Page not found

Comparing the performance of MPI vector datatypes

 


Performance of MPI vector datatypes


 

 

1. write a program to send a vector of 1000 elements of type MPI_DOUBLE, with a stride of 24 between each element. You may want to use these MPI routines in your solution:
2. Time routines and do several interations to get a good average and repeat the test 10 times.

Use the following example and complete the arguments of the calling routines for MPI_Type_vector 
#include < stdio.h>
#include < stdlib.h>
#include "mpi.h"

#define NUMBER_OF_TESTS 10

int main( argc, argv )
int argc;
char **argv;
{
    MPI_Datatype vec1, vec_n;
    int          blocklens[2];
    MPI_Datatype old_types[2];

    double       *buf, *lbuf;
    register double *in_p, *out_p;
    int          rank;
    int          n, stride;
    double       t1, t2, tmin;
    int          i, j, k, nloop;
    MPI_Status   status;

    MPI_Init( &argc, &argv );

    MPI_Comm_rank( MPI_COMM_WORLD, &rank );

    n      = 1000;
    stride = 24;
    nloop  = 100000/n;

    buf = (double *) malloc( n * stride * sizeof(double) );
    if (!buf) {
        fprintf( stderr, "Could not allocate send/recv buffer of size %d
",
                 n * stride );
        MPI_Abort( MPI_COMM_WORLD, 1 );
    }
    lbuf = (double *) malloc( n * sizeof(double) );
    if (!lbuf) {
        fprintf( stderr, "Could not allocated send/recv lbuffer of size %d
",
                 n );
        MPI_Abort( MPI_COMM_WORLD, 1 );
    }

    if (rank == 0) 
        printf( "Kind	n	stride	time (sec)	Rate (MB/sec)
" );

    /* Use a fixed vector type */
    MPI_Type_vector( ******************************* );
    MPI_Type_commit( ***** );

    tmin = 1000;
    for (k=0; k < NUMBER_OF_TESTS; k++) {
        if (rank == 0) {
            t1 = MPI_Wtime();
            for (j=0; j < nloop; j++) {
                MPI_Send(*********************************** );
                MPI_Recv( ******************************************* );
            }
            t2 = (MPI_Wtime() - t1) / nloop;
            if (t2 < tmin) tmin = t2;
        }
        else if (rank == 1) {
            for (j=0; j < nloop; j++) {
                MPI_Recv( buf, 1, vec1, 0, k, MPI_COMM_WORLD, &status );
                MPI_Send( buf, 1, vec1, 0, k, MPI_COMM_WORLD );
            }
        }
    }
    /* Convert to half the round-trip time */
    tmin = tmin / 2.0;
    if (rank == 0) {
        printf( "Vector	%d	%d	%f	%f
", 
                n, stride, tmin, n * sizeof(double) * 1.0e-6 / tmin );
    }
    MPI_Type_free( &vec1 );


    MPI_Finalize( );
    return 0;
}

Galaxy-P Case Study (Proteomics)

Professor Tim Griffin collaborated with MSI to develop and deploy tools to automate proteomics analysis tasks. The result was a set of tools incorporated into the Galaxy framework, and additional funding to expand his group’s work into related fields. The first objective was to provide an...
Results image: 

Genetic Studies of Epstein-Barr Virus Infection

Epstein-Barr Virus (EBV) is a common virus in the human herpes family. It is well known as the cause of infectious mononucleosis, but it also is associated with some forms of cancer and with autoimmune diseases such as systemic lupus erythematosus and multiple sclerosis. Much of the population is...

