mirror of
https://sourceware.org/git/binutils-gdb.git
synced 2024-12-14 04:43:38 +08:00
3634f66996
This splits BINOP_ASSIGN into a new function for future use. gdb/ChangeLog 2021-03-08 Tom Tromey <tom@tromey.com> * opencl-lang.c (eval_opencl_assign): New function. (evaluate_subexp_opencl): Use it.
1101 lines
32 KiB
C
1101 lines
32 KiB
C
/* OpenCL language support for GDB, the GNU debugger.
|
|
Copyright (C) 2010-2021 Free Software Foundation, Inc.
|
|
|
|
Contributed by Ken Werner <ken.werner@de.ibm.com>.
|
|
|
|
This file is part of GDB.
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation; either version 3 of the License, or
|
|
(at your option) any later version.
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>. */
|
|
|
|
#include "defs.h"
|
|
#include "gdbtypes.h"
|
|
#include "symtab.h"
|
|
#include "expression.h"
|
|
#include "parser-defs.h"
|
|
#include "language.h"
|
|
#include "varobj.h"
|
|
#include "c-lang.h"
|
|
#include "gdbarch.h"
|
|
|
|
/* Returns the corresponding OpenCL vector type from the given type code,
|
|
the length of the element type, the unsigned flag and the amount of
|
|
elements (N). */
|
|
|
|
static struct type *
|
|
lookup_opencl_vector_type (struct gdbarch *gdbarch, enum type_code code,
|
|
unsigned int el_length, unsigned int flag_unsigned,
|
|
int n)
|
|
{
|
|
unsigned int length;
|
|
|
|
/* Check if n describes a valid OpenCL vector size (2, 3, 4, 8, 16). */
|
|
if (n != 2 && n != 3 && n != 4 && n != 8 && n != 16)
|
|
error (_("Invalid OpenCL vector size: %d"), n);
|
|
|
|
/* Triple vectors have the size of a quad vector. */
|
|
length = (n == 3) ? el_length * 4 : el_length * n;
|
|
|
|
auto filter = [&] (struct type *type)
|
|
{
|
|
LONGEST lowb, highb;
|
|
|
|
return (type->code () == TYPE_CODE_ARRAY && type->is_vector ()
|
|
&& get_array_bounds (type, &lowb, &highb)
|
|
&& TYPE_TARGET_TYPE (type)->code () == code
|
|
&& TYPE_TARGET_TYPE (type)->is_unsigned () == flag_unsigned
|
|
&& TYPE_LENGTH (TYPE_TARGET_TYPE (type)) == el_length
|
|
&& TYPE_LENGTH (type) == length
|
|
&& highb - lowb + 1 == n);
|
|
};
|
|
const struct language_defn *lang = language_def (language_opencl);
|
|
return language_lookup_primitive_type (lang, gdbarch, filter);
|
|
}
|
|
|
|
/* Returns nonzero if the array ARR contains duplicates within
|
|
the first N elements. */
|
|
|
|
static int
|
|
array_has_dups (int *arr, int n)
|
|
{
|
|
int i, j;
|
|
|
|
for (i = 0; i < n; i++)
|
|
{
|
|
for (j = i + 1; j < n; j++)
|
|
{
|
|
if (arr[i] == arr[j])
|
|
return 1;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* The OpenCL component access syntax allows to create lvalues referring to
|
|
selected elements of an original OpenCL vector in arbitrary order. This
|
|
structure holds the information to describe such lvalues. */
|
|
|
|
struct lval_closure
|
|
{
|
|
/* Reference count. */
|
|
int refc;
|
|
/* The number of indices. */
|
|
int n;
|
|
/* The element indices themselves. */
|
|
int *indices;
|
|
/* A pointer to the original value. */
|
|
struct value *val;
|
|
};
|
|
|
|
/* Allocates an instance of struct lval_closure. */
|
|
|
|
static struct lval_closure *
|
|
allocate_lval_closure (int *indices, int n, struct value *val)
|
|
{
|
|
struct lval_closure *c = XCNEW (struct lval_closure);
|
|
|
|
c->refc = 1;
|
|
c->n = n;
|
|
c->indices = XCNEWVEC (int, n);
|
|
memcpy (c->indices, indices, n * sizeof (int));
|
|
value_incref (val); /* Increment the reference counter of the value. */
|
|
c->val = val;
|
|
|
|
return c;
|
|
}
|
|
|
|
static void
|
|
lval_func_read (struct value *v)
|
|
{
|
|
struct lval_closure *c = (struct lval_closure *) value_computed_closure (v);
|
|
struct type *type = check_typedef (value_type (v));
|
|
struct type *eltype = TYPE_TARGET_TYPE (check_typedef (value_type (c->val)));
|
|
LONGEST offset = value_offset (v);
|
|
LONGEST elsize = TYPE_LENGTH (eltype);
|
|
int n, i, j = 0;
|
|
LONGEST lowb = 0;
|
|
LONGEST highb = 0;
|
|
|
|
if (type->code () == TYPE_CODE_ARRAY
|
|
&& !get_array_bounds (type, &lowb, &highb))
|
|
error (_("Could not determine the vector bounds"));
|
|
|
|
/* Assume elsize aligned offset. */
|
|
gdb_assert (offset % elsize == 0);
|
|
offset /= elsize;
|
|
n = offset + highb - lowb + 1;
|
|
gdb_assert (n <= c->n);
|
|
