JP4867962B2 - Multi-body problem calculation apparatus, two-body interaction calculation circuit, and multi-body problem calculation method - Google Patents

Description

  The present invention relates to a many-body problem calculation technique.

  In the field of molecular dynamics, the behavior of liquids, solids, and polymers is considered to be the result of the movement of the atoms or molecules that compose them, and research is conducted by simulating the movement of these particles. The problem of dealing with a system consisting of many particles that interact with each other is called a many-body problem, and molecular dynamics calculates many-body problems with atoms and molecules as particles.

  Since interaction works between the particles of the system handled in the many-body problem, the force acting between the particles becomes an important physical quantity in the many-body problem and is always calculated. The force acting on a particle is the sum of the forces exerted by all other particles.

  In addition, since the pressure and volume of a substance change with temperature change, and the volume change means the change of the position of particles, pressure is also an important physical quantity. Furthermore, the virial used to determine the pressure is frequently calculated as an important physical quantity. Virial is a product obtained by multiplying the force acting on each particle by the particle coordinate and multiplying the product. The pressure can be obtained from the virial system.

  An example of an apparatus for obtaining virial is disclosed in Patent Document 1. In Patent Document 1, a virial is obtained by multiplying a coordinate difference between two particles and a force acting between the two particles.

  Referring to FIG. 8, the calculation device disclosed in Patent Document 1 includes a coordinate difference calculation unit 81, a distance square calculation unit 82, a function calculation unit 83, a delay unit 84, a force calculation unit 85, a virial. A calculation unit 86, a force total calculation unit 87, and a virial total calculation unit 88 are included.

Here, among the many particles in the system, the particle to be calculated is a particle i, and the other particle that affects the particle i is a particle j. Also, the coordinates of the particle i are (x _i , y _i , z _i ), and the coordinates of the particle j are (x _j , y _j , z _j ). Let F be the magnitude of the force that particle j exerts on particle i.

The coordinate difference calculation unit 81 receives the coordinates of the two particles i and j, and calculates the coordinate differences Δx _j , Δy _j and Δz _j according to the equations (1) to (3).

Δx _j = x _j −x _i (1)
Δy _j = y _j −y _i (2)
Δz _j = z _j −z _i (3)

  The distance square calculation unit 82 calculates the square of the interparticle distance r from the coordinate difference obtained by the coordinate difference calculation unit 81 according to Equation (4).

r ² = (Δx _j ) ² + (Δy _j ) ² + (Δz _j ) ² (4)

The function calculation unit 83 receives the square of the interparticle distance r obtained by the distance square calculation unit 82 as an input, and calculates a value F / r obtained by dividing the force F exerted on the particle i from the particle j by the interparticle distance r. . The force acting between the particles is given in a different function form depending on the type of force and the type of particle. In any case, a value can be uniquely obtained if the interparticle distance r is determined. Therefore, the force F is obtained from the square of the interparticle distance r, and a value F / r obtained by dividing the force by the distance is obtained.

The delay unit 84 delays the coordinate differences Δx _j , Δy _j , Δz _j obtained by the coordinate difference calculation unit 81. This delay is for adjusting the time required for the calculation in the function calculation unit 83.

The force calculation unit 85 receives the coordinate differences Δx _j , Δy _j , Δz _j from the delay unit 84 and the value F / r from the function calculation unit 83 and inputs each coordinate axis direction according to equations (5) to (7). The force components Fx, Fy, and Fz are calculated.

Fx = F × (Δx _j / r) = (F / r) × Δx _j (5)
Fy = F × (Δy _j / r) = (F / r) × Δy _j (6)
Fz = F × (Δz _j / r) = (F / r) × Δz _j (7)

The virial calculation unit 86 receives the coordinate differences Δx _j , Δy _j , Δz _j from the delay unit 84 and the force components Fx, Fy, Fz from the force calculation unit 85 as inputs, and the equations (8) to (10) The virial components Vx, Vy, and Vz in the coordinate axis directions are calculated according to

Vx = Fx × Δx _j (8)
Vy = Fy × Δy _j (9)
Vz = Fz × Δz _j (10)

  The force summation calculation unit 87 receives all the particles other than the particle i from the force calculation unit 85 as the particle j, inputs the force components Fx, Fy, and Fz of the particle j, and converts all the particles j into the particle i. Find the total force exerted. The virial sum calculation unit 88 calculates the virial of the entire system.

JP-A-8-287042

According to the technique described in Patent Document 1, it is possible to obtain a force and its sum, and a virial and its sum. However, in the technique described in Patent Literature 1, (F / r) is multiplied by the coordinate difference in the force calculation unit, and the force and the coordinate difference are multiplied in the virial calculation unit. In addition, since the virial calculation section also requires a multiplication circuit, there is a problem that the amount of hardware increases. In order to solve such problems, it can be considered that the virial calculation unit is realized by software. However, when calculating the virial of the entire system, the order of the calculation amount with respect to the number N of particles in the system is O (N ² ), and if it is calculated by software, the calculation time becomes enormous.

  Therefore, the object of the present invention is to increase the amount of hardware when the virial of the entire system is calculated using hardware, and when the virial of the entire system is calculated using software. Is to provide a technique for solving the problem of enormous calculation time.

