This partially fixes bug 061 reported by Victor Liu.

Some compilers allocate local arrays on the heap when their size is
above a particular threshold.
This leads to wrong results when multiple threads call the same routine.

The bug fix consists in using a larger workspace, as proposed by Victor Liu
(Tue Nov 13, 2012).
Some routines still have large local arrays and cannot be fixed that way
because they have no workspace parameter: xGBTRF, xBPTRF, and xHSEQR.

1 files changed, 17 insertions, 26 deletions

diff --git a/SRC/dormql.f b/SRC/dormql.f
index 96c6f195..512f234e 100644
--- a/SRC/dormql.f
+++ b/SRC/dormql.f
@@ -136,9 +136,7 @@
 *>          The dimension of the array WORK.
 *>          If SIDE = 'L', LWORK >= max(1,N);
 *>          if SIDE = 'R', LWORK >= max(1,M).
-*>          For optimum performance LWORK >= N*NB if SIDE = 'L', and
-*>          LWORK >= M*NB if SIDE = 'R', where NB is the optimal
-*>          blocksize.
+*>          For good performance, LWORK should generally be larger.
 *>
 *>          If LWORK = -1, then a workspace query is assumed; the routine
 *>          only calculates the optimal size of the WORK array, returns
@@ -185,17 +183,15 @@
 *  =====================================================================
 *
 *     .. Parameters ..
-      INTEGER            NBMAX, LDT
-      PARAMETER          ( NBMAX = 64, LDT = NBMAX+1 )
+      INTEGER            NBMAX, LDT, TSIZE
+      PARAMETER          ( NBMAX = 64, LDT = NBMAX+1,
+     $                     TSIZE = LDT*NBMAX )
 *     ..
 *     .. Local Scalars ..
       LOGICAL            LEFT, LQUERY, NOTRAN
-      INTEGER            I, I1, I2, I3, IB, IINFO, IWS, LDWORK, LWKOPT,
+      INTEGER            I, I1, I2, I3, IB, IINFO, IWT, LDWORK, LWKOPT,
      $                   MI, NB, NBMIN, NI, NQ, NW
 *     ..
-*     .. Local Arrays ..
-      DOUBLE PRECISION   T( LDT, NBMAX )
-*     ..
 *     .. External Functions ..
       LOGICAL            LSAME
       INTEGER            ILAENV
@@ -239,25 +235,22 @@
          INFO = -7
       ELSE IF( LDC.LT.MAX( 1, M ) ) THEN
          INFO = -10
+      ELSE IF( LWORK.LT.NW .AND. .NOT.LQUERY ) THEN
+         INFO = -12
       END IF
 *
       IF( INFO.EQ.0 ) THEN
+*
+*        Compute the workspace requirements
+*
          IF( M.EQ.0 .OR. N.EQ.0 ) THEN
             LWKOPT = 1
          ELSE
-*
-*           Determine the block size.  NB may be at most NBMAX, where
-*           NBMAX is used to define the local array T.
-*
             NB = MIN( NBMAX, ILAENV( 1, 'DORMQL', SIDE // TRANS, M, N,
      $                               K, -1 ) )
-            LWKOPT = NW*NB
+            LWKOPT = NW*NB + TSIZE
          END IF
          WORK( 1 ) = LWKOPT
-*
-         IF( LWORK.LT.NW .AND. .NOT.LQUERY ) THEN
-            INFO = -12
-         END IF
       END IF
 *
       IF( INFO.NE.0 ) THEN
@@ -276,14 +269,11 @@
       NBMIN = 2
       LDWORK = NW
       IF( NB.GT.1 .AND. NB.LT.K ) THEN
-         IWS = NW*NB
-         IF( LWORK.LT.IWS ) THEN
-            NB = LWORK / LDWORK
+         IF( LWORK.LT.NW*NB+TSIZE ) THEN
+            NB = (LWORK-TSIZE) / LDWORK
             NBMIN = MAX( 2, ILAENV( 2, 'DORMQL', SIDE // TRANS, M, N, K,
      $              -1 ) )
          END IF
-      ELSE
-         IWS = NW
       END IF
 *
       IF( NB.LT.NBMIN .OR. NB.GE.K ) THEN
@@ -296,6 +286,7 @@
 *
 *        Use blocked code
 *
+         IWT = 1 + NW*NB
          IF( ( LEFT .AND. NOTRAN ) .OR.
      $       ( .NOT.LEFT .AND. .NOT.NOTRAN ) ) THEN
             I1 = 1
@@ -320,7 +311,7 @@
 *           H = H(i+ib-1) . . . H(i+1) H(i)
 *
             CALL DLARFT( 'Backward', 'Columnwise', NQ-K+I+IB-1, IB,
-     $                   A( 1, I ), LDA, TAU( I ), T, LDT )
+     $                   A( 1, I ), LDA, TAU( I ), WORK( IWT ), LDT )
             IF( LEFT ) THEN
 *
 *              H or H**T is applied to C(1:m-k+i+ib-1,1:n)
@@ -336,8 +327,8 @@
 *           Apply H or H**T
 *
             CALL DLARFB( SIDE, TRANS, 'Backward', 'Columnwise', MI, NI,
-     $                   IB, A( 1, I ), LDA, T, LDT, C, LDC, WORK,
-     $                   LDWORK )
+     $                   IB, A( 1, I ), LDA, WORK( IWT ), LDT, C, LDC,
+     $                   WORK, LDWORK )
    10    CONTINUE
       END IF
       WORK( 1 ) = LWKOPT