Information processing program, information processing method, and information processing apparatus
By obtaining the candidate setting values calculated by the theoretical calculation of density function in parallel processing and estimating costs using model formulas, the problem of determining the number of nodes is solved, and cost optimization and efficiency improvement are achieved.
Patent Information
- Application Number
- JP2023072807
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-04-26
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to effectively reduce the cost of parallel processing for material density function theoretical calculations, especially to determine the number of nodes required for parallel processing.
By obtaining candidate setting values calculated by the theory of parallel processing density function, the cost is estimated using the model formula, and the optimal setting value is determined based on the sum of costs to optimize the resource configuration of parallel processing.
It effectively reduces the cost of theoretical calculation of density function in parallel processing, improves calculation efficiency and resource utilization.
Smart Images

Figure 2024158022000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an information processing program, an information processing method, and an information processing device. [Background technology]
[0002] Traditionally, in the fields of materials engineering, materials science, and materials development, density functional theory calculations for materials are sometimes performed in parallel to analyze the ground state energy of a material or the change in energy due to the movement of atoms within a material.
[0003] The prior art, for example, involves following a set of lubricant performance requirements and using atomistic modeling tools to design a lubricant formulation that substantially meets a set of performance requirements. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Special Publication No. 2008-523472 Summary of the Invention [Problem to be solved by the invention]
[0005] However, in the conventional technology, it is difficult to reduce the cost of parallel processing of density functional theory calculations for a material. For example, it is not possible to determine how many nodes are required for parallel processing of density functional theory calculations for a material.
[0006] In one aspect, the present invention aims to reduce the cost involved in parallel processing of density functional theory calculations for a substance. [Means for solving the problem]
[0007] According to one embodiment, an information processing program, an information processing method, and an information processing device are proposed that obtain one or more candidates for setting values for an execution environment for parallel processing of density functional theory calculations on a material, calculate a sum of costs corresponding to each of one or more calculation processes for the density functional theory calculations on the material based on a model formula that corresponds to each of the one or more calculation processes for the density functional theory calculations on the material and outputs an estimated value of the cost required to perform the calculation process in response to an input of the setting value, and determine the setting value based on the calculated sum. Effect of the Invention
[0008] According to one embodiment, it is possible to reduce the cost involved in parallel processing of density functional theory calculations for a substance. [Brief description of the drawings]
[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of an information processing method according to an embodiment. [Diagram 2] FIG. 2 is an explanatory diagram illustrating an example of an information processing system 200. As shown in FIG. [Diagram 3] FIG. 3 is a block diagram showing an example of a hardware configuration of the information processing device 100. As shown in FIG. [Figure 4] FIG. 4 is a block diagram showing an example of a functional configuration of the information processing device 100. As shown in FIG. [Diagram 5] FIG. 5 is an explanatory diagram (part 1) showing an example of the operation of the information processing device 100. [Figure 6] FIG. 6 is an explanatory diagram (part 2) showing an example of the operation of the information processing device 100. [Figure 7] FIG. 7 is an explanatory diagram (part 3) showing an example of the operation of the information processing device 100. [Figure 8] FIG. 8 is an explanatory diagram (part 4) showing an example of the operation of the information processing device 100. [Figure 9] FIG. 9 is an explanatory diagram (part 5) showing an example of the operation of the information processing device 100. [Figure 10]FIG. 10 is an explanatory diagram (part 1) showing a specific example of the operation of the information processing device 100. [Figure 11] FIG. 11 is an explanatory diagram (part 2) showing a specific example of the operation of the information processing device 100. [Figure 12] FIG. 12 is an explanatory diagram showing an embodiment of the information processing device 100. As shown in FIG. [Figure 13] FIG. 13 is a flowchart (part 1) illustrating an example of an overall processing procedure. [Figure 14] FIG. 14 is a flowchart (part 2) illustrating an example of the overall processing procedure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0010] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment of an information processing program, an information processing method, and an information processing device according to the present invention will be described in detail with reference to the drawings.
[0011] (An example of an information processing method according to an embodiment) 1 is an explanatory diagram showing an example of an information processing method according to an embodiment. The information processing device 100 is a computer for facilitating parallel processing of density functional theory calculations for a substance. The information processing device 100 is, for example, a server or a PC (Personal Computer).
[0012] 2. Description of the Related Art In the fields of materials engineering, materials science, and materials development, it is desirable to verify the physical properties of a substance. The substance is, for example, a material. For example, in order to design, research, or develop a practical material, it is desirable to verify the physical properties of the material. The material is, for example, a catalyst.
[0013] Here, for example, it is conceivable to verify the physical properties of the material by manufacturing an actual material and carrying out a predetermined experiment on the actual material, but this has the problem of increasing the human, time, or monetary costs involved in verifying the physical properties of the material, etc. Also, for example, it is difficult to accurately manufacture an actual material in a desired state in order to verify the physical properties of a material in a different state.
[0014] For this reason, there is a tendency to desire to carry out a simulation to verify the physical properties of a substance. Density functional theory calculations for a substance are used in the simulation. For example, when carrying out a simulation, density functional theory calculations for a substance may be carried out in order to analyze the ground state energy of the substance or changes in energy according to the movement of atoms within the substance. Specifically, it is conceivable to carry out a simulation to verify the rate of a chemical reaction by calculating the energy used in a chemical reaction on the surface of the substance by carrying out density functional theory calculations for the substance.
[0015] Here, the larger the scale of a substance, the longer the processing time required to perform density functional theory calculations on the substance. The scale of a substance may include, for example, 10,000 atoms. For example, the amount of calculations required to perform density functional theory calculations on a substance with respect to the number of atoms, N, is O(N^3). Furthermore, when performing a simulation to verify the physical properties of a plurality of substances, density functional theory calculations are performed on each substance. In this way, density functional theory calculations on a substance may become a bottleneck when verifying the physical properties of a material.
[0016] Therefore, it is sometimes desirable to perform parallel processing of density functional theory calculations for a material. However, in the past, there was a problem that it was difficult to reduce the time, power, or monetary costs involved in performing parallel processing of density functional theory calculations for a material.
[0017] For example, there is a problem in that it is not possible to determine how many nodes are preferable for parallel processing of density functional theory calculations on a substance. Specifically, there is a problem in that it is not possible to determine how many nodes are preferable for parallel processing of density functional theory calculations on a substance, how many processes are preferable for parallel processing, or how many threads are preferable for parallel processing of density functional theory calculations on a substance.
[0018] For example, it is conceivable that an operator should determine how many nodes are preferable for parallel processing of density functional theory calculations for a material, but this leads to a problem of increasing the workload of the operator. In addition, for example, if an operator is not familiar with density functional theory calculations for materials, it is difficult for the operator to appropriately determine how many nodes are preferable for parallel processing of density functional theory calculations for a material.
[0019] When density functional theory calculations for a material are parallel-processed without determining how many nodes are preferable for parallel processing of the density functional theory calculations for the material, there is a problem that the time, power, or monetary costs increase.
[0020] Therefore, in this embodiment, an information processing method capable of reducing the cost required for parallel processing of density functional theory calculations for a substance will be described. In the following description, density functional theory may be referred to as "DFT (Density Functional Theory)". In the following description, DFT may be Hybrid DFT.
[0021] 1, an information processing device 100 stores model formulas 110 corresponding to each of one or more calculation steps related to a DFT calculation for a substance. The DFT calculation for a substance is defined by, for example, a combination of predetermined attribute values. The predetermined attribute values are, for example, set in advance by a user.
[0022] The attribute values include, for example, the type of atoms forming the substance, the number of atoms forming the substance, the positions of the atoms forming the substance, the type of density functional for the substance, the type of basis function used for the density functional for the substance, or the termination conditions for DFT calculations for the substance.
[0023] A model formula 110 corresponding to any of the calculation processes has a function of outputting an estimate of the cost required to execute any of the calculation processes in response to an input of a setting value 111 related to an execution environment for parallel processing of DFT calculations for a substance. The cost is, for example, a time cost, a power cost, or a monetary cost. The setting value 111 is, for example, the number of processes for parallel processing of DFT calculations for a substance, or the number of threads in a process. The setting value 111 is, for example, a combination of the number of processes and the number of threads in a process.
