portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-09 23:45:36 +00:00
Code Releases Activity

Merge branch 'feature/contractor' into develop

Browse Source
This commit is contained in:
Antonin Portelli 2018-11-22 15:44:46 +00:00
commit 6364aa8acf
20 changed files with 1446 additions and 90 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
Grid/serialisation/Hdf5IO.cc
View File

@ -61,9 +61,9 @@ Group & Hdf5Writer::getGroup(void)
}
// Reader implementation ///////////////////////////////////////////////////////
Hdf5Reader::Hdf5Reader(const std::string &fileName)
Hdf5Reader::Hdf5Reader(const std::string &fileName, const bool readOnly)
: fileName_(fileName)
, file_(fileName.c_str(), H5F_ACC_RDWR)
, file_(fileName.c_str(), readOnly ? H5F_ACC_RDONLY : H5F_ACC_RDWR)
{
group_ = file_.openGroup("/");
readSingleAttribute(dataSetThres_, HDF5_GRID_GUARD "dataset_threshold",

2
Grid/serialisation/Hdf5IO.h
View File

@ -54,7 +54,7 @@ namespace Grid
class Hdf5Reader: public Reader<Hdf5Reader>
{
public:
Hdf5Reader(const std::string &fileName);
Hdf5Reader(const std::string &fileName, const bool readOnly = true);
virtual ~Hdf5Reader(void) = default;
bool push(const std::string &s);
void pop(void);

2
Grid/threads/Threads.h
View File

@ -47,6 +47,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#else
#define PARALLEL_FOR_LOOP
#define PARALLEL_FOR_LOOP_INTERN
#define PARALLEL_FOR_LOOP_REDUCE(op, var)
#define PARALLEL_NESTED_LOOP2
#define PARALLEL_NESTED_LOOP5
#define PARALLEL_REGION
@ -58,6 +59,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#define parallel_for_internal PARALLEL_FOR_LOOP_INTERN for
#define parallel_for_nest2 PARALLEL_NESTED_LOOP2 for
#define parallel_for_nest5 PARALLEL_NESTED_LOOP5 for
#define parallel_critical PARALLEL_CRITICAL
namespace Grid {

17
Grid/util/Sha.h
View File

@ -28,16 +28,31 @@
extern "C" {
#include <openssl/sha.h>
}
#ifdef USE_IPP
#include "ipp.h"
#endif
#pragma once
class GridChecksum
{
public:
static inline uint32_t crc32(void *data,size_t bytes)
static inline uint32_t crc32(const void *data, size_t bytes)
{
return ::crc32(0L,(unsigned char *)data,bytes);
}
#ifdef USE_IPP
static inline uint32_t crc32c(const void* data, size_t bytes)
{
uint32_t crc32c = ~(uint32_t)0;
ippsCRC32C_8u(reinterpret_cast<const unsigned char *>(data), bytes, &crc32c);
ippsSwapBytes_32u_I(&crc32c, 1);
return ~crc32c;
}
#endif
template <typename T>
static inline std::string sha256_string(const std::vector<T> &hash)
{

