International Business Machines Corporation

David R Terry

Dung Q Nguyen

Sundeep Chadha

Brian David Barrick

Michael Joseph Genden

Joshua W Bowman

Cliff Kucharski

Steven J Battle

Dhivya Jeganathan

Hung Qui Le

Eula A Tolentino

Sam G Chu

Brian William Thompto

Saiful Islam

Jing Zhang

Salma Ayub

James Wilson Bishop

Owen Chiang

Marcy Evelyn Byers

Susan Elizabeth Eisen

Maarten Jakob Boersma

Jose Angel Paredes

Maureen Anne Delaney

Thao T Doan

Rokesh Jayasundar

Albert James Van Norstrand, Jr

Kurt Alan Feiste

Jentje Leenstra

Michael Klaus Kroener

Christopher M Mueller

David S Walder

Jerry Y Lu

Nasrin Sultana

Kenneth Lundy Ward

Niels Fricke

Tu-An T Nguyen

Glenn Owen Kincaid

Binglong Zhang

Howard Lawrence Levy

Eula Faye Tolentino

An approach is provided in which a system writes a set of data into a register file entry that includes a first memory array and a second memory array. The register file entry also includes a set of first write ports corresponding to the first memory array and a set of second write ports corresponding to the second memory array. The system configures a selection bit based on determining that a selected one of the set of first write ports is utilized to store the set of data in the first memory array. In turn, the system reads the set of data out of the first memory array based on the configured selection bit.

Implementing write ports in register-file array cell

Operation of a multi-slice processor that includes execution slices and a dispatch network of the multi-slice processor implementing a hardware level transfer of an execution thread between execution slices. Operation of such a multi-slice processor includes responsive to a thread switch signal: halting dispatch of one or more instructions retrieved from an instruction cache; generating a plurality of instructions to transfer an execution thread from a first execution slice to a second execution slice; and dispatching the plurality of instructions instead of the one or more instructions retrieved from the instruction cache; and transferring, in dependence upon execution of the plurality of instructions from the thread switching instruction generator, the execution thread from the first execution slice to the second execution slice.

Operation of a multi-slice processor implementing a hardware level transfer of an execution thread

A non-limiting example of a computer-implemented method for file register writes using pointers includes, responsive to a dispatch instruction, storing, at a location in a history buffer, an instruction tag and first data associated with the instruction tag. The method further includes storing a pointer in an issue queue. The pointer points to the location in the history buffer. The method further includes performing a write back of second data using the pointer stored in the issue queue. The write back writes the second data into the location of the history buffer associated with the pointer.

Register file write using pointers

Techniques are disclosed for performing issue queue snooping for an asynchronous flush and restore of a history buffer (HB) in a processing unit. One technique includes identifying an entry of the HB to restore to a register file in the processing unit. A restore ITAG of the HB entry is sent to the register file via a first restore bus, and restore data of the HB entry and the restore ITAG is sent to the register file via a second restore bus. After the restore ITAG and restore data are sent, an instruction is dispatched before the register file obtains the restore data. After it is determined that the restore data is still available via the second restore bus, a snooping operation is performed to obtain the restore data from the second restore bus for the dispatched instruction.

Issue queue snooping for asynchronous flush and restore of distributed history buffer

A computer processing system is provided. The computer processing system includes a processor configured to insert a move instruction in an instruction scheduling dependency graph generated for operands of the instruction operating on register pairs to ensure operand readiness for all of the operands used by the instruction operating on register pairs.

Low-overhead, low-latency operand dependency tracking for instructions operating on register pairs in a processor core

Reducing power consumption in a multi-slice computer processor that includes a re-order buffer and an architected register file, including: designating an entry in the re-order buffer as being invalid and unwritten; assigning a pending instruction to the entry in the re-order buffer; responsive to assigning the pending instruction to the entry in the re-order buffer, designating the entry as being valid; writing data generated by executing the pending instruction into the re-order buffer; and responsive to writing data generated by executing the pending instruction into the re-order buffer, designating the entry as being written.

