Patch: Support IA-64 speculation [3/5]

[Vladimir N. Makarov]

Daniel Berlin wrote:

>On Fri, 2005-12-23 at 18:13 +0300, Maxim Kuvyrkov wrote:
>  
>
>>Hi,
>>
>>This patch is an addition to sched-rgn.c, that provides finding of more 
>>non-trivial regions for scheduling. It finds twice as much regions 
>>compared to traditional find_rgns procedure.
>>    
>>
>
>There has been *real* region finding code posted to the list before, and
>it should still work.
>
>  
>
Daniel, what code? Could you give a reference.  If you are talking about 
Bob Kid's forming superblocks or treeregion scheduling from NCSU,  they 
have advantages (forming bigger blocks or tree of blocks for scheduler) 
and disadvantages (code increase).  But may be it is another code and I 
missed it.

>Why not simply use that instead of trying to hack up the current regions
>code (which really isn't very good)?
>
>Otherwise, what's next? find_even_more_regions?
>
>  
>