Bacterial Communities in Distribution Systems

Abstract: 
<h3 class="red">Simulating Mechanical Response of Bacterial Biofilms</h3><p>In nature bacteria have a dual mode of existence. Bacteria exist either in individual free-floating (or planktonic) form or as sessile communities on moist surfaces, known as biofilms. The biofilm mode of existence has been found to be the dominant one for bacteria in nature (more than 99%). While these biofilms are helpful in natural geochemical cycling of essential elements and nutrients in nature, they are a nuisance in variety of engineered systems. The scientific community woefully lacks the expertise to be able to control bacterial biofilms. Therefore, the Hozalski research group studies the mechanical properties of bacterial biofilms, devising strategies to measure the biofilm properties and performing related simulations in order to get a better understanding of biofilm control. Specifically, they are working with biofilm cohesive strength data collected in their laboratory and applying it in real-world systems (such as water distribution pipelines) to understand the stability of biofilms in engineered systems. To this end, they conduct Monte Carlo simulations in MATLAB to predict aggregate response of biofilms under various shear conditions in pipelines. Specifically they are looking at biomass eroded as a function of wall shear stress for different biofilm thicknesses, and conducting time-lapse studies under different flow conditions.</p><p>Another aspect of this research with involves conducting FEM-based Abaqus simulations of micro-cantilever test to determine cohesive strength and other properties. While the experiments assume homogenous stress and strain fields along with cylindrical shape assumption about the biofilm specimen, the FEM modeling allows the researchers to test the validity of their assumptions and derive realistic stress and strain fields with different biofilm specimen shapes in the micro-cantilever test. Many of these simulations in MATLAB and Abaqus are extremely complicated and time-consuming, requiring the use of MSI resources.</p><p>Return to this PI&#39;s <a href="https://www.msi.umn.edu/pi/1e734512ac2d526c06e95c94460384bc/44100">main page</a>.</p>
Group name: 
hozalski

IGF signaling in breast cancer

Abstract: 
<h4>IGF Signaling in Breast Cancer</h4><p><span style="font-size: 14px; line-height: 1.5;">Over the past 15 years, the Yee laboratory has shown that the insulin-like growth factors (IGFs) play an important role in regulating breast cancer proliferation, survival, and metastasis. This work, along with other laboratories, has established that the type I IGF receptor (IGF1R) is a viable target for breast cancer therapy. Currently, there are at least 10 targeted anti-IGF drugs currently in clinical trial with several ongoing and proposed studies in breast cancer. In order to optimize the use of targeted therapies, tools must be developed to predict clinically relevant outcomes. For example, expression of estrogen receptor-alpha (ER) is an important predictor of response to anti-estrogen therapy but mere expression of ER alone is imperfect. Indeed, the first clinical trials are being performed that are designed to improve outcomes of anti-estrogen therapy based on gene expression profiling of tumors. Thus, the idea that gene expression profiling can reveal complex biologic and clinical phenotypes is now established.</span></p><p>These researchers will take advantage of a series of breast cancer cells they have created to develop gene expression signatures that correlate with IGF responsiveness. They hypothesize that expression of specific insulin receptor substrate (IRS) adaptor proteins link IGF1R to identifiable gene signatures and cancer phenotypes. To test this hypothesis, they will: develop gene expression profiles to predict phenotypes driven by IGF1R signaling; validate these signatures <em>in vitro</em>, <em>in vivo</em>, and <em>in silico</em>; and examine changes in gene expression signatures <em>in vivo</em> after anti-IGF1R therapy. Given the vast number of new drugs active in breast cancer, developing tools to personalize therapeutic decisions will be critical to continued success in decreasing breast cancer mortality. The long-term goal of this research is to use the discoveries made in the laboratory to optimize the use of anti-IGF drugs.</p>
Group name: 
yeed

Center for Mass Spectrometry and Proteomics

Abstract: 
<h3 class="red">Center for Mass Spectrometry and Proteomics</h3><p>The <a href="http://cbs.umn.edu/cmsp/home">Center for Mass Spectrometry and Proteomics</a> (CMSP) is a core facility with the purpose of serving the research needs of the life sciences community at the University of Minnesota. The facility provides comprehensive service and expert advice in virtually all areas of mass spectrometry (MS) in the field of proteomics and metabolomics.</p><p>The Center has collaborated with MSI to extend the set of Galaxy tools and datatypes to perform proteomics analysis of data generated by mass spectrometry. Several UMN researchers used data generated from CMSP and workflows generated by Galaxy-P team to analyze data. Galaxy is a web-platform that offers an integrated informatics solution for analytical research in biological and medical science. It is designed so that wet-bench biologists can use it. It also allows researchers to share, collaborate, and publish their analyses.</p><p>Professor&nbsp;<a href="https://www.msi.umn.edu/pi/7dc3ead7f6ba21a3b94053a765a95da0/1006">Tim Griffin</a>&nbsp;(Biochemistry, Molecular Biology, and Biophysics; Faculty Director of CMSP) and Pratik Jagtap (Managing Director&nbsp;CMSP), along with MSI staff worked on tools to analyze proteomics data gathered from mass spectrometry. These included analysis tools, visualizers, and file-conversion tools.&nbsp;</p><p>The Galaxy-P team is promoting this work through publications in research journals and oral presentations at conferences, in order to inform the larger proteomics researcher community about these new tools.</p><p>A <a href="https://www.msi.umn.edu/content/expanding-galaxy-platform">Research Spotlight</a> about this work appeared on the MSI website in November 2015.</p><p>Return to this PI&#39;s <a href="https://www.msi.umn.edu/pi/c73beac7106f98e38b8f32ceb0625a58/39189">main page</a>.</p>
Group name: 
mscls

