void
bounds_reduced_extents (array_t *a, array_t *b, int which, const char *a_name,
const char *intrinsic)
{
index_type i, n, a_size, b_size;
assert (GFC_DESCRIPTOR_RANK(a) == GFC_DESCRIPTOR_RANK(b) - 1);
a_size = size0 (a);
b_size = size0 (b);
if (b_size == 0)
{
if (a_size != 0)
runtime_error ("Incorrect size in %s of %s"
" intrinsic: should not be zero-sized",
a_name, intrinsic);
}
else
{
if (a_size == 0)
runtime_error ("Incorrect size of %s of %s"
" intrinsic: should be zero-sized",
a_name, intrinsic);
i = 0;
for (n = 0; n < GFC_DESCRIPTOR_RANK (b); n++)
{
index_type a_extent, b_extent;
if (n != which)
{
a_extent = GFC_DESCRIPTOR_EXTENT(a, i);
b_extent = GFC_DESCRIPTOR_EXTENT(b, n);
if (a_extent != b_extent)
runtime_error("Incorrect extent in %s of %s"
" intrinsic in dimension %ld: is %ld,"
" should be %ld", a_name, intrinsic, (long int) i + 1,
(long int) a_extent, (long int) b_extent);
i++;
}
}
}
}
void
bounds_equal_extents (array_t *a, array_t *b, const char *a_name,
const char *intrinsic)
{
index_type a_size, b_size, n;
assert (GFC_DESCRIPTOR_RANK(a) == GFC_DESCRIPTOR_RANK(b));
a_size = size0 (a);
b_size = size0 (b);
if (b_size == 0)
{
if (a_size != 0)
runtime_error ("Incorrect size of %s in %s"
" intrinsic: should be zero-sized",
a_name, intrinsic);
}
else
{
if (a_size == 0)
runtime_error ("Incorrect size of %s of %s"
" intrinsic: Should not be zero-sized",
a_name, intrinsic);
for (n = 0; n < GFC_DESCRIPTOR_RANK (b); n++)
{
index_type a_extent, b_extent;
a_extent = GFC_DESCRIPTOR_EXTENT(a, n);
b_extent = GFC_DESCRIPTOR_EXTENT(b, n);
if (a_extent != b_extent)
runtime_error("Incorrect extent in %s of %s"
" intrinsic in dimension %ld: is %ld,"
" should be %ld", a_name, intrinsic, (long int) n + 1,
(long int) a_extent, (long int) b_extent);
}
}
}
dgVector dgCollisionBox::SupportVertexSpecial(const dgVector& dir, dgInt32* const vertexIndex) const
{
dgAssert(dgAbsf(dir.DotProduct3(dir) - dgFloat32(1.0f)) < dgFloat32(1.0e-3f));
dgAssert(dir.m_w == dgFloat32(0.0f));
dgVector mask(dir < dgVector(dgFloat32(0.0f)));
if (vertexIndex) {
dgVector index(m_indexMark.CompProduct4(mask & dgVector::m_one));
index = (index.AddHorizontal()).GetInt();
*vertexIndex = dgInt32 (index.m_ix);
}
dgVector padd (D_BOX_SKIN_THINCKNESS);
padd = padd & dgVector::m_triplexMask;
dgVector size0 (m_size[0] - padd);
dgVector size1 (m_size[1] + padd);
return (size1 & mask) + size0.AndNot(mask);
}
void
bounds_ifunction_return (array_t * a, const index_type * extent,
const char * a_name, const char * intrinsic)
{
int empty;
int n;
int rank;
index_type a_size;
rank = GFC_DESCRIPTOR_RANK (a);
a_size = size0 (a);
empty = 0;
for (n = 0; n < rank; n++)
{
if (extent[n] == 0)
empty = 1;
}
if (empty)
{
if (a_size != 0)
runtime_error ("Incorrect size in %s of %s"
" intrinsic: should be zero-sized",
a_name, intrinsic);
}
else
{
if (a_size == 0)
runtime_error ("Incorrect size of %s in %s"
" intrinsic: should not be zero-sized",
a_name, intrinsic);
for (n = 0; n < rank; n++)
{
index_type a_extent;
a_extent = GFC_DESCRIPTOR_EXTENT(a, n);
if (a_extent != extent[n])
runtime_error("Incorrect extent in %s of %s"
" intrinsic in dimension %ld: is %ld,"
" should be %ld", a_name, intrinsic, (long int) n + 1,
(long int) a_extent, (long int) extent[n]);
}
}
}
static void
unpack_bounds (gfc_array_char *ret, const gfc_array_char *vector,
const gfc_array_l1 *mask, const gfc_array_char *field)
{
index_type vec_size, mask_count;
vec_size = size0 ((array_t *) vector);
mask_count = count_0 (mask);
if (vec_size < mask_count)
runtime_error ("Incorrect size of return value in UNPACK"
" intrinsic: should be at least %ld, is"
" %ld", (long int) mask_count,
(long int) vec_size);
if (field != NULL)
bounds_equal_extents ((array_t *) field, (array_t *) mask,
"FIELD", "UNPACK");
if (ret->base_addr != NULL)
bounds_equal_extents ((array_t *) ret, (array_t *) mask,
"return value", "UNPACK");
}
void NonUniformScaledCollision(DemoEntityManager* const scene)
{
// load the skybox
scene->CreateSkyBox();
// load the scene from a ngd file format
CreateLevelMesh(scene, "flatPlane.ngd", 1);
// CreateLevelMesh (scene, "sponza.ngd", 1);
// CreateLevelMesh (scene, "cattle.ngd", fileName);
// CreateLevelMesh (scene, "playground.ngd", 1);
dMatrix camMatrix(dRollMatrix(-20.0f * 3.1416f / 180.0f) * dYawMatrix(-45.0f * 3.1416f / 180.0f));
dQuaternion rot(camMatrix);
// dVector origin (-30.0f, 40.0f, -15.0f, 0.0f);
dVector origin(-10.0f, 5.0f, -15.0f, 0.0f);
scene->SetCameraMatrix(rot, origin);
NewtonWorld* const world = scene->GetNewton();
int defaultMaterialID = NewtonMaterialGetDefaultGroupID(world);
dVector location(0.0f, 0.0f, 0.0f, 0.0f);
dVector size0(0.5f, 0.5f, 1.0f, 0.0f);
dVector size1(0.5f, 0.5f, 0.5f, 0.0f);
int count = 5;
dMatrix shapeOffsetMatrix(dRollMatrix(3.141592f / 2.0f));
shapeOffsetMatrix.m_posit.m_y = 0.0f;
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _SPHERE_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _BOX_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _CAPSULE_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size1, count, count, 4.0f, _CAPSULE_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size1, count, count, 4.0f, _CYLINDER_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _CYLINDER_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _CHAMFER_CYLINDER_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _CONE_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _REGULAR_CONVEX_HULL_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _RANDOM_CONVEX_HULL_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 4.0f, _COMPOUND_CONVEX_CRUZ_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
}
static void
eoshift0 (gfc_array_char * ret, const gfc_array_char * array,
int shift, const char * pbound, int which, index_type size,
char filler)
{
/* r.* indicates the return array. */
index_type rstride[GFC_MAX_DIMENSIONS];
index_type rstride0;
index_type roffset;
char *rptr;
char *dest;
/* s.* indicates the source array. */
index_type sstride[GFC_MAX_DIMENSIONS];
index_type sstride0;
index_type soffset;
const char *sptr;
const char *src;
index_type count[GFC_MAX_DIMENSIONS];
index_type extent[GFC_MAX_DIMENSIONS];
index_type dim;
index_type len;
index_type n;
/* The compiler cannot figure out that these are set, initialize
them to avoid warnings. */
len = 0;
soffset = 0;
roffset = 0;