354
Hadrons/A2AMatrix.hpp
View File

@ -32,11 +32,19 @@ See the full license in the file "LICENSE" in the top level distribution directo
#include <Hadrons/Global.hpp>
#include <Hadrons/TimerArray.hpp>
#include <Grid/Eigen/unsupported/CXX11/Tensor>
#ifdef USE_MKL
#include "mkl.h"
#include "mkl_cblas.h"
#endif
#ifndef HADRONS_A2AM_NAME
#define HADRONS_A2AM_NAME "a2aMatrix"
#endif
#ifndef HADRONS_A2AM_IO_TYPE
#define HADRONS_A2AM_IO_TYPE ComplexF
#endif
#define HADRONS_A2AM_PARALLEL_IO
BEGIN_HADRONS_NAMESPACE
@ -51,6 +59,12 @@ BEGIN_HADRONS_NAMESPACE
template <typename T>
using A2AMatrixSet = Eigen::TensorMap<Eigen::Tensor<T, 5, Eigen::RowMajor>>;
template <typename T>
using A2AMatrix = Eigen::Matrix<T, -1, -1, Eigen::RowMajor>;
template <typename T>
using A2AMatrixTr = Eigen::Matrix<T, -1, -1, Eigen::ColMajor>;
/******************************************************************************
* Abstract class for A2A kernels *
******************************************************************************/
@ -76,10 +90,15 @@ public:
// constructors
A2AMatrixIo(void) = default;
A2AMatrixIo(std::string filename, std::string dataname,
const unsigned int nt, const unsigned int ni,
const unsigned int nj);
const unsigned int nt, const unsigned int ni = 0,
const unsigned int nj = 0);
// destructor
~A2AMatrixIo(void) = default;
// access
unsigned int getNi(void) const;
unsigned int getNj(void) const;
unsigned int getNt(void) const;
size_t getSize(void) const;
// file allocation
template <typename MetadataType>
void initFile(const MetadataType &d, const unsigned int chunkSize);
@ -88,9 +107,11 @@ public:
const unsigned int blockSizei, const unsigned int blockSizej);
void saveBlock(const A2AMatrixSet<T> &m, const unsigned int ext, const unsigned int str,
const unsigned int i, const unsigned int j);
template <template <class> class Vec, typename VecT>
void load(Vec<VecT> &v, double *tRead = nullptr);
private:
std::string filename_, dataname_;
unsigned int nt_, ni_, nj_;
std::string filename_{""}, dataname_{""};
unsigned int nt_{0}, ni_{0}, nj_{0};
};
/******************************************************************************
@ -136,6 +157,226 @@ private:
std::vector<IoHelper> nodeIo_;
};
/******************************************************************************
* A2A matrix contraction kernels *
******************************************************************************/
class A2AContraction
{
public:
// accTrMul(acc, a, b): acc += tr(a*b)
template <typename C, typename MatLeft, typename MatRight>
static inline void accTrMul(C &acc, const MatLeft &a, const MatRight &b)
{
if ((MatLeft::Options == Eigen::RowMajor) and
(MatRight::Options == Eigen::ColMajor))
{
parallel_for (unsigned int r = 0; r < a.rows(); ++r)
{
C tmp;
#ifdef USE_MKL
dotuRow(tmp, r, a, b);
#else
tmp = a.row(r).conjugate().dot(b.col(r));
#endif
parallel_critical
{
acc += tmp;
}
}
}
else
{
parallel_for (unsigned int c = 0; c < a.cols(); ++c)
{
C tmp;
#ifdef USE_MKL
dotuCol(tmp, c, a, b);
#else
tmp = a.col(c).conjugate().dot(b.row(c));
#endif
parallel_critical
{
acc += tmp;
}
}
}
}
template <typename MatLeft, typename MatRight>
static inline double accTrMulFlops(const MatLeft &a, const MatRight &b)
{
double n = a.rows()*a.cols();
return 8.*n;
}
// mul(res, a, b): res = a*b
#ifdef USE_MKL
template <template <class, int...> class Mat, int... Opts>
static inline void mul(Mat<ComplexD, Opts...> &res,
const Mat<ComplexD, Opts...> &a,
const Mat<ComplexD, Opts...> &b)
{
static const ComplexD one(1., 0.), zero(0., 0.);
if ((res.rows() != a.rows()) or (res.cols() != b.cols()))
{
res.resize(a.rows(), b.cols());
}
if (Mat<ComplexD, Opts...>::Options == Eigen::RowMajor)
{
cblas_zgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
a.cols(), &one, a.data(), a.cols(), b.data(), b.cols(), &zero,
res.data(), res.cols());
}
else if (Mat<ComplexD, Opts...>::Options == Eigen::ColMajor)
{
cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
a.cols(), &one, a.data(), a.rows(), b.data(), b.rows(), &zero,
res.data(), res.rows());
}
}
template <template <class, int...> class Mat, int... Opts>
static inline void mul(Mat<ComplexF, Opts...> &res,
const Mat<ComplexF, Opts...> &a,
const Mat<ComplexF, Opts...> &b)
{
static const ComplexF one(1., 0.), zero(0., 0.);
if ((res.rows() != a.rows()) or (res.cols() != b.cols()))
{
res.resize(a.rows(), b.cols());
}
if (Mat<ComplexF, Opts...>::Options == Eigen::RowMajor)
{
cblas_cgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
a.cols(), &one, a.data(), a.cols(), b.data(), b.cols(), &zero,
res.data(), res.cols());
}
else if (Mat<ComplexF, Opts...>::Options == Eigen::ColMajor)
{
cblas_cgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
a.cols(), &one, a.data(), a.rows(), b.data(), b.rows(), &zero,
res.data(), res.rows());
}
}
#else
template <typename Mat>
static inline void mul(Mat &res, const Mat &a, const Mat &b)
{
res = a*b;
}
#endif
template <typename Mat>
static inline double mulFlops(const Mat &a, const Mat &b)
{
double nr = a.rows(), nc = a.cols();
return nr*nr*(6.*nc + 2.*(nc - 1.));
}
private:
template <typename C, typename MatLeft, typename MatRight>
static inline void makeDotRowPt(C * &aPt, unsigned int &aInc, C * &bPt,
unsigned int &bInc, const unsigned int aRow,
const MatLeft &a, const MatRight &b)
{
if (MatLeft::Options == Eigen::RowMajor)
{
aPt = a.data() + aRow*a.cols();
aInc = 1;
}
else if (MatLeft::Options == Eigen::ColMajor)
{
aPt = a.data() + aRow;
aInc = a.rows();
}
if (MatRight::Options == Eigen::RowMajor)
{
bPt = b.data() + aRow;
bInc = b.cols();
}
else if (MatRight::Options == Eigen::ColMajor)
{
bPt = b.data() + aRow*b.rows();
bInc = 1;
}
}
#ifdef USE_MKL
template <typename C, typename MatLeft, typename MatRight>
static inline void makeDotColPt(C * &aPt, unsigned int &aInc, C * &bPt,
unsigned int &bInc, const unsigned int aCol,
const MatLeft &a, const MatRight &b)
{
if (MatLeft::Options == Eigen::RowMajor)
{
aPt = a.data() + aCol;
aInc = a.cols();
}
else if (MatLeft::Options == Eigen::ColMajor)
{
aPt = a.data() + aCol*a.rows();
aInc = 1;
}
if (MatRight::Options == Eigen::RowMajor)
{
bPt = b.data() + aCol*b.cols();
bInc = 1;
}
else if (MatRight::Options == Eigen::ColMajor)
{
bPt = b.data() + aCol;
bInc = b.rows();
}
}
template <typename MatLeft, typename MatRight>
static inline void dotuRow(ComplexF &res, const unsigned int aRow,
const MatLeft &a, const MatRight &b)
{
const ComplexF *aPt, *bPt;
unsigned int aInc, bInc;
makeDotRowPt(aPt, aInc, bPt, bInc, aRow, a, b);
cblas_cdotu_sub(a.cols(), aPt, aInc, bPt, bInc, &res);
}
template <typename MatLeft, typename MatRight>
static inline void dotuCol(ComplexF &res, const unsigned int aCol,
const MatLeft &a, const MatRight &b)
{
const ComplexF *aPt, *bPt;
unsigned int aInc, bInc;
makeDotColPt(aPt, aInc, bPt, bInc, aCol, a, b);
cblas_cdotu_sub(a.rows(), aPt, aInc, bPt, bInc, &res);
}
template <typename MatLeft, typename MatRight>
static inline void dotuRow(ComplexD &res, const unsigned int aRow,
const MatLeft &a, const MatRight &b)
{
const ComplexD *aPt, *bPt;
unsigned int aInc, bInc;
makeDotRowPt(aPt, aInc, bPt, bInc, aRow, a, b);
cblas_zdotu_sub(a.cols(), aPt, aInc, bPt, bInc, &res);
}
template <typename MatLeft, typename MatRight>
static inline void dotuCol(ComplexD &res, const unsigned int aCol,
const MatLeft &a, const MatRight &b)
{
const ComplexD *aPt, *bPt;
unsigned int aInc, bInc;
makeDotColPt(aPt, aInc, bPt, bInc, aCol, a, b);
cblas_zdotu_sub(a.rows(), aPt, aInc, bPt, bInc, &res);
}
#endif
};
/******************************************************************************
* A2AMatrixIo template implementation *
******************************************************************************/
@ -148,6 +389,31 @@ A2AMatrixIo<T>::A2AMatrixIo(std::string filename, std::string dataname,
, nt_(nt), ni_(ni), nj_(nj)
{}
// access //////////////////////////////////////////////////////////////////////
template <typename T>
unsigned int A2AMatrixIo<T>::getNt(void) const
{
return nt_;
}
template <typename T>
unsigned int A2AMatrixIo<T>::getNi(void) const
{
return ni_;
}
template <typename T>
unsigned int A2AMatrixIo<T>::getNj(void) const
{
return nj_;
}
template <typename T>
size_t A2AMatrixIo<T>::getSize(void) const
{
return nt_*ni_*nj_*sizeof(T);
}
// file allocation /////////////////////////////////////////////////////////////
template <typename T>
template <typename MetadataType>
@ -171,7 +437,7 @@ void A2AMatrixIo<T>::initFile(const MetadataType &d, const unsigned int chunkSiz
}
// create the dataset
Hdf5Reader reader(filename_);
Hdf5Reader reader(filename_, false);
push(reader, dataname_);
auto &group = reader.getGroup();
@ -191,7 +457,7 @@ void A2AMatrixIo<T>::saveBlock(const T *data,
const unsigned int blockSizej)
{
#ifdef HAVE_HDF5
Hdf5Reader reader(filename_);
Hdf5Reader reader(filename_, false);
std::vector<hsize_t> count = {nt_, blockSizei, blockSizej},
offset = {0, static_cast<hsize_t>(i),
static_cast<hsize_t>(j)},
@ -226,6 +492,82 @@ void A2AMatrixIo<T>::saveBlock(const A2AMatrixSet<T> &m,
saveBlock(m.data() + offset, i, j, blockSizei, blockSizej);
}
template <typename T>
template <template <class> class Vec, typename VecT>
void A2AMatrixIo<T>::load(Vec<VecT> &v, double *tRead)
{
#ifdef HAVE_HDF5
Hdf5Reader reader(filename_);
std::vector<hsize_t> hdim;
H5NS::DataSet dataset;
H5NS::DataSpace dataspace;
H5NS::CompType datatype;
H5NS::DSetCreatPropList plist;
push(reader, dataname_);
auto &group = reader.getGroup();
dataset = group.openDataSet(HADRONS_A2AM_NAME);
datatype = dataset.getCompType();
dataspace = dataset.getSpace();
plist = dataset.getCreatePlist();
hdim.resize(dataspace.getSimpleExtentNdims());
dataspace.getSimpleExtentDims(hdim.data());
if ((nt_*ni_*nj_ != 0) and
((hdim[0] != nt_) or (hdim[1] != ni_) or (hdim[2] != nj_)))
{
HADRONS_ERROR(Size, "all-to-all matrix size mismatch (got "
+ std::to_string(hdim[0]) + "x" + std::to_string(hdim[1]) + "x"
+ std::to_string(hdim[2]) + ", expected "
+ std::to_string(nt_) + "x" + std::to_string(ni_) + "x"
+ std::to_string(nj_));
}
else if (ni_*nj_ == 0)
{
if (hdim[0] != nt_)
{
HADRONS_ERROR(Size, "all-to-all time size mismatch (got "
+ std::to_string(hdim[0]) + ", expected "
+ std::to_string(nt_) + ")");
}
ni_ = hdim[1];
nj_ = hdim[2];
}
A2AMatrix<T> buf(ni_, nj_);
std::vector<hsize_t> count = {1, static_cast<hsize_t>(ni_),
static_cast<hsize_t>(nj_)},
stride = {1, 1, 1},
block = {1, 1, 1},
memCount = {static_cast<hsize_t>(ni_),
static_cast<hsize_t>(nj_)};
H5NS::DataSpace memspace(memCount.size(), memCount.data());
std::cout << "Loading timeslice";
std::cout.flush();
*tRead = 0.;
for (unsigned int tp1 = nt_; tp1 > 0; --tp1)
{
unsigned int t = tp1 - 1;
std::vector<hsize_t> offset = {static_cast<hsize_t>(t), 0, 0};
if (t % 10 == 0)
{
std::cout << " " << t;
std::cout.flush();
}
dataspace.selectHyperslab(H5S_SELECT_SET, count.data(), offset.data(),
stride.data(), block.data());
if (tRead) *tRead -= usecond();
dataset.read(buf.data(), datatype, memspace, dataspace);
if (tRead) *tRead += usecond();
v[t] = buf.template cast<VecT>();
}
std::cout << std::endl;
#else
HADRONS_ERROR(Implementation, "all-to-all matrix I/O needs HDF5 library");
#endif
}
/******************************************************************************
* A2AMatrixBlockComputation template implementation *
******************************************************************************/

