Apple Inc.

P.a. Semi, Inc.

Broadcom Corporation

Po-Yung Chang

Sudarshan Kadambi

Tse-Yu Yeh

Ramesh B Gunna

Wei-Han Lien

Eric Hao

Puneet Kumar

James B Keller

Daniel C Murray

Zongjian Chen

Mark Pearce

Anup S Mehta

Jesse Pan

Hari Kannan

Sridhar Subramanian

Pradeep Kanapathipillai

Peter J Bannon

Rajat Goel

Tse-Yuh Yeh

Ming-Ta Hsu

Techniques are disclosed relating to ordering of load instructions in a weakly-ordered memory model. In one embodiment, a processor includes a cache with multiple cache lines and a store queue configured to maintain status information associated with a store instruction that targets a location in one of the cache lines. In this embodiment, the processor is configured to set an indicator in the status information in response to migration of the targeted cache line. The indicator may be usable to sequence performance of load instructions that are younger than the store instruction. For example, the processor may be configured to wait, based on the indicator, to perform a younger load instruction that targets the same location as the store instruction until the store instruction is removed from the store queue. This may prevent forwarding of the value of the store instruction to the younger load and preserve load-load ordering.

Load ordering in a weakly-ordered processor

In one embodiment, a processor may be configured to write ECC granular stores into the data cache, while non-ECC granular stores may be merged with cache data in a memory request buffer. In one embodiment, a processor may be configured to detect that a victim block writeback hits one or more stores in a memory request buffer (or vice versa) and may convert the victim block writeback to a fill. In one embodiment, a processor may speculatively issue stores that are subsequent to a load from a load/store queue, but prevent the update for the stores in response to a snoop hit on the load.

Store handling in a processor

In one embodiment, a processor comprises a prefetch unit coupled to a data cache. The prefetch unit is configured to concurrently maintain a plurality of separate, active prefetch streams. Each prefetch stream is either software initiated via execution by the processor of a dedicated prefetch instruction or hardware initiated via detection of a data cache miss by one or more load/store memory operations. The prefetch unit is further configured to generate prefetch requests responsive to the plurality of prefetch streams to prefetch data in to the data cache. In an embodiment, the prefetch unit is configured to check for a cache hit for a prefetch request by checking a duplicate cache tags.

Data prefetch unit utilizing duplicate cache tags

In an embodiment, a processor may be configured to detect a store exclusive operation followed by a memory barrier operation in a speculative instruction stream being executed by the processor. The processor may fuse the store exclusive operation and the memory barrier operation, creating a fused operation. The fused operation may be transmitted and globally ordered, and the processor may complete both the store exclusive operation and the memory barrier operation in response to the fused operation. As the fused operation progresses through the processor and one or more other components (e.g. caches in the cache hierarchy) to the ordering point in the system, the fused operation may push previous memory operations to effect the memory barrier operation. In some embodiments, the latency for completing the store exclusive operation and the subsequent data memory barrier operation may be reduced if the store exclusive operation is successful at the ordering point.

Fused store exclusive/memory barrier operation

In one embodiment, a processor comprises a scheduler configured to issue a first instruction operation to be executed and an execution core coupled to the scheduler. Configured to execute the first instruction operation, the execution core comprises a plurality of replay sources configured to cause a replay of the first instruction operation responsive to detecting at least one of a plurality of replay cases. The scheduler is configured to inhibit issuance of the first instruction operation subsequent to the replay for a subset of the plurality of replay cases. The scheduler is coupled to receive an acknowledgement indication corresponding to each of the plurality of replay cases in the subset, and is configured to inhibit issuance of the first instruction operation until the acknowledgement indication is asserted that corresponds to an identified replay case of the subset.

Replay reduction for power saving

In one embodiment, a processor comprises a retire unit and a load/store unit coupled thereto. The retire unit is configured to retire a first store memory operation responsive to the first store memory operation having been processed at least to a pipeline stage at which exceptions are reported for the first store memory operation. The load/store unit comprises a queue having a first entry assigned to the first store memory operation. The load/store unit is configured to retain the first store memory operation in the first entry subsequent to retirement of the first store memory operation if the first store memory operation is not complete. The queue may have multiple entries, and more than one store may be retained in the queue after being retired by the retire unit.

Early release of resources by proceeding to retire store operations from exception reporting stage but keeping in load/store queue

In one embodiment, a processor comprises a circuit coupled to receive an indication of a memory operation to be executed in the processor. The circuit is configured to predict whether or not the memory operation is misaligned. A number of accesses performed by the processor to execute the memory operation is dependent on whether or not the circuit predicts the memory operation as misaligned. In another embodiment, a misalignment predictor is coupled to receive an indication of a memory operation, and comprises a memory and a control circuit coupled to the memory. The memory is configured to store a plurality of indications of memory operations previously detected as misaligned during execution in a processor. The control circuit is configured to predict whether or not a memory operation is misaligned responsive to a comparison of the received indication and the plurality of indications stored in the memory.

Misalignment predictor

In one embodiment, an apparatus comprises a queue comprising a plurality of entries and a control unit coupled to the queue. The control unit is configured to allocate a first queue entry to a store memory operation, and is configured to write a first even offset, a first even mask, a first odd offset, and a first odd mask corresponding to the store memory operation to the first entry. A group of contiguous memory locations are logically divided into alternately-addressed even and odd byte ranges. A given store memory operation writes at most one even byte range and one adjacent odd byte range. The first even offset identifies a first even byte range that is potentially written by the store memory operation, and the first odd offset identifies a first odd byte range that is potentially written by the store memory operation. The first even mask identifies bytes within the first even byte range that are written by the store memory operation, and wherein the first odd mask identifies bytes within the first odd byte range that are written by the store memory operation.

Efficient encoding for detecting load dependency on store with misalignment

In one embodiment, a processor comprises a prefetch unit coupled to a data cache. The prefetch unit is configured to concurrently maintain a plurality of separate, active prefetch streams. Each prefetch stream is either software initiated via execution by the processor of a dedicated prefetch instruction or hardware initiated via detection of a data cache miss by one or more load/store memory operations. The prefetch unit is further configured to generate prefetch requests responsive to the plurality of prefetch streams to prefetch data in to the data cache.

Prefetch unit

In one embodiment, a processor comprises a prediction circuit and another circuit coupled to the prediction circuit. The prediction circuit is configured to predict whether or not a first load instruction will experience a partial store to load forward (PSTLF) event during execution. A PSTLF event occurs if a plurality of bytes, accessed responsive to the first load instruction during execution, include at least a first byte updated responsive to a previous uncommitted store operation and also include at least a second byte not updated responsive to the previous uncommitted store operation. Coupled to receive the first load instruction, the circuit is configured to generate one or more load operations responsive to the first load instruction. The load operations are to be executed in the processor to execute the first load instruction, and a number of the load operations is dependent on the prediction by the prediction circuit.

Partial load/store forward prediction

Replaying memory operation assigned a load/store buffer entry occupied by store operation processed beyond exception reporting stage and retired from scheduler

Growing community of inventors

Saratoga, CA, United States of America

Po-Yung Chang