if (ret->data == NULL)
{
int i;
ret->data = internal_malloc_size (size * size0 ((array_t *)array));
ret->offset = 0;
ret->dtype = array->dtype;
for (i = 0; i < GFC_DESCRIPTOR_RANK (array); i++)
{
ret->dim[i].lbound = 0;
ret->dim[i].ubound = array->dim[i].ubound - array->dim[i].lbound;
if (i == 0)
ret->dim[i].stride = 1;
else
ret->dim[i].stride = (ret->dim[i-1].ubound + 1) * ret->dim[i-1].stride;
}
}
which = which - 1;
extent[0] = 1;
count[0] = 0;
sstride[0] = -1;
rstride[0] = -1;
n = 0;
for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++)
{
if (dim == which)
{
roffset = ret->dim[dim].stride * size;
if (roffset == 0)
roffset = size;
soffset = array->dim[dim].stride * size;
if (soffset == 0)
soffset = size;
len = array->dim[dim].ubound + 1 - array->dim[dim].lbound;
}
else
{
count[n] = 0;
extent[n] = array->dim[dim].ubound + 1 - array->dim[dim].lbound;
rstride[n] = ret->dim[dim].stride * size;
sstride[n] = array->dim[dim].stride * size;
n++;
}
}
if (sstride[0] == 0)
sstride[0] = size;
if (rstride[0] == 0)
rstride[0] = size;
dim = GFC_DESCRIPTOR_RANK (array);
rstride0 = rstride[0];
sstride0 = sstride[0];
rptr = ret->data;
sptr = array->data;
if ((shift >= 0 ? shift : -shift) > len)
{
shift = len;
len = 0;
}
else
{
if (shift > 0)
len = len - shift;
else
len = len + shift;
}
while (rptr)
{
/* Do the shift for this dimension. */
if (shift > 0)
{
src = &sptr[shift * soffset];
dest = rptr;
}
else
{
src = sptr;
dest = &rptr[-shift * roffset];
}
for (n = 0; n < len; n++)
{
memcpy (dest, src, size);
dest += roffset;
src += soffset;
}
if (shift >= 0)
{
n = shift;
}
else
{
dest = rptr;
n = -shift;
}
if (pbound)
while (n--)
{
memcpy (dest, pbound, size);
dest += roffset;
}
else
while (n--)
{
memset (dest, filler, size);
dest += roffset;
}
/* Advance to the next section. */
rptr += rstride0;
sptr += sstride0;
count[0]++;
n = 0;
while (count[n] == extent[n])
{
/* When we get to the end of a dimension, reset it and increment
the next dimension. */
count[n] = 0;
/* We could precalculate these products, but this is a less
frequently used path so probably not worth it. */
rptr -= rstride[n] * extent[n];
sptr -= sstride[n] * extent[n];
n++;
if (n >= dim - 1)
{
/* Break out of the loop. */
rptr = NULL;
break;
}
else
{
count[n]++;
rptr += rstride[n];
sptr += sstride[n];
}
}
}
}
static void
transpose_internal (gfc_array_char *ret, gfc_array_char *source,
index_type size)
{
/* r.* indicates the return array. */
index_type rxstride, rystride;
char *rptr;
/* s.* indicates the source array. */
index_type sxstride, systride;
const char *sptr;
index_type xcount, ycount;
index_type x, y;
assert (GFC_DESCRIPTOR_RANK (source) == 2
&& GFC_DESCRIPTOR_RANK (ret) == 2);
if (ret->data == NULL)
{
assert (ret->dtype == source->dtype);
ret->dim[0].lbound = 0;
ret->dim[0].ubound = source->dim[1].ubound - source->dim[1].lbound;
ret->dim[0].stride = 1;
ret->dim[1].lbound = 0;
ret->dim[1].ubound = source->dim[0].ubound - source->dim[0].lbound;
ret->dim[1].stride = ret->dim[0].ubound+1;
ret->data = internal_malloc_size (size * size0 ((array_t*)ret));
ret->offset = 0;
}
else if (unlikely (compile_options.bounds_check))
{
index_type ret_extent, src_extent;
ret_extent = ret->dim[0].ubound + 1 - ret->dim[0].lbound;
src_extent = source->dim[1].ubound + 1 - source->dim[1].lbound;
if (src_extent != ret_extent)
runtime_error ("Incorrect extent in return value of TRANSPOSE"
" intrinsic in dimension 1: is %ld,"
" should be %ld", (long int) src_extent,
(long int) ret_extent);
ret_extent = ret->dim[1].ubound + 1 - ret->dim[1].lbound;
src_extent = source->dim[0].ubound + 1 - source->dim[0].lbound;
if (src_extent != ret_extent)
runtime_error ("Incorrect extent in return value of TRANSPOSE"
" intrinsic in dimension 2: is %ld,"
" should be %ld", (long int) src_extent,
(long int) ret_extent);
}
sxstride = source->dim[0].stride * size;
systride = source->dim[1].stride * size;
xcount = source->dim[0].ubound + 1 - source->dim[0].lbound;
ycount = source->dim[1].ubound + 1 - source->dim[1].lbound;
rxstride = ret->dim[0].stride * size;
rystride = ret->dim[1].stride * size;
rptr = ret->data;
sptr = source->data;
for (y = 0; y < ycount; y++)
{
for (x = 0; x < xcount; x++)
{
memcpy (rptr, sptr, size);
sptr += sxstride;
rptr += rystride;
}
sptr += systride - (sxstride * xcount);
rptr += rxstride - (rystride * xcount);
}
}
static void
cshift0 (gfc_array_char * ret, const gfc_array_char * array,
ptrdiff_t shift, int which, index_type size)
{
/* r.* indicates the return array. */
index_type rstride[GFC_MAX_DIMENSIONS];
index_type rstride0;
index_type roffset;
char *rptr;
/* s.* indicates the source array. */
index_type sstride[GFC_MAX_DIMENSIONS];
index_type sstride0;
index_type soffset;
const char *sptr;
index_type count[GFC_MAX_DIMENSIONS];
index_type extent[GFC_MAX_DIMENSIONS];
index_type dim;
index_type len;
index_type n;
index_type arraysize;
index_type type_size;
if (which < 1 || which > GFC_DESCRIPTOR_RANK (array))
runtime_error ("Argument 'DIM' is out of range in call to 'CSHIFT'");
arraysize = size0 ((array_t *) array);
if (ret->base_addr == NULL)
{
int i;
ret->offset = 0;
GFC_DTYPE_COPY(ret,array);
for (i = 0; i < GFC_DESCRIPTOR_RANK (array); i++)
{
index_type ub, str;
ub = GFC_DESCRIPTOR_EXTENT(array,i) - 1;
if (i == 0)
str = 1;
else
str = GFC_DESCRIPTOR_EXTENT(ret,i-1) *
GFC_DESCRIPTOR_STRIDE(ret,i-1);
GFC_DIMENSION_SET(ret->dim[i], 0, ub, str);
}
/* xmallocarray allocates a single byte for zero size. */
ret->base_addr = xmallocarray (arraysize, size);
}
else if (unlikely (compile_options.bounds_check))
{
bounds_equal_extents ((array_t *) ret, (array_t *) array,
"return value", "CSHIFT");
}
if (arraysize == 0)
return;
type_size = GFC_DTYPE_TYPE_SIZE (array);
switch(type_size)
{
case GFC_DTYPE_LOGICAL_1:
case GFC_DTYPE_INTEGER_1:
cshift0_i1 ((gfc_array_i1 *)ret, (gfc_array_i1 *) array, shift, which);
return;
case GFC_DTYPE_LOGICAL_2:
case GFC_DTYPE_INTEGER_2:
cshift0_i2 ((gfc_array_i2 *)ret, (gfc_array_i2 *) array, shift, which);
return;
case GFC_DTYPE_LOGICAL_4:
case GFC_DTYPE_INTEGER_4:
cshift0_i4 ((gfc_array_i4 *)ret, (gfc_array_i4 *) array, shift, which);
return;
case GFC_DTYPE_LOGICAL_8:
case GFC_DTYPE_INTEGER_8:
cshift0_i8 ((gfc_array_i8 *)ret, (gfc_array_i8 *) array, shift, which);
return;
#ifdef HAVE_GFC_INTEGER_16
case GFC_DTYPE_LOGICAL_16:
case GFC_DTYPE_INTEGER_16:
