gallo/atrip
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Compare commits

base: gallo:20c29ed815ff161966d5c6ee08dcd9b58984a36a
Branches Tags
gallo:cuda
gallo:cuda-energy
gallo:openacc
gallo:newtuples
gallo:cuda-test-doubles
gallo:cuda-energy-working
..
compare: gallo:4277c07cc22790fbc4e3fa4ea76ddb9dd494e586
Branches Tags
gallo:cuda-energy
gallo:cuda
gallo:openacc
gallo:newtuples
gallo:cuda-test-doubles
gallo:cuda-energy-working

No commits in common. "20c29ed815ff161966d5c6ee08dcd9b58984a36a" and "4277c07cc22790fbc4e3fa4ea76ddb9dd494e586" have entirely different histories.
20c29ed815 ... 4277c07cc2

13 changed files with 150 additions and 811 deletions
Show Stats Expand all files Collapse all files

10
bench/Makefile.am
View File

@ -1,4 +1,3 @@
AUTOMAKE_OPTIONS = subdir-objects
include $(top_srcdir)/atrip.mk include $(top_srcdir)/atrip.mk
AM_CPPFLAGS = -I$(top_srcdir)/include/ $(CTF_CPPFLAGS) AM_CPPFLAGS = -I$(top_srcdir)/include/ $(CTF_CPPFLAGS)
@ -20,13 +19,4 @@ endif
if WITH_CUDA if WITH_CUDA
test_main_CXXFLAGS = $(CUDA_CXXFLAGS) test_main_CXXFLAGS = $(CUDA_CXXFLAGS)
test_main_LDADD += $(CUDA_LDFLAGS) test_main_LDADD += $(CUDA_LDFLAGS)
AM_CXXFLAGS = $(CUDA_CXXFLAGS)
AM_LDFLAGS += $(CUDA_LDFLAGS)
bin_PROGRAMS += test-cublas-parallel-atrip
test_cublas_parallel_SOURCES = test-cublas-parallel-atrip.cxx
bin_PROGRAMS += test-cuda-sanity
test_cuda_sanity_SOURCES = test-cuda-sanity.cxx
endif endif

239
bench/test-cublas-parallel-atrip.cxx
View File

@ -1,239 +0,0 @@
#include <mpi.h>
#include <iostream>
#include <cassert>
#include <string.h>
#include <sstream>
#include <string>
#include <vector>
#include <cuda.h>
#include <chrono>
#include <map>
#include <cstdlib>
#include <unistd.h>
#include <CLI11.hpp>
#define CUBLASAPI
#include <cublas_api.h>
#include <cublas_v2.h>
struct Timer {
using Clock = std::chrono::high_resolution_clock;
using Event = std::chrono::time_point<Clock>;
std::chrono::duration<double> duration;
Event _start;
inline void start() noexcept { _start = Clock::now(); }
inline void stop() noexcept { duration += Clock::now() - _start; }
inline void clear() noexcept { duration *= 0; }
inline double count() const noexcept { return duration.count(); }
};
using Timings = std::map<std::string, Timer>;
#define _FORMAT(_fmt, ...) \
([&] (void) -> std::string { \
int _sz = std::snprintf(nullptr, 0, _fmt, __VA_ARGS__); \
std::vector<char> _out(_sz + 1); \
std::snprintf(&_out[0], _out.size(), _fmt, __VA_ARGS__); \
return std::string(_out.data()); \
})()
#define _CHECK_CUDA_SUCCESS(message, ...) \
do { \
CUresult result = __VA_ARGS__; \
printf("doing %s\n", message); \
if (result != CUDA_SUCCESS) { \
printf("\t!!CUDA_ERROR(%d): %s:%d %s\n", \
result, \
__FILE__, \
__LINE__, \
message); \
return 1; \
} \
} while (0)
#define _CHECK_CUBLAS_SUCCESS(message, ...) \
do { \
cublasStatus_t result = __VA_ARGS__; \
if (result != 0) { \
printf("\t!!CUBLAS_ERROR(%d): %s:%d %s\n", \
result, \
__FILE__, \
__LINE__, \
message); \
return 1; \
} \
} while (0)
int main(int argc, char** argv) {
using std::vector;
MPI_Init(NULL, NULL);
int rank, np, ngcards;
size_t no(10), nv(no * 10), its(2);
bool barrier = false;
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &np);
CLI::App app{"Main bench for atrip"};
app.add_option("--no", no, "Occupied orbitals");
app.add_option("--nv", nv, "Virtual orbitals");
app.add_option("--its", its, "Number of iterations to be done");
app.add_option("--barrier", barrier, "Call a MPI_Barrier in every iteration?");
CLI11_PARSE(app, argc, argv);
const size_t oo = no * no, ooo = no * oo;
Timings timings;
MPI_Barrier(MPI_COMM_WORLD);
_CHECK_CUDA_SUCCESS("init for cuda",
cuInit(0));
_CHECK_CUDA_SUCCESS("get ncards",
cuDeviceGetCount(&ngcards));
CUcontext ctx;
CUdevice dev;
char hostname[256];
gethostname(hostname, 256);
printf("%s with rank %d gets card %d\n",
hostname,
rank,
rank % ngcards);
// set contexts
_CHECK_CUDA_SUCCESS("device get", cuDeviceGet(&dev, rank % ngcards));
_CHECK_CUDA_SUCCESS("creating context", cuCtxCreate(&ctx, 0, dev));
_CHECK_CUDA_SUCCESS("setting context", cuCtxSetCurrent(ctx));
_CHECK_CUDA_SUCCESS("synchronizing", cuCtxSynchronize());
MPI_Barrier(MPI_COMM_WORLD);
using host_slice_t = vector<double>;
vector<size_t> sizes = {nv * oo, nv * no , oo, oo * no, oo * no};
vector<CUdeviceptr> P_phh(3), P_ph(6) , H_hh(3), H_hhh(3), T_hhh(1);
vector<vector<CUdeviceptr>*> slices_d = {&P_phh, &P_ph , &H_hh, &H_hhh, &T_hhh};
vector<vector<host_slice_t>> slices_h(slices_d.size());
{
int i = -1;
for (auto& v: slices_d) {
i++;
for (auto& ptr: *v) {
_CHECK_CUDA_SUCCESS("malloc",
cuMemAlloc(&ptr,
sizes[i] * sizeof(double)));
slices_h[i].push_back(std::move(std::vector<double>(sizes[i])));
}
}
}
const double one = 1.0, zero = 0.0;
printf("its: %d\n", its);
printf("barrier: %d\n", barrier);
printf("no: %ld\n", no);
printf("nv: %ld\n", nv);
printf("SIZE %f GB\n", (3 * nv * oo
+ 6 * oo * nv
+ 3 * oo
+ 3 * ooo
+ 1 * ooo
) * sizeof(double) / 1024.0 / 1024.0 / 1024.0);
std::map<std::string, double> tflopss
{{ "dgemm", ooo * (no + nv) * 6.0 * 2.0 * its / 1e12},
{ "holes", ooo * no * 6.0 * 2.0 * its / 1e12},
{ "particles", ooo * nv * 6.0 * 2.0 * its / 1e12}};
cublasHandle_t handle;
_CHECK_CUBLAS_SUCCESS("handle create", cublasCreate(&handle));
printf("handle %ld\n", handle);
timings["dgemm"].start();
for (size_t i = 0; i < its; i++) {
if (barrier) {
MPI_Barrier(MPI_COMM_WORLD);
timings["memcpy"].start();
for (size_t _s = 0; _s < slices_d.size(); _s++) {
// for (size_t _b = 0; _b < slices_h[_s].size(); _b++) {
for (size_t _b = 0; _b < 1 ; _b++) {
auto device = (*slices_d[_s])[_b];
auto host = slices_h[_s][_b].data();
cuMemcpyHtoD(device, host, sizes[_s]);
}
}
timings["memcpy"].stop();
}
timings["holes"].start();
for (size_t j = 0; j < 3; j++) {
_CHECK_CUBLAS_SUCCESS(" > 'geming ...",
cublasDgemm(handle,
CUBLAS_OP_N,
CUBLAS_OP_N,
oo, no, no,
&one,
(double*)H_hhh[j], oo,
(double*)H_hh[j], no,
&zero,
(double*)T_hhh[0], oo));
_CHECK_CUBLAS_SUCCESS(" > 'geming ...",
cublasDgemm(handle,
CUBLAS_OP_N,
CUBLAS_OP_T,
oo, no, no,
&one,
(double*)H_hhh[j], oo,
(double*)H_hh[j], no,
&zero,
(double*)T_hhh[0], oo));
}
timings["holes"].stop();
timings["particles"].start();
for (size_t j = 0; j < 6; j++) {
_CHECK_CUBLAS_SUCCESS(" > 'geming ...",
cublasDgemm(handle,
CUBLAS_OP_T,
CUBLAS_OP_N,
oo, no, nv,
&one,
(double*)P_phh[j % 3], nv,
(double*)P_ph[j], nv,
&zero,
(double*)T_hhh[0], oo));
}
timings["particles"].stop();
cuCtxSynchronize();
}
timings["dgemm"].stop();
printf("Performance: \n");
for (auto name: {"holes", "particles", "dgemm"})
printf("%10s TFlops: %4.1f\n",
name,
tflopss[name]
/ timings[name].count());
printf("Timings: \n");
for (auto const& kv: timings)
printf("%10s: %10f\n", kv.first.c_str(), kv.second.count());
MPI_Finalize();
return 0;
}

