R help - Aug 2004 - precision problems in testing with Intel compilers

I compiled the 1.9.1 src.rpm with the standard gnu tools and it works.
I tried compiling the 1.9.1 src.rpm with the Intel 8 C and FORTRAN
compilers and it bombs out during the testing phase:

    comparing 'd-p-q-r-tests.Rout' to
'./d-p-q-r-tests.Rout.save' ...267c267
    < df = 0.5[1] "Mean relative  difference: 5.001647e-10"
    ---
    > df = 0.5[1] TRUE
    make[3]: *** [d-p-q-r-tests.Rout] Error 1
    make[3]: Leaving directory `/usr/src/redhat/BUILD/R-1.9.1/tests'
    make[2]: *** [test-Specific] Error 2
    make[2]: Leaving directory `/usr/src/redhat/BUILD/R-1.9.1/tests'
    make[1]: *** [test-all-basics] Error 1
    make[1]: Leaving directory `/usr/src/redhat/BUILD/R-1.9.1/tests'
    make: *** [check-all] Error 2
    error: Bad exit status from /var/tmp/rpm-tmp.63044 (%build)

looking at the differences between the failed file and the standard, I get:

    fws wolf tests] diff  d-p-q-r-tests.Rout.save d-p-q-r-tests.Rout.fail 
    3c3
    < Version 1.9.0 Patched (2004-04-19), ISBN 3-900051-00-3
    ---
    > Version 1.9.1  (2004-06-21), ISBN 3-900051-00-3
    281c281
    < df = 0.5[1] TRUE
    ---
    > df = 0.5[1] "Mean relative  difference: 5.001647e-10"
    935c935
    < Time elapsed:  7.83 0.04 16.1 0 0 
    ---
    > Time elapsed:  2.49 0.01 2.55 0 0 

Besides being 3 times faster, it's stopping on the following code:

   for(df in c(0.1, 0.5, 1.5, 4.7, 10, 20,50,100)) {
     cat("df =", formatC(df, wid=3))
     xx <- c(10^-(5:1), .9, 1.2, df + c(3,7,20,30,35,38))
     pp <- pchisq(xx, df=df, ncp = 1) #print(pp)
     dtol <- 1e-12 *(if(2 < df && df <= 50) 64 else if(df >
50) 20000 else
500)
     print(all.equal(xx, qchisq(pp, df=df, ncp=1), tol = dtol))# TRUE
     ##or print(mapply(rErr, xx, qchisq(pp, df=df,ncp=1)), digits = 3)
   }

Where dtol used by all.equal is set to be 5e-10,
which the intel compiler misses by 1.6e-13.
This tolerance value seems a bit arbitrary.
The gcc compiled version's passes the test with a 9.3e-11 error.


I am using the -mp option for the intel compilers, which was recommended
on this mailing list previously and would make sense given the docs:

   Floating Point Optimization Options
       -mp    Maintain floating-point  precision  (disables  some
              optimizations).  The -mp option restricts optimiza-
              tion to maintain declared precision and  to  ensure
              that   floating-point   arithmetic   conforms  more
              closely to the ANSI and IEEE  standards.  For  most
              programs,  specifying this option adversely affects
              performance. If  you  are  not  sure  whether  your
              application  needs  this  option, try compiling and
              running your program both with and  without  it  to
              evaluate the effects on both performance and preci-
              sion.


Has anyone else encountered this?  

-Frank

Thank you for the report.
Your points are quite valid (see below).

Note that these precision tests were a bit stringent on purpose
because pchisq() eventually needs some accuracy improvements
which I had been investigating (when I did some minor
improvements).
>>>>> "FrankSa" == Samuelson, Frank* <Samuelson>
>>>>>     on Thu, 19 Aug 2004 12:11:00 -0400 writes:
    FrankSa> I compiled the 1.9.1 src.rpm with the standard gnu tools and it
works.
    FrankSa> I tried compiling the 1.9.1 src.rpm with the Intel 8 C and
FORTRAN
    FrankSa> compilers and it bombs out during the testing phase:

