[R-pkg-devel] Solaris SPARC, Fortran, and logical errors?

I completely agree that testing on SPARC Solaris is valuable, however
much of a nuisance it is.  But I also agree that it would be great if
we could find a way to provide a publicly accessible SPARC Solaris
testing framework.

On Thu, Mar 16, 2017 at 6:49 AM, Uwe Ligges
<ligges at statistik.tu-dortmund.de> wrote:

On 15.03.2017 18:30, Ben Bolker wrote:

On 17-03-15 11:09 AM, J C Nash wrote:

Possibly tangential, but has there been any effort to set up a Sparc
testbed? It
seems we could use a network-available (virtual?) machine, since this
platform is
often the unfortunate one. Unless, of course, there's a sunset date.

For information, I mentioned SPARC at our local linux group, and
apparently there
are a couple of folk who have them running, but I didn't find out the
state of the
OS etc.

JN

  The virtual machine platforms I know of (admittedly not a complete
list!) only support Solaris on x86, e.g.

Yes, you cannot emulate a Sparc in an efficient way on an amd64 platform.

I take the opportunity to repeat why testing on *Sparc Solaris* gives many
benefits:

- this way we cover big- and little-endian platforms (i.e. for future
stability so that it works on what appear to be still esoteric such as ARM
based architectures or so)
- we cover one of the commercial unixes, i.e. we see
  + how stuff works on the the typically rather old toolchains
  + and what happens in on gnu/gcc-setups and how much GNUisms are used

Best,
Uwe Ligges

https://community.oracle.com/thread/2569292

On 2017-03-15 10:40 AM, Avraham Adler wrote:

Hello.

The Delaporte package works properly on all R-core platforms except
Solaris SPARC, where it  compiles properly but fails a number of its
tests [1]. Not having access to a SPARC testbed, I'm limited in what
kind of diagnostics I can do. One thing I have noticed is that a lot
of the failures occur when I am passing non-default logicals (like
lower tail or log). For example, the first failure at that link is
when "log = true" is supposed to be passed, but the SPARC answers are
the unlogged values. Of the 22 failed tests, 12 of them pass logicals.

I'll bring an example of how it is coded below, and if anyone
recognizes where SPARC specifically goes wrong, I'd appreciate. I
guess, if I absolutely had to, I could convert the logical to an
integer in C and pass the integer to Fortran which should work even
for SPARC, but I'd prefer not to if I could help it.

Thank you,

Avi

[1] https://cran.r-project.org/web/checks/check_results_Delaporte.html

*****Example Code*****

R code:

ddelap <- function(x, alpha, beta, lambda, log = FALSE){
  if(!is.double(x)) {storage.mode(x) <- 'double'}
  if(!is.double(alpha)) {storage.mode(alpha) <- 'double'}
  if(!is.double(beta)) {storage.mode(beta) <- 'double'}
  if(!is.double(lambda)) {storage.mode(lambda) <- 'double'}
  if(any(x > floor(x))) {
    warning("Non-integers passed to ddelap. These will have 0
probability.")
  }
  .Call(ddelap_C, x, alpha, beta, lambda, log)
}

C code:

void ddelap_f(double *x, int nx, double *a, int na, double *b, int nb,
double *l, int nl,
              int *lg, double *ret);

extern SEXP ddelap_C(SEXP x, SEXP alpha, SEXP beta, SEXP lambda, SEXP
lg){
  const int nx = LENGTH(x);
  const int na = LENGTH(alpha);
  const int nb = LENGTH(beta);
  const int nl = LENGTH(lambda);
  SEXP ret;
  PROTECT(ret = allocVector(REALSXP, nx));
  ddelap_f(REAL(x), nx, REAL(alpha), na, REAL(beta), nb, REAL(lambda),
nl, LOGICAL(lg), REAL(ret));
  UNPROTECT(1);
  return(ret);
}

Fortran: (not posting ddelap_f_s as that doesn't handle the logging)

    subroutine ddelap_f(x, nx, a, na, b, nb, l, nl, lg, pmfv) bind(C,
name="ddelap_f")

    integer(kind = c_int), intent(in), value         :: nx, na, nb, nl
    ! Sizes
    real(kind = c_double), intent(in), dimension(nx) :: x
    ! Observations
    real(kind = c_double), intent(out), dimension(nx):: pmfv
    ! Result
    real(kind = c_double), intent(in)                :: a(na), b(nb),
l(nl)! Parameters
    logical(kind = c_bool), intent(in)               :: lg
    ! Log flag
    integer                                          :: i
    ! Integer

        !$omp parallel do default(shared) private(i)
        do i = 1, nx
            if (x(i) > floor(x(i))) then
                pmfv(i) = ZERO
            else
                pmfv(i) = ddelap_f_s(x(i), a(mod(i - 1, na) + 1), &
                                     b(mod(i - 1, nb) + 1), l(mod(i -
1, nl) + 1))
            end if
        end do
        !$omp end parallel do

        if (lg) then
            pmfv = log(pmfv)
        end if

    end subroutine ddelap_f