163
bench/test-cublas-parallel.cxx
View File

@ -1,163 +0,0 @@
#include <mpi.h>
#include <iostream>
#include <cassert>
#include <string.h>
#include <sstream>
#include <string>
#include <vector>
#include <cuda.h>
#include <chrono>
#include <map>
#include <cstdlib>
#include <unistd.h>
#define CUBLASAPI
#include <cublas_api.h>
#include <cublas_v2.h>
struct Timer {
using Clock = std::chrono::high_resolution_clock;
using Event = std::chrono::time_point<Clock>;
std::chrono::duration<double> duration;
Event _start;
inline void start() noexcept { _start = Clock::now(); }
inline void stop() noexcept { duration += Clock::now() - _start; }
inline void clear() noexcept { duration *= 0; }
inline double count() const noexcept { return duration.count(); }
};
using Timings = std::map<std::string, Timer>;
#define _FORMAT(_fmt, ...) \
([&] (void) -> std::string { \
int _sz = std::snprintf(nullptr, 0, _fmt, __VA_ARGS__); \
std::vector<char> _out(_sz + 1); \
std::snprintf(&_out[0], _out.size(), _fmt, __VA_ARGS__); \
return std::string(_out.data()); \
})()
#define _CHECK_CUDA_SUCCESS(message, ...) \
do { \
CUresult result = __VA_ARGS__; \
printf("doing %s\n", message); \
if (result != CUDA_SUCCESS) { \
printf("\t!!CUDA_ERROR(%d): %s:%d %s\n", \
result, \
__FILE__, \
__LINE__, \
message); \
return 1; \
} \
} while (0)
#define _CHECK_CUBLAS_SUCCESS(message, ...) \
do { \
cublasStatus_t result = __VA_ARGS__; \
printf("CUBLAS: doing %s\n", message); \
if (result != 0) { \
printf("\t!!CUBLAS_ERROR(%d): %s:%d %s\n", \
result, \
__FILE__, \
__LINE__, \
message); \
return 1; \
} \
} while (0)
int main(int argc, char** argv) {
MPI_Init(NULL, NULL);
int rank, np, ngcards;
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &np);
Timings timings;
MPI_Barrier(MPI_COMM_WORLD);
_CHECK_CUDA_SUCCESS("init for cuda",
cuInit(0));
_CHECK_CUDA_SUCCESS("get ncards",
cuDeviceGetCount(&ngcards));
const size_t N
= argc > 1
? atoi(argv[1])
: 30000
, dgemms
= argc > 2
? atoi(argv[2])
: 2
, flops = 2 * N * N * N * dgemms
;
CUcontext ctx;
CUdevice dev;
char hostname[256];
gethostname(hostname, 256);
printf("%s with rank %d gets card %d\n",
hostname,
rank,
rank % ngcards);
// set contexts
_CHECK_CUDA_SUCCESS("device get",
cuDeviceGet(&dev, rank % ngcards));
_CHECK_CUDA_SUCCESS("creating context",
cuCtxCreate(&ctx, 0, dev));
_CHECK_CUDA_SUCCESS("setting context",
cuCtxSetCurrent(ctx));
_CHECK_CUDA_SUCCESS("synchronizing",
cuCtxSynchronize());
MPI_Barrier(MPI_COMM_WORLD);
CUdeviceptr A, B, C;
const double one = 1.0;
printf("SIZE %f GB\n", N * N * sizeof(double) / 1024.0 / 1024.0 / 1024.0);
_CHECK_CUDA_SUCCESS("A", cuMemAlloc(&A, N * N * sizeof(double)));
_CHECK_CUDA_SUCCESS("B", cuMemAlloc(&B, N * N * sizeof(double)));
_CHECK_CUDA_SUCCESS("C", cuMemAlloc(&C, N * N * sizeof(double)));
cublasHandle_t handle;
cublasStatus_t stat;
_CHECK_CUBLAS_SUCCESS("handle create", cublasCreate(&handle));
printf("handle %ld\n", handle);
timings["dgemm"].start();
for (size_t i = 0; i < dgemms; i++) {
_CHECK_CUBLAS_SUCCESS(_FORMAT(" > 'geming %ld ...", i).c_str(),
cublasDgemm(handle,
CUBLAS_OP_N,
CUBLAS_OP_N,
N, N, N,
&one,
(double*)A, N,
(double*)B, N,
&one,
(double*)C, N));
}
cuCtxSynchronize();
timings["dgemm"].stop();
printf("dgemm Gflops: %f\n",
flops
/ timings["dgemm"].count()
/ 1024.0
/ 1024.0
/ 1024.0);
MPI_Finalize();
return 0;
}
// Local Variables:
// compile-command: "mpic++ \
// -pedantic -std=c++11 \
// -L./cudaroot/lib64 -lcuda \
// -L./cudaroot/lib64 -lcudart \
// -L./cudaroot/lib64 -lcublas \
// ./test-cublas-parallel.cxx -o test-cublas-parallel"
// End:

