Wed 28 Aug 21:38:52 CEST 2024

This commit is contained in:
sbosse 2024-08-28 21:43:37 +02:00
parent d9f4cc608a
commit b98e96edf7

575
src/SimNDT/engine/efit2d.py Normal file
View File

@ -0,0 +1,575 @@
from SimNDT.engine.engineBase import EngineBase
try:
import SimNDT.engine.efit2dcython as EFIT2DCython
except:
print("error in importation for Serial EFIT2D")
from SimNDT.core.material import Material
from SimNDT.core.constants import BC
import copy
import numpy as np
global ErrorImportCL
try:
import pyopencl as cl
ErrorImportCL = False
except ImportError:
ErrorImportCL = True
c11 = 4350 * 4350 * 7750
c12 = 4350 * 4350 * 7750 - 2260 * 2260 * 7750
c22 = c11
c44 = 2260 * 2260 * 7750
pzt = Material("pzt", 7750.0, c11, c12, c22, c44)
class EFIT2D(EngineBase):
def __init__(self, simPack, Platform):
EngineBase.__init__(self, simPack, Platform)
self.max_value = 0.0
self._Receiver = False
def setup_CL(self):
if self.Platform == "OpenCL":
self.initFieldsCL()
def materialSetup(self):
Materials = copy.deepcopy(self.simPack.Materials)
MRI, NRI = np.shape(self.simPack.Simulation.Im)
# Condicion de vacio si uno de los materiales es aire
for n in range(len(Materials)):
if Materials[n].Rho < 2.0:
Materials[n].Rho = 10e23
Materials[n].C11 = 1e-20
Materials[n].C12 = 1e-20
Materials[n].C22 = 1e-20
Materials[n].C44 = 1e-20
Materials[n].Eta_v = 1e-20
Materials[n].Eta_s = 1e-20
# verificar que no sea cero exactamente
if Materials[n].C44 == 0:
Materials[n].C44 = 1e-30
if Materials[n].Eta_s == 0:
Materials[n].Eta_s = 1e-30
self.Rho = np.zeros((MRI, NRI), dtype=np.float32)
self.C11 = np.zeros((MRI, NRI), dtype=np.float32)
self.C12 = np.zeros((MRI, NRI), dtype=np.float32)
self.C22 = np.zeros((MRI, NRI), dtype=np.float32)
self.C44 = np.ones((MRI, NRI), dtype=np.float32) * 1e-30
self.Eta_v = np.zeros((MRI, NRI), dtype=np.float32)
self.Eta_s = np.zeros((MRI, NRI), dtype=np.float32)
for n in range(len(Materials)):
ind = np.nonzero(self.simPack.Simulation.Im == Materials[n].Label)
self.Rho[ind] = Materials[n].Rho
self.C11[ind] = Materials[n].C11
self.C12[ind] = Materials[n].C12
self.C22[ind] = Materials[n].C22
self.C44[ind] = Materials[n].C44
self.Eta_v[ind] = Materials[n].Eta_v
self.Eta_s[ind] = Materials[n].Eta_s
# Find the base materials
for i, item in enumerate(Materials):
if item.Label == 0:
base_material = i
# Set material in boundaries
ind = np.nonzero(self.simPack.Simulation.Im == 255.0)
self.Rho[ind] = Materials[base_material].Rho
self.C11[ind] = Materials[base_material].C11
self.C12[ind] = Materials[base_material].C12
self.C22[ind] = Materials[base_material].C22
self.C44[ind] = Materials[base_material].C44
self.Eta_v[ind] = Materials[base_material].Eta_v
self.Eta_s[ind] = Materials[base_material].Eta_s
def initFields(self):
MRI, NRI = np.shape(self.simPack.Simulation.Im)
Signal = self.simPack.Signal
t = self.simPack.Simulation.t
self.source = Signal.generate(t)
self.Vx = np.zeros((MRI, NRI), dtype=np.float32)
self.Vy = np.zeros((MRI, NRI), dtype=np.float32)
