Re: [R-pkg-devel] Package submission issue - OMP reduction (flang-new)
Thanks again for your help.
>Have you tried explaining this in the package submission comment? What
>was the response? Does the reviewer need more evidence? Linking to the
>LLVM issue on GitHub and/or the previous R-package-devel thread may
>help.
We indicated the possible bug in the comment section, with a link on the GitHub
issue.
They didn't consider this, and just refused the package, sending us the .log
file explaining the issue on .omp reduction.
>1. In your ./configure script, check for the Fortran compiler being
>flang-new version 17. If the check succeeds, disable OpenMP altogether.
How can I do that ? I can create a .configure file on the root folder, but how
do I check the compiler version ? As in .R file, with FC = ... ? And what about
OpenMP deactivation ?
>2. Rewrite your Fortran code so that a variable is never touched before
>it is used in an OpenMP reduction. You may have to declare a new
>variable for every OpenMP reduction loop inside a function.
I'll try this, thanks. I'm still unable to install flang-new on my device, as
it is too voluminate and my computer systematically crash. So i'll do the
changes and send it on CRAN, hopefully it'll be ok.
Many thanks for all your help !
Best regards,
Romain Pierlot
- Mail original -
De: "Ivan Krylov"
À: "Romain Pierlot"
Cc: "r-package-devel"
Envoyé: Lundi 13 Novembre 2023 19:18:32
Objet: Re: [R-pkg-devel] Package submission issue - OMP reduction (flang-new)
On Mon, 13 Nov 2023 16:39:45 +0100 (CET)
Romain Pierlot wrote:
> Here is the error message, and the adequate code part is joint in the
> mail :
>
> error:
> loc("/data/gannet/ripley/R/packages/incoming/frailtypack.Rcheck/00_pkg_src/frailtypack/src/Integrale_mult_scl.f90":1811:17):
> 'omp.reduction' op must be used within an operation supporting
> reduction clause interface error: verification of lowering to FIR
> failed
Have you tried explaining this in the package submission comment? What
was the response? Does the reviewer need more evidence? Linking to the
LLVM issue on GitHub and/or the previous R-package-devel thread may
help.
(LLVM giveth and LLVM taketh away. While I'm sure that their
diagnostics helped improve many packages, I have encountered a few
places in the R code where clang suggested to replace boolean
expressions && and || with bitwise operations & and |. The code still
works, but now relies on boolean constants having certain exact bits
set. Replace a 1 with a 2 and things will break.)
It might be the case that CRAN cannot afford to make an exception for
this compiler bug. I can imagine a number of workarounds:
1. In your ./configure script, check for the Fortran compiler being
flang-new version 17. If the check succeeds, disable OpenMP altogether.
2. Rewrite your Fortran code so that a variable is never touched before
it is used in an OpenMP reduction. You may have to declare a new
variable for every OpenMP reduction loop inside a function.
If you don't succeed in convincing CRAN regarding the compiler bug,
which one of the two workarounds would you prefer?
--
Best regards,
Ivan
--
[ https://www.u-bordeaux.fr/ ] [
http://www.aquitaine-poitou-charentes.inserm.fr/ ]
[ mailto: | Romain Pierlot ]
Ingénieur de recherches Bio-Informatique
Équipe BIOSTAT
[ https://www.u-bordeaux.fr/ | https://www.u-bordeaux.fr ]
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
Re: [R-pkg-devel] Package submission issue - OMP reduction (flang-new)
On Tue, 14 Nov 2023 10:25:03 +0100 (CET)
Romain Pierlot wrote:
> We indicated the possible bug in the comment section, with a link on
> the GitHub issue. They didn't consider this, and just refused the
> package, sending us the .log file explaining the issue on .omp
> reduction.
Sorry about that. Was there an option to reply and explain? I once
argued my way out of a rejection after I explained a mistake on the
reviewer's part, but one has to be careful not to take too much time
from one of the seven people keeping ~2 packages working.
Let's do the workaround then.
> >1. In your ./configure script, check for the Fortran compiler being
> >flang-new version 17. If the check succeeds, disable OpenMP
> >altogether.
> How can I do that ? I can create a .configure file on the root
> folder, but how do I check the compiler version ?
#!/bin/sh
# taken from Writing R Extensions, 1.2. Configure and cleanup
: ${R_HOME=`R RHOME`}
if test -z "${R_HOME}"; then
echo "could not determine R_HOME"
exit 1
fi
# determine the Fortran compiler
FC="`"${R_HOME}/bin/R" CMD config FC`"
# Use --version output to determine the compiler
# A different compiler will either accept --version and print something
# else or fail due to "unknown argument". In both cases the branch will
# not be taken
if "$FC" --version 2>/dev/null | grep -q 'flang-new version 17'; then
# deactivate OpenMP, see below
fi
> And what about OpenMP deactivation ?
In your src/Makevars, you use Make variables SHLIB_OPENMP_FFLAGS and
SHLIB_OPENMP_CFLAGS to enable OpenMP support. The configure script can
edit this file in various ways, for example, setting the variables to
an empty string:
echo 'SHLIB_OPENMP_FFLAGS =' >>src/Makevars
echo 'SHLIB_OPENMP_CFLAGS =' >>src/Makevars
This is untested, but should work.