104
bench/test-cuda-sanity.cxx
View File

@ -1,104 +0,0 @@
#include <mpi.h>
#include <iostream>
#include <cassert>
#include <string.h>
#include <sstream>
#include <string>
#include <vector>
#include <cuda.h>
#define _CHECK_CUDA_SUCCESS(message, ...) \
do { \
CUresult result = __VA_ARGS__; \
printf("doing %s\n", message); \
if (result != CUDA_SUCCESS) { \
printf("\t!!CUDA_ERROR(%d): %s:%d %s\n", \
result, \
__FILE__, \
__LINE__, \
message); \
return 1; \
} \
} while (0)
int main() {
int rank, np, ngcards;
MPI_Init(NULL, NULL);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &np);
_CHECK_CUDA_SUCCESS("init for cuda",
cuInit(0));
_CHECK_CUDA_SUCCESS("get ncards",
cuDeviceGetCount(&ngcards));
for (size_t rank = 0; rank < ngcards; rank++) {
CUcontext ctx;
CUdevice dev;
CUdevprop_st prop;
size_t _free, total, total2;
char *name = (char*)malloc(256);
printf("Setting contexts\n");
// set contexts
_CHECK_CUDA_SUCCESS("device get",
cuDeviceGet(&dev, rank));
_CHECK_CUDA_SUCCESS("creating context",
cuCtxCreate(&ctx, 0, dev));
_CHECK_CUDA_SUCCESS("setting context",
cuCtxSetCurrent(ctx));
_CHECK_CUDA_SUCCESS("synchronizing",
cuCtxSynchronize());
_CHECK_CUDA_SUCCESS("prop get",
cuDeviceGetProperties(&prop, dev));
_CHECK_CUDA_SUCCESS("meminfo get",
cuMemGetInfo(&_free, &total));
_CHECK_CUDA_SUCCESS("name get",
cuDeviceGetName(name, 256, dev));
_CHECK_CUDA_SUCCESS("totalmem get",
cuDeviceTotalMem(&total2, dev));
printf("\n"
"CUDA CARD RANK %d\n"
"=================\n"
"\tname: %s\n"
"\tShared Mem Per Block (KB): %f\n"
"\tFree/Total mem (GB): %f/%f\n"
"\total2 mem (GB): %f\n"
"\n",
dev,
name,
prop.sharedMemPerBlock / 1024.0,
_free / 1024.0 / 1024.0 / 1024.0 ,
total / 1024.0 / 1024.0 / 1024.0 ,
total2 / 1024.0 / 1024.0 / 1024.0
);
if (_free == 0 || total == 0 || total2 == 0)
return 1;
CUdeviceptr data;
_CHECK_CUDA_SUCCESS("memalloc 1",
cuMemAlloc(&data, sizeof(double) * 10000));
_CHECK_CUDA_SUCCESS("memalloc 2",
cuMemAlloc(&data, sizeof(double) * 10000));
}
MPI_Finalize();
return 0;
}
// Local Variables:
// compile-command: "mpic++ \
// -pedantic -std=c++11 \
// -L./cudaroot/lib64 -lcuda \
// -L./cudaroot/lib64 -lcudart \
// -L./cudaroot/lib64 -lcublas \
// ./mem.cxx -o mem"
// End:

6
bench/test_main.cxx
View File

@ -142,10 +142,8 @@ int main(int argc, char** argv) {
+ /* tai */ f * nranks * no * nv + /* tai */ f * nranks * no * nv
; ;
if (rank == 0) { if (atrip::Atrip::rank == 0)
std::cout << "Tentative MEMORY USAGE (GB): " std::cout << "Tentative MEMORY USAGE: " << atrip_memory << "\n";
<< double(atrip_memory) / 1024.0 / 1024.0 / 1024.0 << "\n";
}
std::vector<int> symmetries(4, NS) std::vector<int> symmetries(4, NS)

2
bootstrap.sh
View File

@ -1,7 +1,5 @@
#!/usr/bin/env bash #!/usr/bin/env bash
cd `git rev-parse --show-toplevel`
type -a autoreconf > /dev/null || type -a autoreconf > /dev/null ||
{ {
cat <<EOF && exit cat <<EOF && exit

36
include/atrip/CUDA.hpp
View File

@ -1,13 +1,5 @@
#pragma once #pragma once
#include <atrip/Utils.hpp>
#if defined(HAVE_CUDA)
#include <cuda.h>
#endif
#include <sstream>
#if defined(HAVE_CUDA) && defined(__CUDACC__) #if defined(HAVE_CUDA) && defined(__CUDACC__)
# define __MAYBE_GLOBAL__ __global__ # define __MAYBE_GLOBAL__ __global__
# define __MAYBE_DEVICE__ __device__ # define __MAYBE_DEVICE__ __device__
@ -15,31 +7,3 @@
# define __MAYBE_GLOBAL__ # define __MAYBE_GLOBAL__
# define __MAYBE_DEVICE__ # define __MAYBE_DEVICE__
#endif #endif
#define _CHECK_CUDA_SUCCESS(message, ...) \
do { \
CUresult result = __VA_ARGS__; \
if (result != CUDA_SUCCESS) { \
auto msg = _FORMAT("\t!!CUDA_ERROR(%d): %s:%d\v%s", \
result, \
__FILE__, \
__LINE__, \
message); \
std::cerr << msg; \
throw msg; \
} \
} while (0)
#define _CHECK_CUBLAS_SUCCESS(message, ...) \
do { \
cublasStatus_t result = __VA_ARGS__; \
if (result != 0) { \
auto msg = _FORMAT("\t!!CUBLAS_ERROR(%d): %s:%d\v%s", \
result, \
__FILE__, \
__LINE__, \
message); \
std::cerr << msg; \
throw msg; \
} \
} while (0)

