ANNOUNCEMENT: Live Wireshark University & Allegro Packets online APAC Wireshark Training Session
April 17th, 2024 | 14:30-16:00 SGT (UTC+8) | Online

Wireshark-dev: Re: [Wireshark-dev] Profiling results with optimized function ws_mempbrk_sse42()

From: Anders Broman <anders.broman@xxxxxxxxxxxx>
Date: Wed, 4 Jun 2014 11:24:59 +0000

Hi,

I did a test with

static inline const guint8*

guint8_pbrk(const guint8* haystack, size_t haystacklen _U_, const guint8 *needles, guchar *found_needle)

{

                const guint8 *result = (const guint8 *)strpbrk(haystack, needles);

 

                if (result && found_needle)

                                *found_needle = *result;

 

                return result;

}

 

Which seems to give as good results on Ubuntu 13.10, it ends up using __strpbrk_sse42 (libc-2.17.so: strcspn-c.c, ...)

==29458==

==29458== Events    : Ir

==29458== Collected : 21130728727

==29458==

==29458== I   refs:      21,130,728,727

Load time in WS 0:19.300

 

 

From: wireshark-dev-bounces@xxxxxxxxxxxxx [mailto:wireshark-dev-bounces@xxxxxxxxxxxxx] On Behalf Of Anders Broman
Sent: den 2 juni 2014 08:36
To: wireshark-dev@xxxxxxxxxxxxx
Subject: [Wireshark-dev] Profileing results with optimized function ws_mempbrk_sse42()

 

Hi,

I think this looks very good. With filter

22,212,643,979 reduced to 21,418,572,385. Could we use something similar for format_text() and

Address equal functions? Other string functions?

Regards

Anders

Attachment: callgrind.out.strpbrk.29458.bz2
Description: callgrind.out.strpbrk.29458.bz2