/* Intrinsic translation Copyright (C) 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc. Contributed by Paul Brook and Steven Bosscher This file is part of GCC. GCC 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, or (at your option) any later version. GCC 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 GCC; see the file COPYING3. If not see . */ /* trans-intrinsic.c-- generate GENERIC trees for calls to intrinsics. */ #include "config.h" #include "system.h" #include "coretypes.h" #include "tm.h" /* For UNITS_PER_WORD. */ #include "tree.h" #include "ggc.h" #include "diagnostic-core.h" /* For internal_error. */ #include "toplev.h" /* For rest_of_decl_compilation. */ #include "flags.h" #include "gfortran.h" #include "arith.h" #include "intrinsic.h" #include "trans.h" #include "trans-const.h" #include "trans-types.h" #include "trans-array.h" /* Only for gfc_trans_assign and gfc_trans_pointer_assign. */ #include "trans-stmt.h" /* This maps Fortran intrinsic math functions to external library or GCC builtin functions. */ typedef struct GTY(()) gfc_intrinsic_map_t { /* The explicit enum is required to work around inadequacies in the garbage collection/gengtype parsing mechanism. */ enum gfc_isym_id id; /* Enum value from the "language-independent", aka C-centric, part of gcc, or END_BUILTINS of no such value set. */ enum built_in_function float_built_in; enum built_in_function double_built_in; enum built_in_function long_double_built_in; enum built_in_function complex_float_built_in; enum built_in_function complex_double_built_in; enum built_in_function complex_long_double_built_in; /* True if the naming pattern is to prepend "c" for complex and append "f" for kind=4. False if the naming pattern is to prepend "_gfortran_" and append "[rc](4|8|10|16)". */ bool libm_name; /* True if a complex version of the function exists. */ bool complex_available; /* True if the function should be marked const. */ bool is_constant; /* The base library name of this function. */ const char *name; /* Cache decls created for the various operand types. */ tree real4_decl; tree real8_decl; tree real10_decl; tree real16_decl; tree complex4_decl; tree complex8_decl; tree complex10_decl; tree complex16_decl; } gfc_intrinsic_map_t; /* ??? The NARGS==1 hack here is based on the fact that (c99 at least) defines complex variants of all of the entries in mathbuiltins.def except for atan2. */ #define DEFINE_MATH_BUILTIN(ID, NAME, ARGTYPE) \ { GFC_ISYM_ ## ID, BUILT_IN_ ## ID ## F, BUILT_IN_ ## ID, \ BUILT_IN_ ## ID ## L, END_BUILTINS, END_BUILTINS, END_BUILTINS, \ true, false, true, NAME, NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE, \ NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE}, #define DEFINE_MATH_BUILTIN_C(ID, NAME, ARGTYPE) \ { GFC_ISYM_ ## ID, BUILT_IN_ ## ID ## F, BUILT_IN_ ## ID, \ BUILT_IN_ ## ID ## L, BUILT_IN_C ## ID ## F, BUILT_IN_C ## ID, \ BUILT_IN_C ## ID ## L, true, true, true, NAME, NULL_TREE, NULL_TREE, \ NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE}, #define LIB_FUNCTION(ID, NAME, HAVE_COMPLEX) \ { GFC_ISYM_ ## ID, END_BUILTINS, END_BUILTINS, END_BUILTINS, \ END_BUILTINS, END_BUILTINS, END_BUILTINS, \ false, HAVE_COMPLEX, true, NAME, NULL_TREE, NULL_TREE, NULL_TREE, \ NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE } #define OTHER_BUILTIN(ID, NAME, TYPE, CONST) \ { GFC_ISYM_NONE, BUILT_IN_ ## ID ## F, BUILT_IN_ ## ID, \ BUILT_IN_ ## ID ## L, END_BUILTINS, END_BUILTINS, END_BUILTINS, \ true, false, CONST, NAME, NULL_TREE, NULL_TREE, \ NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE}, static GTY(()) gfc_intrinsic_map_t gfc_intrinsic_map[] = { /* Functions built into gcc itself (DEFINE_MATH_BUILTIN and DEFINE_MATH_BUILTIN_C), then the built-ins that don't correspond to any GFC_ISYM id directly, which use the OTHER_BUILTIN macro. */ #include "mathbuiltins.def" /* Functions in libgfortran. */ LIB_FUNCTION (ERFC_SCALED, "erfc_scaled", false), /* End the list. */ LIB_FUNCTION (NONE, NULL, false) }; #undef OTHER_BUILTIN #undef LIB_FUNCTION #undef DEFINE_MATH_BUILTIN #undef DEFINE_MATH_BUILTIN_C enum rounding_mode { RND_ROUND, RND_TRUNC, RND_CEIL, RND_FLOOR }; /* Find the correct variant of a given builtin from its argument. */ static tree builtin_decl_for_precision (enum built_in_function base_built_in, int precision) { enum built_in_function i = END_BUILTINS; gfc_intrinsic_map_t *m; for (m = gfc_intrinsic_map; m->double_built_in != base_built_in ; m++) ; if (precision == TYPE_PRECISION (float_type_node)) i = m->float_built_in; else if (precision == TYPE_PRECISION (double_type_node)) i = m->double_built_in; else if (precision == TYPE_PRECISION (long_double_type_node)) i = m->long_double_built_in; else if (precision == TYPE_PRECISION (float128_type_node)) { /* Special treatment, because it is not exactly a built-in, but a library function. */ return m->real16_decl; } return (i == END_BUILTINS ? NULL_TREE : builtin_decl_explicit (i)); } tree gfc_builtin_decl_for_float_kind (enum built_in_function double_built_in, int kind) { int i = gfc_validate_kind (BT_REAL, kind, false); if (gfc_real_kinds[i].c_float128) { /* For __float128, the story is a bit different, because we return a decl to a library function rather than a built-in. */ gfc_intrinsic_map_t *m; for (m = gfc_intrinsic_map; m->double_built_in != double_built_in ; m++) ; return m->real16_decl; } return builtin_decl_for_precision (double_built_in, gfc_real_kinds[i].mode_precision); } /* Evaluate the arguments to an intrinsic function. The value of NARGS may be less than the actual number of arguments in EXPR to allow optional "KIND" arguments that are not included in the generated code to be ignored. */ static void gfc_conv_intrinsic_function_args (gfc_se *se, gfc_expr *expr, tree *argarray, int nargs) { gfc_actual_arglist *actual; gfc_expr *e; gfc_intrinsic_arg *formal; gfc_se argse; int curr_arg; formal = expr->value.function.isym->formal; actual = expr->value.function.actual; for (curr_arg = 0; curr_arg < nargs; curr_arg++, actual = actual->next, formal = formal ? formal->next : NULL) { gcc_assert (actual); e = actual->expr; /* Skip omitted optional arguments. */ if (!e) { --curr_arg; continue; } /* Evaluate the parameter. This will substitute scalarized references automatically. */ gfc_init_se (&argse, se); if (e->ts.type == BT_CHARACTER) { gfc_conv_expr (&argse, e); gfc_conv_string_parameter (&argse); argarray[curr_arg++] = argse.string_length; gcc_assert (curr_arg < nargs); } else gfc_conv_expr_val (&argse, e); /* If an optional argument is itself an optional dummy argument, check its presence and substitute a null if absent. */ if (e->expr_type == EXPR_VARIABLE && e->symtree->n.sym->attr.optional && formal && formal->optional) gfc_conv_missing_dummy (&argse, e, formal->ts, 0); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); argarray[curr_arg] = argse.expr; } } /* Count the number of actual arguments to the intrinsic function EXPR including any "hidden" string length arguments. */ static unsigned int gfc_intrinsic_argument_list_length (gfc_expr *expr) { int n = 0; gfc_actual_arglist *actual; for (actual = expr->value.function.actual; actual; actual = actual->next) { if (!actual->expr) continue; if (actual->expr->ts.type == BT_CHARACTER) n += 2; else n++; } return n; } /* Conversions between different types are output by the frontend as intrinsic functions. We implement these directly with inline code. */ static void gfc_conv_intrinsic_conversion (gfc_se * se, gfc_expr * expr) { tree type; tree *args; int nargs; nargs = gfc_intrinsic_argument_list_length (expr); args = XALLOCAVEC (tree, nargs); /* Evaluate all the arguments passed. Whilst we're only interested in the first one here, there are other parts of the front-end that assume this and will trigger an ICE if it's not the case. */ type = gfc_typenode_for_spec (&expr->ts); gcc_assert (expr->value.function.actual->expr); gfc_conv_intrinsic_function_args (se, expr, args, nargs); /* Conversion between character kinds involves a call to a library function. */ if (expr->ts.type == BT_CHARACTER) { tree fndecl, var, addr, tmp; if (expr->ts.kind == 1 && expr->value.function.actual->expr->ts.kind == 4) fndecl = gfor_fndecl_convert_char4_to_char1; else if (expr->ts.kind == 4 && expr->value.function.actual->expr->ts.kind == 1) fndecl = gfor_fndecl_convert_char1_to_char4; else gcc_unreachable (); /* Create the variable storing the converted value. */ type = gfc_get_pchar_type (expr->ts.kind); var = gfc_create_var (type, "str"); addr = gfc_build_addr_expr (build_pointer_type (type), var); /* Call the library function that will perform the conversion. */ gcc_assert (nargs >= 2); tmp = build_call_expr_loc (input_location, fndecl, 3, addr, args[0], args[1]); gfc_add_expr_to_block (&se->pre, tmp); /* Free the temporary afterwards. */ tmp = gfc_call_free (var); gfc_add_expr_to_block (&se->post, tmp); se->expr = var; se->string_length = args[0]; return; } /* Conversion from complex to non-complex involves taking the real component of the value. */ if (TREE_CODE (TREE_TYPE (args[0])) == COMPLEX_TYPE && expr->ts.type != BT_COMPLEX) { tree artype; artype = TREE_TYPE (TREE_TYPE (args[0])); args[0] = fold_build1_loc (input_location, REALPART_EXPR, artype, args[0]); } se->expr = convert (type, args[0]); } /* This is needed because the gcc backend only implements FIX_TRUNC_EXPR, which is the same as INT() in Fortran. FLOOR(x) = INT(x) <= x ? INT(x) : INT(x) - 1 Similarly for CEILING. */ static tree build_fixbound_expr (stmtblock_t * pblock, tree arg, tree type, int up) { tree tmp; tree cond; tree argtype; tree intval; argtype = TREE_TYPE (arg); arg = gfc_evaluate_now (arg, pblock); intval = convert (type, arg); intval = gfc_evaluate_now (intval, pblock); tmp = convert (argtype, intval); cond = fold_build2_loc (input_location, up ? GE_EXPR : LE_EXPR, boolean_type_node, tmp, arg); tmp = fold_build2_loc (input_location, up ? PLUS_EXPR : MINUS_EXPR, type, intval, build_int_cst (type, 1)); tmp = fold_build3_loc (input_location, COND_EXPR, type, cond, intval, tmp); return tmp; } /* Round to nearest integer, away from zero. */ static tree build_round_expr (tree arg, tree restype) { tree argtype; tree fn; int argprec, resprec; argtype = TREE_TYPE (arg); argprec = TYPE_PRECISION (argtype); resprec = TYPE_PRECISION (restype); /* Depending on the type of the result, choose the int intrinsic (iround, available only as a builtin, therefore cannot use it for __float128), long int intrinsic (lround family) or long long intrinsic (llround). We might also need to convert the result afterwards. */ if (resprec <= INT_TYPE_SIZE && argprec <= LONG_DOUBLE_TYPE_SIZE) fn = builtin_decl_for_precision (BUILT_IN_IROUND, argprec); else if (resprec <= LONG_TYPE_SIZE) fn = builtin_decl_for_precision (BUILT_IN_LROUND, argprec); else if (resprec <= LONG_LONG_TYPE_SIZE) fn = builtin_decl_for_precision (BUILT_IN_LLROUND, argprec); else gcc_unreachable (); return fold_convert (restype, build_call_expr_loc (input_location, fn, 1, arg)); } /* Convert a real to an integer using a specific rounding mode. Ideally we would just build the corresponding GENERIC node, however the RTL expander only actually supports FIX_TRUNC_EXPR. */ static tree build_fix_expr (stmtblock_t * pblock, tree arg, tree type, enum rounding_mode op) { switch (op) { case RND_FLOOR: return build_fixbound_expr (pblock, arg, type, 0); break; case RND_CEIL: return build_fixbound_expr (pblock, arg, type, 1); break; case RND_ROUND: return build_round_expr (arg, type); break; case RND_TRUNC: return fold_build1_loc (input_location, FIX_TRUNC_EXPR, type, arg); break; default: gcc_unreachable (); } } /* Round a real value using the specified rounding mode. We use a temporary integer of that same kind size as the result. Values larger than those that can be represented by this kind are unchanged, as they will not be accurate enough to represent the rounding. huge = HUGE (KIND (a)) aint (a) = ((a > huge) || (a < -huge)) ? a : (real)(int)a */ static void gfc_conv_intrinsic_aint (gfc_se * se, gfc_expr * expr, enum rounding_mode op) { tree type; tree itype; tree arg[2]; tree tmp; tree cond; tree decl; mpfr_t huge; int n, nargs; int kind; kind = expr->ts.kind; nargs = gfc_intrinsic_argument_list_length (expr); decl = NULL_TREE; /* We have builtin functions for some cases. */ switch (op) { case RND_ROUND: decl = gfc_builtin_decl_for_float_kind (BUILT_IN_ROUND, kind); break; case RND_TRUNC: decl = gfc_builtin_decl_for_float_kind (BUILT_IN_TRUNC, kind); break; default: gcc_unreachable (); } /* Evaluate the argument. */ gcc_assert (expr->value.function.actual->expr); gfc_conv_intrinsic_function_args (se, expr, arg, nargs); /* Use a builtin function if one exists. */ if (decl != NULL_TREE) { se->expr = build_call_expr_loc (input_location, decl, 1, arg[0]); return; } /* This code is probably redundant, but we'll keep it lying around just in case. */ type = gfc_typenode_for_spec (&expr->ts); arg[0] = gfc_evaluate_now (arg[0], &se->pre); /* Test if the value is too large to handle sensibly. */ gfc_set_model_kind (kind); mpfr_init (huge); n = gfc_validate_kind (BT_INTEGER, kind, false); mpfr_set_z (huge, gfc_integer_kinds[n].huge, GFC_RND_MODE); tmp = gfc_conv_mpfr_to_tree (huge, kind, 0); cond = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, arg[0], tmp); mpfr_neg (huge, huge, GFC_RND_MODE); tmp = gfc_conv_mpfr_to_tree (huge, kind, 0); tmp = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, arg[0], tmp); cond = fold_build2_loc (input_location, TRUTH_AND_EXPR, boolean_type_node, cond, tmp); itype = gfc_get_int_type (kind); tmp = build_fix_expr (&se->pre, arg[0], itype, op); tmp = convert (type, tmp); se->expr = fold_build3_loc (input_location, COND_EXPR, type, cond, tmp, arg[0]); mpfr_clear (huge); } /* Convert to an integer using the specified rounding mode. */ static void gfc_conv_intrinsic_int (gfc_se * se, gfc_expr * expr, enum rounding_mode op) { tree type; tree *args; int nargs; nargs = gfc_intrinsic_argument_list_length (expr); args = XALLOCAVEC (tree, nargs); /* Evaluate the argument, we process all arguments even though we only use the first one for code generation purposes. */ type = gfc_typenode_for_spec (&expr->ts); gcc_assert (expr->value.function.actual->expr); gfc_conv_intrinsic_function_args (se, expr, args, nargs); if (TREE_CODE (TREE_TYPE (args[0])) == INTEGER_TYPE) { /* Conversion to a different integer kind. */ se->expr = convert (type, args[0]); } else { /* Conversion from complex to non-complex involves taking the real component of the value. */ if (TREE_CODE (TREE_TYPE (args[0])) == COMPLEX_TYPE && expr->ts.type != BT_COMPLEX) { tree artype; artype = TREE_TYPE (TREE_TYPE (args[0])); args[0] = fold_build1_loc (input_location, REALPART_EXPR, artype, args[0]); } se->expr = build_fix_expr (&se->pre, args[0], type, op); } } /* Get the imaginary component of a value. */ static void gfc_conv_intrinsic_imagpart (gfc_se * se, gfc_expr * expr) { tree arg; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); se->expr = fold_build1_loc (input_location, IMAGPART_EXPR, TREE_TYPE (TREE_TYPE (arg)), arg); } /* Get the complex conjugate of a value. */ static void gfc_conv_intrinsic_conjg (gfc_se * se, gfc_expr * expr) { tree arg; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); se->expr = fold_build1_loc (input_location, CONJ_EXPR, TREE_TYPE (arg), arg); } static tree define_quad_builtin (const char *name, tree type, bool is_const) { tree fndecl; fndecl = build_decl (input_location, FUNCTION_DECL, get_identifier (name), type); /* Mark the decl as external. */ DECL_EXTERNAL (fndecl) = 1; TREE_PUBLIC (fndecl) = 1; /* Mark it __attribute__((const)). */ TREE_READONLY (fndecl) = is_const; rest_of_decl_compilation (fndecl, 1, 0); return fndecl; } /* Initialize function decls for library functions. The external functions are created as required. Builtin functions are added here. */ void gfc_build_intrinsic_lib_fndecls (void) { gfc_intrinsic_map_t *m; tree quad_decls[END_BUILTINS + 1]; if (gfc_real16_is_float128) { /* If we have soft-float types, we create the decls for their C99-like library functions. For now, we only handle __float128 q-suffixed functions. */ tree type, complex_type, func_1, func_2, func_cabs, func_frexp; tree func_iround, func_lround, func_llround, func_scalbn, func_cpow; memset (quad_decls, 0, sizeof(tree) * (END_BUILTINS + 1)); type = float128_type_node; complex_type = complex_float128_type_node; /* type (*) (type) */ func_1 = build_function_type_list (type, type, NULL_TREE); /* int (*) (type) */ func_iround = build_function_type_list (integer_type_node, type, NULL_TREE); /* long (*) (type) */ func_lround = build_function_type_list (long_integer_type_node, type, NULL_TREE); /* long long (*) (type) */ func_llround = build_function_type_list (long_long_integer_type_node, type, NULL_TREE); /* type (*) (type, type) */ func_2 = build_function_type_list (type, type, type, NULL_TREE); /* type (*) (type, &int) */ func_frexp = build_function_type_list (type, type, build_pointer_type (integer_type_node), NULL_TREE); /* type (*) (type, int) */ func_scalbn = build_function_type_list (type, type, integer_type_node, NULL_TREE); /* type (*) (complex type) */ func_cabs = build_function_type_list (type, complex_type, NULL_TREE); /* complex type (*) (complex type, complex type) */ func_cpow = build_function_type_list (complex_type, complex_type, complex_type, NULL_TREE); #define DEFINE_MATH_BUILTIN(ID, NAME, ARGTYPE) #define DEFINE_MATH_BUILTIN_C(ID, NAME, ARGTYPE) #define LIB_FUNCTION(ID, NAME, HAVE_COMPLEX) /* Only these built-ins are actually needed here. These are used directly from the code, when calling builtin_decl_for_precision() or builtin_decl_for_float_type(). The others are all constructed by gfc_get_intrinsic_lib_fndecl(). */ #define OTHER_BUILTIN(ID, NAME, TYPE, CONST) \ quad_decls[BUILT_IN_ ## ID] = define_quad_builtin (NAME "q", func_ ## TYPE, CONST); #include "mathbuiltins.def" #undef OTHER_BUILTIN #undef LIB_FUNCTION #undef DEFINE_MATH_BUILTIN #undef DEFINE_MATH_BUILTIN_C } /* Add GCC builtin functions. */ for (m = gfc_intrinsic_map; m->id != GFC_ISYM_NONE || m->double_built_in != END_BUILTINS; m++) { if (m->float_built_in != END_BUILTINS) m->real4_decl = builtin_decl_explicit (m->float_built_in); if (m->complex_float_built_in != END_BUILTINS) m->complex4_decl = builtin_decl_explicit (m->complex_float_built_in); if (m->double_built_in != END_BUILTINS) m->real8_decl = builtin_decl_explicit (m->double_built_in); if (m->complex_double_built_in != END_BUILTINS) m->complex8_decl = builtin_decl_explicit (m->complex_double_built_in); /* If real(kind=10) exists, it is always long double. */ if (m->long_double_built_in != END_BUILTINS) m->real10_decl = builtin_decl_explicit (m->long_double_built_in); if (m->complex_long_double_built_in != END_BUILTINS) m->complex10_decl = builtin_decl_explicit (m->complex_long_double_built_in); if (!gfc_real16_is_float128) { if (m->long_double_built_in != END_BUILTINS) m->real16_decl = builtin_decl_explicit (m->long_double_built_in); if (m->complex_long_double_built_in != END_BUILTINS) m->complex16_decl = builtin_decl_explicit (m->complex_long_double_built_in); } else if (quad_decls[m->double_built_in] != NULL_TREE) { /* Quad-precision function calls are constructed when first needed by builtin_decl_for_precision(), except for those that will be used directly (define by OTHER_BUILTIN). */ m->real16_decl = quad_decls[m->double_built_in]; } else if (quad_decls[m->complex_double_built_in] != NULL_TREE) { /* Same thing for the complex ones. */ m->complex16_decl = quad_decls[m->double_built_in]; } } } /* Create a fndecl for a simple intrinsic library function. */ static tree gfc_get_intrinsic_lib_fndecl (gfc_intrinsic_map_t * m, gfc_expr * expr) { tree type; VEC(tree,gc) *argtypes; tree fndecl; gfc_actual_arglist *actual; tree *pdecl; gfc_typespec *ts; char name[GFC_MAX_SYMBOL_LEN + 3]; ts = &expr->ts; if (ts->type == BT_REAL) { switch (ts->kind) { case 4: pdecl = &m->real4_decl; break; case 8: pdecl = &m->real8_decl; break; case 10: pdecl = &m->real10_decl; break; case 16: pdecl = &m->real16_decl; break; default: gcc_unreachable (); } } else if (ts->type == BT_COMPLEX) { gcc_assert (m->complex_available); switch (ts->kind) { case 4: pdecl = &m->complex4_decl; break; case 8: pdecl = &m->complex8_decl; break; case 10: pdecl = &m->complex10_decl; break; case 16: pdecl = &m->complex16_decl; break; default: gcc_unreachable (); } } else gcc_unreachable (); if (*pdecl) return *pdecl; if (m->libm_name) { int n = gfc_validate_kind (BT_REAL, ts->kind, false); if (gfc_real_kinds[n].c_float) snprintf (name, sizeof (name), "%s%s%s", ts->type == BT_COMPLEX ? "c" : "", m->name, "f"); else if (gfc_real_kinds[n].c_double) snprintf (name, sizeof (name), "%s%s", ts->type == BT_COMPLEX ? "c" : "", m->name); else if (gfc_real_kinds[n].c_long_double) snprintf (name, sizeof (name), "%s%s%s", ts->type == BT_COMPLEX ? "c" : "", m->name, "l"); else if (gfc_real_kinds[n].c_float128) snprintf (name, sizeof (name), "%s%s%s", ts->type == BT_COMPLEX ? "c" : "", m->name, "q"); else gcc_unreachable (); } else { snprintf (name, sizeof (name), PREFIX ("%s_%c%d"), m->name, ts->type == BT_COMPLEX ? 