7
include/atrip/Slice.hpp
View File

@ -21,7 +21,6 @@
#include <atrip/Tuples.hpp> #include <atrip/Tuples.hpp>
#include <atrip/Utils.hpp> #include <atrip/Utils.hpp>
#include <atrip/CUDA.hpp>
namespace atrip { namespace atrip {
@ -459,11 +458,7 @@ void unwrapAndMarkReady() {
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
// copy the retrieved mpi data to the device // copy the retrieved mpi data to the device
WITH_CHRONO("cuda:memcpy", WITH_CHRONO("cuda:memcpy",
_CHECK_CUDA_SUCCESS("copying mpi data to device", cuMemcpyHtoD(data, (void*)mpi_data, sizeof(F) * size);)
cuMemcpyHtoD(data,
(void*)mpi_data,
sizeof(F) * size));
)
std::free(mpi_data); std::free(mpi_data);
#endif #endif

43
include/atrip/SliceUnion.hpp
View File

@ -195,28 +195,7 @@ template <typename F=double>
; ;
if (blank.info.state == Slice<F>::SelfSufficient) { if (blank.info.state == Slice<F>::SelfSufficient) {
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
const size_t _size = sizeof(F) * sources[from.source].size(); blank.mpi_data = sources[from.source].data();
// TODO: this is code duplication with downstairs
if (freePointers.size() == 0) {
std::stringstream stream;
stream << "No more free pointers "
<< "for type " << type
<< " and name " << name
;
throw std::domain_error(stream.str());
}
auto dataPointer = freePointers.begin();
freePointers.erase(dataPointer);
blank.data = *dataPointer;
//
//
// TODO [#A]: do cuMemcpy of
// sources[from.source].data() ⇒ blank.data
// Do this when everything else is working.
// This will probably be a bottleneck of the H-to-D communication,
// as most slices are SelfSufficient.
//
//
#else #else
blank.data = sources[from.source].data(); blank.data = sources[from.source].data();
#endif #endif
@ -337,16 +316,6 @@ template <typename F=double>
} }
} }
#if defined(HAVE_CUDA)
// In cuda, SelfSufficient slices have an ad-hoc pointer
// since it is a pointer on the device and has to be
// brought back to the free pointer bucket of the SliceUnion.
// Therefore, only in the Recycled case it should not be
// put back the pointer.
if (slice.info.state == Slice<F>::Recycled) {
freeSlicePointer = false;
}
#else
// if the slice is self sufficient, do not dare touching the // if the slice is self sufficient, do not dare touching the
// pointer, since it is a pointer to our sources in our rank. // pointer, since it is a pointer to our sources in our rank.
if ( slice.info.state == Slice<F>::SelfSufficient if ( slice.info.state == Slice<F>::SelfSufficient
@ -354,7 +323,6 @@ template <typename F=double>
) { ) {
freeSlicePointer = false; freeSlicePointer = false;
} }
#endif
// make sure we get its data pointer to be used later // make sure we get its data pointer to be used later
// only for non-recycled, since it can be that we need // only for non-recycled, since it can be that we need
@ -378,7 +346,7 @@ template <typename F=double>
// << " info " << slice.info // << " info " << slice.info
<< "\n"; << "\n";
slice.free(); slice.free();
} // we did not find the slice }
} }
} }
@ -412,17 +380,10 @@ template <typename F=double>
for (auto& ptr: sliceBuffers) { for (auto& ptr: sliceBuffers) {
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
const CUresult error =
cuMemAlloc(&ptr, sizeof(F) * sources[0].size()); cuMemAlloc(&ptr, sizeof(F) * sources[0].size());
if (ptr == 0UL) { if (ptr == 0UL) {
throw "UNSUFICCIENT MEMORY ON THE GRAPHIC CARD FOR FREE POINTERS"; throw "UNSUFICCIENT MEMORY ON THE GRAPHIC CARD FOR FREE POINTERS";
} }
if (error != CUDA_SUCCESS) {
std::stringstream s;
s << "Error allocating memory for slice buffers "
<< "code " << error << "\n";
throw s.str();
}
#else #else
ptr = (DataPtr<F>)malloc(sizeof(F) * sources[0].size()); ptr = (DataPtr<F>)malloc(sizeof(F) * sources[0].size());
#endif #endif

10
include/atrip/Utils.hpp
View File

@ -80,16 +80,6 @@ struct Timer {
using Timings = std::map<std::string, Timer>; using Timings = std::map<std::string, Timer>;
// Chrono:1 ends here // Chrono:1 ends here
// A nice handy macro to do formatting
#define _FORMAT(_fmt, ...) \
([&] (void) -> std::string { \
int _sz = std::snprintf(nullptr, 0, _fmt, __VA_ARGS__); \
std::vector<char> _out(_sz + 1); \
std::snprintf(&_out[0], _out.size(), _fmt, __VA_ARGS__); \
return std::string(_out.data()); \
})()
// [[file:~/cuda/atrip/atrip.org::*Epilog][Epilog:1]] // [[file:~/cuda/atrip/atrip.org::*Epilog][Epilog:1]]
} }
// Epilog:1 ends here // Epilog:1 ends here

