f90 -O3 -mips4 -r12000 -64 -r8 -extend_source -flist mat_mul.f

C ***********************************************************
C Fortran file translated from WHIRL Thu Aug 16 15:21:34 2001
C ***********************************************************

	PROGRAM MAIN
	IMPLICIT NONE
C
C	**** Variables and functions ****
C
	CHARACTER*8 ENV_VALUE
	INTEGER(4) N
	INTEGER(4) OMP_NUM_THREADS
	INTEGER(8) t$1(1_8)
	SAVE t$1
	
	EXTERNAL getenv
	INTEGER(4) getenv
C
C	**** Temporary variables ****
C
	INTEGER(4) tmp0
C
C	**** Initializers ****
C
	DATA t$1 / 2902913183561809920_8 /
C
C	**** statements ****
C
	READ(5, *) N
	tmp0 = getenv('OMP_NUM_THREADS', ENV_VALUE)
	READ(ENV_VALUE(1_8 : 1), t$1) OMP_NUM_THREADS
	CALL array_mult(N, OMP_NUM_THREADS)
	STOP
	END

	SUBROUTINE array_mult(N, N_CPUS)
	IMPLICIT NONE
	INTEGER(4) N
	INTEGER(4) N_CPUS
C
C	**** Variables and functions ****
C
	REAL(8) deref_p_AA(t$4, t$4)
	POINTER(p_AA, deref_p_AA)
	INTEGER(8) t$4
	INTEGER(8) t$5
	REAL(8) ANG_INCR
	REAL(8) deref_p_BB(t$4, t$4)
	POINTER(p_BB, deref_p_BB)
	REAL(8) deref_p_CC(t$4, t$4)
	POINTER(p_CC, deref_p_CC)
	INTEGER(4) J
	INTEGER(4) K
	INTEGER(4) NSZ
	REAL(8) SECS
	REAL(8) TM1
	INTEGER(8) t$6
	INTEGER(8) t$7
	REAL(8) t$11
	REAL(8) deref_se1_F8(*)
	POINTER(se1_F8, deref_se1_F8)
	REAL(8) deref_se2_F8(*)
	POINTER(se2_F8, deref_se2_F8)
	EXTERNAL timef
	REAL(8) timef
C
C	**** Temporary variables ****
C
	INTEGER(8) f90sp_1
	INTEGER(8) f90sp_2
	INTEGER(8) f90sp_3
	INTEGER(8) se1__$stk
	INTEGER(8) se2__$stk
	INTEGER(4) seonly0I
	INTEGER(4) I4
	INTEGER(4) tile2J
	INTEGER(4) tile1I
	INTEGER(4) J0
	REAL(8) mi0
	REAL(8) mi1
	REAL(8) mi2
	REAL(8) mi3
	REAL(8) mi4
	REAL(8) mi5
	REAL(8) mi6
	REAL(8) mi7
	INTEGER(4) I0
	INTEGER(4) wd_J0
	REAL(8) mi8
	REAL(8) mi9
	REAL(8) mi10
	REAL(8) mi11
	INTEGER(4) I1
	INTEGER(4) wd_K
	INTEGER(4) J1
	REAL(8) mi12
	REAL(8) mi13
	INTEGER(4) I2
	INTEGER(4) wd_J
	REAL(8) mi14
	INTEGER(4) I3
	REAL(8) t$
C
C	**** statements ****
C
	t$4 = N
	t$5 = MAX(N, 0)
	t$6 = (MAX(N, 0) * 2)
	t$7 = (MAX(N, 0) * MAX(N, 0))
	f90sp_1 = OPR_ALLOCA(0)
	p_AA = OPR_ALLOCA(((MAX(N, 0) * MAX(N, 0)) * 8))
	f90sp_2 = OPR_ALLOCA(0)
	p_BB = OPR_ALLOCA(((MAX(N, 0) * MAX(N, 0)) * 8))
	f90sp_3 = OPR_ALLOCA(0)
	p_CC = OPR_ALLOCA(((MAX(N, 0) * MAX(N, 0)) * 8))
	ANG_INCR = (1.1769021394429878D+03 / DBLE((N * N)))
	DO J = 1, N, 1
	  se1__$stk = OPR_ALLOCA(0)
	  se1_F8 = OPR_ALLOCA((MIN(N, 1000) * 8))
	  se2__$stk = OPR_ALLOCA(0)
	  se2_F8 = OPR_ALLOCA((MIN(N, 1000) * 8))
	  DO seonly0I = 1, N, 1000
	    DO I4 = seonly0I, MIN(N, (seonly0I + 999)), 1
	      deref_se2_F8((I4 - seonly0I) + 1) = (DBLE((I4 +(NSZ *(J + -1)))) * ANG_INCR)
	      deref_se1_F8((I4 - seonly0I) + 1) = (DBLE((I4 +(NSZ *(J + -1)))) * ANG_INCR)
	    END DO
	    CALL vcos$(deref_se2_F8(1), deref_p_AA(seonly0I, J), %val(((MIN(N, (seonly0I + 999)) - seonly0I) + 1)), %val(1_8), %val(1_8))
	    CALL vsin$(deref_se1_F8(1), deref_p_BB(seonly0I, J), %val(((MIN(N, (seonly0I + 999)) - seonly0I) + 1)), %val(1_8), %val(1_8))
	  END DO
	  CALL OPR_DEALLOCA(se2__$stk, se2_F8)
	  CALL OPR_DEALLOCA(se1__$stk, se1_F8)