'c' : 'r', ts->kind); } argtypes = NULL; for (actual = expr->value.function.actual; actual; actual = actual->next) { type = gfc_typenode_for_spec (&actual->expr->ts); VEC_safe_push (tree, gc, argtypes, type); } type = build_function_type_vec (gfc_typenode_for_spec (ts), argtypes); fndecl = build_decl (input_location, FUNCTION_DECL, get_identifier (name), type); /* Mark the decl as external. */ DECL_EXTERNAL (fndecl) = 1; TREE_PUBLIC (fndecl) = 1; /* Mark it __attribute__((const)), if possible. */ TREE_READONLY (fndecl) = m->is_constant; rest_of_decl_compilation (fndecl, 1, 0); (*pdecl) = fndecl; return fndecl; } /* Convert an intrinsic function into an external or builtin call. */ static void gfc_conv_intrinsic_lib_function (gfc_se * se, gfc_expr * expr) { gfc_intrinsic_map_t *m; tree fndecl; tree rettype; tree *args; unsigned int num_args; gfc_isym_id id; id = expr->value.function.isym->id; /* Find the entry for this function. */ for (m = gfc_intrinsic_map; m->id != GFC_ISYM_NONE || m->double_built_in != END_BUILTINS; m++) { if (id == m->id) break; } if (m->id == GFC_ISYM_NONE) { internal_error ("Intrinsic function %s(%d) not recognized", expr->value.function.name, id); } /* Get the decl and generate the call. */ num_args = gfc_intrinsic_argument_list_length (expr); args = XALLOCAVEC (tree, num_args); gfc_conv_intrinsic_function_args (se, expr, args, num_args); fndecl = gfc_get_intrinsic_lib_fndecl (m, expr); rettype = TREE_TYPE (TREE_TYPE (fndecl)); fndecl = build_addr (fndecl, current_function_decl); se->expr = build_call_array_loc (input_location, rettype, fndecl, num_args, args); } /* If bounds-checking is enabled, create code to verify at runtime that the string lengths for both expressions are the same (needed for e.g. MERGE). If bounds-checking is not enabled, does nothing. */ void gfc_trans_same_strlen_check (const char* intr_name, locus* where, tree a, tree b, stmtblock_t* target) { tree cond; tree name; /* If bounds-checking is disabled, do nothing. */ if (!(gfc_option.rtcheck & GFC_RTCHECK_BOUNDS)) return; /* Compare the two string lengths. */ cond = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, a, b); /* Output the runtime-check. */ name = gfc_build_cstring_const (intr_name); name = gfc_build_addr_expr (pchar_type_node, name); gfc_trans_runtime_check (true, false, cond, target, where, "Unequal character lengths (%ld/%ld) in %s", fold_convert (long_integer_type_node, a), fold_convert (long_integer_type_node, b), name); } /* The EXPONENT(s) intrinsic function is translated into int ret; frexp (s, &ret); return ret; */ static void gfc_conv_intrinsic_exponent (gfc_se *se, gfc_expr *expr) { tree arg, type, res, tmp, frexp; frexp = gfc_builtin_decl_for_float_kind (BUILT_IN_FREXP, expr->value.function.actual->expr->ts.kind); gfc_conv_intrinsic_function_args (se, expr, &arg, 1); res = gfc_create_var (integer_type_node, NULL); tmp = build_call_expr_loc (input_location, frexp, 2, arg, gfc_build_addr_expr (NULL_TREE, res)); gfc_add_expr_to_block (&se->pre, tmp); type = gfc_typenode_for_spec (&expr->ts); se->expr = fold_convert (type, res); } static void trans_this_image (gfc_se * se, gfc_expr *expr) { stmtblock_t loop; tree type, desc, dim_arg, cond, tmp, m, loop_var, exit_label, min_var, lbound, ubound, extent, ml; gfc_se argse; int rank, corank; /* The case -fcoarray=single is handled elsewhere. */ gcc_assert (gfc_option.coarray != GFC_FCOARRAY_SINGLE); gfc_init_coarray_decl (false); /* Argument-free version: THIS_IMAGE(). */ if (expr->value.function.actual->expr == NULL) { se->expr = fold_convert (gfc_get_int_type (gfc_default_integer_kind), gfort_gvar_caf_this_image); return; } /* Coarray-argument version: THIS_IMAGE(coarray [, dim]). */ type = gfc_get_int_type (gfc_default_integer_kind); corank = gfc_get_corank (expr->value.function.actual->expr); rank = expr->value.function.actual->expr->rank; /* Obtain the descriptor of the COARRAY. */ gfc_init_se (&argse, NULL); argse.want_coarray = 1; gfc_conv_expr_descriptor (&argse, expr->value.function.actual->expr); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); desc = argse.expr; if (se->ss) { /* Create an implicit second parameter from the loop variable. */ gcc_assert (!expr->value.function.actual->next->expr); gcc_assert (corank > 0); gcc_assert (se->loop->dimen == 1); gcc_assert (se->ss->info->expr == expr); dim_arg = se->loop->loopvar[0]; dim_arg = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, dim_arg, build_int_cst (TREE_TYPE (dim_arg), 1)); gfc_advance_se_ss_chain (se); } else { /* Use the passed DIM= argument. */ gcc_assert (expr->value.function.actual->next->expr); gfc_init_se (&argse, NULL); gfc_conv_expr_type (&argse, expr->value.function.actual->next->expr, gfc_array_index_type); gfc_add_block_to_block (&se->pre, &argse.pre); dim_arg = argse.expr; if (INTEGER_CST_P (dim_arg)) { int hi, co_dim; hi = TREE_INT_CST_HIGH (dim_arg); co_dim = TREE_INT_CST_LOW (dim_arg); if (hi || co_dim < 1 || co_dim > GFC_TYPE_ARRAY_CORANK (TREE_TYPE (desc))) gfc_error ("'dim' argument of %s intrinsic at %L is not a valid " "dimension index", expr->value.function.isym->name, &expr->where); } else if (gfc_option.rtcheck & GFC_RTCHECK_BOUNDS) { dim_arg = gfc_evaluate_now (dim_arg, &se->pre); cond = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, dim_arg, build_int_cst (TREE_TYPE (dim_arg), 1)); tmp = gfc_rank_cst[GFC_TYPE_ARRAY_CORANK (TREE_TYPE (desc))]; tmp = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, dim_arg, tmp); cond = fold_build2_loc (input_location, TRUTH_ORIF_EXPR, boolean_type_node, cond, tmp); gfc_trans_runtime_check (true, false, cond, &se->pre, &expr->where, gfc_msg_fault); } } /* Used algorithm; cf. Fortran 2008, C.10. Note, due to the scalarizer, one always has a dim_arg argument. m = this_image() - 1 if (corank == 1) { sub(1) = m + lcobound(corank) return; } i = rank min_var = min (rank + corank - 2, rank + dim_arg - 1) for (;;) { extent = gfc_extent(i) ml = m m = m/extent if (i >= min_var) goto exit_label i++ } exit_label: sub(dim_arg) = (dim_arg < corank) ? ml - m*extent + lcobound(dim_arg) : m + lcobound(corank) */ /* this_image () - 1. */ tmp = fold_convert (type, gfort_gvar_caf_this_image); tmp = fold_build2_loc (input_location, MINUS_EXPR, type, tmp, build_int_cst (type, 1)); if (corank == 1) { /* sub(1) = m + lcobound(corank). */ lbound = gfc_conv_descriptor_lbound_get (desc, build_int_cst (TREE_TYPE (gfc_array_index_type), corank+rank-1)); lbound = fold_convert (type, lbound); tmp = fold_build2_loc (input_location, PLUS_EXPR, type, tmp, lbound); se->expr = tmp; return; } m = gfc_create_var (type, NULL); ml = gfc_create_var (type, NULL); loop_var = gfc_create_var (integer_type_node, NULL); min_var = gfc_create_var (integer_type_node, NULL); /* m = this_image () - 1. */ gfc_add_modify (&se->pre, m, tmp); /* min_var = min (rank + corank-2, rank + dim_arg - 1). */ tmp = fold_build2_loc (input_location, PLUS_EXPR, integer_type_node, fold_convert (integer_type_node, dim_arg), build_int_cst (integer_type_node, rank - 1)); tmp = fold_build2_loc (input_location, MIN_EXPR, integer_type_node, build_int_cst (integer_type_node, rank + corank - 2), tmp); gfc_add_modify (&se->pre, min_var, tmp); /* i = rank. */ tmp = build_int_cst (integer_type_node, rank); gfc_add_modify (&se->pre, loop_var, tmp); exit_label = gfc_build_label_decl (NULL_TREE); TREE_USED (exit_label) = 1; /* Loop body. */ gfc_init_block (&loop); /* ml = m. */ gfc_add_modify (&loop, ml, m); /* extent = ... */ lbound = gfc_conv_descriptor_lbound_get (desc, loop_var); ubound = gfc_conv_descriptor_ubound_get (desc, loop_var); extent = gfc_conv_array_extent_dim (lbound, ubound, NULL); extent = fold_convert (type, extent); /* m = m/extent. */ gfc_add_modify (&loop, m, fold_build2_loc (input_location, TRUNC_DIV_EXPR, type, m, extent)); /* Exit condition: if (i >= min_var) goto exit_label. */ cond = fold_build2_loc (input_location, GE_EXPR, boolean_type_node, loop_var, min_var); tmp = build1_v (GOTO_EXPR, exit_label); tmp = fold_build3_loc (input_location, COND_EXPR, void_type_node, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&loop, tmp); /* Increment loop variable: i++. */ gfc_add_modify (&loop, loop_var, fold_build2_loc (input_location, PLUS_EXPR, integer_type_node, loop_var, build_int_cst (integer_type_node, 1))); /* Making the loop... actually loop! */ tmp = gfc_finish_block (&loop); tmp = build1_v (LOOP_EXPR, tmp); gfc_add_expr_to_block (&se->pre, tmp); /* The exit label. */ tmp = build1_v (LABEL_EXPR, exit_label); gfc_add_expr_to_block (&se->pre, tmp); /* sub(co_dim) = (co_dim < corank) ? ml - m*extent + lcobound(dim_arg) : m + lcobound(corank) */ cond = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, dim_arg, build_int_cst (TREE_TYPE (dim_arg), corank)); lbound = gfc_conv_descriptor_lbound_get (desc, fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, dim_arg, build_int_cst (TREE_TYPE (dim_arg), rank-1))); lbound = fold_convert (type, lbound); tmp = fold_build2_loc (input_location, MINUS_EXPR, type, ml, fold_build2_loc (input_location, MULT_EXPR, type, m, extent)); tmp = fold_build2_loc (input_location, PLUS_EXPR, type, tmp, lbound); se->expr = fold_build3_loc (input_location, COND_EXPR, type, cond, tmp, fold_build2_loc (input_location, PLUS_EXPR, type, m, lbound)); } static void trans_image_index (gfc_se * se, gfc_expr *expr) { tree num_images, cond, coindex, type, lbound, ubound, desc, subdesc, tmp, invalid_bound; gfc_se argse, subse; int rank, corank, codim; type = gfc_get_int_type (gfc_default_integer_kind); corank = gfc_get_corank (expr->value.function.actual->expr); rank = expr->value.function.actual->expr->rank; /* Obtain the descriptor of the COARRAY. */ gfc_init_se (&argse, NULL); argse.want_coarray = 1; gfc_conv_expr_descriptor (&argse, expr->value.function.actual->expr); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); desc = argse.expr; /* Obtain a handle to the SUB argument. */ gfc_init_se (&subse, NULL); gfc_conv_expr_descriptor (&subse, expr->value.function.actual->next->expr); gfc_add_block_to_block (&se->pre, &subse.pre); gfc_add_block_to_block (&se->post, &subse.post); subdesc = build_fold_indirect_ref_loc (input_location, gfc_conv_descriptor_data_get (subse.expr)); /* Fortran 2008 does not require that the values remain in the cobounds, thus we need explicitly check this - and return 0 if they are exceeded. */ lbound = gfc_conv_descriptor_lbound_get (desc, gfc_rank_cst[rank+corank-1]); tmp = gfc_build_array_ref (subdesc, gfc_rank_cst[corank-1], NULL); invalid_bound = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, fold_convert (gfc_array_index_type, tmp), lbound); for (codim = corank + rank - 2; codim >= rank; codim--) { lbound = gfc_conv_descriptor_lbound_get (desc, gfc_rank_cst[codim]); ubound = gfc_conv_descriptor_ubound_get (desc, gfc_rank_cst[codim]); tmp = gfc_build_array_ref (subdesc, gfc_rank_cst[codim-rank], NULL); cond = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, fold_convert (gfc_array_index_type, tmp), lbound); invalid_bound = fold_build2_loc (input_location, TRUTH_OR_EXPR, boolean_type_node, invalid_bound, cond); cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, fold_convert (gfc_array_index_type, tmp), ubound); invalid_bound = fold_build2_loc (input_location, TRUTH_OR_EXPR, boolean_type_node, invalid_bound, cond); } invalid_bound = gfc_unlikely (invalid_bound); /* See Fortran 2008, C.10 for the following algorithm. */ /* coindex = sub(corank) - lcobound(n). */ coindex = fold_convert (gfc_array_index_type, gfc_build_array_ref (subdesc, gfc_rank_cst[corank-1], NULL)); lbound = gfc_conv_descriptor_lbound_get (desc, gfc_rank_cst[rank+corank-1]); coindex = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, fold_convert (gfc_array_index_type, coindex), lbound); for (codim = corank + rank - 2; codim >= rank; codim--) { tree extent, ubound; /* coindex = coindex*extent(codim) + sub(codim) - lcobound(codim). */ lbound = gfc_conv_descriptor_lbound_get (desc, gfc_rank_cst[codim]); ubound = gfc_conv_descriptor_ubound_get (desc, gfc_rank_cst[codim]); extent = gfc_conv_array_extent_dim (lbound, ubound, NULL); /* coindex *= extent. */ coindex = fold_build2_loc (input_location, MULT_EXPR, gfc_array_index_type, coindex, extent); /* coindex += sub(codim). */ tmp = gfc_build_array_ref (subdesc, gfc_rank_cst[codim-rank], NULL); coindex = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, coindex, fold_convert (gfc_array_index_type, tmp)); /* coindex -= lbound(codim). */ lbound = gfc_conv_descriptor_lbound_get (desc, gfc_rank_cst[codim]); coindex = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, coindex, lbound); } coindex = fold_build2_loc (input_location, PLUS_EXPR, type, fold_convert(type, coindex), build_int_cst (type, 1)); /* Return 0 if "coindex" exceeds num_images(). */ if (gfc_option.coarray == GFC_FCOARRAY_SINGLE) num_images = build_int_cst (type, 1); else { gfc_init_coarray_decl (false); num_images = fold_convert (type, gfort_gvar_caf_num_images); } tmp = gfc_create_var (type, NULL); gfc_add_modify (&se->pre, tmp, coindex); cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, tmp, num_images); cond = fold_build2_loc (input_location, TRUTH_OR_EXPR, boolean_type_node, cond, fold_convert (boolean_type_node, invalid_bound)); se->expr = fold_build3_loc (input_location, COND_EXPR, type, cond, build_int_cst (type, 0), tmp); } static void trans_num_images (gfc_se * se) { gfc_init_coarray_decl (false); se->expr = fold_convert (gfc_get_int_type (gfc_default_integer_kind), gfort_gvar_caf_num_images); } static void gfc_conv_intrinsic_rank (gfc_se *se, gfc_expr *expr) { gfc_se argse; gfc_init_se (&argse, NULL); argse.data_not_needed = 1; argse.descriptor_only = 1; gfc_conv_expr_descriptor (&argse, expr->value.function.actual->expr); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); se->expr = gfc_conv_descriptor_rank (argse.expr); } /* Evaluate a single upper or lower bound. */ /* TODO: bound intrinsic generates way too much unnecessary code. */ static void gfc_conv_intrinsic_bound (gfc_se * se, gfc_expr * expr, int upper) { gfc_actual_arglist *arg; gfc_actual_arglist *arg2; tree desc; tree type; tree bound; tree tmp; tree cond, cond1, cond3, cond4, size; tree ubound; tree lbound; gfc_se argse; gfc_array_spec * as; bool assumed_rank_lb_one; arg = expr->value.function.actual; arg2 = arg->next; if (se->ss) { /* Create an implicit second parameter from the loop variable. */ gcc_assert (!arg2->expr); gcc_assert (se->loop->dimen == 1); gcc_assert (se->ss->info->expr == expr); gfc_advance_se_ss_chain (se); bound = se->loop->loopvar[0]; bound = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, bound, se->loop->from[0]); } else { /* use the passed argument. */ gcc_assert (arg2->expr); gfc_init_se (&argse, NULL); gfc_conv_expr_type (&argse, arg2->expr, gfc_array_index_type); gfc_add_block_to_block (&se->pre, &argse.pre); bound = argse.expr; /* Convert from one based to zero based. */ bound = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, bound, gfc_index_one_node); } /* TODO: don't re-evaluate the descriptor on each iteration. */ /* Get a descriptor for the first parameter. */ gfc_init_se (&argse, NULL); gfc_conv_expr_descriptor (&argse, arg->expr); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); desc = argse.expr; as = gfc_get_full_arrayspec_from_expr (arg->expr); if (INTEGER_CST_P (bound)) { int hi, low; hi = TREE_INT_CST_HIGH (bound); low = TREE_INT_CST_LOW (bound); if (hi || low < 0 || ((!as || as->type != AS_ASSUMED_RANK) && low >= GFC_TYPE_ARRAY_RANK (TREE_TYPE (desc))) || low > GFC_MAX_DIMENSIONS) gfc_error ("'dim' argument of %s intrinsic at %L is not a valid " "dimension index", upper ? "UBOUND" : "LBOUND", &expr->where); } if (!INTEGER_CST_P (bound) || (as && as->type == AS_ASSUMED_RANK)) { if (gfc_option.rtcheck & GFC_RTCHECK_BOUNDS) { bound = gfc_evaluate_now (bound, &se->pre); cond = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, bound, build_int_cst (TREE_TYPE (bound), 0)); if (as && as->type == AS_ASSUMED_RANK) tmp = gfc_conv_descriptor_rank (desc); else tmp = gfc_rank_cst[GFC_TYPE_ARRAY_RANK (TREE_TYPE (desc))]; tmp = fold_build2_loc (input_location, GE_EXPR, boolean_type_node, bound, fold_convert(TREE_TYPE (bound), tmp)); cond = fold_build2_loc (input_location, TRUTH_ORIF_EXPR, boolean_type_node, cond, tmp); gfc_trans_runtime_check (true, false, cond, &se->pre, &expr->where, gfc_msg_fault); } } /* Take care of the lbound shift for assumed-rank arrays, which are nonallocatable and nonpointers. Those has a lbound of 1. */ assumed_rank_lb_one = as && as->type == AS_ASSUMED_RANK && ((arg->expr->ts.type != BT_CLASS && !arg->expr->symtree->n.sym->attr.allocatable && !arg->expr->symtree->n.sym->attr.pointer) || (arg->expr->ts.type == BT_CLASS && !CLASS_DATA (arg->expr)->attr.allocatable && !CLASS_DATA (arg->expr)->attr.class_pointer)); ubound = gfc_conv_descriptor_ubound_get (desc, bound); lbound = gfc_conv_descriptor_lbound_get (desc, bound); /* 13.14.53: Result value for LBOUND Case (i): For an array section or for an array expression other than a whole array or array structure component, LBOUND(ARRAY, DIM) has the value 1. For a whole array or array structure component, LBOUND(ARRAY, DIM) has the value: (a) equal to the lower bound for subscript DIM of ARRAY if dimension DIM of ARRAY does not have extent zero or if ARRAY is an assumed-size array of rank DIM, or (b) 1 otherwise. 13.14.113: Result value for UBOUND Case (i): For an array section or for an array expression other than a whole array or array structure component, UBOUND(ARRAY, DIM) has the value equal to the number of elements in the given dimension; otherwise, it has a value equal to the upper bound for subscript DIM of ARRAY if dimension DIM of ARRAY does not have size zero and has value zero if dimension DIM has size zero. */ if (!upper && assumed_rank_lb_one) se->expr = gfc_index_one_node; else if (as) { tree stride = gfc_conv_descriptor_stride_get (desc, bound); cond1 = fold_build2_loc (input_location, GE_EXPR, boolean_type_node, ubound, lbound); cond3 = fold_build2_loc (input_location, GE_EXPR, boolean_type_node, stride, gfc_index_zero_node); cond3 = fold_build2_loc (input_location, TRUTH_AND_EXPR, boolean_type_node, cond3, cond1); cond4 = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, stride, gfc_index_zero_node); if (upper) { tree cond5; cond = fold_build2_loc (input_location, TRUTH_OR_EXPR, boolean_type_node, cond3, cond4); cond5 = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, gfc_index_one_node, lbound); cond5 = fold_build2_loc (input_location, TRUTH_AND_EXPR, boolean_type_node, cond4, cond5); cond = fold_build2_loc (input_location, TRUTH_OR_EXPR, boolean_type_node, cond, cond5); if (assumed_rank_lb_one) { tmp = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, ubound, lbound); tmp = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, tmp, gfc_index_one_node); } else tmp = ubound; se->expr = fold_build3_loc (input_location, COND_EXPR, gfc_array_index_type, cond, tmp, gfc_index_zero_node); } else { if (as->type == AS_ASSUMED_SIZE) cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, bound, build_int_cst (TREE_TYPE (bound), arg->expr->rank - 1)); else cond = boolean_false_node; cond1 = fold_build2_loc (input_location, TRUTH_OR_EXPR, boolean_type_node, cond3, cond4); cond = fold_build2_loc (input_location, TRUTH_OR_EXPR, boolean_type_node, cond, cond1); se->expr = fold_build3_loc (input_location, COND_EXPR, gfc_array_index_type, cond, lbound, gfc_index_one_node); } } else { if (upper) { size = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, ubound, lbound); se->expr = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, size, gfc_index_one_node); se->expr = fold_build2_loc (input_location, MAX_EXPR, gfc_array_index_type, se->expr, gfc_index_zero_node); } else se->expr = gfc_index_one_node; } type = gfc_typenode_for_spec (&expr->ts); se->expr = convert (type, se->expr); } static void conv_intrinsic_cobound (gfc_se * se, gfc_expr * expr) { gfc_actual_arglist *arg; gfc_actual_arglist *arg2; gfc_se argse; tree bound, resbound, resbound2, desc, cond, tmp; tree type; int corank; gcc_assert (expr->value.function.isym->id == GFC_ISYM_LCOBOUND || expr->value.function.isym->id == GFC_ISYM_UCOBOUND || expr->value.function.isym->id == GFC_ISYM_THIS_IMAGE); arg = expr->value.function.actual; arg2 = arg->next; gcc_assert (arg->expr->expr_type == EXPR_VARIABLE); corank = gfc_get_corank (arg->expr); gfc_init_se (&argse, NULL); argse.want_coarray = 1; gfc_conv_expr_descriptor (&argse, arg->expr); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); desc = argse.expr; if (se->ss) { /* Create an implicit second parameter from the loop variable. */ gcc_assert (!arg2->expr); gcc_assert (corank > 0); gcc_assert (se->loop->dimen == 1); gcc_assert (se->ss->info->expr == expr); bound = se->loop->loopvar[0]; bound = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, bound, gfc_rank_cst[arg->expr->rank]); gfc_advance_se_ss_chain (se); } else { /* use the passed argument. */ gcc_assert (arg2->expr); gfc_init_se (&argse, NULL); gfc_conv_expr_type (&argse, arg2->expr, gfc_array_index_type); gfc_add_block_to_block (&se->pre, &argse.pre); bound = argse.expr; if (INTEGER_CST_P (bound)) { int hi, low; hi = TREE_INT_CST_HIGH (bound); low = TREE_INT_CST_LOW (bound); if (hi || low < 1 || low > GFC_TYPE_ARRAY_CORANK (TREE_TYPE (desc))) gfc_error ("'dim' argument of %s intrinsic at %L is not a valid " "dimension index", expr->value.function.isym->name, &expr->where); } else if (gfc_option.rtcheck & GFC_RTCHECK_BOUNDS) { bound = gfc_evaluate_now (bound, &se->pre); cond = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, bound, build_int_cst (TREE_TYPE (bound), 1)); tmp = gfc_rank_cst[GFC_TYPE_ARRAY_CORANK (TREE_TYPE (desc))]; tmp = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, bound, tmp); cond = fold_build2_loc (input_location, TRUTH_ORIF_EXPR, boolean_type_node, cond, tmp); gfc_trans_runtime_check (true, false, cond, &se->pre, &expr->where, gfc_msg_fault); } /* Subtract 1 to get to zero based and add dimensions. */ switch (arg->expr->rank) { case 0: bound = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, bound, gfc_index_one_node); case 1: break; default: bound = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, bound, gfc_rank_cst[arg->expr->rank - 1]); } } resbound = gfc_conv_descriptor_lbound_get (desc, bound); /* Handle UCOBOUND with special handling of the last codimension. */ if (expr->value.function.isym->id == GFC_ISYM_UCOBOUND) { /* Last codimension: For -fcoarray=single just return the lcobound - otherwise add ceiling (real (num_images ()) / real (size)) - 1 = (num_images () + size - 1) / size - 1 = (num_images - 1) / size(), where size is the product of the extent of all but the last codimension. */ if (gfc_option.coarray != GFC_FCOARRAY_SINGLE && corank > 1) { tree cosize; gfc_init_coarray_decl (false); cosize = gfc_conv_descriptor_cosize (desc, arg->expr->rank, corank); tmp = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, fold_convert (gfc_array_index_type, gfort_gvar_caf_num_images), build_int_cst (gfc_array_index_type, 1)); tmp = fold_build2_loc (input_location, TRUNC_DIV_EXPR, gfc_array_index_type, tmp, fold_convert (gfc_array_index_type, cosize)); resbound = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, resbound, tmp); } else if (gfc_option.coarray != GFC_FCOARRAY_SINGLE) { /* ubound = lbound + num_images() - 1. */ gfc_init_coarray_decl (false); tmp = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, fold_convert (gfc_array_index_type, gfort_gvar_caf_num_images), build_int_cst (gfc_array_index_type, 1)); resbound = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, resbound, tmp); } if (corank > 1) { cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, bound, build_int_cst (TREE_TYPE (bound), arg->expr->rank + corank - 1)); resbound2 = gfc_conv_descriptor_ubound_get (desc, bound); se->expr = fold_build3_loc (input_location, COND_EXPR, gfc_array_index_type, cond, resbound, resbound2); } else se->expr = resbound; } else se->expr = resbound; type = gfc_typenode_for_spec (&expr->ts); se->expr = convert (type, se->expr); } static void gfc_conv_intrinsic_abs (gfc_se * se, gfc_expr * expr) { tree arg, cabs; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); switch (expr->value.function.actual->expr->ts.type) { case BT_INTEGER: case BT_REAL: se->expr = fold_build1_loc (input_location, ABS_EXPR, TREE_TYPE (arg), arg); break; case BT_COMPLEX: cabs = gfc_builtin_decl_for_float_kind (BUILT_IN_CABS, expr->ts.kind); se->expr = build_call_expr_loc (input_location, cabs, 1, arg); break; default: gcc_unreachable (); } } /* Create a complex value from one or two real components. */ static void gfc_conv_intrinsic_cmplx (gfc_se * se, gfc_expr * expr, int both) { tree real; tree imag; tree type; tree *args; unsigned int num_args; num_args = gfc_intrinsic_argument_list_length (expr); args = XALLOCAVEC (tree, num_args); type = gfc_typenode_for_spec (&expr->ts); gfc_conv_intrinsic_function_args (se, expr, args, num_args); real = convert (TREE_TYPE (type), args[0]); if (both) imag = convert (TREE_TYPE (type), args[1]); else if (TREE_CODE (TREE_TYPE (args[0])) == COMPLEX_TYPE) { imag = fold_build1_loc (input_location, IMAGPART_EXPR, TREE_TYPE (TREE_TYPE (args[0])), args[0]); imag = convert (TREE_TYPE (type), imag); } else imag = build_real_from_int_cst (TREE_TYPE (type), integer_zero_node); se->expr = fold_build2_loc (input_location, COMPLEX_EXPR, type, real, imag); } /* Remainder function MOD(A, P) = A - INT(A / P) * P MODULO(A, P) = A - FLOOR (A / P) * P The obvious algorithms above are numerically instable for large arguments, hence these intrinsics are instead implemented via calls to the fmod family of functions. It is the responsibility of the user to ensure that the second argument is non-zero. */ static void gfc_conv_intrinsic_mod (gfc_se * se, gfc_expr * expr, int modulo) { tree type; tree tmp; tree test; tree test2; tree fmod; tree zero; tree args[2]; gfc_conv_intrinsic_function_args (se, expr, args, 2); switch (expr->ts.type) { case BT_INTEGER: /* Integer case is easy, we've got a builtin op. */ type = TREE_TYPE (args[0]); if (modulo) se->expr = fold_build2_loc (input_location, FLOOR_MOD_EXPR, type, args[0], args[1]); else se->expr = fold_build2_loc (input_location, TRUNC_MOD_EXPR, type, args[0], args[1]); break; case BT_REAL: fmod = NULL_TREE; /* Check if we have a builtin fmod. */ fmod = gfc_builtin_decl_for_float_kind (BUILT_IN_FMOD, expr->ts.kind); /* The builtin should always be available. */ gcc_assert (fmod != NULL_TREE); tmp = build_addr (fmod, current_function_decl); se->expr = build_call_array_loc (input_location, TREE_TYPE (TREE_TYPE (fmod)), tmp, 2, args); if (modulo == 0) return; type = TREE_TYPE (args[0]); args[0] = gfc_evaluate_now (args[0], &se->pre); args[1] = gfc_evaluate_now (args[1], &se->pre); /* Definition: modulo = arg - floor (arg/arg2) * arg2 In order to calculate the result accurately, we use the fmod function as follows. res = fmod (arg, arg2); if (res) { if ((arg < 0) xor (arg2 < 0)) res += arg2; } else res = copysign (0., arg2); => As two nested ternary exprs: res = res ? (((arg < 0) xor (arg2 < 0)) ? res + arg2 : res) : copysign (0., arg2); */ zero = gfc_build_const (type, integer_zero_node); tmp = gfc_evaluate_now (se->expr, &se->pre); if (!flag_signed_zeros) { test = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, args[0], zero); test2 = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, args[1], zero); test2 = fold_build2_loc (input_location, TRUTH_XOR_EXPR, boolean_type_node, test, test2); test = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, tmp, zero); test = fold_build2_loc (input_location, TRUTH_AND_EXPR, boolean_type_node, test, test2); test = gfc_evaluate_now (test, &se->pre); se->expr = fold_build3_loc (input_location, COND_EXPR, type, test, fold_build2_loc (input_location, PLUS_EXPR, type, tmp, args[1]), tmp); } else { tree expr1, copysign, cscall; copysign = gfc_builtin_decl_for_float_kind (BUILT_IN_COPYSIGN, expr->ts.kind); test = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, args[0], zero); test2 = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, args[1], zero); test2 = fold_build2_loc (input_location, TRUTH_XOR_EXPR, boolean_type_node, test, test2); expr1 = fold_build3_loc (input_location, COND_EXPR, type, test2, fold_build2_loc (input_location, PLUS_EXPR, type, tmp, args[1]), tmp); test = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, tmp, zero); cscall = build_call_expr_loc (input_location, copysign, 2, zero, args[1]); se->expr = fold_build3_loc (input_location, COND_EXPR, type, test, expr1, cscall); } return; default: gcc_unreachable (); } } /* DSHIFTL(I,J,S) = (I << S) | (J >> (BITSIZE(J) - S)) DSHIFTR(I,J,S) = (I << (BITSIZE(I) - S)) | (J >> S) where the right shifts are logical (i.e. 0's are shifted in). Because SHIFT_EXPR's want shifts strictly smaller than the integral type width, we have to special-case both S == 0 and S == BITSIZE(J): DSHIFTL(I,J,0) = I DSHIFTL(I,J,BITSIZE) = J DSHIFTR(I,J,0) = J DSHIFTR(I,J,BITSIZE) = I. */ static void gfc_conv_intrinsic_dshift (gfc_se * se, gfc_expr * expr, bool dshiftl) { tree type, utype, stype, arg1, arg2, shift, res, left, right; tree args[3], cond, tmp; int bitsize; gfc_conv_intrinsic_function_args (se, expr, args, 3); gcc_assert (TREE_TYPE (args[0]) == TREE_TYPE (args[1])); type = TREE_TYPE (args[0]); bitsize = TYPE_PRECISION (type); utype = unsigned_type_for (type); stype = TREE_TYPE (args[2]); arg1 = gfc_evaluate_now (args[0], &se->pre); arg2 = gfc_evaluate_now (args[1], &se->pre); shift = gfc_evaluate_now (args[2], &se->pre); /* The generic case. */ tmp = fold_build2_loc (input_location, MINUS_EXPR, stype, build_int_cst (stype, bitsize), shift); left = fold_build2_loc (input_location, LSHIFT_EXPR, type, arg1, dshiftl ? shift : tmp); right = fold_build2_loc (input_location, RSHIFT_EXPR, utype, fold_convert (utype, arg2), dshiftl ? tmp : shift); right = fold_convert (type, right); res = fold_build2_loc (input_location, BIT_IOR_EXPR, type, left, right); /* Special cases. */ cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, shift, build_int_cst (stype, 0)); res = fold_build3_loc (input_location, COND_EXPR, type, cond, dshiftl ? arg1 : arg2, res); cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, shift, build_int_cst (stype, bitsize)); res = fold_build3_loc (input_location, COND_EXPR, type, cond, dshiftl ? arg2 : arg1, res); se->expr = res; } /* Positive difference DIM (x, y) = ((x - y) < 0) ? 0 : x - y. */ static void gfc_conv_intrinsic_dim (gfc_se * se, gfc_expr * expr) { tree val; tree tmp; tree type; tree zero; tree args[2]; gfc_conv_intrinsic_function_args (se, expr, args, 2); type = TREE_TYPE (args[0]); val = fold_build2_loc (input_location, MINUS_EXPR, type, args[0], args[1]); val = gfc_evaluate_now (val, &se->pre); zero = gfc_build_const (type, integer_zero_node); tmp = fold_build2_loc (input_location, LE_EXPR, boolean_type_node, val, zero); se->expr = fold_build3_loc (input_location, COND_EXPR, type, tmp, zero, val); } /* SIGN(A, B) is absolute value of A times sign of B. The real value versions use library functions to ensure the correct handling of negative zero. Integer case implemented as: SIGN(A, B) = { tmp = (A ^ B) >> C; (A + tmp) ^ tmp } */ static void gfc_conv_intrinsic_sign (gfc_se * se, gfc_expr * expr) { tree tmp; tree type; tree args[2]; gfc_conv_intrinsic_function_args (se, expr, args, 2); if (expr->ts.type == BT_REAL) { tree abs; tmp = gfc_builtin_decl_for_float_kind (BUILT_IN_COPYSIGN, expr->ts.kind); abs = gfc_builtin_decl_for_float_kind (BUILT_IN_FABS, expr->ts.kind); /* We explicitly have to ignore the minus sign. We do so by using result = (arg1 == 0) ? abs(arg0) : copysign(arg0, arg1). */ if (!gfc_option.flag_sign_zero && MODE_HAS_SIGNED_ZEROS (TYPE_MODE (TREE_TYPE (args[1])))) { tree cond, zero; zero = build_real_from_int_cst (TREE_TYPE (args[1]), integer_zero_node); cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, args[1], zero); se->expr = fold_build3_loc (input_location, COND_EXPR, TREE_TYPE (args[0]), cond, build_call_expr_loc (input_location, abs, 1, args[0]), build_call_expr_loc (input_location, tmp, 2, args[0], args[1])); } else se->expr = build_call_expr_loc (input_location, tmp, 2, args[0], args[1]); return; } /* Having excluded floating point types, we know we are now dealing with signed integer types. */ type = TREE_TYPE (args[0]); /* Args[0] is used multiple times below. */ args[0] = gfc_evaluate_now (args[0], &se->pre); /* Construct (A ^ B) >> 31, which generates a bit mask of all zeros if the signs of A and B are the same, and of all ones if they differ. */ tmp = fold_build2_loc (input_location, BIT_XOR_EXPR, type, args[0], args[1]); tmp = fold_build2_loc (input_location, RSHIFT_EXPR, type, tmp, build_int_cst (type, TYPE_PRECISION (type) - 1)); tmp = gfc_evaluate_now (tmp, &se->pre); /* Construct (A + tmp) ^ tmp, which is A if tmp is zero, and -A if tmp] is all ones (i.e. -1). */ se->expr = fold_build2_loc (input_location, BIT_XOR_EXPR, type, fold_build2_loc (input_location, PLUS_EXPR, type, args[0], tmp), tmp); } /* Test for the presence of an optional argument. */ static void gfc_conv_intrinsic_present (gfc_se * se, gfc_expr * expr) { gfc_expr *arg; arg = expr->value.function.actual->expr; gcc_assert (arg->expr_type == EXPR_VARIABLE); se->expr = gfc_conv_expr_present (arg->symtree->n.sym); se->expr = convert (gfc_typenode_for_spec (&expr->ts), se->expr); } /* Calculate the double precision product of two single precision values. */ static void gfc_conv_intrinsic_dprod (gfc_se * se, gfc_expr * expr) { tree type; tree args[2]; gfc_conv_intrinsic_function_args (se, expr, args, 2); /* Convert the args to double precision before multiplying. */ type = gfc_typenode_for_spec (&expr->ts); args[0] = convert (type, args[0]); args[1] = convert (type, args[1]); se->expr = fold_build2_loc (input_location, MULT_EXPR, type, args[0], args[1]); } /* Return a length one character string containing an ascii character. */ static void gfc_conv_intrinsic_char (gfc_se * se, gfc_expr * expr) { tree arg[2]; tree var; tree type; unsigned int num_args; num_args = gfc_intrinsic_argument_list_length (expr); gfc_conv_intrinsic_function_args (se, expr, arg, num_args); type = gfc_get_char_type (expr->ts.kind); var = gfc_create_var (type, "char"); arg[0] = fold_build1_loc (input_location, NOP_EXPR, type, arg[0]); gfc_add_modify (&se->pre, var, arg[0]); se->expr = gfc_build_addr_expr (build_pointer_type (type), var); se->string_length = build_int_cst (gfc_charlen_type_node, 1); } static void gfc_conv_intrinsic_ctime (gfc_se * se, gfc_expr * expr) { tree var; tree len; tree tmp; tree cond; tree fndecl; tree *args; unsigned int num_args; num_args = gfc_intrinsic_argument_list_length (expr) + 2; args = XALLOCAVEC (tree, num_args); var = gfc_create_var (pchar_type_node, "pstr"); len = gfc_create_var (gfc_charlen_type_node, "len"); gfc_conv_intrinsic_function_args (se, expr, &args[2], num_args - 2); args[0] = gfc_build_addr_expr (NULL_TREE, var); args[1] = gfc_build_addr_expr (NULL_TREE, len); fndecl = build_addr (gfor_fndecl_ctime, current_function_decl); tmp = build_call_array_loc (input_location, TREE_TYPE (TREE_TYPE (gfor_fndecl_ctime)), fndecl, num_args, args); gfc_add_expr_to_block (&se->pre, tmp); /* Free the temporary afterwards, if necessary. */ cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, len, build_int_cst (TREE_TYPE (len), 0)); tmp = gfc_call_free (var); tmp = build3_v (COND_EXPR, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&se->post, tmp); se->expr = var; se->string_length = len; } static void gfc_conv_intrinsic_fdate (gfc_se * se, gfc_expr * expr) { tree var; tree len; tree tmp; tree cond; tree fndecl; tree *args; unsigned int num_args; num_args = gfc_intrinsic_argument_list_length (expr) + 2; args = XALLOCAVEC (tree, num_args); var = gfc_create_var (pchar_type_node, "pstr"); len = gfc_create_var (gfc_charlen_type_node, "len"); gfc_conv_intrinsic_function_args (se, expr, &args[2], num_args - 2); args[0] = gfc_build_addr_expr (NULL_TREE, var); args[1] = gfc_build_addr_expr (NULL_TREE, len); fndecl = build_addr (gfor_fndecl_fdate, current_function_decl); tmp = build_call_array_loc (input_location, TREE_TYPE (TREE_TYPE (gfor_fndecl_fdate)), fndecl, num_args, args); gfc_add_expr_to_block (&se->pre, tmp); /* Free the temporary afterwards, if necessary. */ cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, len, build_int_cst (TREE_TYPE (len), 0)); tmp = gfc_call_free (var); tmp = build3_v (COND_EXPR, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&se->post, tmp); se->expr = var; se->string_length = len; } /* Return a character string containing the tty name. */ static void gfc_conv_intrinsic_ttynam (gfc_se * se, gfc_expr * expr) { tree var; tree len; tree tmp; tree cond; tree fndecl; tree *args; unsigned int num_args; num_args = gfc_intrinsic_argument_list_length (expr) + 2; args = XALLOCAVEC (tree, num_args); var = gfc_create_var (pchar_type_node, "pstr"); len = gfc_create_var (gfc_charlen_type_node, "len"); gfc_conv_intrinsic_function_args (se, expr, &args[2], num_args - 2); args[0] = gfc_build_addr_expr (NULL_TREE, var); args[1] = gfc_build_addr_expr (NULL_TREE, len); fndecl = build_addr (gfor_fndecl_ttynam, current_function_decl); tmp = build_call_array_loc (input_location, TREE_TYPE (TREE_TYPE (gfor_fndecl_ttynam)), fndecl, num_args, args); gfc_add_expr_to_block (&se->pre, tmp); /* Free the temporary afterwards, if necessary. */ cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, len, build_int_cst (TREE_TYPE (len), 0)); tmp = gfc_call_free (var); tmp = build3_v (COND_EXPR, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&se->post, tmp); se->expr = var; se->string_length = len; } /* Get the minimum/maximum value of all the parameters. minmax (a1, a2, a3, ...) { mvar = a1; if (a2 .op. mvar || isnan(mvar)) mvar = a2; if (a3 .op. mvar || isnan(mvar)) mvar = a3; ... return mvar } */ /* TODO: Mismatching types can occur when specific names are used. These should be handled during resolution. */ static void gfc_conv_intrinsic_minmax (gfc_se * se, gfc_expr * expr, enum tree_code op) { tree tmp; tree mvar; tree val; tree thencase; tree *args; tree type; gfc_actual_arglist *argexpr; unsigned int i, nargs; nargs = gfc_intrinsic_argument_list_length (expr); args = XALLOCAVEC (tree, nargs); gfc_conv_intrinsic_function_args (se, expr, args, nargs); type = gfc_typenode_for_spec (&expr->ts); argexpr = expr->value.function.actual; if (TREE_TYPE (args[0]) != type) args[0] = convert (type, args[0]); /* Only evaluate the argument once. */ if (TREE_CODE (args[0]) != VAR_DECL && !TREE_CONSTANT (args[0])) args[0] = gfc_evaluate_now (args[0], &se->pre); mvar = gfc_create_var (type, "M"); gfc_add_modify (&se->pre, mvar, args[0]); for (i = 1, argexpr = argexpr->next; i < nargs; i++) { tree cond, isnan; val = args[i]; /* Handle absent optional arguments by ignoring the comparison. */ if (argexpr->expr->expr_type == EXPR_VARIABLE && argexpr->expr->symtree->n.sym->attr.optional && TREE_CODE (val) == INDIRECT_REF) cond = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, TREE_OPERAND (val, 0), build_int_cst (TREE_TYPE (TREE_OPERAND (val, 0)), 0)); else { cond = NULL_TREE; /* Only evaluate the argument once. */ if (TREE_CODE (val) != VAR_DECL && !TREE_CONSTANT (val)) val = gfc_evaluate_now (val, &se->pre); } thencase = build2_v (MODIFY_EXPR, mvar, convert (type, val)); tmp = fold_build2_loc (input_location, op, boolean_type_node, convert (type, val), mvar); /* FIXME: When the IEEE_ARITHMETIC module is implemented, the call to __builtin_isnan might be made dependent on that module being loaded, to help performance of programs that don't rely on IEEE semantics. */ if (FLOAT_TYPE_P (TREE_TYPE (mvar))) { isnan = build_call_expr_loc (input_location, builtin_decl_explicit (BUILT_IN_ISNAN), 1, mvar); tmp = fold_build2_loc (input_location, TRUTH_OR_EXPR, boolean_type_node, tmp, fold_convert (boolean_type_node, isnan)); } tmp = build3_v (COND_EXPR, tmp, thencase, build_empty_stmt (input_location)); if (cond != NULL_TREE) tmp = build3_v (COND_EXPR, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&se->pre, tmp); argexpr = argexpr->next; } se->expr = mvar; } /* Generate library calls for MIN and MAX intrinsics for character variables. */ static void gfc_conv_intrinsic_minmax_char (gfc_se * se, gfc_expr * expr, int op) { tree *args; tree var, len, fndecl, tmp, cond, function; unsigned int nargs; nargs = gfc_intrinsic_argument_list_length (expr); args = XALLOCAVEC (tree, nargs + 4); gfc_conv_intrinsic_function_args (se, expr, &args[4], nargs); /* Create the result variables. */ len = gfc_create_var (gfc_charlen_type_node, "len"); args[0] = gfc_build_addr_expr (NULL_TREE, len); var = gfc_create_var (gfc_get_pchar_type (expr->ts.kind), "pstr"); args[1] = gfc_build_addr_expr (ppvoid_type_node, var); args[2] = build_int_cst (integer_type_node, op); args[3] = build_int_cst (integer_type_node, nargs / 2); if (expr->ts.kind == 1) function = gfor_fndecl_string_minmax; else if (expr->ts.kind == 4) function = gfor_fndecl_string_minmax_char4; else gcc_unreachable (); /* Make the function call. */ fndecl = build_addr (function, current_function_decl); tmp = build_call_array_loc (input_location, TREE_TYPE (TREE_TYPE (function)), fndecl, nargs + 4, args); gfc_add_expr_to_block (&se->pre, tmp); /* Free the temporary afterwards, if necessary. */ cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, len, build_int_cst (TREE_TYPE (len), 0)); tmp = gfc_call_free (var); tmp = build3_v (COND_EXPR, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&se->post, tmp); se->expr = var; se->string_length = len; } /* Create a symbol node for this intrinsic. The symbol from the frontend has the generic name. */ static gfc_symbol * gfc_get_symbol_for_expr (gfc_expr * expr) { gfc_symbol *sym; /* TODO: Add symbols for intrinsic function to the global namespace. */ gcc_assert (strlen (expr->value.function.name) <= GFC_MAX_SYMBOL_LEN - 5); sym = gfc_new_symbol (expr->value.function.name, NULL); sym->ts = expr->ts; sym->attr.external = 1; sym->attr.function = 1; sym->attr.always_explicit = 1; sym->attr.proc = PROC_INTRINSIC; sym->attr.flavor = FL_PROCEDURE; sym->result = sym; if (expr->rank > 0) { sym->attr.dimension = 1; sym->as = gfc_get_array_spec (); sym->as->type = AS_ASSUMED_SHAPE; sym->as->rank = expr->rank; } gfc_copy_formal_args_intr (sym, expr->value.function.isym); return sym; } /* Generate a call to an external intrinsic function. */ static void gfc_conv_intrinsic_funcall (gfc_se * se, gfc_expr * expr) { gfc_symbol *sym; VEC(tree,gc) *append_args; gcc_assert (!se->ss || se->ss->info->expr == expr); if (se->ss) gcc_assert (expr->rank > 0); else gcc_assert (expr->rank == 0); sym = gfc_get_symbol_for_expr (expr); /* Calls to libgfortran_matmul need to be appended special arguments, to be able to call the BLAS ?gemm functions if required and possible. */ append_args = NULL; if (expr->value.function.isym->id == GFC_ISYM_MATMUL && sym->ts.type != BT_LOGICAL) { tree cint = gfc_get_int_type (gfc_c_int_kind); if (gfc_option.flag_external_blas && (sym->ts.type == BT_REAL || sym->ts.type == BT_COMPLEX) && (sym->ts.kind == 4 || sym->ts.kind == 8)) { tree gemm_fndecl; if (sym->ts.type == BT_REAL) { if (sym->ts.kind == 4) gemm_fndecl = gfor_fndecl_sgemm; else gemm_fndecl = gfor_fndecl_dgemm; } else { if (sym->ts.kind == 4) gemm_fndecl = gfor_fndecl_cgemm; else gemm_fndecl = gfor_fndecl_zgemm; } append_args = VEC_alloc (tree, gc, 3); VEC_quick_push (tree, append_args, build_int_cst (cint, 1)); VEC_quick_push (tree, append_args, build_int_cst (cint, gfc_option.blas_matmul_limit)); VEC_quick_push (tree, append_args, gfc_build_addr_expr (NULL_TREE, gemm_fndecl)); } else { append_args = VEC_alloc (tree, gc, 3); VEC_quick_push (tree, append_args, build_int_cst (cint, 0)); VEC_quick_push (tree, append_args, build_int_cst (cint, 0)); VEC_quick_push (tree, append_args, null_pointer_node); } } gfc_conv_procedure_call (se, sym, expr->value.function.actual, expr, append_args); gfc_free_symbol (sym); } /* ANY and ALL intrinsics. ANY->op == NE_EXPR, ALL->op == EQ_EXPR. Implemented as any(a) { forall (i=...) if (a[i] != 0) return 1 end forall return 0 } all(a) { forall (i=...) if (a[i] == 0) return 0 end forall return 1 } */ static void