Skip to end of metadata
Go to start of metadata

Project Summary

Sector 34 uses a depth reconstruction algorithm for reconstructing data. This page is used for tracking SSG's progress at improving the performance of the code. Information about progress can be found on the Progress page. Documentation of meetings and discussions are kept on the Meeting Minutes page.

Source Code

The source code can be found in SVN here.

Development tasks can be found on SSG's Jira Depth Reconstruction page here.

Release

CPU

reconSingle1.2

  • Updated CPU version for single HDF5 input

reconMultiple1.2

  • Updated CPU version for multiple HDF5 input
GPU with multiple input

cudaReconMult1.0 

  • GPU code with multiple hdf5 input, can set the video card number for the code to run.(double precision)

cudaReconMult1.1  

  • New feature for GPU code with multiple hdf5 input image to read both txt and xml geometry files .(double precision)

cudaReconMult1.2

  • Replace atomicAdd to float version (single precision)
  • Change cuda part from double to float
  • Increase performance by at least 2 times compare to GPU code and 10 times to CPU code
GPU with single input

cudaReconSingle1.0 

  • GPU code with single hdf5 input, can set the video card number for the code to run. (double precision)

cudaReconSingle1.1  

  • New feature for GPU code with single hdf5 input image to read both txt and xml geometry files .(double precision)

cudaReconSingle1.2 

  • Replace atomicAdd to float version (single precision)
  • Change cuda part from double to float
  • Increase performance by at least 2 times compare to GPU code and 10 times to CPU code

 

 

 

 

 

 

 

 

  • No labels