void GameScene::guiUpdate(float dt)
{
// Points handling
pointsLabel->setString("Score: " + std::to_string(points));
pointsLabel->setPosition(10 + pointsLabel->getBoundingBox().size.width / 2, 20);
// Bars update
mHpBarFill->setTextureRect(cocos2d::Rect(0.f, 0.f,
_MAX(_MIN(mPlayer->getHp() / (float)Entities::maxHP * mHpBarBorder->getBoundingBox().size.width,
mHpBarBorder->getBoundingBox().size.width), 0),
28.f));
mRageBarFill->setTextureRect(cocos2d::Rect(0.f, 0.f,
_MAX(_MIN(mPlayer->getRage() / (float)Entities::rageCharging * mRageBarBorder->getBoundingBox().size.width,
mRageBarBorder->getBoundingBox().size.width), 0),
28.f));
// Change difficulty if condition true
switch (currentDifficulty)
{
case Enemies::EASY: {
if (points >= Enemies::MEDIUM_CONDITION)
currentDifficulty = Enemies::MEDIUM;
break;
}
case Enemies::MEDIUM: {
if (points >= Enemies::HARD_CONDITION)
currentDifficulty = Enemies::HARD;
break;
}
}
Entities::playerShootingFreq = _MAX(0.1f, 0.2f - (float)((points / 2000) / 100.f));
}
void InitADERDG(ADERDG* adg, double xmin, double xmax)
{
adg->xmin = xmin;
adg->xmax = xmax;
for(int i = 0; i < _NBFACES; i++) {
double xh = (double)i / _NBELEMS_IN;
double x = stretching(xh);
adg->face[i] = xmin + x * (xmax - xmin);
//printf("i=%d x=%f\n",i, adg->face[i]);
}
adg->dx = 0.0;
for(int ie = 1; ie <= _NBELEMS_IN; ie++) {
adg->dx = _MAX(adg->dx, adg->face[ie] - adg->face[ie - 1]);
}
adg->cfl = _CFL;
adg->dt = adg->cfl * adg->dx; // TODO: put the velocity
adg->ncfl = 0;
for(int ie = 1; ie <= _NBELEMS_IN; ie++) {
adg->cell_level[ie] =
(int) (log(adg->dx / (adg->face[ie] - adg->face[ie-1])) / log(2));
adg->ncfl = _MAX ( adg->ncfl , adg->cell_level[ie] );
}
// convention: first and last cell are among the biggest elements
adg->cell_level[0] = 0;
adg->cell_level[_NBELEMS_IN + 1] = 0;
printf("found %d cfl levels\n", adg->ncfl);
adg->dt_small = adg->dt / (1 << adg->ncfl);
printf("small dt=%f max cell size=%f\n", adg->dt_small, adg->dx);
for(int ie = 1; ie <= _NBELEMS_IN; ie++) {
for(int j = 0;j < _NGLOPS; j++) {
double h = adg->face[ie] - adg->face[ie-1];
double x = adg->face[ie-1]
+ h * gauss_lob_point[gauss_lob_offset[_D] + j];
ExactSol(x, 0, adg->wnow[ie][j]);
}
}
for(int ie = 1;ie <= _NBELEMS; ie++){
for(int j = 0;j < _NGLOPS; j++){
for(int iv = 0; iv < _M; iv++){
adg->wnext[ie][j][iv] = adg->wnow[ie][j][iv];
}
}
}
for(int i=0;i<_NBFACES;i++){
adg->face_level[i]=_MAX(adg->cell_level[i],adg->cell_level[i+1]);
}
}
bool TimeManager::IsDiffDay20(time_t uPrevANSITime, time_t uNextANSITime)
{
time_t tTime1 = uPrevANSITime - 72000;
time_t tTime2 = uNextANSITime - 72000;
tTime1 = _MAX(tTime1, 0);
tTime2 = _MAX(tTime2, 0);
return IsDiffDay00(tTime1, tTime2);
}
bool TimeManager::IsDiffDay18(time_t uPrevANSITime, time_t uNextANSITime)
{
//都减去18小时，判断是否跨0点
time_t tTime1 = uPrevANSITime - 64800;
time_t tTime2 = uNextANSITime - 64800;
tTime1 = _MAX(tTime1, 0);
tTime2 = _MAX(tTime2, 0);
return IsDiffDay00(tTime1, tTime2);
}
BOOL
q_layoutStructNextEncode(StructLayout layout) {
unsigned int desired_align;
const char* type;
// add size of previous layout
if (layout->prev_type) {
type = q_skipVarQualifiersEncode(layout->prev_type);
if (*type == Q_C_BFLD) while (isdigit(*++type)); // empty loop
layout->offset += q_sizeOfTypeEncode(type) * kBitsPerUnit;
}
if (*layout->cur_type == Q_C_STRUCT_E) return NO;
layout->cur_type = q_skipVarNameEncode(layout->cur_type);
type = q_skipVarQualifiersEncode(layout->cur_type);
if (*type == Q_C_BFLD) while (isdigit(*++type)); // empty loop
desired_align = q_alignOfTypeEncode(type) * kBitsPerUnit;
layout->align = _MAX(layout->align, desired_align);
if (layout->offset % desired_align != 0)
// skip space before this field
// bump the cumulative size to multiple of field alignment
layout->offset = _ROUND(layout->offset, desired_align);
layout->prev_type = layout->cur_type;
layout->cur_type = q_skipTypeSpecEncode(layout->cur_type);
return YES;
}
/**
* @function GUI_W_UsrEntryDelete
* @brief Delete the selected string part (if any), or the car corresponding to insert line position
* @param void *_g_obj: generic object
* @return none
*/
void GUI_W_UsrEntryDelete(g_obj_st *obj) {
uint16_t from, to;
usr_entry_st *entry;
if(obj != NULL && obj->draw == EntryDraw) {
entry = (usr_entry_st *) obj->obj;
if(entry->bEditable) {
from = _MIN(entry->cursStart, entry->cursStop);
to = _MAX(entry->cursStart, entry->cursStop);
/*delecte the selection, or car located before the insert line*/
StrDelete(entry->buffer, entry->sizeMax, from, to);
/*if there was no user selection, decrease the position of the insert bar*/
if(from == to && from > 0) {
entry->cursStart--;
entry->cursStop--;
}
/*always ensure that the resulting insert line is bounded into the string size*/
entry->len = gstrlen(entry->buffer);
if(entry->len < entry->cursStart) entry->cursStart = entry->len;
/*after a deletion, set insert line to the lower cursor; since insert line -> entry->cursStop = entry->cursStart*/
entry->cursStart = _MIN(entry->cursStart, entry->cursStop);
entry->cursStop = entry->cursStart;
/*force refresh*/
GUI_ObjSetNeedRefresh(obj, true);
}
}
}
static void collect_results(struct benchmark_config *config,
struct work_queue *queue, unsigned long long start,
unsigned long long elapsed)
{
int i;
unsigned long long sum[2] = { 0, 0 }, min[2] = { ULONG_MAX, ULONG_MAX };
unsigned long long max[2] = { 0, 0 }, avg[2];
for (i = 0; i < config->num_works; i++) {
struct work *work = work_queue_pop(queue);
sum[0] += work->elapsed[0];
sum[1] += work->elapsed[1];
min[0] = _MIN(min[0], work->elapsed[0]);
min[1] = _MIN(min[1], work->elapsed[1]);
max[0] = _MAX(max[0], work->elapsed[0]);
max[1] = _MAX(max[1], work->elapsed[1]);
if (config->verbose > 1) {
printf(
"%lld.%03lld %lld.%03lld %lld.%03lld %lld.%03lld\n",
(work->start[0] - start) / 1000000,
(work->start[0] - start) / 1000 % 1000,
work->elapsed[0] / 1000000,
work->elapsed[0] / 1000 % 1000,
(work->start[1] - start) / 1000000,
(work->start[1] - start) / 1000 % 1000,
work->elapsed[1] / 1000000,
work->elapsed[1] / 1000 % 1000);
}
free(work);
}
avg[0] = sum[0] / config->num_works;
avg[1] = sum[1] / config->num_works;
if (config->verbose > 0) {
printf(
"# %lld.%03lld %lld.%03lld %lld.%03lld %lld.%03lld %lld.%03lld %lld.%03lld\n",
avg[0] / 1000000, avg[0] / 1000 % 1000,
min[0] / 1000000, min[0] / 1000 % 1000,
max[0] / 1000000, max[0] / 1000 % 1000,
avg[1] / 1000000, avg[1] / 1000 % 1000,
min[1] / 1000000, min[1] / 1000 % 1000,
max[1] / 1000000, max[1] / 1000 % 1000);
}
}
static void benchmark(void)
{
int i;
long sum = 0;
long min_ms = LONG_MAX, max_ms = 0, avg_ms;
pthread_t *tid;
struct benchmark_thread_data *data;
#ifdef CHUNKD_BENCHMARK
stc_init();
#endif
tid = xmalloc(sizeof(tid[0]) * threads);
data = xmalloc(sizeof(data[0]) * threads);
for (i = 0; i < threads; i++) {
data[i].id = i;
xpthread_create(&tid[i], NULL, benchmark_thread, &data[i]);
}
wait_threads(tid, threads);
#define _MIN(a, b) ((a) < (b) ? (a) : (b))
#define _MAX(a, b) ((a) < (b) ? (b) : (a))
for (i = 0; i < threads; i++) {
long ms = data[i].time_ms;
sum += ms;
min_ms = _MIN(min_ms, ms);
max_ms = _MAX(max_ms, ms);
}
avg_ms = sum / threads;
printf("%d %ld.%03ld %ld.%03ld %ld.%03ld\n", threads,
avg_ms / 1000, avg_ms % 1000,
min_ms / 1000, min_ms % 1000,
max_ms / 1000, max_ms % 1000);
if (verbose) {
unsigned long long total_bytes;
unsigned long long bytes_per_msec;
total_bytes = value_length;
total_bytes *= threads;
total_bytes *= requests;
bytes_per_msec = total_bytes / avg_ms;
printf("Throughput: %llu KB/sec\n",
bytes_per_msec * 1000UL / 1024UL);
}
free(data);
free(tid);
}
void
q_layoutStructEndEncode(StructLayout layout, int* size, unsigned int* align) {
if (layout->cur_type && *layout->cur_type == Q_C_STRUCT_E) {
layout->align = _MAX(1, layout->align);
layout->offset = _ROUND(layout->offset, layout->align);
layout->cur_type = nil;
}
if (size) *size = layout->offset / kBitsPerUnit;
if (align) *align = layout->align / kBitsPerUnit;
}
CBotNeuralNet :: CBotNeuralNet ( unsigned short int numinputs, unsigned short int numhiddenlayers,
unsigned short int neuronsperhiddenlayer, unsigned short int numoutputs,
ga_nn_value learnrate)
{
register unsigned short int i;
register unsigned short int j;
m_pOutputs = new CLogisticalNeuron[numoutputs];
m_pHidden = new CLogisticalNeuron*[numhiddenlayers];
m_layerinput = new ga_nn_value[_MAX(numinputs,neuronsperhiddenlayer)];
m_layeroutput = new ga_nn_value[_MAX(numoutputs,_MAX(numinputs,neuronsperhiddenlayer))];
for ( j = 0; j < numhiddenlayers; j ++ )
{
m_pHidden[j] = new CLogisticalNeuron[neuronsperhiddenlayer];
for ( i = 0; i < neuronsperhiddenlayer; i ++ )
{
if ( j == 0 )
m_pHidden[j][i].init(numinputs,learnrate);
else
m_pHidden[j][i].init(neuronsperhiddenlayer,learnrate);
}
}
for ( i = 0; i < numoutputs; i ++ )
m_pOutputs[i].init(neuronsperhiddenlayer,learnrate);
//m_transferFunction = new CSigmoidTransfer ();
m_numInputs = numinputs;
m_numOutputs = numoutputs;
m_numHidden = neuronsperhiddenlayer;
m_numHiddenLayers = numhiddenlayers;
}
void
_hbuf_reserve(struct hdfs_heap_buf *h, size_t space)
{
int remain, toalloc;
remain = h->size - h->used;
if ((size_t)remain >= space)
return;
toalloc = _MAX(32, space - remain + 16);
h->buf = realloc(h->buf, h->size + toalloc);
ASSERT(h->buf);
h->size += toalloc;
}
void
q_layoutStructBeginEncode(const char* type, StructLayout layout) {
const char* ntype;
if (*type++ != Q_C_STRUCT_B) q_throwError(er2, type);
ntype = type;
while (*ntype != Q_C_STRUCT_E && *ntype != Q_C_STRUCT_B && *ntype != Q_C_UNION_B && *ntype++ != '='); // empty loop
if (*(ntype - 1) == '=')
type = ntype;
layout->cur_type = type;
layout->prev_type = nil;
layout->offset = 0;
layout->align = _MAX(kBitsPerUnit, STRUCT_SIZE_BOUNDARY);
}
void NavDialog::CreateBitmap()
{
RECT r;
::GetClientRect(_hSelf, &r);
const int maxLines = _MAX(m_view[0].m_lines, m_view[1].m_lines);
const int maxHeight = (r.bottom - r.top) - 2 * cSpace - 2;
int reductionRatio = m_compact ? maxLines / maxHeight : 0;
if (reductionRatio && (maxLines % maxHeight))
++reductionRatio;
m_view[0].create(m_clr, reductionRatio);
m_view[1].create(m_clr, reductionRatio);
SetScalingFactor();
}
bool IsPointInContour(int x, int y, TVAZone* contour)
{
int count = 0;
for (int i = 0; i < contour->NumPoints; i++)
{
int j = (i+1)%contour->NumPoints;
//горизонтальный отрезок.