cshift0_i16 ((gfc_array_i16 *)ret, (gfc_array_i16 *) array, shift,
which);
return;
#endif
case GFC_DTYPE_REAL_4:
cshift0_r4 ((gfc_array_r4 *)ret, (gfc_array_r4 *) array, shift, which);
return;
case GFC_DTYPE_REAL_8:
cshift0_r8 ((gfc_array_r8 *)ret, (gfc_array_r8 *) array, shift, which);
return;
/* FIXME: This here is a hack, which will have to be removed when
the array descriptor is reworked. Currently, we don't store the
kind value for the type, but only the size. Because on targets with
__float128, we have sizeof(logn double) == sizeof(__float128),
we cannot discriminate here and have to fall back to the generic
handling (which is suboptimal). */
#if !defined(GFC_REAL_16_IS_FLOAT128)
# ifdef HAVE_GFC_REAL_10
case GFC_DTYPE_REAL_10:
cshift0_r10 ((gfc_array_r10 *)ret, (gfc_array_r10 *) array, shift,
which);
return;
# endif
# ifdef HAVE_GFC_REAL_16
case GFC_DTYPE_REAL_16:
cshift0_r16 ((gfc_array_r16 *)ret, (gfc_array_r16 *) array, shift,
which);
return;
# endif
#endif
case GFC_DTYPE_COMPLEX_4:
cshift0_c4 ((gfc_array_c4 *)ret, (gfc_array_c4 *) array, shift, which);
return;
case GFC_DTYPE_COMPLEX_8:
cshift0_c8 ((gfc_array_c8 *)ret, (gfc_array_c8 *) array, shift, which);
return;
/* FIXME: This here is a hack, which will have to be removed when
the array descriptor is reworked. Currently, we don't store the
kind value for the type, but only the size. Because on targets with
__float128, we have sizeof(logn double) == sizeof(__float128),
we cannot discriminate here and have to fall back to the generic
handling (which is suboptimal). */
#if !defined(GFC_REAL_16_IS_FLOAT128)
# ifdef HAVE_GFC_COMPLEX_10
case GFC_DTYPE_COMPLEX_10:
cshift0_c10 ((gfc_array_c10 *)ret, (gfc_array_c10 *) array, shift,
which);
return;
# endif
# ifdef HAVE_GFC_COMPLEX_16
case GFC_DTYPE_COMPLEX_16:
cshift0_c16 ((gfc_array_c16 *)ret, (gfc_array_c16 *) array, shift,
which);
return;
# endif
#endif
default:
break;
}
switch (size)
{
/* Let's check the actual alignment of the data pointers. If they
are suitably aligned, we can safely call the unpack functions. */
case sizeof (GFC_INTEGER_1):
cshift0_i1 ((gfc_array_i1 *) ret, (gfc_array_i1 *) array, shift,
which);
break;
case sizeof (GFC_INTEGER_2):
if (GFC_UNALIGNED_2(ret->base_addr) || GFC_UNALIGNED_2(array->base_addr))
break;
else
{
cshift0_i2 ((gfc_array_i2 *) ret, (gfc_array_i2 *) array, shift,
which);
return;
}
case sizeof (GFC_INTEGER_4):
if (GFC_UNALIGNED_4(ret->base_addr) || GFC_UNALIGNED_4(array->base_addr))
break;
else
{
cshift0_i4 ((gfc_array_i4 *)ret, (gfc_array_i4 *) array, shift,
which);
return;
}
case sizeof (GFC_INTEGER_8):
if (GFC_UNALIGNED_8(ret->base_addr) || GFC_UNALIGNED_8(array->base_addr))
{
/* Let's try to use the complex routines. First, a sanity
check that the sizes match; this should be optimized to
a no-op. */
if (sizeof(GFC_INTEGER_8) != sizeof(GFC_COMPLEX_4))
break;
if (GFC_UNALIGNED_C4(ret->base_addr)
|| GFC_UNALIGNED_C4(array->base_addr))
break;
cshift0_c4 ((gfc_array_c4 *) ret, (gfc_array_c4 *) array, shift,
which);
return;
}
else
{
cshift0_i8 ((gfc_array_i8 *)ret, (gfc_array_i8 *) array, shift,
which);
return;
}
#ifdef HAVE_GFC_INTEGER_16
case sizeof (GFC_INTEGER_16):
if (GFC_UNALIGNED_16(ret->base_addr)
|| GFC_UNALIGNED_16(array->base_addr))
{
/* Let's try to use the complex routines. First, a sanity
check that the sizes match; this should be optimized to
a no-op. */
if (sizeof(GFC_INTEGER_16) != sizeof(GFC_COMPLEX_8))
break;
if (GFC_UNALIGNED_C8(ret->base_addr)
|| GFC_UNALIGNED_C8(array->base_addr))
break;
cshift0_c8 ((gfc_array_c8 *) ret, (gfc_array_c8 *) array, shift,
which);
return;
}
else
{
cshift0_i16 ((gfc_array_i16 *) ret, (gfc_array_i16 *) array,
shift, which);
return;
}
#else
case sizeof (GFC_COMPLEX_8):
if (GFC_UNALIGNED_C8(ret->base_addr)
|| GFC_UNALIGNED_C8(array->base_addr))
break;
else
{
cshift0_c8 ((gfc_array_c8 *) ret, (gfc_array_c8 *) array, shift,
which);
return;
}
#endif
default:
break;
}
which = which - 1;
sstride[0] = 0;
rstride[0] = 0;
extent[0] = 1;
count[0] = 0;
n = 0;
/* Initialized for avoiding compiler warnings. */
roffset = size;
soffset = size;
len = 0;
for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++)
{
if (dim == which)
{
roffset = GFC_DESCRIPTOR_STRIDE_BYTES(ret,dim);
if (roffset == 0)
roffset = size;
soffset = GFC_DESCRIPTOR_STRIDE_BYTES(array,dim);
if (soffset == 0)
soffset = size;
len = GFC_DESCRIPTOR_EXTENT(array,dim);
}
else
{
count[n] = 0;
extent[n] = GFC_DESCRIPTOR_EXTENT(array,dim);
rstride[n] = GFC_DESCRIPTOR_STRIDE_BYTES(ret,dim);
sstride[n] = GFC_DESCRIPTOR_STRIDE_BYTES(array,dim);
n++;
}
}
if (sstride[0] == 0)
sstride[0] = size;
if (rstride[0] == 0)
rstride[0] = size;
dim = GFC_DESCRIPTOR_RANK (array);
rstride0 = rstride[0];
sstride0 = sstride[0];
rptr = ret->base_addr;
sptr = array->base_addr;
shift = len == 0 ? 0 : shift % (ptrdiff_t)len;
if (shift < 0)
shift += len;
while (rptr)
{
/* Do the shift for this dimension. */
/* If elements are contiguous, perform the operation
in two block moves. */
if (soffset == size && roffset == size)
{
size_t len1 = shift * size;
size_t len2 = (len - shift) * size;
memcpy (rptr, sptr + len1, len2);
memcpy (rptr + len2, sptr, len1);
}
else
{
/* Otherwise, we'll have to perform the copy one element at
a time. */
char *dest = rptr;
const char *src = &sptr[shift * soffset];
for (n = 0; n < len - shift; n++)
{
memcpy (dest, src, size);
dest += roffset;
src += soffset;
}
for (src = sptr, n = 0; n < shift; n++)
{
memcpy (dest, src, size);
dest += roffset;
src += soffset;
}
}
/* Advance to the next section. */
rptr += rstride0;
sptr += sstride0;
count[0]++;
n = 0;
while (count[n] == extent[n])
{
/* When we get to the end of a dimension, reset it and increment
the next dimension. */
count[n] = 0;
/* We could precalculate these products, but this is a less
frequently used path so probably not worth it. */
rptr -= rstride[n] * extent[n];
sptr -= sstride[n] * extent[n];
n++;
if (n >= dim - 1)
{
/* Break out of the loop. */
rptr = NULL;
break;
}
else
{
count[n]++;
rptr += rstride[n];
sptr += sstride[n];
}
}
}
}
index_type dim;
index_type len;
index_type n;
index_type size;
index_type arraysize;
int which;
GFC_INTEGER_16 sh;
GFC_INTEGER_16 delta;
/* The compiler cannot figure out that these are set, initialize
them to avoid warnings. */
len = 0;
soffset = 0;
roffset = 0;
arraysize = size0 ((array_t *) array);
size = GFC_DESCRIPTOR_SIZE(array);
