| [9017] | 1 | This is the GPU based ANUGA |
|---|
| 2 | =========================== |
|---|
| 3 | |
|---|
| 4 | |
|---|
| 5 | |
|---|
| 6 | Documentation |
|---|
| 7 | ------------- |
|---|
| 8 | |
|---|
| 9 | Documentation is under doc directory, and the html version generated by **Sphinx** under the directory *doc/sphinx/build/html/* |
|---|
| 10 | |
|---|
| 11 | |
|---|
| 12 | |
|---|
| 13 | |
|---|
| 14 | |
|---|
| 15 | Install Guide |
|---|
| 16 | ------------- |
|---|
| 17 | |
|---|
| 18 | 1. Original [ANUGA](anuga.anu.edu.au) is required |
|---|
| 19 | - For CUDA version, [PyCUDA](documen.tician.de/pycuda) is required |
|---|
| 20 | - For OpenHMPP version, current implementation is based on the [CAPS OpenHMPP Compiler](www.caps-entreprise.com) |
|---|
| 21 | * When compiling the code with **Makefile**, the NVIDIA device architecture and compute capability need to be specified |
|---|
| 22 | * For example, GTX480 with 2.0 compute capability </br> |
|---|
| 23 | ```HMPP_FLAGS13 = -e --nvcc-options -Xptxas=-v,-arch=sm_20 -c --force``` |
|---|
| 24 | * For GTX680 with 3.0 compute capability </br> |
|---|
| 25 | ```HMPP_FLAGS13 = -e --nvcc-options -Xptxas=-v,-arch=sm_30 -c --force``` |
|---|
| 26 | * Also the path for **python**, **numpy**, and **ANUGA/utilities** packages need to be specified |
|---|
| 27 | * Defining the macro **USING_MIRROR_DATA** in **hmpp_fun.h** </br> |
|---|
| 28 | ```#define USING_MIRROR_DATA``` |
|---|
| 29 | </br>will enable the advanced version, which uses OpenHMPP Mirrored Data technology so that data transmission costs can be effectively cut down, otherwise basic version is enabled. |
|---|
| 30 | |
|---|
| 31 | |
|---|
| 32 | |
|---|
| 33 | |
|---|
| 34 | |
|---|
| 35 | Environment Vars |
|---|
| 36 | ---------------- |
|---|
| 37 | |
|---|
| 38 | Please add following vars to you **.bashrc** or **.bash_profile** file |
|---|
| 39 | |
|---|
| 40 | ``` |
|---|
| 41 | export ANUGA_CUDA=/where_the_anuga-cuda/src |
|---|
| 42 | |
|---|
| 43 | export $PYTHONPATH=$PYTHONPATH:$ANUGA_CUDA |
|---|
| 44 | ``` |
|---|
| 45 | |
|---|
| 46 | |
|---|
| 47 | |
|---|
| 48 | Basic Code Structure |
|---|
| 49 | -------------------- |
|---|
| 50 | |
|---|
| 51 |  |
|---|
| 52 | |
|---|
| 53 | * **README** (What you are reading) |
|---|
| 54 | * **docs/** Documentation directory |
|---|
| 55 | * **codeStructure.pdf** The diagram above |
|---|
| 56 | * **Evolve workflow.pdf** The overall workflow of the evolve procedure. This includes all the function dependency and function interfaces, which is helpful to understand the evolve procedure of ANUGA |
|---|
| 57 | * **device_spe/** Some device specifications of our working station |
|---|
| 58 | * **profiling/** Profiling results |
|---|
| 59 | * **CUDA/** All the profiling results on CUDA implementation |
|---|
| 60 | * **sphinx/** The [Sphinx](http://sphinx-doc.org/) generated documentation |
|---|
| 61 | * **source/** The source files for Sphinx based documents |
|---|
| 62 | * **build/** The generated documents |
|---|
| 63 | * **html/** HTML version documentation |
|---|
| 64 | * **src/** Source code directory |
|---|
| 65 | * **anuga_cuda/** The CUDA implementation |
|---|
| 66 | * **config.py** Detail configuration for the CUDA implementation, including the path for all the kernel functions, optimal CUDA thread block configuration, etc. |
|---|
| 67 | * **gpu_domain_advanced.py** Python Class for CUDA implementation in advanced version |
|---|
| 68 | * **gpu_domain_basic.py** Python Class for CUDA implementation in basic version |
|---|
| 69 | * **anuga_HMPP/** The OpenHMPP implementation |
|---|
| 70 | * **Makefile** The Makefile |
|---|
| 71 | * **hmpp_dimain.py** Python Class for OpenHMPP implementation |
|---|
| 72 | * **hmpp_python_glue.c** The Python/C API to set up communication between Python ANUGA and OpenHMPP. |
|---|
| 73 | * **sw_domain.h** The C Struct type **domain** used in C implementation to access mesh information generated in Python ANUGA |
|---|
| 74 | * **sw_domain_fun.h** Connect C Struct type **domain** to all mesh information |
|---|
| 75 | * **hmpp_fun.h** All function declarations. |
|---|
| 76 | * **evolve.c** The evolve procedure |
|---|
| 77 | * **scripts/** Some useful bash script |
|---|
| 78 | * **utilities/** Utilities for sorting mesh information, checking results, etc. |
|---|
| 79 | * **test/** Testing cases |
|---|
| 80 | * **CUDA/** Testing cases for CUDA implementation |
|---|
| 81 | * **merimbula/** Merimbula testing case directory |
|---|
| 82 | * **merimbula.py** Merimbula testing case |
|---|
| 83 | * **OpenHMPP/** Testing cases for OpenHMPP implementation |
|---|
| 84 | * **merimbula.py** Merimbula testing case |
|---|
| 85 | |
|---|
| 86 | |
|---|
| 87 | Examples |
|---|
| 88 | -------- |
|---|
| 89 | |
|---|
| 90 | Running Merimbula model with CUDA implementation. |
|---|
| 91 | |
|---|
| 92 | ```$ python merimbula.py -gpu``` |
|---|
| 93 | |
|---|
| 94 | With pair-testing. |
|---|
| 95 | |
|---|
| 96 | ```$ python merimbula.py -gpu -test``` |
|---|
| 97 | |
|---|
| 98 | With rearranged mesh information. |
|---|
| 99 | |
|---|
| 100 | ```$ python merimbula.py -gpu -rg``` |
|---|
| 101 | |
|---|
| 102 | |
|---|
| 103 | |
|---|
| 104 | Author |
|---|
| 105 | ------ |
|---|
| 106 | |
|---|
| 107 | Mail to [Zhe Weng (John)](wengcsyz@gmail.com) |
|---|