[0024] (1-1) The information processing device 100 acquires one or more candidates for the setting value 111 related to the execution environment for parallel processing of DFT calculations for a substance. The candidates are, for example, an example of a combination of the number of processes and the number of threads in the process. The information processing device 100 acquires one or more candidates for the setting value 111 by accepting an input of one or more candidates for the setting value 111 based on an operational input by a user.
[0025] (1-2) The information processing device 100 calculates a sum 112 of costs corresponding to each of the one or more acquired candidates based on the model formula 110. For example, for each candidate, the information processing device 100 acquires an estimate of the cost output by the model formula 110 corresponding to each calculation process. For example, the information processing device 100 calculates a sum 112 by adding up the acquired costs for each candidate. This enables the information processing device 100 to evaluate how favorable each candidate is in terms of the cost required for parallel processing of DFT calculations for a substance.
[0026] (1-3) The information processing device 100 determines the setting value 111 based on the calculated sum 112. For example, the information processing device 100 determines, as the setting value 111, one of the one or more candidates whose calculated sum 112 is the smallest. For example, the information processing device 100 may determine, as the setting value 111, one of the one or more candidates whose calculated sum 112 is equal to or less than a threshold. For example, the information processing device 100 may determine, as the setting value 111, a statistical value of a candidate whose calculated sum 112 is equal to or less than a threshold, as the setting value 111, among the one or more candidates.
[0027] This allows the information processing device 100 to appropriately set the setting value 111 in terms of the cost required for parallel processing of DFT calculations on a substance. The information processing device 100 can reduce the cost required for parallel processing of DFT calculations on a substance. This allows the information processing device 100 to easily perform a simulation for verifying the physical properties of a substance.
[0028] Here, the case where the predetermined attribute value is set by the user in advance has been described, but the present invention is not limited to this. For example, the information processing device 100 may receive an input of a combination of the predetermined attribute values. In this case, the information processing device 100 may generate a model formula 110 corresponding to each calculation step of one or more calculation steps related to the DFT calculation for a substance, which are defined by the combination of the predetermined attribute values that have been received as input.
[0029] Here, the case where the functions of the information processing device 100 are realized by a single computer has been described, but this is not limiting. For example, the functions of the information processing device 100 may be realized by a plurality of computers working together. Specifically, the functions of the information processing device 100 may be realized on a cloud.
[0030] (An example of the information processing system 200) Next, an example of an information processing system 200 to which the information processing device 100 shown in FIG. 1 is applied will be described with reference to FIG.
[0031] 2 is an explanatory diagram showing an example of an information processing system 200. In FIG. 2, the information processing system 200 includes an information processing device 100, one or more parallel processing devices 201, and one or more client devices 202.
[0032] In the information processing system 200, the information processing device 100 and the parallel processing device 201 are connected via a wired or wireless network 210. The network 210 is, for example, a local area network (LAN), a wide area network (WAN), the Internet, etc. In the information processing system 200, the information processing device 100 and the client device 202 are connected via the wired or wireless network 210.
[0033] The information processing device 100 is a computer for facilitating parallel processing of DFT calculations on a substance. The information processing device 100 receives a processing request for parallel processing of DFT calculations on a substance from a client device 202. The processing request includes, for example, a combination of predetermined attribute values. The processing request includes, for example, one or more candidates for setting values related to the execution environment of the DFT calculations on the substance.
[0034] The information processing device 100 acquires a combination of predetermined attribute values from the processing request. The information processing device 100 generates a model formula corresponding to each of one or more calculation steps related to the DFT calculation for the substance, which are defined by the acquired combination of predetermined attribute values. A specific example of generating the model formula will be described later with reference to, for example, Figs. 5 to 12.
[0035] The information processing device 100 acquires one or more candidates for setting values related to the execution environment of the DFT calculation for the substance from the processing request. The information processing device 100 calculates the sum of costs corresponding to each of the acquired one or more candidates based on the model formula, as in Fig. 1. The information processing device 100 determines the setting values based on the calculated sum, as in Fig. 1.
[0036] The information processing device 100 controls one or more parallel processing devices 201 to perform parallel processing of the DFT calculation on the substance based on the determined setting values. The information processing device 100 obtains a result of the parallel processing of the DFT calculation on the substance as a result of controlling the one or more parallel processing devices 201. The information processing device 100 transmits the result of the parallel processing of the DFT calculation on the substance to the client device 202. The information processing device 100 is, for example, a server or a PC.
[0037] The parallel processing device 201 is a computer for performing parallel processing of DFT calculations on a substance. The parallel processing device 201 shares the DFT calculations on a substance under the control of the information processing device 100. The parallel processing device 201 is, for example, a server or a PC.
[0038] The client device 202 is a computer used by an operator who is to verify the physical properties of a substance. The client device 202 accepts input of a combination of predetermined attribute values based on the operator's operation input. The client device 202 accepts input of one or more candidates for setting values related to the execution environment of a DFT calculation for a substance based on the operator's operation input.
[0039] The client device 202 generates a processing request including the combination of the inputted predetermined attribute values and one or more candidates of setting values related to the execution environment of the DFT calculation for the substance for which the inputted input was received. The client device 202 transmits the generated processing request to the information processing device 100.
[0040] The client device 202 receives the result of the parallel processing of the DFT calculation for the substance from the information processing device 100. The client device 202 outputs the result of the parallel processing of the DFT calculation for the substance so that the operator can refer to it. The client device 202 is, for example, a PC, a tablet terminal, or a smartphone.
[0041] Here, the case where the information processing device 100 is a computer different from the parallel processing device 201 has been described, but this is not limiting. For example, the information processing device 100 may have a function as the parallel processing device 201 and may also operate as the parallel processing device 201.
[0042] Here, the case where the information processing device 100 is a computer different from the client device 202 has been described, but this is not limited thereto. For example, the information processing device 100 may have a function as the client device 202 and may also operate as the client device 202.
[0043] (Example of hardware configuration of information processing device 100) Next, an example of the hardware configuration of the information processing device 100 will be described with reference to FIG.
[0044] Fig. 3 is a block diagram showing an example of a hardware configuration of the information processing device 100. In Fig. 3, the information processing device 100 has a CPU (Central Processing Unit) 301, a memory 302, a network I / F (Interface) 303, a recording medium I / F 304, and a recording medium 305. In addition, each component is connected to each other via a bus 300.
[0045] Here, the CPU 301 is responsible for the overall control of the information processing device 100. The memory 302 has, for example, a read only memory (ROM), a random access memory (RAM), and a flash ROM. Specifically, for example, the flash ROM and the ROM store various programs, and the RAM is used as a work area for the CPU 301. The programs stored in the memory 302 are loaded into the CPU 301, causing the CPU 301 to execute the coded processes.
[0046] The network I / F 303 is connected to the network 210 through a communication line, and is connected to other computers via the network 210. The network I / F 303 manages an internal interface with the network 210, and controls input and output of data from other computers. The network I / F 303 is, for example, a modem or a LAN adapter.
[0047] The recording medium I / F 304 controls reading / writing data from / to the recording medium 305 under the control of the CPU 301. The recording medium I / F 304 is, for example, a disk drive, a solid state drive (SSD), a universal serial bus (USB) port, etc. The recording medium 305 is a non-volatile memory that stores data written under the control of the recording medium I / F 304. The recording medium 305 is, for example, a disk, a semiconductor memory, a USB memory, etc. The recording medium 305 may be detachable from the information processing device 100.
[0048] In addition to the above-mentioned components, the information processing device 100 may have, for example, a keyboard, a mouse, a display, a printer, a scanner, a microphone, a speaker, etc. Furthermore, the information processing device 100 may have a plurality of recording medium I / Fs 304 and recording media 305. Furthermore, the information processing device 100 may not have the recording medium I / Fs 304 and the recording media 305.
[0049] (Example of hardware configuration of parallel processing device 201) Specifically, an example of the hardware configuration of the parallel processing device 201 is similar to the example of the hardware configuration of the information processing device 100 shown in FIG. 3, and therefore a description thereof will be omitted.
[0050] (Example of Hardware Configuration of Client Device 202) Specifically, an example of the hardware configuration of the client device 202 is similar to the example of the hardware configuration of the information processing device 100 shown in FIG. 3, and therefore a description thereof will be omitted.
[0051] (Example of functional configuration of information processing device 100) Next, an example of a functional configuration of the information processing device 100 will be described with reference to FIG.