163
Hadrons/DiskVector.hpp
View File

@ -29,6 +29,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
#define Hadrons_DiskVector_hpp_
#include <Hadrons/Global.hpp>
#include <Hadrons/A2AMatrix.hpp>
#include <deque>
#include <sys/stat.h>
#include <ftw.h>
@ -59,14 +60,18 @@ public:
: master_(master), cmaster_(master), i_(i) {}
// operator=: somebody is trying to store a vector element
// write to disk and cache
// write to cache and tag as modified
T &operator=(const T &obj) const
{
auto &cache = *master_.cachePtr_;
auto &modified = *master_.modifiedPtr_;
auto &index = *master_.indexPtr_;
DV_DEBUG_MSG(&master_, "writing to " << i_);
master_.cacheInsert(i_, obj);
master_.save(master_.filename(i_), obj);
modified[index.at(i_)] = true;
return master_.cachePtr_->at(i_);
return cache[index.at(i_)];
}
// implicit cast to const object reference and redirection
@ -83,6 +88,7 @@ public:
public:
DiskVectorBase(const std::string dirname, const unsigned int size = 0,
const unsigned int cacheSize = 1, const bool clean = true);
DiskVectorBase(DiskVectorBase<T> &&v) = default;
virtual ~DiskVectorBase(void);
const T & operator[](const unsigned int i) const;
RwAccessHelper operator[](const unsigned int i);
@ -97,14 +103,17 @@ private:
void cacheInsert(const unsigned int i, const T &obj) const;
void clean(void);
private:
std::string dirname_;
unsigned int size_, cacheSize_;
double access_{0.}, hit_{0.};
bool clean_;
std::string dirname_;
unsigned int size_, cacheSize_;
double access_{0.}, hit_{0.};
bool clean_;
// using pointers to allow modifications when class is const
// semantic: const means data unmodified, but cache modification allowed
std::unique_ptr<std::map<unsigned int, T>> cachePtr_;
std::unique_ptr<std::deque<unsigned int>> loadsPtr_;
std::unique_ptr<std::vector<T>> cachePtr_;
std::unique_ptr<std::vector<bool>> modifiedPtr_;
std::unique_ptr<std::map<unsigned int, unsigned int>> indexPtr_;
std::unique_ptr<std::stack<unsigned int>> freePtr_;
std::unique_ptr<std::deque<unsigned int>> loadsPtr_;
};
/******************************************************************************
@ -135,7 +144,7 @@ private:
* Specialisation for Eigen matrices *
******************************************************************************/
template <typename T>
using EigenDiskVectorMat = Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic>;
using EigenDiskVectorMat = A2AMatrix<T>;
template <typename T>
class EigenDiskVector: public DiskVectorBase<EigenDiskVectorMat<T>>
@ -152,24 +161,31 @@ public:
private:
virtual void load(EigenDiskVectorMat<T> &obj, const std::string filename) const
{
std::ifstream f(filename, std::ios::binary);
std::vector<unsigned char> hash(SHA256_DIGEST_LENGTH);
Eigen::Index nRow, nCol;
size_t matSize;
double t;
std::ifstream f(filename, std::ios::binary);
uint32_t crc, check;
Eigen::Index nRow, nCol;
size_t matSize;
double tRead, tHash;
f.read(reinterpret_cast<char *>(hash.data()), hash.size()*sizeof(unsigned char));
f.read(reinterpret_cast<char *>(&nRow), sizeof(Eigen::Index));
f.read(reinterpret_cast<char *>(&nCol), sizeof(Eigen::Index));
f.read(reinterpret_cast<char *>(&crc), sizeof(crc));
f.read(reinterpret_cast<char *>(&nRow), sizeof(nRow));
f.read(reinterpret_cast<char *>(&nCol), sizeof(nCol));
obj.resize(nRow, nCol);
matSize = nRow*nCol*sizeof(T);
t = -usecond();
tRead = -usecond();
f.read(reinterpret_cast<char *>(obj.data()), matSize);
t += usecond();
DV_DEBUG_MSG(this, "Eigen read " << matSize/t*1.0e6/1024/1024 << " MB/s");
auto check = GridChecksum::sha256(obj.data(), matSize);
DV_DEBUG_MSG(this, "Eigen sha256 " << GridChecksum::sha256_string(check));
if (hash != check)
tRead += usecond();
tHash = -usecond();
#ifdef USE_IPP
check = GridChecksum::crc32c(obj.data(), matSize);
#else
check = GridChecksum::crc32(obj.data(), matSize);
#endif
tHash += usecond();
DV_DEBUG_MSG(this, "Eigen read " << tRead/1.0e6 << " sec " << matSize/tRead*1.0e6/1024/1024 << " MB/s");
DV_DEBUG_MSG(this, "Eigen crc32 " << std::hex << check << std::dec
<< " " << tHash/1.0e6 << " sec " << matSize/tHash*1.0e6/1024/1024 << " MB/s");
if (crc != check)
{
HADRONS_ERROR(Io, "checksum failed")
}
@ -177,24 +193,31 @@ private:
virtual void save(const std::string filename, const EigenDiskVectorMat<T> &obj) const
{
std::ofstream f(filename, std::ios::binary);
std::vector<unsigned char> hash(SHA256_DIGEST_LENGTH);
Eigen::Index nRow, nCol;
size_t matSize;
double t;
std::ofstream f(filename, std::ios::binary);
uint32_t crc;
Eigen::Index nRow, nCol;
size_t matSize;
double tWrite, tHash;
nRow = obj.rows();
nCol = obj.cols();
matSize = nRow*nCol*sizeof(T);
hash = GridChecksum::sha256(obj.data(), matSize);
DV_DEBUG_MSG(this, "Eigen sha256 " << GridChecksum::sha256_string(hash));
f.write(reinterpret_cast<char *>(hash.data()), hash.size()*sizeof(unsigned char));
f.write(reinterpret_cast<char *>(&nRow), sizeof(Eigen::Index));
f.write(reinterpret_cast<char *>(&nCol), sizeof(Eigen::Index));
t = -usecond();
tHash = -usecond();
#ifdef USE_IPP
crc = GridChecksum::crc32c(obj.data(), matSize);
#else
crc = GridChecksum::crc32(obj.data(), matSize);
#endif
tHash += usecond();
f.write(reinterpret_cast<char *>(&crc), sizeof(crc));
f.write(reinterpret_cast<char *>(&nRow), sizeof(nRow));
f.write(reinterpret_cast<char *>(&nCol), sizeof(nCol));
tWrite = -usecond();
f.write(reinterpret_cast<const char *>(obj.data()), matSize);
t += usecond();
DV_DEBUG_MSG(this, "Eigen write " << matSize/t*1.0e6/1024/1024 << " MB/s");
tWrite += usecond();
DV_DEBUG_MSG(this, "Eigen write " << tWrite/1.0e6 << " sec " << matSize/tWrite*1.0e6/1024/1024 << " MB/s");
DV_DEBUG_MSG(this, "Eigen crc32 " << std::hex << crc << std::dec
<< " " << tHash/1.0e6 << " sec " << matSize/tHash*1.0e6/1024/1024 << " MB/s");
}
};
@ -207,7 +230,10 @@ DiskVectorBase<T>::DiskVectorBase(const std::string dirname,
const unsigned int cacheSize,
const bool clean)
: dirname_(dirname), size_(size), cacheSize_(cacheSize), clean_(clean)
, cachePtr_(new std::map<unsigned int, T>())
, cachePtr_(new std::vector<T>(size))
, modifiedPtr_(new std::vector<bool>(size, false))
, indexPtr_(new std::map<unsigned int, unsigned int>())
, freePtr_(new std::stack<unsigned int>)
, loadsPtr_(new std::deque<unsigned int>())
{
struct stat s;
@ -217,6 +243,10 @@ DiskVectorBase<T>::DiskVectorBase(const std::string dirname,
HADRONS_ERROR(Io, "directory '" + dirname + "' already exists")
}
mkdir(dirname);
for (unsigned int i = 0; i < cacheSize_; ++i)
{
freePtr_->push(i);
}
}
template <typename T>
@ -231,8 +261,10 @@ DiskVectorBase<T>::~DiskVectorBase(void)
template <typename T>
const T & DiskVectorBase<T>::operator[](const unsigned int i) const
{
auto &cache = *cachePtr_;
auto &loads = *loadsPtr_;
auto &cache = *cachePtr_;
auto &index = *indexPtr_;
auto &freeInd = *freePtr_;
auto &loads = *loadsPtr_;
DV_DEBUG_MSG(this, "accessing " << i << " (RO)");
@ -241,7 +273,7 @@ const T & DiskVectorBase<T>::operator[](const unsigned int i) const
HADRONS_ERROR(Size, "index out of range");
}
const_cast<double &>(access_)++;
if (cache.find(i) == cache.end())
if (index.find(i) == index.end())
{
// cache miss
DV_DEBUG_MSG(this, "cache miss");
@ -268,7 +300,7 @@ const T & DiskVectorBase<T>::operator[](const unsigned int i) const
DV_DEBUG_MSG(this, "in cache: " << msg);
#endif
return cache.at(i);
return cache[index.at(i)];
}
template <typename T>
@ -306,13 +338,24 @@ std::string DiskVectorBase<T>::filename(const unsigned int i) const
template <typename T>
void DiskVectorBase<T>::evict(void) const
{
auto &cache = *cachePtr_;
auto &loads = *loadsPtr_;
auto &cache = *cachePtr_;
auto &modified = *modifiedPtr_;
auto &index = *indexPtr_;
auto &freeInd = *freePtr_;
auto &loads = *loadsPtr_;
if (cache.size() >= cacheSize_)
if (index.size() >= cacheSize_)
{
DV_DEBUG_MSG(this, "evicting " << loads.front());
cache.erase(loads.front());
unsigned int i = loads.front();
DV_DEBUG_MSG(this, "evicting " << i);
if (modified[index.at(i)])
{
DV_DEBUG_MSG(this, "element " << i << " modified, saving to disk");
save(filename(i), cache[index.at(i)]);
}
freeInd.push(index.at(i));
index.erase(i);
loads.pop_front();
}
}
@ -320,30 +363,44 @@ void DiskVectorBase<T>::evict(void) const
template <typename T>
void DiskVectorBase<T>::fetch(const unsigned int i) const
{
auto &cache = *cachePtr_;
auto &loads = *loadsPtr_;
auto &cache = *cachePtr_;
auto &modified = *modifiedPtr_;
auto &index = *indexPtr_;
auto &freeInd = *freePtr_;
auto &loads = *loadsPtr_;
struct stat s;
DV_DEBUG_MSG(this, "loading " << i << " from disk");
evict();
if(stat(filename(i).c_str(), &s) != 0)
{
HADRONS_ERROR(Io, "disk vector element " + std::to_string(i) + " uninitialised");
}
load(cache[i], filename(i));
index[i] = freeInd.top();
freeInd.pop();
load(cache[index.at(i)], filename(i));
loads.push_back(i);
modified[index.at(i)] = false;
}
template <typename T>
void DiskVectorBase<T>::cacheInsert(const unsigned int i, const T &obj) const
{
auto &cache = *cachePtr_;
auto &loads = *loadsPtr_;
auto &cache = *cachePtr_;
auto &modified = *modifiedPtr_;
auto &index = *indexPtr_;
auto &freeInd = *freePtr_;
auto &loads = *loadsPtr_;
evict();
cache[i] = obj;
index[i] = freeInd.top();
freeInd.pop();
cache[index.at(i)] = obj;
loads.push_back(i);
modified[index.at(i)] = false;
#ifdef DV_DEBUG
std::string msg;