A first many-body problem calculation apparatus according to the present invention is:
One of the particles in the basic cell in the system to which the periodic boundary condition is applied is selected as a calculation target particle, and the image cells of the periodic boundary condition are sequentially set as processing units, and the particles in the image cell are the calculation target. The process of obtaining the sum of the forces acting on the particles is repeated until all the particles in the basic cell are selected as calculation target particles, and the sum of the forces acting on the particles for each particle in the basic cell is calculated. A total force calculation unit to find and hold;
A selection circuit that sequentially selects each coordinate component of the sum of forces each time the sum of forces in an image cell is obtained in the force sum calculator;
Based on the positional relationship between the image cell and the basic cell, which is a processing unit in the force sum calculation unit, the coordinate component of the sum of the forces selected by the selection circuit is output as it is, or the sign is inverted. An addition / subtraction control circuit for outputting or setting the output to 0;
A holding circuit that holds the addition result for each component of the virial; and
An addition for adding the output of the addition / subtraction control circuit and the addition result of the component corresponding to the output of the addition / subtraction control circuit among the addition results held in the holding circuit and storing the addition result in the holding circuit And
For each virial component, using the particle coordinates of the particles in the basic cell and the sum of the forces acting on the particles in the basic cell held in the force sum calculation unit, A first process for obtaining a product sum with the particle coordinates, a second process for multiplying the addition result of the component held in the holding circuit by a constant, and a product sum obtained by the first process. A software execution unit that executes software for performing a third process for obtaining a difference between the result and the multiplication result obtained in the second process;

A first two-body interaction calculation circuit according to the present invention includes:
A two-body interaction calculation circuit that is a component of a multi-body problem calculation device,
One of the particles in the basic cell in the system to which the periodic boundary condition is applied is selected as a calculation target particle, and the image cells of the periodic boundary condition are sequentially set as processing units, and the particles in the image cell are the calculation target. The process of obtaining the sum of the forces acting on the particles is repeated until all the particles in the basic cell are selected as calculation target particles, and the sum of the forces acting on the particles for each particle in the basic cell is calculated. A total force calculation unit to find and hold;
A selection circuit that sequentially selects each coordinate component of the sum of forces each time the sum of forces in an image cell is obtained in the force sum calculator;
Based on the positional relationship between the image cell and the basic cell, which is a processing unit in the force sum calculation unit, the coordinate component of the sum of the forces selected by the selection circuit is output as it is, or the sign is inverted. An addition / subtraction control circuit for outputting or setting the output to 0;
A holding circuit that holds the addition result for each component of the virial; and
An addition for adding the output of the addition / subtraction control circuit and the addition result of the component corresponding to the output of the addition / subtraction control circuit among the addition results held in the holding circuit and storing the addition result in the holding circuit With a vessel.

The first many-body problem calculation method according to the present invention is:
The force summation calculation unit selects one of the particles in the basic cell in the system to which the periodic boundary condition is applied as a calculation target particle, and sequentially sets the image cell of the periodic boundary condition as a processing unit. The process of obtaining the sum of the forces exerted by the particles on the calculation target particles is repeated until all the particles in the basic cell are selected as calculation target particles, and for each particle in the basic cell, the particle To find and hold the sum of the forces
Each time the selection circuit calculates the total force of the image cell unit in the force total calculation unit, it sequentially selects each coordinate component of the total force,
The addition / subtraction control circuit outputs, as it is, the coordinate component of the sum of the forces selected by the selection circuit based on the positional relationship between the image cells and the basic cells that are processing units in the force sum calculation unit, Invert the sign to output, or set the output to 0,
An adder adds the output of the addition / subtraction control circuit and the addition result of the component corresponding to the output of the addition / subtraction control circuit among the addition results for each virial component held in the holding circuit, and adds Storing the result in the holding circuit;
For each component of the virial component, the software execution unit uses the particle coordinates of the particles in the basic cell and the sum of the forces acting on the particles in the basic cell held in the force sum calculation unit. A first process for obtaining the sum of products of the total force and the particle coordinates, a second process for multiplying the addition result for the component held in the holding circuit and a constant, and the first process The software for performing the third process for obtaining the difference between the product-sum result obtained in step 1 and the multiplication result obtained in the second process is executed.

  ADVANTAGE OF THE INVENTION According to this invention, the amount of hardware of a many-body problem calculation apparatus can be reduced, and the increase in calculation time can be prevented.

  Next, the best mode for carrying out the present invention will be described in detail with reference to the drawings.

First, the basic concept of the present invention will be described. The present invention surrounds the image cell indicated by around the basic cell = lattice vector _n represented by the lattice vector n = (0,0,0), as shown in FIG. _{1 (n x, n y,} n z) Under the periodic boundary condition, the virial calculation formula is expressed by the second term on the right side where the order of the calculation amount with respect to the number N of particles in the system is O (N ² ), as shown in Equation (11), and the calculation amount Is expressed in a form separated into the first term on the right side where the order of O (N) is O (N).

