MemoCode 2009 Team BarraCuda

Team Name

We selected BarraCuda because it is a very fearsome and voracious predator. You can find more at http://en.wikipedia.org/wiki/Barracuda

CUDA Documentation

http://www.nvidia.com/object/cuda_develop.html
http://www.nvidia.com/object/cuda_home.html#

Ported C Code to CPU

C Code (zip)

Equipment

Tesla C870

• http://www.nvidia.com/object/tesla_gpu_server.html
• Host Computer: argyle.ece.iastate.edu
  • Contact Mike Steffen (steffma@iastate.edu) for an account.

Minimum Memory Requirements

The green points ("critical values") in the graph below are the fewest points necessary to complete the polar to cartesian interpolation.

File:Critical Points1.png

Each graph below corresponds to a specific N value. For each graph, the x-axis is in terms of theta and ranges from pi/256 to pi/4. The y-axis shows the minimum number of KB's required to hold the critical values assuming each critical value requires four bytes of memory. There are ten lines on each graph corresponding to specific R values, R = 10, 20, ... , 90, 100 going from the top line to the bottom.

File:Graph N10.pngFile:Graph N50.png
File:Graph N100.pngFile:Graph N200.png
File:Graph N300.pngFile:Graph N400.png
File:Graph N500.pngFile:Graph N600.png
File:Graph N700.pngFile:Graph N800.png
File:Graph N900.pngFile:Graph N1000.png

Final Results

Benchmark Results (Runtime)

Test Cases	N	Theta	R	Ref Run Time (usec)	Cuda Run Time (usec)	Speedup
0	10	asin(0.1)	10	10958	6	1826.33
1	700	asin(0.1)	10	55619461	691	80491.26
2	1000	asin(0.1)	10	113455229	1202	94388.71
3	10	asin(0.0123)	100	11033	5	2206.6
4	700	asin(0.0123)	100	56893226	717	79348.99
5	1000	asin(0.0123)	100	116084611	1215	95542.89
6	10	asin(0.2)	100	11021	7	1574.43
7	700	asin(0.2)	100	55963647	665	84155.86
8	1000	asin(0.2)	100	112892176	1352	83500.13
					Geometric Mean	23924.63
					Average	58115.02