if ((int)contour->Points[i].Y == (int)contour->Points[j].Y)
continue;
//отрезок выше от луча
else if (contour->Points[i].Y > y && contour->Points[j].Y > y)
continue;
//отрезок ниже от луча
else if (contour->Points[i].Y < y && contour->Points[j].Y < y)
continue;
// отрезок справа от луча
else if ((int)_MIN(contour->Points[i].X,contour->Points[j].X) > x)
continue;
//нижняя граница на луче
else if ((int)_MIN(contour->Points[i].Y,contour->Points[j].Y) == (int)y)
continue;
//верхняя граница отрезка на луче
else if ((int)_MAX(contour->Points[i].Y,contour->Points[j].Y) == (int)y)
{
double x1 = contour->Points[i].Y > contour->Points[j].Y ? contour->Points[i].X : contour->Points[j].X;
if (x > x1)
count++;
}
else
{
double k,b;
k = (contour->Points[j].Y - contour->Points[i].Y) / (contour->Points[j].X - contour->Points[i].X);
b = contour->Points[j].Y - k*contour->Points[j].X;
// точка пересечения.
double t;
t = (y - b) / k;
if (t < x )
count++;
}
}
return count & 1;
}
void NavDialog::SetScalingFactor()
{
m_view[0].m_lines = ::SendMessage(m_view[0].m_hView, SCI_GETLINECOUNT, 0, 0);
m_view[1].m_lines = ::SendMessage(m_view[1].m_hView, SCI_GETLINECOUNT, 0, 0);
m_view[0].m_firstVisible = ::SendMessage(m_view[0].m_hView, SCI_GETFIRSTVISIBLELINE, 0, 0);
m_view[1].m_firstVisible = ::SendMessage(m_view[1].m_hView, SCI_GETFIRSTVISIBLELINE, 0, 0);
m_maxBmpLines = _MAX(m_view[0].maxBmpLines(), m_view[1].maxBmpLines());
m_syncView = (m_maxBmpLines == m_view[0].maxBmpLines()) ? &m_view[0] : &m_view[1];
RECT r;
::GetClientRect(_hSelf, &r);
m_navViewWidth = ((r.right - r.left) - 3 * cSpace - 4) / 2;
m_navHeight = (r.bottom - r.top) - 2 * cSpace - 2;
m_pixelsPerLine = m_navHeight / m_maxBmpLines;
if (m_pixelsPerLine == 0)
{
m_pixelsPerLine = 1;
ShowScroller(r);
}
else
{
if (m_pixelsPerLine > 5)
m_pixelsPerLine = 5;
m_navHeight = m_pixelsPerLine * m_maxBmpLines;
if (m_hScroll)
::ShowScrollBar(m_hScroll, SB_CTL, FALSE);
}
updateScroll();
updateDockingDlg();
if (isVisible())
::InvalidateRect(_hSelf, NULL, TRUE);
}
/* accepts password file */
static int
find_strchr (char *username, char *file)
{
FILE *fd;
char *pos;
char line[5 * 1024];
unsigned int i;
fd = fopen (file, "r");
if (fd == NULL)
{
fprintf (stderr, "Cannot open file '%s'\n", file);
return -1;
}
while (fgets (line, sizeof (line), fd) != NULL)
{
/* move to first ':' */
i = 0;
while ((line[i] != ':') && (line[i] != '\0') && (i < sizeof (line)))
{
i++;
}
if (strncmp (username, line, _MAX (i, strlen (username))) == 0)
{
/* find the index */
pos = strrchr (line, ':');
pos++;
fclose (fd);
return atoi (pos);
}
}
fclose (fd);
return -1;
}
/**
* @function GUI_W_UsrEntryGetSelection
* @brief copy the selected string part into a given buffer
* @param void *_g_obj: generic object
* @param uint8_t *buffer: user buffer; will be \0 terminated
* @param uint16_t bufSize: size of the user buffer
* @param void *_g_obj: generic object
* @return none
*/
void GUI_W_UsrEntryGetSelection(g_obj_st *obj, uint8_t *buffer, uint16_t bufSize) {
usr_entry_st *entry;
uint16_t pos, end;
if(obj != NULL && obj->draw == EntryDraw && buffer != NULL && bufSize > 0) {
entry = (usr_entry_st *) obj->obj;
/*copy the highlighted text of the entry in the user buffer*/
pos = _MIN(entry->cursStart, entry->cursStop);
end = _MAX(entry->cursStart, entry->cursStop);
while(bufSize > 1 && pos < end) {
*buffer = entry->buffer[pos];
pos++;
buffer++;
bufSize--;
}
/*always be sure that str is '\0' terminated*/
*buffer = 0;
}
}
int
q_sizeOfTypeEncode(const char* type) {
type = q_skipVarNameEncode(type);
type = q_skipVarQualifiersEncode(type);
switch (*type) {
case Q_C_ID:
return sizeof(id);
case Q_C_CLASS:
return sizeof(Class);
case Q_C_SEL:
return sizeof(SEL);
case Q_C_CHR:
return sizeof(char);
case Q_C_UCHR:
return sizeof(unsigned char);
case Q_C_SHT:
return sizeof(short);
case Q_C_USHT:
return sizeof(unsigned short);
case Q_C_INT:
return sizeof(int);
case Q_C_UINT:
return sizeof(unsigned int);
case Q_C_LNG:
return sizeof(long);
case Q_C_ULNG:
return sizeof(unsigned long);
case Q_C_FLT:
return sizeof(float);
case Q_C_DBL:
return sizeof(double);
case Q_C_VOID:
return 0;
case Q_C_PTR:
case Q_C_CHARPTR:
return sizeof(char*);
case Q_C_ARY_B: {
int len = atoi(type + 1);
while (isdigit(*++type)); // empty loop
return len * q_sizeOfTypeEncode(type);
}
case Q_C_BFLD:
return atoi(type + 1) / kBitsPerUnit;
case Q_C_STRUCT_B: {
struct _StructLayout layout;
unsigned int size;
q_layoutStructBeginEncode(type, &layout);
while (q_layoutStructNextEncode(&layout)); // empty loop
q_layoutStructEndEncode(&layout, &size, nil);
return size;
}
case Q_C_UNION_B: {
int max_size = 0;
while (*type != Q_C_UNION_E && *type++ != '='); // empty loop
while (*type != Q_C_UNION_E) {
type = q_skipVarNameEncode(type);
max_size = _MAX(max_size, q_sizeOfTypeEncode(type));
type = q_skipTypeSpecEncode(type);
}
return max_size;
}
default:
q_throwError(er1, type);
}
return 0;
}
// Symmetric tridiagonal QL algorithm.
static void tql2(double V[3][3], double d[3], double e[3]) {
// This is derived from the Algol procedures tql2, by
// Bowdler, Martin, Reinsch, and Wilkinson, Handbook for
// Auto. Comp., Vol.ii-Linear Algebra, and the corresponding
// Fortran subroutine in EISPACK.
e[0] = e[1]; e[1] = e[2]; e[2] = 0.0;
double f = 0.0;
double tst1 = 0.0;
double eps = pow(2.0,-52.0);
for (int l = 0; l < n; l++) {
// Find small subdiagonal element
tst1 = _MAX(tst1, fabs(d[l]) + fabs(e[l]));
int m = l;
while (m < n) {
if ( fabs(e[m]) <= eps*tst1 ) break;
m++;
}
// If m == l, d[l] is an eigenvalue,
// otherwise, iterate.
if (m > l) {
int iter = 0;
do {
iter = iter + 1; // (Could check iteration count here.)
// Compute implicit shift
double g = d[l];
double p = (d[l+1] - g) / (2.0 * e[l]);
double r = hypot2(p,1.0);
if (p < 0) {
r = -r;
}
d[l] = e[l] / (p + r);
d[l+1] = e[l] * (p + r);
double dl1 = d[l+1];
double h = g - d[l];
for (int i = l+2; i < n; i++) {
d[i] -= h;
}
f = f + h;
// Implicit QL transformation.
p = d[m];
double c = 1.0;
double c2 = c;
double c3 = c;
double el1 = e[l+1];
double s = 0.0;
double s2 = 0.0;
for (int i = m-1; i >= l; --i) {
c3 = c2;
c2 = c;
s2 = s;
g = c * e[i];
h = c * p;
r = hypot2(p,e[i]);
e[i+1] = s * r;
s = e[i] / r;
c = p / r;
p = c * d[i] - s * g;
d[i+1] = h + s * (c * g + s * d[i]);
// Accumulate transformation.
for (int k = 0; k < n; ++k) {
h = V[k][i+1];
V[k][i+1] = s * V[k][i] + c * h;
V[k][i] = c * V[k][i] - s * h;
}
}
p = -s * s2 * c3 * el1 * e[l] / dl1;
e[l] = s * p;
d[l] = c * p;
// Check for convergence.
