It is a branch of PRESTO. We are working to accelerate the program with GPU. The work is based on Scott Ransom's C version,https://github.com/scottransom/presto About presto, please refer to http://www.cv.nrao.edu/~sransom/presto/
#2013-12-02 ####what's new####
- Texture memory is used to reduce the time required for memory read on GPU.
- cuFFT is used in a smarter way. cuFFT plans are stored independently outside functions that actually use them.
####test results####
- CPU platform: Intel Xeon E5-1650, 3.20GHz, 12 Cores, 62GByte memory
- GPU platform: Nvidia GeForce GTX780, with CUDA 5.5
#####data point = 83886080, dt per bin(s) = 0.00064#####
-numharm | -zmax | CPU runtime(sec) | GPU runtime(sec) | acceleration |
---|---|---|---|---|
16 | 256 | 2075.00 | 93.3 | 22.24 |
8 | 256 | 1089.82 | 53.96 | 20.19 |
1 | 256 | 247.44 | 9.79 | 25.27 |
#2013-10-23#
####The 1st release####
- You need a CUDA library to compile the code.
- Only accelsearch is GPUed. To run accelsearch on GPU, use the -cuda option. Or the program will run on CPU. For example: accelsearch -numharm 16 -zmax 256 ur_data.dat -cuda 0, 0 means the 1st GPU in your machine.