Parallel Option Pricing in the GPU - Final Project CSC 213

Students:

Diogo Tandeta Tartarotti and Vinicius Ono Sant'anna

Source code

• main.cu: has the GPU kernel and main function.
• black.cuh: has the functions used to calculate D1 and D2
• file_handling.cuh: has the function to process input from .csv files
• time_util.cuh: has the function to time the kernel

Input Data Format

.csv file with the following comma-separated columns in this order: Stock Price, Strike Price, Maturity, Dividends, Volatility, Risk-free, Call Price. Also, the first line of the file must not contain more than 1024 characters.

Usage

This program uses the European options solution of the Black-Scholes model to find the call and put prices of provided options. The program then writes the prices to an output file and reports statistics about its execution to the terminal. There are input data files of varying sizes in the data folder.

To build the program, run make. To run the program, run ./main path/to/datafile To see the results of the program, open prices_output.csv in a text editor

Example:

onosanta@wang:parallel_black_scholes-main$ make
nvcc -g  -o main main.cu 
onosanta@wang:parallel_black_scholes-main$ ./main data/SNP.csv
Number of options: 57516
Total computation time: 688μs
Computation rate: 83,598,837.21 options per second
onosanta@wang:parallel_black_scholes-main$ ./main data/SNP-medium.csv
Number of options: 103250
Total computation time: 1244μs
Computation rate: 82,998,392.28 options per second
onosanta@wang:parallel_black_scholes-main$ ./main data/SNP-large.csv
Number of options: 516250
Total computation time: 5408μs
Computation rate: 95,460,428.99 options per second

Note that for each run the program should create a file in the current directory named prices_output.csv, which contains the output for put and call premiums prices for each of input options inputs as shown in the example above.