Cosolute Interactions With Nucleic Acids

Abstract: 
<h3 class="red">Cosolute Interactions With Nucleic Acids</h3><p>Biochemical reactions involving DNA, RNA, or proteins often involve large changes in the surface area of these biopolymers. The surface area changes expose (or bury) chemical functional groups that interact with the solvent and all its solute components. The thermodynamic favorability of the interactions between solvent and the surface area exposed or buried during DNA, RNA, and protein biochemical reactions can dramatically attenuate or enhance the rate of these reactions. Thus, we have the potential to use solvent and its solutes to probe the chemical composition of surface area changes during DNA, RNA, and protein biochemical reactions and develop a general method to ascertain biopolymer structural changes.</p><p>For several years this research group has quantified the interaction of neutral organic molecules like urea or amino acids (which are generically called cosolutes) with the surface area of nucleic acids. The researchers use a mix of uv-absorbance, differential scanning calorimetry, vapor pressure osmometry, solubility measurements, and molecular dynamics simulations to determine the excess (or deficiency) of these cosolutes near the nucleic acid surface area. If the interactions between cosolutes and the chemical functional groups in the newly exposed nucleic acid surface area after unfolding are thermodynamically favorable, the stability of the folded nucleic acid will be lower in aqueous cosolute solutions relative to water alone.</p><p>The group has recently begun exploring cosolute interactions with nucleobases, model compounds, and nucleosides using a novel partition assay at St. Olaf College. Briefly, they determine the partition coefficient of a model compound between an organic hexanol layer and an aqueous layer containing cosolute. Favorable model compound-cosolute interactions result in an increase in model compound concentration in the aqueous layer. The ultimate goal is to quantify cosolute interactions with specific chemical functional groups on the model compounds so that they can predict the magnitude and chemical composition of the surface area exposed or buried during biopolymer binding.</p><p>In order to this analysis, the researchers need reliable surface area calculations for their model compounds. They use software available through MSI to perform these calculations.</p><p>Return to this PI&#39;s <a href="https://www.msi.umn.edu/pi/a704087f69fb101f2e8823cb7a3811e0/31248">main page</a>.</p>
Group name: 
schwinef

Molecular Mechanisms of NRAS-Mediated Leukemia Stem Cell Self-Renewal

Abstract: 
<h3 class="red">Molecular Mechanisms of NRAS-Mediated Leukemia Stem Cell Self-Renewal</h3><p>Acute myeloid leukemia (AML) is frequently fatal because patients who initially respond to chemotherapy eventually relapse. Leukemia stem cells (LSCs) recapitulate the disease by self-renewal. LSC self-renewal is therefore critical to relapse. NRAS<sup>G12V</sup> is required for self-renewal in a murine AML model. Most anticancer therapies are designed to inhibit proliferation. Yet, in hematopoietic stem cells, the mechanisms of NRAS-mediated proliferation are distinct from self-renewal. Consequently, targeting proliferation may explain the failure of traditional chemotherapy to eradicate this disease. To study NRAS-mediated leukemia self-renewal, these researchers use a transgenic mouse model of AML with an&nbsp;<em>Mll-AF9</em>&nbsp;fusion and a tetracycline repressible, <em>NRAS<sup>G12V</sup></em>. Doxycycline abolishes <em>NRAS<sup>G12V</sup></em> expression leading to leukemia remission. These researchers hypothesize that NRAS-activated pathways required for self-renewal are limited to a subpopulation of cells with the LSC immunophenotype.</p><p>During 2015, the researchers examined the NRAS-activated stem-cell containing subpopulation of their mouse model by capturing single cells (~200 cells) and performing single cell RNA sequencing (along with bulk RNA population controls). They also started to capture cells from human AML samples (~100 cells) to investigate if subpopulations exist in these cells and if a self-renewing signature could be detected. They were able to implement this analysis using MSI HPC and storage resources. During 2016, they plan to capture and sequence more mouse model cells (~300) and human AML patient cells (~500). Both of these projects will include the proper population controls. They also plan to process a small set (20 samples) of bulk RNA sequencing on sorted subpopulations that they have identified in the mouse model.</p><p>The group is actively investigating if these important biomarkers of self-renewal using this NRAS AML mouse model are present in the human AML cells. Ultimately, the goal of this research is to target this self-renewing population therapeutically so that relapse of AML can be dramatically reduced.</p><p>Return to this PI&#39;s <a href="https://www.msi.umn.edu/pi/b587b88a60143917d5a28fd1e9a4c84c/10670">main page</a>.</p>
Group name: 
sachsz2