} while (fabs(e[l]) > eps*tst1);
}
d[l] = d[l] + f;
e[l] = 0.0;
}
// Sort eigenvalues and corresponding vectors.
for (int i = 0; i < n-1; i++) {
int k = i;
double p = d[i];
for (int j = i+1; j < n; j++) {
if (d[j] < p) {
k = j;
p = d[j];
}
}
if (k != i) {
d[k] = d[i];
d[i] = p;
for (int j = 0; j < n; j++) {
p = V[j][i];
V[j][i] = V[j][k];
V[j][k] = p;
}
}
}
}
// Sort the bunch in ascending z (ct) order, and return
// a vector of iterators which point to the equal-spaced
// bin boundaries defines by zmin to zmax in steps of dz
//
// Returns the number of particles removed from tails
// i.e. z<zmin || z>=zmax
//
// hdp contains the derivative of the distribution
// calculated using the Savitzky-Golay filter c
// If c is empty, then the derivative will be zero.
size_t ParticleBinList(ParticleBunch& bunch, double zmin, double zmax, size_t nbins,
vector<ParticleBunch::iterator>& pbins,
vector<double>& hd, vector<double>& hdp, vector<double>* c)
{
//cout << "In ParticleBinList" << endl;
//cout << zmin << "\t" << zmax << "\t" << nbins << endl;
double dz = (zmax-zmin)/double(nbins);
vector<ParticleBunch::iterator> bins;
vector<double> hbins(nbins,0);
bins.reserve(nbins+1);
bunch.SortByCT();
size_t lost = TruncateZ(bunch,zmin,zmax);
ParticleBunch::iterator p = bunch.begin();
bins.push_back(p);
double z=zmin;
double total=0;
size_t n;
for(n=0; n<nbins; n++)
{
z+=dz;
while(p!=bunch.end() && p->ct()<z)
{
total++;
hbins[n]++;
p++;
}
bins.push_back(p);
}
if(p!=bunch.end())
{
#ifdef ENABLE_MPI
cerr << "bad slicing in rank: " << MPI::COMM_WORLD.Get_rank() << endl;
#endif
#ifndef ENABLE_MPI
cerr << "bad slicing" << endl;
#endif
cerr << "z = " << z << " ct = " << p->ct() << " zmax = " << zmax << endl;
//Dump out the bad bunch
/*
ofstream* badbunch = new ofstream("badbunch.bunch");
bunch.Output(*badbunch);
badbunch->close();
delete badbunch;
cerr << "Output of the current bunch is to badbunch.bunch" << endl;
*/
#ifndef ENABLE_MPI
abort();
#endif
#ifdef ENABLE_MPI
MPI::COMM_WORLD.Abort(1);
#endif
}
// bins.push_back(p); // should be end()
// normalise distribution
// and apply filter
vector<double> fbins(nbins,0);
vector<double> fpbins(nbins,0);
double a = 1/total/dz;
int w = c ? (c->size()-1)/2 : 0;
size_t m;
for(n=0; n<nbins; n++)
{
fbins[n] = hbins[n]*a;
if(c)
//for(m=_MAX(0,int(n)-w); m<=_MIN(nbins,int(n)+w); m++)// ERROR! m can be set to nbins -> out of range!
for(m=_MAX(0,int(n)-w); m<_MIN(nbins,size_t(n)+w); m++) // This needs to be checked!
{
fpbins[n] += hbins[m]*(*c)[m-n+w]*a;
}
}
pbins.swap(bins);
hd.swap(fbins);
hdp.swap(fpbins);
return lost;
}
void process_main()
{
create("start_proc");
g_sim.m_num_CSIM_process++;
fprintf(stderr, "started simulation.\n");
// simulation progress verbose
if (g_cfg.sim_show_progress)
process_sim_progress();
// router
for (unsigned int i=0; i<g_Router_vec.size(); i++) {
process_router(g_Router_vec[i]);
}
// input/output NI
for (unsigned int n=0; n<g_NIInput_vec.size(); n++) {
switch (g_cfg.NIin_type) {
case NI_INPUT_TYPE_PER_PC:
process_NI_input(g_NIInput_vec[n], 0);
break;
case NI_INPUT_TYPE_PER_VC:
for (int NI_vc=0; NI_vc<g_cfg.router_num_vc; NI_vc++)
process_NI_input(g_NIInput_vec[n], NI_vc);
break;
default:
assert(0);
}
}
for (unsigned int n=0; n<g_NIOutput_vec.size(); n++) {
process_NI_output(g_NIOutput_vec[n]);
}
// profile
if (g_cfg.profile_perf || g_cfg.profile_power) {
if (g_cfg.profile_interval_cycle)
process_profile_cycle();
else
process_profile_instr();
}
#ifdef LINK_DVS
// link-dvs
process_link_dvs_link_speedup();
process_link_dvs_link_slowdown();
process_link_dvs_set();
#endif
// injection
switch (g_cfg.wkld_type) {
case WORKLOAD_TRIPS_TRACE:
case WORKLOAD_TILED_CMP_TRACE:
case WORKLOAD_TILED_CMP_VALUE_TRACE:
case WORKLOAD_SNUCA_CMP_VALUE_TRACE:
process_parse_trace();
break;
case WORKLOAD_SYNTH_SPATIAL:
case WORKLOAD_SYNTH_TRAFFIC_MATRIX:
for (unsigned int c=0; c<g_Core_vec.size(); c++)
process_gen_synth_traffic(c);
break;
default:
assert(0);
}
// control simulation for warmup and finalize
process_control_sim();
g_ev_sim_done->wait();
// Now the simulation is done.
fprintf(stderr, "finished simulation at clk=%.0lf.\n", simtime());
// Find the simulation end time
g_sim.m_end_time = time((time_t *)NULL);
g_sim.m_elapsed_time = _MAX(g_sim.m_end_time - g_sim.m_start_time, 1);
#ifdef _DEBUG_ROUTER_PROCESS
printf("PROCESS COMPLETE: process_main()\n");
#endif
}
/* Parses the tpasswd files, in order to verify the given
* username/password pair.
*/
int
verify_passwd (char *conffile, char *tpasswd, char *username,
const char *passwd)
{
FILE *fd;
char line[5 * 1024];
unsigned int i;
gnutls_datum_t g, n;
int iindex;
char *p, *pos;
iindex = find_strchr (username, tpasswd);
if (iindex == -1)
{
fprintf (stderr, "Cannot find '%s' in %s\n", username, tpasswd);
return -1;
}
fd = fopen (conffile, "r");
if (fd == NULL)
{
fprintf (stderr, "Cannot find %s\n", conffile);
return -1;
}
do
{
p = fgets (line, sizeof (line) - 1, fd);
}
while (p != NULL && atoi (p) != iindex);
if (p == NULL)
{
fprintf (stderr, "Cannot find entry in %s\n", conffile);
return -1;
}
line[sizeof (line) - 1] = 0;
fclose (fd);
if ((iindex = read_conf_values (&g, &n, line)) < 0)
{
fprintf (stderr, "Cannot parse conf file '%s'\n", conffile);
return -1;
}
fd = fopen (tpasswd, "r");
if (fd == NULL)
{
fprintf (stderr, "Cannot open file '%s'\n", tpasswd);
return -1;
}
while (fgets (line, sizeof (line), fd) != NULL)
{
/* move to first ':'
* This is the actual verifier.
*/
i = 0;
while ((line[i] != ':') && (line[i] != '\0') && (i < sizeof (line)))
{
i++;
}
if (strncmp (username, line, _MAX (i, strlen (username))) == 0)
{
char *verifier_pos, *salt_pos;
pos = strchr (line, ':');
fclose (fd);
if (pos == NULL)
{
fprintf (stderr, "Cannot parse conf file '%s'\n", conffile);
return -1;
}
pos++;
verifier_pos = pos;
/* Move to the salt */
pos = strchr (pos, ':');
if (pos == NULL)
{
fprintf (stderr, "Cannot parse conf file '%s'\n", conffile);
return -1;
}
pos++;
salt_pos = pos;
return _verify_passwd_int (username, passwd,
verifier_pos, salt_pos, &g, &n);
}
}
fclose (fd);
return -1;
}
int
crypt_int (const char *username, const char *passwd, int salt_size,
char *tpasswd_conf, char *tpasswd, int uindex)
{
FILE *fd;
char *cr;
gnutls_datum_t g, n;
char line[5 * 1024];
char *p, *pp;
int iindex;
char tmpname[1024];
fd = fopen (tpasswd_conf, "r");
if (fd == NULL)
{
fprintf (stderr, "Cannot find %s\n", tpasswd_conf);
return -1;
}
do
{ /* find the specified uindex in file */
p = fgets (line, sizeof (line) - 1, fd);
iindex = atoi (p);
}
while (p != NULL && iindex != uindex);
if (p == NULL)
{
fprintf (stderr, "Cannot find entry in %s\n", tpasswd_conf);
return -1;
}
line[sizeof (line) - 1] = 0;
fclose (fd);
if ((iindex = read_conf_values (&g, &n, line)) < 0)
{
fprintf (stderr, "Cannot parse conf file '%s'\n", tpasswd_conf);
return -1;
}
cr = _srp_crypt (username, passwd, salt_size, &g, &n);
if (cr == NULL)
{
fprintf (stderr, "Cannot _srp_crypt()...\n");
return -1;
}
else
{
/* delete previous entry */
struct stat st;
FILE *fd2;
int put;
if (strlen (tpasswd) > sizeof (tmpname) + 5)
{
fprintf (stderr, "file '%s' is tooooo long\n", tpasswd);
return -1;
}
strcpy (tmpname, tpasswd);
strcat (tmpname, ".tmp");
if (stat (tmpname, &st) != -1)
{
fprintf (stderr, "file '%s' is locked\n", tpasswd);
return -1;
}
if (filecopy (tpasswd, tmpname) != 0)
{
fprintf (stderr, "Cannot copy '%s' to '%s'\n", tpasswd, tmpname);
return -1;
}
fd = fopen (tpasswd, "w");
if (fd == NULL)
{
fprintf (stderr, "Cannot open '%s' for write\n", tpasswd);
remove (tmpname);
return -1;
}
fd2 = fopen (tmpname, "r");
if (fd2 == NULL)
{
fprintf (stderr, "Cannot open '%s' for read\n", tmpname);
remove (tmpname);
return -1;
}
put = 0;
do
{
p = fgets (line, sizeof (line) - 1, fd2);
if (p == NULL)
break;
pp = strchr (line, ':');
if (pp == NULL)
continue;
if (strncmp
(p, username,
_MAX (strlen (username), (unsigned int) (pp - p))) == 0)
{
put = 1;
fprintf (fd, "%s:%s:%u\n", username, cr, iindex);
}
else
{
fputs (line, fd);
}
}
while (1);
if (put == 0)
{
fprintf (fd, "%s:%s:%u\n", username, cr, iindex);
}
fclose (fd);
fclose (fd2);
remove (tmpname);
}
return 0;
}
/**
* @function EntryDraw
* @brief user entry draw function
* @param void *_g_obj: generic object
* @param void *_obj: frame object
* @return none
*/
static void EntryDraw(void *_g_obj, void *_obj) {
g_obj_st *g_obj;
usr_entry_st *entry;
uint8_t glyph;
uint16_t ii, selStart, selStop;
coord_t x, xInsertLine, xMin, yMin, xMax, yMax;
rect_st rec;
color_t colBack, colText;
/*retreive generic & specific object*/
if(_g_obj != NULL && _obj != NULL) {
g_obj = (g_obj_st *) _g_obj;
entry = (usr_entry_st*) _obj;
/*P2D configuration*/
P2D_SetDisplayMode(DISPLAY_SOLID);
P2D_SetLineType(LINE_SOLID);
SetFont(entry->font);
if(GUI_ObjIsDisabled(g_obj)) {
colBack = GetColor(G_COL_BACKGROUND);
colText = GetColor(G_COL_D_TEXT);
