Advanced Micro Devices Corporation

David Oliver

Willard S Briggs

Scott Hilker

Debjit Das Sarma

Hanbing Liu

Debjit DasSarma

The disclosed embodiments relate to methods and apparatus for accurately, efficiently and quickly executing a fused multiply-and-accumulate instruction with respect to floating-point operands that have packed-single-precision format. The disclosed embodiments can speed up computation of a high-part of a result during a fused multiply-and-accumulate operation so that cycle delay can be reduced and so that power consumption can be reduced.

Methods and apparatus for compressing partial products during a fused multiply-and-accumulate (FMAC) operation on operands having a packed-single-precision format

A device and methods are disclosed for communicating an unrounded result from one arithmetic calculation for use in a second, subsequent calculation. For example, an unrounded result of a first calculation can be forwarded to provide a multiplier, a multiplicand or an addend operand for the subsequent operation. The operand can be forwarded to the input of the same fused multiply addition module (FMAM) that supplied the result, or to another FMAM, and do so without regard to the precision of the forwarded operand, the precision of the subsequent operation, or the native precision of the FMAM.

Arithmetic processing device and methods thereof

A floating point unit includes a floating point adder to perform a floating point addition operation between first and second floating point numbers each having an exponent and a mantissa. The floating point unit also includes an alignment shifter that may calculate a shift value corresponding to a number of bit positions to shift the second mantissa such that the second exponent value is the same as the first exponent value. The alignment shifter may detect an overshift condition, in which the shift value is greater than or equal to a selected overshift threshold value. The selected overshift threshold value comprises a base 2 number in a range of overshift values including a minimum overshift threshold value and a maximum overshift threshold value, and which has a largest number of a consecutive of bits that are zero beginning at a least significant bit.

Mechanism for fast detection of overshift in a floating point unit of a processing device

An apparatus and method is disclosed for updating a status register in an out of order execution pipeline. In one embodiment a dispatch unit in a floating point unit sets a MRI bit flag that indicates that an instruction is the most recently issued instruction. The dispatch unit resets the MRI bit flag for all other instructions. Each execution stage of the execution pipeline keeps track of the MRI bit flag information for the instruction. A writeback unit updates the status register after the execution of the instruction that has its MRI bit flag set. The writeback unit does not update the status register for instructions that have their MRI bit flag reset. This allows the instruction to be identified that is the most recent instruction to enter the dispatch unit.

Apparatus and method for updating a status register in an out of order execution pipeline based on most recently issued instruction information

A variable speed floating point unit comprising: 1) an execution pipeline comprising a plurality of execution stages capable of executing floating point operations in a series of sequential steps; and 2) a clock controller capable of receiving an input clock signal and generating a variable speed output clock signal capable of clocking the execution pipeline. The clock controller adjusts a speed of the variable speed output clock signal according to a level of queued opcodes waiting to be executed in the execution pipeline.

Floating point unit with variable speed execution pipeline and method of operation

A floating point unit comprising: 1) an execution pipeline comprising a plurality of execution stages for executing floating point operations in a series of sequential steps; and 2) a try-again reservation station for storing a plurality of instructions to be loaded into the execution pipeline. Detection of a denormal result in the execution pipeline causes the execution pipeline to store the denormal result in a register array associated with the floating point unit and causes the execution pipeline to store a denormal result instruction in the try-again reservation station. The try-again reservation station subsequently re-loads the denormal result instruction into the execution pipeline and the de-normal result instruction retrieves the denormal result from the register array for additional processing.

Growing community of inventors

Longmont, CO, United States of America

David Oliver