文件列表:

.vscode/ (0, 2023-09-13)
.vscode/configurationCache.log (10426, 2023-09-13)
.vscode/dryrun.log (1966, 2023-09-13)
.vscode/launch.json (1014, 2023-09-13)
.vscode/settings.json (1795, 2023-09-13)
.vscode/targets.log (11392, 2023-09-13)
.vscode/tasks.json (667, 2023-09-13)
FWD_PLOT_OUT/ (0, 2023-09-13)
FWD_PLOT_OUT/seis_fwi (140768, 2023-09-13)
LICENSE (35149, 2023-09-13)
Makefile (1478, 2023-09-13)
Output.txt (32357, 2023-09-13)
Output2.txt (82370, 2023-09-13)
bin_fwi_005_250hz_301_601/ (0, 2023-09-13)
bin_fwi_005_250hz_301_601/seis_fwi (132912, 2023-09-13)
dyke_simulation_rtf_-1.50m.csv (306393, 2023-09-13)
dyke_simulation_rtf_1.50m.csv (306113, 2023-09-13)
dyke_simulation_rtf_10.50m.csv (306345, 2023-09-13)
dyke_simulation_rtf_13.50m.csv (305355, 2023-09-13)
dyke_simulation_rtf_4.50m.csv (305875, 2023-09-13)
dyke_simulation_rtf_7.50m.csv (305930, 2023-09-13)
dyke_simulation_rtf_loc (0, 2023-09-13)
forward_dam_vx.pdf (64524, 2023-09-13)
forward_dam_vz.pdf (52670, 2023-09-13)
fwi_pre_proc.png (23021, 2023-09-13)
include/ (0, 2023-09-13)
include/cpu/ (0, 2023-09-13)
include/cpu/n_PCG_PSV.hpp (295, 2023-09-13)
include/cpu/n_alloc_PSV.hpp (2744, 2023-09-13)
include/cpu/n_contiguous_arrays.hpp (2180, 2023-09-13)
include/cpu/n_globvar.hpp (270, 2023-09-13)
include/cpu/n_kernel_PSV.hpp (2428, 2023-09-13)
include/cpu/n_kernel_lib_PSV.hpp (6459, 2023-09-13)
include/cpu/n_simulate_PSV.hpp (2060, 2023-09-13)
include/cpu/n_solvelin.hpp (967, 2023-09-13)
include/cpu/n_step_length_PSV.hpp (2978, 2023-09-13)
include/cuda/ (0, 2023-09-13)
include/cuda/d_PCG_PSV.cuh (298, 2023-09-13)
... ...

# Seismic Full Waveform Inversion with GPU Acceleration ## Overview This is Seismic Full Waveform Inversion (FWI) is a powerful technique used to estimate subsurface parameters by analyzing seismic measurements obtained at the surface. However, due to the large volume of data, complex model sizes, and non-linear iterative procedures involved, numerical computations for FWI are often computationally intensive and time-consuming. This project addresses these challenges by implementing parallel computation techniques with Graphical Processing Units (GPUs) via CUDA to significantly accelerate the FWI process. **Note:** This project is an implementation of the research paper described in [this paper](https://www.mdpi.com/2076-3417/12/17/8844). ## Implementation - Host code is written in C++ to manage the overall project structure and coordinate computations. - Parallel computation codes are written in CUDA C, a language optimized for GPU processing. ## Performance Comparison The project includes a comprehensive performance evaluation: - **Comparing CUDA C and OpenMP:** The computational time and performance achieved through CUDA C and OpenMP parallel computation are compared to a serial code implementation. - **Scaling with Model Dimensions:** The project demonstrates that as model dimensions increase, the performance improvement is enhanced. It remains nearly constant after reaching a certain threshold. - **Impressive GPU Performance Gain:** In our experiments, we achieved a GPU performance boost of up to 90 times compared to the serial code, underscoring the substantial benefits of GPU acceleration. ## Prerequisites To run this project, you will need the following: - NVIDIA GPU with CUDA support. - CUDA Toolkit installed ([Download CUDA Toolkit](https://developer.nvidia.com/cuda-downloads)). - C/C++ Compiler (e.g., `nvcc` for CUDA code and `g++` for CPU code). - Git (optional, for cloning the repository).

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：