gfc_conv_intrinsic_anyall (gfc_se * se, gfc_expr * expr, enum tree_code op) { tree resvar; stmtblock_t block; stmtblock_t body; tree type; tree tmp; tree found; gfc_loopinfo loop; gfc_actual_arglist *actual; gfc_ss *arrayss; gfc_se arrayse; tree exit_label; if (se->ss) { gfc_conv_intrinsic_funcall (se, expr); return; } actual = expr->value.function.actual; type = gfc_typenode_for_spec (&expr->ts); /* Initialize the result. */ resvar = gfc_create_var (type, "test"); if (op == EQ_EXPR) tmp = convert (type, boolean_true_node); else tmp = convert (type, boolean_false_node); gfc_add_modify (&se->pre, resvar, tmp); /* Walk the arguments. */ arrayss = gfc_walk_expr (actual->expr); gcc_assert (arrayss != gfc_ss_terminator); /* Initialize the scalarizer. */ gfc_init_loopinfo (&loop); exit_label = gfc_build_label_decl (NULL_TREE); TREE_USED (exit_label) = 1; gfc_add_ss_to_loop (&loop, arrayss); /* Initialize the loop. */ gfc_conv_ss_startstride (&loop); gfc_conv_loop_setup (&loop, &expr->where); gfc_mark_ss_chain_used (arrayss, 1); /* Generate the loop body. */ gfc_start_scalarized_body (&loop, &body); /* If the condition matches then set the return value. */ gfc_start_block (&block); if (op == EQ_EXPR) tmp = convert (type, boolean_false_node); else tmp = convert (type, boolean_true_node); gfc_add_modify (&block, resvar, tmp); /* And break out of the loop. */ tmp = build1_v (GOTO_EXPR, exit_label); gfc_add_expr_to_block (&block, tmp); found = gfc_finish_block (&block); /* Check this element. */ gfc_init_se (&arrayse, NULL); gfc_copy_loopinfo_to_se (&arrayse, &loop); arrayse.ss = arrayss; gfc_conv_expr_val (&arrayse, actual->expr); gfc_add_block_to_block (&body, &arrayse.pre); tmp = fold_build2_loc (input_location, op, boolean_type_node, arrayse.expr, build_int_cst (TREE_TYPE (arrayse.expr), 0)); tmp = build3_v (COND_EXPR, tmp, found, build_empty_stmt (input_location)); gfc_add_expr_to_block (&body, tmp); gfc_add_block_to_block (&body, &arrayse.post); gfc_trans_scalarizing_loops (&loop, &body); /* Add the exit label. */ tmp = build1_v (LABEL_EXPR, exit_label); gfc_add_expr_to_block (&loop.pre, tmp); gfc_add_block_to_block (&se->pre, &loop.pre); gfc_add_block_to_block (&se->pre, &loop.post); gfc_cleanup_loop (&loop); se->expr = resvar; } /* COUNT(A) = Number of true elements in A. */ static void gfc_conv_intrinsic_count (gfc_se * se, gfc_expr * expr) { tree resvar; tree type; stmtblock_t body; tree tmp; gfc_loopinfo loop; gfc_actual_arglist *actual; gfc_ss *arrayss; gfc_se arrayse; if (se->ss) { gfc_conv_intrinsic_funcall (se, expr); return; } actual = expr->value.function.actual; type = gfc_typenode_for_spec (&expr->ts); /* Initialize the result. */ resvar = gfc_create_var (type, "count"); gfc_add_modify (&se->pre, resvar, build_int_cst (type, 0)); /* Walk the arguments. */ arrayss = gfc_walk_expr (actual->expr); gcc_assert (arrayss != gfc_ss_terminator); /* Initialize the scalarizer. */ gfc_init_loopinfo (&loop); gfc_add_ss_to_loop (&loop, arrayss); /* Initialize the loop. */ gfc_conv_ss_startstride (&loop); gfc_conv_loop_setup (&loop, &expr->where); gfc_mark_ss_chain_used (arrayss, 1); /* Generate the loop body. */ gfc_start_scalarized_body (&loop, &body); tmp = fold_build2_loc (input_location, PLUS_EXPR, TREE_TYPE (resvar), resvar, build_int_cst (TREE_TYPE (resvar), 1)); tmp = build2_v (MODIFY_EXPR, resvar, tmp); gfc_init_se (&arrayse, NULL); gfc_copy_loopinfo_to_se (&arrayse, &loop); arrayse.ss = arrayss; gfc_conv_expr_val (&arrayse, actual->expr); tmp = build3_v (COND_EXPR, arrayse.expr, tmp, build_empty_stmt (input_location)); gfc_add_block_to_block (&body, &arrayse.pre); gfc_add_expr_to_block (&body, tmp); gfc_add_block_to_block (&body, &arrayse.post); gfc_trans_scalarizing_loops (&loop, &body); gfc_add_block_to_block (&se->pre, &loop.pre); gfc_add_block_to_block (&se->pre, &loop.post); gfc_cleanup_loop (&loop); se->expr = resvar; } /* Update given gfc_se to have ss component pointing to the nested gfc_ss struct and return the corresponding loopinfo. */ static gfc_loopinfo * enter_nested_loop (gfc_se *se) { se->ss = se->ss->nested_ss; gcc_assert (se->ss == se->ss->loop->ss); return se->ss->loop; } /* Inline implementation of the sum and product intrinsics. */ static void gfc_conv_intrinsic_arith (gfc_se * se, gfc_expr * expr, enum tree_code op, bool norm2) { tree resvar; tree scale = NULL_TREE; tree type; stmtblock_t body; stmtblock_t block; tree tmp; gfc_loopinfo loop, *ploop; gfc_actual_arglist *arg_array, *arg_mask; gfc_ss *arrayss = NULL; gfc_ss *maskss = NULL; gfc_se arrayse; gfc_se maskse; gfc_se *parent_se; gfc_expr *arrayexpr; gfc_expr *maskexpr; if (expr->rank > 0) { gcc_assert (gfc_inline_intrinsic_function_p (expr)); parent_se = se; } else parent_se = NULL; type = gfc_typenode_for_spec (&expr->ts); /* Initialize the result. */ resvar = gfc_create_var (type, "val"); if (norm2) { /* result = 0.0; scale = 1.0. */ scale = gfc_create_var (type, "scale"); gfc_add_modify (&se->pre, scale, gfc_build_const (type, integer_one_node)); tmp = gfc_build_const (type, integer_zero_node); } else if (op == PLUS_EXPR || op == BIT_IOR_EXPR || op == BIT_XOR_EXPR) tmp = gfc_build_const (type, integer_zero_node); else if (op == NE_EXPR) /* PARITY. */ tmp = convert (type, boolean_false_node); else if (op == BIT_AND_EXPR) tmp = gfc_build_const (type, fold_build1_loc (input_location, NEGATE_EXPR, type, integer_one_node)); else tmp = gfc_build_const (type, integer_one_node); gfc_add_modify (&se->pre, resvar, tmp); arg_array = expr->value.function.actual; arrayexpr = arg_array->expr; if (op == NE_EXPR || norm2) /* PARITY and NORM2. */ maskexpr = NULL; else { arg_mask = arg_array->next->next; gcc_assert (arg_mask != NULL); maskexpr = arg_mask->expr; } if (expr->rank == 0) { /* Walk the arguments. */ arrayss = gfc_walk_expr (arrayexpr); gcc_assert (arrayss != gfc_ss_terminator); if (maskexpr && maskexpr->rank > 0) { maskss = gfc_walk_expr (maskexpr); gcc_assert (maskss != gfc_ss_terminator); } else maskss = NULL; /* Initialize the scalarizer. */ gfc_init_loopinfo (&loop); gfc_add_ss_to_loop (&loop, arrayss); if (maskexpr && maskexpr->rank > 0) gfc_add_ss_to_loop (&loop, maskss); /* Initialize the loop. */ gfc_conv_ss_startstride (&loop); gfc_conv_loop_setup (&loop, &expr->where); gfc_mark_ss_chain_used (arrayss, 1); if (maskexpr && maskexpr->rank > 0) gfc_mark_ss_chain_used (maskss, 1); ploop = &loop; } else /* All the work has been done in the parent loops. */ ploop = enter_nested_loop (se); gcc_assert (ploop); /* Generate the loop body. */ gfc_start_scalarized_body (ploop, &body); /* If we have a mask, only add this element if the mask is set. */ if (maskexpr && maskexpr->rank > 0) { gfc_init_se (&maskse, parent_se); gfc_copy_loopinfo_to_se (&maskse, ploop); if (expr->rank == 0) maskse.ss = maskss; gfc_conv_expr_val (&maskse, maskexpr); gfc_add_block_to_block (&body, &maskse.pre); gfc_start_block (&block); } else gfc_init_block (&block); /* Do the actual summation/product. */ gfc_init_se (&arrayse, parent_se); gfc_copy_loopinfo_to_se (&arrayse, ploop); if (expr->rank == 0) arrayse.ss = arrayss; gfc_conv_expr_val (&arrayse, arrayexpr); gfc_add_block_to_block (&block, &arrayse.pre); if (norm2) { /* if (x(i) != 0.0) { absX = abs(x(i)) if (absX > scale) { val = scale/absX; result = 1.0 + result * val * val; scale = absX; } else { val = absX/scale; result += val * val; } } */ tree res1, res2, cond, absX, val; stmtblock_t ifblock1, ifblock2, ifblock3; gfc_init_block (&ifblock1); absX = gfc_create_var (type, "absX"); gfc_add_modify (&ifblock1, absX, fold_build1_loc (input_location, ABS_EXPR, type, arrayse.expr)); val = gfc_create_var (type, "val"); gfc_add_expr_to_block (&ifblock1, val); gfc_init_block (&ifblock2); gfc_add_modify (&ifblock2, val, fold_build2_loc (input_location, RDIV_EXPR, type, scale, absX)); res1 = fold_build2_loc (input_location, MULT_EXPR, type, val, val); res1 = fold_build2_loc (input_location, MULT_EXPR, type, resvar, res1); res1 = fold_build2_loc (input_location, PLUS_EXPR, type, res1, gfc_build_const (type, integer_one_node)); gfc_add_modify (&ifblock2, resvar, res1); gfc_add_modify (&ifblock2, scale, absX); res1 = gfc_finish_block (&ifblock2); gfc_init_block (&ifblock3); gfc_add_modify (&ifblock3, val, fold_build2_loc (input_location, RDIV_EXPR, type, absX, scale)); res2 = fold_build2_loc (input_location, MULT_EXPR, type, val, val); res2 = fold_build2_loc (input_location, PLUS_EXPR, type, resvar, res2); gfc_add_modify (&ifblock3, resvar, res2); res2 = gfc_finish_block (&ifblock3); cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, absX, scale); tmp = build3_v (COND_EXPR, cond, res1, res2); gfc_add_expr_to_block (&ifblock1, tmp); tmp = gfc_finish_block (&ifblock1); cond = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, arrayse.expr, gfc_build_const (type, integer_zero_node)); tmp = build3_v (COND_EXPR, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&block, tmp); } else { tmp = fold_build2_loc (input_location, op, type, resvar, arrayse.expr); gfc_add_modify (&block, resvar, tmp); } gfc_add_block_to_block (&block, &arrayse.post); if (maskexpr && maskexpr->rank > 0) { /* We enclose the above in if (mask) {...} . */ tmp = gfc_finish_block (&block); tmp = build3_v (COND_EXPR, maskse.expr, tmp, build_empty_stmt (input_location)); } else tmp = gfc_finish_block (&block); gfc_add_expr_to_block (&body, tmp); gfc_trans_scalarizing_loops (ploop, &body); /* For a scalar mask, enclose the loop in an if statement. */ if (maskexpr && maskexpr->rank == 0) { gfc_init_block (&block); gfc_add_block_to_block (&block, &ploop->pre); gfc_add_block_to_block (&block, &ploop->post); tmp = gfc_finish_block (&block); if (expr->rank > 0) { tmp = build3_v (COND_EXPR, se->ss->info->data.scalar.value, tmp, build_empty_stmt (input_location)); gfc_advance_se_ss_chain (se); } else { gcc_assert (expr->rank == 0); gfc_init_se (&maskse, NULL); gfc_conv_expr_val (&maskse, maskexpr); tmp = build3_v (COND_EXPR, maskse.expr, tmp, build_empty_stmt (input_location)); } gfc_add_expr_to_block (&block, tmp); gfc_add_block_to_block (&se->pre, &block); gcc_assert (se->post.head == NULL); } else { gfc_add_block_to_block (&se->pre, &ploop->pre); gfc_add_block_to_block (&se->pre, &ploop->post); } if (expr->rank == 0) gfc_cleanup_loop (ploop); if (norm2) { /* result = scale * sqrt(result). */ tree sqrt; sqrt = gfc_builtin_decl_for_float_kind (BUILT_IN_SQRT, expr->ts.kind); resvar = build_call_expr_loc (input_location, sqrt, 1, resvar); resvar = fold_build2_loc (input_location, MULT_EXPR, type, scale, resvar); } se->expr = resvar; } /* Inline implementation of the dot_product intrinsic. This function is based on gfc_conv_intrinsic_arith (the previous function). */ static void gfc_conv_intrinsic_dot_product (gfc_se * se, gfc_expr * expr) { tree resvar; tree type; stmtblock_t body; stmtblock_t block; tree tmp; gfc_loopinfo loop; gfc_actual_arglist *actual; gfc_ss *arrayss1, *arrayss2; gfc_se arrayse1, arrayse2; gfc_expr *arrayexpr1, *arrayexpr2; type = gfc_typenode_for_spec (&expr->ts); /* Initialize the result. */ resvar = gfc_create_var (type, "val"); if (expr->ts.type == BT_LOGICAL) tmp = build_int_cst (type, 0); else tmp = gfc_build_const (type, integer_zero_node); gfc_add_modify (&se->pre, resvar, tmp); /* Walk argument #1. */ actual = expr->value.function.actual; arrayexpr1 = actual->expr; arrayss1 = gfc_walk_expr (arrayexpr1); gcc_assert (arrayss1 != gfc_ss_terminator); /* Walk argument #2. */ actual = actual->next; arrayexpr2 = actual->expr; arrayss2 = gfc_walk_expr (arrayexpr2); gcc_assert (arrayss2 != gfc_ss_terminator); /* Initialize the scalarizer. */ gfc_init_loopinfo (&loop); gfc_add_ss_to_loop (&loop, arrayss1); gfc_add_ss_to_loop (&loop, arrayss2); /* Initialize the loop. */ gfc_conv_ss_startstride (&loop); gfc_conv_loop_setup (&loop, &expr->where); gfc_mark_ss_chain_used (arrayss1, 1); gfc_mark_ss_chain_used (arrayss2, 1); /* Generate the loop body. */ gfc_start_scalarized_body (&loop, &body); gfc_init_block (&block); /* Make the tree expression for [conjg(]array1[)]. */ gfc_init_se (&arrayse1, NULL); gfc_copy_loopinfo_to_se (&arrayse1, &loop); arrayse1.ss = arrayss1; gfc_conv_expr_val (&arrayse1, arrayexpr1); if (expr->ts.type == BT_COMPLEX) arrayse1.expr = fold_build1_loc (input_location, CONJ_EXPR, type, arrayse1.expr); gfc_add_block_to_block (&block, &arrayse1.pre); /* Make the tree expression for array2. */ gfc_init_se (&arrayse2, NULL); gfc_copy_loopinfo_to_se (&arrayse2, &loop); arrayse2.ss = arrayss2; gfc_conv_expr_val (&arrayse2, arrayexpr2); gfc_add_block_to_block (&block, &arrayse2.pre); /* Do the actual product and sum. */ if (expr->ts.type == BT_LOGICAL) { tmp = fold_build2_loc (input_location, TRUTH_AND_EXPR, type, arrayse1.expr, arrayse2.expr); tmp = fold_build2_loc (input_location, TRUTH_OR_EXPR, type, resvar, tmp); } else { tmp = fold_build2_loc (input_location, MULT_EXPR, type, arrayse1.expr, arrayse2.expr); tmp = fold_build2_loc (input_location, PLUS_EXPR, type, resvar, tmp); } gfc_add_modify (&block, resvar, tmp); /* Finish up the loop block and the loop. */ tmp = gfc_finish_block (&block); gfc_add_expr_to_block (&body, tmp); gfc_trans_scalarizing_loops (&loop, &body); gfc_add_block_to_block (&se->pre, &loop.pre); gfc_add_block_to_block (&se->pre, &loop.post); gfc_cleanup_loop (&loop); se->expr = resvar; } /* Emit code for minloc or maxloc intrinsic. There are many different cases we need to handle. For performance reasons we sometimes create two loops instead of one, where the second one is much simpler. Examples for minloc intrinsic: 1) Result is an array, a call is generated 2) Array mask is used and NaNs need to be supported: limit = Infinity; pos = 0; S = from; while (S <= to) { if (mask[S]) { if (pos == 0) pos = S + (1 - from); if (a[S] <= limit) { limit = a[S]; pos = S + (1 - from); goto lab1; } } S++; } goto lab2; lab1:; while (S <= to) { if (mask[S]) if (a[S] < limit) { limit = a[S]; pos = S + (1 - from); } S++; } lab2:; 3) NaNs need to be supported, but it is known at compile time or cheaply at runtime whether array is nonempty or not: limit = Infinity; pos = 0; S = from; while (S <= to) { if (a[S] <= limit) { limit = a[S]; pos = S + (1 - from); goto lab1; } S++; } if (from <= to) pos = 1; goto lab2; lab1:; while (S <= to) { if (a[S] < limit) { limit = a[S]; pos = S + (1 - from); } S++; } lab2:; 4) NaNs aren't supported, array mask is used: limit = infinities_supported ? Infinity : huge (limit); pos = 0; S = from; while (S <= to) { if (mask[S]) { limit = a[S]; pos = S + (1 - from); goto lab1; } S++; } goto lab2; lab1:; while (S <= to) { if (mask[S]) if (a[S] < limit) { limit = a[S]; pos = S + (1 - from); } S++; } lab2:; 5) Same without array mask: limit = infinities_supported ? Infinity : huge (limit); pos = (from <= to) ? 1 : 0; S = from; while (S <= to) { if (a[S] < limit) { limit = a[S]; pos = S + (1 - from); } S++; } For 3) and 5), if mask is scalar, this all goes into a conditional, setting pos = 0; in the else branch. */ static void gfc_conv_intrinsic_minmaxloc (gfc_se * se, gfc_expr * expr, enum tree_code op) { stmtblock_t body; stmtblock_t block; stmtblock_t ifblock; stmtblock_t elseblock; tree limit; tree type; tree tmp; tree cond; tree elsetmp; tree ifbody; tree offset; tree nonempty; tree lab1, lab2; gfc_loopinfo loop; gfc_actual_arglist *actual; gfc_ss *arrayss; gfc_ss *maskss; gfc_se arrayse; gfc_se maskse; gfc_expr *arrayexpr; gfc_expr *maskexpr; tree pos; int n; if (se->ss) { gfc_conv_intrinsic_funcall (se, expr); return; } /* Initialize the result. */ pos = gfc_create_var (gfc_array_index_type, "pos"); offset = gfc_create_var (gfc_array_index_type, "offset"); type = gfc_typenode_for_spec (&expr->ts); /* Walk the arguments. */ actual = expr->value.function.actual; arrayexpr = actual->expr; arrayss = gfc_walk_expr (arrayexpr); gcc_assert (arrayss != gfc_ss_terminator); actual = actual->next->next; gcc_assert (actual); maskexpr = actual->expr; nonempty = NULL; if (maskexpr && maskexpr->rank != 0) { maskss = gfc_walk_expr (maskexpr); gcc_assert (maskss != gfc_ss_terminator); } else { mpz_t asize; if (gfc_array_size (arrayexpr, &asize) == SUCCESS) { nonempty = gfc_conv_mpz_to_tree (asize, gfc_index_integer_kind); mpz_clear (asize); nonempty = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, nonempty, gfc_index_zero_node); } maskss = NULL; } limit = gfc_create_var (gfc_typenode_for_spec (&arrayexpr->ts), "limit"); switch (arrayexpr->ts.type) { case BT_REAL: tmp = gfc_build_inf_or_huge (TREE_TYPE (limit), arrayexpr->ts.kind); break; case BT_INTEGER: n = gfc_validate_kind (arrayexpr->ts.type, arrayexpr->ts.kind, false); tmp = gfc_conv_mpz_to_tree (gfc_integer_kinds[n].huge, arrayexpr->ts.kind); break; default: gcc_unreachable (); } /* We start with the most negative possible value for MAXLOC, and the most positive possible value for MINLOC. The most negative possible value is -HUGE for BT_REAL and (-HUGE - 1) for BT_INTEGER; the most positive possible value is HUGE in both cases. */ if (op == GT_EXPR) tmp = fold_build1_loc (input_location, NEGATE_EXPR, TREE_TYPE (tmp), tmp); if (op == GT_EXPR && expr->ts.type == BT_INTEGER) tmp = fold_build2_loc (input_location, MINUS_EXPR, TREE_TYPE (tmp), tmp, build_int_cst (type, 1)); gfc_add_modify (&se->pre, limit, tmp); /* Initialize the scalarizer. */ gfc_init_loopinfo (&loop); gfc_add_ss_to_loop (&loop, arrayss); if (maskss) gfc_add_ss_to_loop (&loop, maskss); /* Initialize the loop. */ gfc_conv_ss_startstride (&loop); /* The code generated can have more than one loop in sequence (see the comment at the function header). This doesn't work well with the scalarizer, which changes arrays' offset when the scalarization loops are generated (see gfc_trans_preloop_setup). Fortunately, {min,max}loc are currently inlined in the scalar case only (for which loop is of rank one). As there is no dependency to care about in that case, there is no temporary, so that we can use the scalarizer temporary code to handle multiple loops. Thus, we set temp_dim here, we call gfc_mark_ss_chain_used with flag=3 later, and we use gfc_trans_scalarized_loop_boundary even later to restore offset. TODO: this prevents inlining of rank > 0 minmaxloc calls, so this should eventually go away. We could either create two loops properly, or find another way to save/restore the array offsets between the two loops (without conflicting with temporary management), or use a single loop minmaxloc implementation. See PR 31067. */ loop.temp_dim = loop.dimen; gfc_conv_loop_setup (&loop, &expr->where); gcc_assert (loop.dimen == 1); if (nonempty == NULL && maskss == NULL && loop.from[0] && loop.to[0]) nonempty = fold_build2_loc (input_location, LE_EXPR, boolean_type_node, loop.from[0], loop.to[0]); lab1 = NULL; lab2 = NULL; /* Initialize the position to zero, following Fortran 2003. We are free to do this because Fortran 95 allows the result of an entirely false mask to be processor dependent. If we know at compile time the array is non-empty and no MASK is used, we can initialize to 1 to simplify the inner loop. */ if (nonempty != NULL && !HONOR_NANS (DECL_MODE (limit))) gfc_add_modify (&loop.pre, pos, fold_build3_loc (input_location, COND_EXPR, gfc_array_index_type, nonempty, gfc_index_one_node, gfc_index_zero_node)); else { gfc_add_modify (&loop.pre, pos, gfc_index_zero_node); lab1 = gfc_build_label_decl (NULL_TREE); TREE_USED (lab1) = 1; lab2 = gfc_build_label_decl (NULL_TREE); TREE_USED (lab2) = 1; } /* An offset must be added to the loop counter to obtain the required position. */ gcc_assert (loop.from[0]); tmp = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, gfc_index_one_node, loop.from[0]); gfc_add_modify (&loop.pre, offset, tmp); gfc_mark_ss_chain_used (arrayss, lab1 ? 3 : 1); if (maskss) gfc_mark_ss_chain_used (maskss, lab1 ? 3 : 1); /* Generate the loop body. */ gfc_start_scalarized_body (&loop, &body); /* If we have a mask, only check this element if the mask is set. */ if (maskss) { gfc_init_se (&maskse, NULL); gfc_copy_loopinfo_to_se (&maskse, &loop); maskse.ss = maskss; gfc_conv_expr_val (&maskse, maskexpr); gfc_add_block_to_block (&body, &maskse.pre); gfc_start_block (&block); } else gfc_init_block (&block); /* Compare with the current limit. */ gfc_init_se (&arrayse, NULL); gfc_copy_loopinfo_to_se (&arrayse, &loop); arrayse.ss = arrayss; gfc_conv_expr_val (&arrayse, arrayexpr); gfc_add_block_to_block (&block, &arrayse.pre); /* We do the following if this is a more extreme value. */ gfc_start_block (&ifblock); /* Assign the value to the limit... */ gfc_add_modify (&ifblock, limit, arrayse.expr); if (nonempty == NULL && HONOR_NANS (DECL_MODE (limit))) { stmtblock_t ifblock2; tree ifbody2; gfc_start_block (&ifblock2); tmp = fold_build2_loc (input_location, PLUS_EXPR, TREE_TYPE (pos), loop.loopvar[0], offset); gfc_add_modify (&ifblock2, pos, tmp); ifbody2 = gfc_finish_block (&ifblock2); cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, pos, gfc_index_zero_node); tmp = build3_v (COND_EXPR, cond, ifbody2, build_empty_stmt (input_location)); gfc_add_expr_to_block (&block, tmp); } tmp = fold_build2_loc (input_location, PLUS_EXPR, TREE_TYPE (pos), loop.loopvar[0], offset); gfc_add_modify (&ifblock, pos, tmp); if (lab1) gfc_add_expr_to_block (&ifblock, build1_v (GOTO_EXPR, lab1)); ifbody = gfc_finish_block (&ifblock); if (!lab1 || HONOR_NANS (DECL_MODE (limit))) { if (lab1) cond = fold_build2_loc (input_location, op == GT_EXPR ? GE_EXPR : LE_EXPR, boolean_type_node, arrayse.expr, limit); else cond = fold_build2_loc (input_location, op, boolean_type_node, arrayse.expr, limit); ifbody = build3_v (COND_EXPR, cond, ifbody, build_empty_stmt (input_location)); } gfc_add_expr_to_block (&block, ifbody); if (maskss) { /* We enclose the above in if (mask) {...