CamCASP/Programming/2 - Revision history

import>Am592: /* Integrals that should be rotated */

2009-09-01T13:17:48Z

Integrals that should be rotated

← Older revision		Revision as of 13:17, 1 September 2009
Line 112:		Line 112:
	This is a list of integrals that should be rotated if '''REDO_DF_ON_ROTATION = .FALSE.''':		This is a list of integrals that should be rotated if '''REDO_DF_ON_ROTATION = .FALSE.''':
	# OVRL,XAXB		# OVRL,XAXB
			# KE__,XAXB
	# NUCA,__XB		# NUCA,__XB
	# NUCB,__XA		# NUCB,__XA

2009-03-02T17:27:27Z

Special Cases

← Older revision		Revision as of 17:27, 2 March 2009
Line 133:		Line 133:

	It is unlikely that this will happen as this is not normal usage of the code, but it could and it would be awful to track down. So, I need a work-around.		It is unlikely that this will happen as this is not normal usage of the code, but it could and it would be awful to track down. So, I need a work-around.

			Effects integrals:
			# NUCA,__XA
			# NUCB,__XB
			# SC__,_A_A
			# SC__,_B_B

	One solution is to use the %ForceRotate field to force rotation when this happens. This field will need to be reset after rotation is performed or else a rotation will be performed on a subsequent call even if nothing has changed.		One solution is to use the %ForceRotate field to force rotation when this happens. This field will need to be reset after rotation is performed or else a rotation will be performed on a subsequent call even if nothing has changed.

			A cleaner solution might be to set %Rotate=.true. for these integrals and re-calculate them when the molecules rotate. Then the RECALCULATE-ROTATE operations will always be performed on these integrals. This will be a waste of CPU most of the time.

			'''NEEDS RESOLVING'''

2009-03-02T17:12:23Z

About the code

← Older revision		Revision as of 17:12, 2 March 2009
Line 100:		Line 100:
	** (auxAB,auxA): '''recalculation''', but no rotation.		** (auxAB,auxA): '''recalculation''', but no rotation.

	The advantage of using '''df_redo_df_on_rotation = .FALSE.''' is clear. A lot of recalculations are avoided.		The advantage of using '''df_redo_df_on_rotation = .FALSE.''' is clear. A lot of recalculations are avoided. The biggest saving is the DF is not re-done for each monomer geometry.

	===Integral types and rotation===		===Integral types and rotation===

2009-03-02T17:07:05Z

Integrals that should be rotated

2009-03-02T17:00:39Z

Integrals that should be rotated

2009-03-02T16:52:03Z

Special Cases

← Older revision		Revision as of 16:52, 2 March 2009
Line 128:		Line 128:

	It is unlikely that this will happen as this is not normal usage of the code, but it could and it would be awful to track down. So, I need a work-around.		It is unlikely that this will happen as this is not normal usage of the code, but it could and it would be awful to track down. So, I need a work-around.

			One solution is to use the %ForceRotate field to force rotation when this happens. This field will need to be reset after rotation is performed or else a rotation will be performed on a subsequent call even if nothing has changed.

2009-03-02T16:47:22Z

Special Cases

← Older revision		Revision as of 16:47, 2 March 2009
Line 126:		Line 126:
	# NUCA,__XA is now re-calculated using the ''rotated'' monomer A, but symmetry blocks are '''not rotated''' because %Rotate=.false.		# NUCA,__XA is now re-calculated using the ''rotated'' monomer A, but symmetry blocks are '''not rotated''' because %Rotate=.false.
	# It is now inconsistent with DF solution.		# It is now inconsistent with DF solution.

			It is unlikely that this will happen as this is not normal usage of the code, but it could and it would be awful to track down. So, I need a work-around.

2009-03-02T16:46:24Z

Integrals that should be rotated

← Older revision		Revision as of 16:46, 2 March 2009
Line 111:		Line 111:
	===Integrals that should be rotated===		===Integrals that should be rotated===
	This is a list of integrals that should be rotated if '''REDO_DF_ON_ROTATION = .FALSE.''':		This is a list of integrals that should be rotated if '''REDO_DF_ON_ROTATION = .FALSE.''':
			# OVRL,XAXB
	#
			# NUCA,__XB
			# NUCB,__XA


			===Special Cases===
			Some integrals depend on parameters like the probe charge (PROB), regularization parameter (NUCA,NUCB,PROB) and DF constraints (SC__). So if these parameters change, these integrals will be re-calculated. Now, some of these, for example, NUCA,__XA, have %Rotate=.false., and so are not rotated when calculated. This can lead to an error. Consider the follow sequence of steps:
			# '''REDO_DF_ON_ROTATION=.false.'''
			# DF performed.
			# NUCA,__XA calculated. It is consistent with the DF solution.
			# Monomer A rotated.
			# NUCA,__XA not rotated because %Rotate=.false. Still consistent with DF.
			# '''reg_eta''' changed.
			# NUCA,__XA is now re-calculated using the ''rotated'' monomer A, but symmetry blocks are '''not rotated''' because %Rotate=.false.
			# It is now inconsistent with DF solution.

2009-03-02T15:53:15Z

Integral types and rotation

← Older revision		Revision as of 15:53, 2 March 2009
Line 108:		Line 108:
	# '''I%ForceRotate = .TRUE.''', or		# '''I%ForceRotate = .TRUE.''', or
	# '''I%Rotate = .TRUE. && REDO_DF_ON_ROTATION = .FALSE.'''.		# '''I%Rotate = .TRUE. && REDO_DF_ON_ROTATION = .FALSE.'''.

			===Integrals that should be rotated===
			This is a list of integrals that should be rotated if '''REDO_DF_ON_ROTATION = .FALSE.''':
			#

2009-03-02T14:44:32Z

About the code

← Older revision		Revision as of 14:44, 2 March 2009
Line 101:		Line 101:

	The advantage of using '''df_redo_df_on_rotation = .FALSE.''' is clear. A lot of recalculations are avoided.		The advantage of using '''df_redo_df_on_rotation = .FALSE.''' is clear. A lot of recalculations are avoided.

			===Integral types and rotation===
			Every integral type has the logical fields '''ForceRotate''' and '''Rotate'''. The former is set to FALSE by default. It can be set to TRUE on an integral-by-integral basis using the ''BEGIN DF-INTS...END'' block. The latter, I%Rotate, is set for every integral when they are defined. The above rules are used to decide the value of this field.

			Integrals will have their symmetry components rotated IF:
			# '''I%ForceRotate = .TRUE.''', or
			# '''I%Rotate = .TRUE. && REDO_DF_ON_ROTATION = .FALSE.'''.