Presentation is loading. Please wait.

Presentation is loading. Please wait.

Problems with M things. Review: NULL MPI handles All MPI_*_NULL are handles to invalid MPI objects MPI_REQUEST_NULL is an exception Calling MPI_TEST/MPI_WAIT.

Published byLeonard Slye Modified over 10 years ago

Similar presentations

Presentation on theme: "Problems with M things. Review: NULL MPI handles All MPI_*_NULL are handles to invalid MPI objects MPI_REQUEST_NULL is an exception Calling MPI_TEST/MPI_WAIT."— Presentation transcript:

1 Problems with M things

2 Review: NULL MPI handles All MPI_*_NULL are handles to invalid MPI objects MPI_REQUEST_NULL is an exception Calling MPI_TEST/MPI_WAIT with REQUEST_NULL results in the empty status status.MPI_SOURCE = MPI_ANY_SOURCE status.MPI_TAG = MPI_ANY_TAG status count = zero …otherwise the ANY/SOME array-based TEST/WAIT functions would be kind of a disaster

3 Review: PROC_NULL behavior Call MPI_RECV with MPI_PROC_NULL Get PROC_NULL status Same as empty status, but with MPI_SOURCE=MPI_PROC_NULL Call MPI_IRECV with MPI_PROC_NULL Get PROC_NULL status (…as expected)

4 MPI_MESSAGE_NULL: MRECV Defining PROC_NULL behavior with MPROBE MPI-3 allows calling MPI_[I]MRECV with MPI_MESSAGE_NULL You get PROC_NULL status back MPI_MPROBE(MPI_PROC_NULL, …, &msg, …) Get msg=MPI_MESSAGE_NULL MPI_MRECV(…, &msg, &status) Get status=PROC_NULL status

5 MPI_MESSAGE_NULL: MRECV Defining PROC_NULL behavior with MPROBE MPI-3 allows calling MPI_[I]MRECV with MPI_MESSAGE_NULL You get PROC_NULL status back MPI_MPROBE(MPI_PROC_NULL, …, &msg, …) Get msg=MPI_MESSAGE_NULL MPI_MRECV(…, &msg, &status) Get status=PROC_NULL status Were calling MPI_MRECV with an invalid message!

6 MPI_MESSAGE_NULL: IMRECV Defining PROC_NULL behavior with MPROBE MPI-3 allows calling MPI_[I]MRECV with MPI_MESSAGE_NULL You get PROC_NULL status back MPI_MPROBE(MPI_PROC_NULL, …, &msg, …) Get msg=MPI_MESSAGE_NULL MPI_IMRECV(…, &msg, &req) Get status=PROC_NULL status MPI_WAIT(&req, &status)

7 MPI_MESSAGE_NULL: IMRECV Defining PROC_NULL behavior with MPROBE MPI-3 allows calling MPI_[I]MRECV with MPI_MESSAGE_NULL You get PROC_NULL status back MPI_MPROBE(MPI_PROC_NULL, …, &msg, …) Get msg=MPI_MESSAGE_NULL MPI_IMRECV(…, &msg, &req) Get status=PROC_NULL status MPI_WAIT(&req, &status) Were calling MPI_IMRECV with an invalid message!

8 The Problem Overloading MPI_MESSAGE_NULL: Invalid message, and MPI_PROC_NULL message FAIL

9 Solution: MPI_MESSAGE_NO_PROC When [I]MPROBE with PROC_NULL Get msg = MPI_MESSAGE_NO_PROC MPI_MPROBE(MPI_PROC_NULL, …, &msg, …) Get msg=MPI_MESSAGE_NO_PROC MPI_MRECV(…, &msg, &status) Get status=PROC_NULL status Set msg=MPI_MESSAGE_NULL

10 Rationale Consistency: MPI_*_NULL are invalid handles Can avoid calling MPI_MRECV (or MPI_IMRECV) if desired Same solution applies to both MRECV and IMRECV Alternate names: MPI_MESSAGE_PROC_NULL (BAD) MPI_MPROBE(MPI_PROC_NULL, …, &msg, …) Get msg=MPI_MESSAGE_NO_PROC MPI_MRECV(…, &msg, &status) Get status=PROC_NULL status Set msg=MPI_MESSAGE_NULL

Download ppt "Problems with M things. Review: NULL MPI handles All MPI_*_NULL are handles to invalid MPI objects MPI_REQUEST_NULL is an exception Calling MPI_TEST/MPI_WAIT."

