Compile Failure on Solaris 10 with OpenSSL  ("Randal T. Rioux" <randy@procyonlabs.com>)
Ответы Re: 64-bit Compile Failure on Solaris 10 with OpenSSL
Список pgsql-general
Randal T. Rioux napsal(a):
> I've battled this for a while. I'm finally breaking down and asking for help.
>
> If you're answer to this is "why 64-bit" then don't answer. You wouldn't
> understand. Same if you say "why don't you use packages."
>
> Here is my scenerio:
>
>  - Sun 420R x450Mhz UltraSPARC-II / 4GB RAM
>  - Solaris 10 05/08
>  - OpenSSL 0.9.8h
>  - PostgreSQL 8.3.3
>  - GCC 3.4.6
>  - GNU Make 3.81

Three questions (yeah, you forbided ask, but ...)

1) Why 64

64bit code on SPARC is slower, because SPARC uses 4byte instructions and
processing 64bit data needs more instructions. It is good only if you
need more then 4GB share memory. When you use sunstudio compiler with
best optimization 64bit application has 1%-5% performance degradation.

2) Why you don't use package

You can use Solaris'es packages, which are integrated and optimized for
Solaris.

3)  Why you don't use build-in libssl?

Integrated libssl is not only copy of original open ssl. It has lot of
improvements and it uses crypto hardware accelerator if you have it (for
example Niagara 2).

See more:
http://blogs.sun.com/janp/
http://blogs.sun.com/janp/entry/on_openssl_versions_in_solaris

        Zdenek

В списке pgsql-general по дате отправления:

Предыдущее
От: Martijn van Oosterhout
Дата:
Сообщение: Re: Very weird problem of "order by" in postgresql
Следующее
От: Zdenek Kotala
Дата:
Сообщение: Re: Postgres in a solaris zone - patch 125077-02 needed