International Business Machines Corporation

James Henry Hesson

Stephen J Ciavaglia

Jay LeBlanc

Pamela Anne Wilcox

Walter Thomas Esling

Jay Gerald Heaslip

Stephen William Mahin

Geoffrey Francis Burns

Miles G Canada

Steven Michael Burchfiel

Steven C Espy

Walter Esling

Pamela A Wilcox

An apparatus performs source operand dependency analysis, perform register renaming and provide rapid pipeline recovery for a microprocessor capable of issuing and executing multiple instructions out-of-order in a single machine cycle. The apparatus first provides an enhanced means for rapid pipeline recovery due to a mispredicted branch or other store/load conflict or unsupported store/load forward circumstances. Second, the apparatus provides an improved instruction scheduling means wherein the oldest instructions that have all of their dependencies resolved are executed first. Third, the apparatus provides a means for enabling any execution or memory access instruction to execute out-of-order. Fourth, the apparatus provides a means for handling precise recovery of interrupts when processing instructions in out-of-order sequence.

Apparatus to perform source operand dependency analysis perform register

A floating point arithmetic unit for performing independent multiply and add operations in the execution of a multiply-add instruction AC+B on three operands A, B, and C of p-bit precision includes a multiplier unit, a sticky collect unit, an adder unit, and a rounding unit. In addition, a risk condition detection unit provides detection of a risk condition corresponding to an occurrence of an imprecise resultant quantity prior to being rounded by the rounding unit. Upon detection of a risk condition, a trap is triggered and an extended sequence implementation unit carries out an extended multiply-add sequence and provides a multiply-add output having infinite precision prior to a final rounding. A floating point arithmetic method for performing independent multiply and add operations in the execution of a multiply-add instruction AC+B on three operands A, B, and C of p-bit precision is disclosed also.

Method and apparatus for performing high-precision multiply-add

A parallel multiply accumulator that provides a two's complement and unsigned multiply has a accumulator structure that minimizes the local and global interconnect lengths so that the design readily scales with advanced complementary metal oxide semiconductor (CMOS) integrated circuit (IC) technologies. The multiplier accumulator is formed with a plurality of identical panels, four in the preferred embodiment. Each identical panel implements the unique compensation necessary to form a multiplication/accumulation. The improved panel design is intended to operate in multiple panel designs and achieves a significant performance improvement. The improved multiply accumulator panel compensation is utilized for both two's complement compensation and according to the position of the panel within the overall array structure. Netlist generators are used to compile the netlist for both the main panel and the full adder arrays to sum the panels. This allows rapid generation of multiple arrays suitable for use in single and double precision arithmetic as well as extended precision arithmetic.

Apparatus for performing two's complement and unsigned multiply

An apparatus for use with a computer system for identifying dependencies within a register, which dependencies are established by a succession of instructions for the computer system. The register includes a plurality of cells which may be in a hierarchical arrangement of register storage sets. In its preferred embodiment, the apparatus comprises a storage means for storing a bit map, which bit map is configured to provide bit map identifications identifying designated register storage sets. The bit map represents the hierarchical arrangement. The apparatus further comprises a logic means for logically treating information, which logic means is coupled with the storage means and with the computer system. The logic means receives a first bit map identification from a first instruction (the first bit map identification identifies a first register storage set), and receives a second bit map identification from a second instruction (the second bit map identification identifies a second register storage set. The second instruction is subsequent in the succession to the first instruction. The first register storage set is a first storage locus for storing a target operand of the first instruction; the second register storage set is a second storage locus for storing a source operand for the second instruction. The logical treating effects the identifying of dependencies between the target operand and the source operand.

Method and apparatus for identifying dependencies within a register

An apparatus to dynamically controls the out-of-order execution of load/store instructions by detecting a store violation condition and avoiding the penalty of a pipeline recovery process. The apparatus permits a load and store instruction to issue and execute out of order and incorporates a unique store barrier cache which is used to dynamically predict whether or not a store violation condition is likely to occur and, if so, to restrict the issue of instructions to the load/store unit until the store instruction has been executed and it is once again safe to proceed with out-of-order execution. The method implemented by the apparatus delivers performance within one percent of theoretically possible with apriori knowledge of load and store addresses.

Apparatus to dynamically control the out-of-order execution of

A high speed, compact low power integer adder unit for advanced microprocessors features modular construction, low gate count and a fast add time. A 64-bit implementation is characterized by a unique combination of dual rail logic circuits and dual carry select path within each of four 16-bit adder building blocks to achieve a one gate delay increment for each additional 16-bit adder building block after the first. Each of the 16-bit adder building blocks are composed of modules that receive four of sixteen bits of the operands, and each of the modules are comprised of submodules. The submodules are in turn comprised of dual rail logic circuits with a dual carry select path so as to constitute a nested carry select architecture wherein the nesting of the dual carry select path extends from submodules to a module and from modules to a basic building block. The dual carry select paths are optimized both internal to the submodules and modules and at the submodule and module boundaries to achieve a minimum gate delay number.

Growing community of inventors

Chittenden County, VT, United States of America

James Henry Hesson