    FrankSa> comparing 'd-p-q-r-tests.Rout' to
'./d-p-q-r-tests.Rout.save' ...267c267
    FrankSa> < df = 0.5[1] "Mean relative  difference:
5.001647e-10"
    FrankSa> ---
    >> df = 0.5[1] TRUE
    FrankSa> make[3]: *** [d-p-q-r-tests.Rout] Error 1
    FrankSa> make[3]: Leaving directory
`/usr/src/redhat/BUILD/R-1.9.1/tests'
    FrankSa> make[2]: *** [test-Specific] Error 2
    FrankSa> make[2]: Leaving directory
`/usr/src/redhat/BUILD/R-1.9.1/tests'
    FrankSa> make[1]: *** [test-all-basics] Error 1
    FrankSa> make[1]: Leaving directory
`/usr/src/redhat/BUILD/R-1.9.1/tests'
    FrankSa> make: *** [check-all] Error 2
    FrankSa> error: Bad exit status from /var/tmp/rpm-tmp.63044 (%build)

    FrankSa> looking at the differences between the failed file and the
standard, I get:

    FrankSa> fws wolf tests] diff  d-p-q-r-tests.Rout.save
d-p-q-r-tests.Rout.fail
    FrankSa> 3c3
    FrankSa> < Version 1.9.0 Patched (2004-04-19), ISBN 3-900051-00-3
    FrankSa> ---
    >> Version 1.9.1  (2004-06-21), ISBN 3-900051-00-3
    FrankSa> 281c281
    FrankSa> < df = 0.5[1] TRUE
    FrankSa> ---
    >> df = 0.5[1] "Mean relative  difference: 5.001647e-10"
    FrankSa> 935c935
    FrankSa> < Time elapsed:  7.83 0.04 16.1 0 0 
    FrankSa> ---
    >> Time elapsed:  2.49 0.01 2.55 0 0 

    FrankSa> Besides being 3 times faster, it's stopping on the following
code:

(intel's new C compiler,  what hardware?)

    FrankSa> for(df in c(0.1, 0.5, 1.5, 4.7, 10, 20,50,100)) {
    FrankSa>  cat("df =", formatC(df, wid=3))
    FrankSa>  xx <- c(10^-(5:1), .9, 1.2, df + c(3,7,20,30,35,38))
    FrankSa>  pp <- pchisq(xx, df=df, ncp = 1) #print(pp)
    FrankSa>  dtol <- 1e-12 *(if(2 < df && df <= 50) 64 else
if(df > 50) 20000 else 500)
    FrankSa>  print(all.equal(xx, qchisq(pp, df=df, ncp=1), tol = dtol))#
TRUE
    FrankSa>  ##or print(mapply(rErr, xx, qchisq(pp, df=df,ncp=1)), digits =
3)
    FrankSa> }

    FrankSa> Where dtol used by all.equal is set to be 5e-10,
    FrankSa> which the intel compiler misses by 1.6e-13.
    FrankSa> This tolerance value seems a bit arbitrary.

indeed.  Replacing 500 by 501 above would already let pass your setup

    FrankSa> The gcc compiled version's passes the test with a 9.3e-11
error.


    FrankSa> I am using the -mp option for the intel compilers, which was
recommended
    FrankSa> on this mailing list previously and would make sense given the
docs:

(still be interested to hear what happens with the test when you
 do *not* set "-mp")

    FrankSa> Floating Point Optimization Options
    FrankSa> -mp    Maintain floating-point  precision  (disables  some
    FrankSa> optimizations).  The -mp option restricts optimiza-
    FrankSa> tion to maintain declared precision and  to  ensure
    FrankSa> that   floating-point   arithmetic   conforms  more
    FrankSa> closely to the ANSI and IEEE  standards.  For  most
    FrankSa> programs,  specifying this option adversely affects
    FrankSa> performance. If  you  are  not  sure  whether  your
    FrankSa> application  needs  this  option, try compiling and
    FrankSa> running your program both with and  without  it  to
    FrankSa> evaluate the effects on both performance and preci-
    FrankSa> sion.


    FrankSa> Has anyone else encountered this?