{
return NULL;
}
static void
handler (int sig)
{
}
static void __attribute__ ((noinline))
clobber_lots_of_regs (void)
{
#define X1(n) long r##n = 10##n; __asm __volatile ("" : "+r" (r##n));
#define X2(n) X1(n##0) X1(n##1) X1(n##2) X1(n##3) X1(n##4)
#define X3(n) X2(n##0) X2(n##1) X2(n##2) X2(n##3) X2(n##4)
X3(0) X3(1) X3(2) X3(3) X3(4)
#undef X1
#define X1(n) __asm __volatile ("" : : "r" (r##n));
X3(0) X3(1) X3(2) X3(3) X3(4)
#undef X1
#undef X2
#undef X3
}
static int
do_test (void)
{
pthread_t th;
int old, rc;
int ret = 0;
int fd[2];
void dooptab()
{ int i;
FILE *f;
/* Load optab[] */
#define X1(arr,mask) for(i=0;i<sizeof(arr)/sizeof(int);i++)xptab1[arr[i]]|=mask;
#define X2(arr,mask) for(i=0;i<sizeof(arr)/sizeof(int);i++)xptab2[arr[i]]|=mask;
#define X3(arr,mask) for(i=0;i<sizeof(arr)/sizeof(int);i++)xptab3[arr[i]]|=mask;
X1(_binary,_OTbinary);
X1(_unary,_OTunary);
X1(_commut,_OTcommut);
X1(_assoc,_OTassoc);
X1(_sideff,_OTsideff);
X1(_eop0e,_OTeop0e);
X1(_eop00,_OTeop00);
X1(_eop1e,_OTeop1e);
X2(_logical,_OTlogical);
X2(_wid,_OTwid);
X2(_call,_OTcall);
X2(_rtol,_OTrtol);
X2(_assign,_OTassign);
X2(_def,_OTdef);
X2(_ae,_OTae);
X2(_exp,_OTexp);
X3(_boolnop,_OTboolnop);
#undef X3
#undef X2
#undef X1
f = fopen("optab.c","w");
fprintf(f,"const unsigned char optab1[OPMAX] =\n\t{");
for (i = 0; i < OPMAX; i++)
{ if ((i & 7) == 0)
fprintf(f,"\n\t");
fprintf(f,"0x%x",xptab1[i]);
if (i != OPMAX - 1)
fprintf(f,",");
}
fprintf(f,"\t};\n");
fprintf(f,"const unsigned char optab2[OPMAX] =\n\t{");
for (i = 0; i < OPMAX; i++)
{ if ((i & 7) == 0)
fprintf(f,"\n\t");
fprintf(f,"0x%x",xptab2[i]);
if (i != OPMAX - 1)
fprintf(f,",");
}
fprintf(f,"\t};\n");
fprintf(f,"const unsigned char optab3[OPMAX] =\n\t{");
for (i = 0; i < OPMAX; i++)
{ if ((i & 7) == 0)
fprintf(f,"\n\t");
fprintf(f,"0x%x",xptab3[i]);
if (i != OPMAX - 1)
fprintf(f,",");
}
fprintf(f,"\t};\n");
fprintf(f,"const unsigned char opcost[OPMAX] =\n\t{");
for (i = 0; i < OPMAX; i++)
{ if ((i & 7) == 0)
fprintf(f,"\n\t");
fprintf(f,"0x%x",cost(i));
if (i != OPMAX - 1)
fprintf(f,",");
}
fprintf(f,"\t};\n");
doreltables(f);
fclose(f);
}
static int sort(struct env *env, int M)
{
int sp=0;
struct tree *left=NULL, *right=NULL;
int nBucket = 0;
struct particle **ps = env->p;
struct tree *node = env->tree;
/*========================================================================
* Allocate stack
*======================================================================*/
int ns = (int)MMAX(1, floor(log(((double)(env->N+1))/(M+1))/log(2.0)));
sp = 0;
sortStack = (struct tree **)realloc(sortStack, ns * sizeof(struct tree *));
assert(sortStack != NULL);
if (node->iUpper - node->iLower + 1 <= M)
return 0;
while (1)
{
int i, p, k;
struct particle *t;
int nl, nr;
int d;
real fSplit;
assert(node != NULL);
split(node, &d, &fSplit);
#if 0
i = node->iLower;
int j = node->iUpper;
