SimGrid
3.7
Scalable simulation of distributed systems
|
This section describes the task structure of MSG (m_task_t) and the functions for managing it. See Task Actions to see how to put the tasks in action. More...
Defines | |
#define | MSG_TASK_UNINITIALIZED NULL |
Default value for an uninitialized m_task_t. | |
Typedefs | |
typedef struct m_task * | m_task_t |
Task datatypeA task may then be defined by a computing amount, a message size and some private data. | |
typedef struct m_gpu_task * | m_gpu_task_t |
GPU task datatypeA task may then be defined by a computing amount, a dispatch latency and a collect latency. | |
Functions | |
m_task_t | MSG_parallel_task_create (const char *name, int host_nb, const m_host_t *host_list, double *computation_amount, double *communication_amount, void *data) |
Creates a new m_task_t (a parallel one....). | |
m_task_t | MSG_task_create (const char *name, double compute_duration, double message_size, void *data) |
Creates a new m_task_t. | |
m_gpu_task_t | MSG_gpu_task_create (const char *name, double compute_duration, double dispatch_latency, double collect_latency) |
Creates a new m_gpu_task_t. | |
void * | MSG_task_get_data (m_task_t task) |
Return the user data of a m_task_t. | |
void | MSG_task_set_data (m_task_t task, void *data) |
Sets the user data of a m_task_t. | |
void | MSG_task_set_copy_callback (void(*callback)(m_task_t task, m_process_t sender, m_process_t receiver)) |
Sets a function to be called when a task has just been copied. | |
m_process_t | MSG_task_get_sender (m_task_t task) |
Return the sender of a m_task_t. | |
m_host_t | MSG_task_get_source (m_task_t task) |
Return the source of a m_task_t. | |
const char * | MSG_task_get_name (m_task_t task) |
Return the name of a m_task_t. | |
void | MSG_task_set_name (m_task_t task, const char *name) |
Sets the name of a m_task_t. | |
MSG_error_t | MSG_task_destroy (m_task_t task) |
Destroy a m_task_t. | |
double | MSG_task_get_compute_duration (m_task_t task) |
Returns the computation amount needed to process a task m_task_t. Once a task has been processed, this amount is thus set to 0... | |
void | MSG_task_set_compute_duration (m_task_t task, double computation_amount) |
set the computation amount needed to process a task m_task_t. | |
double | MSG_task_get_remaining_computation (m_task_t task) |
Returns the remaining computation amount of a task m_task_t. | |
double | MSG_task_get_remaining_communication (m_task_t task) |
Returns the total amount received by a task m_task_t. If the communication does not exist it will return 0. So, if the communication has FINISHED or FAILED it returns zero. | |
double | MSG_task_get_data_size (m_task_t task) |
Returns the size of the data attached to a task m_task_t. | |
void | MSG_task_set_priority (m_task_t task, double priority) |
Changes the priority of a computation task. This priority doesn't affect the transfer rate. A priority of 2 will make a task receive two times more cpu power than the other ones. |
This section describes the task structure of MSG (m_task_t) and the functions for managing it. See Task Actions to see how to put the tasks in action.
Since most scheduling algorithms rely on a concept of task that can be either computed locally or transferred on another processor, it seems to be the right level of abstraction for our purposes. A task may then be defined by a computing amount, a message size and some private data.
typedef struct m_gpu_task* m_gpu_task_t |
GPU task datatypeA task may then be defined by a computing amount, a dispatch latency and a collect latency.
m_task_t MSG_parallel_task_create | ( | const char * | name, |
int | host_nb, | ||
const m_host_t * | host_list, | ||
double * | computation_amount, | ||
double * | communication_amount, | ||
void * | data | ||
) |
Creates a new m_task_t (a parallel one....).
A constructor for m_task_t taking six arguments and returning the corresponding object.
name | a name for the object. It is for user-level information and can be NULL. |
host_nb | the number of hosts implied in the parallel task. |
host_list | an array of host_nb m_host_t. |
computation_amount | an array of host_nb doubles. computation_amount[i] is the total number of operations that have to be performed on host_list[i]. |
communication_amount | an array of host_nb* host_nb doubles. |
data | a pointer to any data may want to attach to the new object. It is for user-level information and can be NULL. It can be retrieved with the function MSG_task_get_data. |
m_task_t MSG_task_create | ( | const char * | name, |
double | compute_duration, | ||
double | message_size, | ||
void * | data | ||
) |
Creates a new m_task_t.
A constructor for m_task_t taking four arguments and returning the corresponding object.
name | a name for the object. It is for user-level information and can be NULL. |
compute_duration | a value of the processing amount (in flop) needed to process this new task. If 0, then it cannot be executed with MSG_task_execute(). This value has to be >=0. |
message_size | a value of the amount of data (in bytes) needed to transfer this new task. If 0, then it cannot be transfered with MSG_task_send() and MSG_task_recv(). This value has to be >=0. |
data | a pointer to any data may want to attach to the new object. It is for user-level information and can be NULL. It can be retrieved with the function MSG_task_get_data. |
m_gpu_task_t MSG_gpu_task_create | ( | const char * | name, |
double | compute_duration, | ||
double | dispatch_latency, | ||
double | collect_latency | ||
) |
Creates a new m_gpu_task_t.
A constructor for m_gpu_task_t taking four arguments and returning a pointer to the new created GPU task.
name | a name for the object. It is for user-level information and can be NULL. |
compute_duration | a value of the processing amount (in flop) needed to process this new task. If 0, then it cannot be executed with MSG_gpu_task_execute(). This value has to be >=0. |
dispatch_latency | time in seconds to load this task on the GPU |
collect_latency | time in seconds to transfer result from the GPU back to the CPU (host) when done |
void* MSG_task_get_data | ( | m_task_t | task | ) |
Return the user data of a m_task_t.
This function checks whether task is a valid pointer or not and return the user data associated to task if it is possible.
void MSG_task_set_data | ( | m_task_t | task, |
void * | data | ||
) |
Sets the user data of a m_task_t.
This function allows to associate a new pointer to the user data associated of task.
void MSG_task_set_copy_callback | ( | void(*)(m_task_t task, m_process_t sender, m_process_t receiver) | callback | ) |
Sets a function to be called when a task has just been copied.
callback | a callback function |
m_process_t MSG_task_get_sender | ( | m_task_t | task | ) |
Return the sender of a m_task_t.
This functions returns the m_process_t which sent this task
m_host_t MSG_task_get_source | ( | m_task_t | task | ) |
const char* MSG_task_get_name | ( | m_task_t | task | ) |
void MSG_task_set_name | ( | m_task_t | task, |
const char * | name | ||
) |
Sets the name of a m_task_t.
This functions allows to associate a name to a task
MSG_error_t MSG_task_destroy | ( | m_task_t | task | ) |
Destroy a m_task_t.
Destructor for m_task_t. Note that you should free user data, if any, before calling this function.
Only the process that owns the task can destroy it. The owner changes after a successful send. If a task is successfully sent, the receiver becomes the owner and is supposed to destroy it. The sender should not use it anymore. If the task failed to be sent, the sender remains the owner of the task.
Back to the main Simgrid Documentation page |
The version of SimGrid documented here is v3.7. Documentation of other versions can be found in their respective archive files (directory doc/html). |
Generated by ![]() |