if (pwhich)
which = *pwhich - 1;
else
which = 0;
if (ret->base_addr == NULL)
{
int i;
ret->base_addr = xmallocarray (arraysize, size);
ret->offset = 0;
ret->dtype = array->dtype;
for (i = 0; i < GFC_DESCRIPTOR_RANK (array); i++)
static void
cshift1 (gfc_array_char * ret, const gfc_array_char * array,
const gfc_array_i4 * h, const GFC_INTEGER_4 * pwhich, index_type size)
{
/* r.* indicates the return array. */
index_type rstride[GFC_MAX_DIMENSIONS];
index_type rstride0;
index_type roffset;
char *rptr;
char *dest;
/* s.* indicates the source array. */
index_type sstride[GFC_MAX_DIMENSIONS];
index_type sstride0;
index_type soffset;
const char *sptr;
const char *src;
/* h.* indicates the array. */
index_type hstride[GFC_MAX_DIMENSIONS];
index_type hstride0;
const GFC_INTEGER_4 *hptr;
index_type count[GFC_MAX_DIMENSIONS];
index_type extent[GFC_MAX_DIMENSIONS];
index_type dim;
index_type len;
index_type n;
int which;
GFC_INTEGER_4 sh;
if (pwhich)
which = *pwhich - 1;
else
which = 0;
if (which < 0 || (which + 1) > GFC_DESCRIPTOR_RANK (array))
runtime_error ("Argument 'DIM' is out of range in call to 'CSHIFT'");
if (ret->data == NULL)
{
int i;
ret->data = internal_malloc_size (size * size0 ((array_t *)array));
ret->offset = 0;
ret->dtype = array->dtype;
for (i = 0; i < GFC_DESCRIPTOR_RANK (array); i++)
{
ret->dim[i].lbound = 0;
ret->dim[i].ubound = array->dim[i].ubound - array->dim[i].lbound;
if (i == 0)
ret->dim[i].stride = 1;
else
ret->dim[i].stride = (ret->dim[i-1].ubound + 1) * ret->dim[i-1].stride;
}
}
extent[0] = 1;
count[0] = 0;
n = 0;
/* Initialized for avoiding compiler warnings. */
roffset = size;
soffset = size;
len = 0;
for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++)
{
if (dim == which)
{
roffset = ret->dim[dim].stride * size;
if (roffset == 0)
roffset = size;
soffset = array->dim[dim].stride * size;
if (soffset == 0)
soffset = size;
len = array->dim[dim].ubound + 1 - array->dim[dim].lbound;
}
else
{
count[n] = 0;
extent[n] = array->dim[dim].ubound + 1 - array->dim[dim].lbound;
rstride[n] = ret->dim[dim].stride * size;
sstride[n] = array->dim[dim].stride * size;
hstride[n] = h->dim[n].stride;
n++;
}
}
if (sstride[0] == 0)
sstride[0] = size;
if (rstride[0] == 0)
rstride[0] = size;
if (hstride[0] == 0)
hstride[0] = 1;
dim = GFC_DESCRIPTOR_RANK (array);
rstride0 = rstride[0];
sstride0 = sstride[0];
hstride0 = hstride[0];
rptr = ret->data;
sptr = array->data;
hptr = h->data;
while (rptr)
{
/* Do the for this dimension. */
sh = *hptr;
sh = (div (sh, len)).rem;
if (sh < 0)
sh += len;
src = &sptr[sh * soffset];
dest = rptr;
for (n = 0; n < len; n++)
{
memcpy (dest, src, size);
dest += roffset;
if (n == len - sh - 1)
src = sptr;
else
src += soffset;
}
/* Advance to the next section. */
rptr += rstride0;
sptr += sstride0;
hptr += hstride0;
count[0]++;
n = 0;
while (count[n] == extent[n])
{
/* When we get to the end of a dimension, reset it and increment
the next dimension. */
count[n] = 0;
/* We could precalculate these products, but this is a less
frequently used path so proabably not worth it. */
rptr -= rstride[n] * extent[n];
sptr -= sstride[n] * extent[n];
hptr -= hstride[n] * extent[n];
n++;
if (n >= dim - 1)
{
/* Break out of the loop. */
rptr = NULL;
break;
}
else
{
count[n]++;
rptr += rstride[n];
sptr += sstride[n];
hptr += hstride[n];
}
}
}
}
void
matmul_l8 (gfc_array_l8 * retarray, gfc_array_l4 * a, gfc_array_l4 * b)
{
GFC_INTEGER_4 *abase;
GFC_INTEGER_4 *bbase;
GFC_LOGICAL_8 *dest;
index_type rxstride;
index_type rystride;
index_type xcount;
index_type ycount;
index_type xstride;
index_type ystride;
index_type x;
index_type y;
GFC_INTEGER_4 *pa;
GFC_INTEGER_4 *pb;
index_type astride;
index_type bstride;
index_type count;
index_type n;
assert (GFC_DESCRIPTOR_RANK (a) == 2
|| GFC_DESCRIPTOR_RANK (b) == 2);
if (retarray->data == NULL)
{
if (GFC_DESCRIPTOR_RANK (a) == 1)
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = b->dim[1].ubound - b->dim[1].lbound;
retarray->dim[0].stride = 1;
}
else if (GFC_DESCRIPTOR_RANK (b) == 1)
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
retarray->dim[0].stride = 1;
}
else
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
retarray->dim[0].stride = 1;
retarray->dim[1].lbound = 0;
retarray->dim[1].ubound = b->dim[1].ubound - b->dim[1].lbound;
retarray->dim[1].stride = retarray->dim[0].ubound+1;
}
retarray->data
= internal_malloc_size (sizeof (GFC_LOGICAL_8) * size0 ((array_t *) retarray));
retarray->base = 0;
}
abase = a->data;
if (GFC_DESCRIPTOR_SIZE (a) != 4)
{
assert (GFC_DESCRIPTOR_SIZE (a) == 8);
abase = GFOR_POINTER_L8_TO_L4 (abase);
}
bbase = b->data;
if (GFC_DESCRIPTOR_SIZE (b) != 4)
{
assert (GFC_DESCRIPTOR_SIZE (b) == 8);
bbase = GFOR_POINTER_L8_TO_L4 (bbase);
}
dest = retarray->data;
if (retarray->dim[0].stride == 0)
retarray->dim[0].stride = 1;
if (a->dim[0].stride == 0)
a->dim[0].stride = 1;
if (b->dim[0].stride == 0)
b->dim[0].stride = 1;
if (GFC_DESCRIPTOR_RANK (retarray) == 1)
{
rxstride = retarray->dim[0].stride;
rystride = rxstride;
}
else
{
rxstride = retarray->dim[0].stride;
rystride = retarray->dim[1].stride;
}
/* If we have rank 1 parameters, zero the absent stride, and set the size to
one. */
if (GFC_DESCRIPTOR_RANK (a) == 1)
{
astride = a->dim[0].stride;
count = a->dim[0].ubound + 1 - a->dim[0].lbound;
xstride = 0;
rxstride = 0;
xcount = 1;
}
else
{
astride = a->dim[1].stride;
count = a->dim[1].ubound + 1 - a->dim[1].lbound;
xstride = a->dim[0].stride;
xcount = a->dim[0].ubound + 1 - a->dim[0].lbound;
}
if (GFC_DESCRIPTOR_RANK (b) == 1)
{
bstride = b->dim[0].stride;
assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
ystride = 0;
rystride = 0;
ycount = 1;
}
else
{
bstride = b->dim[0].stride;
assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
ystride = b->dim[1].stride;
ycount = b->dim[1].ubound + 1 - b->dim[1].lbound;
}
for (y = 0; y < ycount; y++)
{
for (x = 0; x < xcount; x++)
{
/* Do the summation for this element. For real and integer types
this is the same as DOT_PRODUCT. For complex types we use do
a*b, not conjg(a)*b. */
pa = abase;
pb = bbase;
*dest = 0;
for (n = 0; n < count; n++)
{
if (*pa && *pb)
{
*dest = 1;
break;
}
pa += astride;
pb += bstride;
}
dest += rxstride;
abase += xstride;
}
abase -= xstride * xcount;
bbase += ystride;
dest += rystride - (rxstride * xcount);
}
}
/* Dimensions: retarray(x,y) a(x, count) b(count,y).