8
Hadrons/Global.cc
View File

@ -166,7 +166,13 @@ std::string Hadrons::dirname(const std::string &s)
void Hadrons::makeFileDir(const std::string filename, GridBase *g)
{
if (g->IsBoss())
bool doIt = true;
if (g)
{
doIt = g->IsBoss();
}
if (doIt)
{
std::string dir = dirname(filename);
int status = mkdir(dir);

21
Hadrons/Global.hpp
View File

@ -32,6 +32,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
#include <set>
#include <stack>
#include <regex>
#include <Grid/Grid.h>
#include <cxxabi.h>
@ -217,15 +218,15 @@ typedef XmlReader ResultReader;
typedef XmlWriter ResultWriter;
#endif
#define RESULT_FILE_NAME(name) \
name + "." + std::to_string(vm().getTrajectory()) + "." + resultFileExt
#define RESULT_FILE_NAME(name, traj) \
name + "." + std::to_string(traj) + "." + resultFileExt
// recursive mkdir
#define MAX_PATH_LENGTH 512u
int mkdir(const std::string dirName);
std::string basename(const std::string &s);
std::string dirname(const std::string &s);
void makeFileDir(const std::string filename, GridBase *g);
void makeFileDir(const std::string filename, GridBase *g = nullptr);
// default Schur convention
#ifndef HADRONS_DEFAULT_SCHUR
@ -248,6 +249,20 @@ void makeFileDir(const std::string filename, GridBase *g);
// pretty print time profile
void printTimeProfile(const std::map<std::string, GridTime> &timing, GridTime total);
// token replacement utility
template <typename T>
void tokenReplace(std::string &str, const std::string token,
const T &x, const std::string mark = "@")
{
std::string fullToken = mark + token + mark;
auto pos = str.find(fullToken);
if (pos != std::string::npos)
{
str.replace(pos, fullToken.size(), std::to_string(x));
}
}
END_HADRONS_NAMESPACE
#include <Hadrons/Exceptions.hpp>

9
Hadrons/Makefile.am
View File

@ -5,17 +5,17 @@ lib_LIBRARIES = libHadrons.a
include modules.inc
libHadrons_a_SOURCES = \
$(modules_cc) \
Application.cc \
Environment.cc \
Exceptions.cc \
Global.cc \
Module.cc \
TimerArray.cc \
VirtualMachine.cc
VirtualMachine.cc \
$(modules_cc)
libHadrons_adir = $(includedir)/Hadrons
nobase_libHadrons_a_HEADERS = \
$(modules_hpp) \
A2AVectors.hpp \
A2AMatrix.hpp \
Application.hpp \
@ -33,4 +33,5 @@ nobase_libHadrons_a_HEADERS = \
ModuleFactory.hpp \
Solver.hpp \
TimerArray.hpp \
VirtualMachine.hpp
VirtualMachine.hpp \
$(modules_hpp)

2
Hadrons/Module.hpp
View File

@ -144,7 +144,7 @@ if (env().getGrid()->IsBoss() and !ioStem.empty())\
{\
makeFileDir(ioStem, env().getGrid());\
{\
ResultWriter _writer(RESULT_FILE_NAME(ioStem));\
ResultWriter _writer(RESULT_FILE_NAME(ioStem, vm().getTrajectory()));\
write(_writer, name, result);\
}\
}

8
Hadrons/Modules/MContraction/A2AAslashField.hpp
View File

@ -33,10 +33,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
#include <Hadrons/ModuleFactory.hpp>
#include <Hadrons/A2AMatrix.hpp>
#ifndef ASF_IO_TYPE
#define ASF_IO_TYPE ComplexF
#endif
BEGIN_HADRONS_NAMESPACE
/******************************************************************************
@ -113,7 +109,7 @@ public:
typedef A2AMatrixBlockComputation<Complex,
FermionField,
A2AAslashFieldMetadata,
ASF_IO_TYPE> Computation;
HADRONS_A2AM_IO_TYPE> Computation;
typedef AslashFieldKernel<Complex, FImpl> Kernel;
public:
// constructor
@ -196,7 +192,7 @@ void TA2AAslashField<FImpl, PhotonImpl>::execute(void)
LOG(Message) << " " << name << std::endl;
}
LOG(Message) << "A-slash field size: " << nt << "*" << N_i << "*" << N_j
<< " (filesize " << sizeString(nt*N_i*N_j*sizeof(ASF_IO_TYPE))
<< " (filesize " << sizeString(nt*N_i*N_j*sizeof(HADRONS_A2AM_IO_TYPE))
<< "/EM field)" << std::endl;
// preparing "B" complexified fields

