A unified IR plus ML-based scheduling for CNN inference on multi-vendor integrated GPUs matches or exceeds vendor libraries (up to 1.62x) on image models while supporting more models.
Halide: a language and compiler for optimizing parallelism, locality, and recomputation in image processing pipelines
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Unified Optimization Approach for CNN Model Inference on Integrated GPUs
A unified IR plus ML-based scheduling for CNN inference on multi-vendor integrated GPUs matches or exceeds vendor libraries (up to 1.62x) on image models while supporting more models.