|
|
for (i = offset; i < n; i++)
|
|
memcpy (value_contents_raw (v) + j++ * elsize,
|
|
value_contents (c->val) + c->indices[i] * elsize,
|
|
elsize);
|
|
}
|
|
|
|
static void
|
|
lval_func_write (struct value *v, struct value *fromval)
|
|
{
|
|
struct value *mark = value_mark ();
|
|
struct lval_closure *c = (struct lval_closure *) value_computed_closure (v);
|
|
struct type *type = check_typedef (value_type (v));
|
|
struct type *eltype = TYPE_TARGET_TYPE (check_typedef (value_type (c->val)));
|
|
LONGEST offset = value_offset (v);
|
|
LONGEST elsize = TYPE_LENGTH (eltype);
|
|
int n, i, j = 0;
|
|
LONGEST lowb = 0;
|
|
LONGEST highb = 0;
|
|
|
|
if (type->code () == TYPE_CODE_ARRAY
|
|
&& !get_array_bounds (type, &lowb, &highb))
|
|
error (_("Could not determine the vector bounds"));
|
|
|
|
/* Assume elsize aligned offset. */
|
|
gdb_assert (offset % elsize == 0);
|
|
offset /= elsize;
|
|
n = offset + highb - lowb + 1;
|
|
|
|
/* Since accesses to the fourth component of a triple vector is undefined we
|
|
just skip writes to the fourth element. Imagine something like this:
|
|
int3 i3 = (int3)(0, 1, 2);
|
|
i3.hi.hi = 5;
|
|
In this case n would be 4 (offset=12/4 + 1) while c->n would be 3. */
|
|
if (n > c->n)
|
|
n = c->n;
|
|
|
|
for (i = offset; i < n; i++)
|
|
{
|
|
struct value *from_elm_val = allocate_value (eltype);
|
|
struct value *to_elm_val = value_subscript (c->val, c->indices[i]);
|
|
|
|
memcpy (value_contents_writeable (from_elm_val),
|
|
value_contents (fromval) + j++ * elsize,
|
|
elsize);
|
|
value_assign (to_elm_val, from_elm_val);
|
|
}
|
|
|
|
value_free_to_mark (mark);
|
|
}
|
|
|
|
/* Return nonzero if bits in V from OFFSET and LENGTH represent a
|
|
synthetic pointer. */
|
|
|
|
static int
|
|
lval_func_check_synthetic_pointer (const struct value *v,
|
|
LONGEST offset, int length)
|
|
{
|
|
struct lval_closure *c = (struct lval_closure *) value_computed_closure (v);
|
|
/* Size of the target type in bits. */
|
|
int elsize =
|
|
TYPE_LENGTH (TYPE_TARGET_TYPE (check_typedef (value_type (c->val)))) * 8;
|
|
int startrest = offset % elsize;
|
|
int start = offset / elsize;
|
|
int endrest = (offset + length) % elsize;
|
|
int end = (offset + length) / elsize;
|
|
int i;
|
|
|
|
if (endrest)
|
|
end++;
|
|
|
|
if (end > c->n)
|
|
return 0;
|
|
|
|
for (i = start; i < end; i++)
|
|
{
|
|
int comp_offset = (i == start) ? startrest : 0;
|
|
int comp_length = (i == end) ? endrest : elsize;
|
|
|
|
if (!value_bits_synthetic_pointer (c->val,
|
|
c->indices[i] * elsize + comp_offset,
|
|
comp_length))
|
|
return 0;
|
|
}
|
|
|
|
return 1;
|
|
}
|
|
|
|
static void *
|
|
lval_func_copy_closure (const struct value *v)
|
|
{
|
|
struct lval_closure *c = (struct lval_closure *) value_computed_closure (v);
|
|
|
|
++c->refc;
|
|
|
|
return c;
|
|
}
|
|
|
|
static void
|
|
lval_func_free_closure (struct value *v)
|
|
{
|
|
struct lval_closure *c = (struct lval_closure *) value_computed_closure (v);
|
|
|
|
--c->refc;
|
|
|
|
if (c->refc == 0)
|
|
{
|
|
value_decref (c->val); /* Decrement the reference counter of the value. */
|
|
xfree (c->indices);
|
|
xfree (c);
|
|
}
|
|
}
|
|
|
|
static const struct lval_funcs opencl_value_funcs =
|
|
{
|
|
lval_func_read,
|
|
lval_func_write,
|
|
NULL, /* indirect */
|
|
NULL, /* coerce_ref */
|
|
lval_func_check_synthetic_pointer,
|
|
lval_func_copy_closure,
|
|
lval_func_free_closure
|
|
};
|
|
|
|
/* Creates a sub-vector from VAL. The elements are selected by the indices of
|
|
an array with the length of N. Supported values for NOSIDE are
|
|
EVAL_NORMAL and EVAL_AVOID_SIDE_EFFECTS. */
|
|
|
|
static struct value *
|
|
create_value (struct gdbarch *gdbarch, struct value *val, enum noside noside,
|
|
int *indices, int n)
|
|
{
|
|
struct type *type = check_typedef (value_type (val));
|
|
struct type *elm_type = TYPE_TARGET_TYPE (type);
|
|
struct value *ret;
|
|
|
|
/* Check if a single component of a vector is requested which means
|
|
the resulting type is a (primitive) scalar type. */
|
|
if (n == 1)
|
|
{
|
|
if (noside == EVAL_AVOID_SIDE_EFFECTS)
|
|
ret = value_zero (elm_type, not_lval);
|
|
else
|
|
ret = value_subscript (val, indices[0]);
|
|
}
|
|
else
|
|
{
|
|
/* Multiple components of the vector are requested which means the