[0052] 4 is a block diagram showing an example of a functional configuration of the information processing device 100. The information processing device 100 includes a storage unit 400, an acquisition unit 401, a measurement unit 402, a setting unit 403, a calculation unit 404, a determination unit 405, an implementation unit 406, and an output unit 407.
[0053] The storage unit 400 is realized by, for example, a storage area such as the memory 302 or the recording medium 305 shown in Fig. 3. In the following, a case where the storage unit 400 is included in the information processing device 100 will be described, but this is not limiting. For example, the storage unit 400 may be included in a device different from the information processing device 100, and the stored contents of the storage unit 400 may be accessible from the information processing device 100.
[0054] The acquiring unit 401 to the output unit 407 function as an example of a control unit. Specifically, the acquiring unit 401 to the output unit 407 realize their functions by, for example, causing the CPU 301 to execute a program stored in a storage area such as the memory 302 or the recording medium 305 shown in Fig. 3, or by the network I / F 303. The processing results of each functional unit are stored in, for example, a storage area such as the memory 302 or the recording medium 305 shown in Fig. 3.
[0055] The storage unit 400 stores various information that is referred to or updated in the processing of each functional unit. The storage unit 400 stores, for example, one or more attribute values that define a DFT calculation for a substance. The attribute value indicates, for example, the type of atoms that form the substance, the number of atoms that form the substance, or the positions of atoms that form the substance. The attribute value indicates, for example, the type of density functional for the substance, or the type of basis function used for the density functional for the substance. The attribute value indicates, for example, a termination condition for a DFT calculation for a substance. The attribute value is set, for example, by a user in advance. The attribute value is acquired, for example, by an acquisition unit 401.
[0056] The storage unit 400 stores, for example, one or more samples of setting values related to an execution environment for parallel processing of DFT calculations on a substance. The setting values are, for example, information that specifies the number of processes that share the DFT calculations on a substance and the number of threads that the process has. The setting values are, for example, a combination of the number of processes and the number of threads in the process. The samples are, for example, set in advance by a user. The samples are, for example, acquired by the acquisition unit 401.
[0057] The storage unit 400 stores, for example, one or more candidates for setting values related to an execution environment for parallel processing of DFT calculations for a substance. The candidates are set in advance by, for example, a user. The candidates are acquired by, for example, the acquisition unit 401.
[0058] The storage unit 400 stores, for example, a model formula corresponding to each of one or more calculation steps related to DFT calculation for a substance. The model formula corresponding to any of the calculation steps has a function of outputting an estimated value of the cost required to perform any of the calculation steps in response to an input of a setting value related to an execution environment for parallel processing of DFT calculation for a substance. The cost is, for example, a time cost, an electric power cost, or a monetary cost. The model formula is acquired by, for example, the acquisition unit 401. The model formula is generated by, for example, the setting unit 403.
[0059] The acquisition unit 401 acquires various information used for processing by each functional unit. The acquisition unit 401 stores the acquired various information in the storage unit 400 or outputs it to each functional unit. The acquisition unit 401 may also output the various information stored in the storage unit 400 to each functional unit. The acquisition unit 401 acquires various information based on, for example, an operation input by a user. The acquisition unit 401 may receive various information from, for example, a device different from the information processing device 100.
[0060] The acquiring unit 401 acquires, for example, a processing request requesting parallel processing of DFT calculations for a substance. Specifically, the acquiring unit 401 acquires the processing request by accepting an input of the processing request based on an operation input by a user. Specifically, the acquiring unit 401 acquires the processing request by receiving it from another computer. The other computer is, for example, the client device 202.
[0061] The processing request may include, for example, one or more attribute values that define a DFT calculation for a substance. The processing request may include, for example, one or more samples of setting values related to an execution environment for parallel processing of a DFT calculation for a substance. The processing request may include, for example, one or more candidates of setting values related to an execution environment for parallel processing of a DFT calculation for a substance. The processing request may include, for example, a model formula corresponding to each of one or more calculation steps related to a DFT calculation for a substance.
[0062] The acquiring unit 401 acquires, for example, one or more attribute values that define a DFT calculation for a substance. Specifically, the acquiring unit 401 acquires one or more attribute values by accepting input of one or more attribute values based on an operation input by a user. Specifically, the acquiring unit 401 acquires one or more attribute values by receiving them from another computer. The other computer is, for example, the client device 202. Specifically, the acquiring unit 401 acquires one or more attribute values by extracting them from a processing request.
[0063] The acquiring unit 401 acquires, for example, one or more samples of setting values related to an execution environment for parallel processing of DFT calculations for a substance. Specifically, the acquiring unit 401 acquires the one or more samples by accepting input of the one or more samples based on an operation input by a user. Specifically, the acquiring unit 401 acquires the one or more samples by receiving them from another computer. The other computer is, for example, the client device 202. Specifically, the acquiring unit 401 acquires the one or more samples by extracting them from a processing request.
[0064] The acquiring unit 401 acquires, for example, one or more candidates for setting values related to an execution environment for parallel processing of DFT calculations on a substance. Specifically, the acquiring unit 401 acquires the one or more candidates by accepting an input of the one or more candidates based on an operation input by a user. Specifically, the acquiring unit 401 acquires the one or more candidates by receiving them from another computer. The other computer is, for example, the client device 202. Specifically, the acquiring unit 401 acquires the one or more candidates by extracting them from a processing request.
[0065] The acquiring unit 401 acquires, for example, a model formula corresponding to each of one or more calculation steps related to a DFT calculation for a substance. Specifically, the acquiring unit 401 acquires the model formula by accepting an input of the model formula based on an operation input by a user. Specifically, the acquiring unit 401 acquires the model formula by receiving it from another computer. The other computer is, for example, the client device 202. Specifically, the acquiring unit 401 acquires the model formula by extracting it from a processing request.
[0066] The acquisition unit 401 may receive a start trigger for starting processing of any of the functional units. The start trigger may be, for example, a predetermined operation input by a user. The start trigger may be, for example, reception of predetermined information from another computer. The start trigger may be, for example, output of predetermined information by any of the functional units.
[0067] Specifically, the acquiring unit 401 accepts the acquisition of one or more attribute values defining the DFT calculation and one or more samples of setting values related to the execution environment for parallel processing of the DFT calculation as a start trigger for starting the processing of the measuring unit 402 and the setting unit 403. Specifically, the acquiring unit 401 accepts the acquisition of one or more candidates of setting values related to the execution environment for parallel processing of the DFT calculation for a substance as a start trigger for starting the processing of the calculating unit 404, the determining unit 405, and the implementing unit 406.
[0068] The measurement unit 402 measures an actual value of a cost required for performing each calculation step related to a predetermined DFT calculation for a substance for each of the one or more samples acquired by the acquisition unit 401. The predetermined DFT calculation for a substance is set in advance by, for example, a user. The measurement unit 402 submits, for example, a job related to a predetermined DFT calculation for a substance to an execution environment indicated by each sample. The measurement unit 402 measures an actual value of a cost required for performing each calculation step related to a predetermined DFT calculation for a substance based on, for example, a result of executing the submitted job in the execution environment indicated by each sample. This allows the measurement unit 402 to obtain a guideline for generating a model formula in the setting unit 403.
[0069] The measurement unit 402 measures an actual value of a cost required for performing each calculation step related to a DFT calculation for a substance for each of the one or more samples acquired by the acquisition unit 401 based on one or more attribute values acquired by the acquisition unit 401. The measurement unit 402 sets a DFT calculation for a substance based on, for example, one or more attribute values acquired by the acquisition unit 401. The measurement unit 402 submits a job related to the set DFT calculation for, for example, an execution environment indicated by each sample. The measurement unit 402 measures an actual value of a cost required for performing each calculation step related to the set DFT calculation based on, for example, a result of executing the submitted job in the execution environment indicated by each sample. This allows the measurement unit 402 to obtain a guideline for generating a model formula in the setting unit 403.
[0070] The setting unit 403 generates a model formula corresponding to each calculation process related to the DFT calculation for the substance based on the measured actual values. For example, the setting unit 403 generates a model formula by setting parameters of the model formula corresponding to each calculation process related to the DFT calculation for the substance by linear regression using the least squares method based on the measured actual values. This allows the setting unit 403 to obtain a guideline for determining the setting value in the determination unit 405.
