Accelerator for processing inference tasks in parallel and operating method thereof
AssigneesSamsung Electronics Co., Ltd., Industry-Academic Cooperation Foundation, Yonsei University
InventorsHo Young Kim, Won Woo Ro, Sung Ji Choi
Life of this patent
Estimating the term…
Abstract
Disclosed is an accelerator and a method of operating the accelerator including determining whether any group shares weights of a first group from among groups, determining a presence of an idle processing element (PE) array, in response to no group sharing the weights of the first group, and selecting a second group having a memory time overlapping a computation time of the first group from among the groups, in response to the idle PE array being present.
Claims20
A method of operating an accelerator, the method comprising:
- determining whether any group shares weights of a first group from among groups;
- determining a presence of an idle processing element (PE) array, in response to no group sharing the weights of the first group;
- and
- selecting a second group having a memory time overlapping a computation time of the first group from among the groups, in response to the idle PE array being present.
An accelerator, comprising:
- processing element (PE) arrays;
- and
- a scheduler configured to determine whether any group shares weights of a first group among groups, to determine a presence of an PE array from among the PE arrays, in response to no group sharing the weights of the first group, and to select a second group having a memory time overlapping a computation time of the first group from among the groups, in response to the idle PE array being present.
Dependencies are inferred from the claim text.