}
else {
colBack = GetColor(G_COL_E_BACKGROUND);
colText = entry->colText;
}
/*retrieve text coord*/
GetTextCoords(&(g_obj->rec), &xMin, &yMin, &xMax, &yMax);
/*display usr_entry glyphs, one by one*/
selStart = _MIN(entry->cursStart, entry->cursStop);
selStop = _MAX(entry->cursStart, entry->cursStop);
x = xMin;
xInsertLine = x;
ii = entry->offsetDisplay;
while(x < xMax && entry->buffer[ii] != 0) {
/*select the color of the text, according to the user selection (i.e. reverse colors if the current glyph is a part of the user selection)*/
if(entry->bEditable == false || selStart == selStop || ii < selStart || ii >= selStop) {
P2D_SetColors(colText, colBack);
}
else {
P2D_SetColors(colBack, GetColor(G_COL_SPECIAL));
}
/*display the glyph*/
glyph = entry->buffer[ii];
P2D_PutGlyph(x, yMin, glyph);
x += P2D_GetGlyphWidth(glyph);
/*if the car corresponds to the user selection bar, store its coord*/
if(ii == entry->cursStop - 1) xInsertLine = x;
/*next car*/
ii++;
}
/*clear from last car to the end of the entry*/
P2D_SetColors(colBack, colBack);
(void) P2D_CoordToRect(&rec, x, yMin, xMax, yMax);
P2D_FillRect(&rec);
/*clear between text and object rect (1 px width)*/
(void) P2D_CoordToRect(&rec, xMin-1, yMin-1, xMax+1, yMax+1);
P2D_Rect(&rec);
/*display the insert line*/
if(entry->bEditable && entry->bBlink) {
P2D_SetColor(colText);
P2D_Line(xInsertLine, yMin, xInsertLine, yMax);
}
/*object rect*/
P2D_SetColor(GetColor(G_COL_LOWER_REC));
P2D_Rect(&(g_obj->rec));
}
}
int main(int argc, char **argv) {
int c, i, mu, status;
int ispin, icol, isc;
int n_c = 3;
int n_s = 4;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int grid_size[4];
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix, iy, is, it, i3;
int sl0, sl1, sl2, sl3, have_source_flag=0;
int source_proc_coords[4], lsl0, lsl1, lsl2, lsl3;
int check_residuum = 0;
unsigned int VOL3, V5;
int do_gt = 0;
int full_orbit = 0;
int smear_source = 0;
char filename[200], source_filename[200], source_filename_write[200];
double ratime, retime;
double plaq_r=0., plaq_m=0., norm, norm2;
double spinor1[24];
double *gauge_qdp[4], *gauge_field_timeslice=NULL, *gauge_field_smeared=NULL;
double _1_2_kappa, _2_kappa, phase;
FILE *ofs;
int mu_trans[4] = {3, 0, 1, 2};
int threadid, nthreads;
int timeslice, source_timeslice;
char rng_file_in[100], rng_file_out[100];
int *source_momentum=NULL;
int source_momentum_class = -1;
int source_momentum_no = 0;
int source_momentum_runs = 1;
int imom;
int num_gpu_on_node=0, rank;
int source_location_5d_iseven;
int convert_sign=0;
#ifdef HAVE_QUDA
int rotate_gamma_basis = 1;
#else
int rotate_gamma_basis = 0;
#endif
omp_lock_t *lck = NULL, gen_lck[1];
int key = 0;
/****************************************************************************/
/* for smearing parallel to inversion */
double *smearing_spinor_field[] = {NULL,NULL};
int dummy_flag = 0;
/****************************************************************************/
/****************************************************************************/
#if (defined HAVE_QUDA) && (defined MULTI_GPU)
int x_face_size, y_face_size, z_face_size, t_face_size, pad_size;
#endif
/****************************************************************************/
/************************************************/
int qlatt_nclass;
int *qlatt_id=NULL, *qlatt_count=NULL, **qlatt_rep=NULL, **qlatt_map=NULL;
double **qlatt_list=NULL;
/************************************************/
/************************************************/
double boundary_condition_factor;
int boundary_condition_factor_set = 0;
/************************************************/
//#ifdef MPI
// kernelPackT = true;
//#endif
/***********************************************
* QUDA parameters
***********************************************/
#ifdef HAVE_QUDA
QudaPrecision cpu_prec = QUDA_DOUBLE_PRECISION;
QudaPrecision cuda_prec = QUDA_DOUBLE_PRECISION;
QudaPrecision cuda_prec_sloppy = QUDA_SINGLE_PRECISION;
QudaGaugeParam gauge_param = newQudaGaugeParam();
QudaInvertParam inv_param = newQudaInvertParam();
#endif
while ((c = getopt(argc, argv, "soch?vgf:p:b:S:R:")) != -1) {
switch (c) {
case 'v':
g_verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'c':
check_residuum = 1;
fprintf(stdout, "# [invert_dw_quda] will check residuum again\n");
break;
case 'p':
n_c = atoi(optarg);
fprintf(stdout, "# [invert_dw_quda] will use number of colors = %d\n", n_c);
break;
case 'o':
full_orbit = 1;
fprintf(stdout, "# [invert_dw_quda] will invert for full orbit, if source momentum set\n");
case 's':
smear_source = 1;
fprintf(stdout, "# [invert_dw_quda] will smear the sources if they are read from file\n");
break;
case 'b':
boundary_condition_factor = atof(optarg);
boundary_condition_factor_set = 1;
fprintf(stdout, "# [invert_dw_quda] const. boundary condition factor set to %e\n", boundary_condition_factor);
break;
case 'S':
convert_sign = atoi(optarg);
fprintf(stdout, "# [invert_dw_quda] using convert sign %d\n", convert_sign);
break;
case 'R':
rotate_gamma_basis = atoi(optarg);
fprintf(stdout, "# [invert_dw_quda] rotate gamma basis %d\n", rotate_gamma_basis);
break;
case 'h':
case '?':
default:
usage();
break;
}
}
// get the time stamp
g_the_time = time(NULL);
/**************************************
* set the default values, read input
**************************************/
if(filename_set==0) strcpy(filename, "cvc.input");
if(g_proc_id==0) fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
#ifdef MPI
#ifdef HAVE_QUDA
grid_size[0] = g_nproc_x;
grid_size[1] = g_nproc_y;
grid_size[2] = g_nproc_z;
grid_size[3] = g_nproc_t;
fprintf(stdout, "# [] g_nproc = (%d,%d,%d,%d)\n", g_nproc_x, g_nproc_y, g_nproc_z, g_nproc_t);
initCommsQuda(argc, argv, grid_size, 4);
#else
MPI_Init(&argc, &argv);
#endif
#endif
#if (defined PARALLELTX) || (defined PARALLELTXY)
EXIT_WITH_MSG(1, "[] Error, 2-dim./3-dim. MPI-Version not yet implemented");
#endif
// some checks on the input data
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stderr, "[invert_dw_quda] Error, T and L's must be set\n");
usage();
}
// set number of openmp threads
// initialize MPI parameters
mpi_init(argc, argv);
// the volume of a timeslice
VOL3 = LX*LY*LZ;
V5 = T*LX*LY*LZ*L5;
g_kappa5d = 0.5 / (5. + g_m5);
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] kappa5d = %e\n", g_kappa5d);
fprintf(stdout, "# [%2d] parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] L5 = %3d\n",\
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, L5);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "[invert_dw_quda] Error from init_geometry\n");
EXIT(1);
}
geometry();
if( init_geometry_5d() != 0 ) {
fprintf(stderr, "[invert_dw_quda] Error from init_geometry_5d\n");
EXIT(2);
}
geometry_5d();
/**************************************
* initialize the QUDA library
**************************************/
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] initializing quda\n");
#ifdef HAVE_QUDA
// cudaGetDeviceCount(&num_gpu_on_node);
if(g_gpu_per_node<0) {
if(g_cart_id==0) fprintf(stderr, "[] Error, number of GPUs per node not set\n");
EXIT(106);
} else {
num_gpu_on_node = g_gpu_per_node;
}
#ifdef MPI
rank = comm_rank();
#else
rank = 0;
#endif
g_gpu_device_number = rank % num_gpu_on_node;
fprintf(stdout, "# [] process %d/%d uses device %d\n", rank, g_cart_id, g_gpu_device_number);
initQuda(g_gpu_device_number);
#endif
/**************************************
* prepare the gauge field
**************************************/
// read the gauge field from file
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
if(strcmp( gaugefilename_prefix, "identity")==0 ) {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] Setting up unit gauge field\n");
for(ix=0;ix<VOLUME; ix++) {
for(mu=0;mu<4;mu++) {
_cm_eq_id(g_gauge_field+_GGI(ix,mu));
}
}
} else if(strcmp( gaugefilename_prefix, "random")==0 ) {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] Setting up random gauge field with seed = %d\n", g_seed);
init_rng_state(g_seed, &g_rng_state);
random_gauge_field(g_gauge_field, 1.);
plaquette(&plaq_m);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
check_error(write_lime_gauge_field(filename, plaq_m, Nconf, 64), "write_lime_gauge_field", NULL, 12);
} else {
if(g_gauge_file_format == 0) {
// ILDG
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "# Reading gauge field from file %s\n", filename);
status = read_lime_gauge_field_doubleprec(filename);
} else if(g_gauge_file_format == 1) {
// NERSC
sprintf(filename, "%s.%.5d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "# Reading gauge field from file %s\n", filename);
status = read_nersc_gauge_field(g_gauge_field, filename, &plaq_r);
//status = read_nersc_gauge_field_3x3(g_gauge_field, filename, &plaq_r);
}
if(status != 0) {
fprintf(stderr, "[invert_dw_quda] Error, could not read gauge field");
EXIT(12);
