L16 - Parallel Computing Flashcards
1
Q
How can parallel computing be applied to matrix multiplication?
A
Split into sub-matrices and split operations across processors.
2
Q
What is an issue with sequential training in Big Data?
A
Bottlenecking
3
Q
Why does Parallel Computing not make P == NP?
A
Because an arbitrary input size would require an arbitrary number of processors.
4
Q
What is the PRAM model?
A
An abstract computational model used for analysing efficiency of parallel algorthims.
5
Q
What does the PRAM model consist of?
A
- Unbounded number of processors
- Unbounded memory
- A finite program
6
Q
How is the running time of PRAM measured?
A
Time -> Running time of last processor
Hardware -> Max. number of active processors
7
Q
How are the individual processors measured?
A
Time - Number of instructions exectured
Space - Number of cells accessed