[0071] The calculation unit 404 calculates the sum of costs corresponding to each of the one or more candidates acquired by the acquisition unit 401 based on a model formula corresponding to each calculation step of a predetermined DFT calculation for a substance. The predetermined DFT calculation for a substance is set in advance by a user, for example. The calculation unit 404, for example, inputs the candidate into a model formula corresponding to each calculation step for each of the one or more candidates acquired by the acquisition unit 401, thereby acquiring a cost output by the model formula. The calculation unit 404, for example, adds up the acquired costs for each candidate to calculate the sum of costs corresponding to the candidate. This allows the calculation unit 404 to obtain a guideline for determining a setting value in the determination unit 405.
[0072] The calculation unit 404 calculates the sum of costs corresponding to each of the one or more acquired candidates based on the one or more acquired attribute values and the model formula corresponding to each calculation process related to the DFT calculation for the substance. The calculation unit 404 sets the DFT calculation for the substance based on, for example, one or more attribute values acquired by the acquisition unit 401. For example, the calculation unit 404 inputs the candidate into the model formula corresponding to each calculation process related to the set DFT calculation for each of the one or more candidates acquired by the acquisition unit 401, thereby acquiring the cost output by the model formula. For example, the calculation unit 404 calculates the sum of costs corresponding to the candidate by adding up the acquired costs for each candidate. As a result, the calculation unit 404 can obtain a guideline for determining a setting value in the determination unit 405.
[0073] The determining unit 405 determines the set value based on the calculated sum. For example, the determining unit 405 determines, as the set value, one of the candidates having the smallest calculated sum among the one or more candidates acquired by the acquiring unit 401. For example, the determining unit 405 determines, as the set value, one of the candidates having the calculated sum equal to or less than a threshold among the one or more candidates acquired by the acquiring unit 401. For example, the determining unit 405 determines, as the set value, the statistical value of the candidate having the calculated sum equal to or less than a threshold among the one or more candidates acquired by the acquiring unit 401. This allows the determining unit 405 to appropriately set the set value in terms of the cost required for parallel processing of DFT calculations on the substance. Therefore, the determining unit 405 can easily reduce the cost required for parallel processing of DFT calculations on the substance.
[0074] For example, if the cost is an index value that indicates a more favorable state as the value increases, the determination unit 405 may determine, as the set value, any candidate having the largest calculated sum among the one or more candidates acquired by the acquisition unit 401. For example, if the cost is an index value that indicates a more favorable state as the value increases, the determination unit 405 may determine, as the set value, any candidate having a calculated sum equal to or greater than a threshold among the one or more candidates acquired by the acquisition unit 401.
[0075] The implementation unit 406 controls one or more processes or one or more threads in a process to perform parallel processing of DFT calculations on a substance based on the determined setting values. The process may be included in, for example, the information processing device 100. The process may be included in, for example, another computer. The other computer is, for example, the parallel processing device 201.
[0076] The implementation unit 406 identifies a combination of the number A of processes and the number B of threads in the process based on, for example, the determined setting value. The implementation unit 406 prepares, for example, a specified number A of processes including the specified number B of threads on one or more parallel processing devices 201, thereby preparing an execution environment for parallel processing of DFT calculations on a substance. The implementation unit 406 controls, for example, the prepared execution environment so that the DFT calculations on the substance are parallel processed in the prepared execution environment. The implementation unit 406 acquires a result of parallel processing of the DFT calculations on the substance from the prepared execution environment. In this way, the implementation unit 406 can reduce the cost incurred when parallel processing the DFT calculations on the substance.
[0077] The output unit 407 outputs the processing result of at least one of the functional units. The output format is, for example, display on a display, printout on a printer, transmission to an external device via the network I / F 303, or storage in a storage area such as the memory 302 or the recording medium 305. In this way, the output unit 407 can notify the user of the processing result of at least one of the functional units, thereby improving the convenience of the information processing device 100.
[0078] The output unit 407 outputs, for example, the setting value determined by the determination unit 405. Specifically, the output unit 407 outputs the setting value determined by the determination unit 405 so that the user can refer to it. Specifically, the output unit 407 may transmit the setting value determined by the determination unit 405 to another computer. Specifically, the other computer is the client device 202 or the like. In this way, the output unit 407 can reduce the cost required for parallel processing of DFT calculations on a substance externally.
[0079] The output unit 407 outputs, for example, a result of parallel processing of the DFT calculation for the substance acquired by the execution unit 406. Specifically, the output unit 407 outputs the result of parallel processing of the DFT calculation for the substance acquired by the execution unit 406 so that the user can refer to it. Specifically, the output unit 407 may transmit the result of parallel processing of the DFT calculation for the substance acquired by the execution unit 406 to another computer. Specifically, the other computer is the client device 202 or the like. In this way, the output unit 407 can make the result of parallel processing of the DFT calculation for the substance available externally.
[0080] Here, the case where the information processing device 100 includes the acquisition unit 401, the measurement unit 402, the setting unit 403, the calculation unit 404, the determination unit 405, the implementation unit 406, and the output unit 407 has been described, but is not limited to this. For example, the information processing device 100 may not include any of the functional units. Specifically, the information processing device 100 may not include the implementation unit 406. In this case, the information processing device 100 may transmit the setting value determined by the determination unit 405 to another computer including the implementation unit 406.
[0081] (An example of the operation of the information processing device 100) Next, an example of the operation of the information processing device 100 will be described with reference to FIGS.
[0082] 5 to 9 are explanatory diagrams showing an example of the operation of the information processing device 100. For example, in Fig. 5 to 9, the information processing device 100 searches for machine settings 501 that minimize an execution cost 504. The machine settings 501 are information that affects the processing time of a DFT calculation.
[0083] The machine settings 501 are, for example, a combination of the number of parallel processing devices 201, the number of processes, or the number of threads used when performing parallel processing of the DFT calculation. The parallel processing is, for example, MPI or OpenMP. The parallel processing may be, for example, a combination of MPI and OpenMP. The machine settings 501 may include, for example, the CPU clock frequency of the parallel processing devices 201 or the number of accelerators used when performing parallel processing of the DFT calculation.
[0084] The execution cost 504 is, for example, the processing time of the DFT calculation. The execution cost 504 may be, for example, a node time product or a node time. The node is, for example, a computer that shares the DFT calculation. Specifically, the node is the parallel processing device 201.
[0085] For example, when a DFT calculation is performed using one or more parallel processing devices 201 that are supercomputers, the usage fee for the supercomputer tends to increase as the processing time of the DFT calculation increases. For this reason, it is desirable to search for a machine setting 501 that minimizes the execution cost 504.
[0086] First, specifically, in FIG. 5, (5-1) the information processing device 100 receives input of a plurality of samples of the machine settings 501. For example, the information processing device 100 receives N combinations of the number p of processes and the number t of threads as the plurality of samples of the machine settings 501 (p i ,t i ), where i is an integer number of 1, . . . , N. The lower limit of N depends on, for example, the number of values of the parameters 512 of one or more sub-model equations 511 forming the model 510.
[0087] The information processing device 100 accepts an input of a DFT setting 502. The DFT setting 502 includes, for example, information on atoms. The atomic information includes, for example, the number of atoms. The atomic information includes, for example, the type of atom, basis function, potential, three-dimensional position, and the like for each atom. The DFT setting 502 includes, for example, information on lattice. The lattice information includes, for example, lattice size, iteration boundary condition, or symmetry, and the like. The DFT setting 502 includes, for example, information on functionals. The functional information includes, for example, the type of functional, or parameters of the functional, and the like. The DFT setting 502 includes, for example, information on the setting of an SCF loop. The setting information includes, for example, a truncation condition, a maximum iteration condition, a type of minimizer, and a type of preprocessing, and the like.
[0088] The information processing device 100 calculates the processing time required to perform each calculation step of the DFT calculation for each of the input samples of the multiple samples, based on the input received DFT settings 502. The information processing device 100 generates the sub-model formula 511 by setting parameters 512 of the sub-model formula 511 corresponding to each calculation step, based on a combination 503 of each sample and the processing time calculated for that sample.
[0089] The sub-model formula 511 has a function of calculating the cost required to perform the calculation process. The information processing device 100 generates the sub-model formula 511 to generate a model 510. The model 510 has a function of calculating the total sum of the costs represented by the sub-model formula 511. A specific example in which the information processing device 100 sets the parameters 512 will be described later with reference to FIG. 7. Next, moving to the description of FIG. 6, an example of the sub-model formula 511 will be described.
