[patch, libfortran] Speed up cshift for dim > 1

[Thomas Koenig]

Hello world,

the attached patch implements a blocked algorithm for
improving the speed of cshift for dim > 1.

It uses the fact that

   integer, dimension (n1,n2,n3) :: a, b

   b = cshift(a,shift,3)

is identical, as far as the memory locations is concerned.

   integer, dimension (n1*n2*n3) :: c, d
   d = cshift(c, shift*n1*n2, 1)

The speedup is quite large; from being really slow for
dim > 1, this patch makes it go even faster.

Below there are some comparisons for the attached benchmark,
do-1.f90. gfortran-7 uses the old library version.

Interestingly, the library version is also much faster
than an implementation of straight DO loops.

Regression-tested.  OK for trunk?

Regards

	Thomas


$ gfortran-7 -static-libgfortran -O3 do-1.f90 && ./a.out
  Testing explicit DO loops
  Dim =            1  Elapsed CPU time =    5.71363878
  Dim =            2  Elapsed CPU time =    5.40494061
  Dim =            3  Elapsed CPU time =    5.40769291
  Testing built-in cshift
  Dim =            1  Elapsed CPU time =    3.43479729
  Dim =            2  Elapsed CPU time =    11.7110386
  Dim =            3  Elapsed CPU time =    31.0966301
$ gfortran -static-libgfortran -O3 do-1.f90 && ./a.out
  Testing explicit DO loops
  Dim =            1  Elapsed CPU time =    5.73881340
  Dim =            2  Elapsed CPU time =    5.38435745
  Dim =            3  Elapsed CPU time =    5.38971329
  Testing built-in cshift
  Dim =            1  Elapsed CPU time =    3.42018127
  Dim =            2  Elapsed CPU time =    2.24075317
  Dim =            3  Elapsed CPU time =    2.23136330


2017-06-14  Thomas Koenig  <tkoenig@gcc.gnu.org>

         PR fortran/52473
         * m4/cshift0.m4:  For arrays that are contiguous up to
         shift, implement blocked algorighm for cshift.
         * generated/cshift0_c10.c:  Regenerated.
         * generated/cshift0_c16.c:  Regenerated.
         * generated/cshift0_c4.c:  Regenerated.
         * generated/cshift0_c8.c:  Regenerated.
         * generated/cshift0_i1.c:  Regenerated.
         * generated/cshift0_i16.c:  Regenerated.
         * generated/cshift0_i2.c:  Regenerated.
         * generated/cshift0_i4.c:  Regenerated.
         * generated/cshift0_i8.c:  Regenerated.
         * generated/cshift0_r10.c:  Regenerated.
         * generated/cshift0_r16.c:  Regenerated.
         * generated/cshift0_r4.c:  Regenerated.
         * generated/cshift0_r8.c:  Regenerated.

2017-06-14  Thomas Koenig  <tkoenig@gcc.gnu.org>

         PR fortran/52473
         * gfortran.dg/cshift_1.f90:  New test.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: p6.diff
Type: text/x-patch
Size: 48036 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20170614/a3f07b86/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cshift_1.f90
Type: text/x-fortran
Size: 3148 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20170614/a3f07b86/attachment-0001.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: do-1.f90
Type: text/x-fortran
Size: 2681 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20170614/a3f07b86/attachment-0002.bin>