Loading...
Please wait, while we are loading the content...
High-performance scientific code for GPUs
| Content Provider | CiteSeerX |
|---|---|
| Abstract | •Thousands of threads executing the same instructions •Multi-level memory hierachy •context switch-free hardware instruction scheduling •But very simple cores: no branch prediction, no out-of-order scheduling •Limited on-chip resources to be shared by threads Figure: GPU architecture 2 Research objective Is it possible to use high-level representation to generate high-performance scientific code for GPUs? Our intention is use this work to generate high-performance code for performing ultrasound simulation on GPUs. 3 Auto-tuning •Auto-tuning refers to software that is able to modify itself |
| File Format | |
| Access Restriction | Open |
| Subject Keyword | High-performance Scientific Code Simple Core Limited On-chip Resource Branch Prediction Gpu Architecture High-performance Code Thread Figure Auto-tuning Auto-tuning Refers High-level Representation Research Objective Ultrasound Simulation |
| Content Type | Text |
