linux-kernel - Re: all syscalls initially taking 4usec on a P4? Re: nonblocking UDPv4 recvfrom() taking 4usec @ 3GHz?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20070221110634.GA30394@2ka.mipt.ru>
Date:	Wed, 21 Feb 2007 14:06:34 +0300
From:	Evgeniy Polyakov <johnpol@....mipt.ru>
To:	bert hubert <bert.hubert@...herlabs.nl>
Cc:	Andi Kleen <andi@...stfloor.org>, netdev@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: all syscalls initially taking 4usec on a P4? Re: nonblocking UDPv4 recvfrom() taking 4usec @ 3GHz?

Here is data for 50 bytes reading for essentially idle machine 
(core duo 2.4 ghz):

delta for syscall: 3326961404-3326969261: 7857 cycles = 3.273750 us
delta for syscall: 3326975687-3326980979: 5292 cycles = 2.205000 us
delta for syscall: 3327199967-3327205583: 5616 cycles = 2.340000 us
delta for syscall: 3327208706-3327214817: 6111 cycles = 2.546250 us
delta for syscall: 3327442427-3327450446: 8019 cycles = 3.341250 us
delta for syscall: 3327453281-3327459104: 5823 cycles = 2.426250 us
delta for syscall: 3327462002-3327466844: 4842 cycles = 2.017500 us
delta for syscall: 3327726809-3327733451: 6642 cycles = 2.767500 us
delta for syscall: 3327736178-3327741947: 5769 cycles = 2.403750 us
delta for syscall: 3327968972-3327974264: 5292 cycles = 2.205000 us
delta for syscall: 3327977981-3327984506: 6525 cycles = 2.718750 us
delta for syscall: 3327987215-3327991958: 4743 cycles = 1.976250 us
delta for syscall: 3328245416-3328252058: 6642 cycles = 2.767500 us
delta for syscall: 3328258646-3328264172: 5526 cycles = 2.302500 us

and sometimes (frequently enough)

delta for syscall: 1937450129-1937469011: 18882 cycles = 7.867500 us
delta for syscall: 1956633332-1956654248: 20916 cycles = 8.715000 us
delta for syscall: 1995042065-1995061703: 19638 cycles = 8.182500 us
delta for syscall: 2014239488-2014257992: 18504 cycles = 7.710000 us
delta for syscall: 2016294674-2016316202: 21528 cycles = 8.970000 us
delta for syscall: 2054518259-2054537447: 19188 cycles = 7.995000 us
delta for syscall: 2071839668-2071857146: 17478 cycles = 7.282500 us

But there is even following sometimes:

delta for syscall: 3271734859-3271872037: 137178 cycles = 57.157501 us

Queue of the non-blocking socket is being kept full by another host
which sends udp with 40k chunks per socket write.

Reading /dev/zero in chunks of 50 bytes ends up with about 190
nanosecond on that machine.

-- 
	Evgeniy Polyakov
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/