I think that workaround (1) should be easier to test (you can try
disabling OpenMP when the Fortran compiler is gfortran, verify that it
works, then switch the test to detect flang-new which you don't have
installed), but if you're more comfortable editing Fortran code, (2)
should also be viable.
--
Best regards,
Ivan
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
Re: [R-pkg-devel] Package submission issue - OMP reduction (flang-new)
On 11/13/23 19:18, Ivan Krylov wrote:
On Mon, 13 Nov 2023 16:39:45 +0100 (CET)
Romain Pierlot wrote:
Here is the error message, and the adequate code part is joint in the
mail :
error:
loc("/data/gannet/ripley/R/packages/incoming/frailtypack.Rcheck/00_pkg_src/frailtypack/src/Integrale_mult_scl.f90":1811:17):
'omp.reduction' op must be used within an operation supporting
reduction clause interface error: verification of lowering to FIR
failed
Have you tried explaining this in the package submission comment? What
was the response? Does the reviewer need more evidence? Linking to the
LLVM issue on GitHub and/or the previous R-package-devel thread may
help.
(LLVM giveth and LLVM taketh away. While I'm sure that their
diagnostics helped improve many packages, I have encountered a few
places in the R code where clang suggested to replace boolean
expressions && and || with bitwise operations & and |. The code still
works, but now relies on boolean constants having certain exact bits
set. Replace a 1 with a 2 and things will break.)
It might be the case that CRAN cannot afford to make an exception for
this compiler bug. I can imagine a number of workarounds:
1. In your ./configure script, check for the Fortran compiler being
flang-new version 17. If the check succeeds, disable OpenMP altogether.
2. Rewrite your Fortran code so that a variable is never touched before
it is used in an OpenMP reduction. You may have to declare a new
variable for every OpenMP reduction loop inside a function.
If you don't succeed in convincing CRAN regarding the compiler bug,
which one of the two workarounds would you prefer?
I would definitely look for a work-around. In the end we want to give
users software that works, rather than something that doesn't, but
telling them, possibly with good evidence, than it is not our fault.
Tomas
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
[R-pkg-devel] Trouble with greek letter in figure title in R-devel
Dear package developers,
I am struggling with an error on R devel (all flavors) that I cannot
reproduce with a freshly installed R-devel on my machine: Function
halfnormal in package DoE.base places the greek letter alpha in the
title of a figure. I changed the code to using plotmath about a month
ago (on CRAN request), and everything seemed fine then. Now, however,
R-devel produces an error message
(https://www.r-project.org/nosvn/R.check/r-devel-windows-x86_64/DoE.base-00check.html):
Error in title(...) :
conversion failure on 'Plot for rnorm.12., method = Lenth, α = 0.05'
in 'mbcsToSbcs': for α (U+03B1)
Calls: halfnormal ... -> plot -> plot.default -> localTitle
-> title
Execution halted
CRAN threaten to archive the package because of that error - so I would
really appreciate help on finding out what I should do (apart from nasty
things like writing alpha instead of using the greek letter).
The code in function halfnormal that creates the title in question looks
as follows:
## added August 1 5 2014, modified Oct 17 2023
if (is.null(titel)){
titel <- as.expression(bquote("Plot for "*.(xnam)*", "*alpha ==
.(alpha)))
dots$main <- titel
}
The list dots is later used in "do.call(plotfun, dots)".
Any recommendations?
Thanks and regards,
Ulrike
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
Re: [R-pkg-devel] Trouble with greek letter in figure title in R-devel
On 14.11.2023 15:45, Ulrike Groemping wrote:
Dear package developers,
I am struggling with an error on R devel (all flavors) that I cannot
reproduce with a freshly installed R-devel on my machine: Function
halfnormal in package DoE.base places the greek letter alpha in the
title of a figure. I changed the code to using plotmath about a month
ago (on CRAN request), and everything seemed fine then. Now, however,
R-devel produces an error message
(https://www.r-project.org/nosvn/R.check/r-devel-windows-x86_64/DoE.base-00check.html):
Error in title(...) :
conversion failure on 'Plot for rnorm.12., method = Lenth, α = 0.05'
in 'mbcsToSbcs': for α (U+03B1)
Yes, and your code contains
titel <- bquote(paste("Plot for ", paste(xnam), ", method = ",
.(method),
", \U03B1 = ", .(alpha), sep=""))
So please change this to plotmath as in the instance further above in
your code where you already wrote
titel <- as.expression(bquote("Plot for "*.(xnam)*", "*alpha ==
.(alpha)))
Best,
Uwe Ligges
Calls: halfnormal ... -> plot -> plot.default -> localTitle
-> title
Execution halted
CRAN threaten to archive the package because of that error - so I would
really appreciate help on finding out what I should do (apart from nasty
things like writing alpha instead of using the greek letter).
The code in function halfnormal that creates the title in question looks
as follows:
## added August 1 5 2014, modified Oct 17 2023
if (is.null(titel)){
titel <- as.expression(bquote("Plot for "*.(xnam)*", "*alpha ==
.(alpha)))
dots$main <- titel
}
The list dots is later used in "do.call(plotfun, dots)".