}
}
#ifdef MPI
xchange_gauge();
#endif
// measure the plaquette
plaquette(&plaq_m);
if(g_cart_id==0) fprintf(stdout, "# Measured plaquette value: %25.16e\n", plaq_m);
if(g_cart_id==0) fprintf(stdout, "# Read plaquette value : %25.16e\n", plaq_r);
#ifndef HAVE_QUDA
if(N_Jacobi>0) {
#endif
// allocate the smeared / qdp ordered gauge field
alloc_gauge_field(&gauge_field_smeared, VOLUMEPLUSRAND);
for(i=0;i<4;i++) {
gauge_qdp[i] = gauge_field_smeared + i*18*VOLUME;
}
#ifndef HAVE_QUDA
}
#endif
#ifdef HAVE_QUDA
// transcribe the gauge field
omp_set_num_threads(g_num_threads);
#pragma omp parallel for private(ix,iy,mu)
for(ix=0;ix<VOLUME;ix++) {
iy = g_lexic2eot[ix];
for(mu=0;mu<4;mu++) {
_cm_eq_cm(gauge_qdp[mu_trans[mu]]+18*iy, g_gauge_field+_GGI(ix,mu));
}
}
// multiply timeslice T-1 with factor of -1 (antiperiodic boundary condition)
if(g_proc_coords[0]==g_nproc_t-1) {
if(!boundary_condition_factor_set) boundary_condition_factor = -1.;
fprintf(stdout, "# [] process %d multiplies gauge-field timeslice T_global-1 with boundary condition factor %e\n", g_cart_id,
boundary_condition_factor);
omp_set_num_threads(g_num_threads);
#pragma omp parallel for private(ix,iy)
for(ix=0;ix<VOL3;ix++) {
iix = (T-1)*VOL3 + ix;
iy = g_lexic2eot[iix];
_cm_ti_eq_re(gauge_qdp[mu_trans[0]]+18*iy, -1.);
}
}
// QUDA precision parameters
switch(g_cpu_prec) {
case 0: cpu_prec = QUDA_HALF_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] CPU prec = half\n"); break;
case 1: cpu_prec = QUDA_SINGLE_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] CPU prec = single\n"); break;
case 2: cpu_prec = QUDA_DOUBLE_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] CPU prec = double\n"); break;
default: cpu_prec = QUDA_DOUBLE_PRECISION; break;
}
switch(g_gpu_prec) {
case 0: cuda_prec = QUDA_HALF_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] GPU prec = half\n"); break;
case 1: cuda_prec = QUDA_SINGLE_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] GPU prec = single\n"); break;
case 2: cuda_prec = QUDA_DOUBLE_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] GPU prec = double\n"); break;
default: cuda_prec = QUDA_DOUBLE_PRECISION; break;
}
switch(g_gpu_prec_sloppy) {
case 0: cuda_prec_sloppy = QUDA_HALF_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] GPU sloppy prec = half\n"); break;
case 1: cuda_prec_sloppy = QUDA_SINGLE_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] GPU sloppy prec = single\n"); break;
case 2: cuda_prec_sloppy = QUDA_DOUBLE_PRECISION; if(g_cart_id==0) fprintf(stdout, "# [] GPU sloppy prec = double\n"); break;
default: cuda_prec_sloppy = QUDA_SINGLE_PRECISION; break;
}
// QUDA gauge parameters
gauge_param.X[0] = LX;
gauge_param.X[1] = LY;
gauge_param.X[2] = LZ;
gauge_param.X[3] = T;
inv_param.Ls = L5;
gauge_param.anisotropy = 1.0;
gauge_param.type = QUDA_WILSON_LINKS;
gauge_param.gauge_order = QUDA_QDP_GAUGE_ORDER;
gauge_param.t_boundary = QUDA_ANTI_PERIODIC_T;
gauge_param.cpu_prec = cpu_prec;
gauge_param.cuda_prec = cuda_prec;
gauge_param.reconstruct = QUDA_RECONSTRUCT_12;
gauge_param.cuda_prec_sloppy = cuda_prec_sloppy;
gauge_param.reconstruct_sloppy = QUDA_RECONSTRUCT_12;
gauge_param.gauge_fix = QUDA_GAUGE_FIXED_NO;
gauge_param.ga_pad = 0;
inv_param.sp_pad = 0;
inv_param.cl_pad = 0;
// For multi-GPU, ga_pad must be large enough to store a time-slice
#ifdef MULTI_GPU
x_face_size = inv_param.Ls * gauge_param.X[1]*gauge_param.X[2]*gauge_param.X[3]/2;
y_face_size = inv_param.Ls * gauge_param.X[0]*gauge_param.X[2]*gauge_param.X[3]/2;
z_face_size = inv_param.Ls * gauge_param.X[0]*gauge_param.X[1]*gauge_param.X[3]/2;
t_face_size = inv_param.Ls * gauge_param.X[0]*gauge_param.X[1]*gauge_param.X[2]/2;
pad_size = _MAX(x_face_size, y_face_size);
pad_size = _MAX(pad_size, z_face_size);
pad_size = _MAX(pad_size, t_face_size);
gauge_param.ga_pad = pad_size;
if(g_cart_id==0) printf("# [invert_dw_quda] pad_size = %d\n", pad_size);
#endif
// load the gauge field
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] loading gauge field\n");
loadGaugeQuda((void*)gauge_qdp, &gauge_param);
gauge_qdp[0] = NULL;
gauge_qdp[1] = NULL;
gauge_qdp[2] = NULL;
gauge_qdp[3] = NULL;
#endif
/*********************************************
* APE smear the gauge field
*********************************************/
if(N_Jacobi>0) {
memcpy(gauge_field_smeared, g_gauge_field, 72*VOLUMEPLUSRAND*sizeof(double));
fprintf(stdout, "# [invert_dw_quda] APE smearing gauge field with paramters N_APE=%d, alpha_APE=%e\n", N_ape, alpha_ape);
APE_Smearing_Step_threads(gauge_field_smeared, N_ape, alpha_ape);
xchange_gauge_field(gauge_field_smeared);
}
// allocate memory for the spinor fields
#ifdef HAVE_QUDA
no_fields = 3+2;
#else
no_fields = 6+2;
#endif
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND*L5);
smearing_spinor_field[0] = g_spinor_field[no_fields-2];
smearing_spinor_field[1] = g_spinor_field[no_fields-1];
switch(g_source_type) {
case 0:
case 5:
// the source locaton
sl0 = g_source_location / (LX_global*LY_global*LZ);
sl1 = ( g_source_location % (LX_global*LY_global*LZ) ) / ( LY_global*LZ);
sl2 = ( g_source_location % ( LY_global*LZ) ) / ( LZ);
sl3 = g_source_location % LZ;
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] global sl = (%d, %d, %d, %d)\n", sl0, sl1, sl2, sl3);
source_proc_coords[0] = sl0 / T;
source_proc_coords[1] = sl1 / LX;
source_proc_coords[2] = sl2 / LY;
source_proc_coords[3] = sl3 / LZ;
#ifdef MPI
MPI_Cart_rank(g_cart_grid, source_proc_coords, &g_source_proc_id);
#else
g_source_proc_id = 0;
#endif
have_source_flag = g_source_proc_id == g_cart_id;
lsl0 = sl0 % T;
lsl1 = sl1 % LX;
lsl2 = sl2 % LY;
lsl3 = sl3 % LZ;
if(have_source_flag) {
fprintf(stdout, "# [invert_dw_quda] process %d has the source at (%d, %d, %d, %d)\n", g_cart_id, lsl0, lsl1, lsl2, lsl3);
}
break;
case 2:
case 3:
case 4:
// the source timeslice
#ifdef MPI
source_proc_coords[0] = g_source_timeslice / T;
source_proc_coords[1] = 0;
source_proc_coords[2] = 0;
source_proc_coords[3] = 0;
MPI_Cart_rank(g_cart_grid, source_proc_coords, &g_source_proc_id);
have_source_flag = ( g_source_proc_id == g_cart_id );
source_timeslice = have_source_flag ? g_source_timeslice % T : -1;
#else
g_source_proc_id = 0;
have_source_flag = 1;
source_timeslice = g_source_timeslice;
#endif
break;
}
#ifdef HAVE_QUDA
/*************************************************************
* QUDA inverter parameters
*************************************************************/
inv_param.dslash_type = QUDA_DOMAIN_WALL_DSLASH;
if(strcmp(g_inverter_type_name, "cg") == 0) {
inv_param.inv_type = QUDA_CG_INVERTER;
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] using cg inverter\n");
} else if(strcmp(g_inverter_type_name, "bicgstab") == 0) {
inv_param.inv_type = QUDA_BICGSTAB_INVERTER;
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] using bicgstab inverter\n");
#ifdef MULTI_GPU
} else if(strcmp(g_inverter_type_name, "gcr") == 0) {
inv_param.inv_type = QUDA_GCR_INVERTER;
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] using gcr inverter\n");
#endif
} else {
if(g_cart_id==0) fprintf(stderr, "[invert_dw_quda] Error, unrecognized inverter type %s\n", g_inverter_type_name);
EXIT(123);
}
if(inv_param.inv_type == QUDA_CG_INVERTER) {
inv_param.solution_type = QUDA_MAT_SOLUTION;
inv_param.solve_type = QUDA_NORMEQ_PC_SOLVE;
} else if(inv_param.inv_type == QUDA_BICGSTAB_INVERTER) {
inv_param.solution_type = QUDA_MAT_SOLUTION;
inv_param.solve_type = QUDA_DIRECT_PC_SOLVE;
} else {
inv_param.solution_type = QUDA_MATPC_SOLUTION;
inv_param.solve_type = QUDA_DIRECT_PC_SOLVE;
}
inv_param.m5 = g_m5;
inv_param.kappa = 0.5 / (5. + inv_param.m5);
inv_param.mass = g_m0;
inv_param.tol = solver_precision;
inv_param.maxiter = niter_max;
inv_param.reliable_delta = reliable_delta;
#ifdef MPI
// domain decomposition preconditioner parameters
if(inv_param.inv_type == QUDA_GCR_INVERTER) {
if(g_cart_id == 0) printf("# [] settup DD parameters\n");
inv_param.gcrNkrylov = 15;
inv_param.inv_type_precondition = QUDA_MR_INVERTER;
inv_param.tol_precondition = 1e-6;
inv_param.maxiter_precondition = 200;
inv_param.verbosity_precondition = QUDA_VERBOSE;
inv_param.prec_precondition = cuda_prec_sloppy;
inv_param.omega = 0.7;
}
#endif
inv_param.matpc_type = QUDA_MATPC_EVEN_EVEN;
inv_param.dagger = QUDA_DAG_NO;
inv_param.mass_normalization = QUDA_KAPPA_NORMALIZATION; //;QUDA_MASS_NORMALIZATION;
inv_param.cpu_prec = cpu_prec;
inv_param.cuda_prec = cuda_prec;
inv_param.cuda_prec_sloppy = cuda_prec_sloppy;
inv_param.verbosity = QUDA_VERBOSE;
inv_param.preserve_source = QUDA_PRESERVE_SOURCE_NO;
inv_param.dirac_order = QUDA_DIRAC_ORDER;
#ifdef MPI
inv_param.preserve_dirac = QUDA_PRESERVE_DIRAC_YES;
inv_param.prec_precondition = cuda_prec_sloppy;
inv_param.gamma_basis = QUDA_DEGRAND_ROSSI_GAMMA_BASIS;
inv_param.dirac_tune = QUDA_TUNE_NO;
#endif
#endif