self.Txx = np.zeros((MRI, NRI), dtype=np.float32)
self.Txy = np.zeros((MRI, NRI), dtype=np.float32)
self.Tyy = np.zeros((MRI, NRI), dtype=np.float32)
self.DVx = np.zeros((MRI, NRI), dtype=np.float32)
self.DVy = np.zeros((MRI, NRI), dtype=np.float32)
self.ABS = np.ones((MRI, NRI), dtype=np.float32)
self.SV = np.zeros((MRI, NRI), dtype=np.float32)
def staggeredProp(self):
MRI, NRI = np.shape(self.simPack.Simulation.Im)
BXtemp = np.zeros((MRI, NRI), dtype=np.float32)
BYtemp = np.zeros((MRI, NRI), dtype=np.float32)
self.BX = np.zeros((MRI, NRI), dtype=np.float32)
self.BY = np.zeros((MRI, NRI), dtype=np.float32)
self.C44_Eff = np.zeros((MRI, NRI), dtype=np.float32)
BXtemp[:, :] = 1.0 / self.Rho[:, :]
BYtemp[:, :] = 1.0 / self.Rho[:, :]
self.BX[:-2, :] = 0.5 * (BXtemp[1:-1, :] + BXtemp[:-2, :])
self.BX[-2, :] = np.copy(BXtemp[-2, :])
self.BY[:, :-2] = 0.5 * (BYtemp[:, 1:-1] + BYtemp[:, :-2])
self.BY[:, -2] = np.copy(BYtemp[:, -2])
self.C44_Eff[:-2, :-2] = 4. / (
(1. / self.C44[:-2, :-2]) + (1. / self.C44[1:-1, :-2]) + (1. / self.C44[:-2, 1:-1]) + (
1. / self.C44[1:-1, 1:-1]))
self.Eta_vs = np.zeros((MRI, NRI), dtype=np.float32)
self.Eta_ss = np.zeros((MRI, NRI), dtype=np.float32)
self.Eta_vs = self.Eta_v + 2 * self.Eta_s
self.Eta_ss[:-2, :-2] = 4. / (
(1. / self.Eta_s[:-2, :-2]) + (1. / self.Eta_s[1:-1, :-2]) + (1. / self.Eta_s[:-2, 1:-1]) + (
1. / self.Eta_s[1:-1, 1:-1]))
def applyBoundaries(self):
Materials = copy.deepcopy(self.simPack.Materials)
MRI, NRI = np.shape(self.simPack.Simulation.Im)
APARA = 0.015
Top = int(self.simPack.Simulation.TapGrid[0])
Bottom = int(self.simPack.Simulation.TapGrid[1])
Left = int(self.simPack.Simulation.TapGrid[2])
Right = int(self.simPack.Simulation.TapGrid[3])
_Top = Top - np.arange(0, Top)
_Bottom = np.arange(MRI - Bottom + 1, MRI) - MRI + Bottom - 1
_Left = Left - np.arange(0, Left)
_Right = np.arange(NRI - Right + 1, NRI) - NRI + Right - 1
Boundaries = self.simPack.Boundary
for boundary in Boundaries:
if boundary.Name == "Top" and boundary.BC == BC.AirLayer:
self.BX[0:Top, :] = 0.0
self.BY[0:Top, :] = 0.0
self.C11[0:Top, :] = 0.0
self.C12[0:Top, :] = 0.0
self.C22[0:Top, :] = 0.0
self.C44_Eff[0:Top, :] = 0.0
self.Eta_vs[0:Top, :] = 0.0
self.Eta_ss[0:Top, :] = 0.0
if boundary.Name == "Bottom" and boundary.BC == BC.AirLayer:
self.BX[MRI - Bottom:, :] = 0.0
self.BY[MRI - Bottom:, :] = 0.0
self.C11[MRI - Bottom:, :] = 0.0
self.C12[MRI - Bottom:, :] = 0.0
self.C22[MRI - Bottom:, :] = 0.0
self.C44_Eff[MRI - Bottom:, :] = 0.0
self.Eta_vs[MRI - Bottom:, :] = 0.0
self.Eta_ss[MRI - Bottom:, :] = 0.0
if boundary.Name == "Left" and boundary.BC == BC.AirLayer:
self.BX[:, 0:Left] = 0.0
self.BY[:, 0:Left] = 0.0
self.C11[:, 0:Left] = 0.0
self.C12[:, 0:Left] = 0.0
self.C22[:, 0:Left] = 0.0
self.C44_Eff[:, 0:Left] = 0.0
self.Eta_vs[:, 0:Left] = 0.0
self.Eta_ss[:, 0:Left] = 0.0
if boundary.Name == "Right" and boundary.BC == BC.AirLayer:
self.BX[:, NRI - Right:] = 0.0
