[openib-general] [PATCHv2] osm: improving FatTree routing engine

[Yevgeny Kliteynik]

Hi Hal,

Hal Rosenstock wrote:
> Hi Yevgeny,
> 
> On Wed, 2006-12-20 at 03:54, Yevgeny Kliteynik wrote:
>> Hi Hal
>>
>> [V2 of the patch - not using max_lid_ho setter]
>>
>> FatTree routing engine improvemets:
>> 1. Improved building of LFTs
>> 2. Setting max lid on osm switches
>> 3. Using ucast manager LFT dump function
>> 4. Stoped using global variable 'osm'
>> 5. Improved logging
>> 6. Some cosmetics
> 
> In general, it should be one "thought" per patch 
> but since this is so new
 
This is the reason why all these changes are in one patch - they 
are not appearing in the code in an incremental manner, but rather
as a bunch of changes all over the code.

> I will incorporate this all in one patch.
> 
>> Signed-off-by: Yevgeny Kliteynik <kliteyn at dev.mellanox.co.il>
> 
> Thanks. Applied.
> 
> One minor comment below.
> 
>> ---
>>  osm/opensm/osm_ucast_ftree.c |  439 +++++++++++++++++++++++++++---------------
>>  1 files changed, 281 insertions(+), 158 deletions(-)
>>
>> diff --git a/osm/opensm/osm_ucast_ftree.c b/osm/opensm/osm_ucast_ftree.c
>> index 15e4cd0..0d7188a 100644
>> --- a/osm/opensm/osm_ucast_ftree.c
>> +++ b/osm/opensm/osm_ucast_ftree.c
> 
> [snip...]
> 
>> +void
>> +ucast_mgr_dump_to_file(
>> +   IN  osm_ucast_mgr_t *p_mgr,
>> +   IN  const char *file_name,
>> +   IN  void (*func)(cl_map_item_t *, void *));
>> +
>> +void
>> +ucast_mgr_dump_lfts(
>> +   IN  cl_map_item_t *p_map_item,
>> +   void *cxt);
>> +
> 
> Rather than declaring these here, should these go into osm_ucast_mgr.h ?
 
I thought about it, but was reluctant to do it because osm_ucast_mgr.h contains
only "important" functions. But now that the dump function is one of the routing
engine capabilities, I guess you're right - it's better to declare these functions
in the header file.

-- Yevgeny

> -- Hal
>