/*******************************************
* write initial rng state to file
*******************************************/
if( g_source_type==2 && g_coherent_source==2 ) {
sprintf(rng_file_out, "%s.0", g_rng_filename);
status = init_rng_stat_file (g_seed, rng_file_out);
if( status != 0 ) {
fprintf(stderr, "[invert_dw_quda] Error, could not write rng status\n");
EXIT(210);
}
} else if( (g_source_type==2 /*&& g_coherent_source==1*/) || g_source_type==3 || g_source_type==4) {
if( init_rng_state(g_seed, &g_rng_state) != 0 ) {
fprintf(stderr, "[invert_dw_quda] Error, could initialize rng state\n");
EXIT(211);
}
}
/*******************************************
* prepare locks for openmp
*******************************************/
nthreads = g_num_threads - 1;
lck = (omp_lock_t*)malloc(nthreads * sizeof(omp_lock_t));
if(lck == NULL) {
EXIT_WITH_MSG(97, "[invert_dw_quda] Error, could not allocate lck\n");
}
// init locks
for(i=0;i<nthreads;i++) {
omp_init_lock(lck+i);
}
omp_init_lock(gen_lck);
// check the source momenta
if(g_source_momentum_set) {
source_momentum = (int*)malloc(3*sizeof(int));
if(g_source_momentum[0]<0) g_source_momentum[0] += LX_global;
if(g_source_momentum[1]<0) g_source_momentum[1] += LY_global;
if(g_source_momentum[2]<0) g_source_momentum[2] += LZ_global;
fprintf(stdout, "# [invert_dw_quda] using final source momentum ( %d, %d, %d )\n", g_source_momentum[0], g_source_momentum[1], g_source_momentum[2]);
if(full_orbit) {
status = make_qcont_orbits_3d_parity_avg( &qlatt_id, &qlatt_count, &qlatt_list, &qlatt_nclass, &qlatt_rep, &qlatt_map);
if(status != 0) {
if(g_cart_id==0) fprintf(stderr, "\n[invert_dw_quda] Error while creating O_3-lists\n");
EXIT(4);
}
source_momentum_class = qlatt_id[g_ipt[0][g_source_momentum[0]][g_source_momentum[1]][g_source_momentum[2]]];
source_momentum_no = qlatt_count[source_momentum_class];
source_momentum_runs = source_momentum_class==0 ? 1 : source_momentum_no + 1;
if(g_cart_id==0) fprintf(stdout, "# [] source momentum belongs to class %d with %d members, which means %d runs\n",
source_momentum_class, source_momentum_no, source_momentum_runs);
}
}
if(g_source_type == 5) {
if(g_seq_source_momentum_set) {
if(g_seq_source_momentum[0]<0) g_seq_source_momentum[0] += LX_global;
if(g_seq_source_momentum[1]<0) g_seq_source_momentum[1] += LY_global;
if(g_seq_source_momentum[2]<0) g_seq_source_momentum[2] += LZ_global;
} else if(g_source_momentum_set) {
g_seq_source_momentum[0] = g_source_momentum[0];
g_seq_source_momentum[1] = g_source_momentum[1];
g_seq_source_momentum[2] = g_source_momentum[2];
}
fprintf(stdout, "# [invert_dw_quda] using final sequential source momentum ( %d, %d, %d )\n",
g_seq_source_momentum[0], g_seq_source_momentum[1], g_seq_source_momentum[2]);
}
/***********************************************
* loop on spin-color-index
***********************************************/
for(isc=g_source_index[0]; isc<=g_source_index[1]; isc++)
// for(isc=g_source_index[0]; isc<=g_source_index[0]; isc++)
{
ispin = isc / n_c;
icol = isc % n_c;
for(imom=0; imom<source_momentum_runs; imom++) {
/***********************************************
* set source momentum
***********************************************/
if(g_source_momentum_set) {
if(imom == 0) {
if(full_orbit) {
source_momentum[0] = 0;
source_momentum[1] = 0;
source_momentum[2] = 0;
} else {
source_momentum[0] = g_source_momentum[0];
source_momentum[1] = g_source_momentum[1];
source_momentum[2] = g_source_momentum[2];
}
} else {
source_momentum[0] = qlatt_map[source_momentum_class][imom-1] / (LY_global*LZ_global);
source_momentum[1] = ( qlatt_map[source_momentum_class][imom-1] % (LY_global*LZ_global) ) / LZ_global;
source_momentum[2] = qlatt_map[source_momentum_class][imom-1] % LZ_global;
}
if(g_cart_id==0) fprintf(stdout, "# [] run no. %d, source momentum (%d, %d, %d)\n",
imom, source_momentum[0], source_momentum[1], source_momentum[2]);
}
/***********************************************
* prepare the souce
***********************************************/
if(g_read_source == 0) { // create source
switch(g_source_type) {
case 0:
// point source
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] Creating point source\n");
for(ix=0;ix<L5*VOLUME;ix++) { _fv_eq_zero(g_spinor_field[0]+ix); }
if(have_source_flag) {
if(g_source_momentum_set) {
phase = 2*M_PI*( source_momentum[0]*sl1/(double)LX_global + source_momentum[1]*sl2/(double)LY_global + source_momentum[2]*sl3/(double)LZ_global );
g_spinor_field[0][_GSI(g_ipt[lsl0][lsl1][lsl2][lsl3]) + 2*(n_c*ispin+icol) ] = cos(phase);
g_spinor_field[0][_GSI(g_ipt[lsl0][lsl1][lsl2][lsl3]) + 2*(n_c*ispin+icol)+1] = sin(phase);
} else {
g_spinor_field[0][_GSI(g_ipt[lsl0][lsl1][lsl2][lsl3]) + 2*(n_c*ispin+icol) ] = 1.;
}
}
if(g_source_momentum_set) {
sprintf(source_filename, "%s.%.4d.t%.2dx%.2dy%.2dz%.2d.%.2d.qx%.2dqy%.2dqz%.2d",
filename_prefix, Nconf, sl0, sl1, sl2, sl3, n_c*ispin+icol, source_momentum[0], source_momentum[1], source_momentum[2]);
} else {
sprintf(source_filename, "%s.%.4d.t%.2dx%.2dy%.2dz%.2d.%.2d", filename_prefix, Nconf, sl0, sl1, sl2, sl3, n_c*ispin+icol);
}
#ifdef HAVE_QUDA
// set matpc_tpye
source_location_5d_iseven = ( (g_iseven[g_ipt[lsl0][lsl1][lsl2][lsl3]] && ispin<n_s/2) || (!g_iseven[g_ipt[lsl0][lsl1][lsl2][lsl3]] && ispin>=n_s/2) ) ? 1 : 0;
if(source_location_5d_iseven) {
inv_param.matpc_type = QUDA_MATPC_EVEN_EVEN;
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] matpc type is MATPC_EVEN_EVEN\n");
} else {
inv_param.matpc_type = QUDA_MATPC_ODD_ODD;
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] matpc type is MATPC_ODD_ODD\n");
}
#endif
break;
case 2:
// timeslice source
if(g_coherent_source==1) {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] Creating coherent timeslice source\n");
status = prepare_coherent_timeslice_source(g_spinor_field[0], gauge_field_smeared, g_coherent_source_base, g_coherent_source_delta, VOLUME, g_rng_state, 1);
if(status != 0) {
fprintf(stderr, "[invert_dw_quda] Error from prepare source, status was %d\n", status);
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 123);
MPI_Finalize();
#endif
exit(123);
}
check_error(prepare_coherent_timeslice_source(g_spinor_field[0], gauge_field_smeared, g_coherent_source_base, g_coherent_source_delta, VOLUME, g_rng_state, 1),
"prepare_coherent_timeslice_source", NULL, 123);
timeslice = g_coherent_source_base;
} else {
if(g_coherent_source==2) {
timeslice = (g_coherent_source_base+isc*g_coherent_source_delta)%T_global;
fprintf(stdout, "# [invert_dw_quda] Creating timeslice source\n");
check_error(prepare_timeslice_source(g_spinor_field[0], gauge_field_smeared, timeslice, VOLUME, g_rng_state, 1),
"prepare_timeslice_source", NULL, 123);
} else {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] Creating timeslice source\n");
check_error(prepare_timeslice_source(g_spinor_field[0], gauge_field_smeared, g_source_timeslice, VOLUME, g_rng_state, 1),
"prepare_timeslice_source", NULL, 124);
timeslice = g_source_timeslice;
}
}
if(g_source_momentum_set) {
sprintf(source_filename, "%s.%.4d.%.2d.%.5d.qx%.2dqy%.2dqz%.2d", filename_prefix, Nconf,
timeslice, isc, source_momentum[0], source_momentum[1], source_momentum[2]);
} else {
sprintf(source_filename, "%s.%.4d.%.2d.%.5d", filename_prefix, Nconf, timeslice, isc);
}
break;
case 3:
// timeslice sources for one-end trick (spin dilution)
fprintf(stdout, "# [invert_dw_quda] Creating timeslice source for one-end-trick\n");
check_error( prepare_timeslice_source_one_end(g_spinor_field[0], gauge_field_smeared, source_timeslice, source_momentum, isc%n_s, g_rng_state, \
( isc%n_s==(n_s-1) && imom==source_momentum_runs-1 )), "prepare_timeslice_source_one_end", NULL, 125 );
c = N_Jacobi > 0 ? isc%n_s + n_s : isc%n_s;
if(g_source_momentum_set) {
sprintf(source_filename, "%s.%.4d.%.2d.%.2d.qx%.2dqy%.2dqz%.2d", filename_prefix, Nconf,
g_source_timeslice, c, source_momentum[0], source_momentum[1], source_momentum[2]);
} else {
sprintf(source_filename, "%s.%.4d.%.2d.%.2d", filename_prefix, Nconf, g_source_timeslice, c);
}
break;
case 4:
// timeslice sources for one-end trick (spin and color dilution )
fprintf(stdout, "# [invert_dw_quda] Creating timeslice source for one-end-trick\n");
check_error(prepare_timeslice_source_one_end_color(g_spinor_field[0], gauge_field_smeared, source_timeslice, source_momentum,\
isc%(n_s*n_c), g_rng_state, ( isc%(n_s*n_c)==(n_s*n_c-1) && imom==source_momentum_runs-1 )), "prepare_timeslice_source_one_end_color", NULL, 126);
c = N_Jacobi > 0 ? isc%(n_s*n_c) + (n_s*n_c) : isc%(n_s*n_c);
if(g_source_momentum_set) {
sprintf(source_filename, "%s.%.4d.%.2d.%.2d.qx%.2dqy%.2dqz%.2d", filename_prefix, Nconf,
g_source_timeslice, c, source_momentum[0], source_momentum[1], source_momentum[2]);
} else {
sprintf(source_filename, "%s.%.4d.%.2d.%.2d", filename_prefix, Nconf, g_source_timeslice, c);
