Welcome to LZN's Blog!

Wind extinguishes a candle but energizes fire.

Start to play with C++ in Linux

2020-05-29

LZN

technology

Now we spin-up for the next journey! We try to compile the first C++ program in the linux environment.

First we review the Big Three command line to invoke compilers:

	GNU	INTEL	PGI
FORTRAN	gfortran	ifort	pgfortran
C	gcc	icc	pgcc
C++	g++	icpc	pgc++

Now we try a very simple program, do a loooong loop and see the CPU lapsed time!

#include <iostream>
#include <time.h>
 
using namespace std;
 
int main()
{
    for (int i = 0; i < 1000000000; i++)
    {
        i++;
    }
    cout << "Totle Time : " << (double)clock() /CLOCKS_PER_SEC<< "s" << endl;
    return 0;
}

G++: 1.53s
ICPC: 0s

Interesting! Here you see the optimization of the Intel compilers!

Updated 2020-05-29

Read All

Install cfgrib for xarray

2020-05-19

LZN

technology

python cfgrib eccodes
Here we archive the process of building cfgrib engine for xarray in python3.7 (Anaconda). Although there still seems to be some problems, the eccodes and cfgrib have been installed sucessfully.

First, we found using conda install cannot install eccodes due to dependency problems (might comes from python version issue, we did not check then).

Then we install the eccodes from source. The process was smooth, just install cmake prior to the eccodes installation.

When install python interface, try both --install-option and export ECCODES_DIR=/path/to/where/you/install/eccodes.

After that:
```
pip install --force-reinstall cfgrib
```
Note conda does not work in my case.

Set all possible environmental varibales:
```
export ECCODES_DIR=/home/yhuangci/lee/soft/eccodes-2.17.0
export PATH=$ECCODES_DIR:$PATH
export LIBECCODES=$ECCODES_DIR/lib
export LD_LIBRARY_PATH=$ECCODES_DIR/lib:$LD_LIBRARY_PATH
```
xr.open_dataset('example.grib', engine='cfgrib') works with some data-caused errors.

Error things like:
```
cfgrib.dataset.DatasetBuildError: key present and new value is different
```
See this post. Add a backend_kwargs to overwrite the “strict” option will fix the problem:
```
ds_grib = xr.open_dataset('../../data/era5/sfc202005.grib', engine='cfgrib', backend_kwargs={'errors': 'ignore'})
```
Updated 2020-05-19
Read All
Install metpy in a new conda environment

2020-05-11

LZN

technology

python metpy anaconda
I used the default environment to install metpy with:
```
conda install metpy
```
got errors.
```
TypeError: find_intersections takes 5 parameters, but 3 units were passed
```
Following this post. The issue seems to be a specific version dependency issue. In conda, the metpy is version 0.11, while version 0.12 solves the problem.

Here we just create a new conda environment for metpy.
```
conda create -n metpy python=3.7
```
Use pip install to install the metpy. After that, we found the cartopy need to be installed. Then, everything goes fine.

Example figure:

CartoPy figure:

Updated 2020-05-11
Read All

11/115

Welcome to LZN's Blog!

Start to play with C++ in Linux

Install cfgrib for xarray

Install metpy in a new conda environment