linux-mips
[Top] [All Lists]

Re: cpu_idle and cpu_wait

To: ralf@linux-mips.org
Subject: Re: cpu_idle and cpu_wait
From: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Date: Thu, 17 Nov 2005 16:08:33 +0900 (JST)
Cc: linux-mips@linux-mips.org
In-reply-to: <20051117.112144.108306652.nemoto@toshiba-tops.co.jp>
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <20051117.011906.25910026.anemo@mba.ocn.ne.jp> <20051116184201.GJ3229@linux-mips.org> <20051117.112144.108306652.nemoto@toshiba-tops.co.jp>
Sender: linux-mips-bounce@linux-mips.org
>>>>> On Thu, 17 Nov 2005 11:21:45 +0900 (JST), Atsushi Nemoto 
>>>>> <anemo@mba.ocn.ne.jp> said:
anemo> And I checked x86 implemetation and found that HLT or MWAIT
anemo> instruction also must be used with interrupts enabled.  So IIUC
anemo> it seems x86 have same latency problem too.

No, I was wrong.  MWAIT (and MONITOR) instruction provides something
like "test and wait" mechanism.  mwait_idle() is using them for
thread_flag, so there is no latency problem on processors which have
MWAIT/MONITOR.

---
Atsushi Nemoto

<Prev in Thread] Current Thread [Next in Thread>