Re: Speed and Topology map

[Andreas B. <and...@mu...>]

On Thu, Mar 16, 2000 at 04:37:27PM -0700, Bowles, Karl wrote:
> Unibrain: 
> speed_map[x]   | speed setting
> where x =         |
> -----------------------|------------------------
> 56                   |    S400
> 61                   |    S400

[big snip losing context]

> 127                 |    S400
> 184                 |    S400
> all others         |    0x00       Our drivers indicate all other's are at
> S100
> 
> When I figure it out:
> speed_code[i]=speed_code[j]; where [i]=64 x m + n & [j] = 64 x n + m
> with
>     m  |    n         =>    i    |    j
> ---------|----------         --------|-------
>     0   |    0                 0   |   0
>     0   |    1                 1   |   64
>     0   |    2                 2   |   128
>     1   |    0                 64 |   1
>     1   |    1                 65 |   65
>     1   |    2                 66 |   129
>     2   |    0                128|   2
>     2   |    1                129|   66
>     2   |    2                130|   130

These i and j indexes are the only legal to read in your 3 node
setup.  I'm not sure what you accessed in your two examples and I
don't know what the problem is.
 
> What did I do wrong? Our software agrees better with Unibrain's, but none of
> the three possibilities agree fully. Also, the way I did it looks like it
> will work until, but it is not that simple.

Do they disagree anywhere inside the legal indexes?

> I know you HAVE to implement the
> topology tree so that you can set the speed at each of the locations to the
> speed of the slowest node on the path from one point in the tree to the
> next.

My algorithm presets the map to S400 and eliminates impossible speeds.
Without explicitly creating a tree and using the information in place instead.

I hand-verified my algorithm with a number of test cases (9 node
setups) through a test program which displays the speed map and it
gives correct results.

> In my case, all the nodes are S400 capable so it is a trivial case.

And there certainly shouldn't be a problem with my algorithm in that case.

> Should the Linux driver software use the 0x0f convention of indicating there
> is not node at a location? 

It doesn't matter.  Speed codes outside those referring to existing
nodes are undefined by 1394-1995.  In Linux they start out as zero and
then just contain what they had last written to.

-- 
          Andreas E. Bombe <and...@mu...>
http://home.pages.de/~andreas.bombe/                DSA key 0x04880A44