Re: [Beowulf] [xcat-user] Cannot run offload codes with Intel Xeon Phi cards deployed via xCAT

2013-08-22 Thread Christopher Samuel
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Hi Andrew, On 23/08/13 02:42, Andrew Howard wrote: > Did you configure the cards with micctrl like the docs tell you to? > Stop mpss, run micctrl --initdefaults, then start MPSS. Otherwise > micuser won't exist. xCAT does that for you, the problem w

Re: [Beowulf] [xcat-user] Cannot run offload codes with Intel Xeon Phi cards deployed via xCAT

2013-08-22 Thread Andrew Howard
Did you configure the cards with micctrl like the docs tell you to? Stop mpss, run micctrl --initdefaults, then start MPSS. Otherwise micuser won't exist. -Andy -- Andrew Howard HPC Systems Engineer Purdue University (765) 889-2523 On Wed, Aug 21, 2013 at 12:05 AM, Christopher Samuel wrote: >

Re: [Beowulf] Good demo applications for small, slow cluster

2013-08-22 Thread Andrew M.A. Cater
On Thu, Aug 22, 2013 at 02:03:57PM -0400, Prentice Bisbal wrote: > > On 08/21/2013 12:45 PM, Douglas Eadline wrote: > >> Sorts in general.. Good idea. > >> > >> Yes, we'll do a distributed computing bubble sort. > >> > >> Interesting, though.. There are probably simple algorithms which are > >> ef

Re: [Beowulf] Good demo applications for small, slow cluster

2013-08-22 Thread Lux, Jim (337C)
Yes.. another good choice.. Simple algorithm, lots of ways to divide it up. Simple enough that the focus would be on the parallelizing, rather than numerical convergence issues or something like that. Jim Lux -Original Message- From: Max R. Dechantsreiter [mailto:m...@performancejones.c

Re: [Beowulf] Restricting users from ssh into nodes

2013-08-22 Thread Prentice Bisbal
On 08/22/2013 02:05 PM, Reuti wrote: > Am 22.08.2013 um 20:00 schrieb Prentice Bisbal: > >> On 08/20/2013 02:03 PM, Mark Hahn wrote: 1. Many people's job scripts use ssh either directly (to say clean up /tmp) or indirectly from mpirun. >>> sure. >>> (good mpirun's use the batch engin

Re: [Beowulf] Restricting users from ssh into nodes

2013-08-22 Thread Reuti
Am 22.08.2013 um 20:00 schrieb Prentice Bisbal: > On 08/20/2013 02:03 PM, Mark Hahn wrote: >>> 1. Many people's job scripts use ssh either directly (to say clean up /tmp) >>> or indirectly from mpirun. >> sure. >> >>> (good mpirun's use the batch engine's per-node >>> daemon to launch the binarie

Re: [Beowulf] Good demo applications for small, slow cluster

2013-08-22 Thread Prentice Bisbal
On 08/21/2013 12:45 PM, Douglas Eadline wrote: >> Sorts in general.. Good idea. >> >> Yes, we'll do a distributed computing bubble sort. >> >> Interesting, though.. There are probably simple algorithms which are >> efficient in a single processor environment, but become egregiously >> inefficient

Re: [Beowulf] Restricting users from ssh into nodes

2013-08-22 Thread Prentice Bisbal
On 08/20/2013 02:03 PM, Mark Hahn wrote: >> 1. Many people's job scripts use ssh either directly (to say clean up /tmp) >> or indirectly from mpirun. > sure. > >> (good mpirun's use the batch engine's per-node >> daemon to launch the binaries not ssh). > why should a scheduler have daemons clutteri

Re: [Beowulf] Good demo applications for small, slow cluster

2013-08-22 Thread Lux, Jim (337C)
Yes.. Like the example in Dan McCracken's FORTRAN textbook. A 2D plate with a square hole in the middle and different boundary conditions on the edges. Simple algorithm, so the person doesn't get wound up in the details of the FEM calculation. Lots of potential ways to partition the problem (di

Re: [Beowulf] Good demo applications for small, slow cluster

2013-08-22 Thread John Hearns
Jim - how about heat transport? Simple model of a cold flat plate heated to a constant N degrees at one point or edge. You split into square cells and solve heat transport equation in each cell. Temperature plot vs time makes nice movie. You add more MPI processes and the model runs faster or you c