|
|
resulting type is a vector as well. */
|
|
struct type *dst_type =
|
|
lookup_opencl_vector_type (gdbarch, elm_type->code (),
|
|
TYPE_LENGTH (elm_type),
|
|
elm_type->is_unsigned (), n);
|
|
|
|
if (dst_type == NULL)
|
|
dst_type = init_vector_type (elm_type, n);
|
|
|
|
make_cv_type (TYPE_CONST (type), TYPE_VOLATILE (type), dst_type, NULL);
|
|
|
|
if (noside == EVAL_AVOID_SIDE_EFFECTS)
|
|
ret = allocate_value (dst_type);
|
|
else
|
|
{
|
|
/* Check whether to create a lvalue or not. */
|
|
if (VALUE_LVAL (val) != not_lval && !array_has_dups (indices, n))
|
|
{
|
|
struct lval_closure *c = allocate_lval_closure (indices, n, val);
|
|
ret = allocate_computed_value (dst_type, &opencl_value_funcs, c);
|
|
}
|
|
else
|
|
{
|
|
int i;
|
|
|
|
ret = allocate_value (dst_type);
|
|
|
|
/* Copy src val contents into the destination value. */
|
|
for (i = 0; i < n; i++)
|
|
memcpy (value_contents_writeable (ret)
|
|
+ (i * TYPE_LENGTH (elm_type)),
|
|
value_contents (val)
|
|
+ (indices[i] * TYPE_LENGTH (elm_type)),
|
|
TYPE_LENGTH (elm_type));
|
|
}
|
|
}
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
/* OpenCL vector component access. */
|
|
|
|
static struct value *
|
|
opencl_component_ref (struct expression *exp, struct value *val,
|
|
const char *comps, enum noside noside)
|
|
{
|
|
LONGEST lowb, highb;
|
|
int src_len;
|
|
struct value *v;
|
|
int indices[16], i;
|
|
int dst_len;
|
|
|
|
if (!get_array_bounds (check_typedef (value_type (val)), &lowb, &highb))
|
|
error (_("Could not determine the vector bounds"));
|
|
|
|
src_len = highb - lowb + 1;
|
|
|
|
/* Throw an error if the amount of array elements does not fit a
|
|
valid OpenCL vector size (2, 3, 4, 8, 16). */
|
|
if (src_len != 2 && src_len != 3 && src_len != 4 && src_len != 8
|
|
&& src_len != 16)
|
|
error (_("Invalid OpenCL vector size"));
|
|
|
|
if (strcmp (comps, "lo") == 0 )
|
|
{
|
|
dst_len = (src_len == 3) ? 2 : src_len / 2;
|
|
|
|
for (i = 0; i < dst_len; i++)
|
|
indices[i] = i;
|
|
}
|
|
else if (strcmp (comps, "hi") == 0)
|
|
{
|
|
dst_len = (src_len == 3) ? 2 : src_len / 2;
|
|
|
|
for (i = 0; i < dst_len; i++)
|
|
indices[i] = dst_len + i;
|
|
}
|
|
else if (strcmp (comps, "even") == 0)
|
|
{
|
|
dst_len = (src_len == 3) ? 2 : src_len / 2;
|
|
|
|
for (i = 0; i < dst_len; i++)
|
|
indices[i] = i*2;
|
|
}
|
|
else if (strcmp (comps, "odd") == 0)
|
|
{
|
|
dst_len = (src_len == 3) ? 2 : src_len / 2;
|
|
|
|
for (i = 0; i < dst_len; i++)
|
|
indices[i] = i*2+1;
|
|
}
|
|
else if (strncasecmp (comps, "s", 1) == 0)
|
|
{
|
|
#define HEXCHAR_TO_INT(C) ((C >= '0' && C <= '9') ? \
|
|
C-'0' : ((C >= 'A' && C <= 'F') ? \
|
|
C-'A'+10 : ((C >= 'a' && C <= 'f') ? \
|
|
C-'a'+10 : -1)))
|
|
|
|
dst_len = strlen (comps);
|
|
/* Skip the s/S-prefix. */
|
|
dst_len--;
|
|
|
|
for (i = 0; i < dst_len; i++)
|
|
{
|
|
indices[i] = HEXCHAR_TO_INT(comps[i+1]);
|
|
/* Check if the requested component is invalid or exceeds
|
|
the vector. */
|
|
if (indices[i] < 0 || indices[i] >= src_len)
|
|
error (_("Invalid OpenCL vector component accessor %s"), comps);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
dst_len = strlen (comps);
|
|
|
|
for (i = 0; i < dst_len; i++)
|
|
{
|
|
/* x, y, z, w */
|
|
switch (comps[i])
|
|
{
|
|
case 'x':
|
|
indices[i] = 0;
|
|
break;
|
|
case 'y':
|
|
indices[i] = 1;
|
|
break;
|
|
case 'z':
|
|
if (src_len < 3)
|
|
error (_("Invalid OpenCL vector component accessor %s"), comps);
|
|
indices[i] = 2;
|
|
break;
|
|
case 'w':
|
|
if (src_len < 4)
|
|
error (_("Invalid OpenCL vector component accessor %s"), comps);
|
|
indices[i] = 3;
|
|
break;
|
|
default:
|
|
error (_("Invalid OpenCL vector component accessor %s"), comps);
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Throw an error if the amount of requested components does not
|
|
result in a valid length (1, 2, 3, 4, 8, 16). */
|
|
if (dst_len != 1 && dst_len != 2 && dst_len != 3 && dst_len != 4
|
|
&& dst_len != 8 && dst_len != 16)
|
|
error (_("Invalid OpenCL vector component accessor %s"), comps);
|
|
|
|
v = create_value (exp->gdbarch, val, noside, indices, dst_len);
|
|
|
|
return v;
|
|
}
|
|
|
|
/* Perform the unary logical not (!) operation. */
|
|
|
|
static struct value *
|
|
opencl_logical_not (struct expression *exp, struct value *arg)
