AIX7863 PROGRAMMING GPU KERNELS FOR DEEP LEARNING
| Course Code: |
8877863 |
| METU Credit (Theoretical-Laboratory hours/week): |
3(3-0) |
| ECTS Credit: |
8.0 |
| Department: |
Artificial Intelligence Engineering |
| Language of Instruction: |
English |
| Level of Study: |
Graduate |
| Course Coordinator: |
|
| Offered Semester: |
Fall or Spring Semesters. |
Course Content
GPU computing using C++ and industry standard compute frameworks. Differences between CPUs and GPUs, modern GPU architecture, compute frameworks and memory hierarchies, CPU?GPU interaction, data transfer, execution, synchronization, shared memory, profiling, and performance analysis using speed-of-light and roofline models. Parallel algorithms for GPUs, compute ecosystem, improving code performance using kernel-level optimization techniques and grid-level optimization techniques.
Production-grade compute frameworks and running compute kernels from Python code. Case studies of compute-based GPU algorithms.
