Re: Handling transient data

[Thomas H. <th...@tu...>]

Jesse Barnes wrote:
> On Monday, March 03, 2008 10:34 am Thomas Hellström wrote:
>   
>> 1) Allocating all pages at once:
>> Yes, I think this might improve  performance in some cases. The reason
>> it hasn't been done already is the added complexity needed to keep track
>> of  the different allocation sizes. One optimization that´s already in
>> the pipeline is to page in more that a single page (for example 16) when
>> we hit a pagefault in nopfn().
>>     
>
> Can you elaborate on the added complexity?  In looking at the code the other 
> day, it seemed like the bits were already there to check for previously 
> allocated pages in the fault handler, as well as a place where we could 
> easily add in a real allocation call, drm_ttm_alloc_pages (which should 
> probably be called drm_ttm_alloc_pagedir or something instead, since it just 
> allocates the page pointers).
>
> Do you just mean the code to try allocating with different orders until the 
> allocation succeeds?
>   
Yes, that and the info needed to call free_pages with the correct order. 
However I'm not saying this should stop us from implementing this.

/Thomas

> Jesse
>