Apple Inc.

Hewlett-packard Development Company, L.p.

Andrew J Beaumont-Smith

Aditya Kesiraju

Ran A Chachick

Peter J Bannon

Ramesh B Gunna

Brian P Lilly

John H Mylius

Tyler J Huberty

Krishna C Potnuru

Rajdeep L Bhuyar

Daniel C Murray

Pradeep Kanapathipillai

Boris S Alvarez-Heredia

Sandeep Kumar Gupta

Tse-Yu Yeh

Wei-Han Lien

James Vash

Derek R Kumar

John D Pape

Shih-Chieh Wen

Matthias Knoth

Honkai Tam

Jason M Kassoff

Jeffry E Gonion

Gerard Richard Williams

Ali Sazegari

Toshinari Takayanagi

Eric Jason Furbish

Stephen G Meier

Jaidev Prasad Patwardhan

Brandon H Dwiel

Jung Wook Cho

Eric Bainville

Jong-Suk Lee

Deepankar Duggal

Sridhar Samudrala

Srikanth Balasubramanian

Toshi Takayanagi

An apparatus includes a plurality of processor circuits, a cache memory circuit, and a trace control circuit. The trace control circuit may be configured, in response to activation of a mode to record information indicative of program execution of at least one processor circuit of the plurality of processor circuits, to monitor memory requests transmitted between ones of the plurality of processor circuits and the cache memory circuit, and then to select a particular memory request of monitored memory requests using an arbitration algorithm. The trace control circuit may be further configured to allocate space in a trace buffer to the particular memory request, and to store, in the trace buffer, information associated with the particular memory request.

Debug trace of cache memory requests

In an embodiment, a coprocessor may include a bypass indication which identifies execution circuitry that is not used by a given processor instruction, and thus may be bypassed. The corresponding circuitry may be disabled during execution, preventing evaluation when the output of the circuitry will not be used for the instruction. In another embodiment, the coprocessor may implement a grid of processing elements in rows and columns, where a given coprocessor instruction may specify an operation that causes up to all of the processing elements to operate on vectors of input operands to produce results. Implementations of the coprocessor may implement a portion of the processing elements. The coprocessor control circuitry may be designed to operate with the full grid or partial grid, reissuing instructions in the partial grid case to perform the requested operation. In still another embodiment, the coprocessor may be able to fuse vector mode operations.

Coprocessors with bypass optimization, variable grid architecture, and fused vector operations

A prefetcher for a coprocessor is disclosed. An apparatus includes a processor and a coprocessor that are configured to execute processor and coprocessor instructions, respectively. The processor and coprocessor instructions appear together in code sequences fetched by the processor, with the coprocessor instructions being provided to the coprocessor by the processor. The apparatus further includes a coprocessor prefetcher configured to monitor a code sequence fetched by the processor and, in response to identifying a presence of coprocessor instructions in the code sequence, capture the memory addresses, generated by the processor, of operand data for coprocessor instructions. The coprocessor is further configured to issue, for a cache memory accessible to the coprocessor, prefetches for data associated with the memory addresses prior to execution of the coprocessor instructions by the coprocessor.

Coprocessor prefetcher

A coprocessor with register renaming is disclosed. An apparatus includes a plurality of processors and a coprocessor respectively configured to execute processor instructions and coprocessor instructions. The coprocessor receives coprocessor instructions from ones of the processors. The coprocessor includes an array of processing elements and a result register set comprising storage elements respectively distributed within the array of processing elements. For a given member of the array of processing elements, a corresponding storage element is configured to store coprocessor instruction results generated by the given member. The result register set implements a plurality of contexts to store respective coprocessor states corresponding to coprocessor instructions received from different processors. Based on a determination that one of the contexts is inactive, the coprocessor is configured to store coprocessor instruction results corresponding to an active context within storage elements of the result register set corresponding to the inactive context.

Coprocessor register renaming using registers associated with an inactive context to store results from an active context

A system may include a plurality of processors and a coprocessor. A plurality of coprocessor context priority registers corresponding to a plurality of contexts supported by the coprocessor may be included. The plurality of processors may use the plurality of contexts, and may program the coprocessor context priority register corresponding to a context with a value specifying a priority of the context relative to other contexts. An arbiter may arbitrate among instructions issued by the plurality of processors based on the priorities in the plurality of coprocessor context priority registers. In one embodiment, real-time threads may be assigned higher priorities than bulk processing tasks, improving bandwidth allocated to the real-time threads as compared to the bulk tasks.

Coprocessor context priority

An instruction set architecture including instructions for a processor and instructions for a coprocessor may include synchronizing instructions that may be used to begin and end instruction sequences that include coprocessor instructions (coprocessor sequences). If a terminating synchronizing instruction is followed by an initial synchronizing instruction and the pair are detected in the coprocessor concurrently, the coprocessor may suppress execution of the pair of instructions.

Coprocessor synchronizing instruction suppression

In an embodiment, a coprocessor includes multiple processing elements arranged in a grid of one or more rows and one or more columns. A given processing element includes an arithmetic/logic unit (ALU) circuit configured to perform an ALU operation specified by an instruction executable by the coprocessor, wherein the ALU circuit is configured to produce a result. The given processing element further comprises a first memory coupled to the execute circuit. The first memory is configured to store results generated by the given processing element. The first memory includes a portion of a result memory implemented by the coprocessor, wherein locations in the result memory are specifiable as destination operands of instructions executable by the coprocessor. The portion of the result memory implemented by the first memory is the portion of the result memory that the given processing element is capable of updating.

Coprocessor with distributed register

In an embodiment, a computation engine may offload work from a processor (e.g. a CPU) and efficiently perform computations such as those used in LSTM and other workloads at high performance. In an embodiment, the computation engine may perform computations on input vectors from input memories in the computation engine, and may accumulate results in an output memory within the computation engine. The input memories may be loaded with initial vector data from memory, incurring the memory latency that may be associated with reading the operands. Compute instructions may be performed on the operands, generating results in an output memory. One or more extract instructions may be supported to move data from the output memory to the input memory, permitting additional computation on the data in the output memory without moving the results to main memory.

Computation engine with extract instructions to minimize memory access

A processor and method for maintaining the integrity of an execution return address stack (RAS). The execution RAS is maintained in an accurate state by storing information regarding branch instructions in a branch information table. The first time a branch instruction is executed, an entry is allocated and populated in the table. If the branch instruction is re-executed, a pointer address is retrieved from the corresponding table entry and the execution RAS pointer is repositioned to the retrieved pointer address. The execution RAS can also be used to restore a speculative RAS due to a mis-speculation.

Maintaining the integrity of an execution return address stack

Growing community of inventors

Cambridge, MA, United States of America

Andrew J Beaumont-Smith