Qualcomm Incorporated

International Business Machines Corporation

Kenneth Alan Dockser

James Norris Dieffenderfer

Thomas Andrew Sartorius

Michael Scott McIlvaine

Rodney Wayne Smith

Jeffrey Todd Bridges

Pathik Sunil Lall

Thomas Philip Speier

Victor Roberts Augsburg

Yusuf Cagatay Tekmen

Robert Douglas Clancy

Manish Satyapal Gupta

Brian Stempel

Michael W Morrow

Anil Krishna

Daren Eugene Streett

Gerald Paul Michalak

Bonnie Collett Sexton

Siddhartha Chatterjee

Sandeep Suresh Navada

Fred Gehrung Gustayson

Michael Thomas Dibrino

Henry Arthur Pellerin, Iii

Niket Kumar Choudhary

Michael Scott Mcilvaine

Michael William Morrow

Iii Henry Arthur Pellerin

A processor includes a queue for storing instructions processed within the context of a current value of a register field, where for some embodiments the instruction is undefined or defined, depending upon the register field at time of processing. After a write instruction (an instruction that writes to the register field) executes, the queue is searched for any entries that contain instructions that depend upon the executed write instruction. Each such entry stores the value of the register field at the time the instruction in the entry was processed. If such an entry is found in the queue and its stored value of the register field does not match the value that the write instruction wrote to the register field, then the processor flushes the pipeline and restarts at a state so as to correctly execute the instruction.

Optimizing performance for context-dependent instructions

Systems and methods relate to division of a dividend by a divisor, with fast result formatting. Counts of leading sign bits of the dividend and the divisor are determined. The dividend and the divisor are normalized based on their respective counts of leading sign bits to obtain a normalized dividend and a normalized divisor, respectively. An exact number of significant quotient bits of a quotient of the division, based on the normalized dividend, the normalized divisor, and the counts of leading sign bits of the dividend and the divisor and used to determine a correct position of a leading bit of the quotient based on this exact number. The quotient is developed by placing the leading bit at or near the correct position and appending less significant bits to the right of the leading bit. Thus, left-shifts in each iteration and large final shifts are avoided in formatting the result.

Division and root computation with fast result formatting

A memory structure compresses a portion of a memory tag using an indexed tag compression structure. A set of higher order bits of the memory tag may be stored in the indexed tag compression structure, where the set of higher order bits are identified by an index value. A tag array stores a set of lower order bits of the memory tag and the index value identifying the entry in the tag compression structure storing the set of higher order bits of the memory tag. The memory tag may comprise at least a portion of a memory address of a data element stored in a data array.

Method and apparatus for cache tag compression

A method and apparatus for allowing an out-of-order processor to reuse an in-use physical register is disclosed herein. The method and apparatus uses identifiers, such as tokens and/or other identifiers in a rename map table (RMT) and a physical register file (PRF), to indicate whether an instruction result is allowed or disallowed to be written into a physical register.

Method and apparatus for selective renaming in a microprocessor

Apparatus and methods provide early access of instructions. A fetch queue is coupled to an instruction cache and configured to store a mix of processor instructions for a first processor and coprocessor instructions for a second processor. A coprocessor instruction selector is coupled to the fetch queue and configured to copy coprocessor instructions from the fetch queue. A queue is coupled to the coprocessor instruction selector and from which coprocessor instructions are accessed for execution before the coprocessor instruction is issued to the first processor. Execution of the copied coprocessor instruction is started in the coprocessor before the coprocessor instruction is issued to a processor. The execution of the copied coprocessor instruction is completed based on information received from the processor after the coprocessor instruction has been issued to the processor.

Processor with a coprocessor having early access to not-yet issued instructions

A queuing apparatus having a hierarchy of queues, in one of a number of aspects, is configured to control backpressure between processors in a multiprocessor system. A fetch queue is coupled to an instruction cache and configured to store first instructions for a first processor and second instructions for a second processor in an order fetched from the instruction cache. An in-order queue is coupled to the fetch queue and configured to store the second instructions accepted from the fetch queue in response to a write indication. An out-of-order queue is coupled to the fetch queue and to the in-order queue and configured to store the second instructions accepted from the fetch queue in response to an indication that space is available in the out-of-order queue, wherein the second instructions may be accessed out-of-order with respect to other second instructions executing on different execution pipelines.

