公开(公告)号：US20200272520A1

公开(公告)日：2020-08-27

申请号：US16801776

申请日：2020-02-26

Applicant: QUALCOMM Incorporated

Inventor： Richard SENIOR ,  Sundeep KUSHWAHA ,  Harsha Gordhan JAGASIA ,  Christopher AHN ,  Gurvinder Singh CHHABRA ,  Nieyan GENG ,  Maksim KRASNYANSKIY ,  UNNI PRASAD

IPC: G06F9/50 ,  G06F9/48 ,  G06F12/0806

Abstract: A method of managing a stack includes detecting, by a stack manager of a processor, that a size of a frame to be allocated exceeds available space of a first stack. The first stack is used by a particular task executing at the processor. The method also includes designating a second stack for use by the particular task. The method further includes copying metadata associated with the first stack to the second stack. The metadata enables the stack manager to transition from the second stack to the first stack upon detection that the second stack is no longer in use by the particular task. The method also includes allocating the frame in the second stack.

公开(公告)号：US20250156187A1

公开(公告)日：2025-05-15

申请号：US18510088

申请日：2023-11-15

Applicant: QUALCOMM Incorporated

Inventor： Richard SENIOR ,  Venkatarami MORA ,  Suresh Kumar VENKUMAHANTI ,  Avinash GADDE ,  Christopher AHN ,  Sangchol LEE ,  Sanjana RAJEEV ,  Dhairya AGRAWAL

IPC: G06F9/38 ,  G06F9/30

Abstract: Aspects of the disclosure are directed to reduction of data transfer overhead. In accordance with one aspect, increment a function call counter for an executed function call; decrement the function call counter for an executed return from function call; infer the indirect branch address based on a function call counter value and when a cumulative count of received executed atoms indicates a return from function call has been executed; oversaturate the function call counter at a maximum counter value if the function call counter contains the maximum counter value and a subsequent function call is executed; and undersaturate the function call counter at a minimum counter value if the function call counter contains the minimum counter value and a subsequent return from function call is executed.

公开(公告)号：US20190303158A1

公开(公告)日：2019-10-03

申请号：US15940896

申请日：2018-03-29

Applicant: QUALCOMM Incorporated

Inventor： Gurkanwal BRAR ,  Christopher AHN ,  Gurvinder Singh CHHABRA

IPC: G06F9/38 ,  G06F8/41 ,  G06F8/658 ,  G06N3/08

Abstract: Systems and methods for branch prediction include identifying a subset of branch instructions executable by a processor as a neural subset of branch instructions, based on information obtained from using an execution trace, wherein the neural subset of branch instructions are determined to have larger benefit from a neural branch predictor than a non-neural branch predictor. The neural branch predictor is pre-trained for the neural subset based on the execution trace. Annotations are added to the neural subset of branch instructions, wherein the annotations are preserved across software revisions. At runtime, when the neural subset of branch instructions are encountered during any future software revision, the branch instructions thereof are detected as belonging to the neural subset of branch instructions based on the annotations, and the pre-trained neural branch predictor is used for making their branch predictions.

公开(公告)号：US20190087193A1

公开(公告)日：2019-03-21

申请号：US15712112

申请日：2017-09-21

Applicant: QUALCOMM Incorporated

Inventor： Gurkanwal BRAR ,  Christopher AHN ,  Gurvinder Singh CHHABRA

IPC: G06F9/38

CPC classification number: G06F9/3848 ,  G06F9/3806

Abstract: Systems and methods for branch prediction include identifying a subset of branch instructions from an execution trace of instructions executed by a processor. The identified subset of branch instructions have greater benefit from branch predictions made by a neural branch predictor than branch predictions made by a non-neural branch predictor. During runtime, the neural branch predictor is selectively used for obtaining branch predictions of the identified subset of branch instructions. For remaining branch instructions outside the identified subset of branch instructions, branch predictions are obtained from a non-neural branch predictor. Further, a weight vector matrix comprising weight vectors for the identified subset of branch instructions of the neural branch predictor is pre-trained based on the execution trace.