8
Hadrons/Modules/MContraction/A2AMesonField.hpp
View File

@ -35,10 +35,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
#include <Hadrons/ModuleFactory.hpp>
#include <Hadrons/A2AMatrix.hpp>
#ifndef MF_IO_TYPE
#define MF_IO_TYPE ComplexF
#endif
BEGIN_HADRONS_NAMESPACE
/******************************************************************************
@ -118,7 +114,7 @@ public:
typedef A2AMatrixBlockComputation<Complex,
FermionField,
A2AMesonFieldMetadata,
MF_IO_TYPE> Computation;
HADRONS_A2AM_IO_TYPE> Computation;
typedef MesonFieldKernel<Complex, FImpl> Kernel;
public:
// constructor
@ -248,7 +244,7 @@ void TA2AMesonField<FImpl>::execute(void)
LOG(Message) << " " << g << std::endl;
}
LOG(Message) << "Meson field size: " << nt << "*" << N_i << "*" << N_j
<< " (filesize " << sizeString(nt*N_i*N_j*sizeof(MF_IO_TYPE))
<< " (filesize " << sizeString(nt*N_i*N_j*sizeof(HADRONS_A2AM_IO_TYPE))
<< "/momentum/bilinear)" << std::endl;
auto &ph = envGet(std::vector<ComplexField>, momphName_);

2
Hadrons/Modules/MScalar/ChargedProp.cc
View File

@ -146,7 +146,7 @@ void TChargedProp::execute(void)
std::vector<int> siteCoor;
LOG(Message) << "Saving momentum-projected propagator to '"
<< RESULT_FILE_NAME(par().output) << "'..."
<< RESULT_FILE_NAME(par().output, vm().getTrajectory()) << "'..."
<< std::endl;
result.projection.resize(par().outputMom.size());
result.lattice_size = env().getGrid()->_fdimensions;

2
Hadrons/Modules/MScalar/ScalarVP.cc
View File

@ -462,7 +462,7 @@ void TScalarVP::execute(void)
if (!par().output.empty())
{
LOG(Message) << "Saving momentum-projected HVP to '"
<< RESULT_FILE_NAME(par().output) << "'..."
<< RESULT_FILE_NAME(par().output, vm().getTrajectory()) << "'..."
<< std::endl;
saveResult(par().output, "HVP", outputData);
}

2
Hadrons/Modules/MScalar/VPCounterTerms.cc
View File

@ -239,7 +239,7 @@ void TVPCounterTerms::execute(void)
if (!par().output.empty())
{
LOG(Message) << "Saving momentum-projected correlators to '"
<< RESULT_FILE_NAME(par().output) << "'..."
<< RESULT_FILE_NAME(par().output, vm().getTrajectory()) << "'..."
<< std::endl;
saveResult(par().output, "scalar_loops", outputData);
}