self.BY[:, NRI - Right:] = 0.0
self.C11[:, NRI - Right:] = 0.0
self.C12[:, NRI - Right:] = 0.0
self.C22[:, NRI - Right:] = 0.0
self.C44_Eff[:, NRI - Right:] = 0.0
self.Eta_vs[:, NRI - Right:] = 0.0
self.Eta_ss[:, NRI - Right:] = 0.0
for j in range(0, NRI):
self.ABS[0:Top, j] = np.exp(- ((APARA * _Top) ** 2))
self.ABS[MRI - Bottom + 1:, j] = np.exp(- ((APARA * _Bottom) ** 2))
for i in range(0, MRI):
self.ABS[i, 0:Left] = np.exp(- ((APARA * _Left) ** 2))
self.ABS[i, NRI - Right + 1:] = np.exp(- ((APARA * _Right) ** 2))
def sourceSetup(self):
longitudinal = self.simPack.Source.Longitudinal
shear = self.simPack.Source.Shear
Pressure = self.simPack.Source.Pressure
Displacement = self.simPack.Source.Displacement
if Pressure and not Displacement:
self.typeSource = 0
elif not Pressure and Displacement:
self.typeSource = 1
elif Pressure and Displacement:
self.typeSource = 2
else:
self.typeSource = 0
if longitudinal and not shear:
self.typeWave = np.int32(0)
elif not longitudinal and shear:
self.typeWave = np.int32(1)
elif longitudinal and shear:
self.typeWave = np.int32(2)
else:
self.typeWave = np.int32(0)
self.Amplitude = self.simPack.Signal.Amplitude
XL = self.simPack.Inspection.XL
YL = self.simPack.Inspection.YL
NX = np.size(XL, 0)
IsPZT = self.simPack.Transducers[0].PZT
size1 = int((self.simPack.Simulation.TapGrid[0]))
size2 = int((self.simPack.Simulation.TapGrid[1]))
for IT in range(-size1, 0):
for m in range(0, NX):
xl = int(XL[m, 0])
yl = int(YL[m, 0])
self.BX[xl + IT, yl] = 1.0 / pzt.Rho if IsPZT else 0
self.BY[xl + IT, yl] = 1.0 / pzt.Rho if IsPZT else 0
self.C11[xl + IT, yl] = pzt.C11 if IsPZT else 0
self.C12[xl + IT, yl] = pzt.C12 if IsPZT else 0
self.C22[xl + IT, yl] = pzt.C22 if IsPZT else 0
self.C44[xl + IT, yl] = pzt.C44 if IsPZT else 0
if self.simPack.Inspection.Name == "Transmission":
for IT in range(1, size2):
for m in range(0, NX):
xl = int(XL[m, 1])
yl = int(YL[m, 1])
self.BX[xl + IT, yl] = 1.0 / pzt.Rho if IsPZT else 0
self.BY[xl + IT, yl] = 1.0 / pzt.Rho if IsPZT else 0
self.C11[xl + IT, yl] = pzt.C11 if IsPZT else 0
self.C12[xl + IT, yl] = pzt.C12 if IsPZT else 0
self.C22[xl + IT, yl] = pzt.C22 if IsPZT else 0
self.C44[xl + IT, yl] = pzt.C44 if IsPZT else 0
def simSetup(self):
self.MRI, self.NRI = np.shape(self.simPack.Simulation.Im)
XL = self.simPack.Inspection.XL
YL = self.simPack.Inspection.YL
self.NX = np.size(XL, 0)
self.XL = np.copy(np.int32(XL[:, 0]))
self.YL = np.copy(np.int32(YL[:, 0]))
if self.simPack.Transducers[0].Window and not self.simPack.Transducers[0].PointSource:
self.win = np.float32(1.0 - np.cos(2 * np.pi * np.arange(0, self.NX) / (self.NX + 1)))
else:
self.win = np.ones((self.NX,), dtype=np.float32)
def initFieldsCL(self):
self.Txx_buf = cl.Buffer(self.ctx, self.mf.READ_WRITE | self.mf.COPY_HOST_PTR, hostbuf=self.Txx)
self.Tyy_buf = cl.Buffer(self.ctx, self.mf.READ_WRITE | self.mf.COPY_HOST_PTR, hostbuf=self.Tyy)
self.Txy_buf = cl.Buffer(self.ctx, self.mf.READ_WRITE | self.mf.COPY_HOST_PTR, hostbuf=self.Txy)