86
src/atrip/Atrip.cxx
View File

@ -24,7 +24,7 @@
using namespace atrip; using namespace atrip;
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
#include <atrip/CUDA.hpp> #include <cuda.h>
#endif #endif
template <typename F> bool RankMap<F>::RANK_ROUND_ROBIN; template <typename F> bool RankMap<F>::RANK_ROUND_ROBIN;
@ -49,6 +49,11 @@ void Atrip::init(MPI_Comm world) {
Atrip::communicator = world; Atrip::communicator = world;
MPI_Comm_rank(world, (int*)&Atrip::rank); MPI_Comm_rank(world, (int*)&Atrip::rank);
MPI_Comm_size(world, (int*)&Atrip::np); MPI_Comm_size(world, (int*)&Atrip::np);
#if defined(HAVE_CUDA)
Atrip::cuda.status = cublasCreate(&Atrip::cuda.handle);
#endif
} }
template <typename F> template <typename F>
@ -66,24 +71,18 @@ Atrip::Output Atrip::run(Atrip::Input<F> const& in) {
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
int ngcards; int ngcards;
_CHECK_CUDA_SUCCESS("initializing cuda",
cuInit(0));
_CHECK_CUDA_SUCCESS("getting device count",
cuDeviceGetCount(&ngcards));
const auto clusterInfo = getClusterInfo(Atrip::communicator); const auto clusterInfo = getClusterInfo(Atrip::communicator);
cuDeviceGetCount(&ngcards);
LOG(0,"Atrip") << "ngcards: " << ngcards << "\n"; LOG(0,"Atrip") << "ngcards: " << ngcards << "\n";
if (clusterInfo.ranksPerNode > ngcards) { if (clusterInfo.ranksPerNode > ngcards) {
const auto msg std::cerr << "ATRIP: You are running on more ranks per node than the number of graphic cards\n"
= _FORMAT("ATRIP: You are running on more ranks per node than the number of graphic cards\n" << "You have " << ngcards << " cards at your disposal\n";
"You have %d cards at your disposal\n", ngcards); throw "";
std::cerr << msg; }
throw msg; if (clusterInfo.ranksPerNode < ngcards) {
} else if (clusterInfo.ranksPerNode < ngcards) { std::cerr << "You have " << ngcards << " cards at your disposal\n"
const auto msg << "You will be only using " << clusterInfo.ranksPerNode
= _FORMAT("You have %d cards at your disposal.\n" << ", i.e., the nubmer of ranks.\n";
"You will be only using %d, i.e, the number of ranks\n",
ngcards, clusterInfo.ranksPerNode);
std::cerr << msg;
} }
@ -95,27 +94,16 @@ Atrip::Output Atrip::run(Atrip::Input<F> const& in) {
struct { struct { size_t free, total; } avail; size_t total; } memory; struct { struct { size_t free, total; } avail; size_t total; } memory;
char *name = (char*)malloc(256); char *name = (char*)malloc(256);
// - TODO :: we should check that the Zuweisung of graphic cards
// to nodes works as expected, i.e., node k should get from 0
// to ngcards with the formula =rank % ngcards=.
// set current device // set current device
_CHECK_CUDA_SUCCESS("getting device for index <rank>", cuDeviceGet(&dev, rank);
cuDeviceGet(&dev, rank % ngcards)); cuCtxCreate(&ctx, 0, dev);
_CHECK_CUDA_SUCCESS("creating a cuda context", cuCtxSetCurrent(ctx);
cuCtxCreate(&ctx, 0, dev));
_CHECK_CUDA_SUCCESS("setting the context",
cuCtxSetCurrent(ctx));
// get information of the device // get information of the device
_CHECK_CUDA_SUCCESS("getting properties of current device", cuDeviceGetProperties(&prop, dev);
cuDeviceGetProperties(&prop, dev)); cuMemGetInfo(&memory.avail.free, &memory.avail.total);
_CHECK_CUDA_SUCCESS("getting memory information", cuDeviceGetName(name, 256, dev);
cuMemGetInfo(&memory.avail.free, &memory.avail.total)); cuDeviceTotalMem(&memory.total, dev);
_CHECK_CUDA_SUCCESS("getting name",
cuDeviceGetName(name, 256, dev));
_CHECK_CUDA_SUCCESS("getting total memory",
cuDeviceTotalMem(&memory.total, dev));
printf("\n" printf("\n"
"CUDA CARD RANK %d\n" "CUDA CARD RANK %d\n"
@ -136,10 +124,6 @@ Atrip::Output Atrip::run(Atrip::Input<F> const& in) {
memory.total / 1024.0 / 1024.0 / 1024.0 memory.total / 1024.0 / 1024.0 / 1024.0
); );
std::free((void*)name); std::free((void*)name);
_CHECK_CUBLAS_SUCCESS("creating a cublas handle",
cublasCreate(&Atrip::cuda.handle));
} }
MPI_Barrier(universe); MPI_Barrier(universe);
} }
@ -179,27 +163,17 @@ Atrip::Output Atrip::run(Atrip::Input<F> const& in) {
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
DataPtr<F> Tai, epsi, epsa; DataPtr<F> Tai, epsi, epsa;
//TODO: free memory pointers in the end of the algorithm //TODO: free memory pointers in the end of the algorithm
cuMemAlloc(&Tai, sizeof(F) * _Tai.size());
cuMemAlloc(&epsi, sizeof(F) * _epsi.size());
cuMemAlloc(&epsa, sizeof(F) * _epsa.size());
_CHECK_CUDA_SUCCESS("Tai", cuMemcpyHtoD(Tai, (void*)_Tai.data(), sizeof(F) * _Tai.size());
cuMemAlloc(&Tai, sizeof(F) * _Tai.size())); cuMemcpyHtoD(epsi,(void*)_epsi.data(), sizeof(F) * _epsi.size());
_CHECK_CUDA_SUCCESS("epsi", cuMemcpyHtoD(epsa, (void*)_epsa.data(), sizeof(F) * _epsa.size());
cuMemAlloc(&epsi, sizeof(F) * _epsi.size()));
_CHECK_CUDA_SUCCESS("epsa",
cuMemAlloc(&epsa, sizeof(F) * _epsa.size()));
_CHECK_CUDA_SUCCESS("memcpy Tai", cuMemAlloc(&Tijk, sizeof(F) * No * No * No);
cuMemcpyHtoD(Tai, (void*)_Tai.data(), sizeof(F) * _Tai.size())); cuMemAlloc(&Zijk, sizeof(F) * No * No * No);
_CHECK_CUDA_SUCCESS("memcpy epsi",
cuMemcpyHtoD(epsi,(void*)_epsi.data(), sizeof(F) * _epsi.size()));
_CHECK_CUDA_SUCCESS("memcpy epsa",
cuMemcpyHtoD(epsa, (void*)_epsa.data(), sizeof(F) * _epsa.size()));
_CHECK_CUDA_SUCCESS("Tijk",
cuMemAlloc(&Tijk, sizeof(F) * No * No * No));
_CHECK_CUDA_SUCCESS("Zijk",
cuMemAlloc(&Zijk, sizeof(F) * No * No * No));
#else #else
std::vector<F> &Tai = _Tai, &epsi = _epsi, &epsa = _epsa; std::vector<F> &Tai = _Tai, &epsi = _epsi, &epsa = _epsa;
Zijk = (DataFieldType<F>*)malloc(No*No*No * sizeof(DataFieldType<F>)); Zijk = (DataFieldType<F>*)malloc(No*No*No * sizeof(DataFieldType<F>));

