C MATRIX DETERMINANT
	SUBROUTINE MDET(XVBLS,I1,I2,J1,J2,DET)
	INCLUDE 'VKLUGPRM.FTN'
	REAL*8 XVBLS(1),DET,SUMA,SUMB
C NOTE XVBLS IS RRW BY RCL MATRIX IN PORTACALC
C I1,I2 ARE TOP COL,ROW COORD; J1,J2 ARE BOTTOM
C STORAGE OF XVBLS IS (COL,ROW) SO LOCATIONS INSIDE
C IT ARE
C  ADDR=(ROW-1)*RRW+COL (RRW IS # OF COLS)
	DET=0.
	N=J1-I1+1
	M=J2-I2+1
	IF(N.NE.M)RETURN
	IF(N.LE.1)RETURN
C ONLY SQUARE MATRICES MAY HAVE NONZERO DETERMINANTS
C ALSO, DIMENSION HAS TO BE > 1
	NN=N
C  FIXUP... (OK FOR N=2,3 ANYHOW)
	IF(N.EQ.2)NN=N-1
C  SUM OVER DIAGS...
C MULTIPLY DIAGONALS FROM TOP AND BOTTOM ROWS OF MATRIX AND GET
C DIFFERENCE EACH TIME FOR ACCURACY
	DO 1 N1=1,NN
	SUMA=1.
	SUMB=1.
	DO 2 N2=1,N
	NCL=N1+N2-1
	N2L=N+1-N2
	IF(NCL.GT.N)NCL=NCL-N
C NOW MULTIPLY SUMA (POSITIVE TERMS) BY X(NCL,N2) AND SUMB(NEG TERMS)
C BY X(NCL,N2L)
	LA=(N2-2+I2)*RRW+I1+NCL-1
	LB=(N2L-2+I2)*RRW+I1+NCL-1
	CALL XVBLGT(LA,0,XVBLS(1))
	SUMA=SUMA*XVBLS(1)
	CALL XVBLGT(LB,0,XVBLS(1))
	SUMB=SUMB*XVBLS(1)
2	CONTINUE
C NOW ACCUMULATE TERMS IN DETERMINANT
	DET=DET+SUMA-SUMB
C DO IN THIS ORDER TO AVOID EXCESSIVE LOSS OF PRECISION DUE TO
C DIFFERENCES OF LARGE TERMS. THIS IS BAD ENOUGH AS IT IS...
1	CONTINUE
	RETURN
	END