[0090] 6, a table 600 shows an example of the sub-model formula 511. The table 600 shows a sub-model s and a sub-model formula T (s) (p,t;x (s) ) and the parameter x (s) It has fields with and.
[0091] In the field of the sub-model s, the name of the calculation process corresponding to the sub-model s is set as a name for identifying the sub-model s. (s) (p,t;x (s) ) field is a submodel formula 511 representing the above submodel s. (s) (p,t;x (s) ) formula is set. p is the number of processes. t is the number of threads. x (s) The field contains the parameter 512, x (s) is set.
[0092] In the table 600, for convenience, the parameters 512 of each sub-model formula 511 are (s) Although the same symbols a, b, c, and d are used, x (s) The symbols a, b, c, and d indicate different values.
[0093] Here, the two-electron integral is, for example, a calculation process for calculating the Hartree-Fock exchange term of the hybrid DFT. The two-electron integral, for example, has a process-parallel calculation cost of O(p -1 ) or thread-parallel computation cost O(t -1 ) is included. The computation cost is, for example, the computation time. Specifically, when a computation with a fixed computation amount is divided into parts without dependencies and parallelized by processes or threads, the computation time is O(p -1 ) or O(t -1 ).
[0094] The potential integral is, for example, a calculation process for calculating the integral of a potential. The potential integral has a process-parallel calculation cost of O(p -1 ) and the communication cost O(p) of exchanging the calculation results between processes by all-to-all communication. The communication cost is, for example, the communication time. Specifically, when exchanging data whose total size is fixed between processes, a one-dimensional torus is formed between the processes, and each size is O(p -1) in p steps, the communication cost is O(p).
[0095] Sparse matrix multiplication is a computational process that calculates the matrix multiplication of block sparse matrices. For example, the time required to share sparse matrix information between processes by all-reduce communication is O(logp), and the computational cost of the Cannon algorithm is O(t -1 / 2 p 1 / 2 ) and communication cost O(p 1 / 2 Specifically, when data of a fixed size is shared between processes, if a binary tree is constructed between the processes and messages of a fixed size are communicated in logp steps, the communication cost is O(logp).
[0096] Dense matrix multiplication is a computational process that calculates the matrix multiplication of block dense matrices. For example, the computational cost of the Cannon algorithm is O(t -1 / 2 p 1 / 2 ) and communication cost O(p 1 / 2 ) is included. Specifically, when multiplying fixed-size block square matrices, the blocks are evenly distributed and held among two-dimensional processes. Here, the number of blocks in the row and column directions is O(p 1 / 2 ) The number of rows and columns in each block is O(p -1 / 2 ). O(p 1 / 2 If there is a fixed communication cost for each step, then the total communication cost is O(p 1 / 2 ). O(p 1 / 2 ) steps, thread-parallel O(t -1 / 2 ), the total computational cost is O(t -1 / 2 p 1 / 2 ).
[0097] The eigenvalue calculation is, for example, a calculation process for calculating the eigenvalues of a symmetric dense matrix. It is assumed that the eigenvalue calculation includes a communication cost of O(logp), assuming that the communication cost of broadcast communication from the root process to other processes becomes dominant when the number of processes is equal to or greater than a certain value. The FFT is, for example, a calculation process for performing a 3D FFT or an inverse transformation of the 3D FFT to calculate a plane wave. The FFT, for example, has a process-parallel calculation cost of O(p -1 ) and the communication cost O(p) of exchanging the calculation results between processes via all-to-all communication.
[0098] Others, for example, have computational costs of O(logp), O(p -1 ), O(t -1 ) and so on. The two-electron integral corresponds specifically to integrate_four_center in the CP2K code. The potential integral corresponds specifically to integrate_v_rspace in the CP2K code. The sparse matrix multiplication corresponds specifically to dbcsr_multiply_generic in the CP2K code. The dense matrix multiplication corresponds specifically to cp_gemm in the CP2K code. The eigenvalue calculation corresponds specifically to cp_fm_syevd in the CP2K code. The FFT corresponds specifically to fft_wrap_pw1pw2 in the CP2K code.
[0099] The information processing device 100, for example, selects T that becomes one of the sub-model formulas 511 shown in the table 600 based on an operational input by a user. (s) (p,t;x (s) ) may be changed or deleted. For example, the information processing device 100 may change or delete T (s) (p,t;x (s) ) parameter 512 x (s) may be modified or deleted.
[0100] The information processing apparatus 100 may newly store, for example, a new sub-model formula 511 other than the sub-model formula 511 shown in Table 600 based on the operation input of the user. When not considering Hybrid DFT, the information processing apparatus 100 may not store the sub-model formula 511 of the two-electron integral.
[0101] Next, moving on to the description of FIG. 7, the information processing apparatus 100, based on a plurality of input samples, becomes the sub-model formula 511 corresponding to each calculation process, T (s) (p,t;x (s) ) becomes the parameter 512 of x (s) A specific example of setting will be described.
[0102] The information processing apparatus 100 sets the DFT calculation based on the received DFT setting 502. The information processing apparatus 100 performs the set DFT calculation for each (p i ,t i ) of {(p i=1 N . Here, the information processing apparatus 100 may perform only a part of the set DFT calculation instead of performing the entire set DFT calculation. Specifically, if the DFT calculation is to repeat a predetermined operation X times, the information processing apparatus 100 may repeat the predetermined operation only Y (<X) times.
[0103] The information processing apparatus 100 measures the processing time {T^ i (s)}>0 required when performing the calculation process corresponding to the sub-model s based on the result of the set DFT calculation. S is a set of sub-models s. The measurement can be realized, for example, by an existing profiler, an existing performance counter, or an instruction statement manually inserted into the code of the DFT calculation software.
[0104] For example, when only a part of the set DFT calculation is performed, the information processing device 100 calculates the processing time {T^ i (s)} s∈S Specifically, the information processing device 100 may estimate X / Y times the processing time corresponding to the calculation process corresponding to the sub-model s when a predetermined calculation is repeated Y times as the processing time {T^ i (s)} s∈S It is estimated as follows.
[0105] The information processing device 100 is i ,t i ) and the (p i ,t i ) was measured for {T^ i (s)} s∈S and {(p i ,t i ,T^ i (s) )} i=1 N The information processing device 100 stores the stored {(p i ,t i ,T^ i (s) )} i=1 N Based on this, the parameters x corresponding to the sub-model s are calculated by linear regression. (s) At this time, the information processing device 100 sets the parameter x (s) If any of the values are negative, they may be reset to zero.
[0106] Here, the information processing device 100 calculates the parameter x (s) A specific example of setting the following will be described. Specifically, the information processing device 100 sets three (n i ,p i ,t i) (i=1,2,3) and three (n i ,p i ,t i )(i=4,5,6) and {T^ i (二電子積分)} i=1 6 The {T^ i (二電子積分)} i=1 6 is shown in Table 700.
[0107] n i is the number of nodes. i are 1,2,4. Flat MPI parallelism is a format that includes one MPI process per core. Flat MPI parallelism is a format that includes, for example, 10 processes per node. Hybrid parallelism is a format that includes one MPI process per node and 10 threads per process.
[0108] The information processing device 100 calculates the parameter x by the least squares method based on the table 700. (二電子積分) =(a (二電子積分) ,b (二電子積分) ,c (二電子積分) )=(5.04, 6.50, 1.05). In this way, the information processing device 100 sets an arbitrary (n i ,p i ,t i ), T (二電子積分) (p,t;x (二電子積分) For example, in the case of MPI parallelism, the information processing device 100 can calculate T (二電子積分) (80,80;x (二電子積分) )=1.194 [s]. Next, we will return to the explanation of FIG.
[0109] 5, (5-2) the information processing device 100 accepts input of multiple candidates for machine setting 501. For example, the information processing device 100 accepts input of a set M of multiple combinations (p, t) of the number p of processes and the number t of threads as multiple candidates for the machine setting 501. Specifically, the information processing device 100 may accept input of the set M including all configurable combinations (p, t) as a population of targets for searching for the machine setting 501 with the smallest execution cost.
[0110] Based on the generated model 510, the information processing device 100 generates T corresponding to each combination (p, t) in M for which input is received. (s) (p,t;x (s) ) the sum of Σ s∈S T (s) The information processing device 100 calculates T (s) (p,t;x (s) ), T corresponding to each combination (p,t) in M (s) (p,t;x (s) ) the sum of Σ s∈S T (s) The information processing device 100 calculates the calculated sum Σ s∈S T (s) Based on this, the execution cost 504 is calculated.
