src

src

folder containing all libraries that are used by main.cpp

---

vi_processor_base.cpp

Base Class for all following Implementations of the desired task. Responsible for parsing the data from the given dataset from .npy files into Eigen vectors.

Given functionality:

• VI_Processor_Base - reads all files .npy that are needed for the value iteration, transforms them to Eigen vectors/SparseMatrix
• void Process() - function that calls value_iteration_impl of respective processor
• std::string GetName() - get name of implementation
• bool SetParameter() - set parameter of implementation
• std::map<std::string, float> GetParameters() - get parameter of implementation
• float iteration_step() - performs the value iteration for given amount of steps (before remerging whole probability matrix) on given subset of the probability matrix
• void debug_message() - prints given debugging messages for respective processor

---

vi_processor_base.h

Header of vi_processor_base.cpp

---

vi_processor_impl_distr_01.cpp

first implementation of a communication scheme. Using MPI_Allgatherv,MPI_Allreduce and MPI_Gatherv. Waits for given amount of periods before re-gathering the updated J, inherits functionality from VI_Processor_Base.

Given functionality:

• std::string GetName() - get name of implementation
• bool SetParameter() - set parameter of implementation
• std::map<std::string, float> GetParameters() - get parameter of implementation
• void value_iteration_impl() - concrete implementation of communication scheme, doing the value iteration

---

vi_processor_impl_distr_01.h

Header of vi_processor_impl_distr_01.cpp

---

vi_processor_impl_distr_02.cpp

second implementation of a communication scheme. Using MPI_Send, MPI_Recv and MPI_Bcast. Waits for given amount of periods before re-gathering the updated J, inherits functionality from VI_Processor_Base. Logic of communication scheme is similar to VI_Processor_Impl_Distr_01

Given functionality:

• std::string GetName() - get name of implementation
• bool SetParameter() - set parameter of implementation
• std::map<std::string, float> GetParameters() - get parameter of implementation
• void value_iteration_impl() - concrete implementation of communication scheme, doing the value iteration

---

vi_processor_impl_distr_02.h

Header of vi_processor_impl_distr_02.cpp

---

vi_processor_impl_distr_03.cpp

third implementation of a communication scheme. Using MPI_Sendrecv and MPI_Gatherv. Updates J every period completely, inherits functionality from VI_Processor_Base. Works synchronely

Given functionality:

• std::string GetName() - get name of implementation
• bool SetParameter() - set parameter of implementation
• std::map<std::string, float> GetParameters() - get parameter of implementation
• void value_iteration_impl() - concrete implementation of communication scheme, doing the value iteration

---

vi_processor_impl_distr_03.h

Header of vi_processor_impl_distr_03.cpp

---

vi_processor_impl_distr_04.cpp

third implementation of a communication scheme. Using MPI_Isend and MPI_Irecv, MPI_Ibcast and MPI_Igatherv. Waits for given amount of periods before re-gathering the updated J, inherits functionality from VI_Processor_Base. Logic of communication scheme is similar to VI_Processor_Impl_Distr_02 but non-blocking

Given functionality:

• std::string GetName() - get name of implementation
• bool SetParameter() - set parameter of implementation
• std::map<std::string, float> GetParameters() - get parameter of implementation
• void value_iteration_impl() - concrete implementation of communication scheme, doing the value iteration

---

vi_processor_impl_distr_04.h

Header of vi_processor_impl_distr_04.cpp

---

vi_processor_impl_distr_05.cpp

third implementation of a communication scheme. Using MPI_Igather, MPI_Bcast and MPI_Gatherv. Waits for given amount of periods before re-gathering the updated J, inherits functionality from VI_Processor_Base. Logic of communication scheme is similar to VI_Processor_Impl_Distr_01 but non-blocking

Given functionality:

• std::string GetName() - get name of implementation
• bool SetParameter() - set parameter of implementation
• std::map<std::string, float> GetParameters() - get parameter of implementation
• void value_iteration_impl() - concrete implementation of communication scheme, doing the value iteration

---

vi_processor_impl_distr_05.h

Header of vi_processor_impl_distr_05.cpp

---

vi_processor_impl_local.cpp

local implementation of value iteration without any communication scheme or parallelization. Implemented to use as reference for performance comparisons. Updates J every period completely, inherits functionality from VI_Processor_Base. Works synchronely

Given functionality:

• std::string GetName() - get name of implementation
• void value_iteration_impl() - concrete implementation of value iteration

---

vi_processor_impl_local.h

Header of vi_processor_impl_local.cpp