Ufdjrw

Professor being mistaken for a grad student

Official degrees of earth’s rotation per day

Are all passive ability checks floors for active ability checks?

Why does a Star of David appear at a rally with Francisco Franco?

Instead of a Universal Basic Income program, why not implement a "Universal Basic Needs" program?

How to terminate ping <dest> &

How to write cleanly even if my character uses expletive language?

How are passwords stolen from companies if they only store hashes?

combinatorics floor summation

What is the significance behind "40 days" that often appears in the Bible?

Recruiter wants very extensive technical details about all of my previous work

What is the Japanese sound word for the clinking of money?

What is a ^ b and (a & b) << 1?

What is the adequate fee for a reveal operation?

Describing a chess game in a novel

What is "focus distance lower/upper" and how is it different from depth of field?

Is a party consisting of only a bard, a cleric, and a warlock functional long-term?

Why one should not leave fingerprints on bulbs and plugs?

Are Roman Catholic priests ever addressed as pastor

Why did it take so long to abandon sail after steamships were demonstrated?

How to make healing in an exploration game interesting

How could an airship be repaired midflight?

Is there a symmetric-key algorithm which we can use for creating a signature?

What do you call the act of removing a part of a word and replacing it with an apostrophe

MPI gather data into different indices for each rank

0

Pretend I have 5 x 3 array that I need to fill with data from other ranks. Each rank has a different number of rows (n x 3) that they need to send, but ideally I would like to specify the row that the data should occupy in the final array. For example (sorry for the Pythony syntax):

Rank 0
Data = [[A1,A2,A3],
 [A4,A5,A6]]
Indices = [1, 4]

Rank 1
Data = [[B1,B2,B3]]
Indices = [2]

Rank 2
Data = [[C1,C2,C3],
 [C4,C5,C6]]
Indices = [0,3]


Post Gather Data
Rank 0 = [[C1,C2,C3],
 [A1,A2,A3],
 [B1,B2,B3],
 [C4,C5,C6],
 [A4,A5,A6]]

The mapping of which rank is responsible for which indices is available on all ranks, so I can definitely do a gatherv() and map it back into the final array, but I was hoping I could fill it up correctly as part of the communication.

When doing IO, I do a very similar operation by having each rank define their own indexed data type and then calling MPI_File_write_all(), but I'm having trouble finding an equivalent to fill an buffer. Sorry if I'm just missing something obvious.

mpi

share|improve this question

asked Mar 6 at 20:40

rpmcnallyrpmcnally

1098

add a comment | 

0

Pretend I have 5 x 3 array that I need to fill with data from other ranks. Each rank has a different number of rows (n x 3) that they need to send, but ideally I would like to specify the row that the data should occupy in the final array. For example (sorry for the Pythony syntax):

Rank 0
Data = [[A1,A2,A3],
 [A4,A5,A6]]
Indices = [1, 4]

Rank 1
Data = [[B1,B2,B3]]
Indices = [2]

Rank 2
Data = [[C1,C2,C3],
 [C4,C5,C6]]
Indices = [0,3]


Post Gather Data
Rank 0 = [[C1,C2,C3],
 [A1,A2,A3],
 [B1,B2,B3],
 [C4,C5,C6],
 [A4,A5,A6]]

The mapping of which rank is responsible for which indices is available on all ranks, so I can definitely do a gatherv() and map it back into the final array, but I was hoping I could fill it up correctly as part of the communication.

When doing IO, I do a very similar operation by having each rank define their own indexed data type and then calling MPI_File_write_all(), but I'm having trouble finding an equivalent to fill an buffer. Sorry if I'm just missing something obvious.

mpi

share|improve this question

asked Mar 6 at 20:40

rpmcnallyrpmcnally

1098

add a comment | 

Pretend I have 5 x 3 array that I need to fill with data from other ranks. Each rank has a different number of rows (n x 3) that they need to send, but ideally I would like to specify the row that the data should occupy in the final array. For example (sorry for the Pythony syntax):

Rank 0
Data = [[A1,A2,A3],
 [A4,A5,A6]]
Indices = [1, 4]

Rank 1
Data = [[B1,B2,B3]]
Indices = [2]

Rank 2
Data = [[C1,C2,C3],
 [C4,C5,C6]]
Indices = [0,3]


Post Gather Data
Rank 0 = [[C1,C2,C3],
 [A1,A2,A3],
 [B1,B2,B3],
 [C4,C5,C6],
 [A4,A5,A6]]

The mapping of which rank is responsible for which indices is available on all ranks, so I can definitely do a gatherv() and map it back into the final array, but I was hoping I could fill it up correctly as part of the communication.

When doing IO, I do a very similar operation by having each rank define their own indexed data type and then calling MPI_File_write_all(), but I'm having trouble finding an equivalent to fill an buffer. Sorry if I'm just missing something obvious.

mpi

share|improve this question

asked Mar 6 at 20:40

rpmcnallyrpmcnally

1098

Rank 0
Data = [[A1,A2,A3],
 [A4,A5,A6]]
Indices = [1, 4]

Rank 1
Data = [[B1,B2,B3]]
Indices = [2]

Rank 2
Data = [[C1,C2,C3],
 [C4,C5,C6]]
Indices = [0,3]


Post Gather Data
Rank 0 = [[C1,C2,C3],
 [A1,A2,A3],
 [B1,B2,B3],
 [C4,C5,C6],
 [A4,A5,A6]]

share|improve this question

asked Mar 6 at 20:40

rpmcnallyrpmcnally

1098

share|improve this question

asked Mar 6 at 20:40

rpmcnallyrpmcnally

1098

asked Mar 6 at 20:40

rpmcnallyrpmcnally

1098

asked Mar 6 at 20:40

rpmcnallyrpmcnally

1098