Here, Formula (11) will be described in detail. Equation (11) is obtained when the length of each side indicated by the lattice vector n = (0, 0, 0) is L _x , L in FIG. 2 in which FIG. 1 is viewed in the XY plane where the z-axis lattice vector is 0. The position of a central particle i in the basic cell of _y and L _z is expressed by coordinates (x _i , y _i , z _i ), and the position of a particle j other than the particle i in the basic cell is expressed by coordinates (x _j , y _j , expressed as z _j) represents the lattice vector n = (n _x, n _y, the particles jn in the image cell n represented by n _z) coordinates (x _jn, y _jn, with z _jn), with respect to the particle i Te x-coordinate component of the force f _ij exerted by particle j and f _xij, the x-coordinate component of the force f _ijn exerted by particles jn and _f xijn, the sum of the forces exerted by all particles to the particle i This represents the virial V _xx of the xx component of the pressure tensor when the x coordinate component is F _xi .

In the present invention, F _xi is calculated by hardware in the order O (N) of Expression (11), and the multiplication of F _xi and xi and the multiplication result are performed by software. Further, the part of ΣΣn _x Σf _{xijn in} the part of the order O (N ² ) in the expression (11) is indicated by this lattice vector n based on the value of the lattice vector n indicating the position of the image cell with respect to the basic cell. The total sum of the forces in the image cell is calculated using hardware having a circuit configuration for adding to or subtracting from the previous integrated value. This circuit configuration does not require a multiplier and can reduce the amount of hardware and power consumption. Further, the multiplication of the constant 1/2 in the portion of the order O (N ² ) and the length Lx in the x-axis direction of the basic cell, and the multiplication result (1/2 · L _x ) of this multiplication and ΣΣn _x Σf _{Multiplication} with _xijn is calculated by software.

The components of the pressure tensor include the nine components yy, zz, xy, yz, zz, yz, yz, and xz including the xx component shown in the equation (11). 11) The coordinate x _i , the total force F _xi , the length L _x of one side of the basic cell, the lattice vector n _x , and the sum Σf _xijn of the force exerted from the particle jn are simply changed to the y coordinate component and the z coordinate component. And the form of the calculation formula is the same. For the xy, yx, yz, zy, zx, and xz components, the calculation formula of the xy component is given as an example, and the calculation formula shown in Formula (12) is obtained.

Equation (12) changes the coordinate x _i in equation (11) to the y coordinate y _i , changes the length L _x in the x-axis direction of the basic cell to the length L _y in the y-axis direction, and the x-coordinate component n _x is replaced by a y-coordinate component n _y. Partial _{ΣΣn x} Σf xijn be computed in _{hardware, ΣΣn y Σf xijn, ΣΣn z} Σf xijn, paying attention ... to a circuit for calculating this part is the sum of each coordinate component of the force by the particles j included in the image cell .SIGMA.f _{By inputting xijn} , Σf _yijn , Σf _zijn and each coordinate component n _x , n _y , _{nz of} the lattice vector, and selecting the input according to the component to be calculated, calculation is performed for each component. It can be seen that all components can be calculated using a hardware amount calculation circuit for one component without providing a circuit. The force is calculated for each central particle i like F _xi , F _yi , and F _zi , whereas Virial calculates the sum for the entire system unlike the force, so prepare multiple pipelines to increase the degree of parallelism. Even if they are raised, all are added eventually, so there is only one virial adder circuit and one addition result holding means. By having one virial calculation circuit in the multi-body problem calculation device as described above, it is not necessary to have an adder circuit and a holding circuit for all components of the virial, and increase the pipeline to increase the degree of parallelism. However, it is possible to reduce the amount of hardware without having to increase the number of virial calculation circuits.

[First embodiment of the present invention]
Referring to FIG. 3, a multi-body problem calculation apparatus 1 according to the first embodiment of the present invention includes a two-body interaction calculation circuit 2, a software execution unit 3 realized by a CPU, and a control unit 4. And.

  The two-body interaction calculation circuit 2 includes a center particle coordinate holding unit 21, a basic cell information holding unit 22, a lattice vector holding unit 23, a coordinate difference calculation unit 24, a coordinate difference delay unit 25, and a distance square calculation. A unit 26, a function calculator 27, a force calculator 28, a force sum calculator 29, a virial calculator 30, and a lattice vector delay unit 31.

In the central particle coordinate holding unit 21, coordinates r _i (x _i , y _i , z _i ) of the central particle i to be calculated by the control unit 4 are set.

In the basic cell information holding unit 22, the size (L _x , L _y , L _z ) of the basic cell is set by the control unit 4.

A lattice vector n ( _nx , _ny , _nz ) indicating an image cell to be calculated by the control unit 4 is set in the lattice vector holding unit 23.

The coordinate difference calculation unit 24 receives the coordinates r _jn (x _jn , y _jn , z _jn ) of the particle j to be calculated input from the control unit 4 and the central particle i held in the central particle coordinate holding unit 21. The coordinate difference from the coordinate r _in (x _in , y _in , z _in ) is calculated. Specifically, since it is a periodic boundary condition, the coordinates r _i (x _i , y _i , z _i ) of the central particle i and the coordinates r _jn (x _jn , y _jn , z _jn ) of the particle j are mirror images. transformed coordinates _{r jn '= (x jn +} n x L x, y jn + n y L y, z jn + n z L z) difference between _{r ijn = (x i - (} x jn + n x L x), y i _{- (y jn + n y L} y), z i - to calculate the _{(z jn + n z L z} )).

The coordinate difference delay unit 25 has the same configuration as the delay unit 84 shown in FIG. 8, and the coordinate difference r _ijn calculated by the coordinate difference calculation unit 24 is used in the force calculation unit 28 by the coordinate difference r _ijn. Delay (hold) until the timing is reached.