self.Vx_buf = cl.Buffer(self.ctx, self.mf.READ_WRITE | self.mf.COPY_HOST_PTR, hostbuf=self.Vx)
self.Vy_buf = cl.Buffer(self.ctx, self.mf.READ_WRITE | self.mf.COPY_HOST_PTR, hostbuf=self.Vy)
self.DVx_buf = cl.Buffer(self.ctx, self.mf.READ_WRITE | self.mf.COPY_HOST_PTR, hostbuf=self.DVx)
self.DVy_buf = cl.Buffer(self.ctx, self.mf.READ_WRITE | self.mf.COPY_HOST_PTR, hostbuf=self.DVy)
self.ABS_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.ABS)
self.BX_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.BX)
self.BY_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.BY)
self.C11_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.C11)
self.C12_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.C12)
self.C44_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.C44_Eff)
self.ETA_vs_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.Eta_vs)
self.ETA_s_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.Eta_s)
self.ETA_ss_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.Eta_ss)
self.XL_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.XL)
self.YL_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.YL)
self.WIN_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.win)
if not self.simPack.Inspection.Name == "Tomography":
self._Receiver = True
XL = self.simPack.Inspection.XL
YL = self.simPack.Inspection.YL
self.receiver_buf = cl.Buffer(self.ctx, self.mf.WRITE_ONLY | self.mf.COPY_HOST_PTR,
hostbuf=self.receiver_signals)
self.XXL = np.copy(np.int32(XL[:, 1]))
self.YYL = np.copy(np.int32(YL[:, 1]))
self.XXL_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.XXL)
self.YYL_buf = cl.Buffer(self.ctx, self.mf.READ_ONLY | self.mf.COPY_HOST_PTR, hostbuf=self.YYL)
self.program = cl.Program(self.ctx, self.kernel()).build()
def receiverSetup(self):
TimeSteps = int(self.simPack.Simulation.TimeSteps)
N_IR = np.size(self.simPack.Inspection.IR, 1)
self.receiver_signals = np.zeros((TimeSteps, N_IR - 1), dtype=np.float32)
# @blab+
def clEnqueue(self, Var, Var_buf):
# external access to Txx .. for snapshoting needs enqueueing of CL buffers
cl.enqueue_copy(self.queue, Var, Var_buf)
def receivers(self, Var, Var_buf):
if self._Receiver:
self.program.Receiver_EFIT2D(self.queue, (self.NX,), None, self.Txx_buf, self.receiver_buf,
np.int32(self.n),
self.XXL_buf, self.YYL_buf).wait()
else:
cl.enqueue_copy(self.queue, Var, Var_buf)
self.receiver_signals[self.n, :] = self.simPack.Inspection.getReceivers(Var)
def saveOutput(self):
if self._Receiver:
cl.enqueue_copy(self.queue, self.receiver_signals, self.receiver_buf).wait()
def receiversSerial(self, Var):
self.receiver_signals[self.n, :] = self.simPack.Inspection.getReceivers(Var)
def run(self):
y = np.float32(self.source[self.n])
self.program.Velocity_EFIT2D_Voigt(self.queue, (self.NRI, self.MRI,), None,
self.Txx_buf, self.Txy_buf, self.Tyy_buf,
self.Vx_buf, self.Vy_buf, self.DVx_buf, self.DVy_buf,
self.BX_buf, self.BY_buf, self.ABS_buf, self.ddx).wait()