}
break;
case 5:
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] preparing sequential point source\n");
check_error( prepare_sequential_point_source (g_spinor_field[0], isc, sl0, g_seq_source_momentum,
smear_source, g_spinor_field[1], gauge_field_smeared), "prepare_sequential_point_source", NULL, 33);
sprintf(source_filename, "%s.%.4d.t%.2dx%.2d.y%.2d.z%.2d.%.2d.qx%.2dqy%.2dqz%.2d", filename_prefix2, Nconf,
sl0, sl1, sl2, sl3, isc, g_source_momentum[0], g_source_momentum[1], g_source_momentum[2]);
break;
default:
fprintf(stderr, "\nError, unrecognized source type\n");
exit(32);
break;
}
} else { // read source
switch(g_source_type) {
case 0: // point source
if(g_source_momentum_set) {
sprintf(source_filename, "%s.%.4d.t%.2dx%.2dy%.2dz%.2d.%.2d.qx%.2dqy%.2dqz%.2d", \
filename_prefix2, Nconf, sl0, sl1, sl2, sl3, isc, source_momentum[0], source_momentum[1], source_momentum[2]);
} else {
sprintf(source_filename, "%s.%.4d.t%.2dx%.2dy%.2dz%.2d.%.2d", filename_prefix2, Nconf, sl0, sl1, sl2, sl3, isc);
}
fprintf(stdout, "# [invert_dw_quda] reading source from file %s\n", source_filename);
check_error(read_lime_spinor(g_spinor_field[0], source_filename, 0), "read_lime_spinor", NULL, 115);
break;
case 2: // timeslice source
if(g_source_momentum_set) {
sprintf(source_filename, "%s.%.4d.%.2d.%.5d.qx%.2dqy%.2dqz%.2d", filename_prefix2, Nconf, g_source_timeslice,
isc, source_momentum[0], source_momentum[1], source_momentum[2]);
} else {
sprintf(source_filename, "%s.%.4d.%.2d.%.5d", filename_prefix2, Nconf, g_source_timeslice, isc);
}
fprintf(stdout, "# [invert_dw_quda] reading source from file %s\n", source_filename);
check_error(read_lime_spinor(g_spinor_field[0], source_filename, 0), "read_lime_spinor", NULL, 115);
break;
default:
check_error(1, "source type", NULL, 104);
break;
case -1: // timeslice source
sprintf(source_filename, "%s", filename_prefix2);
fprintf(stdout, "# [invert_dw_quda] reading source from file %s\n", source_filename);
check_error(read_lime_spinor(g_spinor_field[0], source_filename, 0), "read_lime_spinor", NULL, 115);
break;
}
} // of if g_read_source
if(g_write_source) {
check_error(write_propagator(g_spinor_field[0], source_filename, 0, g_propagator_precision), "write_propagator", NULL, 27);
}
/***********************************************************************************************
* here threads split:
***********************************************************************************************/
if(dummy_flag==0) strcpy(source_filename_write, source_filename);
memcpy((void*)(smearing_spinor_field[0]), (void*)(g_spinor_field[0]), 24*VOLUME*sizeof(double));
if(dummy_flag>0) {
// copy only if smearing has been done; otherwise do not copy, do not invert
if(g_cart_id==0) fprintf(stdout, "# [] copy smearing field -> g field\n");
memcpy((void*)(g_spinor_field[0]), (void*)(smearing_spinor_field[1]), 24*VOLUME*sizeof(double));
}
omp_set_num_threads(g_num_threads);
#pragma omp parallel private(threadid, _2_kappa, is, ix, iy, iix, ratime, retime) shared(key,g_read_source, smear_source, N_Jacobi, kappa_Jacobi, smearing_spinor_field, g_spinor_field, nthreads, convert_sign, VOLUME, VOL3, T, L5, isc, rotate_gamma_basis, g_cart_id) firstprivate(inv_param, gauge_param, ofs)
{
threadid = omp_get_thread_num();
if(threadid < nthreads) {
fprintf(stdout, "# [] proc%.2d thread%.2d starting source preparation\n", g_cart_id, threadid);
// smearing
if( ( !g_read_source || (g_read_source && smear_source ) ) && N_Jacobi > 0 ) {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] smearing source with N_Jacobi=%d, kappa_Jacobi=%e\n", N_Jacobi, kappa_Jacobi);
Jacobi_Smearing_threaded(gauge_field_smeared, smearing_spinor_field[0], smearing_spinor_field[1], kappa_Jacobi, N_Jacobi, threadid, nthreads);
}
/***********************************************
* create the 5-dim. source field
***********************************************/
if(convert_sign == 0) {
spinor_4d_to_5d_threaded(smearing_spinor_field[0], smearing_spinor_field[0], threadid, nthreads);
} else if(convert_sign == 1 || convert_sign == -1) {
spinor_4d_to_5d_sign_threaded(smearing_spinor_field[0], smearing_spinor_field[0], convert_sign, threadid, nthreads);
}
for(is=0; is<L5; is++) {
for(it=threadid; it<T; it+=nthreads) {
memcpy((void*)(g_spinor_field[0]+_GSI(g_ipt_5d[is][it][0][0][0])), (void*)(smearing_spinor_field[0]+_GSI(g_ipt_5d[is][it][0][0][0])), VOL3*24*sizeof(double));
}
}
// reorder, multiply with g2
for(is=0; is<L5; is++) {
for(it=threadid; it<T; it+=nthreads) {
for(i3=0; i3<VOL3; i3++) {
ix = (is*T+it)*VOL3 + i3;
_fv_eq_zero(smearing_spinor_field[1]+_GSI(ix));
}}}
if(rotate_gamma_basis) {
for(it=threadid; it<T; it+=nthreads) {
for(i3=0; i3<VOL3; i3++) {
ix = it * VOL3 + i3;
iy = lexic2eot_5d(0, ix);
_fv_eq_gamma_ti_fv(smearing_spinor_field[1]+_GSI(iy), 2, smearing_spinor_field[0]+_GSI(ix));
}}
for(it=threadid; it<T; it+=nthreads) {
for(i3=0; i3<VOL3; i3++) {
ix = it * VOL3 + i3;
iy = lexic2eot_5d(L5-1, ix);
_fv_eq_gamma_ti_fv(smearing_spinor_field[1]+_GSI(iy), 2, smearing_spinor_field[0]+_GSI(ix+(L5-1)*VOLUME));
}}
} else {
for(it=threadid; it<T; it+=nthreads) {
for(i3=0; i3<VOL3; i3++) {
ix = it * VOL3 + i3;
iy = lexic2eot_5d(0, ix);
_fv_eq_fv(smearing_spinor_field[1]+_GSI(iy), smearing_spinor_field[0]+_GSI(ix));
}}
for(it=threadid; it<T; it+=nthreads) {
for(i3=0; i3<VOL3; i3++) {
ix = it * VOL3 + i3;
iy = lexic2eot_5d(L5-1, ix);
_fv_eq_fv(smearing_spinor_field[1]+_GSI(iy), smearing_spinor_field[0]+_GSI(ix+(L5-1)*VOLUME));
}}
}
fprintf(stdout, "# [] proc%.2d thread%.2d finished source preparation\n", g_cart_id, threadid);
} else if(threadid == g_num_threads-1 && dummy_flag > 0) { // else branch on threadid
fprintf(stdout, "# [] proc%.2d thread%.2d starting inversion for dummy_flag = %d\n", g_cart_id, threadid, dummy_flag);
/***********************************************
* perform the inversion
***********************************************/
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] starting inversion\n");
xchange_field_5d(g_spinor_field[0]);
memset(g_spinor_field[1], 0, (VOLUME+RAND)*L5*24*sizeof(double));
ratime = CLOCK;
#ifdef MPI
if(inv_param.inv_type == QUDA_BICGSTAB_INVERTER || inv_param.inv_type == QUDA_GCR_INVERTER) {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] calling invertQuda\n");
invertQuda(g_spinor_field[1], g_spinor_field[0], &inv_param);
} else if(inv_param.inv_type == QUDA_CG_INVERTER) {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] calling testCG\n");
testCG(g_spinor_field[1], g_spinor_field[0], &inv_param);
} else {
if(g_cart_id==0) fprintf(stderr, "# [invert_dw_quda] unrecognized inverter\n");
}
#else
invertQuda(g_spinor_field[1], g_spinor_field[0], &inv_param);
#endif
retime = CLOCK;
if(g_cart_id==0) {
fprintf(stdout, "# [invert_dw_quda] QUDA time: %e seconds\n", inv_param.secs);
fprintf(stdout, "# [invert_dw_quda] QUDA Gflops: %e\n", inv_param.gflops/inv_param.secs);
fprintf(stdout, "# [invert_dw_quda] wall time: %e seconds\n", retime-ratime);
fprintf(stdout, "# [invert_dw_quda] Device memory used:\n\tSpinor: %f GiB\n\tGauge: %f GiB\n",
inv_param.spinorGiB, gauge_param.gaugeGiB);
}
} // of if threadid
// wait till all threads are here
#pragma omp barrier
if(inv_param.mass_normalization == QUDA_KAPPA_NORMALIZATION) {
_2_kappa = 2. * g_kappa5d;
for(ix=threadid; ix<VOLUME*L5;ix+=g_num_threads) {
_fv_ti_eq_re(g_spinor_field[1]+_GSI(ix), _2_kappa );
}
}
#pragma omp barrier
// reorder, multiply with g2
for(is=0;is<L5;is++) {
for(ix=threadid; ix<VOLUME; ix+=g_num_threads) {
iy = lexic2eot_5d(is, ix);
iix = is*VOLUME + ix;
_fv_eq_fv(g_spinor_field[0]+_GSI(iix), g_spinor_field[1]+_GSI(iy));
}}
#pragma omp barrier
if(rotate_gamma_basis) {
for(ix=threadid; ix<VOLUME*L5; ix+=g_num_threads) {
_fv_eq_gamma_ti_fv(g_spinor_field[1]+_GSI(ix), 2, g_spinor_field[0]+_GSI(ix));
}
} else {
for(ix=threadid; ix<VOLUME*L5;ix+=g_num_threads) {
_fv_eq_fv(g_spinor_field[1]+_GSI(ix), g_spinor_field[0]+_GSI(ix));
}
}
if(g_cart_id==0 && threadid==g_num_threads-1) fprintf(stdout, "# [invert_dw_quda] inversion done in %e seconds\n", retime-ratime);
#pragma omp single
{
#ifdef MPI
xchange_field_5d(g_spinor_field[1]);
#endif
/***********************************************
* check residuum
***********************************************/
if(check_residuum && dummy_flag>0) {
// apply the Wilson Dirac operator in the gamma-basis defined in cvc_linalg,
// which uses the tmLQCD conventions (same as in contractions)
// without explicit boundary conditions
#ifdef MPI
xchange_field_5d(g_spinor_field[2]);
xchange_field_5d(g_spinor_field[1]);
#endif
memset(g_spinor_field[0], 0, 24*(VOLUME+RAND)*L5*sizeof(double));
//sprintf(filename, "%s.inverted.ascii.%.2d", source_filename, g_cart_id);
//ofs = fopen(filename, "w");
//printf_spinor_field_5d(g_spinor_field[1], ofs);
//fclose(ofs);