//partitionCount++;
//PARTITION(ps, t, ->r[d], i,j, < fSplit,> fSplit);
while (i <= j && ((ps[i] ->r[d]) < fSplit)) { ++i; }
#if 0
while (i <= j)
{
switch ((i-j) & 0x7)
case 7: if
}
#endif
while (i <= j && ((ps[j] ->r[d]) > fSplit)) { --j; }
while (i < j) {
SWAP(ps[i], ps[j], t);
while ((ps[++i] ->r[d]) < fSplit) { }
while ((ps[--j] ->r[d]) > fSplit) { }
}
#else
# define PART(A, i, j, t, d, CMPL) \
do { \
int k = i--; \
do { \
if (A[k]->r[d] CMPL) { i++; SWAP(A[i], A[k], t); } \
} while (++k <= j); \
i++; \
} while(0)
i = node->iLower;
const int j = node->iUpper;
switch (d)
{
case 0: PART(ps, i, j, t, 0, < fSplit); break;
case 1: PART(ps, i, j, t, 1, < fSplit); break;
case 2: PART(ps, i, j, t, 2, < fSplit); break;
}
#endif
//fprintf(err, "(%i %i)\n", i, node->iUpper);
nl = i - node->iLower;
nr = node->iUpper - i + 1;
//fprintf(err, "nl=%i nr=%i\n", nl, nr);
/*========================================================================
* If both sides of the partition are not empty then create two new nodes
* and crop the bounding boxes. Otherwise, undo the split operation.
*======================================================================*/
if (nl > 0 || nr > 0) { // jpc changed this from && to || so that a split is always created.
NEW_NODE(node, left, node->iLower, i - 1);
NEW_NODE(node, right, i, node->iUpper);
nodeCount += 2;
X3( node->left->bnd.r [_] = node->bnd.r [_] );
X3( node->left->bnd.max [_] = node->bnd.max [_] );
node->left->bnd.max[d] *= 0.5;
node->left->bnd.r[d] -= node->left->bnd.max[d];
left = node->left;
X3( node->right->bnd.r [_] = node->bnd.r [_] );
X3( node->right->bnd.max [_] = node->bnd.max [_] );
node->right->bnd.max[d] *= 0.5;
node->right->bnd.r[d] += node->right->bnd.max[d];
right = node->right;
}
else
{
node->bnd.max[d] *= 0.5;
if (nl > 0) {
node->bnd.r[d] -= node->bnd.max[d];
left = node;
}
else {
node->bnd.r[d] += node->bnd.max[d];
right = node;
}
}
/*========================================================================
* Now figure out which subfile to process next
*======================================================================*/
if (nl > M && nr > M)
{
if (nr > nl) sortStack[sp++] = right, node = left;
else sortStack[sp++] = left, node = right;
}
else
{
if (nl > M) node = left;
//else if (nl > 0) left->iLower = 0;
if (nr > M) node = right;
//else if (nr > 0) right->iLower = 0;
}
if (nl <= M && nr <= M) {
if (sp) node = sortStack[--sp]; /* pop tn */
else break;
}
}
return 0;
}
void test() {
g1(X1());
g2(X2());
g3(X3());
g4(X4<int>());
}
struct X4
{
constexpr operator double() const { return 1.0; }
};
struct X5
{
constexpr operator int() const { return __INT_MAX__; }
constexpr operator unsigned() const { return __INT_MAX__ * 2U + 1; }
};
enum E0 { e0 = X0() };
enum E1 { e1 = X1() };
enum E2 { e2 = X2() };
enum E3 { e3 = X3() };