}. */ tmp = gfc_finish_block (&block); tmp = build3_v (COND_EXPR, maskse.expr, tmp, build_empty_stmt (input_location)); } else tmp = gfc_finish_block (&block); gfc_add_expr_to_block (&body, tmp); if (lab1) { gfc_trans_scalarized_loop_boundary (&loop, &body); if (HONOR_NANS (DECL_MODE (limit))) { if (nonempty != NULL) { ifbody = build2_v (MODIFY_EXPR, pos, gfc_index_one_node); tmp = build3_v (COND_EXPR, nonempty, ifbody, build_empty_stmt (input_location)); gfc_add_expr_to_block (&loop.code[0], tmp); } } gfc_add_expr_to_block (&loop.code[0], build1_v (GOTO_EXPR, lab2)); gfc_add_expr_to_block (&loop.code[0], build1_v (LABEL_EXPR, lab1)); /* If we have a mask, only check this element if the mask is set. */ if (maskss) { gfc_init_se (&maskse, NULL); gfc_copy_loopinfo_to_se (&maskse, &loop); maskse.ss = maskss; gfc_conv_expr_val (&maskse, maskexpr); gfc_add_block_to_block (&body, &maskse.pre); gfc_start_block (&block); } else gfc_init_block (&block); /* Compare with the current limit. */ gfc_init_se (&arrayse, NULL); gfc_copy_loopinfo_to_se (&arrayse, &loop); arrayse.ss = arrayss; gfc_conv_expr_val (&arrayse, arrayexpr); gfc_add_block_to_block (&block, &arrayse.pre); /* We do the following if this is a more extreme value. */ gfc_start_block (&ifblock); /* Assign the value to the limit... */ gfc_add_modify (&ifblock, limit, arrayse.expr); tmp = fold_build2_loc (input_location, PLUS_EXPR, TREE_TYPE (pos), loop.loopvar[0], offset); gfc_add_modify (&ifblock, pos, tmp); ifbody = gfc_finish_block (&ifblock); cond = fold_build2_loc (input_location, op, boolean_type_node, arrayse.expr, limit); tmp = build3_v (COND_EXPR, cond, ifbody, build_empty_stmt (input_location)); gfc_add_expr_to_block (&block, tmp); if (maskss) { /* We enclose the above in if (mask) {...}. */ tmp = gfc_finish_block (&block); tmp = build3_v (COND_EXPR, maskse.expr, tmp, build_empty_stmt (input_location)); } else tmp = gfc_finish_block (&block); gfc_add_expr_to_block (&body, tmp); /* Avoid initializing loopvar[0] again, it should be left where it finished by the first loop. */ loop.from[0] = loop.loopvar[0]; } gfc_trans_scalarizing_loops (&loop, &body); if (lab2) gfc_add_expr_to_block (&loop.pre, build1_v (LABEL_EXPR, lab2)); /* For a scalar mask, enclose the loop in an if statement. */ if (maskexpr && maskss == NULL) { gfc_init_se (&maskse, NULL); gfc_conv_expr_val (&maskse, maskexpr); gfc_init_block (&block); gfc_add_block_to_block (&block, &loop.pre); gfc_add_block_to_block (&block, &loop.post); tmp = gfc_finish_block (&block); /* For the else part of the scalar mask, just initialize the pos variable the same way as above. */ gfc_init_block (&elseblock); gfc_add_modify (&elseblock, pos, gfc_index_zero_node); elsetmp = gfc_finish_block (&elseblock); tmp = build3_v (COND_EXPR, maskse.expr, tmp, elsetmp); gfc_add_expr_to_block (&block, tmp); gfc_add_block_to_block (&se->pre, &block); } else { gfc_add_block_to_block (&se->pre, &loop.pre); gfc_add_block_to_block (&se->pre, &loop.post); } gfc_cleanup_loop (&loop); se->expr = convert (type, pos); } /* Emit code for minval or maxval intrinsic. There are many different cases we need to handle. For performance reasons we sometimes create two loops instead of one, where the second one is much simpler. Examples for minval intrinsic: 1) Result is an array, a call is generated 2) Array mask is used and NaNs need to be supported, rank 1: limit = Infinity; nonempty = false; S = from; while (S <= to) { if (mask[S]) { nonempty = true; if (a[S] <= limit) goto lab; } S++; } limit = nonempty ? NaN : huge (limit); lab: while (S <= to) { if(mask[S]) limit = min (a[S], limit); S++; } 3) NaNs need to be supported, but it is known at compile time or cheaply at runtime whether array is nonempty or not, rank 1: limit = Infinity; S = from; while (S <= to) { if (a[S] <= limit) goto lab; S++; } limit = (from <= to) ? NaN : huge (limit); lab: while (S <= to) { limit = min (a[S], limit); S++; } 4) Array mask is used and NaNs need to be supported, rank > 1: limit = Infinity; nonempty = false; fast = false; S1 = from1; while (S1 <= to1) { S2 = from2; while (S2 <= to2) { if (mask[S1][S2]) { if (fast) limit = min (a[S1][S2], limit); else { nonempty = true; if (a[S1][S2] <= limit) { limit = a[S1][S2]; fast = true; } } } S2++; } S1++; } if (!fast) limit = nonempty ? NaN : huge (limit); 5) NaNs need to be supported, but it is known at compile time or cheaply at runtime whether array is nonempty or not, rank > 1: limit = Infinity; fast = false; S1 = from1; while (S1 <= to1) { S2 = from2; while (S2 <= to2) { if (fast) limit = min (a[S1][S2], limit); else { if (a[S1][S2] <= limit) { limit = a[S1][S2]; fast = true; } } S2++; } S1++; } if (!fast) limit = (nonempty_array) ? NaN : huge (limit); 6) NaNs aren't supported, but infinities are. Array mask is used: limit = Infinity; nonempty = false; S = from; while (S <= to) { if (mask[S]) { nonempty = true; limit = min (a[S], limit); } S++; } limit = nonempty ? limit : huge (limit); 7) Same without array mask: limit = Infinity; S = from; while (S <= to) { limit = min (a[S], limit); S++; } limit = (from <= to) ? limit : huge (limit); 8) Neither NaNs nor infinities are supported (-ffast-math or BT_INTEGER): limit = huge (limit); S = from; while (S <= to) { limit = min (a[S], limit); S++); } (or while (S <= to) { if (mask[S]) limit = min (a[S], limit); S++; } with array mask instead). For 3), 5), 7) and 8), if mask is scalar, this all goes into a conditional, setting limit = huge (limit); in the else branch. */ static void gfc_conv_intrinsic_minmaxval (gfc_se * se, gfc_expr * expr, enum tree_code op) { tree limit; tree type; tree tmp; tree ifbody; tree nonempty; tree nonempty_var; tree lab; tree fast; tree huge_cst = NULL, nan_cst = NULL; stmtblock_t body; stmtblock_t block, block2; gfc_loopinfo loop; gfc_actual_arglist *actual; gfc_ss *arrayss; gfc_ss *maskss; gfc_se arrayse; gfc_se maskse; gfc_expr *arrayexpr; gfc_expr *maskexpr; int n; if (se->ss) { gfc_conv_intrinsic_funcall (se, expr); return; } type = gfc_typenode_for_spec (&expr->ts); /* Initialize the result. */ limit = gfc_create_var (type, "limit"); n = gfc_validate_kind (expr->ts.type, expr->ts.kind, false); switch (expr->ts.type) { case BT_REAL: huge_cst = gfc_conv_mpfr_to_tree (gfc_real_kinds[n].huge, expr->ts.kind, 0); if (HONOR_INFINITIES (DECL_MODE (limit))) { REAL_VALUE_TYPE real; real_inf (&real); tmp = build_real (type, real); } else tmp = huge_cst; if (HONOR_NANS (DECL_MODE (limit))) { REAL_VALUE_TYPE real; real_nan (&real, "", 1, DECL_MODE (limit)); nan_cst = build_real (type, real); } break; case BT_INTEGER: tmp = gfc_conv_mpz_to_tree (gfc_integer_kinds[n].huge, expr->ts.kind); break; default: gcc_unreachable (); } /* We start with the most negative possible value for MAXVAL, and the most positive possible value for MINVAL. The most negative possible value is -HUGE for BT_REAL and (-HUGE - 1) for BT_INTEGER; the most positive possible value is HUGE in both cases. */ if (op == GT_EXPR) { tmp = fold_build1_loc (input_location, NEGATE_EXPR, TREE_TYPE (tmp), tmp); if (huge_cst) huge_cst = fold_build1_loc (input_location, NEGATE_EXPR, TREE_TYPE (huge_cst), huge_cst); } if (op == GT_EXPR && expr->ts.type == BT_INTEGER) tmp = fold_build2_loc (input_location, MINUS_EXPR, TREE_TYPE (tmp), tmp, build_int_cst (type, 1)); gfc_add_modify (&se->pre, limit, tmp); /* Walk the arguments. */ actual = expr->value.function.actual; arrayexpr = actual->expr; arrayss = gfc_walk_expr (arrayexpr); gcc_assert (arrayss != gfc_ss_terminator); actual = actual->next->next; gcc_assert (actual); maskexpr = actual->expr; nonempty = NULL; if (maskexpr && maskexpr->rank != 0) { maskss = gfc_walk_expr (maskexpr); gcc_assert (maskss != gfc_ss_terminator); } else { mpz_t asize; if (gfc_array_size (arrayexpr, &asize) == SUCCESS) { nonempty = gfc_conv_mpz_to_tree (asize, gfc_index_integer_kind); mpz_clear (asize); nonempty = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, nonempty, gfc_index_zero_node); } maskss = NULL; } /* Initialize the scalarizer. */ gfc_init_loopinfo (&loop); gfc_add_ss_to_loop (&loop, arrayss); if (maskss) gfc_add_ss_to_loop (&loop, maskss); /* Initialize the loop. */ gfc_conv_ss_startstride (&loop); /* The code generated can have more than one loop in sequence (see the comment at the function header). This doesn't work well with the scalarizer, which changes arrays' offset when the scalarization loops are generated (see gfc_trans_preloop_setup). Fortunately, {min,max}val are currently inlined in the scalar case only. As there is no dependency to care about in that case, there is no temporary, so that we can use the scalarizer temporary code to handle multiple loops. Thus, we set temp_dim here, we call gfc_mark_ss_chain_used with flag=3 later, and we use gfc_trans_scalarized_loop_boundary even later to restore offset. TODO: this prevents inlining of rank > 0 minmaxval calls, so this should eventually go away. We could either create two loops properly, or find another way to save/restore the array offsets between the two loops (without conflicting with temporary management), or use a single loop minmaxval implementation. See PR 31067. */ loop.temp_dim = loop.dimen; gfc_conv_loop_setup (&loop, &expr->where); if (nonempty == NULL && maskss == NULL && loop.dimen == 1 && loop.from[0] && loop.to[0]) nonempty = fold_build2_loc (input_location, LE_EXPR, boolean_type_node, loop.from[0], loop.to[0]); nonempty_var = NULL; if (nonempty == NULL && (HONOR_INFINITIES (DECL_MODE (limit)) || HONOR_NANS (DECL_MODE (limit)))) { nonempty_var = gfc_create_var (boolean_type_node, "nonempty"); gfc_add_modify (&se->pre, nonempty_var, boolean_false_node); nonempty = nonempty_var; } lab = NULL; fast = NULL; if (HONOR_NANS (DECL_MODE (limit))) { if (loop.dimen == 1) { lab = gfc_build_label_decl (NULL_TREE); TREE_USED (lab) = 1; } else { fast = gfc_create_var (boolean_type_node, "fast"); gfc_add_modify (&se->pre, fast, boolean_false_node); } } gfc_mark_ss_chain_used (arrayss, lab ? 3 : 1); if (maskss) gfc_mark_ss_chain_used (maskss, lab ? 3 : 1); /* Generate the loop body. */ gfc_start_scalarized_body (&loop, &body); /* If we have a mask, only add this element if the mask is set. */ if (maskss) { gfc_init_se (&maskse, NULL); gfc_copy_loopinfo_to_se (&maskse, &loop); maskse.ss = maskss; gfc_conv_expr_val (&maskse, maskexpr); gfc_add_block_to_block (&body, &maskse.pre); gfc_start_block (&block); } else gfc_init_block (&block); /* Compare with the current limit. */ gfc_init_se (&arrayse, NULL); gfc_copy_loopinfo_to_se (&arrayse, &loop); arrayse.ss = arrayss; gfc_conv_expr_val (&arrayse, arrayexpr); gfc_add_block_to_block (&block, &arrayse.pre); gfc_init_block (&block2); if (nonempty_var) gfc_add_modify (&block2, nonempty_var, boolean_true_node); if (HONOR_NANS (DECL_MODE (limit))) { tmp = fold_build2_loc (input_location, op == GT_EXPR ? GE_EXPR : LE_EXPR, boolean_type_node, arrayse.expr, limit); if (lab) ifbody = build1_v (GOTO_EXPR, lab); else { stmtblock_t ifblock; gfc_init_block (&ifblock); gfc_add_modify (&ifblock, limit, arrayse.expr); gfc_add_modify (&ifblock, fast, boolean_true_node); ifbody = gfc_finish_block (&ifblock); } tmp = build3_v (COND_EXPR, tmp, ifbody, build_empty_stmt (input_location)); gfc_add_expr_to_block (&block2, tmp); } else { /* MIN_EXPR/MAX_EXPR has unspecified behavior with NaNs or signed zeros. */ if (HONOR_SIGNED_ZEROS (DECL_MODE (limit))) { tmp = fold_build2_loc (input_location, op, boolean_type_node, arrayse.expr, limit); ifbody = build2_v (MODIFY_EXPR, limit, arrayse.expr); tmp = build3_v (COND_EXPR, tmp, ifbody, build_empty_stmt (input_location)); gfc_add_expr_to_block (&block2, tmp); } else { tmp = fold_build2_loc (input_location, op == GT_EXPR ? MAX_EXPR : MIN_EXPR, type, arrayse.expr, limit); gfc_add_modify (&block2, limit, tmp); } } if (fast) { tree elsebody = gfc_finish_block (&block2); /* MIN_EXPR/MAX_EXPR has unspecified behavior with NaNs or signed zeros. */ if (HONOR_NANS (DECL_MODE (limit)) || HONOR_SIGNED_ZEROS (DECL_MODE (limit))) { tmp = fold_build2_loc (input_location, op, boolean_type_node, arrayse.expr, limit); ifbody = build2_v (MODIFY_EXPR, limit, arrayse.expr); ifbody = build3_v (COND_EXPR, tmp, ifbody, build_empty_stmt (input_location)); } else { tmp = fold_build2_loc (input_location, op == GT_EXPR ? MAX_EXPR : MIN_EXPR, type, arrayse.expr, limit); ifbody = build2_v (MODIFY_EXPR, limit, tmp); } tmp = build3_v (COND_EXPR, fast, ifbody, elsebody); gfc_add_expr_to_block (&block, tmp); } else gfc_add_block_to_block (&block, &block2); gfc_add_block_to_block (&block, &arrayse.post); tmp = gfc_finish_block (&block); if (maskss) /* We enclose the above in if (mask) {...}. */ tmp = build3_v (COND_EXPR, maskse.expr, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&body, tmp); if (lab) { gfc_trans_scalarized_loop_boundary (&loop, &body); tmp = fold_build3_loc (input_location, COND_EXPR, type, nonempty, nan_cst, huge_cst); gfc_add_modify (&loop.code[0], limit, tmp); gfc_add_expr_to_block (&loop.code[0], build1_v (LABEL_EXPR, lab)); /* If we have a mask, only add this element if the mask is set. */ if (maskss) { gfc_init_se (&maskse, NULL); gfc_copy_loopinfo_to_se (&maskse, &loop); maskse.ss = maskss; gfc_conv_expr_val (&maskse, maskexpr); gfc_add_block_to_block (&body, &maskse.pre); gfc_start_block (&block); } else gfc_init_block (&block); /* Compare with the current limit. */ gfc_init_se (&arrayse, NULL); gfc_copy_loopinfo_to_se (&arrayse, &loop); arrayse.ss = arrayss; gfc_conv_expr_val (&arrayse, arrayexpr); gfc_add_block_to_block (&block, &arrayse.pre); /* MIN_EXPR/MAX_EXPR has unspecified behavior with NaNs or signed zeros. */ if (HONOR_NANS (DECL_MODE (limit)) || HONOR_SIGNED_ZEROS (DECL_MODE (limit))) { tmp = fold_build2_loc (input_location, op, boolean_type_node, arrayse.expr, limit); ifbody = build2_v (MODIFY_EXPR, limit, arrayse.expr); tmp = build3_v (COND_EXPR, tmp, ifbody, build_empty_stmt (input_location)); gfc_add_expr_to_block (&block, tmp); } else { tmp = fold_build2_loc (input_location, op == GT_EXPR ? MAX_EXPR : MIN_EXPR, type, arrayse.expr, limit); gfc_add_modify (&block, limit, tmp); } gfc_add_block_to_block (&block, &arrayse.post); tmp = gfc_finish_block (&block); if (maskss) /* We enclose the above in if (mask) {...}. */ tmp = build3_v (COND_EXPR, maskse.expr, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&body, tmp); /* Avoid initializing loopvar[0] again, it should be left where it finished by the first loop. */ loop.from[0] = loop.loopvar[0]; } gfc_trans_scalarizing_loops (&loop, &body); if (fast) { tmp = fold_build3_loc (input_location, COND_EXPR, type, nonempty, nan_cst, huge_cst); ifbody = build2_v (MODIFY_EXPR, limit, tmp); tmp = build3_v (COND_EXPR, fast, build_empty_stmt (input_location), ifbody); gfc_add_expr_to_block (&loop.pre, tmp); } else if (HONOR_INFINITIES (DECL_MODE (limit)) && !lab) { tmp = fold_build3_loc (input_location, COND_EXPR, type, nonempty, limit, huge_cst); gfc_add_modify (&loop.pre, limit, tmp); } /* For a scalar mask, enclose the loop in an if statement. */ if (maskexpr && maskss == NULL) { tree else_stmt; gfc_init_se (&maskse, NULL); gfc_conv_expr_val (&maskse, maskexpr); gfc_init_block (&block); gfc_add_block_to_block (&block, &loop.pre); gfc_add_block_to_block (&block, &loop.post); tmp = gfc_finish_block (&block); if (HONOR_INFINITIES (DECL_MODE (limit))) else_stmt = build2_v (MODIFY_EXPR, limit, huge_cst); else else_stmt = build_empty_stmt (input_location); tmp = build3_v (COND_EXPR, maskse.expr, tmp, else_stmt); gfc_add_expr_to_block (&block, tmp); gfc_add_block_to_block (&se->pre, &block); } else { gfc_add_block_to_block (&se->pre, &loop.pre); gfc_add_block_to_block (&se->pre, &loop.post); } gfc_cleanup_loop (&loop); se->expr = limit; } /* BTEST (i, pos) = (i & (1 << pos)) != 0. */ static void gfc_conv_intrinsic_btest (gfc_se * se, gfc_expr * expr) { tree args[2]; tree type; tree tmp; gfc_conv_intrinsic_function_args (se, expr, args, 2); type = TREE_TYPE (args[0]); tmp = fold_build2_loc (input_location, LSHIFT_EXPR, type, build_int_cst (type, 1), args[1]); tmp = fold_build2_loc (input_location, BIT_AND_EXPR, type, args[0], tmp); tmp = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, tmp, build_int_cst (type, 0)); type = gfc_typenode_for_spec (&expr->ts); se->expr = convert (type, tmp); } /* Generate code for BGE, BGT, BLE and BLT intrinsics. */ static void gfc_conv_intrinsic_bitcomp (gfc_se * se, gfc_expr * expr, enum tree_code op) { tree args[2]; gfc_conv_intrinsic_function_args (se, expr, args, 2); /* Convert both arguments to the unsigned type of the same size. */ args[0] = fold_convert (unsigned_type_for (TREE_TYPE (args[0])), args[0]); args[1] = fold_convert (unsigned_type_for (TREE_TYPE (args[1])), args[1]); /* If they have unequal type size, convert to the larger one. */ if (TYPE_PRECISION (TREE_TYPE (args[0])) > TYPE_PRECISION (TREE_TYPE (args[1]))) args[1] = fold_convert (TREE_TYPE (args[0]), args[1]); else if (TYPE_PRECISION (TREE_TYPE (args[1])) > TYPE_PRECISION (TREE_TYPE (args[0]))) args[0] = fold_convert (TREE_TYPE (args[1]), args[0]); /* Now, we compare them. */ se->expr = fold_build2_loc (input_location, op, boolean_type_node, args[0], args[1]); } /* Generate code to perform the specified operation. */ static void gfc_conv_intrinsic_bitop (gfc_se * se, gfc_expr * expr, enum tree_code op) { tree args[2]; gfc_conv_intrinsic_function_args (se, expr, args, 2); se->expr = fold_build2_loc (input_location, op, TREE_TYPE (args[0]), args[0], args[1]); } /* Bitwise not. */ static void gfc_conv_intrinsic_not (gfc_se * se, gfc_expr * expr) { tree arg; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); se->expr = fold_build1_loc (input_location, BIT_NOT_EXPR, TREE_TYPE (arg), arg); } /* Set or clear a single bit. */ static void gfc_conv_intrinsic_singlebitop (gfc_se * se, gfc_expr * expr, int set) { tree args[2]; tree type; tree tmp; enum tree_code op; gfc_conv_intrinsic_function_args (se, expr, args, 2); type = TREE_TYPE (args[0]); tmp = fold_build2_loc (input_location, LSHIFT_EXPR, type, build_int_cst (type, 1), args[1]); if (set) op = BIT_IOR_EXPR; else { op = BIT_AND_EXPR; tmp = fold_build1_loc (input_location, BIT_NOT_EXPR, type, tmp); } se->expr = fold_build2_loc (input_location, op, type, args[0], tmp); } /* Extract a sequence of bits. IBITS(I, POS, LEN) = (I >> POS) & ~((~0) << LEN). */ static void gfc_conv_intrinsic_ibits (gfc_se * se, gfc_expr * expr) { tree args[3]; tree type; tree tmp; tree mask; gfc_conv_intrinsic_function_args (se, expr, args, 3); type = TREE_TYPE (args[0]); mask = build_int_cst (type, -1); mask = fold_build2_loc (input_location, LSHIFT_EXPR, type, mask, args[2]); mask = fold_build1_loc (input_location, BIT_NOT_EXPR, type, mask); tmp = fold_build2_loc (input_location, RSHIFT_EXPR, type, args[0], args[1]); se->expr = fold_build2_loc (input_location, BIT_AND_EXPR, type, tmp, mask); } static void gfc_conv_intrinsic_shift (gfc_se * se, gfc_expr * expr, bool right_shift, bool arithmetic) { tree args[2], type, num_bits, cond; gfc_conv_intrinsic_function_args (se, expr, args, 2); args[0] = gfc_evaluate_now (args[0], &se->pre); args[1] = gfc_evaluate_now (args[1], &se->pre); type = TREE_TYPE (args[0]); if (!arithmetic) args[0] = fold_convert (unsigned_type_for (type), args[0]); else gcc_assert (right_shift); se->expr = fold_build2_loc (input_location, right_shift ? RSHIFT_EXPR : LSHIFT_EXPR, TREE_TYPE (args[0]), args[0], args[1]); if (!arithmetic) se->expr = fold_convert (type, se->expr); /* The Fortran standard allows shift widths <= BIT_SIZE(I), whereas gcc requires a shift width < BIT_SIZE(I), so we have to catch this special case. */ num_bits = build_int_cst (TREE_TYPE (args[1]), TYPE_PRECISION (type)); cond = fold_build2_loc (input_location, GE_EXPR, boolean_type_node, args[1], num_bits); se->expr = fold_build3_loc (input_location, COND_EXPR, type, cond, build_int_cst (type, 0), se->expr); } /* ISHFT (I, SHIFT) = (abs (shift) >= BIT_SIZE (i)) ? 