The distance square calculation unit 26 has the same configuration as the distance square calculation unit 82 shown in FIG. 8, and calculates the square of distance | r _ijn | ² from the coordinate difference r _ijn .

The function calculation unit 27 has the same configuration as the function calculation unit 83 shown in FIG. 8, and calculates f _ijn / r _ijn .

The force calculation unit 28 has the same configuration as that of the force calculation unit 85 shown in FIG. 8, and the force f _ijn = (f that the particle j exerts on the central particle i from f _ijn / r _ijn and the coordinate difference r _{ijn. xijn,} f _yijn, to calculate the f _zijn).

The force sum calculation unit 29 calculates the sum Σf _ijn of the force exerted on the center particle i to be calculated from the force f _ijn calculated by the force calculation unit 28 for each image cell, and outputs it to the virial calculation unit 30. . Further, the force total calculation unit 29 calculates the total force F _i = (F _xi , F _yi , F _zi ) exerted on the particle i for each central particle i to be calculated, and the force holding circuit 291 (FIG. 4).

  The lattice vector delay unit 31 delays (holds) the lattice vector n held in the lattice vector holding unit 23 until the timing at which the lattice vector n is used in the virial calculation unit 30.

The virial calculation unit 30 calculates the virial from the total force Σf _ijn calculated by the total force calculation unit 29 and the lattice vector n.

  Referring to FIG. 4, the virial calculation unit 30 includes selection circuits 301, 302, and 306, an addition / subtraction control circuit 303, an adder 304, and a virial holding circuit 305.

The selection circuit 301 sequentially selects each component Σfx _ijn , Σf _yijn , Σf _zijn of the sum of forces in cell units input from the force sum calculation unit 29 and outputs the selected components to the addition / subtraction control circuit 303. The selection circuit 302 sequentially selects the coordinate components _nx , _ny , _nz of the lattice vector n input from the lattice vector delay unit 31 and outputs them to the addition / subtraction control circuit 303.

The addition / subtraction control circuit 303 performs the following processing depending on whether the value of the coordinate component of the lattice vector n selected by the selection circuit 302 is “positive”, “negative”, or “0”.
In the case of “positive”; the coordinate component of the sum of forces in cell units selected by the selection circuit 301 is output to the adder 304 as it is.
In the case of “negative”: the coordinate component of the total force of the cell unit selected by the selection circuit 301 is inverted and output to the adder 304.
In the case of “0”: 0 is output to the adder 304.

  The virial holding circuit 305 holds an integrated value of each coordinate component of the virial. The selection circuit 306 sequentially selects the integrated value of each coordinate component of the virial held in the virial holding circuit 305 and outputs it to the adder 304. The adder 304 adds the output of the addition / subtraction control circuit 303 and the output of the selection circuit 306 and stores the result in the virial holding circuit 305.

[Description of Operation of First Embodiment]
Next, the operation of the present embodiment will be described in detail with reference to the flowchart of FIG.

First, the control unit 4 sets the size L (L _x , L _y , L _z ) of the basic cell in the basic cell information holding unit 22 (step S500). Next, the control unit 4 selects one particle to be calculated from the particles included in the basic cell, and uses the coordinates r _i (x _i , y _i , z _i ) of the particle _i as the central particle coordinates. Set to the holding unit 21 (step S501). Thereafter, the control unit 4 selects one of the image cells, and sets a lattice vector n ( _nx , _ny , _nz ) indicating the image cell in the lattice vector holding unit 23 (step S502).

Thereafter, the control unit 4 selects one particle in the image cell selected in step S502, and inputs the coordinates r _jn (x _jn , y _jn , z _jn ) of the selected particle jn to the coordinate difference calculation unit 24. (Step S503).

As a result, the coordinate difference calculation unit 24 calculates the coordinate difference r _ijn between the particle i and the particle jn, and the force calculation unit 28 _{converts the} coordinate difference r _ijn to f _ijn / r _ijn obtained by the function calculation unit 27. Based on this, the force f _ijn exerted on the particle i by the particle jn is calculated, and the calculation result is output to the force sum calculating unit 29 (step S504). The force sum calculation unit 29 integrates the force f _ijn output from the force calculation unit 28 (step S505).

The controller 4 sequentially calculates the coordinates r _jn of the particles jn in the image cell until the particles jn in the image cell selected in step S502 are selected (until Y in step S506). Output to. Then, when all the particles jn in the image cell are selected and the above process is completed, the control unit 4 instructs the virial calculation unit 30 to perform virial calculation (step S507). At this time, each coordinate component of the total force Σf _ijn of the force exerted on the particle i by the particle jn in the image cell selected in step S502 is output from the force total calculation unit 29. The lattice vector delay unit 31 outputs a lattice vector n = ( _nx , _ny , _nz ) indicating the image cell selected in step S502.

The virial calculation unit 30, when receiving the instruction, first, the selection circuit 302 is lattice vector _{n = (n x, n y} , n z) and select the x-coordinate component n _x of, .SIGMA.f sum of the selection circuit 301 forces The x, y, and z coordinate components of _ijn are sequentially selected, and the selection circuit 306 sequentially selects the virial XX, YX, and ZX components held in the virial holding circuit 305. Subtraction control circuit 303, if the value of n _x which is output from the selection circuit 302 is "positive", it outputs x of sum .SIGMA.f _ijn force sequentially output from the selection circuit 301, y, z coordinates component If “negative”, the x, y, z coordinate components of the total force Σf _ijn sequentially output from the selection circuit 301 are inverted and output, and if “0”, 0 is output. The adder 304 adds the values relating to the x, y, z coordinate components of the total force Σf _ijn sequentially output from the addition / subtraction control circuit 303 and the virial XX, YX, ZX components sequentially output from the selection circuit 306. Then, the addition result is stored again in the virial holding circuit 305.

