linux-mips
[Top] [All Lists]

Re: [PATCH] DECstation: Optimised early printk()

To: "Maciej W. Rozycki" <macro@linux-mips.org>
Subject: Re: [PATCH] DECstation: Optimised early printk()
From: Ralf Baechle <ralf@linux-mips.org>
Date: Thu, 7 Jun 2007 19:04:04 +0100
Cc: linux-mips@linux-mips.org
In-reply-to: <Pine.LNX.4.64N.0706051128210.15653@blysk.ds.pg.gda.pl>
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <Pine.LNX.4.64N.0706051128210.15653@blysk.ds.pg.gda.pl>
Sender: linux-mips-bounce@linux-mips.org
User-agent: Mutt/1.5.14 (2007-02-12)
On Tue, Jun 05, 2007 at 11:45:07AM +0100, Maciej W. Rozycki wrote:

>  This is an optimised implementation of early printk() for the DECstation.  
> After the recent conversion to a MIPS-specific generic routine using a 
> character-by-character output the performance dropped significantly.  
> This change reverts to the previous speed -- even at 9600 bps of the 
> serial console the difference is visible with a naked eye; I presume for a 
> framebuffer it is even worse (it may depend on exactly which one is used 
> though).
> 
>  Additionally the change includes a fix for a problem that the old 
> implementation had -- the format used would not actually limit the length 
> of the string output.  This new implementation uses a local buffer to deal 
> with it -- even with this additional copying it is much faster than the 
> generic function.
> 
>  Plus this driver is registered much earlier than the generic one, 
> allowing one to see critical messages, such as one about an incorrect CPU 
> setting used, that are produced beforehand. :-)

Queued up for 2.6.23, too.

>  As a side note, the SYS_HAS_EARLY_PRINTK option could probably be called 
> SYS_HAS_GENERIC_EARLY_PRINTK or something...

I take a patch to rename this :-)

  Ralf

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