0 : ((shift >= 0) ? i << shift : i >> -shift) where all shifts are logical shifts. */ static void gfc_conv_intrinsic_ishft (gfc_se * se, gfc_expr * expr) { tree args[2]; tree type; tree utype; tree tmp; tree width; tree num_bits; tree cond; tree lshift; tree rshift; gfc_conv_intrinsic_function_args (se, expr, args, 2); args[0] = gfc_evaluate_now (args[0], &se->pre); args[1] = gfc_evaluate_now (args[1], &se->pre); type = TREE_TYPE (args[0]); utype = unsigned_type_for (type); width = fold_build1_loc (input_location, ABS_EXPR, TREE_TYPE (args[1]), args[1]); /* Left shift if positive. */ lshift = fold_build2_loc (input_location, LSHIFT_EXPR, type, args[0], width); /* Right shift if negative. We convert to an unsigned type because we want a logical shift. The standard doesn't define the case of shifting negative numbers, and we try to be compatible with other compilers, most notably g77, here. */ rshift = fold_convert (type, fold_build2_loc (input_location, RSHIFT_EXPR, utype, convert (utype, args[0]), width)); tmp = fold_build2_loc (input_location, GE_EXPR, boolean_type_node, args[1], build_int_cst (TREE_TYPE (args[1]), 0)); tmp = fold_build3_loc (input_location, COND_EXPR, type, tmp, lshift, rshift); /* The Fortran standard allows shift widths <= BIT_SIZE(I), whereas gcc requires a shift width < BIT_SIZE(I), so we have to catch this special case. */ num_bits = build_int_cst (TREE_TYPE (args[1]), TYPE_PRECISION (type)); cond = fold_build2_loc (input_location, GE_EXPR, boolean_type_node, width, num_bits); se->expr = fold_build3_loc (input_location, COND_EXPR, type, cond, build_int_cst (type, 0), tmp); } /* Circular shift. AKA rotate or barrel shift. */ static void gfc_conv_intrinsic_ishftc (gfc_se * se, gfc_expr * expr) { tree *args; tree type; tree tmp; tree lrot; tree rrot; tree zero; unsigned int num_args; num_args = gfc_intrinsic_argument_list_length (expr); args = XALLOCAVEC (tree, num_args); gfc_conv_intrinsic_function_args (se, expr, args, num_args); if (num_args == 3) { /* Use a library function for the 3 parameter version. */ tree int4type = gfc_get_int_type (4); type = TREE_TYPE (args[0]); /* We convert the first argument to at least 4 bytes, and convert back afterwards. This removes the need for library functions for all argument sizes, and function will be aligned to at least 32 bits, so there's no loss. */ if (expr->ts.kind < 4) args[0] = convert (int4type, args[0]); /* Convert the SHIFT and SIZE args to INTEGER*4 otherwise we would need loads of library functions. They cannot have values > BIT_SIZE (I) so the conversion is safe. */ args[1] = convert (int4type, args[1]); args[2] = convert (int4type, args[2]); switch (expr->ts.kind) { case 1: case 2: case 4: tmp = gfor_fndecl_math_ishftc4; break; case 8: tmp = gfor_fndecl_math_ishftc8; break; case 16: tmp = gfor_fndecl_math_ishftc16; break; default: gcc_unreachable (); } se->expr = build_call_expr_loc (input_location, tmp, 3, args[0], args[1], args[2]); /* Convert the result back to the original type, if we extended the first argument's width above. */ if (expr->ts.kind < 4) se->expr = convert (type, se->expr); return; } type = TREE_TYPE (args[0]); /* Evaluate arguments only once. */ args[0] = gfc_evaluate_now (args[0], &se->pre); args[1] = gfc_evaluate_now (args[1], &se->pre); /* Rotate left if positive. */ lrot = fold_build2_loc (input_location, LROTATE_EXPR, type, args[0], args[1]); /* Rotate right if negative. */ tmp = fold_build1_loc (input_location, NEGATE_EXPR, TREE_TYPE (args[1]), args[1]); rrot = fold_build2_loc (input_location,RROTATE_EXPR, type, args[0], tmp); zero = build_int_cst (TREE_TYPE (args[1]), 0); tmp = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, args[1], zero); rrot = fold_build3_loc (input_location, COND_EXPR, type, tmp, lrot, rrot); /* Do nothing if shift == 0. */ tmp = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, args[1], zero); se->expr = fold_build3_loc (input_location, COND_EXPR, type, tmp, args[0], rrot); } /* LEADZ (i) = (i == 0) ? BIT_SIZE (i) : __builtin_clz(i) - (BIT_SIZE('int') - BIT_SIZE(i)) The conditional expression is necessary because the result of LEADZ(0) is defined, but the result of __builtin_clz(0) is undefined for most targets. For INTEGER kinds smaller than the C 'int' type, we have to subtract the difference in bit size between the argument of LEADZ and the C int. */ static void gfc_conv_intrinsic_leadz (gfc_se * se, gfc_expr * expr) { tree arg; tree arg_type; tree cond; tree result_type; tree leadz; tree bit_size; tree tmp; tree func; int s, argsize; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); argsize = TYPE_PRECISION (TREE_TYPE (arg)); /* Which variant of __builtin_clz* should we call? */ if (argsize <= INT_TYPE_SIZE) { arg_type = unsigned_type_node; func = builtin_decl_explicit (BUILT_IN_CLZ); } else if (argsize <= LONG_TYPE_SIZE) { arg_type = long_unsigned_type_node; func = builtin_decl_explicit (BUILT_IN_CLZL); } else if (argsize <= LONG_LONG_TYPE_SIZE) { arg_type = long_long_unsigned_type_node; func = builtin_decl_explicit (BUILT_IN_CLZLL); } else { gcc_assert (argsize == 2 * LONG_LONG_TYPE_SIZE); arg_type = gfc_build_uint_type (argsize); func = NULL_TREE; } /* Convert the actual argument twice: first, to the unsigned type of the same size; then, to the proper argument type for the built-in function. But the return type is of the default INTEGER kind. */ arg = fold_convert (gfc_build_uint_type (argsize), arg); arg = fold_convert (arg_type, arg); arg = gfc_evaluate_now (arg, &se->pre); result_type = gfc_get_int_type (gfc_default_integer_kind); /* Compute LEADZ for the case i .ne. 0. */ if (func) { s = TYPE_PRECISION (arg_type) - argsize; tmp = fold_convert (result_type, build_call_expr_loc (input_location, func, 1, arg)); leadz = fold_build2_loc (input_location, MINUS_EXPR, result_type, tmp, build_int_cst (result_type, s)); } else { /* We end up here if the argument type is larger than 'long long'. We generate this code: if (x & (ULL_MAX << ULL_SIZE) != 0) return clzll ((unsigned long long) (x >> ULLSIZE)); else return ULL_SIZE + clzll ((unsigned long long) x); where ULL_MAX is the largest value that a ULL_MAX can hold (0xFFFFFFFFFFFFFFFF for a 64-bit long long type), and ULLSIZE is the bit-size of the long long type (64 in this example). */ tree ullsize, ullmax, tmp1, tmp2, btmp; ullsize = build_int_cst (result_type, LONG_LONG_TYPE_SIZE); ullmax = fold_build1_loc (input_location, BIT_NOT_EXPR, long_long_unsigned_type_node, build_int_cst (long_long_unsigned_type_node, 0)); cond = fold_build2_loc (input_location, LSHIFT_EXPR, arg_type, fold_convert (arg_type, ullmax), ullsize); cond = fold_build2_loc (input_location, BIT_AND_EXPR, arg_type, arg, cond); cond = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, cond, build_int_cst (arg_type, 0)); tmp1 = fold_build2_loc (input_location, RSHIFT_EXPR, arg_type, arg, ullsize); tmp1 = fold_convert (long_long_unsigned_type_node, tmp1); btmp = builtin_decl_explicit (BUILT_IN_CLZLL); tmp1 = fold_convert (result_type, build_call_expr_loc (input_location, btmp, 1, tmp1)); tmp2 = fold_convert (long_long_unsigned_type_node, arg); btmp = builtin_decl_explicit (BUILT_IN_CLZLL); tmp2 = fold_convert (result_type, build_call_expr_loc (input_location, btmp, 1, tmp2)); tmp2 = fold_build2_loc (input_location, PLUS_EXPR, result_type, tmp2, ullsize); leadz = fold_build3_loc (input_location, COND_EXPR, result_type, cond, tmp1, tmp2); } /* Build BIT_SIZE. */ bit_size = build_int_cst (result_type, argsize); cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, arg, build_int_cst (arg_type, 0)); se->expr = fold_build3_loc (input_location, COND_EXPR, result_type, cond, bit_size, leadz); } /* TRAILZ(i) = (i == 0) ? BIT_SIZE (i) : __builtin_ctz(i) The conditional expression is necessary because the result of TRAILZ(0) is defined, but the result of __builtin_ctz(0) is undefined for most targets. */ static void gfc_conv_intrinsic_trailz (gfc_se * se, gfc_expr *expr) { tree arg; tree arg_type; tree cond; tree result_type; tree trailz; tree bit_size; tree func; int argsize; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); argsize = TYPE_PRECISION (TREE_TYPE (arg)); /* Which variant of __builtin_ctz* should we call? */ if (argsize <= INT_TYPE_SIZE) { arg_type = unsigned_type_node; func = builtin_decl_explicit (BUILT_IN_CTZ); } else if (argsize <= LONG_TYPE_SIZE) { arg_type = long_unsigned_type_node; func = builtin_decl_explicit (BUILT_IN_CTZL); } else if (argsize <= LONG_LONG_TYPE_SIZE) { arg_type = long_long_unsigned_type_node; func = builtin_decl_explicit (BUILT_IN_CTZLL); } else { gcc_assert (argsize == 2 * LONG_LONG_TYPE_SIZE); arg_type = gfc_build_uint_type (argsize); func = NULL_TREE; } /* Convert the actual argument twice: first, to the unsigned type of the same size; then, to the proper argument type for the built-in function. But the return type is of the default INTEGER kind. */ arg = fold_convert (gfc_build_uint_type (argsize), arg); arg = fold_convert (arg_type, arg); arg = gfc_evaluate_now (arg, &se->pre); result_type = gfc_get_int_type (gfc_default_integer_kind); /* Compute TRAILZ for the case i .ne. 0. */ if (func) trailz = fold_convert (result_type, build_call_expr_loc (input_location, func, 1, arg)); else { /* We end up here if the argument type is larger than 'long long'. We generate this code: if ((x & ULL_MAX) == 0) return ULL_SIZE + ctzll ((unsigned long long) (x >> ULLSIZE)); else return ctzll ((unsigned long long) x); where ULL_MAX is the largest value that a ULL_MAX can hold (0xFFFFFFFFFFFFFFFF for a 64-bit long long type), and ULLSIZE is the bit-size of the long long type (64 in this example). */ tree ullsize, ullmax, tmp1, tmp2, btmp; ullsize = build_int_cst (result_type, LONG_LONG_TYPE_SIZE); ullmax = fold_build1_loc (input_location, BIT_NOT_EXPR, long_long_unsigned_type_node, build_int_cst (long_long_unsigned_type_node, 0)); cond = fold_build2_loc (input_location, BIT_AND_EXPR, arg_type, arg, fold_convert (arg_type, ullmax)); cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, cond, build_int_cst (arg_type, 0)); tmp1 = fold_build2_loc (input_location, RSHIFT_EXPR, arg_type, arg, ullsize); tmp1 = fold_convert (long_long_unsigned_type_node, tmp1); btmp = builtin_decl_explicit (BUILT_IN_CTZLL); tmp1 = fold_convert (result_type, build_call_expr_loc (input_location, btmp, 1, tmp1)); tmp1 = fold_build2_loc (input_location, PLUS_EXPR, result_type, tmp1, ullsize); tmp2 = fold_convert (long_long_unsigned_type_node, arg); btmp = builtin_decl_explicit (BUILT_IN_CTZLL); tmp2 = fold_convert (result_type, build_call_expr_loc (input_location, btmp, 1, tmp2)); trailz = fold_build3_loc (input_location, COND_EXPR, result_type, cond, tmp1, tmp2); } /* Build BIT_SIZE. */ bit_size = build_int_cst (result_type, argsize); cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, arg, build_int_cst (arg_type, 0)); se->expr = fold_build3_loc (input_location, COND_EXPR, result_type, cond, bit_size, trailz); } /* Using __builtin_popcount for POPCNT and __builtin_parity for POPPAR; for types larger than "long long", we call the long long built-in for the lower and higher bits and combine the result. */ static void gfc_conv_intrinsic_popcnt_poppar (gfc_se * se, gfc_expr *expr, int parity) { tree arg; tree arg_type; tree result_type; tree func; int argsize; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); argsize = TYPE_PRECISION (TREE_TYPE (arg)); result_type = gfc_get_int_type (gfc_default_integer_kind); /* Which variant of the builtin should we call? */ if (argsize <= INT_TYPE_SIZE) { arg_type = unsigned_type_node; func = builtin_decl_explicit (parity ? BUILT_IN_PARITY : BUILT_IN_POPCOUNT); } else if (argsize <= LONG_TYPE_SIZE) { arg_type = long_unsigned_type_node; func = builtin_decl_explicit (parity ? BUILT_IN_PARITYL : BUILT_IN_POPCOUNTL); } else if (argsize <= LONG_LONG_TYPE_SIZE) { arg_type = long_long_unsigned_type_node; func = builtin_decl_explicit (parity ? BUILT_IN_PARITYLL : BUILT_IN_POPCOUNTLL); } else { /* Our argument type is larger than 'long long', which mean none of the POPCOUNT builtins covers it. We thus call the 'long long' variant multiple times, and add the results. */ tree utype, arg2, call1, call2; /* For now, we only cover the case where argsize is twice as large as 'long long'. */ gcc_assert (argsize == 2 * LONG_LONG_TYPE_SIZE); func = builtin_decl_explicit (parity ? BUILT_IN_PARITYLL : BUILT_IN_POPCOUNTLL); /* Convert it to an integer, and store into a variable. */ utype = gfc_build_uint_type (argsize); arg = fold_convert (utype, arg); arg = gfc_evaluate_now (arg, &se->pre); /* Call the builtin twice. */ call1 = build_call_expr_loc (input_location, func, 1, fold_convert (long_long_unsigned_type_node, arg)); arg2 = fold_build2_loc (input_location, RSHIFT_EXPR, utype, arg, build_int_cst (utype, LONG_LONG_TYPE_SIZE)); call2 = build_call_expr_loc (input_location, func, 1, fold_convert (long_long_unsigned_type_node, arg2)); /* Combine the results. */ if (parity) se->expr = fold_build2_loc (input_location, BIT_XOR_EXPR, result_type, call1, call2); else se->expr = fold_build2_loc (input_location, PLUS_EXPR, result_type, call1, call2); return; } /* Convert the actual argument twice: first, to the unsigned type of the same size; then, to the proper argument type for the built-in function. */ arg = fold_convert (gfc_build_uint_type (argsize), arg); arg = fold_convert (arg_type, arg); se->expr = fold_convert (result_type, build_call_expr_loc (input_location, func, 1, arg)); } /* Process an intrinsic with unspecified argument-types that has an optional argument (which could be of type character), e.g. EOSHIFT. For those, we need to append the string length of the optional argument if it is not present and the type is really character. primary specifies the position (starting at 1) of the non-optional argument specifying the type and optional gives the position of the optional argument in the arglist. */ static void conv_generic_with_optional_char_arg (gfc_se* se, gfc_expr* expr, unsigned primary, unsigned optional) { gfc_actual_arglist* prim_arg; gfc_actual_arglist* opt_arg; unsigned cur_pos; gfc_actual_arglist* arg; gfc_symbol* sym; VEC(tree,gc) *append_args; /* Find the two arguments given as position. */ cur_pos = 0; prim_arg = NULL; opt_arg = NULL; for (arg = expr->value.function.actual; arg; arg = arg->next) { ++cur_pos; if (cur_pos == primary) prim_arg = arg; if (cur_pos == optional) opt_arg = arg; if (cur_pos >= primary && cur_pos >= optional) break; } gcc_assert (prim_arg); gcc_assert (prim_arg->expr); gcc_assert (opt_arg); /* If we do have type CHARACTER and the optional argument is really absent, append a dummy 0 as string length. */ append_args = NULL; if (prim_arg->expr->ts.type == BT_CHARACTER && !opt_arg->expr) { tree dummy; dummy = build_int_cst (gfc_charlen_type_node, 0); append_args = VEC_alloc (tree, gc, 1); VEC_quick_push (tree, append_args, dummy); } /* Build the call itself. */ sym = gfc_get_symbol_for_expr (expr); gfc_conv_procedure_call (se, sym, expr->value.function.actual, expr, append_args); gfc_free_symbol (sym); } /* The length of a character string. */ static void gfc_conv_intrinsic_len (gfc_se * se, gfc_expr * expr) { tree len; tree type; tree decl; gfc_symbol *sym; gfc_se argse; gfc_expr *arg; gcc_assert (!se->ss); arg = expr->value.function.actual->expr; type = gfc_typenode_for_spec (&expr->ts); switch (arg->expr_type) { case EXPR_CONSTANT: len = build_int_cst (gfc_charlen_type_node, arg->value.character.length); break; case EXPR_ARRAY: /* Obtain the string length from the function used by trans-array.c(gfc_trans_array_constructor). */ len = NULL_TREE; get_array_ctor_strlen (&se->pre, arg->value.constructor, &len); break; case EXPR_VARIABLE: if (arg->ref == NULL || (arg->ref->next == NULL && arg->ref->type == REF_ARRAY)) { /* This doesn't catch all cases. See http://gcc.gnu.org/ml/fortran/2004-06/msg00165.html and the surrounding thread. */ sym = arg->symtree->n.sym; decl = gfc_get_symbol_decl (sym); if (decl == current_function_decl && sym->attr.function && (sym->result == sym)) decl = gfc_get_fake_result_decl (sym, 0); len = sym->ts.u.cl->backend_decl; gcc_assert (len); break; } /* Otherwise fall through. */ default: /* Anybody stupid enough to do this deserves inefficient code. */ gfc_init_se (&argse, se); if (arg->rank == 0) gfc_conv_expr (&argse, arg); else gfc_conv_expr_descriptor (&argse, arg); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); len = argse.string_length; break; } se->expr = convert (type, len); } /* The length of a character string not including trailing blanks. */ static void gfc_conv_intrinsic_len_trim (gfc_se * se, gfc_expr * expr) { int kind = expr->value.function.actual->expr->ts.kind; tree args[2], type, fndecl; gfc_conv_intrinsic_function_args (se, expr, args, 2); type = gfc_typenode_for_spec (&expr->ts); if (kind == 1) fndecl = gfor_fndecl_string_len_trim; else if (kind == 4) fndecl = gfor_fndecl_string_len_trim_char4; else gcc_unreachable (); se->expr = build_call_expr_loc (input_location, fndecl, 2, args[0], args[1]); se->expr = convert (type, se->expr); } /* Returns the starting position of a substring within a string. */ static void gfc_conv_intrinsic_index_scan_verify (gfc_se * se, gfc_expr * expr, tree function) { tree logical4_type_node = gfc_get_logical_type (4); tree type; tree fndecl; tree *args; unsigned int num_args; args = XALLOCAVEC (tree, 5); /* Get number of arguments; characters count double due to the string length argument. Kind= is not passed to the library and thus ignored. */ if (expr->value.function.actual->next->next->expr == NULL) num_args = 4; else num_args = 5; gfc_conv_intrinsic_function_args (se, expr, args, num_args); type = gfc_typenode_for_spec (&expr->ts); if (num_args == 4) args[4] = build_int_cst (logical4_type_node, 0); else args[4] = convert (logical4_type_node, args[4]); fndecl = build_addr (function, current_function_decl); se->expr = build_call_array_loc (input_location, TREE_TYPE (TREE_TYPE (function)), fndecl, 5, args); se->expr = convert (type, se->expr); } /* The ascii value for a single character. */ static void gfc_conv_intrinsic_ichar (gfc_se * se, gfc_expr * expr) { tree args[2], type, pchartype; gfc_conv_intrinsic_function_args (se, expr, args, 2); gcc_assert (POINTER_TYPE_P (TREE_TYPE (args[1]))); pchartype = gfc_get_pchar_type (expr->value.function.actual->expr->ts.kind); args[1] = fold_build1_loc (input_location, NOP_EXPR, pchartype, args[1]); type = gfc_typenode_for_spec (&expr->ts); se->expr = build_fold_indirect_ref_loc (input_location, args[1]); se->expr = convert (type, se->expr); } /* Intrinsic ISNAN calls __builtin_isnan. */ static void gfc_conv_intrinsic_isnan (gfc_se * se, gfc_expr * expr) { tree arg; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); se->expr = build_call_expr_loc (input_location, builtin_decl_explicit (BUILT_IN_ISNAN), 1, arg); STRIP_TYPE_NOPS (se->expr); se->expr = fold_convert (gfc_typenode_for_spec (&expr->ts), se->expr); } /* Intrinsics IS_IOSTAT_END and IS_IOSTAT_EOR just need to compare their argument against a constant integer value. */ static void gfc_conv_has_intvalue (gfc_se * se, gfc_expr * expr, const int value) { tree arg; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); se->expr = fold_build2_loc (input_location, EQ_EXPR, gfc_typenode_for_spec (&expr->ts), arg, build_int_cst (TREE_TYPE (arg), value)); } /* MERGE (tsource, fsource, mask) = mask ? tsource : fsource. */ static void gfc_conv_intrinsic_merge (gfc_se * se, gfc_expr * expr) { tree tsource; tree fsource; tree mask; tree type; tree len, len2; tree *args; unsigned int num_args; num_args = gfc_intrinsic_argument_list_length (expr); args = XALLOCAVEC (tree, num_args); gfc_conv_intrinsic_function_args (se, expr, args, num_args); if (expr->ts.type != BT_CHARACTER) { tsource = args[0]; fsource = args[1]; mask = args[2]; } else { /* We do the same as in the non-character case, but the argument list is different because of the string length arguments. We also have to set the string length for the result. */ len = args[0]; tsource = args[1]; len2 = args[2]; fsource = args[3]; mask = args[4]; gfc_trans_same_strlen_check ("MERGE intrinsic", &expr->where, len, len2, &se->pre); se->string_length = len; } type = TREE_TYPE (tsource); se->expr = fold_build3_loc (input_location, COND_EXPR, type, mask, tsource, fold_convert (type, fsource)); } /* MERGE_BITS (I, J, MASK) = (I & MASK) | (I & (~MASK)). */ static void gfc_conv_intrinsic_merge_bits (gfc_se * se, gfc_expr * expr) { tree args[3], mask, type; gfc_conv_intrinsic_function_args (se, expr, args, 3); mask = gfc_evaluate_now (args[2], &se->pre); type = TREE_TYPE (args[0]); gcc_assert (TREE_TYPE (args[1]) == type); gcc_assert (TREE_TYPE (mask) == type); args[0] = fold_build2_loc (input_location, BIT_AND_EXPR, type, args[0], mask); args[1] = fold_build2_loc (input_location, BIT_AND_EXPR, type, args[1], fold_build1_loc (input_location, BIT_NOT_EXPR, type, mask)); se->expr = fold_build2_loc (input_location, BIT_IOR_EXPR, type, args[0], args[1]); } /* MASKL(n) = n == 0 ? 