|
|
{
|
|
struct type *type = check_typedef (value_type (arg));
|
|
struct type *rettype;
|
|
struct value *ret;
|
|
|
|
if (type->code () == TYPE_CODE_ARRAY && type->is_vector ())
|
|
{
|
|
struct type *eltype = check_typedef (TYPE_TARGET_TYPE (type));
|
|
LONGEST lowb, highb;
|
|
int i;
|
|
|
|
if (!get_array_bounds (type, &lowb, &highb))
|
|
error (_("Could not determine the vector bounds"));
|
|
|
|
/* Determine the resulting type of the operation and allocate the
|
|
value. */
|
|
rettype = lookup_opencl_vector_type (exp->gdbarch, TYPE_CODE_INT,
|
|
TYPE_LENGTH (eltype), 0,
|
|
highb - lowb + 1);
|
|
ret = allocate_value (rettype);
|
|
|
|
for (i = 0; i < highb - lowb + 1; i++)
|
|
{
|
|
/* For vector types, the unary operator shall return a 0 if the
|
|
value of its operand compares unequal to 0, and -1 (i.e. all bits
|
|
set) if the value of its operand compares equal to 0. */
|
|
int tmp = value_logical_not (value_subscript (arg, i)) ? -1 : 0;
|
|
memset (value_contents_writeable (ret) + i * TYPE_LENGTH (eltype),
|
|
tmp, TYPE_LENGTH (eltype));
|
|
}
|
|
}
|
|
else
|
|
{
|
|
rettype = language_bool_type (exp->language_defn, exp->gdbarch);
|
|
ret = value_from_longest (rettype, value_logical_not (arg));
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
/* Perform a relational operation on two scalar operands. */
|
|
|
|
static int
|
|
scalar_relop (struct value *val1, struct value *val2, enum exp_opcode op)
|
|
{
|
|
int ret;
|
|
|
|
switch (op)
|
|
{
|
|
case BINOP_EQUAL:
|
|
ret = value_equal (val1, val2);
|
|
break;
|
|
case BINOP_NOTEQUAL:
|
|
ret = !value_equal (val1, val2);
|
|
break;
|
|
case BINOP_LESS:
|
|
ret = value_less (val1, val2);
|
|
break;
|
|
case BINOP_GTR:
|
|
ret = value_less (val2, val1);
|
|
break;
|
|
case BINOP_GEQ:
|
|
ret = value_less (val2, val1) || value_equal (val1, val2);
|
|
break;
|
|
case BINOP_LEQ:
|
|
ret = value_less (val1, val2) || value_equal (val1, val2);
|
|
break;
|
|
case BINOP_LOGICAL_AND:
|
|
ret = !value_logical_not (val1) && !value_logical_not (val2);
|
|
break;
|
|
case BINOP_LOGICAL_OR:
|
|
ret = !value_logical_not (val1) || !value_logical_not (val2);
|
|
break;
|
|
default:
|
|
error (_("Attempt to perform an unsupported operation"));
|
|
break;
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
/* Perform a relational operation on two vector operands. */
|
|
|
|
static struct value *
|
|
vector_relop (struct expression *exp, struct value *val1, struct value *val2,
|
|
enum exp_opcode op)
|
|
{
|
|
struct value *ret;
|
|
struct type *type1, *type2, *eltype1, *eltype2, *rettype;
|
|
int t1_is_vec, t2_is_vec, i;
|
|
LONGEST lowb1, lowb2, highb1, highb2;
|
|
|
|
type1 = check_typedef (value_type (val1));
|
|
type2 = check_typedef (value_type (val2));
|
|
|
|
t1_is_vec = (type1->code () == TYPE_CODE_ARRAY && type1->is_vector ());
|
|
t2_is_vec = (type2->code () == TYPE_CODE_ARRAY && type2->is_vector ());
|
|
|
|
if (!t1_is_vec || !t2_is_vec)
|
|
error (_("Vector operations are not supported on scalar types"));
|
|
|
|
eltype1 = check_typedef (TYPE_TARGET_TYPE (type1));
|
|
eltype2 = check_typedef (TYPE_TARGET_TYPE (type2));
|
|
|
|
if (!get_array_bounds (type1,&lowb1, &highb1)
|
|
|| !get_array_bounds (type2, &lowb2, &highb2))
|
|
error (_("Could not determine the vector bounds"));
|
|
|
|
/* Check whether the vector types are compatible. */
|
|
if (eltype1->code () != eltype2->code ()
|
|
|| TYPE_LENGTH (eltype1) != TYPE_LENGTH (eltype2)
|
|
|| eltype1->is_unsigned () != eltype2->is_unsigned ()
|
|
|| lowb1 != lowb2 || highb1 != highb2)
|
|
error (_("Cannot perform operation on vectors with different types"));
|
|
|
|
/* Determine the resulting type of the operation and allocate the value. */
|
|
rettype = lookup_opencl_vector_type (exp->gdbarch, TYPE_CODE_INT,
|
|
TYPE_LENGTH (eltype1), 0,
|
|
highb1 - lowb1 + 1);
|
|
ret = allocate_value (rettype);
|
|
|
|
for (i = 0; i < highb1 - lowb1 + 1; i++)
|
|
{
|
|
/* For vector types, the relational, equality and logical operators shall
|
|
return 0 if the specified relation is false and -1 (i.e. all bits set)
|
|
if the specified relation is true. */
|
|
int tmp = scalar_relop (value_subscript (val1, i),
|
|
value_subscript (val2, i), op) ? -1 : 0;
|
|