Either a or b can be rank 1. In this case x or y is 1. */
void
__matmul_c8 (gfc_array_c8 * retarray, gfc_array_c8 * a, gfc_array_c8 * b)
{
GFC_COMPLEX_8 *abase;
GFC_COMPLEX_8 *bbase;
GFC_COMPLEX_8 *dest;
GFC_COMPLEX_8 res;
index_type rxstride;
index_type rystride;
index_type xcount;
index_type ycount;
index_type xstride;
index_type ystride;
index_type x;
index_type y;
GFC_COMPLEX_8 *pa;
GFC_COMPLEX_8 *pb;
index_type astride;
index_type bstride;
index_type count;
index_type n;
assert (GFC_DESCRIPTOR_RANK (a) == 2
|| GFC_DESCRIPTOR_RANK (b) == 2);
if (retarray->data == NULL)
{
if (GFC_DESCRIPTOR_RANK (a) == 1)
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = b->dim[1].ubound - b->dim[1].lbound;
retarray->dim[0].stride = 1;
}
else if (GFC_DESCRIPTOR_RANK (b) == 1)
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
retarray->dim[0].stride = 1;
}
else
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
retarray->dim[0].stride = 1;
retarray->dim[1].lbound = 0;
retarray->dim[1].ubound = b->dim[1].ubound - b->dim[1].lbound;
retarray->dim[1].stride = retarray->dim[0].ubound+1;
}
retarray->data = internal_malloc (sizeof (GFC_COMPLEX_8) * size0 (retarray));
retarray->base = 0;
}
abase = a->data;
bbase = b->data;
dest = retarray->data;
if (retarray->dim[0].stride == 0)
retarray->dim[0].stride = 1;
if (a->dim[0].stride == 0)
a->dim[0].stride = 1;
if (b->dim[0].stride == 0)
b->dim[0].stride = 1;
if (GFC_DESCRIPTOR_RANK (retarray) == 1)
{
rxstride = retarray->dim[0].stride;
rystride = rxstride;
}
else
{
rxstride = retarray->dim[0].stride;
rystride = retarray->dim[1].stride;
}
/* If we have rank 1 parameters, zero the absent stride, and set the size to
one. */
if (GFC_DESCRIPTOR_RANK (a) == 1)
{
astride = a->dim[0].stride;
count = a->dim[0].ubound + 1 - a->dim[0].lbound;
xstride = 0;
rxstride = 0;
xcount = 1;
}
else
{
astride = a->dim[1].stride;
count = a->dim[1].ubound + 1 - a->dim[1].lbound;
xstride = a->dim[0].stride;
xcount = a->dim[0].ubound + 1 - a->dim[0].lbound;
}
if (GFC_DESCRIPTOR_RANK (b) == 1)
{
bstride = b->dim[0].stride;
assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
ystride = 0;
rystride = 0;
ycount = 1;
}
else
{
bstride = b->dim[0].stride;
assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
ystride = b->dim[1].stride;
ycount = b->dim[1].ubound + 1 - b->dim[1].lbound;
}
for (y = 0; y < ycount; y++)
{
for (x = 0; x < xcount; x++)
{
/* Do the summation for this element. For real and integer types
this is the same as DOT_PRODUCT. For complex types we use do
a*b, not conjg(a)*b. */
pa = abase;
pb = bbase;
res = 0;
for (n = 0; n < count; n++)
{
res += *pa * *pb;
pa += astride;
pb += bstride;
}
*dest = res;
dest += rxstride;
abase += xstride;
}
abase -= xstride * xcount;
bbase += ystride;
dest += rystride - (rxstride * xcount);
}
}
void
eoshift3_4 (gfc_array_char *ret, gfc_array_char *array,
gfc_array_i4 *h, const gfc_array_char *bound,
GFC_INTEGER_4 *pwhich)
{
/* r.* indicates the return array. */
index_type rstride[GFC_MAX_DIMENSIONS];
index_type rstride0;
index_type roffset;
char *rptr;
char *dest;
/* s.* indicates the source array. */
index_type sstride[GFC_MAX_DIMENSIONS];
index_type sstride0;
index_type soffset;
const char *sptr;
const char *src;
/* h.* indicates the shift array. */
index_type hstride[GFC_MAX_DIMENSIONS];
index_type hstride0;
const GFC_INTEGER_4 *hptr;
/* b.* indicates the bound array. */
index_type bstride[GFC_MAX_DIMENSIONS];
index_type bstride0;
const char *bptr;
index_type count[GFC_MAX_DIMENSIONS];
index_type extent[GFC_MAX_DIMENSIONS];
index_type dim;
index_type size;
index_type len;
index_type n;
int which;
GFC_INTEGER_4 sh;
GFC_INTEGER_4 delta;
if (pwhich)
which = *pwhich - 1;
else
which = 0;
size = GFC_DESCRIPTOR_SIZE (ret);
if (ret->data == NULL)
{
int i;
ret->data = internal_malloc_size (size * size0 ((array_t *)array));
ret->base = 0;
ret->dtype = array->dtype;
for (i = 0; i < GFC_DESCRIPTOR_RANK (array); i++)
{
ret->dim[i].lbound = 0;
ret->dim[i].ubound = array->dim[i].ubound - array->dim[i].lbound;
if (i == 0)
ret->dim[i].stride = 1;
else
ret->dim[i].stride = (ret->dim[i-1].ubound + 1) * ret->dim[i-1].stride;
}
}
extent[0] = 1;
count[0] = 0;
size = GFC_DESCRIPTOR_SIZE (array);
n = 0;
for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++)
{
if (dim == which)
{
roffset = ret->dim[dim].stride * size;
if (roffset == 0)
roffset = size;
soffset = array->dim[dim].stride * size;
if (soffset == 0)
soffset = size;
len = array->dim[dim].ubound + 1 - array->dim[dim].lbound;
}
else
{
count[n] = 0;
extent[n] = array->dim[dim].ubound + 1 - array->dim[dim].lbound;
rstride[n] = ret->dim[dim].stride * size;
sstride[n] = array->dim[dim].stride * size;
hstride[n] = h->dim[n].stride;
if (bound)
bstride[n] = bound->dim[n].stride * size;
else
bstride[n] = 0;
n++;
}
}
if (sstride[0] == 0)
sstride[0] = size;
if (rstride[0] == 0)
rstride[0] = size;
if (hstride[0] == 0)
hstride[0] = 1;
if (bound && bstride[0] == 0)
bstride[0] = size;
dim = GFC_DESCRIPTOR_RANK (array);
rstride0 = rstride[0];
sstride0 = sstride[0];
hstride0 = hstride[0];
bstride0 = bstride[0];
rptr = ret->data;
sptr = array->data;
hptr = h->data;
if (bound)
bptr = bound->data;
else
bptr = zeros;
while (rptr)
{
/* Do the shift for this dimension. */
sh = *hptr;
if (( sh >= 0 ? sh : -sh ) > len)
{
delta = len;
sh = len;
}
else
delta = (sh >= 0) ? sh: -sh;
if (sh > 0)
{
src = &sptr[delta * soffset];
dest = rptr;
}
else
{
src = sptr;
dest = &rptr[delta * roffset];
}
for (n = 0; n < len - delta; n++)
{
memcpy (dest, src, size);
dest += roffset;
src += soffset;
}
if (sh < 0)
dest = rptr;
n = delta;
while (n--)
{
memcpy (dest, bptr, size);
dest += roffset;
}
/* Advance to the next section. */
rptr += rstride0;
sptr += sstride0;
hptr += hstride0;
bptr += bstride0;
count[0]++;
n = 0;
while (count[n] == extent[n])
{
/* When we get to the end of a dimension, reset it and increment
the next dimension. */
count[n] = 0;
/* We could precalculate these products, but this is a less
frequently used path so proabably not worth it. */
rptr -= rstride[n] * extent[n];
sptr -= sstride[n] * extent[n];
hptr -= hstride[n] * extent[n];
bptr -= bstride[n] * extent[n];
n++;
if (n >= dim - 1)
{
/* Break out of the loop. */
rptr = NULL;
break;
}
else
{
count[n]++;
rptr += rstride[n];
sptr += sstride[n];
hptr += hstride[n];
bptr += bstride[n];
}
}
}
}
static void
transpose_internal (gfc_array_char *ret, gfc_array_char *source)
{