Hybrid queue for storing instructions from fetch queue directly in out-of-order queue or temporarily in in-order queue until space is available

A controlled-precision Iterative Arithmetic Logic Unit (IALU) included in a processor produces sub-precision results, i.e. results having a bit precision less than full precision. In one embodiment, the controlled-precision IALU comprises an arithmetic logic circuit and a precision control circuit. The arithmetic logic circuit is configured to iteratively process operands of a first bit precision to obtain a result. The precision control circuit is configured to end the iterative operand processing when the result achieves a programmed second bit precision less than the first bit precision. In one embodiment, the precision control circuit causes the arithmetic logic circuit to end the iterative operand processing in response to an indicator received by the control circuit. The controlled-precision IALU further comprises rounding logic configured to round the sub-precision result.

Controlled-precision iterative arithmetic logic unit

Selective power control of one or more processing elements matches a degree of parallelism to requirements of a task performed in a highly parallel programmable data processor. For example, when program operations require less than the full width of the data path, a software instruction of the program sets a mode of operation requiring a subset of the parallel processing capacity. At least one parallel processing element, that is not needed, can be shut down to conserve power. At a later time, when the added capacity is needed, execution of another software instruction sets the mode of operation to that of the wider data path, typically the full width, and the mode change reactivates the previously shut-down processing element.

Software selectable adjustment of SIMD parallelism

A method and apparatus for performing a floating-point operation with a floating-point processor having a given precision is disclosed. A subprecision for the floating-point operation on one or more floating-point numbers is selected. The selection of the subprecision results in one or more excess bits for each of the one or more floating-point numbers. Power may be removed from one or more components in the floating-point processor that would otherwise be used to store or process the one or more excess bits, and the floating-point operation is performed with power removed from the one or more components.

Floating-point processor with reduced power requirements for selectable subprecision

Instruction execution delay is alterable after the system design has been finalized, thus enabling the system to dynamically account for various conditions that impact instruction execution. In some embodiments, the dynamic delay is determined by an application to be executed by the processing system. In other embodiments, the dynamic delay is determined by analyzing the history of previously executed instructions. In yet other embodiments, the dynamic delay is determined by assessing the processing resources available to a given application. Regardless, the delay may be dynamically altered on a per-instruction, multiple instruction, or application basis. Processor instruction execution may be controlled by determining a first delay value for a first set of one or more instructions and a second delay value for a second set of one or more instructions. Execution of the sets of instructions is delayed based on the corresponding delay value.

Method and apparatus for executing processor instructions based on a dynamically alterable delay

In a denormal support mode, the normalization circuit of a floating-point adder is used to normalize or denormalized the output of a floating-point multiplier. Each floating-point multiply instruction is speculatively converted to a multiply-add instruction, with the addend forced to zero. This preserves the value of the product, while normalizing or denormalizing the product using the floating-point adder's normalization circuit. When the operands to the multiply operation are available, they are inspected. If the operands will not generate an unnormal intermediate product or a denormal final product, the add operation is suppressed, such as by operand-forwarding. Additionally, each non-fused floating-point multiply-add instruction is replaced with a multiply-add instruction having a zero addend, and a floating-point add instruction having the addend of the original multiply-add instruction is inserted into the instruction stream. Upon inspection of the operands, if an unnormal intermediate result or a denormal final result will not occur, the addend may be restored to the multiply-add instruction and the add instruction converted to a NOP.

Mode-based multiply-add recoding for denormal operands

A microprocessor includes a branch unit, a load/store unit (LSU), an arithmetic logic unit (ALU), and a vector unit to execute a vector instruction. The vector unit includes a vector register file having a primary vector register and a secondary vector register. The processor preferably further includes a first data bus and a second data bus wherein the first and second data busses couple the vector unit to the data memory. The vector unit includes a first input multiplexer enabling data on the first data bus to be provided to the primary register file or the secondary register file and a second input multiplexer, independent of the first input multiplexer enabling data on the second data bus to be provided to the second data bus. The first and second data busses may comprise first and second portions of a data memory bus.

Vector unit in a processor enabled to replicate data on a first portion of a data bus to primary and secondary registers

A pre-saturating multiplier inspects the operands to a multiply operation prior to performing any multiplication. If the operands will cause an overflow requiring saturation, the multiplier outputs the saturated value without multiplying the original operands. In one embodiment, parameters derived from the operands are altered such that when the multiply operation is performed on the altered parameters, the multiplier produces the saturated result. This may comprise altering a Booth recoded bit group to select a negative zero instead of a zero as a partial product, and suppressing the addition of the value one to the partial products (thus effectively subtracting the value one). In another embodiment, when the operands that will cause an overflow are detected, the output of the multiplier is forced to a predetermined saturation value.

Growing community of inventors

Cary, NC, United States of America

Kenneth Alan Dockser