After that, the selection circuit 302 selects the y coordinate component n _y of the lattice vector n, the selection circuit 301 sequentially selects the x, y, z coordinate components of the total force Σf _ijn , and the selection circuit 306 selects the virial XY, YY. , ZY components are selected. Thereby, the adder 304 updates the XY, YY, and ZY components of the virial held in the virial holding circuit 305.

Finally, the selection circuit 302 selects the z-coordinate components n _z grating vector n, x the sum .SIGMA.f _ijn of the selection circuit 301 forces, y, sequentially selects the z-coordinate components, the selection circuit 306 virial XZ, YZ, and ZZ components are sequentially selected. As a result, the adder 304 updates the XZ, YZ, and ZZ components of the virial held in the virial holding circuit 305 (step S508).

  Thereafter, the control unit 4 determines whether or not the virial calculation has been completed for all the image cells in the system (step S509). If the virial calculation has not been completed for all the image cells, the control unit 4 returns to step S502 and selects the next image cell. On the other hand, if the calculation of virials for all the image cells is completed, it is determined whether the calculation of virials for all the particles i in the basic cell is completed (step S510).

  If the calculation of virials for all the particles i in the basic cell has not been completed, the control unit 4 returns to step S501 and selects the next particle i. On the other hand, if the calculation of virials for all the particles i has been completed, the control unit 4 ends the calculation using the two-body interaction calculation circuit 2 (hardware), and the software execution unit 3 Is activated (step S511).

Thus, the software execution unit 3, virial calculator 30 virial hold circuit 305 to hold has been that the components _{ΣΣn x} Σf xijn _{virial, ΣΣn y Σf yijn, ΣΣn z} Σf zijn, ΣΣn y Σf xijn, ΣΣn z Σf _yijn , ΣΣn _x Σf _zijn , ΣΣn _x Σf _yijn , ΣΣny _y Σf _zijn , and ΣΣn _z Σf _xijn are acquired, and the total force F for each particle i held in the force holding circuit 291 of the force sum calculating unit 29 _i is acquired (step S512).

Thereafter, the software execution unit 3 calculates the virial of the entire system based on the virial acquired in step S512 and the total force (step S513). Here, virials of xx, yy, zz, xy, yz, zx, yz, zy, and xz components are respectively calculated. For example, virial V _{xx of the} xx component is calculated as follows. First, the calculation shown in the first term of equation (1) is performed. That is, for each particle i, the x-coordinate component of the total force acting on the particle i is multiplied by the x-coordinate of the particle i to obtain the sum (first calculation result). Next, the calculation shown in the second term of Expression (2) is performed. That is, ΣΣn _x Σf _xijn acquired from the virial holding circuit 305 in the virial calculation unit 30 is multiplied by a constant ½ and the basic cell length L _x to obtain a multiplication result (second calculation result). . Thereafter, the virial V _xx is obtained by subtracting the second calculation result from the first calculation result.

[Effect of the first embodiment]
According to the present embodiment, it is possible to prevent an increase in calculation time while reducing the amount of hardware of the multi-body problem calculation apparatus. The reason for this is that the processing for obtaining the virial of the entire system is divided into a part where the order of computational complexity is large but multiplication is not required, and a part where the order of computational complexity is small but multiplication is necessary, This is because the portion where the order of the amount is increased is calculated by the virial calculation unit 30 (hardware), and the portion where the order of the amount of calculation is decreased is calculated by the software execution unit 3 (software).

  Further, in the present embodiment, a selection circuit 301 that sequentially selects each component of the total force obtained by the total force calculation unit 29 is provided, and a plurality of coordinate components are calculated using one adder 304. Since this can be done, the amount of hardware can be further reduced.

[Second Embodiment of the Present Invention]
Next, a second embodiment of the present invention will be described in detail. The present embodiment is characterized in that the hardware can be reduced even when the two-body interaction calculation circuit is configured using a plurality of pipelines.

  Referring to FIG. 6, a multi-body problem calculating apparatus 1a according to the second embodiment of the present invention is a two-body mutual calculation instead of the two-body interaction calculation circuit 2, the software execution unit 3, and the control unit 4. The point which is provided with the action calculation circuit 2a, the software execution part 3a, and the control part 4a is different from 1st Embodiment shown in FIG.

  The two-body interaction calculation circuit 2a includes first and second pipelines 2-1 and 2-2, and a virial calculation unit 30a.

  The first pipeline 2-1 has the same configuration as the two-body interaction calculation circuit 2 shown in FIG. 3 except that the virial calculation unit is not provided. The second pipeline 2-2 has a configuration similar to that of the first pipeline 2-1.

