Researchers at North Carolina State University have developed a way for collaboration between the CPU and GPU during the execution of programs, allowing their computing performance to be increased by over 20%. This is possible due to the introduction of 'fused architecture' by chip manufacturers where the CPU and GPU co-exist on the same chip.
AMD was the first to introduce their Fusion APUs to the region in early 2011 while Intel hoped on the bandwagon with their Sandy Bridge offerings at roughly the same time. One of the university researchers, Dr. Huiyang Zhou said that one of the shortcomings of this fused architecture was the lack of collaboration between the CPU and GPU though they are located on the same chip.
His team's approach to this is to allow the GPU cores to execute computational functions, and have CPU cores pre-fetch the data the GPUs will need from off-chip main memory. If the CPU has determined what data a GPU will need in advance, and fetches it from off-chip main memory on behalf of the GPU. Such an action by the former will allow the GPU cores to focus on executing the functions themselves and the overall process is sped up.
In preliminary testing, the researcher discovered that its new approach improved fused processor performance by an average of 21.4%. Their findings will be presented in a paper titled "CPU-Assisted GPGPU on Fused CPU-GPU Architectures" that will be presented on February 27th at the 18th International Symposium on High Performance Computer Architecture in the US.
The research was funded by the National Science Foundation and AMD. Earlier this month, AMD shared its company strategy to focus more on APUs as well as to make inroads into the tablet market with their APUs.
(Source: North Carolina State University)