Ken,

"-bind-to core" gives me the following topology:

process 0 binding: 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 1 binding: 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 2 binding: 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 3 binding: 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 4 binding: 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 5 binding: 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 6 binding: 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0
process 7 binding: 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0
process 8 binding: 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0
process 9 binding: 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0
process 10 binding: 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0
process 11 binding: 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0
process 12 binding: 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0
process 13 binding: 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
process 14 binding: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0
process 15 binding: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0

but I want this:

process 0 binding: 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 1 binding: 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 2 binding: 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 3 binding: 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 4 binding: 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 5 binding: 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0
process 6 binding: 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0
process 7 binding: 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0
process 8 binding: 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0
process 9 binding: 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0
process 10 binding: 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0
process 11 binding: 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
process 12 binding: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0
process 13 binding: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0
process 14 binding: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0
process 15 binding: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0

The latter gives me better performance for my application, and I am guessing it's because I have evenly distribute the processes among the two sockets (sequentially). Which is why I resorted to what I had originally with the custom binding.

Thanks,

On Mon, May 11, 2015 at 4:53 PM, Kenneth Raffenetti <raffenet@mcs.anl.gov> wrote:

Justin,

Try using the "-bind-to core" option instead. It should do exactly what you are wanting. See this page with examples for more details https://wiki.mpich.org/mpich/index.php/Using_the_Hydra_Process_Manager#Process-core_Binding

Ken

On 05/11/2015 04:48 PM, Justin Chang wrote:

Hello everyone,

I am working with an HPC machine that has this configuring for a single
compute node:

Machine (64GB total)
NUMANode L#0 (P#0 32GB)
Socket L#0 + L3 L#0 (25MB)
L2 L#0 (256KB) + L1d L#0 (32KB) + L1i L#0 (32KB) + Core L#0 + PU
L#0 (P#0)
L2 L#1 (256KB) + L1d L#1 (32KB) + L1i L#1 (32KB) + Core L#1 + PU
L#1 (P#1)
L2 L#2 (256KB) + L1d L#2 (32KB) + L1i L#2 (32KB) + Core L#2 + PU
L#2 (P#2)
L2 L#3 (256KB) + L1d L#3 (32KB) + L1i L#3 (32KB) + Core L#3 + PU
L#3 (P#3)
L2 L#4 (256KB) + L1d L#4 (32KB) + L1i L#4 (32KB) + Core L#4 + PU
L#4 (P#4)
L2 L#5 (256KB) + L1d L#5 (32KB) + L1i L#5 (32KB) + Core L#5 + PU
L#5 (P#5)
L2 L#6 (256KB) + L1d L#6 (32KB) + L1i L#6 (32KB) + Core L#6 + PU
L#6 (P#6)
L2 L#7 (256KB) + L1d L#7 (32KB) + L1i L#7 (32KB) + Core L#7 + PU
L#7 (P#7)
L2 L#8 (256KB) + L1d L#8 (32KB) + L1i L#8 (32KB) + Core L#8 + PU
L#8 (P#8)
L2 L#9 (256KB) + L1d L#9 (32KB) + L1i L#9 (32KB) + Core L#9 + PU
L#9 (P#9)
HostBridge L#0
PCIBridge
PCI 1000:0087
Block L#0 "sda"
PCIBridge
PCI 15b3:1003
Net L#1 "eth0"
Net L#2 "ib0"
OpenFabrics L#3 "mlx4_0"
PCIBridge
PCI 8086:1521
Net L#4 "eth1"
PCI 8086:1521
Net L#5 "eth2"
PCIBridge
PCI 102b:0533
PCI 8086:1d02
NUMANode L#1 (P#1 32GB) + Socket L#1 + L3 L#1 (25MB)
L2 L#10 (256KB) + L1d L#10 (32KB) + L1i L#10 (32KB) + Core L#10 +
PU L#10 (P#10)
L2 L#11 (256KB) + L1d L#11 (32KB) + L1i L#11 (32KB) + Core L#11 +
PU L#11 (P#11)
L2 L#12 (256KB) + L1d L#12 (32KB) + L1i L#12 (32KB) + Core L#12 +
PU L#12 (P#12)
L2 L#13 (256KB) + L1d L#13 (32KB) + L1i L#13 (32KB) + Core L#13 +
PU L#13 (P#13)
L2 L#14 (256KB) + L1d L#14 (32KB) + L1i L#14 (32KB) + Core L#14 +
PU L#14 (P#14)
L2 L#15 (256KB) + L1d L#15 (32KB) + L1i L#15 (32KB) + Core L#15 +
PU L#15 (P#15)
L2 L#16 (256KB) + L1d L#16 (32KB) + L1i L#16 (32KB) + Core L#16 +
PU L#16 (P#16)
L2 L#17 (256KB) + L1d L#17 (32KB) + L1i L#17 (32KB) + Core L#17 +
PU L#17 (P#17)
L2 L#18 (256KB) + L1d L#18 (32KB) + L1i L#18 (32KB) + Core L#18 +
PU L#18 (P#18)
L2 L#19 (256KB) + L1d L#19 (32KB) + L1i L#19 (32KB) + Core L#19 +
PU L#19 (P#19)

If I ran my program with 16 processes, I would have the follow batch script:

#!/bin/bash
#SBATCH -N 1
#SBATCH -n 20
#SBATCH -t 0-09:00
#SBATCH -o output.txt

mpiexec.hydra -bind-to user:0,1,2,3,4,5,6,7,10,11,12,13,14,15,16,17 -n
16 ./my_program <args>

This would give me decent speedup. However, what if I want to use 32
processes? Since each node only has 20 cores I would need #SBATCH -N 2
and #SBATCH -n 40. However, I want ranks 0-15 and 16-31 to have the same
mapping as above but on different compute nodes, so how would I do this?
Or would the above line work so long as I have a multiple of 16 processes?

Thanks,

--
Justin Chang
PhD Candidate, Civil Engineering - Computational Sciences
University of Houston, Department of Civil and Environmental Engineering
Houston, TX 77004
(512) 963-3262

_______________________________________________
discuss mailing list discuss@mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss

_______________________________________________
discuss mailing list discuss@mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss