Name
Abstract | ||
|---|---|---|
Internet of Things edge intelligence requires convolutional neural network (CNN) inference to take place in the edge devices itself.
<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ARM big.LITTLE</italic>
architecture is at the heart of prevalent commercial edge devices. It comprises of single-ISA heterogeneous cores grouped into multiple homogeneous clusters that enable power and performance tradeoffs. All cores are expected to be simultaneously employed in inference to attain maximal throughput. However, high communication overhead involved in parallelization of computations from convolution kernels across clusters is detrimental to throughput. We present an alternative framework called
<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Pipe-it</italic>
that employs pipelined design to split convolutional layers across clusters while limiting parallelization of their respective kernels to the assigned cluster. We develop a performance-prediction model that utilizes only the convolutional layer descriptors to predict the execution time of each layer individually on all permitted core configurations (type and count).
<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Pipe-it</italic>
then exploits the predictions to create a balanced pipeline using an efficient design space exploration algorithm.
<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Pipe-it</italic>
on average results in a 39% higher throughput than the highest antecedent throughput. |
Year | DOI | Venue |
|---|---|---|
2020 | 10.1109/TCAD.2019.2944584 | IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems |
Keywords | DocType | Volume |
Multicore processing,Throughput,Kernel,Pipelines,Integrated circuit modeling,Streaming media | Journal | 39 |
Issue | ISSN | Citations |
10 | 0278-0070 | 9 |
PageRank | References | Authors |
0.62 | 0 | 6 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Siqi Wang | 1 | 37 | 5.39 |
| Gayathri Ananthanarayanan | 2 | 16 | 2.24 |
| Yifan Zeng | 3 | 26 | 2.51 |
| Neeraj Goel | 4 | 9 | 3.33 |
| Anuj Pathania | 5 | 181 | 14.97 |
| Tulika Mitra | 6 | 2714 | 135.99 |