Similar presentations

MPI-3 Persistent WG (Point-to-point Persistent Channels) February 8, 2011 Tony Skjellum, Coordinator MPI Forum.

MPI-3 Persistent WG (Point-to-point Persistent Channels) February 8, 2011 Tony Skjellum, Coordinator MPI Forum.
1 Tuning for MPI Protocols l Aggressive Eager l Rendezvous with sender push l Rendezvous with receiver pull l Rendezvous blocking (push or pull)

1 Tuning for MPI Protocols l Aggressive Eager l Rendezvous with sender push l Rendezvous with receiver pull l Rendezvous blocking (push or pull)
MPI Message Passing Interface

MPI Message Passing Interface
1 Computer Science, University of Warwick Accessing Irregularly Distributed Arrays Process 0’s data arrayProcess 1’s data arrayProcess 2’s data array Process.

1 Computer Science, University of Warwick Accessing Irregularly Distributed Arrays Process 0’s data arrayProcess 1’s data arrayProcess 2’s data array Process.
COMMUNICATING SEQUENTIAL PROCESSES C. A. R. Hoare The Queen’s University Belfast, North Ireland.

COMMUNICATING SEQUENTIAL PROCESSES C. A. R. Hoare The Queen’s University Belfast, North Ireland.
1 Non-Blocking Communications. 2 #include int main(int argc, char **argv) { int my_rank, ncpus; int left_neighbor, right_neighbor; int data_received=-1;

1 Non-Blocking Communications. 2 #include int main(int argc, char **argv) { int my_rank, ncpus; int left_neighbor, right_neighbor; int data_received=-1;
Pipeline Computations Increasing Throughput By Using More Processes.

Pipeline Computations Increasing Throughput By Using More Processes.
Asynchronous I/O with MPI Anthony Danalis. Basic Non-Blocking API  MPI_Isend()  MPI_Irecv()  MPI_Wait()  MPI_Waitall()  MPI_Waitany()  MPI_Test()

Asynchronous I/O with MPI Anthony Danalis. Basic Non-Blocking API  MPI_Isend()  MPI_Irecv()  MPI_Wait()  MPI_Waitall()  MPI_Waitany()  MPI_Test()
Parallel Processing1 Parallel Processing (CS 667) Lecture 9: Advanced Point to Point Communication Jeremy R. Johnson *Parts of this lecture was derived.

Parallel Processing1 Parallel Processing (CS 667) Lecture 9: Advanced Point to Point Communication Jeremy R. Johnson *Parts of this lecture was derived.
The Linux Kernel: Memory Management

The Linux Kernel: Memory Management
1 Implementing Master/Slave Algorithms l Many algorithms have one or more master processes that send tasks and receive results from slave processes l Because.

1 Implementing Master/Slave Algorithms l Many algorithms have one or more master processes that send tasks and receive results from slave processes l Because.
1 Buffers l When you send data, where does it go? One possibility is: Process 0Process 1 User data Local buffer the network User data Local buffer.

1 Buffers l When you send data, where does it go? One possibility is: Process 0Process 1 User data Local buffer the network User data Local buffer.
Reference Types. 2 Objectives Introduce reference types –class –array Discuss details of use –declaration –allocation –assignment –null –parameter –aggregation.

Reference Types. 2 Objectives Introduce reference types –class –array Discuss details of use –declaration –allocation –assignment –null –parameter –aggregation.
Презентація за розділом “Гумористичні твори”

Презентація за розділом “Гумористичні твори”
Центр атестації педагогічних працівників 2014

Центр атестації педагогічних працівників 2014
Галактики і квазари.

Галактики і квазари.
Характеристика ІНДІЇ.

Характеристика ІНДІЇ.
Процюк Н.В. вчитель початкових класів Боярської ЗОШ І – ІІІ ст №4

Процюк Н.В. вчитель початкових класів Боярської ЗОШ І – ІІІ ст №4
What a disaster! How bad can it really be?. This is not a disaster!

What a disaster! How bad can it really be?. This is not a disaster!
1 What is message passing? l Data transfer plus synchronization l Requires cooperation of sender and receiver l Cooperation not always apparent in code.

1 What is message passing? l Data transfer plus synchronization l Requires cooperation of sender and receiver l Cooperation not always apparent in code.

Similar presentations

Ads by Google