memset (value_contents_writeable (ret) + i * TYPE_LENGTH (eltype1),
|
|
tmp, TYPE_LENGTH (eltype1));
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
|
|
/* Perform a cast of ARG into TYPE. There's sadly a lot of duplication in
|
|
here from valops.c:value_cast, opencl is different only in the
|
|
behaviour of scalar to vector casting. As far as possibly we're going
|
|
to try and delegate back to the standard value_cast function. */
|
|
|
|
static struct value *
|
|
opencl_value_cast (struct type *type, struct value *arg)
|
|
{
|
|
if (type != value_type (arg))
|
|
{
|
|
/* Casting scalar to vector is a special case for OpenCL, scalar
|
|
is cast to element type of vector then replicated into each
|
|
element of the vector. First though, we need to work out if
|
|
this is a scalar to vector cast; code lifted from
|
|
valops.c:value_cast. */
|
|
enum type_code code1, code2;
|
|
struct type *to_type;
|
|
int scalar;
|
|
|
|
to_type = check_typedef (type);
|
|
|
|
code1 = to_type->code ();
|
|
code2 = check_typedef (value_type (arg))->code ();
|
|
|
|
if (code2 == TYPE_CODE_REF)
|
|
code2 = check_typedef (value_type (coerce_ref(arg)))->code ();
|
|
|
|
scalar = (code2 == TYPE_CODE_INT || code2 == TYPE_CODE_BOOL
|
|
|| code2 == TYPE_CODE_CHAR || code2 == TYPE_CODE_FLT
|
|
|| code2 == TYPE_CODE_DECFLOAT || code2 == TYPE_CODE_ENUM
|
|
|| code2 == TYPE_CODE_RANGE);
|
|
|
|
if (code1 == TYPE_CODE_ARRAY && to_type->is_vector () && scalar)
|
|
{
|
|
struct type *eltype;
|
|
|
|
/* Cast to the element type of the vector here as
|
|
value_vector_widen will error if the scalar value is
|
|
truncated by the cast. To avoid the error, cast (and
|
|
possibly truncate) here. */
|
|
eltype = check_typedef (TYPE_TARGET_TYPE (to_type));
|
|
arg = value_cast (eltype, arg);
|
|
|
|
return value_vector_widen (arg, type);
|
|
}
|
|
else
|
|
/* Standard cast handler. */
|
|
arg = value_cast (type, arg);
|
|
}
|
|
return arg;
|
|
}
|
|
|
|
/* Perform a relational operation on two operands. */
|
|
|
|
static struct value *
|
|
opencl_relop (struct expression *exp, struct value *arg1, struct value *arg2,
|
|
enum exp_opcode op)
|
|
{
|
|
struct value *val;
|
|
struct type *type1 = check_typedef (value_type (arg1));
|
|
struct type *type2 = check_typedef (value_type (arg2));
|
|
int t1_is_vec = (type1->code () == TYPE_CODE_ARRAY
|
|
&& type1->is_vector ());
|
|
int t2_is_vec = (type2->code () == TYPE_CODE_ARRAY
|
|
&& type2->is_vector ());
|
|
|
|
if (!t1_is_vec && !t2_is_vec)
|
|
{
|
|
int tmp = scalar_relop (arg1, arg2, op);
|
|
struct type *type =
|
|
language_bool_type (exp->language_defn, exp->gdbarch);
|
|
|
|
val = value_from_longest (type, tmp);
|
|
}
|
|
else if (t1_is_vec && t2_is_vec)
|
|
{
|
|
val = vector_relop (exp, arg1, arg2, op);
|
|
}
|
|
else
|
|
{
|
|
/* Widen the scalar operand to a vector. */
|
|
struct value **v = t1_is_vec ? &arg2 : &arg1;
|
|
struct type *t = t1_is_vec ? type2 : type1;
|
|
|
|
if (t->code () != TYPE_CODE_FLT && !is_integral_type (t))
|
|
error (_("Argument to operation not a number or boolean."));
|
|
|
|
*v = opencl_value_cast (t1_is_vec ? type1 : type2, *v);
|
|
val = vector_relop (exp, arg1, arg2, op);
|
|
}
|
|
|
|
return val;
|
|
}
|
|
|
|
/* A helper function for BINOP_ASSIGN. */
|
|
|
|
static struct value *
|
|
eval_opencl_assign (struct type *expect_type, struct expression *exp,
|
|
enum noside noside,
|
|
struct value *arg1, struct value *arg2)
|
|
{
|
|
if (noside == EVAL_SKIP || noside == EVAL_AVOID_SIDE_EFFECTS)
|
|
return arg1;
|
|
|
|
struct type *type1 = value_type (arg1);
|
|
if (deprecated_value_modifiable (arg1)
|
|
&& VALUE_LVAL (arg1) != lval_internalvar)
|
|
arg2 = opencl_value_cast (type1, arg2);
|
|
|
|
return value_assign (arg1, arg2);
|
|
}
|
|
|
|
/* Expression evaluator for the OpenCL. Most operations are delegated to
|
|
evaluate_subexp_standard; see that function for a description of the
|
|
arguments. */
|
|
|
|
static struct value *
|
|
evaluate_subexp_opencl (struct type *expect_type, struct expression *exp,
|
|
int *pos, enum noside noside)
|
|
{
|
|
enum exp_opcode op = exp->elts[*pos].opcode;
|
|
struct value *arg1 = NULL;
|
|
struct value *arg2 = NULL;
|
|
struct type *type1, *type2;
|
|
|
|
switch (op)
|
|
{
|
|
/* Handle assignment and cast operators to support OpenCL-style
