US 11,922,178 B2
Methods and apparatus to load data within a machine learning accelerator
Arnab Raha, Santa Clara, CA (US); Deepak Mathaikutty, Chandler, AZ (US); Debabrata Mohapatra, Santa Clara, CA (US); Sang Kyun Kim, San Jose, CA (US); Gautham Chinya, Sunnyvale, CA (US); and Cormac Brick, San Francisco, CA (US)
Assigned to Intel Corporation, Santa Clara, CA (US)
Filed by Intel Corporation, Santa Clara, CA (US)
Filed on Jun. 25, 2021, as Appl. No. 17/359,392.
Prior Publication US 2021/0326144 A1, Oct. 21, 2021
Int. Cl. G06F 9/445 (2018.01); G06F 9/30 (2018.01); G06F 9/50 (2006.01); G06N 20/00 (2019.01); H03K 19/177 (2020.01); H03K 19/20 (2006.01)
CPC G06F 9/445 (2013.01) [G06F 9/3001 (2013.01); G06F 9/5027 (2013.01); G06N 20/00 (2019.01); H03K 19/177 (2013.01); H03K 19/20 (2013.01)] 20 Claims
OG exemplary drawing
 
6. At least one non-transitory machine-readable medium comprising instructions that, when executed, cause at least one processor to:
load compressed machine learning parameter data into a cache accessible by a processor engine, the loaded data including first operable data;
identify that second operable data is present in the loaded data in the cache;
decompress, with the processor engine, the first operable data into first decompressed data;
execute, with the processor engine, a first machine learning operation using the first decompressed data;
decompress, with the processor engine, the second operable data into second decompressed data, the identification of the second operable data based on a comparison of bits in a sparsity bitmap to bits in the first decompressed data and the second decompressed data; and
execute, with the processor engine, a second machine learning operation using the second decompressed data.