# WRF作業
> [name=紀冠宇]

[toc]

# Compile WRF
## module load GCC ==(5%)==
- 於Taiwania 1的clogin1節點上執行
- Check `module list`

[rainchi15@clogin1 ~]$ module list
Currently Loaded Modulefiles:
  1) gcc/7.5.0                2) mpich-3.1.4-t            3) hdf5-1.8.21-t            4) pnetcdf-1.12.0-t         5) netcdf-c-4.7.3-t         6) netcdf-fortran-4.4.5-t

## Install HDF5 + NetCDF ==(15%)==
- Check `which ncdump`

[rainchi15@clogin1 ~]$ which ncdump
~/wrf-hw/opt/netcdf/bin/ncdump

- Report (Why use this flag?)

## Get WRF.exe ==(20%)==
- Check `tail -15 build_wrf.log`

 
==========================================================================
build started:   Tue Feb 20 22:53:56 CST 2024
build completed: Tue Feb 20 22:54:17 CST 2024
 
--->                  Executables successfully built                  <---
 
-rwxr-xr-x 1 rainchi15 TRI113199 52620696 Feb 20 22:54 main/ndown.exe
-rwxr-xr-x 1 rainchi15 TRI113199 52526192 Feb 20 22:54 main/real.exe
-rwxr-xr-x 1 rainchi15 TRI113199 51568200 Feb 20 22:54 main/tc.exe
-rwxr-xr-x 1 rainchi15 TRI113199 58313544 Feb 20 22:54 main/wrf.exe
 
==========================================================================

# Run WRF:
## Just run on headnode ==(5%)==
- Check Screenshot


## Single Node with Slurm ==(15%)==
- Check `./validate.csh ./VALIDATION ./VALIDATE`




## Optimization(Multiple Node, AVX512...) ==(20% ⬆️)==
- Check `./timing.csh TIMING/rsl.out.0000`

- Report (What do you do?)
- AVX512
- Parallel scalability test for
- 1 node
- 2 nodes
- 4 nodes

---

# Others
## Profile WRF ==(15%)==:
- Check Vtune Screenshot


- Report (Based on the results, What are your assumptions?)
猜測最前面20秒左右應該是MPI在複製Memory的內容，所以沒有CPU Usage
VTUNE的分析結果確實也是在MPI_memcpy花非常大量的時間處理
而後面皆為2秒多一個起伏循環
這些時間剛好與timing.csh得出的結果相似(radiation/non-raidation step)
## Display output ==(5%)==:
- Check Screenshot


## 其他你想告訴我們的

- 繳交前:
- [x] HackMD 開頭寫名字
- [x] 自行填寫進度追蹤表
- [x] 刪除沒完成項目的分數
- [x] 可閱讀權限設定為所有人