/* r.* indicates the return array. */
index_type rxstride, rystride;
char *rptr;
/* s.* indicates the source array. */
index_type sxstride, systride;
const char *sptr;
index_type xcount, ycount;
index_type x, y;
index_type size;
assert (GFC_DESCRIPTOR_RANK (source) == 2
&& GFC_DESCRIPTOR_RANK (ret) == 2);
size = GFC_DESCRIPTOR_SIZE(ret);
if (ret->base_addr == NULL)
{
assert (ret->dtype == source->dtype);
GFC_DIMENSION_SET(ret->dim[0], 0, GFC_DESCRIPTOR_EXTENT(source,1) - 1,
1);
GFC_DIMENSION_SET(ret->dim[1], 0, GFC_DESCRIPTOR_EXTENT(source,0) - 1,
GFC_DESCRIPTOR_EXTENT(source, 1));
ret->base_addr = xmallocarray (size0 ((array_t*)ret), size);
ret->offset = 0;
}
else if (unlikely (compile_options.bounds_check))
{
index_type ret_extent, src_extent;
ret_extent = GFC_DESCRIPTOR_EXTENT(ret,0);
src_extent = GFC_DESCRIPTOR_EXTENT(source,1);
if (src_extent != ret_extent)
runtime_error ("Incorrect extent in return value of TRANSPOSE"
" intrinsic in dimension 1: is %ld,"
" should be %ld", (long int) src_extent,
(long int) ret_extent);
ret_extent = GFC_DESCRIPTOR_EXTENT(ret,1);
src_extent = GFC_DESCRIPTOR_EXTENT(source,0);
if (src_extent != ret_extent)
runtime_error ("Incorrect extent in return value of TRANSPOSE"
" intrinsic in dimension 2: is %ld,"
" should be %ld", (long int) src_extent,
(long int) ret_extent);
}
sxstride = GFC_DESCRIPTOR_STRIDE_BYTES(source,0);
systride = GFC_DESCRIPTOR_STRIDE_BYTES(source,1);
xcount = GFC_DESCRIPTOR_EXTENT(source,0);
ycount = GFC_DESCRIPTOR_EXTENT(source,1);
rxstride = GFC_DESCRIPTOR_STRIDE_BYTES(ret,0);
rystride = GFC_DESCRIPTOR_STRIDE_BYTES(ret,1);
rptr = ret->base_addr;
sptr = source->base_addr;
for (y = 0; y < ycount; y++)
{
for (x = 0; x < xcount; x++)
{
memcpy (rptr, sptr, size);
sptr += sxstride;
rptr += rystride;
}
sptr += systride - (sxstride * xcount);
rptr += rxstride - (rystride * xcount);
}
}
void
matmul_i8 (gfc_array_i8 * retarray, gfc_array_i8 * a, gfc_array_i8 * b)
{
GFC_INTEGER_8 *abase;
GFC_INTEGER_8 *bbase;
GFC_INTEGER_8 *dest;
index_type rxstride, rystride, axstride, aystride, bxstride, bystride;
index_type x, y, n, count, xcount, ycount;
assert (GFC_DESCRIPTOR_RANK (a) == 2
|| GFC_DESCRIPTOR_RANK (b) == 2);
/* C[xcount,ycount] = A[xcount, count] * B[count,ycount]
Either A or B (but not both) can be rank 1:
o One-dimensional argument A is implicitly treated as a row matrix
dimensioned [1,count], so xcount=1.
o One-dimensional argument B is implicitly treated as a column matrix
dimensioned [count, 1], so ycount=1.
*/
if (retarray->data == NULL)
{
if (GFC_DESCRIPTOR_RANK (a) == 1)
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = b->dim[1].ubound - b->dim[1].lbound;
retarray->dim[0].stride = 1;
}
else if (GFC_DESCRIPTOR_RANK (b) == 1)
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
retarray->dim[0].stride = 1;
}
else
{
retarray->dim[0].lbound = 0;
retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
retarray->dim[0].stride = 1;
retarray->dim[1].lbound = 0;
retarray->dim[1].ubound = b->dim[1].ubound - b->dim[1].lbound;
retarray->dim[1].stride = retarray->dim[0].ubound+1;
}
retarray->data
= internal_malloc_size (sizeof (GFC_INTEGER_8) * size0 (retarray));
retarray->base = 0;
}
abase = a->data;
bbase = b->data;
dest = retarray->data;
if (retarray->dim[0].stride == 0)
retarray->dim[0].stride = 1;
if (a->dim[0].stride == 0)
a->dim[0].stride = 1;
if (b->dim[0].stride == 0)
b->dim[0].stride = 1;
if (GFC_DESCRIPTOR_RANK (retarray) == 1)
{
/* One-dimensional result may be addressed in the code below
either as a row or a column matrix. We want both cases to
work. */
rxstride = rystride = retarray->dim[0].stride;
}
else
{
rxstride = retarray->dim[0].stride;
rystride = retarray->dim[1].stride;
}
if (GFC_DESCRIPTOR_RANK (a) == 1)
{
/* Treat it as a a row matrix A[1,count]. */
axstride = a->dim[0].stride;
aystride = 1;
xcount = 1;
count = a->dim[0].ubound + 1 - a->dim[0].lbound;
}
else
{
axstride = a->dim[0].stride;
aystride = a->dim[1].stride;
count = a->dim[1].ubound + 1 - a->dim[1].lbound;
xcount = a->dim[0].ubound + 1 - a->dim[0].lbound;
}
assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
if (GFC_DESCRIPTOR_RANK (b) == 1)
{
/* Treat it as a column matrix B[count,1] */
bxstride = b->dim[0].stride;
/* bystride should never be used for 1-dimensional b.
in case it is we want it to cause a segfault, rather than
an incorrect result. */
bystride = 0xDEADBEEF;
ycount = 1;
}
else
{
bxstride = b->dim[0].stride;
bystride = b->dim[1].stride;
ycount = b->dim[1].ubound + 1 - b->dim[1].lbound;
}
abase = a->data;
bbase = b->data;
dest = retarray->data;
if (rxstride == 1 && axstride == 1 && bxstride == 1)
{
GFC_INTEGER_8 *bbase_y;
GFC_INTEGER_8 *dest_y;
GFC_INTEGER_8 *abase_n;
GFC_INTEGER_8 bbase_yn;
if (rystride == ycount)
memset (dest, 0, (sizeof (GFC_INTEGER_8) * size0((array_t *) retarray)));
else
{
for (y = 0; y < ycount; y++)
for (x = 0; x < xcount; x++)
dest[x + y*rystride] = (GFC_INTEGER_8)0;
}
for (y = 0; y < ycount; y++)
{
bbase_y = bbase + y*bystride;
dest_y = dest + y*rystride;
for (n = 0; n < count; n++)
{
abase_n = abase + n*aystride;
bbase_yn = bbase_y[n];
for (x = 0; x < xcount; x++)
{
dest_y[x] += abase_n[x] * bbase_yn;
}
}
}
}
else
{
for (y = 0; y < ycount; y++)
for (x = 0; x < xcount; x++)
dest[x*rxstride + y*rystride] = (GFC_INTEGER_8)0;
for (y = 0; y < ycount; y++)
for (n = 0; n < count; n++)
for (x = 0; x < xcount; x++)
/* dest[x,y] += a[x,n] * b[n,y] */
dest[x*rxstride + y*rystride] += abase[x*axstride + n*aystride] * bbase[n*bxstride + y*bystride];
}
}
static void
eoshift0 (gfc_array_char * ret, const gfc_array_char * array,
int shift, const char * pbound, int which, index_type size,
const char *filler, index_type filler_len)
{
/* r.* indicates the return array. */
index_type rstride[GFC_MAX_DIMENSIONS];
index_type rstride0;
index_type roffset;
char * restrict rptr;
char *dest;
/* s.* indicates the source array. */
index_type sstride[GFC_MAX_DIMENSIONS];
index_type sstride0;
index_type soffset;
const char *sptr;
const char *src;
index_type count[GFC_MAX_DIMENSIONS];
index_type extent[GFC_MAX_DIMENSIONS];
index_type dim;
index_type len;
index_type n;
index_type arraysize;
/* The compiler cannot figure out that these are set, initialize
them to avoid warnings. */
len = 0;
soffset = 0;
roffset = 0;
arraysize = size0 ((array_t *) array);
if (ret->base_addr == NULL)
{
int i;
ret->offset = 0;
ret->dtype = array->dtype;
for (i = 0; i < GFC_DESCRIPTOR_RANK (array); i++)
