-
Notifications
You must be signed in to change notification settings - Fork 3
Open
Description
Sketches can be represented as binary presence/absence vectors. The overlap between two sketches can then be calculated by the dot product of the two vectors representing the sketches. If we consider many sketches, then the problem resolves into a matrix multiplication, which can be accelerated using GPU computation.
For this task, I plan to do the following:
- After reading sketches, represent them using vectors
- Use CUDA programming to implement a matrix multiplication
- Extract overlaps among sketches from the matmult result
- Benchmark the implementation
Metadata
Metadata
Assignees
Labels
No labels