Skip to content

A modified version of rayinvr that can work properly on modern Linux(e.g. Ubuntu 18.04) and Windows(e.g. Windows 10) platform. Rayinvr by Zelt:

License

Notifications You must be signed in to change notification settings

hzhu212/rayinvr

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

36 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

A modified version of rayinvr that can work properly on modern Linux(e.g. Ubuntu 18.04) and Windows(e.g. Windows 10) platform.

How to use

Prepare tool-chains

On Linux, you need to install x11 library and gfortran. For example, on Ubuntu 18.04:

sudo apt install gfortran libx11-dev

On Windows, you need to install GNU Make and mingw-w64 manually.

Build

Assuming we are in root directory of rayinvr source now.

Inside each subdirectory there is a Makefile to make two versions of each program. For example, you can build tramp as the following steps:

cd tramp
make xtramp FC=gfortran CC=gcc
make tramp FC=gfortran CC=gcc
# or just `make FC=gfortran CC=gcc`, will build both the two versions.

PS: the keyword arguments FC=gfortran CC=gcc means setting Fortran Compiler to gfortran and C Compiler to gcc. By default, the two variables are f77 and cc respectively.

The first version, prefixed by the letter x, is based on X11 graphics system. The second version is originally for users who have access to the commercial graphics package called Uniras. However, the second version would also be helpful if you just need to calculate without ploting. On Windows, there is no x11 library so that the second version is the only choice.

After this step, we will get rayinvr.exe, tramp.exe etc. under build directory.

Add to PATH

This step is optional.

Add build directory to PATH, then we will be able to run rayinvr and other commands at any path.

Run examples

In original rayinvr repository, Zelt provided 8 example datasets, which is helpful to test our build results.

But we can't run our commands on these datasets directly, because I have changed main.f to handle higher float precision when reading v.in file (see below). If you have tried running commands at examples directory, you will probably end up with an error msg like this:

At line 236 of file main.f (unit = 20, file = 'v.in')
Fortran runtime error: Bad value during floating point read

Error termination. Backtrace:

But don't worry, I have made a python script which can convert old style v.in file into high precision version, and I have put all converted example datasets in examples-new directory.

Now, let's try example3 (Why not example1? because example3 is neither too simple, nor too complicated):

# On Windows 10 WSL for my case

# go to dataset directory
cd examples-new/e3

# run rayinvr.exe
../../build/rayinvr.exe

The output will be as below, basically consistent with that of original rayinvr:

shot#   1:   ray code  2.2:    59 rays traced
shot#   1:   ray code  3.2:    60 rays traced
shot#  -2:   ray code  2.2:    17 rays traced
shot#  -2:   ray code  3.2:     7 rays traced
shot#   2:   ray code  2.2:    63 rays traced
shot#   2:   ray code  3.2:    73 rays traced
shot#  -3:   ray code  2.2:    34 rays traced
shot#  -3:   ray code  3.2:    26 rays traced
shot#   3:   ray code  2.2:    65 rays traced
shot#   3:   ray code  3.2:    73 rays traced

|----------------------------------------------------------------|
|                                                                |
| total of   2089 rays consisting of    21239 points were traced |
|                                                                |
|           model consists of  4 layers and  40 blocks           |
|                                                                |
|----------------------------------------------------------------|


Number of data points used:      477
RMS traveltime residual:       0.068
Normalized chi-squared:       46.331

Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_DIVIDE_BY_ZERO IEEE_OVERFLOW_FLAG IEEE_DENORMAL

Now, start your work with rayinvr!

Problems in original rayinvr and fix

This section lists what errors the original rayinvr throws and how I fix them.

Makefile: unilink not found

Just abandon unilink command and use $(FC) instead. Open the Makefile that throws this error, and change the line:

unilink ${TRAMP_OBJS}

to

$(FC) -o main ${TRAMP_OBJS}

PS: $(FC) will be replaced with your Fortran Compiler, for example gfortran.

Issue of fortran "Variable FORMAT expression"

Both f77 and gfortran do not support "Variable FORMAT expression", which is used in several rayinvr programs. For example, pltlib.f:198 reads:

5     format(i2/i10/4e15.5,<nchar>a1)

This line includes a "Variable FORMAT expression", in which nchar is an integer variable. The error throwed out is like the following:

gfortran Error: Unexpected element ‘<’ in format string at (1)

To fix this issue, there are 2 common ways:

  1. As GUN Fortran recommended, we can construct a format string dynamicly. See here

  2. Simply replace <nchar> with a number which is large enough, say 10000. Now, the format line will looks like this:

    5     format(i2/i10/4e15.5,10000a1)

Issue of character '$'

gfortran does not allow using $ in variable name by default. Or it will throw the following error:

# when building xrayinvr
# in rayinvr/trc.f:498:17:
Fatal Error: Invalid character ‘$’ at (1). Use ‘-fdollar-ok’ to allow it as an extension compilation terminated

To fix this error, you should add option -fdollar-ok to gfortran. To be more detail, add this option to FFLAGS at the beginning of each Makefile file:

FFLAGS = -O -fdollar-ok

Change default foreground and backgroud color of Xwindow

pltlib/xbuplot.c uses Xlib to plot graphics. Foreground and background settings locates at line 99 and 100:

  /*  create opaque window  */
  win = XCreateSimpleWindow (display, RootWindow (display, screen_num),
                x, y, win_width, win_height, border_width,
        WhitePixel (display, screen_num),    // foreground
        BlackPixel (display, screen_num));   // background

Read v.in with higher precision

The default precision of v.in is f7.2. If you want to read a higher precision version of v.in(say, f8.3), just open the main.f that reads the v.in, change the lines:

15       format(i2,1x,10f7.2)
25       format(3x,10i7)
25       format(3x,10(5x,i2))

to:

15       format(i2,1x,10f8.3)
25       format(3x,10i8)
25       format(3x,10(6x,i2))

Promoted X axis precision from 1m to 0.1m

Changed rayinvr/main.f:890 (and many others):

if(abs(xshotr-xshotf).lt..001.and.idr(is).eq.idf) then

to:

if(abs(xshotr-xshotf).lt..0001.and.idr(is).eq.idf) then

Some syntax errors

Changed pltsyn/pltsec.f:253:

108     if(namp.eq.0) write(*,*, fmt="(/

to:

108     if(namp.eq.0) write(*, fmt="(/

Adaption for large models

Enlarged some parameters in rayinvr/rayinvr.par and tramp/tramp.par for processing larger models.

Promoted prayi to 100000 to enable huge tx.in files (>100000 lines).

promote default real and integer precision from 4 bytes to 8 bytes

Add -fdefault-real-8 -fdefault-double-8 -fdefault-integer-8 options to FFLAGS in Makefile, and replace fixed declaritions (real*4) in the code with flexible ones (real).

About

A modified version of rayinvr that can work properly on modern Linux(e.g. Ubuntu 18.04) and Windows(e.g. Windows 10) platform. Rayinvr by Zelt:

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages