SGI F77 COMPILER BASIC TEST - Origin 2000
                          Updated 99/03/17


OBJECTIVE

First tests on R10000: to see if compiler options are as important
as on R8000.


DESCRIPTION

Machine: Origin 2000, R10000, IP27: 2 x 250 MHz.
Program: Matrix multiplication using nested DO loops.
Matrix dimensions: ND = 800 (reservation).
Timing: dtime (tempd.f)
Program details:
    P25:    I-loop inside (vector style), no directives:

      PROGRAM P25
      ...
      IMPLICIT    REAL*8  (A-H,O-Z)
      PARAMETER   (ND = 800, NIN = 5, NOUT = 6)
      DIMENSION   A(ND,ND), B(ND,ND), C(ND,ND)
      ...
      DO  26  J = 1,N
          DO  24  K = 1,N
              DO  22  I = 1,N
                  C(I,J) = C(I,J) + A(I,K) * B(K,J)
   22         CONTINUE
   24     CONTINUE
   26 CONTINUE
      ...


RESULTS

Pfa was not installed.

Compiler options are less important than on R8000. Even without DGEMM,
one can reach 80 percent of speed; on R8000, this was 50 percent and 
80 percent was reached using DGEMM.

(Complib speedup is also less important than on R8000, but this was
tested for the CFHHM benchmark only.)


Table I.  ND = 1600, N = 1600. Theoretical CPU time lower limit is 16.4
seconds (250 MHz R10000). No DGEMM test.
--------------------------------------------------------------------------
Program   Compiler call                               Threads  CPU time
--------------------------------------------------------------------------
p25       f77 -r10000 -mips4 -64 -O3                            23.7

          f77 -r10000 -mips4 -64 -O3 
              -LNO:ou=2                                         25
              -LNO:ou=4                                         20.2  *1
              -LNO:ou=6                                         31.1

          f77 -Ofast=ip27                                       24.3

          f77 -Ofast=ip27
              -LNO:ou=4                                         21.7
--------------------------------------------------------------------------
*1  R = 16.4/20.2 = 0.81, while on R8000 the best R without DGEMM was 0.53 
    using advanced options.