	END DO
	t$11 = timef()
	TM1 = t$11
	DO tile2J = 1, N, 160
	  DO tile1I = 1, N, 240
	    DO K = 1, (N + -3_8), 4
	      DO J0 = tile2J, MIN((tile2J + 158), (N + -1_8)), 2
	        mi0 = deref_p_BB(K, J0)
	        mi1 = deref_p_BB(K + 3, J0 + 1)
	        mi2 = deref_p_BB(K + 2, J0 + 1)
	        mi3 = deref_p_BB(K + 1, J0 + 1)
	        mi4 = deref_p_BB(K, J0 + 1)
	        mi5 = deref_p_BB(K + 1, J0)
	        mi6 = deref_p_BB(K + 3, J0)
	        mi7 = deref_p_BB(K + 2, J0)
	        DO I0 = tile1I, MIN(N, (tile1I + 239)), 1
	          deref_p_CC(I0, J0) = (deref_p_CC(I0, J0) +(deref_p_AA(I0, K) * mi0))
	          deref_p_CC(I0, J0) = (deref_p_CC(I0, J0) +(deref_p_AA(I0, K + 1) * mi5))
	          deref_p_CC(I0, J0) = (deref_p_CC(I0, J0) +(deref_p_AA(I0, K + 2) * mi7))
	          deref_p_CC(I0, J0) = (deref_p_CC(I0, J0) +(deref_p_AA(I0, K + 3) * mi6))
	          deref_p_CC(I0, J0 + 1) = (deref_p_CC(I0, J0 + 1) +(deref_p_AA(I0, K) * mi4))
	          deref_p_CC(I0, J0 + 1) = (deref_p_CC(I0, J0 + 1) +(deref_p_AA(I0, K + 1) * mi3))
	          deref_p_CC(I0, J0 + 1) = (deref_p_CC(I0, J0 + 1) +(deref_p_AA(I0, K + 2) * mi2))
	          deref_p_CC(I0, J0 + 1) = (deref_p_CC(I0, J0 + 1) +(deref_p_AA(I0, K + 3) * mi1))
	        END DO
	      END DO
	      DO wd_J0 = J0, MIN(N, (tile2J + 159)), 1
	        mi8 = deref_p_BB(K, wd_J0)
	        mi9 = deref_p_BB(K + 3, wd_J0)
	        mi10 = deref_p_BB(K + 1, wd_J0)
	        mi11 = deref_p_BB(K + 2, wd_J0)
	        DO I1 = tile1I, MIN(N, (tile1I + 239)), 1
	          deref_p_CC(I1, wd_J0) = (deref_p_CC(I1, wd_J0) +(deref_p_AA(I1, K) * mi8))
	          deref_p_CC(I1, wd_J0) = (deref_p_CC(I1, wd_J0) +(deref_p_AA(I1, K + 1) * mi10))
	          deref_p_CC(I1, wd_J0) = (deref_p_CC(I1, wd_J0) +(deref_p_AA(I1, K + 2) * mi11))
	          deref_p_CC(I1, wd_J0) = (deref_p_CC(I1, wd_J0) +(deref_p_AA(I1, K + 3) * mi9))
	        END DO
	      END DO
	    END DO
	    DO wd_K = K, N, 1
	      DO J1 = tile2J, MIN((tile2J + 158), (N + -1_8)), 2
	        mi12 = deref_p_BB(wd_K, J1)
	        mi13 = deref_p_BB(wd_K, J1 + 1)
	        DO I2 = tile1I, MIN(N, (tile1I + 239)), 1
	          deref_p_CC(I2, J1) = (deref_p_CC(I2, J1) +(deref_p_AA(I2, wd_K) * mi12))
	          deref_p_CC(I2, J1 + 1) = (deref_p_CC(I2, J1 + 1) +(deref_p_AA(I2, wd_K) * mi13))
	        END DO
	      END DO
	      DO wd_J = J1, MIN(N, (tile2J + 159)), 1
	        mi14 = deref_p_BB(wd_K, wd_J)
	        DO I3 = tile1I, MIN(N, (tile1I + 239)), 1
	          deref_p_CC(I3, wd_J) = (deref_p_CC(I3, wd_J) +(deref_p_AA(I3, wd_K) * mi14))
	        END DO
	      END DO
	    END DO
	  END DO
	END DO
	t$ = timef()
	SECS = (((t$ - TM1)) * 1.0D-03)
	WRITE(6, *)
	WRITE(6, *) 'NUMBER OF CPUS/TASKS ASSIGNED = ', N_CPUS
	WRITE(6, *) 'Parallelized Matrix Multiplication for ', N, 'x', N, ' Size'
	WRITE(6, *)
	WRITE(6, *) 'Elapsed Wall Clock Time ', SECS, ' Seconds'
	WRITE(6, *)
	WRITE(*, *) deref_p_CC(1234, 1), deref_p_CC(8023, 1), deref_p_CC(2345, 1)
	CALL OPR_DEALLOCA(f90sp_3, p_CC)
	CALL OPR_DEALLOCA(f90sp_2, p_BB)
	CALL OPR_DEALLOCA(f90sp_1, p_AA)
	RETURN
	END SUBROUTINE