Vector Neural Network Instructions

URI: https://gptkb.org/entity/Vector_Neural_Network_Instructions
GPTKB entity
Statements (31)
Predicate Object
gptkbp:instanceOf instruction set architecture extension
gptkbp:abbreviation gptkb:VNNI
gptkbp:category SIMD instruction set
x86 instruction set extension
gptkbp:developedBy gptkb:Intel
gptkbp:documentedIn gptkb:Intel_Software_Developer’s_Manual
gptkbp:enables faster convolution operations
faster matrix multiplication
efficient quantized neural network inference
gptkbp:feature fused multiply add instructions
support for INT8 operations
vector dot product instructions
gptkbp:hasType BF16
FP16
INT8
https://www.w3.org/2000/01/rdf-schema#label Vector Neural Network Instructions
gptkbp:introducedIn gptkb:Intel_AVX-512
2018
gptkbp:purpose accelerate deep learning workloads
improve performance of neural network inference
gptkbp:relatedTo gptkb:AVX-512
gptkb:Deep_Learning_Boost
gptkbp:successor AVX2 FMA instructions
gptkbp:usedIn gptkb:Intel_Xeon_Scalable_processors
Intel Core processors (10th Gen and later)
gptkbp:bfsParent gptkb:Intel_Deep_Learning_Boost
gptkb:VNNI
gptkb:DL_Boost
gptkb:Intel_Deep_Learning_Boost_(VNNI)
gptkb:Deep_Learning_Boost_(VNNI)
gptkbp:bfsLayer 6