mirror of
https://github.com/PhasicFlow/phasicFlow.git
synced 2025-06-22 16:28:30 +00:00
194 lines
4.2 KiB
C++
194 lines
4.2 KiB
C++
/*------------------------------- phasicFlow ---------------------------------
|
|
O C enter of
|
|
O O E ngineering and
|
|
O O M ultiscale modeling of
|
|
OOOOOOO F luid flow
|
|
------------------------------------------------------------------------------
|
|
Copyright (C): www.cemf.ir
|
|
email: hamid.r.norouzi AT gmail.com
|
|
------------------------------------------------------------------------------
|
|
Licence:
|
|
This file is part of phasicFlow code. It is a free software for simulating
|
|
granular and multiphase flows. You can redistribute it and/or modify it under
|
|
the terms of GNU General Public License v3 or any other later versions.
|
|
|
|
phasicFlow is distributed to help others in their research in the field of
|
|
granular and multiphase flows, but WITHOUT ANY WARRANTY; without even the
|
|
implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
|
|
|
|
-----------------------------------------------------------------------------*/
|
|
|
|
#include "AdamsMoulton4.hpp"
|
|
|
|
//const real AB2_coef[] = { 3.0 / 2.0, 1.0 / 2.0};
|
|
|
|
pFlow::AdamsMoulton4::AdamsMoulton4
|
|
(
|
|
const word& baseName,
|
|
repository& owner,
|
|
const pointStructure& pStruct,
|
|
const word& method
|
|
)
|
|
:
|
|
integration(baseName, owner, pStruct, method),
|
|
y0_(
|
|
owner.emplaceObject<realx3PointField_D>(
|
|
objectFile(
|
|
groupNames(baseName,"y0"),
|
|
"",
|
|
objectFile::READ_IF_PRESENT,
|
|
objectFile::WRITE_ALWAYS),
|
|
pStruct,
|
|
zero3,
|
|
false
|
|
)
|
|
),
|
|
dy0_(
|
|
owner.emplaceObject<realx3PointField_D>(
|
|
objectFile(
|
|
groupNames(baseName,"dy0"),
|
|
"",
|
|
objectFile::READ_IF_PRESENT,
|
|
objectFile::WRITE_ALWAYS),
|
|
pStruct,
|
|
zero3
|
|
)
|
|
),
|
|
dy1_(
|
|
owner.emplaceObject<realx3PointField_D>(
|
|
objectFile(
|
|
groupNames(baseName,"dy1"),
|
|
"",
|
|
objectFile::READ_IF_PRESENT,
|
|
objectFile::WRITE_ALWAYS),
|
|
pStruct,
|
|
zero3
|
|
)
|
|
),
|
|
dy2_(
|
|
owner.emplaceObject<realx3PointField_D>(
|
|
objectFile(
|
|
groupNames(baseName,"dy2"),
|
|
"",
|
|
objectFile::READ_IF_PRESENT,
|
|
objectFile::WRITE_ALWAYS),
|
|
pStruct,
|
|
zero3
|
|
)
|
|
)
|
|
{
|
|
|
|
}
|
|
|
|
bool pFlow::AdamsMoulton4::predict
|
|
(
|
|
real dt,
|
|
realx3Vector_D& y,
|
|
realx3Vector_D& dy
|
|
)
|
|
{
|
|
|
|
if(this->pStruct().allActive())
|
|
{
|
|
return predictAll(dt, y, dy, this->pStruct().activeRange());
|
|
}
|
|
else
|
|
{
|
|
return predictRange(dt, y, dy, this->pStruct().activePointsMaskD());
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
bool pFlow::AdamsMoulton4::correct
|
|
(
|
|
real dt,
|
|
realx3Vector_D& y,
|
|
realx3Vector_D& dy
|
|
)
|
|
{
|
|
if(this->pStruct().allActive())
|
|
{
|
|
return intAll(dt, y, dy, this->pStruct().activeRange());
|
|
}
|
|
else
|
|
{
|
|
return intRange(dt, y, dy, this->pStruct().activePointsMaskD());
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
bool pFlow::AdamsMoulton4::setInitialVals(
|
|
const int32IndexContainer& newIndices,
|
|
const realx3Vector& y)
|
|
{
|
|
y0_.insertSetElement(newIndices, y);
|
|
|
|
return true;
|
|
}
|
|
|
|
bool pFlow::AdamsMoulton4::predictAll(
|
|
real dt,
|
|
realx3Vector_D& y,
|
|
realx3Vector_D& dy,
|
|
range activeRng)
|
|
{
|
|
|
|
auto d_dy = dy.deviceViewAll();
|
|
auto d_y = y.deviceViewAll();
|
|
|
|
auto d_y0 = y0_.deviceViewAll();
|
|
auto d_dy0 = dy0_.deviceViewAll();
|
|
auto d_dy1 = dy1_.deviceViewAll();
|
|
auto d_dy2 = dy2_.deviceViewAll();
|
|
|
|
Kokkos::parallel_for(
|
|
"AdamsMoulton4::predict",
|
|
rpIntegration (activeRng.first, activeRng.second),
|
|
LAMBDA_HD(int32 i){
|
|
d_dy0[i] = d_dy[i];
|
|
d_y[i] = d_y0[i] + dt*(
|
|
static_cast<real>(23.0 /12.0 ) * d_dy[i]
|
|
- static_cast<real>(16.0 / 12.0) * d_dy1[i]
|
|
+ static_cast<real>( 5.0 / 12.0) * d_dy2[i]);
|
|
});
|
|
Kokkos::fence();
|
|
|
|
return true;
|
|
}
|
|
|
|
bool pFlow::AdamsMoulton4::intAll(
|
|
real dt,
|
|
realx3Vector_D& y,
|
|
realx3Vector_D& dy,
|
|
range activeRng)
|
|
{
|
|
|
|
auto d_dy = dy.deviceViewAll();
|
|
auto d_y = y.deviceViewAll();
|
|
|
|
auto d_dy0 = dy0_.deviceViewAll();
|
|
auto d_y0 = y0_.deviceViewAll();
|
|
auto d_dy1 = dy1_.deviceViewAll();
|
|
auto d_dy2 = dy2_.deviceViewAll();
|
|
|
|
Kokkos::parallel_for(
|
|
"AdamsMoulton4::correct",
|
|
rpIntegration (activeRng.first, activeRng.second),
|
|
LAMBDA_HD(int32 i){
|
|
auto corrct_y = d_y0[i] + dt*(
|
|
static_cast<real>(9.0/24.0)*d_dy[i]
|
|
+ static_cast<real>(19.0/24.0)*d_dy0[i]
|
|
- static_cast<real>( 5.0/24.0)*d_dy1[i]
|
|
+ static_cast<real>( 1.0/24.0)*d_dy2[i]);
|
|
|
|
d_dy2[i]= d_dy1[i];
|
|
d_dy1[i]= d_dy0[i];
|
|
d_y0[i] = corrct_y;
|
|
d_y[i] = corrct_y;
|
|
});
|
|
Kokkos::fence();
|
|
|
|
return true;
|
|
} |