Oracle International Corporation

Sun Microsystems, Inc.

Oracle America, Inc.

Apple Inc.

Yuan C Chou

Santosh Abraham

Lawrence Spracklen

Manish K Shah

Robert T Golla

Deepankar Duggal

Debasish Chandra

Mridul Agarwal

Richard F Russo

John D Pape

Jared C Smolens

Hyunjin Abraham Lee

Paul Joseph Jordan

Jeffrey S Brooks

Eric W Mahurin

Mark Luttrell

Matthew B Smittle

Kulin N Kothari

Niket K Choudhary

Chang Xu

Joseph I Chamdani

Wei-Han Lien

Sorin Iacobovici

Gideon N Levinsky

Sudarshan Kadambi

Jama Ismail Barreh

Douglas C Holman

Viney Gautam

Haoyan Jia

Yanran Yang

Suraj Sudhir

Vijay Sathish

Suryanarayana Murthy Durbhakula

Wei Chung Hsu

Yasuko Watanabe

Richa Aggarwal

Santosh G Abraham

Lawrence A Spracklen

Disclosed techniques relate to re-use of speculative results from an incorrect execution path. In some embodiments, when a first control transfer instruction is mispredicted, a second control transfer instruction may have been executed on the wrong path because of the misprediction. Result storage circuitry may record information indicating a determined direction for the second control transfer instruction. Control flow tracker circuitry may store, for the first control transfer instruction, information indicating a reconvergence point. Re-use control circuitry may track registers written by instructions prior to the reconvergence point, determine, based on the tracked registers, that the second control transfer instruction does not depend on data from any instruction between the first control transfer instruction and the reconvergence point, and use the recorded determined direction for the second control transfer instruction, notwithstanding the misprediction of the first control transfer instruction.

Re-use of speculative control transfer instruction results from wrong path

Disclosed techniques relate to re-use of speculative results from an incorrect execution path. In some embodiments, when a control transfer instruction is mispredicted, a load instruction may have been executed on the wrong path. In disclosed embodiments, result storage circuitry records information that indicates destination registers of speculatively-executed load instructions including a first load instruction. Control flow tracker circuitry may store information indicating a reconvergence point for the control transfer instruction. Re-use control circuitry may track registers written by instructions prior to the reconvergence point, determine that the first load instruction does not depend on data from any instruction between the control transfer instruction and the reconvergence point, and use, as a result of the first load instruction, a value from a recorded destination register that was written based on speculative execution of the first load, notwithstanding the misprediction of the control transfer instruction.

Re-use of speculative load instruction results from wrong path

Techniques are disclosed relating to predicting values for load operations. In some embodiments, front-end circuitry is configured to predict values of load operations based on multiple value tagged geometric length predictor (VTAGE) prediction tables (based on program counter information and branch history information). Training circuitry may adjust multiple VTAGE learning tables based on completed load operations. Control circuitry may pre-compute access information (e.g., an index) for a VTAGE learning table for a load based on branch history information that is available to the front-end circuitry but that is unavailable to the training circuitry, store the pre-computed access information, and provide the pre-computed access information from the first storage circuitry to the training circuitry to access the VTAGE learning table based on completion of the load. This may facilitate VTAGE training without pipelining the branch history information.

Managing table accesses for tagged geometric length (TAGE) load value prediction

Techniques are disclosed relating to load value prediction. In some embodiments, a processor includes learning table circuitry that is shared for both address and value prediction. Loads may be trained for value prediction when they are eligible for both value and address prediction. Entries in the learning table may be promoted to an address prediction table or a load value prediction table for prediction, e.g., when they reach a threshold confidence level in the training table. In some embodiments, the learning table stores a hash of a predicted load value and control circuitry uses a probing load to retrieve the actual predicted load value for the value prediction table.

Shared learning table for load value prediction and load address prediction

A system, e.g., a system on a chip (SOC), may include one or more processors. A processor may execute an instruction synchronization barrier (ISB) instruction to enforce an ordering constraint on instructions. To execute the ISB instruction, the processor may determine whether contexts of the processor required for execution of instructions older than the ISB instruction are consumed for the older instructions. Responsive to determining that the contexts are consumed for the older instructions, the processor may initiate fetching of an instruction younger than the ISB instruction, without waiting for the older instructions to retire.

Processing of synchronization barrier instructions

A system and method for efficiently reducing the latency of load operations. In various embodiments, logic of a processor accesses a prediction table after fetching instructions. For a prediction table hit, the logic executes a load instruction with a retrieved predicted address from the prediction table. For a prediction table miss, when the logic determines the address of the load instruction and hits in a learning table, the logic updates a level of confidence indication to indicate a higher level of confidence when a stored address matches the determined address. When the logic determines the level of confidence indication stored in a given table entry of the learning table meets a threshold, the logic allocates, in the prediction table, information stored in the given entry. Therefore, the predicted address is available during the next lookup of the prediction table.

