Advanced Micro Devices Corporation

Krishnan V Ramani

John Kalamatianos

Kai Troester

Susumu Mashimo

Stuart F Oberman

Norbert Juffa

Marius Evers

David N Suggs

Michael David Achenbach

Frank C Galloway

Betty Ann McDaniel

Fred Weber

Scott Thomas Bingham

Mike Butler

Gary R Lauterbach

Benjamin T Sander

Mitchell K Alsup

Emil Talpes

Jay E Fleischman

Michael Estlick

Debjit Das Sarma

Ganesh Venkataramanan

Chetana Nagendra Keltcher

Ramsey W Haddad

David Hugh McIntyre

Thomas Kunjan

Chitresh Chandra Narasimhaiah

Ravi Krishna Cherukuri

Ravi Krishna

Chun Liu

Address-based filtering for load/store speculation includes maintaining a filtering table including table entries associated with ranges of addresses; in response to receiving an ordering check triggering transaction, querying the filtering table using a target address of the ordering check triggering transaction to determine if an instruction dependent upon the ordering check triggering transaction has previously been generated a physical address; and in response to determining that the filtering table lacks an indication that the instruction dependent upon the ordering check triggering transaction has previously been generated a physical address, bypassing a lookup operation in an ordering violation memory structure to determine whether the instruction dependent upon the ordering check triggering transaction is currently in-flight.

Address-based filtering for load/store speculation

Speculative execution using a page-level tracked load order queue includes: determining that a first load instruction targets a determined memory region; and in response to the first load instruction targeting the determined memory region, adding an entry to a page-level tracked load order queue instead of a load order queue, where the entry indicates a page address of a target of the first load instruction.

Speculative execution using a page-level tracked load order queue

A technique for speculatively executing load-dependent instructions includes detecting that a memory ordering consistency queue is full for a completed load instruction. The technique also includes storing data loaded by the completed load instruction into a storage location for storing data when the memory ordering consistency queue is full. The technique further includes speculatively executing instructions that are dependent on the completed load instruction. The technique also includes in response to a slot becoming available in the memory ordering consistency queue, replaying the load instruction. The technique further includes in response to receiving loaded data for the replayed load instruction, testing for a data mis-speculation by comparing the loaded data for the replayed load instruction with the data loaded by the completed load instruction that is stored in the storage location.

Speculative instruction wakeup to tolerate draining delay of memory ordering violation check buffers

A method and apparatus generates control information that indicates whether to change a counter value associated with a particular load instruction. In response to the control information, the method and apparatus causes a hysteresis effect for operating between a speculative mode and a non-speculative mode based on the counter value. The hysteresis effect is in favor of the non-speculative mode. The method and apparatus causes the hysteresis effect by incrementing the counter value associated with the particular load instruction by a first value or decrementing the counter value by a second value. The first value is greater than the second value.

Apparatus and method for resynchronization prediction with variable upgrade and downgrade capability

A system and method for tracking stores and loads to reduce load latency when forming the same memory address by bypassing a load store unit within an execution unit is disclosed. Store-load pairs which have a strong history of store-to-load forwarding are identified. Once identified, the load is memory renamed to the register stored by the store. The memory dependency predictor may also be used to detect loads that are dependent on a store but cannot be renamed. In such a configuration, the dependence is signaled to the load store unit and the load store unit uses the information to issue the load after the identified store has its physical address.

Tracking stores and loads by bypassing load store units

A processor and method for handling lock instructions identifies which of a plurality of older store instructions relative to a current lock instruction are able to be locked. The method and processor lock the identified older store instructions as an atomic group with the current lock instruction. The method and processor negatively acknowledge probes until all of the older store instructions in the atomic group have written to cache memory. In some implementations, an atomic grouping unit issues an indication to lock identified older store instructions that are retired and lockable, and in some implementations, also issues an indication to lock older stores that are determined to be lockable that are non-retired.

Processor with accelerated lock instruction operation

A system and method for tracking stores and loads to reduce load latency when forming the same memory address by bypassing a load store unit within an execution unit is disclosed. The system and method include storing data in one or more memory dependent architectural register numbers (MdArns), allocating the one or more MdArns to a MEMFILE, writing the allocated one or more MdArns to a map file, wherein the map file contains a MdArn map to enable subsequent access to an entry in the MEMFILE, upon receipt of a load request, checking a base, an index, a displacement and a match/hit via the map file to identify an entry in the MEMFILE and an associated store, and on a hit, providing the entry responsive to the load request from the one or more MdArns.