{
index_type ub, str;
ub = GFC_DESCRIPTOR_EXTENT(array,i) - 1;
if (i == 0)
str = 1;
else
str = GFC_DESCRIPTOR_EXTENT(ret,i-1)
* GFC_DESCRIPTOR_STRIDE(ret,i-1);
GFC_DIMENSION_SET(ret->dim[i], 0, ub, str);
}
/* xmalloc allocates a single byte for zero size. */
ret->base_addr = xmalloc (size * arraysize);
}
else if (unlikely (compile_options.bounds_check))
{
bounds_equal_extents ((array_t *) ret, (array_t *) array,
"return value", "EOSHIFT");
}
if (arraysize == 0)
return;
which = which - 1;
extent[0] = 1;
count[0] = 0;
sstride[0] = -1;
rstride[0] = -1;
n = 0;
for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++)
{
if (dim == which)
{
roffset = GFC_DESCRIPTOR_STRIDE_BYTES(ret,dim);
if (roffset == 0)
roffset = size;
soffset = GFC_DESCRIPTOR_STRIDE_BYTES(array,dim);
if (soffset == 0)
soffset = size;
len = GFC_DESCRIPTOR_EXTENT(array,dim);
}
else
{
count[n] = 0;
extent[n] = GFC_DESCRIPTOR_EXTENT(array,dim);
rstride[n] = GFC_DESCRIPTOR_STRIDE_BYTES(ret,dim);
sstride[n] = GFC_DESCRIPTOR_STRIDE_BYTES(array,dim);
n++;
}
}
if (sstride[0] == 0)
sstride[0] = size;
if (rstride[0] == 0)
rstride[0] = size;
dim = GFC_DESCRIPTOR_RANK (array);
rstride0 = rstride[0];
sstride0 = sstride[0];
rptr = ret->base_addr;
sptr = array->base_addr;
if ((shift >= 0 ? shift : -shift) > len)
{
shift = len;
len = 0;
}
else
{
if (shift > 0)
len = len - shift;
else
len = len + shift;
}
while (rptr)
{
/* Do the shift for this dimension. */
if (shift > 0)
{
src = &sptr[shift * soffset];
dest = rptr;
}
else
{
src = sptr;
dest = &rptr[-shift * roffset];
}
for (n = 0; n < len; n++)
{
memcpy (dest, src, size);
dest += roffset;
src += soffset;
}
if (shift >= 0)
{
n = shift;
}
else
{
dest = rptr;
n = -shift;
}
if (pbound)
while (n--)
{
memcpy (dest, pbound, size);
dest += roffset;
}
else
while (n--)
{
index_type i;
if (filler_len == 1)
memset (dest, filler[0], size);
else
for (i = 0; i < size ; i += filler_len)
memcpy (&dest[i], filler, filler_len);
dest += roffset;
}
/* Advance to the next section. */
rptr += rstride0;
sptr += sstride0;
count[0]++;
n = 0;
while (count[n] == extent[n])
{
/* When we get to the end of a dimension, reset it and increment
the next dimension. */
count[n] = 0;
/* We could precalculate these products, but this is a less
frequently used path so probably not worth it. */
rptr -= rstride[n] * extent[n];
sptr -= sstride[n] * extent[n];
n++;
if (n >= dim - 1)
{
/* Break out of the loop. */
rptr = NULL;
break;
}
else
{
count[n]++;
rptr += rstride[n];
sptr += sstride[n];
}
}
}
}
void ScaledMeshCollision (DemoEntityManager* const scene)
{
// load the skybox
scene->CreateSkyBox();
// load the scene from a ngd file format
CreateLevelMesh (scene, "flatPlane.ngd", 1);
//CreateLevelMesh (scene, "flatPlaneDoubleFace.ngd", 1);
//CreateLevelMesh (scene, "sponza.ngd", 0);
//CreateLevelMesh (scene, "cattle.ngd", fileName);
//CreateLevelMesh (scene, "playground.ngd", 0);
//dMatrix camMatrix (dRollMatrix(-20.0f * 3.1416f /180.0f) * dYawMatrix(-45.0f * 3.1416f /180.0f));
dMatrix camMatrix (dGetIdentityMatrix());
dQuaternion rot (camMatrix);
dVector origin (-15.0f, 5.0f, 0.0f, 0.0f);
//origin = origin.Scale (0.25f);
scene->SetCameraMatrix(rot, origin);
NewtonWorld* const world = scene->GetNewton();
int defaultMaterialID = NewtonMaterialGetDefaultGroupID (world);
dVector location (0.0f, 0.0f, 0.0f, 0.0f);
dMatrix matrix (dGetIdentityMatrix());
matrix.m_posit = location;
matrix.m_posit.m_x = 0.0f;
matrix.m_posit.m_y = 0.0f;
matrix.m_posit.m_z = 0.0f;
matrix.m_posit.m_w = 1.0f;
DemoEntity teaPot (dGetIdentityMatrix(), NULL);
teaPot.LoadNGD_mesh("teapot.ngd", world);
//teaPot.LoadNGD_mesh("box.ngd", world);
NewtonCollision* const staticCollision = CreateCollisionTree (world, &teaPot, 0, true);
CreateScaleStaticMesh (&teaPot, staticCollision, scene, matrix, dVector (1.0f, 1.0f, 1.0f, 0.0f));
// CreateScaleStaticMesh (&teaPot, staticCollision, scene, matrix, dVector (0.5f, 0.5f, 2.0f, 0.0f));
matrix.m_posit.m_z = -5.0f;
CreateScaleStaticMesh (&teaPot, staticCollision, scene, matrix, dVector (0.5f, 0.5f, 2.0f, 0.0f));
matrix.m_posit.m_z = 5.0f;
CreateScaleStaticMesh (&teaPot, staticCollision, scene, matrix, dVector (3.0f, 3.0f, 1.5f, 0.0f));
matrix.m_posit.m_z = 0.0f;
matrix.m_posit.m_x = -5.0f;
CreateScaleStaticMesh (&teaPot, staticCollision, scene, matrix, dVector (0.5f, 0.5f, 0.5f, 0.0f));
matrix.m_posit.m_x = 5.0f;
CreateScaleStaticMesh (&teaPot, staticCollision, scene, matrix, dVector (2.0f, 2.0f, 2.0f, 0.0f));
// do not forget to destroy the collision mesh helper
NewtonDestroyCollision(staticCollision);
dVector size0 (1.0f, 1.0f, 1.0f, 0.0f);
dVector size1 (0.5f, 1.0f, 1.0f, 0.0f);
dMatrix shapeOffsetMatrix (dRollMatrix(3.141592f/2.0f));
int count = 3;
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 5.0f, _SPHERE_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 5.0f, _BOX_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 5.0f, _CAPSULE_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size1, count, count, 5.0f, _CAPSULE_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size1, count, count, 5.0f, _CYLINDER_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 5.0f, _CYLINDER_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 5.0f, _CHAMFER_CYLINDER_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 5.0f, _CONE_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 5.0f, _REGULAR_CONVEX_HULL_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
AddNonUniformScaledPrimitives(scene, 10.0f, location, size0, count, count, 5.0f, _RANDOM_CONVEX_HULL_PRIMITIVE, defaultMaterialID, shapeOffsetMatrix);
origin.m_x -= 4.0f;
origin.m_y += 1.0f;
scene->SetCameraMatrix(rot, origin);
}
void
transpose_i4 (gfc_array_i4 * ret, gfc_array_i4 * source)
{
/* r.* indicates the return array. */
index_type rxstride, rystride;
GFC_INTEGER_4 *rptr;
/* s.* indicates the source array. */
index_type sxstride, systride;
const GFC_INTEGER_4 *sptr;
index_type xcount, ycount;
index_type x, y;
assert (GFC_DESCRIPTOR_RANK (source) == 2);
if (ret->data == NULL)
{
assert (GFC_DESCRIPTOR_RANK (ret) == 2);
assert (ret->dtype == source->dtype);
ret->dim[0].lbound = 0;
ret->dim[0].ubound = source->dim[1].ubound - source->dim[1].lbound;
ret->dim[0].stride = 1;
ret->dim[1].lbound = 0;
ret->dim[1].ubound = source->dim[0].ubound - source->dim[0].lbound;
ret->dim[1].stride = ret->dim[0].ubound+1;
ret->data = internal_malloc_size (sizeof (GFC_INTEGER_4) * size0 ((array_t *) ret));
