Skip to content

Problem with checks on R/4.0.0

3 messages · Simon Michnowicz, Jeff Newmiller, Martin Maechler

Simon Michnowicz
# 
Dear R Group,
I can build a simple R/4.0.0 OK using   gcc/8.1.0, but when I tried to link
it with  the Intel MKL,  'make check' produced this error

tail tests/reg-tests-1d.Rout.fail
[,1] [,2] [,3]
[1,]    0   NA    0
[2,]    0    0   -1
show NA *or* NaN
 O  I  F  M  2
NA NA  1 NA NA
Error: is.na(print(vapply(nTypes, norm, 0, x = m))) are not all TRUE
Execution halted

Is this a significant error?
 There may be differences in how NaN are treated between GNU and MKL that
caused this.

regards


*---Simon Michnowicz *
Senior Application Specialist,  High-Performance Computing

*Research Support Services - eSolutions*
*Monash eResearch Centre*
Monash University
15 Innovation Walk, Building 75, Clayton Campus
Wellington Road, VIC 3800
Australia

T:  +61 3 9902 0794
M: +61 3 0418 302 046
E: simon.michnowicz at monash.edu
monash.edu
Jeff Newmiller
# 
Do read the Posting Guide... you are on the wrong mailing list for this question.
On May 20, 2020 6:46:01 PM PDT, Simon Michnowicz via R-help <r-help at r-project.org> wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Martin Maechler
    

    

      
      #
    

  


  

          

      
      
      
    

        
> Do read the Posting Guide... you are on the wrong mailing list for this question.

        
> On May 20, 2020 6:46:01 PM PDT, Simon Michnowicz via R-help <r-help at r-project.org> wrote:

        
>> Dear R Group,
    >> I can build a simple R/4.0.0 OK using   gcc/8.1.0, but when I tried to
    >> link
    >> it with  the Intel MKL,  'make check' produced this error
    >> 
    >> tail tests/reg-tests-1d.Rout.fail
    >> 
    >>> (m <- cbind(0, c(NA, 0), 0:-1))
    >> [,1] [,2] [,3]
    >> [1,]    0   NA    0
    >> [2,]    0    0   -1
    >>> nTypes <- eval(formals(base::norm)$type) # "O" "I" "F" "M" "2"
    >>> stopifnot(is.na( print(vapply(nTypes, norm, 0., x = m)) )) # print():
    >> show NA *or* NaN
    >> O  I  F  M  2
    >> NA NA  1 NA NA
    >> Error: is.na(print(vapply(nTypes, norm, 0, x = m))) are not all TRUE
    >> Execution halted
    >> 
    >> Is this a significant error?
    >> There may be differences in how NaN are treated between GNU and MKL
    >> that
    >> caused this.

Yes, there are such differences.
This one is indeed new bug in that version of Lapack  which
in the mean time has been fixed AFAIK...
definitely has been fixed in R's builtin version of Lapack.

Martin Maechler
ETH Zurich  and   R Core Team


    >> regards
    >> 
    >> 
    >> *---Simon Michnowicz *
    >> Senior Application Specialist,  High-Performance Computing
    >> 
    >> *Research Support Services - eSolutions*
    >> *Monash eResearch Centre*
    >> Monash University
    >> 15 Innovation Walk, Building 75, Clayton Campus
    >> Wellington Road, VIC 3800
    >> Australia
    >> 
    >> T:  +61 3 9902 0794
    >> M: +61 3 0418 302 046
    >> E: simon.michnowicz at monash.edu
    >> monash.edu
    >> 
    >> [[alternative HTML version deleted]]
    >> 
    >> ______________________________________________
    >> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
    >> https://stat.ethz.ch/mailman/listinfo/r-help
    >> PLEASE do read the posting guide
    >> http://www.R-project.org/posting-guide.html
    >> and provide commented, minimal, self-contained, reproducible code.

    > -- 
    > Sent from my phone. Please excuse my brevity.

    > ______________________________________________
    > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
    > https://stat.ethz.ch/mailman/listinfo/r-help
    > PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
    > and provide commented, minimal, self-contained, reproducible code.