Servidor

parallel_studio_xe_2015 NOT openmp 4.0 compliance?

Hi,

The following code does not work as expected:

------------------------------------------------------------------------

MODULE  SIM_DATA
    !$OMP declare target (GX)
    !     DIR$ ATTRIBUTES OFFLOAD:mic :: GX
    REAL  GX
END MODULE  SIM_DATA

!DIR$ ATTRIBUTES OFFLOAD:mic :: test0
SUBROUTINE  test0(n, b)
    INTEGER n, b(n)
    print *,'test0',  b
END

Draining store buffer on other core

Hello,

I've a weird question:

As I understand, mfence instruction causes draining of the store-buffer on the same core on which it was executed.

Is there some way for thread on core A, to cause draining of the store-buffer of core B, without running on core B? Maybe some dirty tricks like simulating IO or exception interrupts?

 

Thanks!

internal error: backend signals (when compiling an asynchronous offloading program)

Hello,

 

 

 

 

 

 

I tried to compile the following program on Intel Xeon Phi with icpc 13.0.1. But I failed to compile it. The error message says "": internal error: backend signals" This message tell me no useful information to solve the problem. Could anyone please give me an advice about how can I solve it. I'm thinking about going to buy new intel compiler. Is it compiled successfully in the latest version, such as Parallel Studio XE 2015 Composer?

在Intel® MIC多核架构上使用OpenMP*库的编程及调优实践(3)

本系列将主要介绍在Intel® MIC多核架构上运行及优化OpenMP*多线程程序的相关技术,且将围绕offload及native两种运行时执行环境展开详解。OpenMP编程模型包含了众多调优的编程接口及环境变量设置,本文将就此继续介绍如何更好地实现程序的高效运行。

1、使用offload模式时设置MIC_ENV_PREFIX来将Host环境的设置传播至MIC(target)计算节点

2、offload模式的多种关键字

3、使用-openmp-report来了解编译器对程序中OpenMP区域的优化处理

4、在不确定offload区域正在Host还是target上计算运行时,用户可以通过在代码中加入显示当前正执行的线程数目的方法来确定

5、OpenMP在Offload及Native模式下的不同缺省值

6、通过OMP_STACKSIZE设置OpenMP的的栈空间

  • Linux*
  • Servidor
  • C/C++
  • Updated info on virtualized host support?

    I'm looking to install a Phi unit in my workstation which is a virtualized machine; it's running Linux / kvm as the hypervisor with multiple Linux and Windows vm's.  The host machine is a current E5 system with full BIOS support for all the current virtualization features / functions, and the host fully supports large BAR addresses.  I would like to use the Phi from both Linux and Windows vm's (not simultaneously).

    _mm256_blend_epi16 doesn't work as documented

    The documentation for _mm256_blend_epi16 doesn't indicate that it operates on individual 128-bit channels, but this is the behavior I am seeing.  Is this the correct behavior?  Here is a reproducer code below showing the behavior for _mm256_blend_epi16 and _mm256_blend_epi32 where I attempt to insert a value into the first position of a vector using the blend instruction.

    bootup Phi in Xen/Dom0 failed with kernel crash

    Hello.

    I have unusual (and unsupported) config:
    - centos 5.10
    - kernel 3.10 (patched by citrix - XenServer Creedence Release Candidate)
    - xen from git
    - mpss-3.4.2 (some *.rpm without dependencies installed), kernel modules rebuilt from mpss-modules-3.4.2-1.src.rpm, libraries and base management compiled from sources (mpss-micmgmt-3.4.2.tar.bz2,mpss-daemon-3.4.2.tar.bz2,mpss-metadata-3.4.2.tar.bz2)

    cannot run GDB-MIC

    Hello,

    I am unable to debug Xeon Phi code. I am using Intel Parallel Studio 2015 and its debugger. Specifically I am not able to run gdb-mic.

    I follow these steps

    source compilervars.sh intel64
    gdb-mic

    I get following error. "gdb-mic: error while loading shared libraries: libtinfo.so.5: cannot open shared object file: No such file or directory".

    The missing libraries are found at

    performance jpeg on phi

    Hello.

    I am new in Phi world. I tested jpeg encoding performance (tested on image 1680x1050, single thread):

    14 ms - xeon E5-2620 - libjpeg-turbo with SIMD
    39 ms - xeon E5-2620 - libjpeg-turbo without SIMD
    24 ms - xeon E5-2620 - intel ijg80b with IPP
    42 ms - xeon E5-2620 - intel ijg80b without IPP (#undef USE_IPP)
    386 ms - xeon Phi 31S1P - native intel ijg80b with IPP
    557 ms - xeon Phi 31S1P - native intel ijg80b without IPP (#undef USE_IPP)
    435 ms - xeon Phi 31S1P - native libjpeg-turbo without SIMD

    Suscribirse a Servidor