Next: A.2 Multiblock Configuration. Up: A PRELIMINARY SPECIFICATION FOR Previous: A PRELIMINARY SPECIFICATION FOR

A.1 General Description

The Multiblock Mode of operation uses a special data access scheme which significantly increases the over-all data rate when gathering sparse data from multiple Slaves in a Crate Segment. The Slaves are initially configured by FASTBUS commands to form a Multiblock Group which may subsequently be read as one virtual Slave containing a single data buffer. This eliminates the overhead normally required for Sparse Data Scans and Primary and Secondary Addressing for every Slave in the Group. The Slaves forming the Multiblock Group must all be within one crate segment and if not adjacent to each other then intervening slots must be occupied by modules which connect the Daisy-Chain lines.

Multiblock Mode may also be used to write data into a Multiblock Group with the overhead of only one Primary and one Secondary Address Operation. The data buffer of each Slave in the Group will be completely filled in turn.

To read all the data from a Multiblock Group a Master attaches to Data Space Register #2 of that Group's Slave having the highest Geographical Address, using either Logical or Geographical Primary Addressing. This Slave is called the Primary Link The Master then proceeds to read data using Single or Block Transfers.

When the data buffer of the Primary Link is empty, rather than issuing an SS=2 (buffer empty) response it uses the Daisy-Chain lines to pass a token to the Multiblock Slave having the next lower Geographical Address. This Slave is called a Middle Link. Receipt of the token by the Middle Link causes it to respond to ensuing data cycles driven by the Master. Reading continues until the buffer of the Middle Link is empty. The token is then passed to the Multiblock Slave having the next lower Geographical Address indicating that it is now time for it to supply data.

The Multiblock Slave having the lowest Geographical Address in the Multiblock Group is called the End Link. When addressed and its buffer is empty it issues SS=2 in response to the Master's data cycle.

The amount of data contained in a Link does not affect operation in Multiblock Mode. Links containing no data simply pass the token to the next Link.

Since Links may contain variable amounts of data an indicator should be embedded in the data to specify the source of the data. In Multiblock Mode the Master cannot otherwise tell which Slave is supplying the data.

A Master may rapidly load all data buffers in a Multiblock Group by first attaching to Data Space Register (DSR) #2 of the Primary Link using either Logical or Geographical Primary Addressing and writing data using Single or Block Transfers. When the data buffer of the Primary Link is full the token is passed to the next Link. Writing continues until that Slave's buffer is full. When the buffer in the End Link is full it issues SS=2.

Pipeline operations should not be used as the Slaves assert Wait during certain parts of a Multiblock Operation.



Next: A.2 Multiblock Configuration. Up: A PRELIMINARY SPECIFICATION FOR Previous: A PRELIMINARY SPECIFICATION FOR


A.Daviel,TRIUMF