454
Hadrons/Utilities/Contractor.cc Normal file
View File

@ -0,0 +1,454 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: Hadrons/Utilities/Contractor.cc
Copyright (C) 2015-2018
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Hadrons/Global.hpp>
#include <Hadrons/A2AMatrix.hpp>
#include <Hadrons/DiskVector.hpp>
#include <Hadrons/TimerArray.hpp>
using namespace Grid;
using namespace QCD;
using namespace Hadrons;
#define TIME_MOD(t) (((t) + par.global.nt) % par.global.nt)
namespace Contractor
{
class TrajRange: Serializable
{
public:
GRID_SERIALIZABLE_CLASS_MEMBERS(TrajRange,
unsigned int, start,
unsigned int, end,
unsigned int, step);
};
class GlobalPar: Serializable
{
public:
GRID_SERIALIZABLE_CLASS_MEMBERS(GlobalPar,
TrajRange, trajCounter,
unsigned int, nt,
std::string, diskVectorDir,
std::string, output);
};
class A2AMatrixPar: Serializable
{
public:
GRID_SERIALIZABLE_CLASS_MEMBERS(A2AMatrixPar,
std::string, file,
std::string, dataset,
unsigned int, cacheSize,
std::string, name);
};
class ProductPar: Serializable
{
public:
GRID_SERIALIZABLE_CLASS_MEMBERS(ProductPar,
std::string, terms,
std::vector<std::string>, times,
std::string, translations,
bool, translationAverage);
};
class CorrelatorResult: Serializable
{
public:
GRID_SERIALIZABLE_CLASS_MEMBERS(CorrelatorResult,
std::vector<Contractor::A2AMatrixPar>, a2aMatrix,
ProductPar, contraction,
std::vector<unsigned int>, times,
std::vector<ComplexD>, correlator);
};
}
struct ContractorPar
{
Contractor::GlobalPar global;
std::vector<Contractor::A2AMatrixPar> a2aMatrix;
std::vector<Contractor::ProductPar> product;
};
void makeTimeSeq(std::vector<std::vector<unsigned int>> &timeSeq,
const std::vector<std::set<unsigned int>> &times,
std::vector<unsigned int> &current,
const unsigned int depth)
{
if (depth > 0)
{
for (auto t: times[times.size() - depth])
{
current[times.size() - depth] = t;
makeTimeSeq(timeSeq, times, current, depth - 1);
}
}
else
{
timeSeq.push_back(current);
}
}
void makeTimeSeq(std::vector<std::vector<unsigned int>> &timeSeq,
const std::vector<std::set<unsigned int>> &times)
{
std::vector<unsigned int> current(times.size());
makeTimeSeq(timeSeq, times, current, times.size());
}
void saveCorrelator(const Contractor::CorrelatorResult &result, const std::string dir,
const unsigned int dt, const unsigned int traj)
{
std::string fileStem = "", filename;
std::vector<std::string> terms = strToVec<std::string>(result.contraction.terms);
for (unsigned int i = 0; i < terms.size() - 1; i++)
{
fileStem += terms[i] + "_" + std::to_string(result.times[i]) + "_";
}
fileStem += terms.back();
if (!result.contraction.translationAverage)
{
fileStem += "_dt_" + std::to_string(dt);
}
filename = dir + "/" + RESULT_FILE_NAME(fileStem, traj);
std::cout << "Saving correlator to '" << filename << "'" << std::endl;
makeFileDir(dir);
ResultWriter writer(filename);
write(writer, fileStem, result);
}
std::set<unsigned int> parseTimeRange(const std::string str, const unsigned int nt)
{
std::regex rex("([0-9]+)|(([0-9]+)\\.\\.([0-9]+))");
std::smatch sm;
std::vector<std::string> rstr = strToVec<std::string>(str);
std::set<unsigned int> tSet;
for (auto &s: rstr)
{
std::regex_match(s, sm, rex);
if (sm[1].matched)
{
unsigned int t;
t = std::stoi(sm[1].str());
if (t >= nt)
{
HADRONS_ERROR(Range, "time out of range (from expression '" + str + "')");
}
tSet.insert(t);
}
else if (sm[2].matched)
{
unsigned int ta, tb;
ta = std::stoi(sm[3].str());
tb = std::stoi(sm[4].str());
if ((ta >= nt) or (tb >= nt))
{
HADRONS_ERROR(Range, "time out of range (from expression '" + str + "')");
}
for (unsigned int ti = ta; ti <= tb; ++ti)
{
tSet.insert(ti);
}
}
}
return tSet;
}
struct Sec
{
Sec(const double usec)
{
seconds = usec/1.0e6;
}
double seconds;
};
inline std::ostream & operator<< (std::ostream& s, const Sec &&sec)
{
s << std::setw(10) << sec.seconds << " sec";
return s;
}
struct Flops
{
Flops(const double flops, const double fusec)
{
gFlopsPerSec = flops/fusec/1.0e3;
}
double gFlopsPerSec;
};
inline std::ostream & operator<< (std::ostream& s, const Flops &&f)
{
s << std::setw(10) << f.gFlopsPerSec << " GFlop/s";
return s;
}
struct Bytes
{
Bytes(const double bytes, const double busec)
{
gBytesPerSec = bytes/busec*1.0e6/1024/1024/1024;
}
double gBytesPerSec;
};
inline std::ostream & operator<< (std::ostream& s, const Bytes &&b)
{
s << std::setw(10) << b.gBytesPerSec << " GB/s";
return s;
}
int main(int argc, char* argv[])
{
// parse command line
std::string parFilename;
if (argc != 2)
{
std::cerr << "usage: " << argv[0] << " <parameter file>";
std::cerr << std::endl;
return EXIT_FAILURE;
}
parFilename = argv[1];
// parse parameter file
ContractorPar par;
unsigned int nMat, nCont;
XmlReader reader(parFilename);
read(reader, "global", par.global);
read(reader, "a2aMatrix", par.a2aMatrix);
read(reader, "product", par.product);
nMat = par.a2aMatrix.size();
nCont = par.product.size();
// create diskvectors
std::map<std::string, EigenDiskVector<ComplexD>> a2aMat;
unsigned int cacheSize;
for (auto &p: par.a2aMatrix)
{
std::string dirName = par.global.diskVectorDir + "/" + p.name;
a2aMat.emplace(p.name, EigenDiskVector<ComplexD>(dirName, par.global.nt, p.cacheSize));
}
// trajectory loop
for (unsigned int traj = par.global.trajCounter.start;
traj < par.global.trajCounter.end; traj += par.global.trajCounter.step)
{
std::cout << ":::::::: Trajectory " << traj << std::endl;
// load data
for (auto &p: par.a2aMatrix)
{
std::string filename = p.file;
double t, size;
tokenReplace(filename, "traj", traj);
std::cout << "======== Loading '" << filename << "'" << std::endl;
A2AMatrixIo<HADRONS_A2AM_IO_TYPE> a2aIo(filename, p.dataset, par.global.nt);
a2aIo.load(a2aMat.at(p.name), &t);
std::cout << "Read " << a2aIo.getSize() << " bytes in " << t/1.0e6
<< " sec, " << a2aIo.getSize()/t*1.0e6/1024/1024 << " MB/s" << std::endl;
}
// contract
EigenDiskVector<ComplexD>::Matrix buf;
for (auto &p: par.product)
{
std::vector<std::string> term = strToVec<std::string>(p.terms);
std::vector<std::set<unsigned int>> times;
std::vector<std::vector<unsigned int>> timeSeq;
std::set<unsigned int> translations;
std::vector<A2AMatrixTr<ComplexD>> lastTerm(par.global.nt);
A2AMatrix<ComplexD> prod, buf, tmp;
TimerArray tAr;
double fusec, busec, flops, bytes, tusec;
Contractor::CorrelatorResult result;
tAr.startTimer("Total");
std::cout << "======== Contraction tr(";
for (unsigned int g = 0; g < term.size(); ++g)
{
std::cout << term[g] << ((g == term.size() - 1) ? ')' : '*');
}
std::cout << std::endl;
if (term.size() != p.times.size() + 1)
{
HADRONS_ERROR(Size, "number of terms (" + std::to_string(term.size())
+ ") different from number of times ("
+ std::to_string(p.times.size() + 1) + ")");
}
for (auto &s: p.times)
{
times.push_back(parseTimeRange(s, par.global.nt));
}
for (auto &m: par.a2aMatrix)
{
if (std::find(result.a2aMatrix.begin(), result.a2aMatrix.end(), m) == result.a2aMatrix.end())
{
result.a2aMatrix.push_back(m);
tokenReplace(result.a2aMatrix.back().file, "traj", traj);
}
}
result.contraction = p;
result.correlator.resize(par.global.nt, 0.);
translations = parseTimeRange(p.translations, par.global.nt);
makeTimeSeq(timeSeq, times);
std::cout << timeSeq.size()*translations.size()*(term.size() - 2) << " A*B, "
<< timeSeq.size()*translations.size()*par.global.nt << " tr(A*B)"
<< std::endl;
std::cout << "* Caching transposed last term" << std::endl;
for (unsigned int t = 0; t < par.global.nt; ++t)
{
tAr.startTimer("Disk vector overhead");
const A2AMatrix<ComplexD> &ref = a2aMat.at(term.back())[t];
tAr.stopTimer("Disk vector overhead");
tAr.startTimer("Transpose caching");
lastTerm[t].resize(ref.rows(), ref.cols());
parallel_for (unsigned int j = 0; j < ref.cols(); ++j)
for (unsigned int i = 0; i < ref.rows(); ++i)
{
lastTerm[t](i, j) = ref(i, j);
}
tAr.stopTimer("Transpose caching");
}
bytes = par.global.nt*lastTerm[0].rows()*lastTerm[0].cols()*sizeof(ComplexD);
std::cout << Sec(tAr.getDTimer("Transpose caching")) << " "
<< Bytes(bytes, tAr.getDTimer("Transpose caching")) << std::endl;
for (unsigned int i = 0; i < timeSeq.size(); ++i)
{
unsigned int dti = 0;
auto &t = timeSeq[i];
result.times = t;
for (unsigned int tLast = 0; tLast < par.global.nt; ++tLast)
{
result.correlator[tLast] = 0.;
}
for (auto &dt: translations)
{
std::cout << "* Step " << i*translations.size() + dti + 1
<< "/" << timeSeq.size()*translations.size()
<< " -- positions= " << t << ", dt= " << dt << std::endl;
if (term.size() > 2)
{
std::cout << std::setw(8) << "products";
}
flops = 0.;
bytes = 0.;
fusec = tAr.getDTimer("A*B algebra");
busec = tAr.getDTimer("A*B total");
tAr.startTimer("Linear algebra");
tAr.startTimer("Disk vector overhead");
prod = a2aMat.at(term[0])[TIME_MOD(t[0] + dt)];
tAr.stopTimer("Disk vector overhead");
for (unsigned int j = 1; j < term.size() - 1; ++j)
{
tAr.startTimer("Disk vector overhead");
const A2AMatrix<ComplexD> &ref = a2aMat.at(term[j])[TIME_MOD(t[j] + dt)];
tAr.stopTimer("Disk vector overhead");
tAr.startTimer("A*B total");
tAr.startTimer("A*B algebra");
A2AContraction::mul(tmp, prod, ref);
tAr.stopTimer("A*B algebra");
flops += A2AContraction::mulFlops(prod, ref);
prod = tmp;
tAr.stopTimer("A*B total");
bytes += 3.*tmp.rows()*tmp.cols()*sizeof(ComplexD);
}
if (term.size() > 2)
{
std::cout << Sec(tAr.getDTimer("A*B total") - busec) << " "
<< Flops(flops, tAr.getDTimer("A*B algebra") - fusec) << " "
<< Bytes(bytes, tAr.getDTimer("A*B total") - busec) << std::endl;
}
std::cout << std::setw(8) << "traces";
flops = 0.;
bytes = 0.;
fusec = tAr.getDTimer("tr(A*B)");
busec = tAr.getDTimer("tr(A*B)");
for (unsigned int tLast = 0; tLast < par.global.nt; ++tLast)
{
tAr.startTimer("tr(A*B)");
A2AContraction::accTrMul(result.correlator[TIME_MOD(tLast - dt)], prod, lastTerm[tLast]);
tAr.stopTimer("tr(A*B)");
flops += A2AContraction::accTrMulFlops(prod, lastTerm[tLast]);
bytes += 2.*prod.rows()*prod.cols()*sizeof(ComplexD);
}
tAr.stopTimer("Linear algebra");
std::cout << Sec(tAr.getDTimer("tr(A*B)") - busec) << " "
<< Flops(flops, tAr.getDTimer("tr(A*B)") - fusec) << " "
<< Bytes(bytes, tAr.getDTimer("tr(A*B)") - busec) << std::endl;
if (!p.translationAverage)
{
saveCorrelator(result, par.global.output, dt, traj);
for (unsigned int tLast = 0; tLast < par.global.nt; ++tLast)
{
result.correlator[tLast] = 0.;
}
}
dti++;
}
if (p.translationAverage)
{
for (unsigned int tLast = 0; tLast < par.global.nt; ++tLast)
{
result.correlator[tLast] /= translations.size();
}
saveCorrelator(result, par.global.output, 0, traj);
}
}
tAr.stopTimer("Total");
printTimeProfile(tAr.getTimings(), tAr.getTimer("Total"));
}
}
return EXIT_SUCCESS;
}

