Problems Compiling and Running AlphaPhase

[david20011999]

Hello,

I am trying to use AlphaPhase, but I have encountered several issues both during compilation and execution.

1. Compilation Issues from the Repository

I cloned the repository and tried to compile it, but I ran into the following errors:

• When running bash Compile.sh, I got the following error:

  rm: cannot remove 'alphaphase': No such file or directory
  rm: cannot remove '*.mod': No such file or directory
  Compile.sh: line 3: ifort: command not found
  

  It seems that the script tries to remove files that do not exist and requires ifort, but not all systems have the Intel compiler installed.

• I tried compiling with gfortran instead of ifort, but there were issues with the compiler flags.

• I also ran make_zip_file.sh, but encountered this error:

  make: *** No rule to make target 'latexpdf'.  Stop.
  cp: cannot stat 'AlphaPhase-doc/build/latex/AlphaPhase.pdf': No such file or directory
  The manual needs to be built.
  

  It looks like some files or dependencies for generating the documentation are missing.

2. Issues with the Precompiled Version

To avoid compilation issues, I decided to use the executable included in AlphaPhase.zip. This contains version 1.3.8, which runs correctly. However, when checking the PhasingResults folder, I found that only the FinalPhaseX.txt files are generated, but not FinalPhase.txt, which should contain the full results.

Could you help me resolve these issues? Is there a way to obtain FinalPhase.txt with version 1.3.8, or is there an additional step I need to follow?

Thanks in advance!

[gregorgorjanc]

@david20011999 thanks for the report!

Can you modify the script so that it tries to remove the folder and files, but does not complain if they are not there?

ifort creates a much more performant binary than gfortran due to various optimisations. It's been a while since I have done any Fortran work, but there used to be free academic version for ifort from Intel. Can you check!?

You mention compile errors with gfortran. What were those?

Can you remind me what is the difference between files FinalPhaseX.txt and FinalPhase.txt (it's been a while since I used AlphaPhase!)?

[david20011999]

Hi @gregorgorjanc,

Thanks for your response!

1. Regarding the compilation errors:

   • I had an error because before make_zip_file.sh assumes the program and manual are already built.
   • I have tried to correct it and when I try to build it as instructed, I get the following error:

     [  1%] Building Fortran object lib/AH/CMakeFiles/alphahouse.dir/Utilities/baseSpecFileModule.f90.o
     gfortran: error: language SSSE3 not recognized
     gfortran: error: language SSSE3 not recognized
     make[2]: *** [lib/AH/CMakeFiles/alphahouse.dir/build.make:556: lib/AH/CMakeFiles/alphahouse.dir/Utilities/baseSpecFileModule.f90.o] Error 1
     make[1]: *** [CMakeFiles/Makefile2:143: lib/AH/CMakeFiles/alphahouse.dir/all] Error 2
     make: *** [Makefile:136: all] Error 2
     

   • It looks like gfortran does not recognize the SSSE3 instruction set. I am now switching to ifort to see if that resolves the issue and will keep you updated.

2. Difference between FinalPhaseX.txt and FinalPhase.txt:

   • FinalPhaseX.txt contains phased haplotypes in the X-th core. However, I am not sure if this is the final result or if an additional assembly step is done, combining all core results and the haplotype library.
   • FinalPhase.txt should contain the fully phased genotypes, but this file is missing in my output.

I will continue testing with ifort and let you know how it goes. Thanks for your help!

[gregorgorjanc]

• I had an error because before make_zip_file.sh assumes the program and manual are already built.

This is really meant to create a package, once the binary and manual are built (as you say), so we best sort out the compilation first.

[gregorgorjanc]

Can you modify the script so that it tries to remove the folder and files, but does not complain if they are not there?

@david20011999 which script is producing this error? This will be easy to fix.

[gregorgorjanc]

Can you modify the script so that it tries to remove the folder and files, but does not complain if they are not there?

@david20011999 which script is producing this error? This will be easy to fix.

@david20011999 This commit should sort this out - let us know e99fe5c

[gregorgorjanc]

gfortran: error: language SSSE3 not recognized

@david20011999 Do explore ifort, but just so you know, this comes from cmake compilation process. We geared everything towards ifort. Flags for Fortran programs are defined in this repo https://github.com/AlphaGenes/AlphaCMakeFiles and search for SSSE3 gives https://github.com/search?q=repo%3AAlphaGenes%2FAlphaCMakeFiles%20ssse3&type=code, showing that indeed this flag is for Intel compiler. ChatGPT suggests using -mssse3 for gfortran instead of -xSSSE3;)

[gregorgorjanc]

2. Difference between FinalPhaseX.txt and FinalPhase.txt:

@david20011999 can you please study the Fortran code (should be easy to search for FinalPhase) and let's figure this now out together.

[david20011999]

I've been researching this. On the one hand, I can't find any potential errors in the code that could be associated with the lack of printing of FinalPhase.txt. However, I'm also not sure if that code it is the used to the compilated software, as there are some discrepancies (for example, on line 162 of the InputOutput.f90 file, it writes a PhasingResults/information.txt file, but when executed, it doesn't generate it either). It's possible that it was a test version. Anyway, I'll keep checking in case I missed something.

Regarding the Intel Fortran compiler, I've seen that ifort is no longer maintained/ available by the developers and they've now moved on to ifx (https://www.intel.com/content/www/us/en/docs/fortran-compiler/get-started-guide/2025-1/overview.html). I've tried to get it to compile with a simple change, but it seems the transitioning isn't direct. I will work on it to try to compile it.

[gregorgorjanc]

Are there flags/options you need:want for the desired output?

[david20011999]

I finally managed to compile it!! 🎉 🎊

Sorry for some questions that I now understand are basic😅, but I've never done this type of compilation. Perhaps adding these instructions would be beneficial for new users.

About modifications, switching from ifort compiler to ifx is a problem, and some statements became not correct and need modification (mostly pure and, in some write functions, &).

However, the executable still doesn't return the FinalPhase.txt file, but now it's up to me to figure out why 💪💪.

Thanks for the help, I'll keep you informed.

[gregorgorjanc]

Sorry for some questions that I now understand are basic😅, but I've never done this type of compilation. Perhaps adding these instructions would be beneficial for new users.

Could you document what you have done to this effect?

About modifications, switching from ifort compiler to ifx is a problem, and some statements became not correct and need modification (mostly pure and, in some write functions, &).

Hmm, pure stuff in there for a reason. Is ifx expecting you to remove the pure specification? And what are changes for write operations? I suggest you fork this repository, create a branch and do changes in the branch, then open a pull-request so we see the changes and hopefully learn from them ...

Happy hunting for FinalPhase.txt;)

[david20011999]

Sure! I will fork the repository to add all changes I did and you will be able to evaluate if they were correct and should be added or not.

About FinalPhase.txt it seems easy... In the InpputOutput.f90 file params%outputCombined = FALSE and params%outputPerCore = TRUE in all cases and therefore results were printed by cores instead of together. I think this issue comes from these lines in the same file, but I will explore in depth:

  ! Should probably have an input option to always output per core even when a single run - hence two lines here - but not
  ! currently implemented
  params%outputParams%outputPerCore = .not. singleRun
  params%outputParams%outputCombined = singleRun

[gregorgorjanc]

Good job in tracking this one down @david20011999 !

[gregorgorjanc]

@david20011999 FYI https://github.com/klausler/fortran-wringer-tests A collection of non-portable Fortran usage, standard-conformant or otherwise