[0111] For example, if the execution cost 504 is a processing time, the information processing device 100 calculates the execution cost 504=Σ s∈S T (s) For example, if the execution cost 504 is a node time product, the information processing device 100 calculates the execution cost 504=(number of nodes)×Σ s∈S T (s) For example, if the execution cost 504 is a usage fee, the information processing device 100 calculates the execution cost 504=(usage fee per unit node time)×(number of nodes)×Σ s∈S T (s) Calculate.
[0112] The information processing device 100 determines, based on the calculated execution cost 504, any combination (p, t) in M as the machine setting 501. For example, the information processing device 100 determines, as the machine setting 501, the combination (p, t) that results in the smallest calculated execution cost 504.
[0113] Here, a specific example in which the information processing device 100 determines the machine setting 501 will be described. Here, it is assumed that the DFT calculation includes a calculation process of a two-electron integral and other calculation processes. (その他) = (0.198, 0.401, 0.502, 0, 0569). Let M = {(in, 10n) | n = 10, 11, ..., 40, i = 1, 2, 5, 10}. M represents the case where the number of nodes is n, the number of processes in the node is i, and the number of threads in the process is 10 / i. Specifically, the information processing device 100 calculates Σ s∈S T (s) =T (二電子積分) +T (その他) Let (p,t)∈M be the smallest such that T (s) = 1.984, and (p, t) = (31, 310) is identified and determined as machine setting 501.
[0114] This allows the information processing device 100 to determine the machine settings 501 that allow efficient parallel processing of the DFT calculation. Since the information processing device 100 uses the sub-model formula 511 of the polynomial, it is possible to efficiently determine the machine settings 501. Next, moving to the description of FIG. 8, Σ s∈S T (s) An example of the distribution of will be described.
[0115] In FIG. 8, a graph 800 is s∈S T (s) Graph 800 shows the distribution of, for example, Σ s∈S T (s) Specifically, the graph 800 shows the contour of Σ in the above-mentioned case where the DFT calculation includes the calculation process of the two-electron integral and other calculation processes. s∈S T (s) The distribution of is shown.
[0116] The horizontal axis of graph 800 indicates, for example, the number of nodes. The vertical axis of graph 800 indicates the number of processes per node. The stars on graph 800 indicate the Σ s∈S T (s) is the smallest point. As shown in graph 800, Σ s∈S T (s) is not proportional to the number of nodes or the number of processes.
[0117] In response to this, the information processing device 100 s∈S T (s) Taking these factors into consideration, an appropriate machine setting 501 can be determined. Next, referring to Fig. 9, specific definitions of each element shown in Fig. 5 will be summarized. Each element is, for example, a machine setting 501, a DFT setting 502, a combination 503, an execution cost 504, a model 510, a sub-model formula 511, and parameters 512.
[0118] 9 shows a specific definition of each element shown in FIG. 5. As shown in FIG. 9, a machine setting 501 is, for example, p, t. A DFT setting 502 is, for example, {T^ i (s)} and information used to set up the DFT calculation.
[0119] The combination 503 is, for example, {(p i ,t i )} i=1 N and {{T^ i (s)} s∈S} i=1 N The execution cost 504 is, for example, a combination of Σ s∈S T (s) or (number of nodes) × Σ s∈S T (s) And so on.
[0120] The sub-model formula 511 is, for example, T (s) (p,t;x (s) ) The sub-model formula 511 is, for example, Σ s∈S T(s) = (T (二電子積分) + … + T (その他) ) is used when calculating. Parameter 512 is, for example, {x (s)} s∈S = {(a (二電子積分) , b (二電子積分) , c (二電子積分) ), …, (a (その他) , b (その他) , c (その他) , d (その他) )}. Model 510 is, for example, Σ s∈S T (s) = (T (二電子積分) + … + T (その他) ).
[0121] In this way, the information processing apparatus 100 can determine an appropriate machine setting 501 for a certain DFT setting 502. The information processing apparatus 100 can reduce the execution cost involved in parallel processing of DFT calculations.
[0122] When the information processing apparatus 100 changes the DFT setting 502, after regenerating the submodel formula 511, it will determine an appropriate machine setting 501 again.
[0123] At this time, even if the information processing apparatus 100 interrupts without completing the DFT calculation set based on the changed DFT setting 502, it can regenerate the submodel formula 511. For example, if the DFT calculation set based on the changed DFT setting 502 is to repeat a predetermined operation X times, the information processing apparatus 100 may repeat the predetermined operation only Y (<X) times and regenerate the submodel formula 511. Therefore, even if the information processing apparatus 100 changes the DFT setting 502, it can make it easier to determine an appropriate machine setting 501 again.
[0124] (Specific example of the operation of the information processing apparatus 100) Next, a specific example of the operation of the information processing apparatus 100 will be described with reference to FIGS. 10 and 11.
[0125] 10 and 11 are explanatory diagrams showing a specific example of the operation of information processing device 100. In Fig. 10, there are an implementation program 1010 and a job scheduler 1020. The implementation program 1010 is executed by, for example, information processing device 100. The job scheduler 1020 is executed by, for example, information processing device 100.
[0126] (10-1) A user 1001 executes an implementation program 1010. The user 1001 issues an execution request for the implementation program 1010 including a DFT setting 1011, one or more samples 1012 of machine settings, and a space 1013 of machine settings, via, for example, a client device 202. The client device 202 transmits an execution request including the DFT setting 1011, one or more samples 1012 of machine settings, and a space 1013 of machine settings to the information processing device 100, based on, for example, an operation input by the user 1001.
[0127] The DFT settings 1011 are information that defines the DFT calculation. Each of one or more samples 1012 of machine settings represents an example of a machine setting used in determining parameters 1016 of sub-model equations 1015 that form the model 1014. The samples 1012 are i ,t i )} i=1 N The space of machine settings 1013 represents a universe for searching for machine settings that result in the lowest execution cost.
[0128] (10-2) The implementation program 1010 is started in response to an execution request. The implementation program 1010 acquires a DFT setting 1011, one or more samples 1012 of the machine setting, and a space 1013 of the machine setting from the execution request. The implementation program 1010 sets the content of the DFT calculation based on the DFT setting 1011. The implementation program 1010 generates a group of computation jobs 1030 corresponding to each of the one or more samples 1012 of the machine setting based on the set content of the DFT calculation. The computation job 1030 may include, for example, a sample 1012 corresponding to the computation job 1030 itself. The implementation program 1010 submits a group of computation jobs 1030 corresponding to the sample 1012 to the job scheduler 1020 in association with each sample 1012.
[0129] (10-3) The job scheduler 1020 has a queue 1021. The job scheduler 1020 temporarily stores submitted computation jobs 1030 in the queue 1021. The job scheduler 1020 sequentially takes out the computation jobs 1030 from the queue 1021 and executes them. For example, the job scheduler 1020 assigns the computation job 1030 to any of the parallel processing devices 201 cooperating with the job scheduler 1020 and executes the computation job 1030.
[0130] (10-4) The calculation job 1030 includes a DFT calculation software 1031. The calculation job 1030 performs a DFT calculation by the DFT calculation software 1031. The calculation job 1030 transmits to the implementation program 1010 an actual measurement value 1017 of the processing time required to perform each calculation step of the DFT calculation.
[0131] (10-5) The implementation program 1010 receives the actual measurement values 1017 of the processing times required for performing each calculation step of the DFT calculation from the calculation job 1030. The implementation program 1010 receives the combination of the actual measurement values 1017 of the processing times {{T^ i (s)} s∈S} i=1 NBased on the parameter 1016, {x (s)} s∈S The implementation program 1010 determines the sub-model equation 1015 to generate the model 1014.
[0132] (10-6) The implementation program 1010 determines, from the space 1013 of machine settings, a machine setting (p, t) that minimizes the total cost represented by the sub-model formula 1015. The implementation program 1010 outputs the determined machine setting (p, t) so that the user 1001 can refer to it. In this way, the implementation program 1010 can make an appropriate machine setting (p, t) so that the user 1001 can refer to it. Next, we move on to the explanation of FIG. 11 and explain how the implementation program 1010, the job scheduler 1020, and the computation job 1030 correspond to the hardware.
