In a network for transferring a source field in a source word into a destination field in a destination word two basic hardware sub-functions are utilized: rotation and mask vector generation. In the network the destination field of a destination word is masked. Concurrently in the network, a source word is rotated bringing the source field thereof into corresponding alignment with the masked destination field and all but the source field of the source word is masked. Subsequent logical combining of the masked destination word and the rotated and masked source word generates the desired field transference. In one embodiment the required masking operation is accomplished during a single pass of the destination and source words through the network. In an alternate embodiment using less masking hardware only half of the required masking is accomplished during each pass and two passes are required before the logical combining to achieve the desired field transference.

Shift network having a mask generator and a rotator

A shift circuit comprises a plurality of stages of data selectors, a first data selector, a second data selector and a temporary register. The data selectors shift an n-bit input data by m bits which are specified by a shifting amount data and produces a first output data or a second output data. The first and second data selectors selectively output the first and second output data in accordance with shift direction. The temporary register comprises n bits for storing a data which is not selected by said first or second data selector.

Shift circuit having a plurality of cascade-connected data selectors

A barrel shifter for a floating point processing unit can be optimized by having an automatic normalization feature. A multi-stage shifting unit is employed with external circuitry to verify that only leading zeros would be shifted out before activating a given shifting stage. Bit reversers are also used on the input and the output side of the barrel shifter in order to minimize the size and complexity of the circuit while still allowing bi-directional shifting.

Barrel shifter using bit reversers and having automatic normalization

Where an N-bit input data is rotated together with a carry bit by an N-bit or more rotate count, the actual rotate count is obtained as a remainder or modulo of x/N+1 (x: rotate count; N: data bit length). The above remainder will not be obtained by simply masking shift signals. Therefore, the remainder is calculated at high speed through hardware including a subtrahend calculator section for calculating (N+1) (i) (i=0, 1, 2, . . . ) and a subtracter section for calculating {x-(N+1) (i)} to obtain a modulo or a remainder representative of an actual rotate count.

Barrel shifter including rotate operation

A mask programmable logic array (PLA) for producing a particular digital output given a certain digital input. The input signals to the PLA first pass through a series of AND gates resulting in a predetermined number of product terms being formed. The product signals then pass through a set of OR gates to become the final output signals. In the subject invention, the AND gates and OR gates are implemented through the use of NOR-NOR logic. A first set of NOR gates is implemented in an array to receive input signals and to produce product terms. A second and third set of NOR gates form two arrays. These two arrays are then located on either side of the first array to receive selected product signals in order to produce final output signals. In effect the OR portion of the PLA has been split into two arrays. TABLE OF CONTENTS Subject Background of the Invention Summary of the Invention Brief Description of the Drawings Detailed Description of the Preferred Embodiment The System Block Diagram Microprocessor Unit Pin Designations Clock and Timing Signals System Timing The ROM The Stack Area The RAM Area Elimination of Race Conditions in the RAM The ALU and Control Time Slot End Predictor The CROM Bit Manipulation Scheme Data Pad Input/Output Precharged Data Line Driver Bus Control Test Circuitry Split PLA Control The S-Counter Details of Logic Blocks The MOS/LSI Chip The Chip Test Functions The Instruction Set

Split programmable logic array

A parallel shifter consists of a plurality of AND gates arranged in a skewed configuration to simulate the partial products of a multiplication of the binary number to be shifted with a second binary number representing 2 N where N is the number of shift steps. Pseudo multiplication may be accomplished merely by ORing the outputs of the AND gates of any given binary weight since only one partial product of that weight will be non-zero. Left or right shift is accomplished by selection of the most significant or least significant half of the product. Rotation is accomplished by merging of the two halves.

Tich T. Dao

Papers

Expandable parallel binary shifter/rotator