Test which checks the omp parallel sections reduction directive with all its option. 2.0 omp parallel sections reduction INTEGER FUNCTION par_section_reduct() IMPLICIT NONE INTEGER sum, sum2, known_sum, i, i2,diff INTEGER product,known_product,int_const INTEGER MAX_FACTOR DOUBLE PRECISION dsum,dknown_sum,dt,dpt DOUBLE PRECISION rounding_error, ddiff INTEGER DOUBLE_DIGITS LOGICAL logic_and, logic_or, logic_eqv,logic_neqv INTEGER bit_and, bit_or INTEGER exclusiv_bit_or INTEGER min_value, max_value DOUBLE PRECISION dmin, dmax INTEGER result INCLUDE "omp_testsuite.f" LOGICAL logics(LOOPCOUNT) INTEGER int_array(LOOPCOUNT) DOUBLE PRECISION d_array(LOOPCOUNT) PARAMETER (int_const=10,known_product=3628800) PARAMETER (DOUBLE_DIGITS=20,MAX_FACTOR=10) PARAMETER (rounding_error=1.E-6) INTEGER cut1, cut2, cut3, cut4 dt = 1./3. known_sum = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2 product = 1 sum2 = 0 sum = 0 dsum = 0. result =0 logic_and = .TRUE. logic_or = .FALSE. bit_and = 1 bit_or = 0 exclusiv_bit_or = 0 cut1 = NINT(LOOPCOUNT / 3.) cut2 = cut1 + 1 cut3 = 2 * cut1 cut4 = cut3 + 1 !$omp parallel sections private(i) reduction(+:sum) !$omp section DO i =1, cut1 sum = sum + i END DO !$omp section DO i =cut2, cut3 sum = sum + i END DO !$omp section DO i =cut4, LOOPCOUNT sum = sum + i END DO !$omp end parallel sections IF (known_sum .NE. sum) THEN result = result + 1 WRITE(1,*) "Error in sum with integers: Result was ", & sum,"instead of ", known_sum END IF diff = (LOOPCOUNT * (LOOPCOUNT + 1)) / 2 !$omp parallel sections reduction (-: diff) !$omp section DO i =1, cut1 diff = diff - i END DO !$omp section DO i =cut2, cut3 diff = diff - i END DO !$omp section DO i =cut4, LOOPCOUNT diff = diff - i END DO !$omp end parallel sections IF ( diff .NE. 0 ) THEN result = result + 1 WRITE(1,*) "Error in difference with integers: Result was ", & sum,"instead of 0." END IF !... Test for doubles dsum =0. dpt = 1 DO i=1, DOUBLE_DIGITS dpt= dpt * dt END DO dknown_sum = (1-dpt)/(1-dt) !$omp parallel sections reduction(+:dsum) !$omp section DO i=0,6 dsum = dsum + dt**i END DO !$omp section DO i=7,12 dsum = dsum + dt**i END DO !$omp section DO i=13,DOUBLE_DIGITS-1 dsum = dsum + dt**i END DO !$omp end parallel sections IF(dsum .NE. dknown_sum .AND. & ABS(dsum - dknown_sum) .GT. rounding_error ) THEN result = result + 1 WRITE(1,*) "Error in sum with doubles: Result was ", & dsum,"instead of ",dknown_sum,"(Difference: ", & dsum - dknown_sum,")" END IF dpt = 1 DO i=1, DOUBLE_DIGITS dpt = dpt*dt END DO ddiff = ( 1-dpt)/(1-dt) !$omp parallel sections reduction(-:ddiff) !$omp section DO i=0, 6 ddiff = ddiff - dt**i END DO !$omp section DO i=7, 12 ddiff = ddiff - dt**i END DO !$omp section DO i=13, DOUBLE_DIGITS-1 ddiff = ddiff - dt**i END DO !$omp end parallel sections IF ( ABS(ddiff) .GT. rounding_error ) THEN result = result + 1 WRITE(1,*) "Error in Difference with doubles: Result was ", & ddiff,"instead of 0.0" END IF !