1 Hit in 2.2 sec

Efficiently and Transparently Maintaining High SIMD Occupancy in the Presence of Wavefront Irregularity

Stephen V Cole
OpenMP also allows more complex specifications of parallelism, such as differentiation between fine-grained data parallelism and coarse-grained task parallelism, that facilitate processing across SIMD  ...  In 2016, 64 of the supercomputers on the Top500 list used GPUs, with another 35 using other manycore accelerators such as the Xeon Phi [120] .  ...  , and output or discard them;  ... 
doi:10.7936/k7ft8kff fatcat:wo6h6rk6lfgfjnvwnuuvzjxjya