Early load execution via constant address and stride prediction

A system for prefetching data for a processor includes a processor core, a memory configured to store information for use by the processor core, a cache memory configured to fetch and store information from the memory, and a prefetch circuit. The prefetch circuit may be configured to issue a multi-group prefetch request to retrieve information from the memory to store in the cache memory using a predicted address. The multi-group prefetch request may include a depth value indicative of a number of fetch groups to retrieve. The prefetch circuit may also be configured to generate an accuracy value based on a cache hit rate of prefetched information over a particular time interval, and to modify the depth value based on the accuracy value.

Multi-line data prefetching using dynamic prefetch depth

An apparatus includes a branch target cache configured to store one or more branch addresses, a memory configured to store a return target stack, and a circuit. The circuit may be configured to determine, for a group of one or more fetched instructions, a prediction value indicative of whether the group includes a return instruction. In response to the prediction value indicating that the group includes a return instruction, the circuit may be further configured to select a return address from the return target stack. The circuit may also be configured to determine a hit or miss indication in the branch target cache for the group, and to, in response to receiving a miss indication from the branch target cache, select the return address as a target address for the return instruction.

Method for reducing fetch cycles for return-type instructions

An apparatus is disclosed, the apparatus including a branch target cache memory configured to store one or more entries. Each entry of the one or more entries may include an address tag and a corresponding target address. The apparatus may also include a control circuit configured to check for at least one taken branch instruction in a group of one or more instructions fetched using a current address. The control circuit may be further configured to generate an address tag corresponding to the group of one or more instructions using another address used prior to the current address in response to a determination that the group of one or more instructions includes a taken branch instruction. In addition, the control circuit may be configured to store the corresponding address tag and a target address associated with the taken branch instruction in a particular entry in the branch target cache memory.

Hybrid lookahead branch target cache

An system for prefetching data for a processor includes a processor core, a memory, a cache memory, and a prefetch circuit. The memory may be configured to store information for use by the processor core. The cache memory may be configured to issue a fetch request for information from the memory for use by the processor core. The prefetch circuit may be configured to issue a prefetch request for information from the memory to store in the cache memory using a predicted address, and to monitor, over a particular time interval, an amount of fetch requests from the cache memory and prefetch requests from the prefetch circuit. The prefetch circuit may also be configured to disable prefetch requests from the memory for a subsequent time interval in response to a determination that the amount satisfies a threshold amount.

Utilization-based throttling of hardware prefetchers

An apparatus includes a buffer configured to store a plurality of instructions previously fetched from a memory, wherein each instruction of the plurality of instructions may be included in a respective thread of a plurality of threads. The apparatus also includes control circuitry configured to select a given thread of the plurality of threads dependent upon a number of instructions in the buffer that are included in the given thread. The control circuitry is also configured to fetch a respective instruction corresponding to the given thread from the memory, and to store the respective instruction in the buffer.

Optimizing thread selection at fetch, select, and commit stages of processor core pipeline

The disclosed embodiments relate to a method for controlling prefetching in a processor to prevent over-saturation of interfaces in the memory hierarchy of the processor. While the processor is executing, the method determines a bandwidth utilization of an interface from a cache in the processor to a lower level of the memory hierarchy. Next, the method selectively adjusts a prefetch-dropping high-water mark for occupancy of a miss buffer associated with the cache based on the determined bandwidth utilization, wherein the miss buffer stores entries for outstanding demand requests and prefetches that missed in the cache and are waiting for corresponding data to be returned from the lower level of the memory hierarchy, and wherein when the occupancy of the miss buffer exceeds the prefetch-dropping high-water mark, subsequent prefetches that cause a cache miss are dropped.

Prefetch bandwidth throttling by dynamically adjusting miss buffer prefetch-dropping thresholds

A system includes a memory, a cache including multiple cache lines; and a processor. The processor may be configured to retrieve, from a first cache line, a first instruction to store data in a memory location at an address in the memory. The processor may be configured to retrieve, from a second cache line, a second instruction to read the memory location at the address in the memory. The second instruction may be retrieved after the first instruction. The processor may be configured to execute the second instruction at a first time dependent upon a value of a first entry in a table, wherein the first entry is selected dependent upon a value in the second cache line. The processor may be configured to execute the first instruction at a second time, after the first time, and invalidate the second instruction at a third time, after the second time.

Method for load instruction speculation past older store instructions

The disclosed embodiments relate to a system that generates prefetches for a stream of data accesses with multiple strides. During operation, while a processor is generating the stream of data accesses, the system examines a sequence of strides associated with the stream of data accesses. Next, upon detecting a pattern having a single constant stride in the examined sequence of strides, the system issues prefetch instructions to prefetch a sequence of data cache lines consistent with the single constant stride. Similarly, upon detecting a recurring pattern having two or more different strides in the examined sequence of strides, the system issues prefetch instructions to prefetch a sequence of data cache lines consistent with the recurring pattern having two or more different strides.

Facilitating prefetching for data streams with multiple strides

Growing community of inventors

Mountain View, CA, United States of America

Yuan C Chou