$omp parallel sections reduction(*:product) !$omp section DO i=1,3 product = product * i END DO !$omp section DO i=4,6 product = product * i END DO !$omp section DO i=7,10 product = product * i END DO !$omp end parallel sections IF (known_product .NE. product) THEN result = result + 1 WRITE(1,*) "Error in Product with integers: Result was ", & product," instead of",known_product END IF DO i=1,LOOPCOUNT logics(i) = .TRUE. END DO !$omp parallel sections reduction(.AND.:logic_and) !$omp section DO i=1,cut1 logic_and = logic_and .AND. logics(i) END DO !$omp section DO i=cut2,cut3 logic_and = logic_and .AND. logics(i) END DO !$omp section DO i=cut4,LOOPCOUNT logic_and = logic_and .AND. logics(i) END DO !$omp end parallel sections if (.not. logic_and) then result = result + 1 write(1,*) "Error in logic AND part 1" end if logic_and = .TRUE. logics(LOOPCOUNT/2) = .FALSE. !$omp parallel sections reduction(.AND.:logic_and) !$omp section DO i=1,cut1 logic_and = logic_and .AND. logics(i) END DO !$omp section DO i=cut2,cut3 logic_and = logic_and .AND. logics(i) END DO !$omp section DO i=cut4,LOOPCOUNT logic_and = logic_and .AND. logics(i) END DO !$omp end parallel sections IF (logic_and) THEN result = result + 1 WRITE(1,*) "Error in logic AND pass 2" END IF DO i=1, LOOPCOUNT logics(i) = .FALSE. END DO !$omp parallel sections reduction(.OR.:logic_or) !$omp section DO i = 1, cut1 logic_or = logic_or .OR. logics(i) END DO !$omp section DO i = cut2, cut3 logic_or = logic_or .OR. logics(i) END DO !$omp section DO i = cut4, LOOPCOUNT logic_or = logic_or .OR. logics(i) END DO !$omp end parallel sections IF (logic_or) THEN result = result + 1 WRITE(1,*) "Error in logic OR part 1" END IF logic_or = .FALSE. logics(LOOPCOUNT/2) = .TRUE. !$omp parallel sections reduction(.OR.:logic_or) !$omp section DO i=1,cut1 logic_or = logic_or .OR. logics(i) END DO !$omp section DO i=cut2,cut3 logic_or = logic_or .OR. logics(i) END DO !$omp section DO i=cut4,LOOPCOUNT logic_or = logic_or .OR. logics(i) END DO !$omp end parallel sections IF ( .NOT. logic_or ) THEN result = result + 1 WRITE(1,*) "Error in logic OR part 2" END IF !... Test logic EQV, unique in Fortran DO i=1, LOOPCOUNT logics(i) = .TRUE. END DO logic_eqv = .TRUE. !$omp parallel sections reduction(.EQV.:logic_eqv) !$omp section DO i = 1, cut1 logic_eqv = logic_eqv .EQV. logics(i) END DO !$omp section DO i = cut2, cut3 logic_eqv = logic_eqv .EQV. logics(i) END DO !$omp section DO i = cut4, LOOPCOUNT logic_eqv = logic_eqv .EQV. logics(i) END DO !$omp end parallel sections IF (.NOT. logic_eqv) THEN result = result + 1 WRITE(1,*) "Error in logic EQV part 1" END IF logic_eqv = .TRUE. logics(LOOPCOUNT/2) = .FALSE. !$omp parallel sections reduction(.EQV.:logic_eqv) !$omp section DO i=1,cut1 logic_eqv = logic_eqv .EQV. logics(i) END DO !$omp section DO i=cut2,cut3 logic_eqv = logic_eqv .EQV. logics(i) END DO !$omp section DO i=cut4,LOOPCOUNT logic_eqv = logic_eqv .EQV. logics(i) END DO !