[dita-sidsc] Finished updating register element spreadsheet.

I have finished updating the register element 
spreadsheet. I thought it would send out an email but neither Doug nor I has 
received one yet, so maybe I did something wrong. 

 

If you later get the official email, I apologize for 
the duplication. Here is the message I tried to send:

 

 

Background
The SIDSC dimension element is defined to 
be a string that supports a comma-separated list of integers. <dimension> 
can be instantiated multiple times.
The IP-XACT dim element on which the 
SIDSC element was based is defined as an integer. <dim> too can be 
instantiated multiple times.

 

Recommendation
I recommend we adopt the IP-XACT 
convention of using multiply instantiated integer values, then realign with the 
next revision of IP-XACT if we choose. We clearly don't need or want both a 
comma-separated list and multiple instantiation, and I think the comma-separated 
list will be troublesome both to parse and to validate. Using integer values 
also mirrors the current IP-XACT schemas.

 

Details
It looks to me as though an IP-XACT 
annotation on the dim element was misinterpreted when the SIDSC element was 
being developed. This annotation reads "Dimensions a register array, the 
semantics for dim elements are the same as the C language standard for the 
layout of memory in multidimensional arrays."

 

A Freescale person who is involved with IP-XACT  
confirmed that the intent of this annotation was to signal a document order 
dependence. In other words, dim elements are defined in the same order they 
would appear in a C language array declaration, and they should be interpreted 
as such. This is recognized among IP-XACT members as a potential source of 
trouble.

 

One possible solution would be to introduce a list 
structure in which the ordering of dimensions is made explicit. Another solution 
would be to nest dim elements, so that the logical structure of a 
multi-dimension array would be mirrored in the element structure. Both solutions 
are being considered for IP-XACT rev 1.5 due in early 2009.

 

However, my source said that to the best of his 
knowledge, no one has proposed changing the <dim> element into a 
non-integer.

 

 

 

with best regards

 

Tom Cihak

Freescale Semiconductor

512-996-5072