A method for reducing a number of operations replayed in a processor includes decoding an operation to determine a memory address and a command in the operation. If data is not in a way predictor based on the memory address, a suppress wakeup signal is sent to an operation scheduler, and the operation scheduler suppresses waking up other operations that are dependent on the data.

Replay reduction by wakeup suppression using early miss indication

Methods and apparatuses are provided for increased efficiency in a processor via control word prediction. The apparatus comprises an operational unit capable of determining whether an instruction will change a first control word to a second control word for processing dependent instructions. Execution units process the dependent instructions using a predicted control word and compare the second control word to the predicted control word. A scheduling unit causes the execution units to reprocess the dependent instructions when the predicted control word does not match the second control word. The method comprises determining that an instruction will change a first control word to a second control word and processing the dependent instructions using a predicted control word. The second control word is compared to the predicted control word and the dependent instructions are reprocessed using the second control word when the predicted control word does not match the second control word.

Processor with increased efficiency via control word prediction

A method and a processor load/store unit (LSU) are described for performing store-to-load forwarding (STLF) from an interlocking store. STLF is performed when a starting address of the store and the load do not match, or when a data size of the store is smaller than a data size of the load. The LSU detects a load that interlocks with a store, and determines whether all or only a portion of data bytes needed by the load can be provided by the interlocking store. If it is determined that only a portion of the data bytes needed by the load can be provided by the interlocking store, then that portion of the data bytes is provided by a store data buffer (SDB) and the remaining portion of the data bytes needed by the load is provided by a data cache (DC). Otherwise, the SDB provides all of the data bytes.

Method and apparatus for performing store-to-load forwarding from an interlocking store using an enhanced load/store unit in a processor

A method and microprocessor are described for efficiently executing load instructions out-of-order (speculatively). The microprocessor includes an enhanced load store unit (LSU) and an enhanced instruction decoder. The enhanced LSU receives a plurality of out-of-order value addresses, and sends a resync signal to the enhanced instruction decoder when an execution error associated with a particular load instruction occurs. The enhanced instruction decoder stores a specific address associated with the particular load instruction, and increments a counter value that indicates how many times the resync signal was sent by the resync predictor. When the counter value reaches a predetermined threshold, subsequent load instructions from the specific address are executed in order (non-speculatively). When a future execution of the particular load instruction indicates that the probability of an execution error has been reduced, the counter value is decremented, facilitating newer load instructions associated with the same address to again be executed speculatively.

Method and apparatus for processing load instructions in a microprocessor having an enhanced instruction decoder and an enhanced load store unit

A system and method for data forwarding from a store instruction to a load instruction during out-of-order execution, when the load instruction address matches against multiple older uncommitted store addresses or if the forwarding fails during the first pass due to any other reason. In a first pass, the youngest store instruction in program order of all store instructions older than a load instruction is found and an indication to the store buffer entry holding information of the youngest store instruction is recorded. In a second pass, the recorded indication is used to index the store buffer and the store bypass data is forwarded to the load instruction. Simultaneously, it is verified if no new store, younger than the previously identified store and older than the load has not been issued due to out-of-order execution.

System and method of load-store forwarding

A system and method for linking speculative results of load operations to register values. A system includes a memory file including an entry configured to store a first addressing pattern and a first tag. The memory file is configured to compare the first addressing pattern to a second addressing pattern of a load operation, and to link a data value identified by the first tag to a speculative result of the load operation if there is a match. The system further includes an execution core coupled to the memory file and configured to access the speculative result when executing a second operation that is dependent on the load operation, and a load store unit coupled to the memory file and configured to verify the link between the data value and the speculative result of the load operation by performing a comparison between one or more addresses.

System and method for validating a memory file that links speculative results of load operations to register values

A multimedia execution unit configured to perform vectored floating point and integer instructions. The execution unit may include an add/subtract pipeline having far and close data paths. The far path is configured to handle effective addition operations and effective subtraction operations for operands having an absolute exponent difference greater than one. The close path is configured to handle effective subtraction operations for operands having an absolute exponent difference less than or equal to one. The close path is configured to generate two output values, wherein one output value is the first input operand plus an inverted version of the second input operand, while the second output value is equal to the first output value plus one. Selection of the first or second output value in the close path effectuates the round-to-nearest operation for the output of the adder.

Growing community of inventors

Santa Clara, CA, United States of America

Krishnan V Ramani