if self.typeSource == 1 or self.typeSource == 2:
self.program.Source_Vel_EFIT2D(self.queue, (self.NX,), None,
self.Vx_buf,
self.Vy_buf,
self.XL_buf,
self.YL_buf,
y, self.typeWave, self.WIN_buf).wait()
self.program.Stress_EFIT2D_Voigt(self.queue, (self.NRI, self.MRI,), None,
self.Txx_buf, self.Txy_buf, self.Tyy_buf,
self.Vx_buf, self.Vy_buf, self.DVx_buf, self.DVy_buf,
self.C11_buf, self.C12_buf, self.C44_buf, self.ETA_vs_buf,
self.ETA_s_buf, self.ETA_ss_buf, self.ABS_buf).wait()
if self.typeSource == 0 or self.typeSource == 2:
self.program.Source_EFIT2D(self.queue, (self.NX,), None,
self.Txx_buf,
self.Tyy_buf,
self.Txy_buf,
self.XL_buf,
self.YL_buf,
y, self.typeWave, self.WIN_buf).wait()
self.receivers(self.Txx, self.Txx_buf)
def runGL(self):
cl.enqueue_copy(self.queue, self.Vx, self.Vx_buf)
cl.enqueue_copy(self.queue, self.Vy, self.Vy_buf)
self.SV = np.sqrt(self.Vx ** 2 + self.Vy ** 2)
value = np.max(np.abs(self.SV))
self.max_value = self.max_value if self.max_value >= value else value
self.SV = 20. * np.log10(np.abs(self.SV) / self.max_value + 1e-40)
def runGLSerial(self, step=50):
vx = np.copy(self.Vx)
vy = np.copy(self.Vy)
self.SV = np.sqrt(vx ** 2 + vy ** 2)
value = np.max(np.abs(self.SV))
self.max_value = self.max_value if self.max_value >= value else value
self.SV = 20. * np.log10(np.abs(self.SV) / self.max_value + 1e-40)
def runSerial(self):
y = np.float32(self.source[self.n])
self.Vx, self.Vy, self.DVx, self.DVy = EFIT2DCython.velocityVoigt(self.Txx, self.Txy, self.Tyy,
self.Vx, self.Vy, self.DVx, self.DVy,
self.BX, self.BY, self.ABS, self.ddx,
self.dt)
if self.typeSource == 1 or self.typeSource == 2:
self.Vx, self.Vy = EFIT2DCython.sourceVel(self.Vx, self.Vy, self.XL, self.YL, y, self.typeWave, self.win,
self.dtdxx)
self.Txx, self.Tyy, self.Txy = EFIT2DCython.stressVoigt(self.Txx, self.Txy, self.Tyy,
self.Vx, self.Vy, self.DVx, self.DVy,
self.C11, self.C12, self.C44_Eff,
self.Eta_vs, self.Eta_s, self.Eta_ss, self.ABS,
self.dtx)
if self.typeSource == 0 or self.typeSource == 2:
self.Txx, self.Tyy = EFIT2DCython.sourceStress(self.Txx, self.Tyy, self.XL, self.YL, y, self.typeWave,
self.win, self.dtdxx)
self.receiversSerial(self.Txx)
def kernel(self):
macro = """
#define MRI %s
#define NRI %s
#define ind(i, j) ( ( (i)*NRI) + (j) )
#define dtx %gf
#define dtdxx %gf
#define Stencil 2
#define NX %s
#define dt %gf
""" % (str(self.MRI), str(self.NRI), self.dtx, self.dtdxx,
str(self.NX), self.dt)
return macro + """
__kernel void Receiver_EFIT2D(__global float *Buffer, __global float *receiver, const int t,
__global const int *XXL, __global const int *YYL){
__private float _tmp = 0.0f;
for (int i=0; i<get_global_size(0); ++i)
{
_tmp += Buffer[ind(XXL[i],YYL[i])];
}
receiver[t] = _tmp/(float)get_global_size(0);
}
__kernel void Source_Vel_EFIT2D(__global float *Vx, __global float *Vy,
__global const int *XL, __global const int *YL,
const float source, const int Stype, __global const float *win){
uint m = get_global_id(0);