12
Hadrons/Utilities/Contractor.hpp Normal file
View File

@ -0,0 +1,12 @@
#ifndef Hadrons_Contractor_hpp_
#define Hadrons_Contractor_hpp_
#include <Hadrons/Global.hpp>
BEGIN_HADRONS_NAMESPACE
END_HADRONS_NAMESPACE
#endif // Hadrons_Contractor_hpp_

434
Hadrons/Utilities/ContractorBenchmark.cc Normal file
View File

@ -0,0 +1,434 @@
#include <Hadrons/Global.hpp>
#include <Hadrons/A2AMatrix.hpp>
#ifdef USE_MKL
#include "mkl.h"
#include "mkl_cblas.h"
#endif
using namespace Grid;
using namespace Hadrons;
#ifdef GRID_COMMS_MPI3
#define GET_RANK(rank, nMpi) \
MPI_Comm_size(MPI_COMM_WORLD, &(nMpi));\
MPI_Comm_rank(MPI_COMM_WORLD, &(rank))
#define BARRIER() MPI_Barrier(MPI_COMM_WORLD)
#define INIT() MPI_Init(NULL, NULL)
#define FINALIZE() MPI_Finalize()
#else
#define GET_RANK(rank, nMpi) (nMpi) = 1; (rank) = 0
#define BARRIER()
#define INIT()
#define FINALIZE()
#endif
template <typename Function, typename MatLeft, typename MatRight>
inline void trBenchmark(const std::string name, const MatLeft &left,
const MatRight &right, const ComplexD ref, Function fn)
{
double t, flops, bytes, n = left[0].rows()*left[0].cols();
unsigned int nMat = left.size();
int nMpi, rank;
ComplexD buf;
t = 0.;
GET_RANK(rank, nMpi);
t = -usecond();
BARRIER();
for (unsigned int i = rank*nMat/nMpi; i < (rank+1)*nMat/nMpi; ++i)
{
fn(buf, left[i], right[i]);
}
BARRIER();
t += usecond();
flops = nMat*(6.*n + 2.*(n - 1.));
bytes = nMat*(2.*n*sizeof(ComplexD));
if (rank == 0)
{
std::cout << std::setw(34) << name << ": diff= "
<< std::setw(12) << std::norm(buf-ref)
<< std::setw(10) << t/1.0e6 << " sec "
<< std::setw(10) << flops/t/1.0e3 << " GFlop/s "
<< std::setw(10) << bytes/t*1.0e6/1024/1024/1024 << " GB/s "
<< std::endl;
}
::sleep(1);
}
template <typename Function, typename MatV, typename Mat>
inline void mulBenchmark(const std::string name, const MatV &left,
const MatV &right, const Mat &ref, Function fn)
{
double t, flops, bytes;
double nr = left[0].rows(), nc = left[0].cols(), n = nr*nc;
unsigned int nMat = left.size();
int nMpi, rank;
Mat buf(left[0].rows(), left[0].rows());
t = 0.;
GET_RANK(rank, nMpi);
t = -usecond();
BARRIER();
for (unsigned int i = rank*nMat/nMpi; i < (rank+1)*nMat/nMpi; ++i)
{
fn(buf, left[i], right[i]);
}
BARRIER();
t += usecond();
flops = nMat*(nr*nr*(6.*nc + 2.*(nc - 1.)));
bytes = nMat*(2*nc*nr*sizeof(ComplexD));
if (rank == 0)
{
std::cout << std::setw(34) << name << ": diff= "
<< std::setw(12) << (buf-ref).squaredNorm()
<< std::setw(10) << t/1.0e6 << " sec "
<< std::setw(10) << flops/t/1.0e3 << " GFlop/s "
<< std::setw(10) << bytes/t*1.0e6/1024/1024/1024 << " GB/s "
<< std::endl;
}
::sleep(1);
}
#ifdef USE_MKL
template <typename MatLeft, typename MatRight>
static inline void zdotuRow(ComplexD &res, const unsigned int aRow,
const MatLeft &a, const MatRight &b)
{
const ComplexD *aPt, *bPt;
unsigned int aInc, bInc;
if (MatLeft::Options == Eigen::RowMajor)
{
aPt = a.data() + aRow*a.cols();
aInc = 1;
}
else if (MatLeft::Options == Eigen::ColMajor)
{
aPt = a.data() + aRow;
aInc = a.rows();
}
if (MatRight::Options == Eigen::RowMajor)
{
bPt = b.data() + aRow;
bInc = b.cols();
}
else if (MatRight::Options == Eigen::ColMajor)
{
bPt = b.data() + aRow*b.rows();
bInc = 1;
}
cblas_zdotu_sub(a.cols(), aPt, aInc, bPt, bInc, &res);
}
template <typename MatLeft, typename MatRight>
static inline void zdotuCol(ComplexD &res, const unsigned int aCol,
const MatLeft &a, const MatRight &b)
{
const ComplexD *aPt, *bPt;
unsigned int aInc, bInc;
if (MatLeft::Options == Eigen::RowMajor)
{
aPt = a.data() + aCol;
aInc = a.cols();
}
else if (MatLeft::Options == Eigen::ColMajor)
{
aPt = a.data() + aCol*a.rows();
aInc = 1;
}
if (MatRight::Options == Eigen::RowMajor)
{
bPt = b.data() + aCol*b.cols();
bInc = 1;
}
else if (MatRight::Options == Eigen::ColMajor)
{
bPt = b.data() + aCol;
bInc = b.rows();
}
cblas_zdotu_sub(a.rows(), aPt, aInc, bPt, bInc, &res);
}
#endif
template <typename MatLeft, typename MatRight>
void fullTrBenchmark(const unsigned int ni, const unsigned int nj, const unsigned int nMat)
{
std::vector<MatLeft> left;
std::vector<MatRight> right;
MatRight buf;
ComplexD ref;
int rank, nMpi;
left.resize(nMat, MatLeft::Random(ni, nj));
right.resize(nMat, MatRight::Random(nj, ni));
GET_RANK(rank, nMpi);
if (rank == 0)
{
std::cout << "==== tr(A*B) benchmarks" << std::endl;
std::cout << "A matrices use ";
if (MatLeft::Options == Eigen::RowMajor)
{
std::cout << "row-major ordering" << std::endl;
}
else if (MatLeft::Options == Eigen::ColMajor)
{
std::cout << "col-major ordering" << std::endl;
}
std::cout << "B matrices use ";
if (MatRight::Options == Eigen::RowMajor)
{
std::cout << "row-major ordering" << std::endl;
}
else if (MatRight::Options == Eigen::ColMajor)
{
std::cout << "col-major ordering" << std::endl;
}
std::cout << std::endl;
}
BARRIER();
ref = (left.back()*right.back()).trace();
trBenchmark("Hadrons A2AContraction::accTrMul", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
res = 0.;
A2AContraction::accTrMul(res, a, b);
});
trBenchmark("Naive loop rows first", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
auto nr = a.rows(), nc = a.cols();
res = 0.;
parallel_for (unsigned int i = 0; i < nr; ++i)
{
ComplexD tmp = 0.;
for (unsigned int j = 0; j < nc; ++j)
{
tmp += a(i, j)*b(j, i);
}
parallel_critical
{
res += tmp;
}
}
});
trBenchmark("Naive loop cols first", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
auto nr = a.rows(), nc = a.cols();
res = 0.;
parallel_for (unsigned int j = 0; j < nc; ++j)
{
ComplexD tmp = 0.;
for (unsigned int i = 0; i < nr; ++i)
{
tmp += a(i, j)*b(j, i);
}
parallel_critical
{
res += tmp;
}
}
});
trBenchmark("Eigen tr(A*B)", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
res = (a*b).trace();
});
trBenchmark("Eigen row-wise dot", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
res = 0.;
parallel_for (unsigned int r = 0; r < a.rows(); ++r)
{
ComplexD tmp;
tmp = a.row(r).conjugate().dot(b.col(r));
parallel_critical
{
res += tmp;
}
}
});
trBenchmark("Eigen col-wise dot", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
res = 0.;
parallel_for (unsigned int c = 0; c < a.cols(); ++c)
{
ComplexD tmp;
tmp = a.col(c).conjugate().dot(b.row(c));
parallel_critical
{
res += tmp;
}
}
});
trBenchmark("Eigen Hadamard", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
res = a.cwiseProduct(b.transpose()).sum();
});
#ifdef USE_MKL
trBenchmark("MKL row-wise zdotu", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
res = 0.;
parallel_for (unsigned int r = 0; r < a.rows(); ++r)
{
ComplexD tmp;
zdotuRow(tmp, r, a, b);
parallel_critical
{
res += tmp;
}
}
});
trBenchmark("MKL col-wise zdotu", left, right, ref,
[](ComplexD &res, const MatLeft &a, const MatRight &b)
{
res = 0.;
parallel_for (unsigned int c = 0; c < a.cols(); ++c)
{
ComplexD tmp;
zdotuCol(tmp, c, a, b);
parallel_critical
{
res += tmp;
}
}
});
#endif
BARRIER();
if (rank == 0)
{
std::cout << std::endl;
}
}
template <typename Mat>
void fullMulBenchmark(const unsigned int ni, const unsigned int nj, const unsigned int nMat)
{
std::vector<Mat> left, right;
Mat ref;
int rank, nMpi;
left.resize(nMat, Mat::Random(ni, nj));
right.resize(nMat, Mat::Random(nj, ni));
GET_RANK(rank, nMpi);
if (rank == 0)
{
std::cout << "==== A*B benchmarks" << std::endl;
std::cout << "all matrices use ";
if (Mat::Options == Eigen::RowMajor)
{
std::cout << "row-major ordering" << std::endl;
}
else if (Mat::Options == Eigen::ColMajor)
{
std::cout << "col-major ordering" << std::endl;
}
std::cout << std::endl;
}
BARRIER();
ref = left.back()*right.back();
mulBenchmark("Hadrons A2AContraction::mul", left, right, ref,
[](Mat &res, const Mat &a, const Mat &b)
{
A2AContraction::mul(res, a, b);
});
mulBenchmark("Eigen A*B", left, right, ref,
[](Mat &res, const Mat &a, const Mat &b)
{
res = a*b;
});
#ifdef USE_MKL
mulBenchmark("MKL A*B", left, right, ref,
[](Mat &res, const Mat &a, const Mat &b)
{
const ComplexD one(1., 0.), zero(0., 0.);
if (Mat::Options == Eigen::RowMajor)
{
cblas_zgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
a.cols(), &one, a.data(), a.cols(), b.data(), b.cols(), &zero,
res.data(), res.cols());
}
else if (Mat::Options == Eigen::ColMajor)
{
cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
a.cols(), &one, a.data(), a.rows(), b.data(), b.rows(), &zero,
res.data(), res.rows());
}
});
#endif
BARRIER();
if (rank == 0)
{
std::cout << std::endl;
}
}
int main(int argc, char *argv[])
{
// parse command line
Eigen::Index ni, nj, nMat;
int nMpi, rank;
if (argc != 4)
{
std::cerr << "usage: " << argv[0] << " <Ni> <Nj> <#matrices>";
std::cerr << std::endl;
return EXIT_FAILURE;
}
ni = std::stoi(argv[1]);
nj = std::stoi(argv[2]);
nMat = std::stoi(argv[3]);
INIT();
GET_RANK(rank, nMpi);
if (rank == 0)
{
std::cout << "\n*** ALL-TO-ALL MATRIX CONTRACTION BENCHMARK ***\n" << std::endl;
std::cout << nMat << " couples of " << ni << "x" << nj << " matrices\n" << std::endl;
std::cout << nMpi << " MPI processes" << std::endl;
#ifdef GRID_OMP
#pragma omp parallel
{
#pragma omp single
std::cout << omp_get_num_threads() << " threads\n" << std::endl;
}
#else
std::cout << "Single-threaded\n" << std::endl;
#endif
#ifdef EIGEN_USE_MKL_ALL
std::cout << "Eigen uses the MKL" << std::endl;
#endif
std::cout << "Eigen uses " << Eigen::nbThreads() << " threads" << std::endl;
#ifdef USE_MKL
std::cout << "MKL uses " << mkl_get_max_threads() << " threads" << std::endl;
#endif
std::cout << std::endl;
}
fullTrBenchmark<A2AMatrix<ComplexD>, A2AMatrix<ComplexD>>(ni, nj, nMat);
fullTrBenchmark<A2AMatrix<ComplexD>, A2AMatrixTr<ComplexD>>(ni, nj, nMat);
fullTrBenchmark<A2AMatrixTr<ComplexD>, A2AMatrix<ComplexD>>(ni, nj, nMat);
fullTrBenchmark<A2AMatrixTr<ComplexD>, A2AMatrixTr<ComplexD>>(ni, nj, nMat);
fullMulBenchmark<A2AMatrix<ComplexD>>(ni, nj, nMat);
fullMulBenchmark<A2AMatrixTr<ComplexD>>(ni, nj, nMat);
FINALIZE();
return EXIT_SUCCESS;
}