Q_DW_Wilson_phi(g_spinor_field[0], g_spinor_field[1]);
for(ix=0;ix<VOLUME*L5;ix++) {
_fv_mi_eq_fv(g_spinor_field[0]+_GSI(ix), g_spinor_field[2]+_GSI(ix));
}
spinor_scalar_product_re(&norm2, g_spinor_field[2], g_spinor_field[2], VOLUME*L5);
spinor_scalar_product_re(&norm, g_spinor_field[0], g_spinor_field[0], VOLUME*L5);
if(g_cart_id==0) fprintf(stdout, "\n# [invert_dw_quda] absolut residuum squared: %e; relative residuum %e\n", norm, sqrt(norm/norm2) );
}
if(dummy_flag>0) {
/***********************************************
* create 4-dim. propagator
***********************************************/
if(convert_sign == 0) {
spinor_5d_to_4d(g_spinor_field[1], g_spinor_field[1]);
} else if(convert_sign == -1 || convert_sign == +1) {
spinor_5d_to_4d_sign(g_spinor_field[1], g_spinor_field[1], convert_sign);
}
/***********************************************
* write the solution
***********************************************/
sprintf(filename, "%s.inverted", source_filename_write);
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] writing propagator to file %s\n", filename);
check_error(write_propagator(g_spinor_field[1], filename, 0, g_propagator_precision), "write_propagator", NULL, 22);
//sprintf(filename, "prop.ascii.4d.%.2d.%.2d.%.2d", isc, g_nproc, g_cart_id);
//ofs = fopen(filename, "w");
//printf_spinor_field(g_spinor_field[1], ofs);
//fclose(ofs);
}
if(check_residuum) memcpy(g_spinor_field[2], smearing_spinor_field[0], 24*VOLUME*L5*sizeof(double));
} // of omp single
} // of omp parallel region
if(dummy_flag > 0) strcpy(source_filename_write, source_filename);
dummy_flag++;
} // of loop on momenta
} // of isc
#if 0
// last inversion
{
memcpy(g_spinor_field[0], smearing_spinor_field[1], 24*VOLUME*L5*sizeof(double));
if(g_cart_id==0) fprintf(stdout, "# [] proc%.2d starting last inversion\n", g_cart_id);
/***********************************************
* perform the inversion
***********************************************/
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] starting inversion\n");
xchange_field_5d(g_spinor_field[0]);
memset(g_spinor_field[1], 0, (VOLUME+RAND)*L5*24*sizeof(double));
ratime = CLOCK;
#ifdef MPI
if(inv_param.inv_type == QUDA_BICGSTAB_INVERTER || inv_param.inv_type == QUDA_GCR_INVERTER) {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] calling invertQuda\n");
invertQuda(g_spinor_field[1], g_spinor_field[0], &inv_param);
} else if(inv_param.inv_type == QUDA_CG_INVERTER) {
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] calling testCG\n");
testCG(g_spinor_field[1], g_spinor_field[0], &inv_param);
} else {
if(g_cart_id==0) fprintf(stderr, "# [invert_dw_quda] unrecognized inverter\n");
}
#else
invertQuda(g_spinor_field[1], g_spinor_field[0], &inv_param);
#endif
retime = CLOCK;
if(g_cart_id==0) {
fprintf(stdout, "# [invert_dw_quda] QUDA time: %e seconds\n", inv_param.secs);
fprintf(stdout, "# [invert_dw_quda] QUDA Gflops: %e\n", inv_param.gflops/inv_param.secs);
fprintf(stdout, "# [invert_dw_quda] wall time: %e seconds\n", retime-ratime);
fprintf(stdout, "# [invert_dw_quda] Device memory used:\n\tSpinor: %f GiB\n\tGauge: %f GiB\n",
inv_param.spinorGiB, gauge_param.gaugeGiB);
}
omp_set_num_threads(g_num_threads);
#pragma omp parallel private(threadid,_2_kappa,is,ix,iy,iix) shared(VOLUME,L5,g_kappa,g_spinor_field,g_num_threads)
{
threadid = omp_get_thread_num();
if(inv_param.mass_normalization == QUDA_KAPPA_NORMALIZATION) {
_2_kappa = 2. * g_kappa5d;
for(ix=threadid; ix<VOLUME*L5;ix+=g_num_threads) {
_fv_ti_eq_re(g_spinor_field[1]+_GSI(ix), _2_kappa );
}
}
#pragma omp barrier
// reorder, multiply with g2
for(is=0;is<L5;is++) {
for(ix=threadid; ix<VOLUME; ix+=g_num_threads) {
iy = lexic2eot_5d(is, ix);
iix = is*VOLUME + ix;
_fv_eq_fv(g_spinor_field[0]+_GSI(iix), g_spinor_field[1]+_GSI(iy));
}}
#pragma omp barrier
if(rotate_gamma_basis) {
for(ix=threadid; ix<VOLUME*L5; ix+=g_num_threads) {
_fv_eq_gamma_ti_fv(g_spinor_field[1]+_GSI(ix), 2, g_spinor_field[0]+_GSI(ix));
}
} else {
for(ix=threadid; ix<VOLUME*L5;ix+=g_num_threads) {
_fv_eq_fv(g_spinor_field[1]+_GSI(ix), g_spinor_field[0]+_GSI(ix));
}
}
} // end of parallel region
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] inversion done in %e seconds\n", retime-ratime);
#ifdef MPI
xchange_field_5d(g_spinor_field[1]);
#endif
/***********************************************
* check residuum
***********************************************/
if(check_residuum && dummy_flag>0) {
// apply the Wilson Dirac operator in the gamma-basis defined in cvc_linalg,
// which uses the tmLQCD conventions (same as in contractions)
// without explicit boundary conditions
#ifdef MPI
xchange_field_5d(g_spinor_field[2]);
#endif
memset(g_spinor_field[0], 0, 24*(VOLUME+RAND)*L5*sizeof(double));
//sprintf(filename, "%s.inverted.ascii.%.2d", source_filename, g_cart_id);
//ofs = fopen(filename, "w");
//printf_spinor_field_5d(g_spinor_field[1], ofs);
//fclose(ofs);
Q_DW_Wilson_phi(g_spinor_field[0], g_spinor_field[1]);
for(ix=0;ix<VOLUME*L5;ix++) {
_fv_mi_eq_fv(g_spinor_field[0]+_GSI(ix), g_spinor_field[2]+_GSI(ix));
}
spinor_scalar_product_re(&norm, g_spinor_field[0], g_spinor_field[0], VOLUME*L5);
spinor_scalar_product_re(&norm2, g_spinor_field[2], g_spinor_field[2], VOLUME*L5);
if(g_cart_id==0) fprintf(stdout, "\n# [invert_dw_quda] absolut residuum squared: %e; relative residuum %e\n", norm, sqrt(norm/norm2) );
}
/***********************************************
* create 4-dim. propagator
***********************************************/
if(convert_sign == 0) {
spinor_5d_to_4d(g_spinor_field[1], g_spinor_field[1]);
} else if(convert_sign == -1 || convert_sign == +1) {
spinor_5d_to_4d_sign(g_spinor_field[1], g_spinor_field[1], convert_sign);
}
/***********************************************
* write the solution
***********************************************/
sprintf(filename, "%s.inverted", source_filename_write);
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] writing propagator to file %s\n", filename);
check_error(write_propagator(g_spinor_field[1], filename, 0, g_propagator_precision), "write_propagator", NULL, 22);
//sprintf(filename, "prop.ascii.4d.%.2d.%.2d.%.2d", isc, g_nproc, g_cart_id);
//ofs = fopen(filename, "w");
//printf_spinor_field(g_spinor_field[1], ofs);
//fclose(ofs);
} // of last inversion
#endif // of if 0
/***********************************************
* free the allocated memory, finalize
***********************************************/
#ifdef HAVE_QUDA
// finalize the QUDA library
if(g_cart_id==0) fprintf(stdout, "# [invert_dw_quda] finalizing quda\n");
#ifdef MPI
freeGaugeQuda();
#endif
endQuda();
#endif
if(g_gauge_field != NULL) free(g_gauge_field);
if(gauge_field_smeared != NULL) free(gauge_field_smeared);
if(no_fields>0) {
if(g_spinor_field!=NULL) {
for(i=0; i<no_fields; i++) if(g_spinor_field[i]!=NULL) free(g_spinor_field[i]);
free(g_spinor_field);
}
}
free_geometry();
if(g_source_momentum_set && full_orbit) {
finalize_q_orbits(&qlatt_id, &qlatt_count, &qlatt_list, &qlatt_rep);
if(qlatt_map != NULL) {
free(qlatt_map[0]);
free(qlatt_map);
}
}
if(source_momentum != NULL) free(source_momentum);
if(lck != NULL) free(lck);
#ifdef MPI
#ifdef HAVE_QUDA
endCommsQuda();
#else
MPI_Finalize();
#endif
#endif
if(g_cart_id==0) {
g_the_time = time(NULL);
fprintf(stdout, "\n# [invert_dw_quda] %s# [invert_dw_quda] end of run\n", ctime(&g_the_time));
fprintf(stderr, "\n# [invert_dw_quda] %s# [invert_dw_quda] end of run\n", ctime(&g_the_time));
}
return(0);
}
int
q_alignOfTypeEncode(const char* type) {
type = q_skipVarNameEncode(type);
switch (*type) {
case Q_C_ID:
return __alignof(id);
case Q_C_CLASS:
return __alignof(Class);
case Q_C_SEL:
return __alignof(SEL);
case Q_C_CHR:
return __alignof(char);
case Q_C_UCHR:
return __alignof(unsigned char);
case Q_C_SHT:
return __alignof(short);
case Q_C_USHT:
return __alignof(unsigned short);
case Q_C_INT:
return __alignof(int);
case Q_C_UINT:
return __alignof(unsigned int);
case Q_C_LNG:
return __alignof(long);
case Q_C_ULNG:
return __alignof(unsigned long);
case Q_C_FLT:
return __alignof(float);
case Q_C_DBL:
return __alignof(double);
case Q_C_VOID:
return 0;
case Q_C_PTR:
case Q_C_CHARPTR:
return __alignof(char*);
case Q_C_ARY_B:
while (isdigit(*++type)); // empty loop
return q_alignOfTypeEncode(type);
case Q_C_STRUCT_B: {
struct _StructLayout layout;
unsigned int align;
q_layoutStructBeginEncode(type, &layout);
while (q_layoutStructNextEncode(&layout)); // empty loop
q_layoutStructEndEncode(&layout, nil, &align);
return align;
}
case Q_C_UNION_B: {
int maxaling = 0;
while (*type != Q_C_UNION_E && *type++ != '='); // empty loop
while (*type != Q_C_UNION_E) {
type = q_skipVarNameEncode(type);
maxaling = _MAX(maxaling, q_alignOfTypeEncode(type));
type = q_skipTypeSpecEncode(type);
}
return maxaling;
}
default:
q_throwError(er1, type);
}
return 0;
}