The control unit 4a controls the coordinates r _i (center particle i to be calculated in the central particle coordinate holding unit 21, the basic cell information holding unit 22, and the lattice vector holding unit 23 in each of the pipelines 2-1 and 2-2. x _i , y _i , z _i ), basic cell size L (L _x , L _y , L _z ), and lattice vector n ( _nx , _ny , _nz ) indicating the image cell to be calculated The first point is that the coordinates r _jn (x _jn , y _jn , z _jn ) of the particle j to be calculated are supplied to the coordinate difference calculation unit 24 in each of the pipelines 2-1 and 2-2. This is different from the control unit 4 of the embodiment. The control unit 4a sets a common value for each pipeline for the basic cell information holding unit 22 and the lattice vector holding unit 23, but sets a different value for each pipeline for the central particle coordinate holding unit 21. To do. The same value is supplied to the coordinate difference calculation unit 24 of each pipeline.

The software execution unit 3a obtains the total force F _i for each particle i from the force sum calculation unit 29 in each of the pipelines 2-1 and 2-2, and the software execution unit 3 according to the first embodiment. Is different.

  Referring to FIG. 7, the virial calculation unit 30a of the present embodiment includes a selection circuit 301a instead of the selection circuit 301. The virial calculation unit 30a of the first embodiment shown in FIG. It is different.

The selection circuit 301a first sequentially selects each component Σf _xijn , Σf _yijn , and Σf _zijn of the sum of forces in cell units input from the force sum calculation unit 29 in the first pipeline 2-1, and then The difference from the selection circuit 301 in the first embodiment shown in FIG. 4 is that the components Σf _xijn , Σf _yijn , and Σf _zijn input from the second pipeline 2-2 are sequentially selected. The selection circuit 301a performs the above-described process every time the selection circuit 302 selects a certain coordinate component.

[Description of Operation of Second Embodiment]
Next, the operation of the present embodiment will be described with reference to the flowchart of FIG.

First, the control unit 4a sets the size L (L _x , L _y , L _z ) of the basic cell in the basic cell information holding unit 22 in each pipeline 2-1 and 2-2 (step S500). Next, the control unit 4a selects one particle to be calculated in each of the pipelines 2-1 and 2-2 from the particles included in the basic cell (for the pipeline 2-1). Particle i1, particle i2 selected for pipeline 2-2), coordinates r _i1 (x _i1 , y _i1 , z _i1 ), r _i2 (x _i2 ) of particles i1, i2 for each selected pipeline , Y _i2 , z _i2 ) are set in the central particle coordinate holding unit 21 in each of the pipelines 2-1 and 2-2 (step S501). Thereafter, the control unit 4a selects one of the image cell, the point to image cell lattice vectors _{n (n x, n y,} n z) and, for each pipeline 21 and 22 It is set in the lattice vector holding unit 23 (step S502).

Thereafter, the control unit 4a selects one particle in the image cell selected in step S502, and sets the coordinates r _jn (x _jn , y _jn , z _jn ) of the selected particle jn to each pipeline 2-1, 2. -2 is input to the coordinate difference calculation unit 24 (step S503).

Thereby, in the first pipeline 2-1, the force calculation unit 28 calculates the force f _i1jn that the particle jn exerts on the particle i1, and the force sum calculation unit 29 accumulates the force f _i1jn (steps S504 and S505). ). In the second pipeline 2-2, the force calculation unit 28 calculates the force f _i2jn that the particle jn exerts on the particle i2, and the force sum calculation unit 29 accumulates the force f _i2jn (steps S504 and S505). .

The controller 4a sequentially selects the coordinates r _jn of the particles jn in the image cell in order to select all the particles jn in the image cell selected in step S502 (until Y in step S506). It outputs to the coordinate difference calculation part 24 in 1 and 2-2. When all the particles jn in the image cell are selected, the control unit 4a instructs the virial calculation unit 30a to perform virial calculation (step S507). At this time, each coordinate component of the total force Σf _i1jn of the force that the particle jn in the image cell selected in step S502 exerts on the particle i1 is output from the force sum calculation unit 29 in the first pipeline 2-1. From the force total calculation unit 29 in the second pipeline 2-2, each coordinate component of the total force Σf _i2jn of the force exerted on the particle i2 by the particle jn in the image cell selected in step S502 is output. The lattice vector delay unit 31 outputs a lattice vector n = ( _nx , _ny , _nz ) indicating the image cell selected in step S502.

The virial calculation unit 30a, when receiving the instruction, first selects the x-coordinate component n _x of = selection circuit 302 is the grating vector _{n (n x, n y,} n z), the selection circuit 301a is first pipe The x, y, z coordinate components of the total force Σf _i1jn output from the force total calculation unit 29 in the line 2-1 and the force total calculation unit 29 in the second pipeline 2-2 are output. The x, y, z coordinate components of the total force Σf _i2jn are sequentially selected, and the selection circuit 306 sequentially selects the XX, YX, ZX, XX, YX, ZX components of the virial held in the virial holding circuit 305. Subtraction control circuit 303, depending on the value of n _x being output from the selection circuit 302, the sum .SIGMA.f _I1jn force is outputted from the selection circuit 301a x, y, z coordinate component, the power of the sum .SIGMA.f _I2jn The x, y, and z coordinate components are output as they are, or the sign is inverted, or the value is set to 0. The adder 304 includes a value regarding the x, y, z coordinate components of the total force Σf _i1jn sequentially output from the addition / subtraction control circuit 303, a value regarding the x, y, z coordinate components of the total force Σf _i2jn , and a selection circuit 306. Are sequentially added to the XX, YX, ZX, XX, YX, and ZX components of the virial, and the addition result is stored in the virial holding circuit 305 again.

