Creat membership Creat membership
Sign in

Forgot password?

Confirm
  • Forgot password?
    Sign Up
  • Confirm
    Sign In
Creat membership Creat membership
Sign in

Forgot password?

Confirm
  • Forgot password?
    Sign Up
  • Confirm
    Sign In
Collection
For ¥0.57 per day, unlimited downloads CREATE MEMBERSHIP Download

toTop

If you have any feedback, Please follow the official account to submit feedback.

Turn on your phone and scan

home > search >

Fast computation of bare soil surface roughness on a Fermi GPU

Author:
Li, Xiaojie   Song, Changhe   López, Sebastian   Li, Yunsong   López, José F.  


Journal:
Computers & Geosciences


Issue Date:
2015


Abstract(summary):

Highlights

A parallel computing method for 2D surface roughness was implemented by CUDA GPU API.

The experimental input data is collected by our in-house surface roughness tester.

A significant 115x speedup was achieved for the GPU implementation compare with CPU.

Abstract

Surface roughness is an important factor in bare soil microwave radiation for the observation of the Earth. Correlation length and standard deviation of surface height are the two statistical parameters that describe surface roughness. However, when the number of data points is large, the calculation of surface roughness parameters becomes time-consuming. Therefore, it is desired to have a high-performance computing facility to execute this task. A Graphics Processing Unit (GPU) provides hundreds of computing cores along with a high memory bandwidth. To carry out a parallel implementation of the algorithms, Compute Unified Device Architecture (CUDA) provides researchers with an easy way to execute multiple threads in parallel on GPUs. In this paper, we propose a GPU-based parallel computing method for 2D surface roughness estimation. We use an NVIDIA GeForce GTX 590 graphics card to run the CUDA implementation. The experimental input data is collected by our in-house surface roughness tester which is designed based on the laser triangulation principle, giving sample data points of up to 52,040. According to the experimental results, the serial CPU version of the implementation takes 5422 s whereas our GPU implementation takes only 47 s, resulting a significant 115× speedup.



Page:
38-44


VIEW PDF

The preview is over

If you wish to continue, please create your membership or download this.

Create Membership

Similar Literature

Submit Feedback

This function is a member function, members do not limit the number of downloads