|
|
scalar-to-vector widening. */
|
|
case BINOP_ASSIGN:
|
|
(*pos)++;
|
|
arg1 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
type1 = value_type (arg1);
|
|
arg2 = evaluate_subexp (type1, exp, pos, noside);
|
|
|
|
return eval_opencl_assign (expect_type, exp, noside, arg1, arg2);
|
|
|
|
case UNOP_CAST:
|
|
type1 = exp->elts[*pos + 1].type;
|
|
(*pos) += 2;
|
|
arg1 = evaluate_subexp (type1, exp, pos, noside);
|
|
|
|
if (noside == EVAL_SKIP)
|
|
return value_from_longest (builtin_type (exp->gdbarch)->
|
|
builtin_int, 1);
|
|
|
|
return opencl_value_cast (type1, arg1);
|
|
|
|
case UNOP_CAST_TYPE:
|
|
(*pos)++;
|
|
arg1 = evaluate_subexp (NULL, exp, pos, EVAL_AVOID_SIDE_EFFECTS);
|
|
type1 = value_type (arg1);
|
|
arg1 = evaluate_subexp (type1, exp, pos, noside);
|
|
|
|
if (noside == EVAL_SKIP)
|
|
return value_from_longest (builtin_type (exp->gdbarch)->
|
|
builtin_int, 1);
|
|
|
|
return opencl_value_cast (type1, arg1);
|
|
|
|
/* Handle binary relational and equality operators that are either not
|
|
or differently defined for GNU vectors. */
|
|
case BINOP_EQUAL:
|
|
case BINOP_NOTEQUAL:
|
|
case BINOP_LESS:
|
|
case BINOP_GTR:
|
|
case BINOP_GEQ:
|
|
case BINOP_LEQ:
|
|
(*pos)++;
|
|
arg1 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
arg2 = evaluate_subexp (value_type (arg1), exp, pos, noside);
|
|
|
|
if (noside == EVAL_SKIP)
|
|
return value_from_longest (builtin_type (exp->gdbarch)->
|
|
builtin_int, 1);
|
|
|
|
return opencl_relop (exp, arg1, arg2, op);
|
|
|
|
/* Handle the logical unary operator not(!). */
|
|
case UNOP_LOGICAL_NOT:
|
|
(*pos)++;
|
|
arg1 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
|
|
if (noside == EVAL_SKIP)
|
|
return value_from_longest (builtin_type (exp->gdbarch)->
|
|
builtin_int, 1);
|
|
|
|
return opencl_logical_not (exp, arg1);
|
|
|
|
/* Handle the logical operator and(&&) and or(||). */
|
|
case BINOP_LOGICAL_AND:
|
|
case BINOP_LOGICAL_OR:
|
|
(*pos)++;
|
|
arg1 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
|
|
if (noside == EVAL_SKIP)
|
|
{
|
|
evaluate_subexp (nullptr, exp, pos, noside);
|
|
|
|
return value_from_longest (builtin_type (exp->gdbarch)->
|
|
builtin_int, 1);
|
|
}
|
|
else
|
|
{
|
|
/* For scalar operations we need to avoid evaluating operands
|
|
unnecessarily. However, for vector operations we always need to
|
|
evaluate both operands. Unfortunately we only know which of the
|
|
two cases apply after we know the type of the second operand.
|
|
Therefore we evaluate it once using EVAL_AVOID_SIDE_EFFECTS. */
|
|
int oldpos = *pos;
|
|
|
|
arg2 = evaluate_subexp (nullptr, exp, pos, EVAL_AVOID_SIDE_EFFECTS);
|
|
*pos = oldpos;
|
|
type1 = check_typedef (value_type (arg1));
|
|
type2 = check_typedef (value_type (arg2));
|
|
|
|
if ((type1->code () == TYPE_CODE_ARRAY && type1->is_vector ())
|
|
|| (type2->code () == TYPE_CODE_ARRAY && type2->is_vector ()))
|
|
{
|
|
arg2 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
|
|
return opencl_relop (exp, arg1, arg2, op);
|
|
}
|
|
else
|
|
{
|
|
/* For scalar built-in types, only evaluate the right
|
|
hand operand if the left hand operand compares
|
|
unequal(&&)/equal(||) to 0. */
|
|
int res;
|
|
int tmp = value_logical_not (arg1);
|
|
|
|
if (op == BINOP_LOGICAL_OR)
|
|
tmp = !tmp;
|
|
|
|
arg2
|
|
= evaluate_subexp (nullptr, exp, pos, tmp ? EVAL_SKIP : noside);
|
|
type1 = language_bool_type (exp->language_defn, exp->gdbarch);
|
|
|
|
if (op == BINOP_LOGICAL_AND)
|
|
res = !tmp && !value_logical_not (arg2);
|
|
else /* BINOP_LOGICAL_OR */
|
|
res = tmp || !value_logical_not (arg2);
|
|
|
|
return value_from_longest (type1, res);
|
|
}
|
|
}
|
|
|
|
/* Handle the ternary selection operator. */
|
|
case TERNOP_COND:
|
|
(*pos)++;
|
|
arg1 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
type1 = check_typedef (value_type (arg1));
|
|
if (type1->code () == TYPE_CODE_ARRAY && type1->is_vector ())
|
|
{
|
|
struct value *arg3, *tmp, *ret;
|
|
struct type *eltype2, *type3, *eltype3;
|
|
int t2_is_vec, t3_is_vec, i;
|
|
LONGEST lowb1, lowb2, lowb3, highb1, highb2, highb3;
|
|
|
|
arg2 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
arg3 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
type2 = check_typedef (value_type (arg2));
