Parallel Computing

Only one change in latest MPSS release?

Hi there,

The latest MPSS release (mpss_gold_update_3-2.1.6720-16) lists no changes, and only one resolution in its release notes, which is:

Description:        [OFED] Kernel Soft Lockup when stopping mic-ofed service

Is that really the only change in that release?   I'm curious because if so I can skip this update (which would be handy).

Xeon Phi not quite working

I am trying to install a Xeon Phi 3120A in a custom machine with ASUS P9X79-E WS motherboard (compatible with the Xeon Phi according to ASUS), i7-3930K processor and 1000W power supply. I am having similar issues with both Windows 7 (beta drivers) and CentOS 6.4 (latest drivers as well as N-2 version). The machine boots and finds the Xeon Phi. The initial commands work, but everything afterwards causes the machine to freeze: the screen is still on, but no mouse, no keyboard, no Ctrl+Alt+Del.

what is the difference between fourth gen. i7 and extreme edition ? help

Dear Sir
I am Sharif AL-Hamidi , I am a film maker and also I get used to edit my movies
in addition to that I work as a graphic designer.  I need to buy a new Desktop PC to be suitable for
my job, but i read a lot of things in your website about the fourth generation of core i7 CPU's
Could they be enough and suitable  for my work and multitasking jobs because while I am editing also I work with graphic design at the same time
 
Does they better than the extreme edition or not although they are released after the extreme edition?

__sec_reduce_all_nonzero crashing compiler with backend signals

Why does the following code crash the compiler with backend signals?  Am I doing something wrong?

#include <mathimf.h>
#include <float.h>

double avec[] = {0, 1, 2, 3, 4};

__sec_reduce_all_nonzero(_finite(avec[0:sizeof(avec)/sizeof(avec[0])]));

Thanks.

loop unrolling issues on MIC

Hi,

I am trying to play with an Intel MIC.

I have hit an issue which I traced back to unrolling.

It seems that the compiler fails to unroll loops that contain constants in the expressions. I have attached a small code showing the issue. 

the problem goes around this part

t1=omp_get_wtime()
  do i=1,ns
    a=a/real(n1,pr)
  enddo
 t1=omp_get_wtime()-t1

a above is an array. if n1 is a paremeter I get the first time below, if n1 is a normal variable I get the second time.

Compilation of pdsh and other tools

I would like to make pdsh (Parallel Distributed Shell) available to the MICs, but I'm not quite sure how to do it, if it's possible or if anyone else has even tried.

I looked at the PDF "Configuring Intel Xeon Phi coprocessors inside a cluster", which covers cross and native compiling.  Do I need to build a toolchain in order get pdsh on the MICs?

If anyone has any experience of this and/or any pointers it would be much appreciated.

is one rear fan enough for two Xeon Phi cards onboard?

Hi,

We are using the following motherboard (http://www.supermicro.com/products/system/3u/6037/SYS-6037R-72RFT_.cfm?parts=SHOW) to build a system featuring with Xeon Phi cards. We plan to have two Xeon Phi cards onboard. This motherboard only supports one rear GPU fan. We are wondering if one rear fan is enough for cooling two Xeon Phi cards together.

Thanks,

Parallel Computing abonnieren