Xeon Phi: Memory leak?


I write a rpc-like application to run on Xeon Phi and dynamic link with MKL. However after the program finish (exit), the memory on micsmc is still high, for example, at one of the Xeon Phi nodes the memory usage is of 1172MB.  Howver if I reboot the Xeon Phi, the memory usage after reboot would be of 418MB.  It seems to me that the system has leaking memory problem. How can I recover the memory lost without rebooting Xeon Phi?  Thanks,



Linking C++ and Fortran objects

Platform: CentOS release 6.6 (Final)

GCC: 4.9.1

Intel Parallel Studio Cluster Edition 2016

I am working on a bioinformatics application that my team and I have successfully compiled for Intel Xeon processor. Now we're trying to compile the application for MIC but we have a hard time linking it. The error that we get is the following:

x86_64-k1om-linux-ld: cannot find -lstadc++
/opt/mpss/3.2.1/sysroots/x86_64-mpsssdk-linux/lib/libc-2.14.90.so: could not read symbols: File in wrong format

MPSS 3.6 is available


Please note that the new MPSS 3.6 is available at https://software.intel.com/en-us/articles/intel-manycore-platform-software-stack-mpss

MPSS 3.6 will be supported until the end of March 2016. Major changes in MPSS 3.6 include:

  • hStreams Beta

  • k1om gcc update from 4.7.0 experimental to 5.1

  • k1om glibc update to 2.21

  • MLNX_OFED 2.4 support

  • RHEL 6.4 support drop

NFV topics at the Intel® Developer Forum, the DPDK and Network Builders Summit

The Intel® Developer Forum 2015 took place August 18–20 in San Francisco. Intel® also sponsored a day-long Data Plane Development Kit (DPDK) and Network Builders Summit on August 17, 2015. There were several interesting sessions at the conference relating to Network Functions Virtualization (NFV). In this blog post, we look at some of the highlights and share pointers to session slides or presentations.

undefined symbol in running not in linking of executable linked with a static library (offload model)


I am trying to run an offload application on Xeon Phi, and I get the following error.

"On the sink, dlopen() returned NULL. The result of dlerror() is "/var/volatile/tmp/coi_procs/1/205497/load_lib/iccouthqsEw3: undefined symbol: DISTMEM_rank"
On the remote process, dlopen() failed. The error message sent back from the sink is /var/volatile/tmp/coi_procs/1/205497/load_lib/iccouthqsEw3: undefined symbol: DISTMEM_rank
offload error: cannot load library to the device 0 (error code 20)"

HHVM and Haswell

There is no question that PHP is the most popular language in use today to implement server code on web pages. Something like 80% of web sites are implemented using it. The extremely popular site Facebook has implemented their own PHP interpreter called HHVM, and they are doing code development on it as an open source project. Because of its performance and its open source methodology, it is also being adopted by sites other than Facebook.

Naming shared libraries


I am trying to adjust a large project for working on Xeon Phi with the offload model and I have a few questions.

1. My programs should be linked with some of many static libraries of hundreds of lines. I tried to built these libraries with the -offload-attribute-target=mic option to annotate my functions and variables for mic offloading, and then link them with xiar crs -qoffload-build option. Is it a reasonable approach or am I missing something?

Problem compiling LEO_Tutorial


I'm running Microsoft Visual Studio 2013 (u4), with Intel Parallel Studio 2016 C++. I'm trying to compile the LEO_Tutorial project that comes with Parallel Studio in the mic_samples.zip, and having some problems. According to the tutorial, I should load the solution, set it to use Intel C++, and it should just build out of the box.

The error I get in the Build Output is:-

Konzern abonnieren