18
src/atrip/Blas.cxx
View File

@ -15,13 +15,10 @@
// [[file:~/cuda/atrip/atrip.org::*Blas][Blas:2]] // [[file:~/cuda/atrip/atrip.org::*Blas][Blas:2]]
#include <atrip/Blas.hpp> #include <atrip/Blas.hpp>
#include <atrip/Atrip.hpp> #include <atrip/Atrip.hpp>
#include <atrip/CUDA.hpp>
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
# include <cstring> # include <cstring>
static size_t dgem_call = 0;
static inline static inline
cublasOperation_t char_to_cublasOperation(const char* trans) { cublasOperation_t char_to_cublasOperation(const char* trans) {
if (strncmp("N", trans, 1) == 0) if (strncmp("N", trans, 1) == 0)
@ -52,8 +49,6 @@ namespace atrip {
typename DataField<double>::type *C, typename DataField<double>::type *C,
const int *ldc) { const int *ldc) {
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
// TODO: remove this verbose checking
const cublasStatus_t error =
cublasDgemm(Atrip::cuda.handle, cublasDgemm(Atrip::cuda.handle,
char_to_cublasOperation(transa), char_to_cublasOperation(transa),
char_to_cublasOperation(transb), char_to_cublasOperation(transb),
@ -61,14 +56,6 @@ namespace atrip {
alpha, A, *lda, alpha, A, *lda,
B, *ldb, beta, B, *ldb, beta,
C, *ldc); C, *ldc);
if (error != 0) printf(":%-3ld (%4ldth) ERR<%4d> cublasDgemm: "
"A = %20ld "
"B = %20ld "
"C = %20ld "
"\n",
Atrip::rank,
dgem_call++,
error, A, B, C);
#else #else
dgemm_(transa, transb, dgemm_(transa, transb,
m, n, k, m, n, k,
@ -96,17 +83,16 @@ namespace atrip {
cuDoubleComplex cuDoubleComplex
cu_alpha = {std::real(*alpha), std::imag(*alpha)}, cu_alpha = {std::real(*alpha), std::imag(*alpha)},
cu_beta = {std::real(*beta), std::imag(*beta)}; cu_beta = {std::real(*beta), std::imag(*beta)};
_CHECK_CUBLAS_SUCCESS("cublasZgemm",
cublasZgemm(Atrip::cuda.handle, cublasZgemm(Atrip::cuda.handle,
char_to_cublasOperation(transa), char_to_cublasOperation(transa),
char_to_cublasOperation(transb), char_to_cublasOperation(transb),
*m, *n, *k, *m, *n, *k,
&cu_alpha, &cu_alpha,
A, *lda, A, *lda,
B, *ldb, B, *ldb,
&cu_beta, &cu_beta,
C, *ldc)); C, *ldc);
#else #else
zgemm_(transa, transb, zgemm_(transa, transb,
m, n, k, m, n, k,