[0133] 11 , the implementation program 1010 is stored in, for example, a file system 1140. The file system 1140 is specifically realized by the information processing device 100. The implementation program 1010 is executed by, for example, a login node 1110. The login node 1110 is specifically realized by the information processing device 100. The login node 1110 may be specifically realized by the information processing device 100 and the client device 202.
[0134] The job scheduler 1020 is executed by, for example, a management node 1120. The management node 1120 is specifically realized by the information processing device 100 or the parallel processing device 201. The computation job 1030 is executed by the management node 1120 and a computation node 1131. The computation node 1131 is realized by, for example, the parallel processing device 201. The login node 1110, the management node 1120, the computation node 1131, and the file system 1140 may be realized by, for example, one supercomputer.
[0135] Here, data exchange between the execution program 1010 and the computing node 1131 may be realized via a job scheduler 1020 or a shared file system. Furthermore, the computing jobs 1030 corresponding to each sample 1012 may be executed simultaneously. Furthermore, the computing node 1131 may include an existing profiler or an existing performance counter.
[0136] The implementation program 1010 may control the job scheduler 1020 and the computing nodes 1131 to process the DFT calculations in parallel based on the determined machine settings (p, t). The implementation program 1010, for example, generates a group of computing jobs 1030 for processing the DFT calculations in parallel based on the determined machine settings (p, t) and submits them to the job scheduler 1020. This allows the implementation program 1010 to reduce the execution cost incurred when processing the DFT calculations in parallel.
[0137] (Example of information processing device 100) Next, an embodiment of the information processing device 100 will be described with reference to FIG.
[0138] Fig. 12 is an explanatory diagram showing an embodiment of the information processing device 100. In Fig. 12, the information processing device 100 searches for appropriate machine settings for performing CP2K on the ABCI supercomputer. The DFT settings, for example, specify a hybrid DFT calculation of 52 atoms of Sr-Fe-O using the PBE functional.
[0139] The multiple samples of machine configurations include, for example, five samples where the number of nodes is 4, 8, 16, 32, or 64, the number of processes per node is 2, and the number of threads per node is 40. The multiple samples of machine configurations include, for example, three samples where the number of nodes is 16, the number of processes per node is 1, 4, or 10, and the number of threads per node is 40. The actual processing time is measured using the time measurement function of CP2K.
[0140] In FIG. 12, a graph 1200 is s∈S T (s) The graph 1200 shows the distribution of, for example, Σ s∈S T (s) 12 shows the contours of the number of processes per node. The horizontal axis of the graph 1200 shows, for example, the number of nodes. The vertical axis of the graph 1200 shows the number of processes per node.
[0141] As shown in Graph 1200, Σ s∈S T (s) is not proportional to the number of nodes or the number of processes. For example, when the number of nodes is less than 20, the execution cost tends to be smaller as the number of processes per node increases, but when the number of nodes is 20 or more, the execution cost tends to be larger as the number of processes per node increases. Specifically, as the number of nodes increases, the calculation cost O(p 1 / 2 ) are likely to have an impact on execution costs.
[0142] Here, when the number of nodes is 100, the number of threads per node is 40, and the number of processes is 1, 2, 4, 5, 8, 10, 20, and 40, the information processing device 100 determines the machine setting to (p, t) = (200, 4000). When (p, t) = (200, 4000), the execution cost is 136.9 [s].
[0143] For comparison, when the number of processes per node is 1 (p,t)=(100,4000), the execution cost is 142.9 [s]. For comparison, when the number of processes per node is 40 (p,t)=(4000,4000), the execution cost is 192.7 [s].
[0144] Therefore, the information processing device 100 can determine a machine setting capable of speeding up the DFT calculation by 1.04 times compared to when the number of processes per node is 1. The information processing device 100 can determine a machine setting capable of speeding up the DFT calculation by 1.41 times compared to when the number of processes per node is 40.
[0145] (Overall processing procedure) Next, an example of an overall processing procedure executed by the information processing device 100 will be described with reference to Fig. 13 and Fig. 14. The overall processing is realized by, for example, the CPU 301, storage areas such as the memory 302 and the recording medium 305, and the network I / F 303 shown in Fig. 3.
[0146] 13 and 14 are flowcharts showing an example of an overall processing procedure. In Fig. 13, the information processing device 100 sets i=1 (step S1301). Then, the information processing device 100 proceeds to the processing of step S1302.
[0147] In step S1302, the information processing device 100 executes the machine setting (p i ,t i ), and submits job i to the queue manager 100 (step S1302).
[0148] Then, the information processing device 100 determines whether i>N is satisfied (step S1303). If i>N is not satisfied (step S1303: No), the information processing device 100 increments i and returns to the process of step S1302. On the other hand, if i>N is satisfied (step S1303: Yes), the information processing device 100 proceeds to the process of step S1304.
[0149] In step S1304, the information processing device 100 determines whether all submitted jobs i have been completed (step S1304). If there is an incomplete job i remaining (step S1304: No), the information processing device 100 returns to the process of step S1304. On the other hand, if all submitted jobs i have been completed (step S1304: Yes), the information processing device 100 proceeds to the process of step S1305.
[0150] In step S1305, the information processing device 100 sets i to 1 (step S1305). Then, the information processing device 100 proceeds to the process of step S1306.
[0151] In step S1306, the information processing apparatus 100 calculates {T^ i (s)} s∈S (step S1306).
[0152] Then, the information processing device 100 determines whether i>N is satisfied (step S1307). If i>N is not satisfied (step S1307: No), the information processing device 100 increments i and returns to the process of step S1306. On the other hand, if i>N is satisfied (step S1307: Yes), the information processing device 100 proceeds to the process of step S1308.
[0153] In step S1308, the information processing device 100 selects a sub-model s in S (step S1308). Next, the information processing device 100 calculates {(p i ,t i ,T^ i (s) )} i=1 N , the parameters x of the selected submodel s are calculated by linear regression. (s) (step S1309).
[0154] Then, the information processing device 100 determines whether or not all sub-models s in S have been selected (step S1310). If there are still sub-models s that have not been selected (step S1310: No), the information processing device 100 returns to the process of step S1308. On the other hand, if all sub-models s in S have been selected (step S1310: Yes), the information processing device 100 proceeds to the process of step S1401 in FIG.
[0155] In FIG. 14, the information processing device 100 best = (undefined) (step S1401). Next, the information processing device 100 sets t best = (undefined) (step S1402). Then, the information processing device 100 sets C best =∞ (step S1403). After that, the information processing device 100 proceeds to the process of step S1404.
[0156] In step S1404, the information processing device 100 selects (p, t) in M (step S1404). Next, the information processing device 100 sets C=model (p, t, {x (s)} s∈S )=Σ s∈S T (s) (Step S1405).
[0157] Then, the information processing device 100 <C best It is determined whether or not C <C best If so (step S1406: Yes), the information processing device 100 proceeds to the process of step S1407. <C best If not (step S1406: No), information processing device 100 proceeds to the process of step S1410.
[0158] In step S1407, the information processing device 100 best = p (step S1407). Next, the information processing device 100 sets t best =t (step S1408). Then, the information processing device 100 sets Cbest =C (step S1409). After that, information processing device 100 proceeds to the process of step S1410.
[0159] In step S1410, the information processing device 100 determines whether or not all (p, t) in M have been selected (step S1410). If there are still unselected (p, t) remaining (step S1410: No), the information processing device 100 returns to the process of step S1404. On the other hand, if all (p, t) in M have been selected (step S1410: Yes), the information processing device 100 proceeds to the process of step S1411.
[0160] In step S1411, the information processing device 100 best ,t best ) (step S1411). Then, the information processing device 100 ends the entire process. In this way, the information processing device 100 can find (p, t) within M for which the total cost C is the smallest.
[0161] Here, the information processing device 100 may switch the order of the processes of some steps in the flowcharts of Figures 13 and 14. For example, the order of the processes of steps S1401 and S1402 may be switched. Also, the information processing device 100 may omit the processes of some steps in the flowcharts of Figures 13 and 14. For example, (s) If is known, the processes of steps S1301 to S1310 can be omitted.
[0162] As described above, the information processing device 100 can acquire one or more candidates for the setting value related to the execution environment for parallel processing of density functional theory calculations for a material. The information processing device 100 can store a model formula that corresponds to each of one or more calculation processes related to density functional theory calculations for a material and outputs an estimated value of the cost required to perform the calculation process in response to an input of a setting value. The information processing device 100 can calculate the sum of the costs corresponding to each of the acquired one or more candidates based on the model formula. The information processing device 100 can determine the setting value based on the calculated sum. Thereby, the information processing device 100 can determine an appropriate setting value in order to reduce the cost required to parallel process density functional theory calculations for a material.
