Microsoft Open R和Intel Xeon Phi [英] Microsoft Open R and Intel Xeon Phi

问题描述

亲爱的，

           我有一台工作站配有I7 3770处理器和2台Xeon Phi 3120A，我想用它来加速R。

            I have a workstation with an I7 3770 Processor and 2 Xeon Phi 3120A that I want to use to speed up R.

我使用Windows 10和Microsoft R Open 3.3.1与英特尔MKL（由MRO安装）我按照本文的不同方式设置我的环境安德鲁·埃克斯特罗姆和相关链接与英特尔Phi协处理器进行快速并行计算，但我没有管理
来使其工作。

I use Windows 10 and Microsoft R Open 3.3.1 with Intel MKL (installed by MRO) and I set up my enviroment in different ways following this article Fast parallel computing with Intel Phi coprocessors by Andrew Ekstrom and related links but I didn't manage to make it works.

为了测试，我使用了Andrew Ekstrom的文章中的代码，我通过micsmc实用程序检查了PHI的行为。

For testing I used the code from Andrew Ekstrom's article and I checked PHIs behaviour by micsmc utility.

目前我的shell环境设置如下：

At the moment my shell enviroment is set as follow:

INTEL_LICENSE_FILE = C：\Program Files（x86）\ Common Files\Intel \Licenses

INTEL_MPSS_HOME = C：\ Program Files \ Intel \ MPSS \
$
INTEL_MPSS_HOST_SDK = C：\Program Files \ Intel \ MPSS \sdk \
$
KMP_AFFINITY =粒度=精细，平衡

MIC_ENV_PREFIX = MIC

MIC_KMP_AFFINITY =粒度=精细，平衡

MIC_MKL_NUM_THRE ADS = 224

MIC_OMP_NUM_THREADS = 224

MIC_SYSROOT = C：\Program Files\Intel\MPSS \

MKL_HOST_WORKDIVISION = 0.1

MKL_MIC_ENABLE = 1

MKL_MIC_MAX_MEMORY = 8GB

MKL_MIC_WORKDIVISION = 0.9

MKL_NUM_THREADS = 4

MKL_OMP_NUM_THREADS = 4

NUMBER_OF_PROCESSORS = 8
$
OFFLOAD_DEVICES = 0,1
$
OFFLOAD_REPORT = 2

OMP_NUM_THREADS = 4

INTEL_LICENSE_FILE=C:\Program Files (x86)\Common Files\Intel\Licenses
INTEL_MPSS_HOME=C:\Program Files\Intel\MPSS\
INTEL_MPSS_HOST_SDK=C:\Program Files\Intel\MPSS\sdk\
KMP_AFFINITY=granularity=fine,balanced
MIC_ENV_PREFIX=MIC
MIC_KMP_AFFINITY=granularity=fine,balanced
MIC_MKL_NUM_THREADS=224
MIC_OMP_NUM_THREADS=224
MIC_SYSROOT=C:\Program Files\Intel\MPSS\
MKL_HOST_WORKDIVISION=0.1
MKL_MIC_ENABLE=1
MKL_MIC_MAX_MEMORY=8GB
MKL_MIC_WORKDIVISION=0.9
MKL_NUM_THREADS=4
MKL_OMP_NUM_THREADS=4
NUMBER_OF_PROCESSORS=8
OFFLOAD_DEVICES=0,1
OFFLOAD_REPORT=2
OMP_NUM_THREADS=4

有人知道错误吗？

$
$
b $ b

$




$
$






$ b

推荐答案

这可能是您所指的文章，但如果没有，可能会有一些有用的信息：

This may be the article you're referring to, but if not, there may be some useful info at:

http://blog.revolutionanalytics.com/2015/05/behold-t he-power-of-parallel.html

http://blog.revolutionanalytics.com/2015/05/behold-the-power-of-parallel.html

#David Smith

# David Smith

这篇关于Microsoft Open R和Intel Xeon Phi的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！