Name of the program: 2D prestack GPU-based RTM using effective boundary saving

Name of the manuscript: RTM using effective boundary saving: A staggered grid GPU implementation

Author(s) details:

Pengliang Yang(1), Jinghuai Gao(2), Baoli Wang(3)

Address: 	Institute of Wave and Information, 
		School of Electronic and Information Engineering, 
		Xi'an Jiaotong University, 
		Xi'an,  P.R. China
	and 	
		Bureau of Economic Geology, 
		Jackson School of Geosciences, 
		The University of Texas at Austin, TX, USA


Email(s):	(1) [email protected]
		(2) [email protected]
		(3) [email protected]

Notice: There is Matlab code to compute the staggered grid finite difference coefficients with varying even-order (NJ=2N). Assume you are finding the 10-th order finite difference coefficients. In principle, you should run the following commands in Matlab command window!

format long
NJ=10;
c=staggered_fdcoeff(NJ)

Instructions of the program: 1. To test the GPU-based RTM codes, you need CUDA well-installed; 2. You also need development version of Madagascar installed; 3. To reproduce the Marmousi example, rename SConstruct1.txt as SConstruct and run 'scons view' in that directory; 4. To reproduce the Sigsbee example, rename SConstruct2.txt as SConstruct and run 'scons view' in that directory;

The codes has already installed in the Madagascar development version, the cuda codes lie in $RSFSRC/user/pyang and the example scripts lie in $RSFSRC/book/pyang/gpurtm. Make sure the Internet is connected when run SConstructs because the marmousi and sigsbee models need to be downloaded.

Caution: Due to limited GPU resource on my laptop (NVS5400M GDDR3 2048M), I have to store a large amount of boundaries using pinned memory, which may degrade the speed performance a lot! It means that I have to run hours to obtain the images. If you have enough GPU resource, you may not need to do so! The speedup will improve a lot!

Enjoy this code!