Open Computing Language 101 (notes)

Day: 3
Session Time/Number:   2:30PM
Session Space:

Convener: Sushma
Notetaker: Smitha

TAGs for the Session:
Heterogeneous, computing, graphics,

——————————-
Specific resources to share from session:  (links, books, organizations, articles etc.)

——————————-
Discussion Notes:
1. OpenCL – heterogeneous parallel programming model
2. Used in many applications – compute vision, image processing, win zip, speech processing etc.
3. Meant to be cross platform, gives good performance speed up, portable
4. Concept of host & device code – could run on CPU and GPU respectively
5. Host is control logic and resource mgmt – written in C, C++, Java etc.
6. Device is the actual algorithm written in OpenCL – which is a subset of C language
7. Host calls the function that runs on the device
8. Scales well with many processors
9. Read h/w vendors tuning guide to get maximum performance
10. If you have a parallelizable problem and many resources like CPU, GPU, plus you want performance and portability – use OpenCL!