Wave Manipulation in Tunable Mechanical Metamaterials

Abstract: 
<h3 class="red">Wave Manipulation in Tunable Mechanical Metamaterials</h3><p>Mechanical metamaterials are man-made materials consisting of peculiar spatial arrangements of unit blocks, whose properties are superior to those exhibited by conventional materials and superior to those displayed by their individual components. Recent efforts in the applied physics community have been directed towards the development of techniques to provide metamaterials with tunable characteristics, i.e. the ability to modify their behavior in response to changes in certain externally controlled parameters or in the operational conditions. Popular avenues for tunability include methods involving electroelastic or magnetoelastic materials, thermoelastic effects, and buckling-induced lattice reconfigurations.</p><p>This group has recently explored two distinct strategies. The first is based on the correction of the mechanical properties of internal elements of metamaterials by means of piezoelectric phases. A recent &nbsp;extension of this paradigm involves the use of dielectric elastomers (such as electroactive polymers) which can undergo finite macroscopic deformation upon the application of voltage stimuli. They are currently in the process of simulating wave events excited in soft matamaterials undergoing drastic internal shape reconfiguration. The objective is to show that these changes in shape can result in non-negligible modifications of the spatial wave patterns induce in the material.</p><p>A second avenue towards tunability is based on the use of nonlinearity as a mechanism for wave control. Recently these researchers have introduced an approach to use the nonlinearity (material and geometric) of the medium to stretch the response of materials and distribute it intelligently over multiple wave modes, thus activating functionalities that are not achievable (in the same frequency ranges) in the corresponding linear cases. In this context, they are now performing simulations of nonlinear waves in a variety of media exhibiting nonlinear characteristics, including granular crystals, soft materials, and lattice materials with curved microstructural elements.</p><p>Return to this PI&#39;s <a href="https://www.msi.umn.edu/pi/74217a74ba15725098092a7329b2fe7c/10410">main page</a>.</p>
Group name: 
gonellas

Gaussian Graph Estimation and Recommendation System

Abstract: 
<h3 class="red">Gaussian Graph Estimation and Recommendation System</h3><p>Multiple graphical models have been widely used to describe structural changes of a network responding to certain experimental conditions, which are expressed in terms of conditional dependence between interacting units. For instance, time-varying functional connectivity in brain image analysis is described by node connectivity over a dynamic network, with each node corresponding to one region of interest. Motivated from network analysis under different experimental conditions, such as gene networks for disparate cancer subtypes, these researchers model structural changes over multiple networks with possible heterogeneities. In particular, they estimate multiple precision matrices describing dependencies among interacting units through maximum penalized likelihood. Of particular interest are homogeneous groups of similar entries across and zero-entries of these matrices, referred to as clustering and sparseness structures, respectively. A non-convex method is proposed to seek a sparse representation for each matrix and identify clusters of the entries across the matrices. An efficient method is developed on the basis of difference convex programming, the augmented Lagrangian method, and the blockwise coordinate descent method, which is scalable to hundreds of graphs of thousands nodes through a simple necessary and sufficient partition rule, which divides nodes into smaller disjoint subproblems excluding zero-coefficients nodes for arbitrary graphs with convex relaxation. Theoretically, a finite-sample error bound is derived for the proposed method to reconstruct the clustering and sparseness structures. This leads to consistent reconstruction of these two structures simultaneously, permitting the number of unknown parameters to be exponential in the sample size, and yielding the optimal performance of the oracle estimator as if the true structures were given a priori. Simulation studies suggest that the method enjoys the benefit of pursuing these two disparate kinds of structures, and compares favorably against its convex counterpart in the accuracy of structure pursuit and parameter estimation.</p><p>Return to this PI&#39;s <a href="https://www.msi.umn.edu/pi/b2e04ea7dc9ffc7ad657078958ad44c1/10028">main page</a>.</p>
Group name: 
shenx

Pages