$omp end parallel sections IF ( logic_eqv ) THEN result = result + 1 WRITE(1,*) "Error in logic EQV part 2" END IF !... Test logic NEQV, which is unique in Fortran DO i=1, LOOPCOUNT logics(i) = .FALSE. END DO logic_neqv = .FALSE. !$omp parallel sections reduction(.NEQV.:logic_neqv) !$omp section DO i = 1, cut1 logic_neqv = logic_neqv .NEQV. logics(i) END DO !$omp section DO i = cut2, cut3 logic_neqv = logic_neqv .NEQV. logics(i) END DO !$omp section DO i = cut4, LOOPCOUNT logic_neqv = logic_neqv .NEQV. logics(i) END DO !$omp end parallel sections IF (logic_neqv) THEN result = result + 1 WRITE(1,*) "Error in logic NEQV part 1" END IF logic_neqv = .FALSE. logics(LOOPCOUNT/2) = .TRUE. !$omp parallel sections reduction(.NEQV.:logic_neqv) !$omp section DO i=1,cut1 logic_neqv = logic_neqv .NEQV. logics(i) END DO !$omp section DO i=cut2,cut3 logic_neqv = logic_neqv .NEQV. logics(i) END DO !$omp section DO i=cut4,LOOPCOUNT logic_neqv = logic_neqv .NEQV. logics(i) END DO !$omp end parallel sections IF ( .NOT. logic_neqv ) THEN result = result + 1 WRITE(1,*) "Error in logic NEQV part 2" END IF DO i=1, LOOPCOUNT int_array(i) = 1 END DO !$omp parallel sections reduction(IAND:bit_and) !... iand(I,J): Returns value resulting from boolean AND of !... pair of bits in each of I and J. !$omp section DO i=1, cut1 bit_and = IAND(bit_and,int_array(i)) END DO !$omp section DO i=cut2, cut3 bit_and = IAND(bit_and,int_array(i)) END DO !$omp section DO i=cut4, LOOPCOUNT bit_and = IAND(bit_and,int_array(i)) END DO !$omp end parallel sections IF ( bit_and .LT. 1 ) THEN result = result + 1 WRITE(1,*) "Error in IAND part 1" END IF bit_and = 1 int_array(LOOPCOUNT/2) = 0 !$omp parallel sections reduction(IAND:bit_and) !$omp section DO i=1, cut1 bit_and = IAND ( bit_and, int_array(i) ) END DO !$omp section DO i=cut2, cut3 bit_and = IAND ( bit_and, int_array(i) ) END DO !$omp section DO i=cut4, LOOPCOUNT bit_and = IAND ( bit_and, int_array(i) ) END DO !$omp end parallel sections IF( bit_and .GE. 1) THEN result = result + 1 WRITE(1,*) "Error in IAND part 2" END IF DO i=1, LOOPCOUNT int_array(i) = 0 END DO !$omp parallel sections reduction(IOR:bit_or) !... Ior(I,J): Returns value resulting from boolean OR of !... pair of bits in each of I and J. !$omp section DO i=1, cut1 bit_or = IOR(bit_or, int_array(i) ) END DO !$omp section DO i=cut2, cut3 bit_or = IOR(bit_or, int_array(i) ) END DO !$omp section DO i=cut4, LOOPCOUNT bit_or = IOR(bit_or, int_array(i) ) END DO !$omp end parallel sections IF ( bit_or .GE. 1) THEN result = result + 1 WRITE(1,*) "Error in Ior part 1" END IF bit_or = 0 int_array(LOOPCOUNT/2) = 1 !$omp parallel sections reduction(IOR:bit_or) !$omp section DO i=1, cut1 bit_or = IOR(bit_or, int_array(i) ) END DO !$omp section DO i=cut2, cut3 bit_or = IOR(bit_or, int_array(i) ) END DO !$omp section DO i=cut4, LOOPCOUNT bit_or = IOR(bit_or, int_array(i) ) END DO !$omp end parallel sections IF ( bit_or .LE. 0) THEN result = result + 1 WRITE(1,*) "Error in Ior part 2" END IF DO i=1, LOOPCOUNT int_array(i) = 0 END DO !