Thereafter, the selection circuit 302 selects the y coordinate component n _y grating vector n, x the sum .SIGMA.f _I1jn of the selection circuit 301 forces, y, z coordinate component, the sum of the force .SIGMA.f _I2jn of x, y, z coordinate component Are sequentially selected, and the selection circuit 306 sequentially selects the XY, YY, ZY, XY, YY, and ZY components of Virial. As a result, the XY, YY and ZY components of the virial held in the virial holding circuit 305 are updated by the adder 304.

And, finally, select the z coordinate components n _z of the selection circuit 302 is lattice vector n, x the sum .SIGMA.f _I1jn of the selection circuit 301 forces, y, z coordinate component, the sum of the force .SIGMA.f _I2jn of x, y, The z coordinate components are sequentially selected, and the selection circuit 306 sequentially selects the virial XZ, YZ, ZZ, XZ, YZ, and ZZ components. As a result, the XZ, YZ, and ZZ components of the virial held in the virial holding circuit 305 are updated by the adder 304 (step S508).

  Thereafter, the control unit 4a determines whether or not the virial calculation has been completed for all the image cells in the system (step S509). If the virial calculation has not been completed for all the image cells, the control unit 4 returns to step S502 and selects the next image cell. On the other hand, if the calculation of virials for all the image cells is completed, it is determined whether the calculation of virials for all the particles i in the basic cell is completed (step S510).

  If the calculation of virials for all the particles i in the basic cell is not completed, the control unit 4 returns to step S501 and selects the next particle i for each of the pipelines 2-1 and 2-2. To do. On the other hand, if the calculation of virials for all the particles i has been completed, the control unit 4 ends the calculation using the two-body interaction calculation circuit 2a (hardware), and the software execution unit 3a. Is activated (step S511).

Thus, the software execution section 3a virial calculator 30a in virial hold circuit 305 to hold has been that the components _{ΣΣn x} Σf xijn _{virial, ΣΣn y Σf yijn, ΣΣn z} Σf zijn, ΣΣn y Σf xijn, ΣΣn z Σf _yijn , ΣΣn _x Σf _zijn , ΣΣn _x Σf _yijn , ΣΣny _y Σf _zijn , ΣΣn _z Σf _xijn are obtained, and the force for each particle i from the force sum calculation unit 29 in each of the pipelines 2-1 and 2-2. get the total F _i (step S512).

  Thereafter, the software execution unit 3a calculates the virial of the entire system based on the virial acquired in step S512 and the total force (step S513).

[Effects of Second Embodiment]
According to the present embodiment, even when the two-body interaction calculation circuit 2a is configured using a plurality of pipelines 2-1, 2-2, an effect that hardware can be reduced can be obtained. The reason is that it includes a selection circuit 301a that sequentially selects each component of the sum of the forces obtained by the force sum calculation unit 29 in each of the pipelines 2-1 and 2-2, and includes a plurality of adders 304. This is because the pipelines 2-1 and 2-2 can be shared.

  INDUSTRIAL APPLICABILITY The present invention can be used in the field of drug discovery research for applications such as protein structural analysis and functional analysis, and for evaluation of protein-compound binding properties.

It is a figure for demonstrating how this invention utilizes a periodic boundary condition. It is the figure which looked at the periodic boundary condition of FIG. 1 in the XY plane where the lattice vector of Z axis is 0. It is a block diagram of 1st Embodiment of the many-body problem calculation apparatus concerning this invention. 3 is a block diagram illustrating a configuration example of a virial calculation unit 30. FIG. It is a flowchart for demonstrating operation | movement of 1st and 2nd embodiment. It is a block diagram of 2nd Embodiment of the many-body problem calculation apparatus concerning this invention. It is a block diagram which shows the structural example of the virial calculation part 30a. FIG. 11 is a block diagram for explaining a computing device described in Patent Document 1.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 1a ... Many-body problem calculation device 21 ... Center particle coordinate holding part 22 ... Basic cell information holding part 23 ... Lattice vector holding part 24 ... Coordinate difference calculation part 25 ... Coordinate difference delay part 26 ... Distance square calculation part 27 ... Function Calculation unit 28 ... force calculation unit 29 ... force sum calculation unit 291 ... force holding circuit 30, 30a ... virial calculation unit 301, 301a, 302, 306 ... selection circuit 303 ... addition / subtraction control circuit 304 ... adder 305 ... virial holding circuit 31 ... Lattice vector delay unit 2, 2a ... Inter-body interaction calculation circuit 3, 3a ... Software execution unit 4, 4a ... Control unit

Claims (10)