enum E4 { e4 = X4() }; // { dg-error "integer constant" }
enum E5 { e5 = X5() }; // { dg-error "ambiguous" }
enum F0 : int { f0 = X0() };
enum F1 : int { f1 = X1() };
enum F2 : int { f2 = X2() };
enum F3 : int { f3 = X3() };
enum F4 : int { f4 = X4() }; // { dg-error "narrowing" }
enum F5 : int { f5 = X5() };
enum G0 : signed char { g0 = X0() };
enum G1 : signed char { g1 = X1() };
enum G2 : signed char { g2 = X2() }; // { dg-error "narrowing" }
enum G3 : signed char { g3 = X3() }; // { dg-error "narrowing" }
enum G4 : signed char { g4 = X4() }; // { dg-error "narrowing" }
bool
test_fundamentals(Index const dimension)
{
bool
passed = true;
Index const
number_components = integer_power(dimension, Tensor::ORDER);
std::vector<Scalar> const
X = generate_sequence<Scalar>(number_components, 1.0, 1.0);
// Test constructor with pointer
Tensor const
A(dimension, &X[0]);
// Test copy constructor
Tensor
B = A;
Tensor
C;
// Test copy assignment
C = B - A;
Scalar
error = norm_f(C);
bool const
copy_assigned = error <= machine_epsilon<Scalar>();
passed = passed && copy_assigned;
// Test fill with pointer
B.fill(&X[0]);
C = B - A;
error = norm_f(C);
bool const
filled_pointer = error <= machine_epsilon<Scalar>();
passed = passed && filled_pointer;
std::vector<Scalar> const
Y = generate_sequence<Scalar>(number_components, -1.0, -1.0);
C.fill(&Y[0]);
// Test increment
C += A;
error = norm_f(C);
bool const
incremented = error <= machine_epsilon<Scalar>();
passed = passed && incremented;
C.fill(&X[0]);
// Test decrement
C -= A;
error = norm_f(C);
bool const
decremented = error <= machine_epsilon<Scalar>();
passed = passed && decremented;
#ifdef HAVE_INTREPID_KOKKOSCORE
//test Tensor fill and create for Kokkos data types
Kokkos::View<Scalar *, Kokkos::DefaultExecutionSpace>
X1("X1_kokkos", dimension);
Kokkos::View<Scalar **, Kokkos::DefaultExecutionSpace>
X2("X2_kokkos", dimension, dimension);
Kokkos::View<Scalar ***, Kokkos::DefaultExecutionSpace>
X3("X3_kokkos", dimension, dimension, dimension);
Kokkos::View<Scalar ****, Kokkos::DefaultExecutionSpace>
X4("X4_kokkos", dimension, dimension, dimension, dimension);
Kokkos::deep_copy(X1, 3.1);
Kokkos::deep_copy(X2, 3.2);
Kokkos::deep_copy(X3, 3.3);
Kokkos::deep_copy(X4, 3.4);
Tensor
Z(dimension); //(X1_k,0);
Index
rank = 0;
Index
temp = number_components;
while (temp != 1) {
temp = temp / dimension;
rank = rank + 1;
assert(temp > 0);
}
switch (rank) {
default:
assert(false);
break;
case 1:
Z.fill(X1, 0);
break;
case 2:
Z.fill(X2, 0, 0);
break;
case 3:
Z.fill(X3, 0, 0, 0);
break;
case 4:
Z.fill(X4, 0, 0, 0, 0);
break;
}
// Test copy constructor.
Tensor const
U = Z;
// Test copy assignment.
Tensor
V;
V = U - Z;
error = norm_f(V);
bool const
tensor_create_from_1d_kokkos = error <= machine_epsilon<Scalar>();
passed = passed && tensor_create_from_1d_kokkos;
#endif
return passed;
}