0 : (~0) << (BIT_SIZE - n) MASKR(n) = n == BIT_SIZE ? ~0 : ~((~0) << n) */ static void gfc_conv_intrinsic_mask (gfc_se * se, gfc_expr * expr, int left) { tree arg, allones, type, utype, res, cond, bitsize; int i; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); arg = gfc_evaluate_now (arg, &se->pre); type = gfc_get_int_type (expr->ts.kind); utype = unsigned_type_for (type); i = gfc_validate_kind (BT_INTEGER, expr->ts.kind, false); bitsize = build_int_cst (TREE_TYPE (arg), gfc_integer_kinds[i].bit_size); allones = fold_build1_loc (input_location, BIT_NOT_EXPR, utype, build_int_cst (utype, 0)); if (left) { /* Left-justified mask. */ res = fold_build2_loc (input_location, MINUS_EXPR, TREE_TYPE (arg), bitsize, arg); res = fold_build2_loc (input_location, LSHIFT_EXPR, utype, allones, fold_convert (utype, res)); /* Special case arg == 0, because SHIFT_EXPR wants a shift strictly smaller than type width. */ cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, arg, build_int_cst (TREE_TYPE (arg), 0)); res = fold_build3_loc (input_location, COND_EXPR, utype, cond, build_int_cst (utype, 0), res); } else { /* Right-justified mask. */ res = fold_build2_loc (input_location, LSHIFT_EXPR, utype, allones, fold_convert (utype, arg)); res = fold_build1_loc (input_location, BIT_NOT_EXPR, utype, res); /* Special case agr == bit_size, because SHIFT_EXPR wants a shift strictly smaller than type width. */ cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, arg, bitsize); res = fold_build3_loc (input_location, COND_EXPR, utype, cond, allones, res); } se->expr = fold_convert (type, res); } /* FRACTION (s) is translated into frexp (s, &dummy_int). */ static void gfc_conv_intrinsic_fraction (gfc_se * se, gfc_expr * expr) { tree arg, type, tmp, frexp; frexp = gfc_builtin_decl_for_float_kind (BUILT_IN_FREXP, expr->ts.kind); type = gfc_typenode_for_spec (&expr->ts); gfc_conv_intrinsic_function_args (se, expr, &arg, 1); tmp = gfc_create_var (integer_type_node, NULL); se->expr = build_call_expr_loc (input_location, frexp, 2, fold_convert (type, arg), gfc_build_addr_expr (NULL_TREE, tmp)); se->expr = fold_convert (type, se->expr); } /* NEAREST (s, dir) is translated into tmp = copysign (HUGE_VAL, dir); return nextafter (s, tmp); */ static void gfc_conv_intrinsic_nearest (gfc_se * se, gfc_expr * expr) { tree args[2], type, tmp, nextafter, copysign, huge_val; nextafter = gfc_builtin_decl_for_float_kind (BUILT_IN_NEXTAFTER, expr->ts.kind); copysign = gfc_builtin_decl_for_float_kind (BUILT_IN_COPYSIGN, expr->ts.kind); type = gfc_typenode_for_spec (&expr->ts); gfc_conv_intrinsic_function_args (se, expr, args, 2); huge_val = gfc_build_inf_or_huge (type, expr->ts.kind); tmp = build_call_expr_loc (input_location, copysign, 2, huge_val, fold_convert (type, args[1])); se->expr = build_call_expr_loc (input_location, nextafter, 2, fold_convert (type, args[0]), tmp); se->expr = fold_convert (type, se->expr); } /* SPACING (s) is translated into int e; if (s == 0) res = tiny; else { frexp (s, &e); e = e - prec; e = MAX_EXPR (e, emin); res = scalbn (1., e); } return res; where prec is the precision of s, gfc_real_kinds[k].digits, emin is min_exponent - 1, gfc_real_kinds[k].min_exponent - 1, and tiny is tiny(s), gfc_real_kinds[k].tiny. */ static void gfc_conv_intrinsic_spacing (gfc_se * se, gfc_expr * expr) { tree arg, type, prec, emin, tiny, res, e; tree cond, tmp, frexp, scalbn; int k; stmtblock_t block; k = gfc_validate_kind (BT_REAL, expr->ts.kind, false); prec = build_int_cst (integer_type_node, gfc_real_kinds[k].digits); emin = build_int_cst (integer_type_node, gfc_real_kinds[k].min_exponent - 1); tiny = gfc_conv_mpfr_to_tree (gfc_real_kinds[k].tiny, expr->ts.kind, 0); frexp = gfc_builtin_decl_for_float_kind (BUILT_IN_FREXP, expr->ts.kind); scalbn = gfc_builtin_decl_for_float_kind (BUILT_IN_SCALBN, expr->ts.kind); gfc_conv_intrinsic_function_args (se, expr, &arg, 1); arg = gfc_evaluate_now (arg, &se->pre); type = gfc_typenode_for_spec (&expr->ts); e = gfc_create_var (integer_type_node, NULL); res = gfc_create_var (type, NULL); /* Build the block for s /= 0. */ gfc_start_block (&block); tmp = build_call_expr_loc (input_location, frexp, 2, arg, gfc_build_addr_expr (NULL_TREE, e)); gfc_add_expr_to_block (&block, tmp); tmp = fold_build2_loc (input_location, MINUS_EXPR, integer_type_node, e, prec); gfc_add_modify (&block, e, fold_build2_loc (input_location, MAX_EXPR, integer_type_node, tmp, emin)); tmp = build_call_expr_loc (input_location, scalbn, 2, build_real_from_int_cst (type, integer_one_node), e); gfc_add_modify (&block, res, tmp); /* Finish by building the IF statement. */ cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, arg, build_real_from_int_cst (type, integer_zero_node)); tmp = build3_v (COND_EXPR, cond, build2_v (MODIFY_EXPR, res, tiny), gfc_finish_block (&block)); gfc_add_expr_to_block (&se->pre, tmp); se->expr = res; } /* RRSPACING (s) is translated into int e; real x; x = fabs (s); if (x != 0) { frexp (s, &e); x = scalbn (x, precision - e); } return x; where precision is gfc_real_kinds[k].digits. */ static void gfc_conv_intrinsic_rrspacing (gfc_se * se, gfc_expr * expr) { tree arg, type, e, x, cond, stmt, tmp, frexp, scalbn, fabs; int prec, k; stmtblock_t block; k = gfc_validate_kind (BT_REAL, expr->ts.kind, false); prec = gfc_real_kinds[k].digits; frexp = gfc_builtin_decl_for_float_kind (BUILT_IN_FREXP, expr->ts.kind); scalbn = gfc_builtin_decl_for_float_kind (BUILT_IN_SCALBN, expr->ts.kind); fabs = gfc_builtin_decl_for_float_kind (BUILT_IN_FABS, expr->ts.kind); type = gfc_typenode_for_spec (&expr->ts); gfc_conv_intrinsic_function_args (se, expr, &arg, 1); arg = gfc_evaluate_now (arg, &se->pre); e = gfc_create_var (integer_type_node, NULL); x = gfc_create_var (type, NULL); gfc_add_modify (&se->pre, x, build_call_expr_loc (input_location, fabs, 1, arg)); gfc_start_block (&block); tmp = build_call_expr_loc (input_location, frexp, 2, arg, gfc_build_addr_expr (NULL_TREE, e)); gfc_add_expr_to_block (&block, tmp); tmp = fold_build2_loc (input_location, MINUS_EXPR, integer_type_node, build_int_cst (integer_type_node, prec), e); tmp = build_call_expr_loc (input_location, scalbn, 2, x, tmp); gfc_add_modify (&block, x, tmp); stmt = gfc_finish_block (&block); cond = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, x, build_real_from_int_cst (type, integer_zero_node)); tmp = build3_v (COND_EXPR, cond, stmt, build_empty_stmt (input_location)); gfc_add_expr_to_block (&se->pre, tmp); se->expr = fold_convert (type, x); } /* SCALE (s, i) is translated into scalbn (s, i). */ static void gfc_conv_intrinsic_scale (gfc_se * se, gfc_expr * expr) { tree args[2], type, scalbn; scalbn = gfc_builtin_decl_for_float_kind (BUILT_IN_SCALBN, expr->ts.kind); type = gfc_typenode_for_spec (&expr->ts); gfc_conv_intrinsic_function_args (se, expr, args, 2); se->expr = build_call_expr_loc (input_location, scalbn, 2, fold_convert (type, args[0]), fold_convert (integer_type_node, args[1])); se->expr = fold_convert (type, se->expr); } /* SET_EXPONENT (s, i) is translated into scalbn (frexp (s, &dummy_int), i). */ static void gfc_conv_intrinsic_set_exponent (gfc_se * se, gfc_expr * expr) { tree args[2], type, tmp, frexp, scalbn; frexp = gfc_builtin_decl_for_float_kind (BUILT_IN_FREXP, expr->ts.kind); scalbn = gfc_builtin_decl_for_float_kind (BUILT_IN_SCALBN, expr->ts.kind); type = gfc_typenode_for_spec (&expr->ts); gfc_conv_intrinsic_function_args (se, expr, args, 2); tmp = gfc_create_var (integer_type_node, NULL); tmp = build_call_expr_loc (input_location, frexp, 2, fold_convert (type, args[0]), gfc_build_addr_expr (NULL_TREE, tmp)); se->expr = build_call_expr_loc (input_location, scalbn, 2, tmp, fold_convert (integer_type_node, args[1])); se->expr = fold_convert (type, se->expr); } static void gfc_conv_intrinsic_size (gfc_se * se, gfc_expr * expr) { gfc_actual_arglist *actual; tree arg1; tree type; tree fncall0; tree fncall1; gfc_se argse; gfc_init_se (&argse, NULL); actual = expr->value.function.actual; if (actual->expr->ts.type == BT_CLASS) gfc_add_class_array_ref (actual->expr); argse.want_pointer = 1; argse.data_not_needed = 1; gfc_conv_expr_descriptor (&argse, actual->expr); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); arg1 = gfc_evaluate_now (argse.expr, &se->pre); /* Build the call to size0. */ fncall0 = build_call_expr_loc (input_location, gfor_fndecl_size0, 1, arg1); actual = actual->next; if (actual->expr) { gfc_init_se (&argse, NULL); gfc_conv_expr_type (&argse, actual->expr, gfc_array_index_type); gfc_add_block_to_block (&se->pre, &argse.pre); /* Unusually, for an intrinsic, size does not exclude an optional arg2, so we must test for it. */ if (actual->expr->expr_type == EXPR_VARIABLE && actual->expr->symtree->n.sym->attr.dummy && actual->expr->symtree->n.sym->attr.optional) { tree tmp; /* Build the call to size1. */ fncall1 = build_call_expr_loc (input_location, gfor_fndecl_size1, 2, arg1, argse.expr); gfc_init_se (&argse, NULL); argse.want_pointer = 1; argse.data_not_needed = 1; gfc_conv_expr (&argse, actual->expr); gfc_add_block_to_block (&se->pre, &argse.pre); tmp = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, argse.expr, null_pointer_node); tmp = gfc_evaluate_now (tmp, &se->pre); se->expr = fold_build3_loc (input_location, COND_EXPR, pvoid_type_node, tmp, fncall1, fncall0); } else { se->expr = NULL_TREE; argse.expr = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, argse.expr, gfc_index_one_node); } } else if (expr->value.function.actual->expr->rank == 1) { argse.expr = gfc_index_zero_node; se->expr = NULL_TREE; } else se->expr = fncall0; if (se->expr == NULL_TREE) { tree ubound, lbound; arg1 = build_fold_indirect_ref_loc (input_location, arg1); ubound = gfc_conv_descriptor_ubound_get (arg1, argse.expr); lbound = gfc_conv_descriptor_lbound_get (arg1, argse.expr); se->expr = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, ubound, lbound); se->expr = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, se->expr, gfc_index_one_node); se->expr = fold_build2_loc (input_location, MAX_EXPR, gfc_array_index_type, se->expr, gfc_index_zero_node); } type = gfc_typenode_for_spec (&expr->ts); se->expr = convert (type, se->expr); } /* Helper function to compute the size of a character variable, excluding the terminating null characters. The result has gfc_array_index_type type. */ static tree size_of_string_in_bytes (int kind, tree string_length) { tree bytesize; int i = gfc_validate_kind (BT_CHARACTER, kind, false); bytesize = build_int_cst (gfc_array_index_type, gfc_character_kinds[i].bit_size / 8); return fold_build2_loc (input_location, MULT_EXPR, gfc_array_index_type, bytesize, fold_convert (gfc_array_index_type, string_length)); } static void gfc_conv_intrinsic_sizeof (gfc_se *se, gfc_expr *expr) { gfc_expr *arg; gfc_se argse; tree source_bytes; tree type; tree tmp; tree lower; tree upper; int n; arg = expr->value.function.actual->expr; gfc_init_se (&argse, NULL); if (arg->rank == 0) { if (arg->ts.type == BT_CLASS) gfc_add_data_component (arg); gfc_conv_expr_reference (&argse, arg); type = TREE_TYPE (build_fold_indirect_ref_loc (input_location, argse.expr)); /* Obtain the source word length. */ if (arg->ts.type == BT_CHARACTER) se->expr = size_of_string_in_bytes (arg->ts.kind, argse.string_length); else se->expr = fold_convert (gfc_array_index_type, size_in_bytes (type)); } else { source_bytes = gfc_create_var (gfc_array_index_type, "bytes"); argse.want_pointer = 0; gfc_conv_expr_descriptor (&argse, arg); type = gfc_get_element_type (TREE_TYPE (argse.expr)); /* Obtain the argument's word length. */ if (arg->ts.type == BT_CHARACTER) tmp = size_of_string_in_bytes (arg->ts.kind, argse.string_length); else tmp = fold_convert (gfc_array_index_type, size_in_bytes (type)); gfc_add_modify (&argse.pre, source_bytes, tmp); /* Obtain the size of the array in bytes. */ for (n = 0; n < arg->rank; n++) { tree idx; idx = gfc_rank_cst[n]; lower = gfc_conv_descriptor_lbound_get (argse.expr, idx); upper = gfc_conv_descriptor_ubound_get (argse.expr, idx); tmp = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, upper, lower); tmp = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, tmp, gfc_index_one_node); tmp = fold_build2_loc (input_location, MULT_EXPR, gfc_array_index_type, tmp, source_bytes); gfc_add_modify (&argse.pre, source_bytes, tmp); } se->expr = source_bytes; } gfc_add_block_to_block (&se->pre, &argse.pre); } static void gfc_conv_intrinsic_storage_size (gfc_se *se, gfc_expr *expr) { gfc_expr *arg; gfc_se argse,eight; tree type, result_type, tmp; arg = expr->value.function.actual->expr; gfc_init_se (&eight, NULL); gfc_conv_expr (&eight, gfc_get_int_expr (expr->ts.kind, NULL, 8)); gfc_init_se (&argse, NULL); result_type = gfc_get_int_type (expr->ts.kind); if (arg->rank == 0) { if (arg->ts.type == BT_CLASS) { gfc_add_vptr_component (arg); gfc_add_size_component (arg); gfc_conv_expr (&argse, arg); tmp = fold_convert (result_type, argse.expr); goto done; } gfc_conv_expr_reference (&argse, arg); type = TREE_TYPE (build_fold_indirect_ref_loc (input_location, argse.expr)); } else { argse.want_pointer = 0; gfc_conv_expr_descriptor (&argse, arg); type = gfc_get_element_type (TREE_TYPE (argse.expr)); } /* Obtain the argument's word length. */ if (arg->ts.type == BT_CHARACTER) tmp = size_of_string_in_bytes (arg->ts.kind, argse.string_length); else tmp = fold_convert (result_type, size_in_bytes (type)); done: se->expr = fold_build2_loc (input_location, MULT_EXPR, result_type, tmp, eight.expr); gfc_add_block_to_block (&se->pre, &argse.pre); } /* Intrinsic string comparison functions. */ static void gfc_conv_intrinsic_strcmp (gfc_se * se, gfc_expr * expr, enum tree_code op) { tree args[4]; gfc_conv_intrinsic_function_args (se, expr, args, 4); se->expr = gfc_build_compare_string (args[0], args[1], args[2], args[3], expr->value.function.actual->expr->ts.kind, op); se->expr = fold_build2_loc (input_location, op, gfc_typenode_for_spec (&expr->ts), se->expr, build_int_cst (TREE_TYPE (se->expr), 0)); } /* Generate a call to the adjustl/adjustr library function. */ static void gfc_conv_intrinsic_adjust (gfc_se * se, gfc_expr * expr, tree fndecl) { tree args[3]; tree len; tree type; tree var; tree tmp; gfc_conv_intrinsic_function_args (se, expr, &args[1], 2); len = args[1]; type = TREE_TYPE (args[2]); var = gfc_conv_string_tmp (se, type, len); args[0] = var; tmp = build_call_expr_loc (input_location, fndecl, 3, args[0], args[1], args[2]); gfc_add_expr_to_block (&se->pre, tmp); se->expr = var; se->string_length = len; } /* Generate code for the TRANSFER intrinsic: For scalar results: DEST = TRANSFER (SOURCE, MOLD) where: typeof = typeof and: MOLD is scalar. For array results: DEST(1:N) = TRANSFER (SOURCE, MOLD[, SIZE]) where: typeof = typeof and: N = min (sizeof (SOURCE(:)), sizeof (DEST(:)), sizeof (DEST(0) * SIZE). */ static void gfc_conv_intrinsic_transfer (gfc_se * se, gfc_expr * expr) { tree tmp; tree tmpdecl; tree ptr; tree extent; tree source; tree source_type; tree source_bytes; tree mold_type; tree dest_word_len; tree size_words; tree size_bytes; tree upper; tree lower; tree stmt; gfc_actual_arglist *arg; gfc_se argse; gfc_array_info *info; stmtblock_t block; int n; bool scalar_mold; info = NULL; if (se->loop) info = &se->ss->info->data.array; /* Convert SOURCE. The output from this stage is:- source_bytes = length of the source in bytes source = pointer to the source data. */ arg = expr->value.function.actual; /* Ensure double transfer through LOGICAL preserves all the needed bits. */ if (arg->expr->expr_type == EXPR_FUNCTION && arg->expr->value.function.esym == NULL && arg->expr->value.function.isym != NULL && arg->expr->value.function.isym->id == GFC_ISYM_TRANSFER && arg->expr->ts.type == BT_LOGICAL && expr->ts.type != arg->expr->ts.type) arg->expr->value.function.name = "__transfer_in_transfer"; gfc_init_se (&argse, NULL); source_bytes = gfc_create_var (gfc_array_index_type, NULL); /* Obtain the pointer to source and the length of source in bytes. */ if (arg->expr->rank == 0) { gfc_conv_expr_reference (&argse, arg->expr); source = argse.expr; source_type = TREE_TYPE (build_fold_indirect_ref_loc (input_location, argse.expr)); /* Obtain the source word length. */ if (arg->expr->ts.type == BT_CHARACTER) tmp = size_of_string_in_bytes (arg->expr->ts.kind, argse.string_length); else tmp = fold_convert (gfc_array_index_type, size_in_bytes (source_type)); } else { argse.want_pointer = 0; gfc_conv_expr_descriptor (&argse, arg->expr); source = gfc_conv_descriptor_data_get (argse.expr); source_type = gfc_get_element_type (TREE_TYPE (argse.expr)); /* Repack the source if not a full variable array. */ if (arg->expr->expr_type == EXPR_VARIABLE && arg->expr->ref->u.ar.type != AR_FULL) { tmp = gfc_build_addr_expr (NULL_TREE, argse.expr); if (gfc_option.warn_array_temp) gfc_warning ("Creating array temporary at %L", &expr->where); source = build_call_expr_loc (input_location, gfor_fndecl_in_pack, 1, tmp); source = gfc_evaluate_now (source, &argse.pre); /* Free the temporary. */ gfc_start_block (&block); tmp = gfc_call_free (convert (pvoid_type_node, source)); gfc_add_expr_to_block (&block, tmp); stmt = gfc_finish_block (&block); /* Clean up if it was repacked. */ gfc_init_block (&block); tmp = gfc_conv_array_data (argse.expr); tmp = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, source, tmp); tmp = build3_v (COND_EXPR, tmp, stmt, build_empty_stmt (input_location)); gfc_add_expr_to_block (&block, tmp); gfc_add_block_to_block (&block, &se->post); gfc_init_block (&se->post); gfc_add_block_to_block (&se->post, &block); } /* Obtain the source word length. */ if (arg->expr->ts.type == BT_CHARACTER) tmp = size_of_string_in_bytes (arg->expr->ts.kind, argse.string_length); else tmp = fold_convert (gfc_array_index_type, size_in_bytes (source_type)); /* Obtain the size of the array in bytes. */ extent = gfc_create_var (gfc_array_index_type, NULL); for (n = 0; n < arg->expr->rank; n++) { tree idx; idx = gfc_rank_cst[n]; gfc_add_modify (&argse.pre, source_bytes, tmp); lower = gfc_conv_descriptor_lbound_get (argse.expr, idx); upper = gfc_conv_descriptor_ubound_get (argse.expr, idx); tmp = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, upper, lower); gfc_add_modify (&argse.pre, extent, tmp); tmp = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, extent, gfc_index_one_node); tmp = fold_build2_loc (input_location, MULT_EXPR, gfc_array_index_type, tmp, source_bytes); } } gfc_add_modify (&argse.pre, source_bytes, tmp); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); /* Now convert MOLD. The outputs are: mold_type = the TREE type of MOLD dest_word_len = destination word length in bytes. */ arg = arg->next; gfc_init_se (&argse, NULL); scalar_mold = arg->expr->rank == 0; if (arg->expr->rank == 0) { gfc_conv_expr_reference (&argse, arg->expr); mold_type = TREE_TYPE (build_fold_indirect_ref_loc (input_location, argse.expr)); } else { gfc_init_se (&argse, NULL); argse.want_pointer = 0; gfc_conv_expr_descriptor (&argse, arg->expr); mold_type = gfc_get_element_type (TREE_TYPE (argse.expr)); } gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); if (strcmp (expr->value.function.name, "__transfer_in_transfer") == 0) { /* If this TRANSFER is nested in another TRANSFER, use a type that preserves all bits. */ if (arg->expr->ts.type == BT_LOGICAL) mold_type = gfc_get_int_type (arg->expr->ts.kind); } if (arg->expr->ts.type == BT_CHARACTER) { tmp = size_of_string_in_bytes (arg->expr->ts.kind, argse.string_length); mold_type = gfc_get_character_type_len (arg->expr->ts.kind, tmp); } else tmp = fold_convert (gfc_array_index_type, size_in_bytes (mold_type)); dest_word_len = gfc_create_var (gfc_array_index_type, NULL); gfc_add_modify (&se->pre, dest_word_len, tmp); /* Finally convert SIZE, if it is present. */ arg = arg->next; size_words = gfc_create_var (gfc_array_index_type, NULL); if (arg->expr) { gfc_init_se (&argse, NULL); gfc_conv_expr_reference (&argse, arg->expr); tmp = convert (gfc_array_index_type, build_fold_indirect_ref_loc (input_location, argse.expr)); gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); } else tmp = NULL_TREE; /* Separate array and scalar results. */ if (scalar_mold && tmp == NULL_TREE) goto scalar_transfer; size_bytes = gfc_create_var (gfc_array_index_type, NULL); if (tmp != NULL_TREE) tmp = fold_build2_loc (input_location, MULT_EXPR, gfc_array_index_type, tmp, dest_word_len); else tmp = source_bytes; gfc_add_modify (&se->pre, size_bytes, tmp); gfc_add_modify (&se->pre, size_words, fold_build2_loc (input_location, CEIL_DIV_EXPR, gfc_array_index_type, size_bytes, dest_word_len)); /* Evaluate the bounds of the result. If the loop range exists, we have to check if it is too large. If so, we modify loop->to be consistent with min(size, size(source)). Otherwise, size is made consistent with the loop range, so that the right number of bytes is transferred.*/ n = se->loop->order[0]; if (se->loop->to[n] != NULL_TREE) { tmp = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, se->loop->to[n], se->loop->from[n]); tmp = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, tmp, gfc_index_one_node); tmp = fold_build2_loc (input_location, MIN_EXPR, gfc_array_index_type, tmp, size_words); gfc_add_modify (&se->pre, size_words, tmp); gfc_add_modify (&se->pre, size_bytes, fold_build2_loc (input_location, MULT_EXPR, gfc_array_index_type, size_words, dest_word_len)); upper = fold_build2_loc (input_location, PLUS_EXPR, gfc_array_index_type, size_words, se->loop->from[n]); upper = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, upper, gfc_index_one_node); } else { upper = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, size_words, gfc_index_one_node); se->loop->from[n] = gfc_index_zero_node; } se->loop->to[n] = upper; /* Build a destination descriptor, using the pointer, source, as the data field. */ gfc_trans_create_temp_array (&se->pre, &se->post, se->ss, mold_type, NULL_TREE, false, true, false, &expr->where); /* Cast the pointer to the result. */ tmp = gfc_conv_descriptor_data_get (info->descriptor); tmp = fold_convert (pvoid_type_node, tmp); /* Use memcpy to do the transfer. */ tmp = build_call_expr_loc (input_location, builtin_decl_explicit (BUILT_IN_MEMCPY), 3, tmp, fold_convert (pvoid_type_node, source), fold_build2_loc (input_location, MIN_EXPR, gfc_array_index_type, size_bytes, source_bytes)); gfc_add_expr_to_block (&se->pre, tmp); se->expr = info->descriptor; if (expr->ts.type == BT_CHARACTER) se->string_length = fold_convert (gfc_charlen_type_node, dest_word_len); return; /* Deal with scalar results. */ scalar_transfer: extent = fold_build2_loc (input_location, MIN_EXPR, gfc_array_index_type, dest_word_len, source_bytes); extent = fold_build2_loc (input_location, MAX_EXPR, gfc_array_index_type, extent, gfc_index_zero_node); if (expr->ts.type == BT_CHARACTER) { tree direct; tree indirect; ptr = convert (gfc_get_pchar_type (expr->ts.kind), source); tmpdecl = gfc_create_var (gfc_get_pchar_type (expr->ts.kind), "transfer"); /* If source is longer than the destination, use a pointer to the source directly. */ gfc_init_block (&block); gfc_add_modify (&block, tmpdecl, ptr); direct = gfc_finish_block (&block); /* Otherwise, allocate a string with the length of the destination and copy the source into it. */ gfc_init_block (&block); tmp = gfc_get_pchar_type (expr->ts.kind); tmp = gfc_call_malloc (&block, tmp, dest_word_len); gfc_add_modify (&block, tmpdecl, fold_convert (TREE_TYPE (ptr), tmp)); tmp = build_call_expr_loc (input_location, builtin_decl_explicit (BUILT_IN_MEMCPY), 3, fold_convert (pvoid_type_node, tmpdecl), fold_convert (pvoid_type_node, ptr), extent); gfc_add_expr_to_block (&block, tmp); indirect = gfc_finish_block (&block); /* Wrap it up with the condition. */ tmp = fold_build2_loc (input_location, LE_EXPR, boolean_type_node, dest_word_len, source_bytes); tmp = build3_v (COND_EXPR, tmp, direct, indirect); gfc_add_expr_to_block (&se->pre, tmp); se->expr = tmpdecl; se->string_length = dest_word_len; } else { tmpdecl = gfc_create_var (mold_type, "transfer"); ptr = convert (build_pointer_type (mold_type), source); /* Use memcpy to do the transfer. */ tmp = gfc_build_addr_expr (NULL_TREE, tmpdecl); tmp = build_call_expr_loc (input_location, builtin_decl_explicit (BUILT_IN_MEMCPY), 3, fold_convert (pvoid_type_node, tmp), fold_convert (pvoid_type_node, ptr), extent); gfc_add_expr_to_block (&se->pre, tmp); se->expr = tmpdecl; } } /* Generate code for the ALLOCATED intrinsic. Generate inline code that directly check the address of the argument. */ static void gfc_conv_allocated (gfc_se *se, gfc_expr *expr) { gfc_actual_arglist *arg1; gfc_se arg1se; tree tmp; gfc_init_se (&arg1se, NULL); arg1 = expr->value.function.actual; if (arg1->expr->ts.type == BT_CLASS) { /* Make sure that class array expressions have both a _data component