$omp parallel sections reduction(IEOR:exclusiv_bit_or) !$omp section DO i = 1, cut1 exclusiv_bit_or = IEOR(exclusiv_bit_or, int_array(i)) END DO !$omp section DO i = cut2, cut3 exclusiv_bit_or = IEOR(exclusiv_bit_or, int_array(i)) END DO !$omp section DO i = cut4, LOOPCOUNT exclusiv_bit_or = IEOR(exclusiv_bit_or, int_array(i)) END DO !$omp end parallel sections IF ( exclusiv_bit_or .GE. 1) THEN result = result + 1 WRITE(1,*) "Error in Ieor part 1" END IF exclusiv_bit_or = 0 int_array(LOOPCOUNT/2) = 1 !$omp parallel sections reduction(IEOR:exclusiv_bit_or) !$omp section DO i = 1, cut1 exclusiv_bit_or = IEOR(exclusiv_bit_or, int_array(i)) END DO !$omp section DO i = cut2, cut3 exclusiv_bit_or = IEOR(exclusiv_bit_or, int_array(i)) END DO !$omp section DO i = cut4, LOOPCOUNT exclusiv_bit_or = IEOR(exclusiv_bit_or, int_array(i)) END DO !$omp end parallel sections IF ( exclusiv_bit_or .LE. 0) THEN result = result + 1 WRITE(1,*) "Error in Ieor part 2" END IF DO i=1,LOOPCOUNT int_array(i) = 10 - i END DO min_value = 65535 !$omp parallel sections reduction(MIN:min_value) !$omp section DO i = 1, cut1 min_value = MIN(min_value,int_array(i) ) END DO !$omp section DO i = cut2, cut3 min_value = MIN(min_value,int_array(i) ) END DO !$omp section DO i = cut4, LOOPCOUNT min_value = MIN(min_value,int_array(i) ) END DO !$omp end parallel sections IF ( min_value .GT. (10-LOOPCOUNT) )THEN result = result + 1 WRITE(1,*) "Error in integer MIN" END IF DO i=1,LOOPCOUNT int_array(i) = i END DO max_value = -32768 !$omp parallel sections reduction(MAX:max_value) !$omp section DO i = 1, cut1 max_value = MAX(max_value,int_array(i) ) END DO !$omp section DO i = cut2, cut3 max_value = MAX(max_value,int_array(i) ) END DO !$omp section DO i = cut4, LOOPCOUNT max_value = MAX(max_value,int_array(i) ) END DO !$omp end parallel sections IF ( max_value .LT. LOOPCOUNT )THEN result = result + 1 WRITE(1,*) "Error in integer MAX" END IF !... test double min, max DO i=1,LOOPCOUNT d_array(i) = 10 - i*dt END DO dmin = 2**10 dt = 0.5 !$omp parallel sections reduction(MIN:dmin) !$omp section DO i = 1, cut1 dmin= MIN(dmin,d_array(i) ) END DO !$omp section DO i = cut2, cut3 dmin= MIN(dmin,d_array(i) ) END DO !$omp section DO i = cut4, LOOPCOUNT dmin= MIN(dmin,d_array(i) ) END DO !$omp end parallel sections IF ( dmin .GT. (10-dt) )THEN result = result + 1 WRITE(1,*) "Error in double MIN" END IF DO i=1,LOOPCOUNT d_array(i) = i * dt END DO dmax= - (2**10) !$omp parallel sections reduction(MAX:dmax) !$omp section DO i = 1, cut1 dmax= MAX(dmax,d_array(i) ) END DO !$omp section DO i = cut2, cut3 dmax= MAX(dmax,d_array(i) ) END DO !$omp section DO i = cut4, LOOPCOUNT dmax= MAX(dmax,d_array(i) ) END DO !$omp end parallel sections IF ( dmax .lt. LOOPCOUNT*dt )THEN result = result + 1 write(1,*) "Error in double MAX" END IF IF ( result .EQ. 0 ) THEN = 1 ELSE = 0 END IF CLOSE(2) END FUNCTION