JinwoongKim / Massively-Parallel-Query-Processing-on-Heterogeneous-Architecture

Homepage
http://jinwoongkim.github.io/Massively-Parallel-Query-Processing-on-Heterogeneous-Architecture/
1 stars 2 forks source link

Maximize CPU Threads with Shared Queue(TBU) #50

Open JinwoongKim opened 8 years ago

JinwoongKim commented 8 years ago

We often fail to maximize the utilization of the CPU because less CPU threads than the number of threads shows the peak performance.

For example, even if we have 16 threads, 4 or 8 threads show better performance than 16 threads due to the ... corresponding CUDA block threads...?

image

GPU kernel looks through the shared queue to process the query

JinwoongKim commented 8 years ago

Not only CPU thread, CPU-GPU pair ... should bring a query from queue....