IEEE 200X Fast Track Change Proposal

ID:      FT-28

Status:       Proposed
Proposed:     David Bishop, 10/04, dbishop@vhdl.org
Analyzed:     4-April-05 (Jim Lewis from meeting notes dated 2/28/2005)
Resolved:     Date

Related issues:  CP-002 (std_logic_1164)
Relevant LRM sections:   



Enhancement/Requirements Summary:
-------------------------------------------------------- 
Currently ROR, ROL, SLL, SLR, SLA, and SRA are implicitly
defined only for arrays of bit and boolean.  Going further,
the definition for SLA is odd for positive values given 
the current definition of the numeric_std type math packages
with the left-most bit begin the most significant bit.  

This is a roll-up of how to handle these functions for 
the various packages associated with this standard.


Analysis/LRM Changes
----------------------------
   1) Explicitly define non-numeric shifts (sll, srl, rol, ror) for 
      std_logic_1164 same as bit vectors including the value '0'

   2) Explicitly define all shifts for unsigned, signed in numeric_std/numeric_bit.
      SLA, SLL, SRA, and SRL are as defined in the table below.  

                       sla          sll                    sra                            srl
                    --------------------------------------------------------------------------------------------
      +   signed   |   sll          arg(arg'left-1:0)&0    arg(arg'left)&arg(arg'left:1)  _0&arg(arg'left:1)_   
      -   signed   |   sra (-cnt)   srl(-cnt)              sla(-cnt)                      sll(-cnt)  
      +   unsigned |   sll          arg(arg'left-1:0)&0    0&arg(arg'left:1)              0&arg(arg'left:1)   
      -   unsigned |   sra (-cnt)   srl(-cnt)              sla(-cnt)                      sll(-cnt)  


   3) Explicitly define numeric shifts (sla, sra) for slv in numeric_std_unsigned
      identical to unsigned.  Explicitly define numeric shifts (sla and sra) for
      bit_vector in numeric_bit_unsigned


    4) For fixed point, the implementation for ufixed should match unsigned
       and sfixed should match signed.

    5) Do not implement shift functions for floating point



Resolution:
----------------------------
[To be performed by the 200X Fast Track Working Group]