(19)
(11)EP 3 779 681 A3

(12)EUROPEAN PATENT APPLICATION

(88)Date of publication A3:
24.02.2021 Bulletin 2021/08

(43)Date of publication A2:
17.02.2021 Bulletin 2021/07

(21)Application number: 20199012.4

(22)Date of filing:  13.02.2019
(51)International Patent Classification (IPC): 
G06F 9/30(2018.01)
(84)Designated Contracting States:
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

(30)Priority: 28.03.2018 US 201815938924

(62)Application number of the earlier application in accordance with Art. 76 EPC:
19157044.9 / 3547114

(71)Applicant: INTEL Corporation
Santa Clara, CA 95054 (US)

(72)Inventors:
  • Narayanamoorthy, Srinivasan
    Hillsboro, OR Oregon 97124 (US)
  • Satish, Nadathur Rajagopalan
    Santa Clara, CA California 95054 (US)
  • Suprun, Alexey
    Beaverton, OR Oregon 97007 (US)
  • Janik, Kenneth J.
    Hillsboro, OR Oregon 97124 (US)

(74)Representative: Samson & Partner Patentanwälte mbB 
Widenmayerstraße 6
80538 München
80538 München (DE)

  


(54)ACCELERATOR FOR SPARSE-DENSE MATRIX MULTIPLICATION


(57) Disclosed embodiments relate to multiply-accumulate operations. In one example, a processor comprises: a cache to store data; at least one core coupled to the cache. The at least one core comprises: execution circuitry to perform multiply-accumulate operations with a first source matrix and a second source matrix to generate a result matrix responsive to an instruction, wherein the first source matrix is a sparse matrix having non-zero data elements located at certain positions, wherein the first source matrix is stored in a compressed format that identifies the positions of the non-zero data elements in the first source matrix. The execution circuitry further comprises: a plurality of multiply-accumulate circuits to perform a plurality of multiply-add operations to multiply the non-zero data elements of the first source matrix by corresponding data elements of the second source matrix identified based on the positions in the compressed format to generate a plurality of products, and to add the plurality of products to accumulated values to generate data elements of the result matrix.







Search report












Search report