One of the particles in the basic cell in the system to which the periodic boundary condition is applied is selected as a calculation target particle, and the image cells of the periodic boundary condition are sequentially set as processing units, and the particles in the image cell are the calculation target. The process of obtaining the sum of the forces acting on the particles is repeated until all the particles in the basic cell are selected as calculation target particles, and the sum of the forces acting on the particles for each particle in the basic cell is calculated. A total force calculation unit to find and hold;
A selection circuit that sequentially selects each coordinate component of the sum of forces each time the sum of forces in an image cell is obtained in the force sum calculator;
Based on the positional relationship between the image cell and the basic cell, which is a processing unit in the force sum calculation unit, the coordinate component of the sum of the forces selected by the selection circuit is output as it is, or the sign is inverted. An addition / subtraction control circuit for outputting or setting the output to 0;
A holding circuit that holds the addition result for each component of the virial; and
An addition for adding the output of the addition / subtraction control circuit and the addition result of the component corresponding to the output of the addition / subtraction control circuit among the addition results held in the holding circuit and storing the addition result in the holding circuit And
For each virial component, using the particle coordinates of the particles in the basic cell and the sum of the forces acting on the particles in the basic cell held in the force sum calculation unit, A first process for obtaining a product sum with the particle coordinates, a second process for multiplying the addition result of the component held in the holding circuit by a constant, and a product sum obtained by the first process. A multi-body problem calculation apparatus comprising: a software execution unit that executes software for performing a third process for obtaining a difference between a result and a multiplication result obtained in the second process. The multi-body problem calculation apparatus according to claim 1,
A plurality of force summation calculators operating in parallel
The multi-body problem calculation apparatus, wherein the selection circuit sequentially selects each coordinate component of the total force calculated by the plurality of force total calculation units. In the multibody problem calculation apparatus according to claim 2,
The plurality of force sum calculation units select different particles as calculation target particles, and select the same image cell as an image cell as a processing unit. In the multibody problem calculation apparatus according to claim 1 or 3,
A lattice vector holding unit in which a lattice vector indicating an image cell which is the processing unit of the force summation calculation unit is stored;
The addition / subtraction control circuit determines a positional relationship between an image cell and a basic cell as a processing unit in the force sum calculation unit based on a lattice vector held in the lattice vector holding unit. Multi-body problem calculator. In the many-body problem calculation device according to claim 4,
The addition / subtraction control circuit, when each coordinate component of the lattice vector held in the lattice vector holding unit is the same as the coordinate component of the sum of forces selected by the selection circuit is positive, A multi-body problem calculation apparatus characterized in that the coordinate component of the sum of the forces selected by the selection circuit is output as it is, the sign is inverted when it is negative, and the output is 0 when it is 0. In the multibody problem calculation device according to any one of claims 1 to 5,
A constant used in the second process is a half-body value of a side length of the basic cell. A two-body interaction calculation circuit that is a component of a multi-body problem calculation device,
One of the particles in the basic cell in the system to which the periodic boundary condition is applied is selected as a calculation target particle, and the image cells of the periodic boundary condition are sequentially set as processing units, and the particles in the image cell are the calculation target. The process of obtaining the sum of the forces acting on the particles is repeated until all the particles in the basic cell are selected as calculation target particles, and the sum of the forces acting on the particles for each particle in the basic cell is calculated. A total force calculation unit to find and hold;
A selection circuit that sequentially selects each coordinate component of the sum of forces each time the sum of forces in an image cell is obtained in the force sum calculator;
Based on the positional relationship between the image cell and the basic cell, which is a processing unit in the force sum calculation unit, the coordinate component of the sum of the forces selected by the selection circuit is output as it is, or the sign is inverted. An addition / subtraction control circuit for outputting or setting the output to 0;
A holding circuit that holds the addition result for each component of the virial; and
An addition for adding the output of the addition / subtraction control circuit and the addition result of the component corresponding to the output of the addition / subtraction control circuit among the addition results held in the holding circuit and storing the addition result in the holding circuit And a two-body interaction calculation circuit. The two-body interaction calculation circuit according to claim 7,
A plurality of force summation calculators operating in parallel
The two-body interaction calculation circuit, wherein the selection circuit sequentially selects coordinate components of the total force calculated by the plurality of force total calculation units. The force summation calculation unit selects one of the particles in the basic cell in the system to which the periodic boundary condition is applied as a calculation target particle, and sequentially sets the image cell of the periodic boundary condition as a processing unit. The process of obtaining the sum of the forces exerted by the particles on the calculation target particles is repeated until all the particles in the basic cell are selected as calculation target particles, and for each particle in the basic cell, the particle To find and hold the sum of the forces
Each time the selection circuit calculates the total force of the image cell unit in the force total calculation unit, it sequentially selects each coordinate component of the total force,
The addition / subtraction control circuit outputs, as it is, the coordinate component of the sum of the forces selected by the selection circuit based on the positional relationship between the image cells and the basic cells that are processing units in the force sum calculation unit, Invert the sign to output, or set the output to 0,
An adder adds the output of the addition / subtraction control circuit and the addition result of the component corresponding to the output of the addition / subtraction control circuit among the addition results for each virial component held in the holding circuit, and adds Storing the result in the holding circuit;
For each component of the virial component, the software execution unit uses the particle coordinates of the particles in the basic cell and the sum of the forces acting on the particles in the basic cell held in the force sum calculation unit. A first process for obtaining the sum of products of the total force and the particle coordinates, a second process for multiplying the addition result for the component held in the holding circuit and a constant, and the first process A multi-body problem calculation method, comprising: executing software for performing a third process for obtaining a difference between the product-sum result obtained in step 2 and the multiplication result obtained in the second process. In the many-body problem calculation method according to claim 9,
Multiple force summation calculators work in parallel,
The multi-body problem calculation method, wherein the selection circuit sequentially selects coordinate components of the total force calculated by the plurality of total force calculation units.

Images