Any recommendations?
Thanks and regards,
Ulrike
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
Re: [R-pkg-devel] Trouble with greek letter in figure title in R-devel
Thank you for spotting this, I should have found it myself :-(
I suppose I don't get this error even from latest R-devel because of the
settings of my Windows system?
Best, Ulrike
Am 14.11.2023 um 17:16 schrieb Uwe Ligges:
On 14.11.2023 15:45, Ulrike Groemping wrote:
Dear package developers,
I am struggling with an error on R devel (all flavors) that I cannot
reproduce with a freshly installed R-devel on my machine: Function
halfnormal in package DoE.base places the greek letter alpha in the
title of a figure. I changed the code to using plotmath about a month
ago (on CRAN request), and everything seemed fine then. Now, however,
R-devel produces an error message
(https://www.r-project.org/nosvn/R.check/r-devel-windows-x86_64/DoE.base-00check.html):
Error in title(...) :
conversion failure on 'Plot for rnorm.12., method = Lenth, α =
0.05' in 'mbcsToSbcs': for α (U+03B1)
Yes, and your code contains
titel <- bquote(paste("Plot for ", paste(xnam), ", method = ",
.(method),
", \U03B1 = ", .(alpha), sep=""))
So please change this to plotmath as in the instance further above in
your code where you already wrote
titel <- as.expression(bquote("Plot for "*.(xnam)*", "*alpha ==
.(alpha)))
Best,
Uwe Ligges
Calls: halfnormal ... -> plot -> plot.default ->
localTitle -> title
Execution halted
CRAN threaten to archive the package because of that error - so I
would really appreciate help on finding out what I should do (apart
from nasty things like writing alpha instead of using the greek letter).
The code in function halfnormal that creates the title in question
looks as follows:
## added August 1 5 2014, modified Oct 17 2023
if (is.null(titel)){
titel <- as.expression(bquote("Plot for "*.(xnam)*", "*alpha ==
.(alpha)))
dots$main <- titel
}
The list dots is later used in "do.call(plotfun, dots)".
Any recommendations?
Thanks and regards,
Ulrike
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
--
##
## Prof. Ulrike Groemping
## FB II
## BHT
##
## prof.bht-berlin.de/groemping
## Phone: +49(0)30 4504 5127
## Fax: +49(0)30 4504 66 5127
## Home office: +49(0)30 394 04 863
##
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
Re: [R-pkg-devel] Trouble with greek letter in figure title in R-devel
Not sure if it is yet part of the --as-cran checks.
But you can set the env var
_R_CHECK_MBCS_CONVERSION_FAILURE_=true
Best,
Uwe
On 14.11.2023 17:29, Ulrike Grömping wrote:
Thank you for spotting this, I should have found it myself :-(
I suppose I don't get this error even from latest R-devel because of the
settings of my Windows system?
Best, Ulrike
Am 14.11.2023 um 17:16 schrieb Uwe Ligges:
On 14.11.2023 15:45, Ulrike Groemping wrote:
Dear package developers,
I am struggling with an error on R devel (all flavors) that I cannot
reproduce with a freshly installed R-devel on my machine: Function
halfnormal in package DoE.base places the greek letter alpha in the
title of a figure. I changed the code to using plotmath about a month
ago (on CRAN request), and everything seemed fine then. Now, however,
R-devel produces an error message
(https://www.r-project.org/nosvn/R.check/r-devel-windows-x86_64/DoE.base-00check.html):
Error in title(...) :
conversion failure on 'Plot for rnorm.12., method = Lenth, α =
0.05' in 'mbcsToSbcs': for α (U+03B1)
Yes, and your code contains
titel <- bquote(paste("Plot for ", paste(xnam), ", method = ",
.(method),
", \U03B1 = ", .(alpha), sep=""))
So please change this to plotmath as in the instance further above in
your code where you already wrote
titel <- as.expression(bquote("Plot for "*.(xnam)*", "*alpha ==
.(alpha)))
Best,
Uwe Ligges
Calls: halfnormal ... -> plot -> plot.default ->
localTitle -> title
Execution halted
CRAN threaten to archive the package because of that error - so I
would really appreciate help on finding out what I should do (apart
from nasty things like writing alpha instead of using the greek letter).
The code in function halfnormal that creates the title in question
looks as follows:
## added August 1 5 2014, modified Oct 17 2023
if (is.null(titel)){
titel <- as.expression(bquote("Plot for "*.(xnam)*", "*alpha ==
.(alpha)))
dots$main <- titel
}
The list dots is later used in "do.call(plotfun, dots)".
Any recommendations?
Thanks and regards,
Ulrike
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
Re: [R-pkg-devel] Package submission issue - OMP reduction (flang-new)
On 14/11/2023 9:38 a.m., Tomas Kalibera wrote: I would definitely look for a work-around. In the end we want to give users software that works, rather than something that doesn't, but telling them, possibly with good evidence, that it is not our fault. Fortune nomination! __ R-package-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-package-devel