[0163] According to the information processing device 100, it is possible to measure an actual value of the cost required to perform each of one or more calculation steps for each of one or more samples of setting values. According to the information processing device 100, it is possible to set parameters of a model formula that corresponds to each of one or more calculation steps and outputs an estimated value of the cost required to perform the calculation step in response to an input of a setting value based on the measured actual value. This allows the information processing device 100 to generate a model formula. The information processing device 100 can do without preparing a model formula in advance.
[0164] According to the information processing device 100, it is possible to obtain one or more attribute values that define density functional theory calculations for a substance. According to the information processing device 100, it is possible to measure an actual value of the cost required to perform each of one or more calculation steps for each of one or more samples of setting values based on the one or more acquired attribute values. This allows the information processing device 100 to determine an appropriate setting value according to the one or more acquired attribute values.
[0165] According to the information processing device 100, it is possible to adopt information that defines the number of processes that share the density functional theory calculation for a substance and the number of threads that each process has as a setting value. This allows the information processing device 100 to determine appropriate setting values that define the number of processes and the number of threads.
[0166] According to the information processing device 100, the type of atoms forming a substance, the number of atoms forming a substance, or the positions of atoms forming a substance can be adopted as the attribute value. According to the information processing device 100, the type of density functional for a substance, the type of basis function used in the density functional for a substance, or the termination condition of density functional theory calculation for a substance can be adopted as the attribute value. This allows the information processing device 100 to appropriately specify density functional theory calculation.
[0167] The information processing method described in this embodiment can be realized by executing a prepared program on a computer such as a PC or a workstation. The information processing program described in this embodiment is recorded on a computer-readable recording medium, and is executed by being read from the recording medium by the computer. The recording medium may be a hard disk, a flexible disk, a CD (Compact Disc)-ROM, an MO (Magneto Optical disc), a DVD (Digital Versatile Disc), or the like. The information processing program described in this embodiment may also be distributed via a network such as the Internet.
[0168] The following supplementary notes are further disclosed regarding the above-described embodiment.
[0169] (Appendix 1) Obtain one or more candidate settings for the execution environment for parallel processing of density functional theory calculations for materials; calculating a sum of costs corresponding to each of the one or more candidates obtained based on a model formula that corresponds to each of the one or more calculation steps related to the density functional theory calculation for the substance and outputs an estimated value of the cost required to perform the calculation step in response to the input of the setting value; determining the set value based on the calculated sum; An information processing program that causes a computer to execute a process.
[0170] (Appendix 2) The process of determining The information processing program according to claim 1, characterized in that, among the one or more candidates, the candidate whose calculated sum is the smallest or the candidate whose calculated sum is equal to or less than a threshold value is determined as the setting value.
[0171] (Appendix 3) For each sample of the one or more samples of the set value, an actual measurement value of the cost required to perform each of the one or more calculation steps is measured; setting parameters of the model formula corresponding to each of the one or more calculation steps based on the measured actual values, and outputting an estimated value of the cost required to perform the calculation step in response to the input of the setting value; 3. The information processing program according to claim 1 or 2, characterized in that the program causes the computer to execute a process.
[0172] (Appendix 4) Obtaining one or more attribute values that define the density functional theory calculation for the substance; The measuring process includes: The information processing program according to claim 3, further comprising: measuring an actual value of a cost incurred in performing each of the one or more calculation steps for each of the one or more samples of the setting value based on the one or more acquired attribute values.
[0173] (Appendix 5) The information processing program according to any one of Appendices 1 to 4, characterized in that the set value is information specifying the number of processes that share the density functional theory calculation for the substance and the number of threads that the process has.
[0174] (Appendix 6) The information processing program according to appendix 4, characterized in that the attribute values include at least any one of the types of atoms forming the substance, the number of atoms forming the substance, the positions of the atoms forming the substance, the type of density functional for the substance, the type of basis function used for the density functional for the substance, and a termination condition for the density functional theory calculation for the substance.
[0175] (Appendix 7) Controlling the execution environment represented by the determined setting value so as to perform parallel processing of the density functional theory calculation for the substance; 5. The information processing program according to any one of claims 1 to 4, which causes the computer to execute a process.
[0176] (Appendix 8) Obtain one or more candidate settings for the execution environment for parallel processing of density functional theory calculations for materials; calculating a sum of costs corresponding to each of the one or more candidates obtained based on a model formula that corresponds to each of the one or more calculation steps related to the density functional theory calculation for the substance and outputs an estimated value of the cost required to perform the calculation step in response to the input of the setting value; determining the set value based on the calculated sum; An information processing method characterized in that the processing is executed by a computer.
[0177] (Appendix 9) Obtain one or more candidate settings for the execution environment for parallel processing of density functional theory calculations for materials; calculating a sum of costs corresponding to each of the one or more candidates obtained based on a model formula that corresponds to each of the one or more calculation steps related to the density functional theory calculation for the substance and outputs an estimated value of the cost required to perform the calculation step in response to the input of the setting value; determining the set value based on the calculated sum; An information processing device comprising a control unit. [Explanation of symbols]
[0178] 100 Information processing device 110 Model 111 Setting value 112 Sum 200 Information Processing Systems 201 Parallel Processing Unit 202 Client device 210 Network 300 Bus 301 CPU 302 Memory 303 Network Interface 304 Recording media I / F 305 Recording media 400 Storage section 401 Acquisition Department 402 Measurement Division 403 Settings 404 Calculation Unit 405 Decision Section 406 Implementation Department 407 Output section 501 Machine Settings 502,1011 DFT Settings 503 Combinations 504 Execution Cost 511,1015 Submodel formula 512,1016 parameters 600,700 tables 800,1200 graphs 1001 users 1010 Implementation Program 1012 specimen 1013 Space 1017 Actual value 1020 Job Scheduler 1021 Queue 1030 computational jobs 1031 DFT Calculation Software 1110 Login node 1120 Management node 1131 Compute Nodes 1140 File System
Claims
1. Obtaining one or more candidate settings for an execution environment for parallel processing of density functional theory calculations for a material; calculating a sum of costs corresponding to each of the one or more candidates obtained based on a model formula that corresponds to each of the one or more calculation steps related to the density functional theory calculation for the substance and outputs an estimated value of a cost required to perform the calculation step in response to the input of the setting value; determining the set value based on the calculated sum; An information processing program that causes a computer to execute a process.
2. The process of determining includes:
2. The information processing program according to claim 1, wherein, from among the one or more candidates, a candidate for which the calculated sum is the smallest or a candidate for which the calculated sum is equal to or smaller than a threshold value is determined as the setting value.
3. measuring an actual value of a cost required to perform each of the one or more calculation steps for each of the one or more samples of the set value; setting parameters of the model formula corresponding to each of the one or more calculation steps based on the measured actual values, and outputting an estimated value of the cost required to perform the calculation step in response to the input of the setting value; 3. The information processing program according to claim 1, which causes the computer to execute a process.
4. obtaining one or more attribute values that define the density functional theory calculation for the material; The measuring process includes: The information processing program according to claim 3, further comprising: measuring an actual value of a cost incurred in performing each of the one or more calculation processes for each of the one or more samples of the setting value based on the one or more acquired attribute values.
5. Obtaining one or more candidate settings for an execution environment for parallel processing of density functional theory calculations for a material; calculating a sum of costs corresponding to each of the one or more candidates obtained based on a model formula that corresponds to each of the one or more calculation steps related to the density functional theory calculation for the substance and outputs an estimated value of a cost required to perform the calculation step in response to the input of the setting value; determining the set value based on the calculated sum; An information processing method characterized in that the processing is executed by a computer.
6. Obtaining one or more candidate settings for an execution environment for parallel processing of density functional theory calculations for a material; calculating a sum of costs corresponding to each of the one or more candidates obtained based on a model formula that corresponds to each of the one or more calculation steps related to the density functional theory calculation for the substance and outputs an estimated value of a cost required to perform the calculation step in response to the input of the setting value; determining the set value based on the calculated sum; An information processing device comprising a control unit.
Citation Information
Patent Citations
Methods and systems for developing lubricants, lubricant additives, and lubricant basestocks utilizing atomistic modeling tools
JP2008523472A