141
src/atrip/Equations.cxx
View File

@ -14,13 +14,31 @@
// [[file:~/cuda/atrip/atrip.org::*Prolog][Prolog:2]] // [[file:~/cuda/atrip/atrip.org::*Prolog][Prolog:2]]
#include<atrip/Equations.hpp> #include<atrip/Equations.hpp>
#include<atrip/CUDA.hpp> #include<atrip/CUDA.hpp>
#if defined(HAVE_CUDA)
#include <cuda.h>
#endif
namespace atrip { namespace atrip {
// Prolog:2 ends here // Prolog:2 ends here
// These are just help structures
// to help with the templating of reorder
// function
enum reordering_t
{
IJK,
IKJ,
JIK,
JKI,
KIJ,
KJI
};
template <typename F, reordering_t R>
struct reorder_proxy {};
#ifdef HAVE_CUDA #ifdef HAVE_CUDA
namespace cuda { namespace cuda {
@ -90,35 +108,42 @@ namespace cuda {
template <typename F> template <typename F>
__device__ __device__
void sum_in_place(F* to, const F* from); void sum_in_place(F* to, const F* b);
template <> template <>
__device__ __device__
void sum_in_place(double* to, const double *from) { *to += *from; } void sum_in_place(double* to, const double *b) { *to += *b; }
template <> template <>
__device__ __device__
void sum_in_place(cuDoubleComplex* to, const cuDoubleComplex* from) { void sum_in_place(cuDoubleComplex* to, const cuDoubleComplex* b) {
to->x += from->x; to->x += b->x;
to->y += from->y; to->y += b->y;
}
__device__
cuDoubleComplex& operator+=(cuDoubleComplex& lz, cuDoubleComplex const& rz) {
lz.x += rz.x;
lz.y += rz.y;
return lz;
} }
}; };
#endif #endif
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
#define FOR_K() \ #define LIMS_KS() \
for (size_t kmin = blockIdx.x * blockDim.x + threadIdx.x, \ size_t kmin = blockIdx.x * blockDim.x + threadIdx.x, \
k = kmin, \ k = kmin, \
idx = kmin * size * size * size; \ idx = kmin * size * size * size \
k < (kmin < size) ? kmin + 1 : size; \ ; \
k++) k < (kmin < size) ? kmin + 1 : size
#else #else
#define FOR_K() for (size_t k=0, idx=0; k < size; k++) #define LIMS_KS size_t k=0, idx=0; k < size
#endif #endif
#define _IJK_(i, j, k) i + j*size + k*size*size #define _IJK_(i, j, k) i + j*size + k*size*size
#define _REORDER_BODY_(...) \ #define _REORDER_BODY_(...) \
FOR_K() \ for (LIMS_KS() ; k++) \
for (size_t j = 0; j < size; j++) \ for (size_t j = 0; j < size; j++) \
for (size_t i = 0; i < size; i++, idx++) { \ for (size_t i = 0; i < size; i++, idx++) { \
__VA_ARGS__ \ __VA_ARGS__ \
@ -127,9 +152,7 @@ namespace cuda {
template <typename F> \ template <typename F> \
__MAYBE_GLOBAL__ \ __MAYBE_GLOBAL__ \
void reorder(reorder_proxy< F, _enum > p, \ void reorder(reorder_proxy< F, _enum > p, \
size_t size, \ size_t size, F* to, F* from) { \
F* to, \
F* from) { \
_REORDER_BODY_(__VA_ARGS__) \ _REORDER_BODY_(__VA_ARGS__) \
} }
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
@ -138,31 +161,9 @@ namespace cuda {
#define GO(__TO, __FROM) __TO += __FROM; #define GO(__TO, __FROM) __TO += __FROM;
#endif #endif
// These are just help structures
// to help with the templating of reorder
// function
enum reordering_t
{
IJK,
IKJ,
JIK,
JKI,
KIJ,
KJI
};
/*
* Please the c++ type checker and template creator
* in order to have an argument in the signature of
* the function that helps the compiler know which
* instantiation it should take.
*
*/
template <typename F, reordering_t R>
struct reorder_proxy {};
template <typename F, reordering_t R> template <typename F, reordering_t R>
__MAYBE_GLOBAL__ __MAYBE_GLOBAL__ \
void reorder(reorder_proxy<F, R> proxy, void reorder(reorder_proxy<F, R> proxy,
size_t size, F* to, F* from); size_t size, F* to, F* from);
@ -445,13 +446,9 @@ double getEnergySame
#if defined(ATRIP_USE_DGEMM) #if defined(ATRIP_USE_DGEMM)
#if defined(HAVE_CUDA) #if defined(HAVE_CUDA)
#define REORDER(__II, __JJ, __KK) \ #define REORDER(__II, __JJ, __KK) \
reorder<<<bs, ths>>>(reorder_proxy< \ reorder<<<bs, ths>>>(reorder_proxy<DataFieldType<F>, \
DataFieldType<F>, \ __II ## __JJ ## __KK >{}, \
__II ## __JJ ## __KK \ No, Tijk, _t_buffer);
>{}, \
No, \
Tijk, \
_t_buffer);
#define DGEMM_PARTICLES(__A, __B) \ #define DGEMM_PARTICLES(__A, __B) \
atrip::xgemm<F>("T", \ atrip::xgemm<F>("T", \
"N", \ "N", \
@ -465,7 +462,7 @@ double getEnergySame
(int const*)&Nv, \ (int const*)&Nv, \
&zero, \ &zero, \
_t_buffer, \ _t_buffer, \
(int const*)&NoNo) (int const*)&NoNo);
#define DGEMM_HOLES(__A, __B, __TRANSB) \ #define DGEMM_HOLES(__A, __B, __TRANSB) \
atrip::xgemm<F>("N", \ atrip::xgemm<F>("N", \
__TRANSB, \ __TRANSB, \
@ -480,7 +477,7 @@ double getEnergySame
&zero, \ &zero, \
_t_buffer, \ _t_buffer, \
(int const*)&NoNo \ (int const*)&NoNo \
) );
#define MAYBE_CONJ(_conj, _buffer) \ #define MAYBE_CONJ(_conj, _buffer) \
cuda::maybeConjugate<<< \ cuda::maybeConjugate<<< \
Atrip::kernelDimensions.ooo.blocks, \ Atrip::kernelDimensions.ooo.blocks, \
@ -515,7 +512,7 @@ double getEnergySame
&zero, \ &zero, \
_t_buffer, \ _t_buffer, \
(int const*)&NoNo \ (int const*)&NoNo \
) );
#define DGEMM_HOLES(__A, __B, __TRANSB) \ #define DGEMM_HOLES(__A, __B, __TRANSB) \
atrip::xgemm<F>("N", \ atrip::xgemm<F>("N", \
__TRANSB, \ __TRANSB, \
@ -530,7 +527,7 @@ double getEnergySame
&zero, \ &zero, \
_t_buffer, \ _t_buffer, \
(int const*)&NoNo \ (int const*)&NoNo \
) );
#define MAYBE_CONJ(_conj, _buffer) \ #define MAYBE_CONJ(_conj, _buffer) \
for (size_t __i = 0; __i < NoNoNo; ++__i) \ for (size_t __i = 0; __i < NoNoNo; ++__i) \