ret->offset = 0;
}
if (ret->dim[0].stride == 0)
ret->dim[0].stride = 1;
if (source->dim[0].stride == 0)
source->dim[0].stride = 1;
sxstride = source->dim[0].stride;
systride = source->dim[1].stride;
xcount = source->dim[0].ubound + 1 - source->dim[0].lbound;
ycount = source->dim[1].ubound + 1 - source->dim[1].lbound;
rxstride = ret->dim[0].stride;
rystride = ret->dim[1].stride;
rptr = ret->data;
sptr = source->data;
for (y=0; y < ycount; y++)
{
for (x=0; x < xcount; x++)
{
*rptr = *sptr;
sptr += sxstride;
rptr += rystride;
}
sptr += systride - (sxstride * xcount);
rptr += rxstride - (rystride * xcount);
}
}
void
transpose (gfc_array_char *ret, gfc_array_char *source)
{
/* r.* indicates the return array. */
index_type rxstride, rystride;
char *rptr;
/* s.* indicates the source array. */
index_type sxstride, systride;
const char *sptr;
index_type xcount, ycount;
index_type x, y;
index_type size;
assert (GFC_DESCRIPTOR_RANK (source) == 2
&& GFC_DESCRIPTOR_RANK (ret) == 2);
size = GFC_DESCRIPTOR_SIZE (source);
if (ret->data == NULL)
{
assert (ret->dtype == source->dtype);
ret->dim[0].lbound = 0;
ret->dim[0].ubound = source->dim[1].ubound - source->dim[1].lbound;
ret->dim[0].stride = 1;
ret->dim[1].lbound = 0;
ret->dim[1].ubound = source->dim[0].ubound - source->dim[0].lbound;
ret->dim[1].stride = ret->dim[0].ubound+1;
ret->data = internal_malloc_size (size * size0 ((array_t*)ret));
ret->base = 0;
}
sxstride = source->dim[0].stride * size;
if (sxstride == 0)
sxstride = size;
systride = source->dim[1].stride * size;
xcount = source->dim[0].ubound + 1 - source->dim[0].lbound;
ycount = source->dim[1].ubound + 1 - source->dim[1].lbound;
rxstride = ret->dim[0].stride * size;
if (rxstride == 0)
rxstride = size;
rystride = ret->dim[1].stride * size;
rptr = ret->data;
sptr = source->data;
for (y = 0; y < ycount; y++)
{
for (x = 0; x < xcount; x++)
{
memcpy (rptr, sptr, size);
sptr += sxstride;
rptr += rystride;
}
sptr += systride - (sxstride * xcount);
rptr += rxstride - (rystride * xcount);
}
}
static void
cshift0 (gfc_array_char * ret, const gfc_array_char * array,
ssize_t shift, int which, index_type size)
{
/* r.* indicates the return array. */
index_type rstride[GFC_MAX_DIMENSIONS];
index_type rstride0;
index_type roffset;
char *rptr;
/* s.* indicates the source array. */
index_type sstride[GFC_MAX_DIMENSIONS];
index_type sstride0;
index_type soffset;
const char *sptr;
index_type count[GFC_MAX_DIMENSIONS];
index_type extent[GFC_MAX_DIMENSIONS];
index_type dim;
index_type len;
index_type n;
int whichloop;
if (which < 1 || which > GFC_DESCRIPTOR_RANK (array))
runtime_error ("Argument 'DIM' is out of range in call to 'CSHIFT'");
which = which - 1;
extent[0] = 1;
count[0] = 0;
n = 0;
/* The values assigned here must match the cases in the inner loop. */
whichloop = 0;
switch (GFC_DESCRIPTOR_TYPE (array))
{
case GFC_DTYPE_LOGICAL:
case GFC_DTYPE_INTEGER:
case GFC_DTYPE_REAL:
if (size == sizeof (int))
whichloop = 1;
else if (size == sizeof (long))
whichloop = 2;
else if (size == sizeof (double))
whichloop = 3;
else if (size == sizeof (long double))
whichloop = 4;
break;
case GFC_DTYPE_COMPLEX:
if (size == sizeof (_Complex float))
whichloop = 5;
else if (size == sizeof (_Complex double))
whichloop = 6;
break;
default:
break;
}
/* Initialized for avoiding compiler warnings. */
roffset = size;
soffset = size;
len = 0;
if (ret->data == NULL)
{
int i;
index_type arraysize = size0 ((array_t *)array);
ret->offset = 0;
ret->dtype = array->dtype;
for (i = 0; i < GFC_DESCRIPTOR_RANK (array); i++)
{
ret->dim[i].lbound = 0;
ret->dim[i].ubound = array->dim[i].ubound - array->dim[i].lbound;
if (i == 0)
ret->dim[i].stride = 1;
else
ret->dim[i].stride = (ret->dim[i-1].ubound + 1)
* ret->dim[i-1].stride;
}
if (arraysize > 0)
ret->data = internal_malloc_size (size * arraysize);
else
{
ret->data = internal_malloc_size (1);
return;
}
}
for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++)
{
if (dim == which)
{
roffset = ret->dim[dim].stride * size;
if (roffset == 0)
roffset = size;
soffset = array->dim[dim].stride * size;
if (soffset == 0)
soffset = size;
len = array->dim[dim].ubound + 1 - array->dim[dim].lbound;
}
else
{
count[n] = 0;
extent[n] = array->dim[dim].ubound + 1 - array->dim[dim].lbound;
rstride[n] = ret->dim[dim].stride * size;
sstride[n] = array->dim[dim].stride * size;
n++;
}
}
if (sstride[0] == 0)
sstride[0] = size;
if (rstride[0] == 0)
rstride[0] = size;
dim = GFC_DESCRIPTOR_RANK (array);
rstride0 = rstride[0];
sstride0 = sstride[0];
rptr = ret->data;
sptr = array->data;
shift = shift % (ssize_t)len;
if (shift < 0)
shift += len;
while (rptr)
{
/* Do the shift for this dimension. */
/* If elements are contiguous, perform the operation
in two block moves. */
if (soffset == size && roffset == size)
{
size_t len1 = shift * size;
size_t len2 = (len - shift) * size;
memcpy (rptr, sptr + len1, len2);
memcpy (rptr + len2, sptr, len1);
}
else
{
/* Otherwise, we'll have to perform the copy one element at
a time. We can speed this up a tad for common cases of
fundamental types. */
switch (whichloop)
{
case 0:
{
char *dest = rptr;
const char *src = &sptr[shift * soffset];
for (n = 0; n < len - shift; n++)
{
memcpy (dest, src, size);
dest += roffset;
src += soffset;
}
for (src = sptr, n = 0; n < shift; n++)
{
memcpy (dest, src, size);
dest += roffset;
src += soffset;
}
}
break;
case 1:
copy_loop_int (rptr, sptr, roffset, soffset, len, shift);
break;
case 2:
copy_loop_long (rptr, sptr, roffset, soffset, len, shift);
break;
case 3:
copy_loop_double (rptr, sptr, roffset, soffset, len, shift);
break;
case 4:
copy_loop_ldouble (rptr, sptr, roffset, soffset, len, shift);
break;
case 5:
copy_loop_cfloat (rptr, sptr, roffset, soffset, len, shift);
break;
case 6:
copy_loop_cdouble (rptr, sptr, roffset, soffset, len, shift);
break;
default:
abort ();
}
}
/* Advance to the next section. */
rptr += rstride0;
sptr += sstride0;
count[0]++;
n = 0;
while (count[n] == extent[n])
{
/* When we get to the end of a dimension, reset it and increment
the next dimension. */
count[n] = 0;
/* We could precalculate these products, but this is a less
frequently used path so probably not worth it. */
rptr -= rstride[n] * extent[n];
sptr -= sstride[n] * extent[n];
n++;
if (n >= dim - 1)
{
/* Break out of the loop. */
rptr = NULL;
break;
}
else
{
count[n]++;
rptr += rstride[n];
sptr += sstride[n];
}
}
}
}
int avaliablebytestacks() {
return size0();
}
