Stickification using anywhere padding to accelerate data manipulation
Embodiments are provided for efficient realization of memory-bound operations in a computing system by a processor. Data may be read from and written to a memory at a granular level using a stickification operation. One or more regions of activation and weight tensor data on the memory may be annotated by coupling the stickification operation with padding.
1 . A method for providing efficient realization of memory-bound operations a computing environment by one or more processors comprising:
reading or writing data from a memory at a granular level using a stickification operation; and
annotating regions of activation and weight tensor data on the memory by coupling the stickification operation with padding to accelerate data manipulation operations.
2 . The method of claim 1 , further including:
unstickifying one or more inputs of a tensor; and
restickifying the one or more inputs of the tensors.
3 . The method of claim 1 , further including dividing one or more tensors in one or more of a plurality of dimensions.
4 . The method of claim 1 , further including providing a padded region to be spread in a plurality of dimensions on a data stick.
5 . The method of claim 1 , further including transposing a tensor by shuffling a plurality of dimensions on a data stick, wherein the data stick format is unaltered.
6 . The method of claim 1 , further including reshaping one or more dimensions of the tensor into a single dimension or a plurality of dimensions.
7 . The method of claim 1 , further including squeezing a dimension of a tensor having a size equal to a defined value.
8 . A system for providing efficient realization of memory-bound operations in a computing environment, comprising:
one or more computers with executable instructions that when executed cause the system to:
read or writing data from a memory at a granular level using a stickification operation; and
annotate regions of activation and weight tensor data on the memory by coupling the stickification operation with padding.
9 . The system of claim 8 , wherein the executable instructions when executed cause the system to:
unstick one or more inputs of a tensor; and
restick the one or more inputs of the tensors.
10 . The system of claim 8 , wherein the executable instructions when executed cause the system to divide one or more tensors in one or more of a plurality of dimensions.
11 . The system of claim 8 , wherein the executable instructions when executed cause the system to provide a padded region to be spread in a plurality of dimensions on a data stick.
12 . The system of claim 8 , wherein the executable instructions when executed cause the system to transpose a tensor by shuffling a plurality of dimensions on a data stick, wherein the data stick format is unaltered.
13 . The system of claim 8 , wherein the executable instructions when executed cause the system to reshape one or more dimensions of the tensor into a single dimension or a plurality of dimensions.
14 . The system of claim 8 , wherein the executable instructions when executed cause the system to squeeze a dimension of a tensor having a size equal to a defined value.
15 . A computer program product for providing efficient realization of memory-bound operations in a computing environment, the computer program product comprising:
one or more computer readable storage media, and program instructions collectively stored on the one or more computer readable storage media, the program instruction comprising:
program instructions to read or writing data from a memory at a granular level using a stickification operation; and
program instructions to annotate regions of activation and weight tensor data on the memory by coupling the stickification operation with padding to accelerate data manipulation operations.
16 . The computer program product of claim 15 , further including program instructions to:
unstick one or more inputs of a tensor; and
restick the one or more inputs of the tensors.
17 . The computer program product of claim 15 , further including program instructions to divide one or more tensors in one or more of a plurality of dimensions.
18 . The computer program product of claim 15 , further including program instructions to provide a padded region to be spread in a plurality of dimensions on a data stick.
19 . The computer program product of claim 15 , further including program instructions to transpose a tensor by shuffling a plurality of dimensions on a data stick, wherein the data stick format is unaltered.
20 . The computer program product of claim 15 , further including program instructions to:
reshape one or more dimensions of the tensor into a single dimension or a plurality of dimensions; and
squeeze a dimension of a tensor having a size equal to a defined value.