Uploaded image for project: 'SAFe Program'
  1. SAFe Program
  2. SP-3382

Investigate different implementations of the PSS Complex Fourier Transform (CXFT) component

Change Owns to Parent OfsSet start and due date...
    XporterXMLWordPrintable

Details

    • Spike
    • Could have
    • PI19
    • COM PSS SW
    • None
    • Data Processing
    • Hide

      Enables an informed decision for converging into an implementation for CXFT, which will dictate the direction of this module's development and inform PSS hardware procurement ahead of AA2-scale commissioning.

      Show
      Enables an informed decision for converging into an implementation for CXFT, which will dictate the direction of this module's development and inform PSS hardware procurement ahead of AA2-scale commissioning.
    • Hide

      Document pros and cons of the existing (GPU, FPGA) and potential (CPU) CXFT implementations

      Create Features to plan out the development of PoC code to test for the most appropriate CXFT implementation

      Show
      Document pros and cons of the existing (GPU, FPGA) and potential (CPU) CXFT implementations Create Features to plan out the development of PoC code to test for the most appropriate CXFT implementation
    • 2
    • 2
    • 0
    • Team_PSS
    • Sprint 5
    • Hide

      The current Cheetah implementations and availability of FFT algorithms were checked (e.g., version of used FFTW library). A decision was made that we do not proceed with FPGA algorithms for now, but only explore CPU and GPU-based codes. There is currently no CPU-based FFT in Cheetah, but such an algorithm seems worthwhile exploring.

      There was the idea to look for a streaming FFT algorithm for the new CPU implementation. This and continued development of FFT algorithms required an evaluation of recent  literature to identify a potentially more efficient algorithm.

       

      General bechmarking results documenting pro and cons of various CPU and GPU algorithms (including new versions of the Cheetah GPU implemention and considered CPU code) were found and documented (see attachment). A promising new code was identified that may help with a performance improvement of ~factor 2.

       

      Before we can create the features for further development of the CXFT, we have to benchmark the chosen codes in the framework of Cheetah with PSS-tailored parameters. This will be the next step.

       

       

       

       

       

       

       

      Show
      The current Cheetah implementations and availability of FFT algorithms were checked (e.g., version of used FFTW library). A decision was made that we do not proceed with FPGA algorithms for now, but only explore CPU and GPU-based codes. There is currently no CPU-based FFT in Cheetah, but such an algorithm seems worthwhile exploring. There was the idea to look for a streaming FFT algorithm for the new CPU implementation. This and continued development of FFT algorithms required an evaluation of recent  literature to identify a potentially more efficient algorithm.   General bechmarking results documenting pro and cons of various CPU and GPU algorithms (including new versions of the Cheetah GPU implemention and considered CPU code) were found and documented (see attachment). A promising new code was identified that may help with a performance improvement of ~factor 2.   Before we can create the features for further development of the CXFT, we have to benchmark the chosen codes in the framework of Cheetah with PSS-tailored parameters. This will be the next step.              
    • 19.6
    • Stories Completed, Outcomes Reviewed, Satisfies Acceptance Criteria, Accepted by FO

    Description

      An investigation into a potential CPU implementation of the PSS Complex Fourier Transform (CXFT) component will be carried out and compared to the existing GPU, FPGA (OpenCL+Rabbit) implementations.

      Attachments

        1. FFTsum.pdf
          632 kB
          Posselt, Bettina

        Structure

          Activity

            People

              A.Noutsos Noutsos, Aristeidis
              L.Levin-Preston Levin-Preston, Lina
              Votes:
              0 Vote for this issue
              Watchers:
              0 Start watching this issue

              Feature Progress

                Story Point Burn-up: (100.00%)

                Feature Estimate: 2.0

                IssuesStory Points
                To Do00.0
                In Progress   00.0
                Complete19.0
                Total19.0

                Dates

                  Created:
                  Updated:
                  Resolved:

                  Structure Helper Panel