_conj[__i] = maybeConjugate<F>(_buffer[__i]); _conj[__i] = maybeConjugate<F>(_buffer[__i]);
@ -541,14 +538,8 @@ double getEnergySame
#ifdef HAVE_CUDA #ifdef HAVE_CUDA
DataFieldType<F>* _t_buffer; DataFieldType<F>* _t_buffer;
DataFieldType<F>* _vhhh; DataFieldType<F>* _vhhh;
WITH_CHRONO("double:cuda:alloc", cuMemAlloc((CUdeviceptr*)&_t_buffer, NoNoNo * sizeof(DataFieldType<F>));
_CHECK_CUDA_SUCCESS("Allocating _t_buffer", cuMemAlloc((CUdeviceptr*)&_vhhh, NoNoNo * sizeof(DataFieldType<F>));
cuMemAlloc((CUdeviceptr*)&_t_buffer,
NoNoNo * sizeof(DataFieldType<F>)));
_CHECK_CUDA_SUCCESS("Allocating _vhhh",
cuMemAlloc((CUdeviceptr*)&_vhhh,
NoNoNo * sizeof(DataFieldType<F>)));
)
const size_t const size_t
bs = Atrip::kernelDimensions.ooo.blocks, bs = Atrip::kernelDimensions.ooo.blocks,
ths = Atrip::kernelDimensions.ooo.threads; ths = Atrip::kernelDimensions.ooo.threads;
@ -569,6 +560,7 @@ double getEnergySame
WITH_CHRONO("double:reorder", WITH_CHRONO("double:reorder",
cuda::zeroing<<<bs, ths>>>((DataFieldType<F>*)Tijk, cuda::zeroing<<<bs, ths>>>((DataFieldType<F>*)Tijk,
NoNoNo); NoNoNo);
// synchronize all initializations to zero
) )
#else #else
WITH_CHRONO("double:reorder", WITH_CHRONO("double:reorder",
@ -583,36 +575,36 @@ double getEnergySame
// VhhhC[i + k*No + L*NoNo] * TABhh[L + j*No]; H1 // VhhhC[i + k*No + L*NoNo] * TABhh[L + j*No]; H1
MAYBE_CONJ(_vhhh, VhhhC) MAYBE_CONJ(_vhhh, VhhhC)
WITH_CHRONO("doubles:holes:1", WITH_CHRONO("doubles:holes:1",
DGEMM_HOLES(_vhhh, TABhh, "N"); DGEMM_HOLES(_vhhh, TABhh, "N")
REORDER(I, K, J) REORDER(I, K, J)
) )
// VhhhC[j + k*No + L*NoNo] * TABhh[i + L*No]; H0 // VhhhC[j + k*No + L*NoNo] * TABhh[i + L*No]; H0
WITH_CHRONO("doubles:holes:2", WITH_CHRONO("doubles:holes:2",
DGEMM_HOLES(_vhhh, TABhh, "T"); DGEMM_HOLES(_vhhh, TABhh, "T")
REORDER(J, K, I) REORDER(J, K, I)
) )
// VhhhB[i + j*No + L*NoNo] * TAChh[L + k*No]; H5 // VhhhB[i + j*No + L*NoNo] * TAChh[L + k*No]; H5
MAYBE_CONJ(_vhhh, VhhhB) MAYBE_CONJ(_vhhh, VhhhB)
WITH_CHRONO("doubles:holes:3", WITH_CHRONO("doubles:holes:3",
DGEMM_HOLES(_vhhh, TAChh, "N"); DGEMM_HOLES(_vhhh, TAChh, "N")
REORDER(I, J, K) REORDER(I, J, K)
) )
// VhhhB[k + j*No + L*NoNo] * TAChh[i + L*No]; H3 // VhhhB[k + j*No + L*NoNo] * TAChh[i + L*No]; H3
WITH_CHRONO("doubles:holes:4", WITH_CHRONO("doubles:holes:4",
DGEMM_HOLES(_vhhh, TAChh, "T"); DGEMM_HOLES(_vhhh, TAChh, "T")
REORDER(K, J, I) REORDER(K, J, I)
) )
// VhhhA[j + i*No + L*NoNo] * TBChh[L + k*No]; H1 // VhhhA[j + i*No + L*NoNo] * TBChh[L + k*No]; H1
MAYBE_CONJ(_vhhh, VhhhA) MAYBE_CONJ(_vhhh, VhhhA)
WITH_CHRONO("doubles:holes:5", WITH_CHRONO("doubles:holes:5",
DGEMM_HOLES(_vhhh, TBChh, "N"); DGEMM_HOLES(_vhhh, TBChh, "N")
REORDER(J, I, K) REORDER(J, I, K)
) )
// VhhhA[k + i*No + L*NoNo] * TBChh[j + L*No]; H4 // VhhhA[k + i*No + L*NoNo] * TBChh[j + L*No]; H4
WITH_CHRONO("doubles:holes:6", WITH_CHRONO("doubles:holes:6",
DGEMM_HOLES(_vhhh, TBChh, "T"); DGEMM_HOLES(_vhhh, TBChh, "T")
REORDER(K, I, J) REORDER(K, I, J)
) )
} }
@ -624,32 +616,32 @@ double getEnergySame
{ {
// TAphh[E + i*Nv + j*NoNv] * VBCph[E + k*Nv]; P0 // TAphh[E + i*Nv + j*NoNv] * VBCph[E + k*Nv]; P0
WITH_CHRONO("doubles:particles:1", WITH_CHRONO("doubles:particles:1",
DGEMM_PARTICLES(TAphh, VBCph); DGEMM_PARTICLES(TAphh, VBCph)
REORDER(I, J, K) REORDER(I, J, K)
) )
// TAphh[E + i*Nv + k*NoNv] * VCBph[E + j*Nv]; P3 // TAphh[E + i*Nv + k*NoNv] * VCBph[E + j*Nv]; P3
WITH_CHRONO("doubles:particles:2", WITH_CHRONO("doubles:particles:2",
DGEMM_PARTICLES(TAphh, VCBph); DGEMM_PARTICLES(TAphh, VCBph)
REORDER(I, K, J) REORDER(I, K, J)
) )
// TCphh[E + k*Nv + i*NoNv] * VABph[E + j*Nv]; P5 // TCphh[E + k*Nv + i*NoNv] * VABph[E + j*Nv]; P5
WITH_CHRONO("doubles:particles:3", WITH_CHRONO("doubles:particles:3",
DGEMM_PARTICLES(TCphh, VABph); DGEMM_PARTICLES(TCphh, VABph)
REORDER(K, I, J) REORDER(K, I, J)
) )
// TCphh[E + k*Nv + j*NoNv] * VBAph[E + i*Nv]; P2 // TCphh[E + k*Nv + j*NoNv] * VBAph[E + i*Nv]; P2
WITH_CHRONO("doubles:particles:4", WITH_CHRONO("doubles:particles:4",
DGEMM_PARTICLES(TCphh, VBAph); DGEMM_PARTICLES(TCphh, VBAph)
REORDER(K, J, I) REORDER(K, J, I)
) )
// TBphh[E + j*Nv + i*NoNv] * VACph[E + k*Nv]; P1 // TBphh[E + j*Nv + i*NoNv] * VACph[E + k*Nv]; P1
WITH_CHRONO("doubles:particles:5", WITH_CHRONO("doubles:particles:5",
DGEMM_PARTICLES(TBphh, VACph); DGEMM_PARTICLES(TBphh, VACph)
REORDER(J, I, K) REORDER(J, I, K)
) )
// TBphh[E + j*Nv + k*NoNv] * VCAph[E + i*Nv]; P4 // TBphh[E + j*Nv + k*NoNv] * VCAph[E + i*Nv]; P4
WITH_CHRONO("doubles:particles:6", WITH_CHRONO("doubles:particles:6",
DGEMM_PARTICLES(TBphh, VCAph); DGEMM_PARTICLES(TBphh, VCAph)
REORDER(J, K, I) REORDER(J, K, I)
) )
} }
@ -659,12 +651,9 @@ double getEnergySame
#ifdef HAVE_CUDA #ifdef HAVE_CUDA
// we need to synchronize here since we need // we need to synchronize here since we need
// the Tijk for next process in the pipeline // the Tijk for next process in the pipeline
_CHECK_CUDA_SUCCESS("Synchronizing", cuCtxSynchronize();
cuCtxSynchronize()); cuMemFree((CUdeviceptr)_vhhh);
_CHECK_CUDA_SUCCESS("Freeing _vhhh", cuMemFree((CUdeviceptr)_t_buffer);
cuMemFree((CUdeviceptr)_vhhh));
_CHECK_CUDA_SUCCESS("Freeing _t_buffer",
cuMemFree((CUdeviceptr)_t_buffer));
#else #else
free(_vhhh); free(_vhhh);
free(_t_buffer); free(_t_buffer);