|
|
type3 = check_typedef (value_type (arg3));
|
|
t2_is_vec
|
|
= type2->code () == TYPE_CODE_ARRAY && type2->is_vector ();
|
|
t3_is_vec
|
|
= type3->code () == TYPE_CODE_ARRAY && type3->is_vector ();
|
|
|
|
/* Widen the scalar operand to a vector if necessary. */
|
|
if (t2_is_vec || !t3_is_vec)
|
|
{
|
|
arg3 = opencl_value_cast (type2, arg3);
|
|
type3 = value_type (arg3);
|
|
}
|
|
else if (!t2_is_vec || t3_is_vec)
|
|
{
|
|
arg2 = opencl_value_cast (type3, arg2);
|
|
type2 = value_type (arg2);
|
|
}
|
|
else if (!t2_is_vec || !t3_is_vec)
|
|
{
|
|
/* Throw an error if arg2 or arg3 aren't vectors. */
|
|
error (_("\
|
|
Cannot perform conditional operation on incompatible types"));
|
|
}
|
|
|
|
eltype2 = check_typedef (TYPE_TARGET_TYPE (type2));
|
|
eltype3 = check_typedef (TYPE_TARGET_TYPE (type3));
|
|
|
|
if (!get_array_bounds (type1, &lowb1, &highb1)
|
|
|| !get_array_bounds (type2, &lowb2, &highb2)
|
|
|| !get_array_bounds (type3, &lowb3, &highb3))
|
|
error (_("Could not determine the vector bounds"));
|
|
|
|
/* Throw an error if the types of arg2 or arg3 are incompatible. */
|
|
if (eltype2->code () != eltype3->code ()
|
|
|| TYPE_LENGTH (eltype2) != TYPE_LENGTH (eltype3)
|
|
|| eltype2->is_unsigned () != eltype3->is_unsigned ()
|
|
|| lowb2 != lowb3 || highb2 != highb3)
|
|
error (_("\
|
|
Cannot perform operation on vectors with different types"));
|
|
|
|
/* Throw an error if the sizes of arg1 and arg2/arg3 differ. */
|
|
if (lowb1 != lowb2 || lowb1 != lowb3
|
|
|| highb1 != highb2 || highb1 != highb3)
|
|
error (_("\
|
|
Cannot perform conditional operation on vectors with different sizes"));
|
|
|
|
ret = allocate_value (type2);
|
|
|
|
for (i = 0; i < highb1 - lowb1 + 1; i++)
|
|
{
|
|
tmp = value_logical_not (value_subscript (arg1, i)) ?
|
|
value_subscript (arg3, i) : value_subscript (arg2, i);
|
|
memcpy (value_contents_writeable (ret) +
|
|
i * TYPE_LENGTH (eltype2), value_contents_all (tmp),
|
|
TYPE_LENGTH (eltype2));
|
|
}
|
|
|
|
return ret;
|
|
}
|
|
else
|
|
{
|
|
if (value_logical_not (arg1))
|
|
{
|
|
/* Skip the second operand. */
|
|
evaluate_subexp (nullptr, exp, pos, EVAL_SKIP);
|
|
|
|
return evaluate_subexp (nullptr, exp, pos, noside);
|
|
}
|
|
else
|
|
{
|
|
/* Skip the third operand. */
|
|
arg2 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
evaluate_subexp (nullptr, exp, pos, EVAL_SKIP);
|
|
|
|
return arg2;
|
|
}
|
|
}
|
|
|
|
/* Handle STRUCTOP_STRUCT to allow component access on OpenCL vectors. */
|
|
case STRUCTOP_STRUCT:
|
|
{
|
|
int pc = (*pos)++;
|
|
int tem = longest_to_int (exp->elts[pc + 1].longconst);
|
|
|
|
(*pos) += 3 + BYTES_TO_EXP_ELEM (tem + 1);
|
|
arg1 = evaluate_subexp (nullptr, exp, pos, noside);
|
|
type1 = check_typedef (value_type (arg1));
|
|
|
|
if (noside == EVAL_SKIP)
|
|
{
|
|
return value_from_longest (builtin_type (exp->gdbarch)->
|
|
builtin_int, 1);
|
|
}
|
|
else if (type1->code () == TYPE_CODE_ARRAY && type1->is_vector ())
|
|
{
|
|
return opencl_component_ref (exp, arg1, &exp->elts[pc + 2].string,
|
|
noside);
|
|
}
|
|
else
|
|
{
|
|
struct value *v = value_struct_elt (&arg1, NULL,
|
|
&exp->elts[pc + 2].string, NULL,
|
|
"structure");
|
|
|
|
if (noside == EVAL_AVOID_SIDE_EFFECTS)
|
|
v = value_zero (value_type (v), VALUE_LVAL (v));
|
|
return v;
|
|
}
|
|
}
|
|
default:
|
|
break;
|
|
}
|
|
|
|
return evaluate_subexp_c (expect_type, exp, pos, noside);
|
|
}
|
|
|
|
const struct exp_descriptor exp_descriptor_opencl =
|
|
{
|
|
print_subexp_standard,
|
|
operator_length_standard,
|
|
operator_check_standard,
|
|
dump_subexp_body_standard,
|
|
evaluate_subexp_opencl
|
|
};
|
|
|
|
/* Class representing the OpenCL language. */
|
|
|
|
class opencl_language : public language_defn
|
|
{
|
|
public:
|
|
opencl_language ()
|
|
: language_defn (language_opencl)
|
|
{ /* Nothing. */ }
|
|
|
|
/* See language.h. */
|
|
|
|
const char *name () const override
|
|
{ return "opencl"; }
|
|
|
|
/* See language.h. */
|
|
|
|
const char *natural_name () const override
|
|
{ return "OpenCL C"; }
|
|
|
|
/* See language.h. */
|
|
void language_arch_info (struct gdbarch *gdbarch,
|
|
struct language_arch_info *lai) const override
|
|
{
|
|
/* Helper function to allow shorter lines below. */
|
|
