You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
19 lines
1.1 KiB
19 lines
1.1 KiB
The speedup of Part 1's GPU implementation is less impressive than in the previous MP. |
|
|
|
a. On the other hand, the GPU implementation of Part 1's solution solves a different problem than |
|
the CPU. Explain. |
|
|
|
b. Implement Thrust's GPU solution on the CPU (i.e., replace device_vectors with host_vectors) and |
|
compare the performance of the GPU solution to your new CPU solution. What is the speedup? |
|
|
|
Unlike our implementation from MP3, Part 2's black_scholes_functor takes its arguments as a struct |
|
rather than as three separate parameters. It also returns its two results wrapped in a single struct |
|
rather than by modifying two parameters passed as reference. Additionally, the memory storing this |
|
data is organized as an array of structures rather than MP3's strategy of a structure of arrays. |
|
|
|
a. Assume that Thrust implements the Black-Scholes algorithm by mapping the black_scholes_functor to a |
|
contiguous group of threads by passing them a contiguous group of option structures. In general, what |
|
effect, if any, does this memory access pattern have on performance? |
|
|
|
b. Does this effect matter for the Black-Scholes algorithm? Explain. |
|
|
|
|