int ix = XL[m];
int iy = YL[m];
if (Stype ==0){
Vx[ind(ix,iy)] -= (source*dtdxx)*win[m];
}
else if (Stype ==1){
Vy[ind(ix,iy)] -= (source*dtdxx)*win[m];
}
else if (Stype ==2){
Vx[ind(ix,iy)] -= (source*dtdxx)*win[m];
Vy[ind(ix,iy)] -= (source*dtdxx)*win[m];
}
}
__kernel void Source_EFIT2D(__global float *Txx, __global float *Tyy, __global float *Txy,
__global const int *XL, __global const int *YL,
const float source, const int Stype,__global const float *win){
uint m = get_global_id(0);
int ix = XL[m];
int iy = YL[m];
if (Stype ==0){
Txx[ind(ix,iy)] -= (source*dtdxx)*win[m];
//Txy[ind(ix,iy)] = 0;
}
else if (Stype ==1){
Tyy[ind(ix,iy)] -= (source*dtdxx)*win[m];
//Txy[ind(ix,iy)] = 0;
}
else if (Stype ==2){
Txx[ind(ix,iy)] -= (source*dtdxx)*win[m];
Tyy[ind(ix,iy)] -= (source*dtdxx)*win[m];
//Txy[ind(ix,iy)] = 0;
}
}
__kernel void Velocity_EFIT2D_Voigt( __global float *Txx, __global float *Txy, __global float *Tyy,
__global float *vx, __global float *vy,
__global float *dvx, __global float *dvy,
__global const float *BX, __global const float *BY,
__global const float *ABS, const float ddx){
int j = get_global_id(0);
int i = get_global_id(1);
if (i < MRI-1 && j>0){
dvx[ind(i,j)] = (BX[ind(i,j)]*ddx)*( Txx[ind(i+1,j)] - Txx[ind(i,j)] + Txy[ind(i,j)] - Txy[ind(i,j-1)] );
vx[ind(i,j)] += dt*dvx[ind(i,j)];
}
if (i > 0 && j< NRI-1){
dvy[ind(i,j)] = (BY[ind(i,j)]*ddx)*( Txy[ind(i,j)] - Txy[ind(i-1,j)] + Tyy[ind(i,j+1)] - Tyy[ind(i,j)] );
vy[ind(i,j)] += dt*dvy[ind(i,j)];
}
barrier(CLK_GLOBAL_MEM_FENCE);
// Apply absorbing boundary conditions
vx[ind(i,j)] *= ABS[ind(i,j)];
vy[ind(i,j)] *= ABS[ind(i,j)];
}
__kernel void Stress_EFIT2D_Voigt(__global float *Txx, __global float *Txy, __global float *Tyy,
__global float *vx, __global float *vy,
__global float *dvx, __global float *dvy,
__global const float *C11, __global const float *C12, __global const float *C44,
__global const float *ETA_VS, __global const float *ETA_S, __global const float *ETA_SS,
__global const float *ABS) {
int j = get_global_id(0);
int i = get_global_id(1);
if (i>0 && j>0 ){
Txx[ind(i,j)] += ( ( C11[ind(i,j)]* dtx )*(vx[ind(i,j)] - vx[ind(i-1,j)]) +
( C12[ind(i,j)]* dtx )*(vy[ind(i,j)] - vy[ind(i,j-1)]) +
( (ETA_VS[ind(i,j)])*dtx) * (dvx[ind(i,j)] - dvx[ind(i-1,j)]) +
( (ETA_S [ind(i,j)])*dtx) * (dvy[ind(i,j)] - dvy[ind(i,j-1)]) );
Tyy[ind(i,j)] += ( ( C12[ind(i,j)]* dtx )*(vx[ind(i,j)] - vx[ind(i-1,j)]) +
( C11[ind(i,j)]* dtx )*(vy[ind(i,j)] - vy[ind(i,j-1)]) +
( (ETA_VS[ind(i,j)])*dtx) *(dvy[ind(i,j)] - dvy[ind(i,j-1)]) +
( (ETA_S [ind(i,j)])*dtx) * (dvx[ind(i,j)] - dvx[ind(i-1,j)]) );
}
if (i<MRI-1 && j<NRI-1){
Txy[ind(i,j)] += ( ( C44[ind(i,j)] * dtx )*(vx[ind(i,j+1)] - vx[ind(i,j)] + vy[ind(i+1,j)] - vy[ind(i,j)]) +
( ETA_SS[ind(i,j)] * dtx )*(dvx[ind(i,j+1)] - dvx[ind(i,j)] + dvy[ind(i+1,j)] - dvy[ind(i,j)] ) );
}
barrier(CLK_GLOBAL_MEM_FENCE);
Txx[ind(i,j)] *= ABS[ind(i,j)];
Tyy[ind(i,j)] *= ABS[ind(i,j)];
Txy[ind(i,j)] *= ABS[ind(i,j)];
}
"""