ForRF

Time iterative deconvolution in Fortran with CUDA acceleration.

Requirements

• Fortran Compiler (gfortran/ifort)
• CUDA Toolkit (for GPU support)
• CMake >= 3.10

Build

mkdir build && cd build
cmake .. && make

Usage

Run the test program with an optional argument to select the device (default is GPU).

cd src
# Run on GPU (default)
./forrf_test gpu

# Run on CPU
./forrf_test cpu

Comparison with Seispy

<img src=https://user-images.githubusercontent.com/7437523/147818243-8d5bbfd8-0ae4-4d8c-a3fc-a0e6677d6701.png width=60% />

Comparison between CPU and GPU

Performance comparison on a test dataset:

• CPU: ~3.07s
• GPU: ~0.23s (~13x speedup)

Defect

The amplitude at peaks may not be exactly consistent with Seispy.