8
Hadrons/Utilities/Makefile.am
View File

@ -1,4 +1,4 @@
bin_PROGRAMS = HadronsXmlRun HadronsFermionEP64To32
bin_PROGRAMS = HadronsXmlRun HadronsFermionEP64To32 HadronsContractor HadronsContractorBenchmark
HadronsXmlRun_SOURCES = HadronsXmlRun.cc
HadronsXmlRun_LDADD = ../libHadrons.a ../../Grid/libGrid.a
@ -6,3 +6,9 @@ HadronsXmlRun_LDADD = ../libHadrons.a ../../Grid/libGrid.a
HadronsFermionEP64To32_SOURCES = EigenPackCast.cc
HadronsFermionEP64To32_CXXFLAGS = $(AM_CXXFLAGS) -DFIN=WilsonImplD::FermionField -DFOUT=WilsonImplF::FermionField
HadronsFermionEP64To32_LDADD = ../libHadrons.a ../../Grid/libGrid.a
HadronsContractor_SOURCES = Contractor.cc
HadronsContractor_LDADD = ../libHadrons.a ../../Grid/libGrid.a
HadronsContractorBenchmark_SOURCES = ContractorBenchmark.cc
HadronsContractorBenchmark_LDADD = ../libHadrons.a ../../Grid/libGrid.a

24
configure.ac
View File

@ -123,10 +123,13 @@ case ${ac_SFW_FP16} in
AC_MSG_ERROR(["SFW FP16 option not supported ${ac_SFW_FP16}"]);;
esac
############### MKL
############### Intel libraries
AC_ARG_ENABLE([mkl],
[AC_HELP_STRING([--enable-mkl=yes|no|prefix], [enable Intel MKL for LAPACK & FFTW])],
[ac_MKL=${enable_mkl}], [ac_MKL=no])
AC_ARG_ENABLE([ipp],
[AC_HELP_STRING([--enable-ipp=yes|no|prefix], [enable Intel IPP for fast CRC32C])],
[ac_IPP=${enable_mkl}], [ac_IPP=no])
case ${ac_MKL} in
no)
@ -139,6 +142,17 @@ case ${ac_MKL} in
AC_DEFINE([USE_MKL], [1], [Define to 1 if you use the Intel MKL]);;
esac
case ${ac_IPP} in
no)
;;
yes)
AC_DEFINE([USE_IPP], [1], [Define to 1 if you use the Intel IPP]);;
*)
AM_CXXFLAGS="-I$ac_IPP/include $AM_CXXFLAGS"
AM_LDFLAGS="-L$ac_IPP/lib $AM_LDFLAGS"
AC_DEFINE([USE_IPP], [1], [Define to 1 if you use the Intel IPP]);;
esac
############### HDF5
AC_ARG_WITH([hdf5],
[AS_HELP_STRING([--with-hdf5=prefix],
@ -170,7 +184,13 @@ AC_CHECK_FUNCS([gettimeofday])
if test "${ac_MKL}x" != "nox"; then
AC_SEARCH_LIBS([mkl_set_interface_layer], [mkl_rt], [],
[AC_MSG_ERROR("MKL enabled but library not found")])
[AC_MSG_ERROR("Intel MKL enabled but library not found")])
fi
if test "${ac_IPP}x" != "nox"; then
AC_SEARCH_LIBS([ippsCRC32C_8u], [ippdc],
[LIBS="${LIBS} -lippdc -lippvm -lipps -lippcore"],
[AC_MSG_ERROR("Intel IPP enabled but library not found")])
fi
AC_SEARCH_LIBS([__gmpf_init], [gmp],