auto add = [&] (struct type * t) -> struct type *
|
|
{
|
|
lai->add_primitive_type (t);
|
|
return t;
|
|
};
|
|
|
|
/* Helper macro to create strings. */
|
|
#define OCL_STRING(S) #S
|
|
|
|
/* This macro allocates and assigns the type struct pointers
|
|
for the vector types. */
|
|
#define BUILD_OCL_VTYPES(TYPE, ELEMENT_TYPE) \
|
|
do \
|
|
{ \
|
|
struct type *tmp; \
|
|
tmp = add (init_vector_type (ELEMENT_TYPE, 2)); \
|
|
tmp->set_name (OCL_STRING(TYPE ## 2)); \
|
|
tmp = add (init_vector_type (ELEMENT_TYPE, 3)); \
|
|
tmp->set_name (OCL_STRING(TYPE ## 3)); \
|
|
TYPE_LENGTH (tmp) = 4 * TYPE_LENGTH (ELEMENT_TYPE); \
|
|
tmp = add (init_vector_type (ELEMENT_TYPE, 4)); \
|
|
tmp->set_name (OCL_STRING(TYPE ## 4)); \
|
|
tmp = add (init_vector_type (ELEMENT_TYPE, 8)); \
|
|
tmp->set_name (OCL_STRING(TYPE ## 8)); \
|
|
tmp = init_vector_type (ELEMENT_TYPE, 16); \
|
|
tmp->set_name (OCL_STRING(TYPE ## 16)); \
|
|
} \
|
|
while (false)
|
|
|
|
struct type *el_type, *char_type, *int_type;
|
|
|
|
char_type = el_type = add (arch_integer_type (gdbarch, 8, 0, "char"));
|
|
BUILD_OCL_VTYPES (char, el_type);
|
|
el_type = add (arch_integer_type (gdbarch, 8, 1, "uchar"));
|
|
BUILD_OCL_VTYPES (uchar, el_type);
|
|
el_type = add (arch_integer_type (gdbarch, 16, 0, "short"));
|
|
BUILD_OCL_VTYPES (short, el_type);
|
|
el_type = add (arch_integer_type (gdbarch, 16, 1, "ushort"));
|
|
BUILD_OCL_VTYPES (ushort, el_type);
|
|
int_type = el_type = add (arch_integer_type (gdbarch, 32, 0, "int"));
|
|
BUILD_OCL_VTYPES (int, el_type);
|
|
el_type = add (arch_integer_type (gdbarch, 32, 1, "uint"));
|
|
BUILD_OCL_VTYPES (uint, el_type);
|
|
el_type = add (arch_integer_type (gdbarch, 64, 0, "long"));
|
|
BUILD_OCL_VTYPES (long, el_type);
|
|
el_type = add (arch_integer_type (gdbarch, 64, 1, "ulong"));
|
|
BUILD_OCL_VTYPES (ulong, el_type);
|
|
el_type = add (arch_float_type (gdbarch, 16, "half", floatformats_ieee_half));
|
|
BUILD_OCL_VTYPES (half, el_type);
|
|
el_type = add (arch_float_type (gdbarch, 32, "float", floatformats_ieee_single));
|
|
BUILD_OCL_VTYPES (float, el_type);
|
|
el_type = add (arch_float_type (gdbarch, 64, "double", floatformats_ieee_double));
|
|
BUILD_OCL_VTYPES (double, el_type);
|
|
|
|
add (arch_boolean_type (gdbarch, 8, 1, "bool"));
|
|
add (arch_integer_type (gdbarch, 8, 1, "unsigned char"));
|
|
add (arch_integer_type (gdbarch, 16, 1, "unsigned short"));
|
|
add (arch_integer_type (gdbarch, 32, 1, "unsigned int"));
|
|
add (arch_integer_type (gdbarch, 64, 1, "unsigned long"));
|
|
add (arch_integer_type (gdbarch, gdbarch_ptr_bit (gdbarch), 1, "size_t"));
|
|
add (arch_integer_type (gdbarch, gdbarch_ptr_bit (gdbarch), 0, "ptrdiff_t"));
|
|
add (arch_integer_type (gdbarch, gdbarch_ptr_bit (gdbarch), 0, "intptr_t"));
|
|
add (arch_integer_type (gdbarch, gdbarch_ptr_bit (gdbarch), 1, "uintptr_t"));
|
|
add (arch_type (gdbarch, TYPE_CODE_VOID, TARGET_CHAR_BIT, "void"));
|
|
|
|
/* Type of elements of strings. */
|
|
lai->set_string_char_type (char_type);
|
|
|
|
/* Specifies the return type of logical and relational operations. */
|
|
lai->set_bool_type (int_type, "int");
|
|
}
|
|
|
|
/* See language.h. */
|
|
|
|
void print_type (struct type *type, const char *varstring,
|
|
struct ui_file *stream, int show, int level,
|
|
const struct type_print_options *flags) const override
|
|
{
|
|
/* We nearly always defer to C type printing, except that vector types
|
|
are considered primitive in OpenCL, and should always be printed
|
|
using their TYPE_NAME. */
|
|
if (show > 0)
|
|
{
|
|
type = check_typedef (type);
|
|
if (type->code () == TYPE_CODE_ARRAY && type->is_vector ()
|
|
&& type->name () != NULL)
|
|
show = 0;
|
|
}
|
|
|
|
c_print_type (type, varstring, stream, show, level, flags);
|
|
}
|
|
|
|
/* See language.h. */
|
|
|
|
enum macro_expansion macro_expansion () const override
|
|
{ return macro_expansion_c; }
|
|
|
|
/* See language.h. */
|
|
|
|
const struct exp_descriptor *expression_ops () const override
|
|
{ return &exp_descriptor_opencl; }
|
|
|
|
/* See language.h. */
|
|
|
|
const struct op_print *opcode_print_table () const override
|
|
{ return c_op_print_tab; }
|
|
};
|
|
|
|
/* Single instance of the OpenCL language class. */
|
|
|
|
static opencl_language opencl_language_defn;
|