Reducing power consumption in a multi-slice computer processor

Embodiments include systems, methods, and computer program products for using a multi-level history buffer (HB) for a speculative transaction. One method includes after dispatching a first instruction indicating start of the speculative transaction, marking one or more register file (RF) entries as pre-transaction memory (PTM), and after dispatching a second instruction targeting one of the marked RF entries, moving data from the marked RF entry to a first level HB entry and marking the first level HB entry as PTM. The method also includes upon detecting a write back to the first level HB entry, moving data from the first level HB entry to a second level HB entry and marking the second level HB entry as PTM. The method further includes upon determining that the second level HB entry has been completed, moving data from the second level HB entry to a third level HB entry.

Multi-level history buffer for transaction memory in a microprocessor

Embodiments include systems, methods, and computer program products for on-demand error detection and correction of registers in a processor. One method includes detecting, before a first instruction is dispatched to an issue queue in the processor, an error in data, associated with the first instruction, stored in an entry of a register file in the processor. The method also includes, after detecting the error, halting the dispatch of the first instruction to the issue queue, and determining whether the entry of the register file has completed. The method further includes determining whether to perform error correction on the register file based on the determination of whether the entry of the register file has completed.

On-demand GPR ECC error detection and scrubbing for a multi-slice microprocessor

Techniques are disclosed for performing a flush and restore of a history buffer (HB) in a processing unit. One technique includes identifying one or more entries of the HB to restore to a register file in the processing unit. For each of the one or more HB entries, a determination is made whether to send the HB entry to the register file via a first restore bus or via a second restore bus, different from the first restore bus, based on contents of the HB entry. Each of the one or more HB entries is then sent to the register file via one of the first restore bus or the second restore bus, based on the determination.

Asynchronous flush and restore of distributed history buffer

Methods and apparatus for preventing premature reads from a general purpose register (GPR) including receiving an instruction comprising a source operand identifying a source GPR entry; setting a read-enabled flag based on a value in a particular entry of a source ready vector; if the read-enabled flag indicates data in the source GPR entry is ready for reading, dispatching the received instruction, including performing a read operation of the data in the source GPR entry; and if the read-enabled flag indicates data in the source GPR entry is not ready for reading, dispatching the received instruction without performing a read operation of the data in the source GPR entry.

Preventing premature reads from a general purpose register

Method and system for restoring results to a register file of a processing unit is provided. An instruction is dispatched in a processing slice of the processing unit, targeting a register file, wherein the processing unit includes two or more processing slices, each processing slice including a corresponding history buffer and at least a portion of a register file. The processing unit evicts previous result data from the register file entry to a history buffer corresponding to the processing slice, by writing new result data into the register file entry, in response to the instruction. The processing unit detects a trigger condition relating to a rollback of the processing unit to a previous state, and restores the previous result data from the history buffer to the register file entry, in response to the trigger.

Method and apparatus for execution of threads on processing slices using a history buffer for recording architected register data

Method and system for restoring data to a register file of a processing unit are provided. A history buffer entry (HBE) is marked for restoration to a register file entry. Result data and control information is sent from the HBE to an Issue Queue (ISQ). The ISQ issues an instruction for loading the result data into the register file entry based on the control information. A write back operation is performed to restore the result data to the register file entry, in response to issuing of the instruction.

Method and apparatus for execution of threads on processing slices using a history buffer for restoring architected register data via issued instructions

Operation of a multi-slice processor that includes execution slices and load/store slices coupled via a results bus includes: receiving, by an execution slice, a producer instruction, including: storing, in an entry of an issue queue, the producer instruction; and storing, in a register, an issue queue entry identifier representing the entry of the issue queue in which the producer instruction is stored; receiving, by the execution slice, a source instruction, the source instruction dependent upon the result of the producer instruction, including: storing, in another entry of the issue queue, the source instruction and the issue queue entry identifier of the producer instruction; determining in dependence upon the issue queue entry identifier of the producer instruction that the producer instruction has issued from the issue queue; and responsive to the determination that the producer instruction has issued from the issue queue, issuing the source instruction from the issue queue.

Multi-slice processor issue of a dependent instruction in an issue queue based on issue of a producer instruction

Growing community of inventors

Austin, TX, United States of America

David R Terry