RI_MP2

References: DelBen2013

Parameters influencing the RI-MP2 method. RI-MP2 supports gradients. [Edit on GitHub]

Keywords

• SECTION_PARAMETERS

• BLOCK_SIZE

• NUMBER_INTEGRATION_GROUPS

• PRINT_DGEMM_INFO

Keyword descriptions

SECTION_PARAMETERS: logical = F

Lone keyword: T

Usage: &RI_MP2 .TRUE.

Putting the &RI_MP2 section activates RI-MP2 calculation. [Edit on GitHub]

BLOCK_SIZE: integer = -1

Aliases: MESSAGE_SIZE

Usage: BLOCK_SIZE 2

Determines the blocking used for communication in RI-MP2. Larger BLOCK_SIZE reduces communication but requires more memory. The default (-1) is automatic. [Edit on GitHub]

NUMBER_INTEGRATION_GROUPS: integer = -1

Usage: NUMBER_INTEGRATION_GROUPS 2

Sets the number of integration groups of the communication scheme in RI-MP2. Integrals will be replicated such that each integration group has all integrals available. Must be a divisor of the number of subgroups (see GROUP_SIZE keyword in the WF_CORRELATION section. Smaller groups reduce the communication costs but increase the memory developments. If the provided value is non-positive or not a divisor of the number of subgroups, the number of integration groups is determined automatically (default). [Edit on GitHub]

PRINT_DGEMM_INFO: logical = F

Lone keyword: T

Print details about all DGEMM calls. [Edit on GitHub]