reference and an array reference.... */ if (CLASS_DATA (arg1->expr)->attr.dimension) gfc_add_class_array_ref (arg1->expr); /* .... whilst scalars only need the _data component. */ else gfc_add_data_component (arg1->expr); } if (arg1->expr->rank == 0) { /* Allocatable scalar. */ arg1se.want_pointer = 1; gfc_conv_expr (&arg1se, arg1->expr); tmp = arg1se.expr; } else { /* Allocatable array. */ arg1se.descriptor_only = 1; gfc_conv_expr_descriptor (&arg1se, arg1->expr); tmp = gfc_conv_descriptor_data_get (arg1se.expr); } tmp = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, tmp, fold_convert (TREE_TYPE (tmp), null_pointer_node)); se->expr = convert (gfc_typenode_for_spec (&expr->ts), tmp); } /* Generate code for the ASSOCIATED intrinsic. If both POINTER and TARGET are arrays, generate a call to library function _gfor_associated, and pass descriptors of POINTER and TARGET to it. In other cases, generate inline code that directly compare the address of POINTER with the address of TARGET. */ static void gfc_conv_associated (gfc_se *se, gfc_expr *expr) { gfc_actual_arglist *arg1; gfc_actual_arglist *arg2; gfc_se arg1se; gfc_se arg2se; tree tmp2; tree tmp; tree nonzero_charlen; tree nonzero_arraylen; gfc_ss *ss; bool scalar; gfc_init_se (&arg1se, NULL); gfc_init_se (&arg2se, NULL); arg1 = expr->value.function.actual; if (arg1->expr->ts.type == BT_CLASS) gfc_add_data_component (arg1->expr); arg2 = arg1->next; /* Check whether the expression is a scalar or not; we cannot use arg1->expr->rank as it can be nonzero for proc pointers. */ ss = gfc_walk_expr (arg1->expr); scalar = ss == gfc_ss_terminator; if (!scalar) gfc_free_ss_chain (ss); if (!arg2->expr) { /* No optional target. */ if (scalar) { /* A pointer to a scalar. */ arg1se.want_pointer = 1; gfc_conv_expr (&arg1se, arg1->expr); if (arg1->expr->symtree->n.sym->attr.proc_pointer && arg1->expr->symtree->n.sym->attr.dummy) arg1se.expr = build_fold_indirect_ref_loc (input_location, arg1se.expr); tmp2 = arg1se.expr; } else { /* A pointer to an array. */ gfc_conv_expr_descriptor (&arg1se, arg1->expr); tmp2 = gfc_conv_descriptor_data_get (arg1se.expr); } gfc_add_block_to_block (&se->pre, &arg1se.pre); gfc_add_block_to_block (&se->post, &arg1se.post); tmp = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, tmp2, fold_convert (TREE_TYPE (tmp2), null_pointer_node)); se->expr = tmp; } else { /* An optional target. */ if (arg2->expr->ts.type == BT_CLASS) gfc_add_data_component (arg2->expr); nonzero_charlen = NULL_TREE; if (arg1->expr->ts.type == BT_CHARACTER) nonzero_charlen = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, arg1->expr->ts.u.cl->backend_decl, integer_zero_node); if (scalar) { /* A pointer to a scalar. */ arg1se.want_pointer = 1; gfc_conv_expr (&arg1se, arg1->expr); if (arg1->expr->symtree->n.sym->attr.proc_pointer && arg1->expr->symtree->n.sym->attr.dummy) arg1se.expr = build_fold_indirect_ref_loc (input_location, arg1se.expr); arg2se.want_pointer = 1; gfc_conv_expr (&arg2se, arg2->expr); if (arg2->expr->symtree->n.sym->attr.proc_pointer && arg2->expr->symtree->n.sym->attr.dummy) arg2se.expr = build_fold_indirect_ref_loc (input_location, arg2se.expr); gfc_add_block_to_block (&se->pre, &arg1se.pre); gfc_add_block_to_block (&se->post, &arg1se.post); tmp = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, arg1se.expr, arg2se.expr); tmp2 = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, arg1se.expr, null_pointer_node); se->expr = fold_build2_loc (input_location, TRUTH_AND_EXPR, boolean_type_node, tmp, tmp2); } else { /* An array pointer of zero length is not associated if target is present. */ arg1se.descriptor_only = 1; gfc_conv_expr_lhs (&arg1se, arg1->expr); if (arg1->expr->rank == -1) { tmp = gfc_conv_descriptor_rank (arg1se.expr); tmp = fold_build2_loc (input_location, MINUS_EXPR, TREE_TYPE (tmp), tmp, gfc_index_one_node); } else tmp = gfc_rank_cst[arg1->expr->rank - 1]; tmp = gfc_conv_descriptor_stride_get (arg1se.expr, tmp); nonzero_arraylen = fold_build2_loc (input_location, NE_EXPR, boolean_type_node, tmp, build_int_cst (TREE_TYPE (tmp), 0)); /* A pointer to an array, call library function _gfor_associated. */ arg1se.want_pointer = 1; gfc_conv_expr_descriptor (&arg1se, arg1->expr); arg2se.want_pointer = 1; gfc_conv_expr_descriptor (&arg2se, arg2->expr); gfc_add_block_to_block (&se->pre, &arg2se.pre); gfc_add_block_to_block (&se->post, &arg2se.post); se->expr = build_call_expr_loc (input_location, gfor_fndecl_associated, 2, arg1se.expr, arg2se.expr); se->expr = convert (boolean_type_node, se->expr); se->expr = fold_build2_loc (input_location, TRUTH_AND_EXPR, boolean_type_node, se->expr, nonzero_arraylen); } /* If target is present zero character length pointers cannot be associated. */ if (nonzero_charlen != NULL_TREE) se->expr = fold_build2_loc (input_location, TRUTH_AND_EXPR, boolean_type_node, se->expr, nonzero_charlen); } se->expr = convert (gfc_typenode_for_spec (&expr->ts), se->expr); } /* Generate code for the SAME_TYPE_AS intrinsic. Generate inline code that directly checks the vindices. */ static void gfc_conv_same_type_as (gfc_se *se, gfc_expr *expr) { gfc_expr *a, *b; gfc_se se1, se2; tree tmp; gfc_init_se (&se1, NULL); gfc_init_se (&se2, NULL); a = expr->value.function.actual->expr; b = expr->value.function.actual->next->expr; if (a->ts.type == BT_CLASS) { gfc_add_vptr_component (a); gfc_add_hash_component (a); } else if (a->ts.type == BT_DERIVED) a = gfc_get_int_expr (gfc_default_integer_kind, NULL, a->ts.u.derived->hash_value); if (b->ts.type == BT_CLASS) { gfc_add_vptr_component (b); gfc_add_hash_component (b); } else if (b->ts.type == BT_DERIVED) b = gfc_get_int_expr (gfc_default_integer_kind, NULL, b->ts.u.derived->hash_value); gfc_conv_expr (&se1, a); gfc_conv_expr (&se2, b); tmp = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, se1.expr, fold_convert (TREE_TYPE (se1.expr), se2.expr)); se->expr = convert (gfc_typenode_for_spec (&expr->ts), tmp); } /* Generate code for SELECTED_CHAR_KIND (NAME) intrinsic function. */ static void gfc_conv_intrinsic_sc_kind (gfc_se *se, gfc_expr *expr) { tree args[2]; gfc_conv_intrinsic_function_args (se, expr, args, 2); se->expr = build_call_expr_loc (input_location, gfor_fndecl_sc_kind, 2, args[0], args[1]); se->expr = fold_convert (gfc_typenode_for_spec (&expr->ts), se->expr); } /* Generate code for SELECTED_INT_KIND (R) intrinsic function. */ static void gfc_conv_intrinsic_si_kind (gfc_se *se, gfc_expr *expr) { tree arg, type; gfc_conv_intrinsic_function_args (se, expr, &arg, 1); /* The argument to SELECTED_INT_KIND is INTEGER(4). */ type = gfc_get_int_type (4); arg = gfc_build_addr_expr (NULL_TREE, fold_convert (type, arg)); /* Convert it to the required type. */ type = gfc_typenode_for_spec (&expr->ts); se->expr = build_call_expr_loc (input_location, gfor_fndecl_si_kind, 1, arg); se->expr = fold_convert (type, se->expr); } /* Generate code for SELECTED_REAL_KIND (P, R, RADIX) intrinsic function. */ static void gfc_conv_intrinsic_sr_kind (gfc_se *se, gfc_expr *expr) { gfc_actual_arglist *actual; tree type; gfc_se argse; VEC(tree,gc) *args = NULL; for (actual = expr->value.function.actual; actual; actual = actual->next) { gfc_init_se (&argse, se); /* Pass a NULL pointer for an absent arg. */ if (actual->expr == NULL) argse.expr = null_pointer_node; else { gfc_typespec ts; gfc_clear_ts (&ts); if (actual->expr->ts.kind != gfc_c_int_kind) { /* The arguments to SELECTED_REAL_KIND are INTEGER(4). */ ts.type = BT_INTEGER; ts.kind = gfc_c_int_kind; gfc_convert_type (actual->expr, &ts, 2); } gfc_conv_expr_reference (&argse, actual->expr); } gfc_add_block_to_block (&se->pre, &argse.pre); gfc_add_block_to_block (&se->post, &argse.post); VEC_safe_push (tree, gc, args, argse.expr); } /* Convert it to the required type. */ type = gfc_typenode_for_spec (&expr->ts); se->expr = build_call_expr_loc_vec (input_location, gfor_fndecl_sr_kind, args); se->expr = fold_convert (type, se->expr); } /* Generate code for TRIM (A) intrinsic function. */ static void gfc_conv_intrinsic_trim (gfc_se * se, gfc_expr * expr) { tree var; tree len; tree addr; tree tmp; tree cond; tree fndecl; tree function; tree *args; unsigned int num_args; num_args = gfc_intrinsic_argument_list_length (expr) + 2; args = XALLOCAVEC (tree, num_args); var = gfc_create_var (gfc_get_pchar_type (expr->ts.kind), "pstr"); addr = gfc_build_addr_expr (ppvoid_type_node, var); len = gfc_create_var (gfc_charlen_type_node, "len"); gfc_conv_intrinsic_function_args (se, expr, &args[2], num_args - 2); args[0] = gfc_build_addr_expr (NULL_TREE, len); args[1] = addr; if (expr->ts.kind == 1) function = gfor_fndecl_string_trim; else if (expr->ts.kind == 4) function = gfor_fndecl_string_trim_char4; else gcc_unreachable (); fndecl = build_addr (function, current_function_decl); tmp = build_call_array_loc (input_location, TREE_TYPE (TREE_TYPE (function)), fndecl, num_args, args); gfc_add_expr_to_block (&se->pre, tmp); /* Free the temporary afterwards, if necessary. */ cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, len, build_int_cst (TREE_TYPE (len), 0)); tmp = gfc_call_free (var); tmp = build3_v (COND_EXPR, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&se->post, tmp); se->expr = var; se->string_length = len; } /* Generate code for REPEAT (STRING, NCOPIES) intrinsic function. */ static void gfc_conv_intrinsic_repeat (gfc_se * se, gfc_expr * expr) { tree args[3], ncopies, dest, dlen, src, slen, ncopies_type; tree type, cond, tmp, count, exit_label, n, max, largest; tree size; stmtblock_t block, body; int i; /* We store in charsize the size of a character. */ i = gfc_validate_kind (BT_CHARACTER, expr->ts.kind, false); size = build_int_cst (size_type_node, gfc_character_kinds[i].bit_size / 8); /* Get the arguments. */ gfc_conv_intrinsic_function_args (se, expr, args, 3); slen = fold_convert (size_type_node, gfc_evaluate_now (args[0], &se->pre)); src = args[1]; ncopies = gfc_evaluate_now (args[2], &se->pre); ncopies_type = TREE_TYPE (ncopies); /* Check that NCOPIES is not negative. */ cond = fold_build2_loc (input_location, LT_EXPR, boolean_type_node, ncopies, build_int_cst (ncopies_type, 0)); gfc_trans_runtime_check (true, false, cond, &se->pre, &expr->where, "Argument NCOPIES of REPEAT intrinsic is negative " "(its value is %ld)", fold_convert (long_integer_type_node, ncopies)); /* If the source length is zero, any non negative value of NCOPIES is valid, and nothing happens. */ n = gfc_create_var (ncopies_type, "ncopies"); cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, slen, build_int_cst (size_type_node, 0)); tmp = fold_build3_loc (input_location, COND_EXPR, ncopies_type, cond, build_int_cst (ncopies_type, 0), ncopies); gfc_add_modify (&se->pre, n, tmp); ncopies = n; /* Check that ncopies is not too large: ncopies should be less than (or equal to) MAX / slen, where MAX is the maximal integer of the gfc_charlen_type_node type. If slen == 0, we need a special case to avoid the division by zero. */ i = gfc_validate_kind (BT_INTEGER, gfc_charlen_int_kind, false); max = gfc_conv_mpz_to_tree (gfc_integer_kinds[i].huge, gfc_charlen_int_kind); max = fold_build2_loc (input_location, TRUNC_DIV_EXPR, size_type_node, fold_convert (size_type_node, max), slen); largest = TYPE_PRECISION (size_type_node) > TYPE_PRECISION (ncopies_type) ? size_type_node : ncopies_type; cond = fold_build2_loc (input_location, GT_EXPR, boolean_type_node, fold_convert (largest, ncopies), fold_convert (largest, max)); tmp = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, slen, build_int_cst (size_type_node, 0)); cond = fold_build3_loc (input_location, COND_EXPR, boolean_type_node, tmp, boolean_false_node, cond); gfc_trans_runtime_check (true, false, cond, &se->pre, &expr->where, "Argument NCOPIES of REPEAT intrinsic is too large"); /* Compute the destination length. */ dlen = fold_build2_loc (input_location, MULT_EXPR, gfc_charlen_type_node, fold_convert (gfc_charlen_type_node, slen), fold_convert (gfc_charlen_type_node, ncopies)); type = gfc_get_character_type (expr->ts.kind, expr->ts.u.cl); dest = gfc_conv_string_tmp (se, build_pointer_type (type), dlen); /* Generate the code to do the repeat operation: for (i = 0; i < ncopies; i++) memmove (dest + (i * slen * size), src, slen*size); */ gfc_start_block (&block); count = gfc_create_var (ncopies_type, "count"); gfc_add_modify (&block, count, build_int_cst (ncopies_type, 0)); exit_label = gfc_build_label_decl (NULL_TREE); /* Start the loop body. */ gfc_start_block (&body); /* Exit the loop if count >= ncopies. */ cond = fold_build2_loc (input_location, GE_EXPR, boolean_type_node, count, ncopies); tmp = build1_v (GOTO_EXPR, exit_label); TREE_USED (exit_label) = 1; tmp = fold_build3_loc (input_location, COND_EXPR, void_type_node, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&body, tmp); /* Call memmove (dest + (i*slen*size), src, slen*size). */ tmp = fold_build2_loc (input_location, MULT_EXPR, gfc_charlen_type_node, fold_convert (gfc_charlen_type_node, slen), fold_convert (gfc_charlen_type_node, count)); tmp = fold_build2_loc (input_location, MULT_EXPR, gfc_charlen_type_node, tmp, fold_convert (gfc_charlen_type_node, size)); tmp = fold_build_pointer_plus_loc (input_location, fold_convert (pvoid_type_node, dest), tmp); tmp = build_call_expr_loc (input_location, builtin_decl_explicit (BUILT_IN_MEMMOVE), 3, tmp, src, fold_build2_loc (input_location, MULT_EXPR, size_type_node, slen, fold_convert (size_type_node, size))); gfc_add_expr_to_block (&body, tmp); /* Increment count. */ tmp = fold_build2_loc (input_location, PLUS_EXPR, ncopies_type, count, build_int_cst (TREE_TYPE (count), 1)); gfc_add_modify (&body, count, tmp); /* Build the loop. */ tmp = build1_v (LOOP_EXPR, gfc_finish_block (&body)); gfc_add_expr_to_block (&block, tmp); /* Add the exit label. */ tmp = build1_v (LABEL_EXPR, exit_label); gfc_add_expr_to_block (&block, tmp); /* Finish the block. */ tmp = gfc_finish_block (&block); gfc_add_expr_to_block (&se->pre, tmp); /* Set the result value. */ se->expr = dest; se->string_length = dlen; } /* Generate code for the IARGC intrinsic. */ static void gfc_conv_intrinsic_iargc (gfc_se * se, gfc_expr * expr) { tree tmp; tree fndecl; tree type; /* Call the library function. This always returns an INTEGER(4). */ fndecl = gfor_fndecl_iargc; tmp = build_call_expr_loc (input_location, fndecl, 0); /* Convert it to the required type. */ type = gfc_typenode_for_spec (&expr->ts); tmp = fold_convert (type, tmp); se->expr = tmp; } /* The loc intrinsic returns the address of its argument as gfc_index_integer_kind integer. */ static void gfc_conv_intrinsic_loc (gfc_se * se, gfc_expr * expr) { tree temp_var; gfc_expr *arg_expr; gcc_assert (!se->ss); arg_expr = expr->value.function.actual->expr; if (arg_expr->rank == 0) gfc_conv_expr_reference (se, arg_expr); else gfc_conv_array_parameter (se, arg_expr, true, NULL, NULL, NULL); se->expr= convert (gfc_get_int_type (gfc_index_integer_kind), se->expr); /* Create a temporary variable for loc return value. Without this, we get an error an ICE in gcc/expr.c(expand_expr_addr_expr_1). */ temp_var = gfc_create_var (gfc_get_int_type (gfc_index_integer_kind), NULL); gfc_add_modify (&se->pre, temp_var, se->expr); se->expr = temp_var; } /* Generate code for an intrinsic function. Some map directly to library calls, others get special handling. In some cases the name of the function used depends on the type specifiers. */ void gfc_conv_intrinsic_function (gfc_se * se, gfc_expr * expr) { const char *name; int lib, kind; tree fndecl; name = &expr->value.function.name[2]; if (expr->rank > 0) { lib = gfc_is_intrinsic_libcall (expr); if (lib != 0) { if (lib == 1) se->ignore_optional = 1; switch (expr->value.function.isym->id) { case GFC_ISYM_EOSHIFT: case GFC_ISYM_PACK: case GFC_ISYM_RESHAPE: /* For all of those the first argument specifies the type and the third is optional. */ conv_generic_with_optional_char_arg (se, expr, 1, 3); break; default: gfc_conv_intrinsic_funcall (se, expr); break; } return; } } switch (expr->value.function.isym->id) { case GFC_ISYM_NONE: gcc_unreachable (); case GFC_ISYM_REPEAT: gfc_conv_intrinsic_repeat (se, expr); break; case GFC_ISYM_TRIM: gfc_conv_intrinsic_trim (se, expr); break; case GFC_ISYM_SC_KIND: gfc_conv_intrinsic_sc_kind (se, expr); break; case GFC_ISYM_SI_KIND: gfc_conv_intrinsic_si_kind (se, expr); break; case GFC_ISYM_SR_KIND: gfc_conv_intrinsic_sr_kind (se, expr); break; case GFC_ISYM_EXPONENT: gfc_conv_intrinsic_exponent (se, expr); break; case GFC_ISYM_SCAN: kind = expr->value.function.actual->expr->ts.kind; if (kind == 1) fndecl = gfor_fndecl_string_scan; else if (kind == 4) fndecl = gfor_fndecl_string_scan_char4; else gcc_unreachable (); gfc_conv_intrinsic_index_scan_verify (se, expr, fndecl); break; case GFC_ISYM_VERIFY: kind = expr->value.function.actual->expr->ts.kind; if (kind == 1) fndecl = gfor_fndecl_string_verify; else if (kind == 4) fndecl = gfor_fndecl_string_verify_char4; else gcc_unreachable (); gfc_conv_intrinsic_index_scan_verify (se, expr, fndecl); break; case GFC_ISYM_ALLOCATED: gfc_conv_allocated (se, expr); break; case GFC_ISYM_ASSOCIATED: gfc_conv_associated(se, expr); break; case GFC_ISYM_SAME_TYPE_AS: gfc_conv_same_type_as (se, expr); break; case GFC_ISYM_ABS: gfc_conv_intrinsic_abs (se, expr); break; case GFC_ISYM_ADJUSTL: if (expr->ts.kind == 1) fndecl = gfor_fndecl_adjustl; else if (expr->ts.kind == 4) fndecl = gfor_fndecl_adjustl_char4; else gcc_unreachable (); gfc_conv_intrinsic_adjust (se, expr, fndecl); break; case GFC_ISYM_ADJUSTR: if (expr->ts.kind == 1) fndecl = gfor_fndecl_adjustr; else if (expr->ts.kind == 4) fndecl = gfor_fndecl_adjustr_char4; else gcc_unreachable (); gfc_conv_intrinsic_adjust (se, expr, fndecl); break; case GFC_ISYM_AIMAG: gfc_conv_intrinsic_imagpart (se, expr); break; case GFC_ISYM_AINT: gfc_conv_intrinsic_aint (se, expr, RND_TRUNC); break; case GFC_ISYM_ALL: gfc_conv_intrinsic_anyall (se, expr, EQ_EXPR); break; case GFC_ISYM_ANINT: gfc_conv_intrinsic_aint (se, expr, RND_ROUND); break; case GFC_ISYM_AND: gfc_conv_intrinsic_bitop (se, expr, BIT_AND_EXPR); break; case GFC_ISYM_ANY: gfc_conv_intrinsic_anyall (se, expr, NE_EXPR); break; case GFC_ISYM_BTEST: gfc_conv_intrinsic_btest (se, expr); break; case GFC_ISYM_BGE: gfc_conv_intrinsic_bitcomp (se, expr, GE_EXPR); break; case GFC_ISYM_BGT: gfc_conv_intrinsic_bitcomp (se, expr, GT_EXPR); break; case GFC_ISYM_BLE: gfc_conv_intrinsic_bitcomp (se, expr, LE_EXPR); break; case GFC_ISYM_BLT: gfc_conv_intrinsic_bitcomp (se, expr, LT_EXPR); break; case GFC_ISYM_ACHAR: case GFC_ISYM_CHAR: gfc_conv_intrinsic_char (se, expr); break; case GFC_ISYM_CONVERSION: case GFC_ISYM_REAL: case GFC_ISYM_LOGICAL: case GFC_ISYM_DBLE: gfc_conv_intrinsic_conversion (se, expr); break; /* Integer conversions are handled separately to make sure we get the correct rounding mode. */ case GFC_ISYM_INT: case GFC_ISYM_INT2: case GFC_ISYM_INT8: case GFC_ISYM_LONG: gfc_conv_intrinsic_int (se, expr, RND_TRUNC); break; case GFC_ISYM_NINT: gfc_conv_intrinsic_int (se, expr, RND_ROUND); break; case GFC_ISYM_CEILING: gfc_conv_intrinsic_int (se, expr, RND_CEIL); break; case GFC_ISYM_FLOOR: gfc_conv_intrinsic_int (se, expr, RND_FLOOR); break; case GFC_ISYM_MOD: gfc_conv_intrinsic_mod (se, expr, 0); break; case GFC_ISYM_MODULO: gfc_conv_intrinsic_mod (se, expr, 1); break; case GFC_ISYM_CMPLX: gfc_conv_intrinsic_cmplx (se, expr, name[5] == '1'); break; case GFC_ISYM_COMMAND_ARGUMENT_COUNT: gfc_conv_intrinsic_iargc (se, expr); break; case GFC_ISYM_COMPLEX: gfc_conv_intrinsic_cmplx (se, expr, 1); break; case GFC_ISYM_CONJG: gfc_conv_intrinsic_conjg (se, expr); break; case GFC_ISYM_COUNT: gfc_conv_intrinsic_count (se, expr); break; case GFC_ISYM_CTIME: gfc_conv_intrinsic_ctime (se, expr); break; case GFC_ISYM_DIM: gfc_conv_intrinsic_dim (se, expr); break; case GFC_ISYM_DOT_PRODUCT: gfc_conv_intrinsic_dot_product (se, expr); break; case GFC_ISYM_DPROD: gfc_conv_intrinsic_dprod (se, expr); break; case GFC_ISYM_DSHIFTL: gfc_conv_intrinsic_dshift (se, expr, true); break; case GFC_ISYM_DSHIFTR: gfc_conv_intrinsic_dshift (se, expr, false); break; case GFC_ISYM_FDATE: gfc_conv_intrinsic_fdate (se, expr); break; case GFC_ISYM_FRACTION: gfc_conv_intrinsic_fraction (se, expr); break; case GFC_ISYM_IALL: gfc_conv_intrinsic_arith (se, expr, BIT_AND_EXPR, false); break; case GFC_ISYM_IAND: gfc_conv_intrinsic_bitop (se, expr, BIT_AND_EXPR); break; case GFC_ISYM_IANY: gfc_conv_intrinsic_arith (se, expr, BIT_IOR_EXPR, false); break; case GFC_ISYM_IBCLR: gfc_conv_intrinsic_singlebitop (se, expr, 0); break; case GFC_ISYM_IBITS: gfc_conv_intrinsic_ibits (se, expr); break; case GFC_ISYM_IBSET: gfc_conv_intrinsic_singlebitop (se, expr, 1); break; case GFC_ISYM_IACHAR: case GFC_ISYM_ICHAR: /* We assume ASCII character sequence. */ gfc_conv_intrinsic_ichar (se, expr); break; case GFC_ISYM_IARGC: gfc_conv_intrinsic_iargc (se, expr); break; case GFC_ISYM_IEOR: gfc_conv_intrinsic_bitop (se, expr, BIT_XOR_EXPR); break; case GFC_ISYM_INDEX: kind = expr->value.function.actual->expr->ts.kind; if (kind == 1) fndecl = gfor_fndecl_string_index; else if (kind == 4) fndecl = gfor_fndecl_string_index_char4; else gcc_unreachable (); gfc_conv_intrinsic_index_scan_verify (se, expr, fndecl); break; case GFC_ISYM_IOR: gfc_conv_intrinsic_bitop (se, expr, BIT_IOR_EXPR); break; case GFC_ISYM_IPARITY: gfc_conv_intrinsic_arith (se, expr, BIT_XOR_EXPR, false); break; case GFC_ISYM_IS_IOSTAT_END: gfc_conv_has_intvalue (se, expr, LIBERROR_END); break; case GFC_ISYM_IS_IOSTAT_EOR: gfc_conv_has_intvalue (se, expr, LIBERROR_EOR); break; case GFC_ISYM_ISNAN: gfc_conv_intrinsic_isnan (se, expr); break; case GFC_ISYM_LSHIFT: gfc_conv_intrinsic_shift (se, expr, false, false); break; case GFC_ISYM_RSHIFT: gfc_conv_intrinsic_shift (se, expr, true, true); break; case GFC_ISYM_SHIFTA: gfc_conv_intrinsic_shift (se, expr, true, true); break; case GFC_ISYM_SHIFTL: gfc_conv_intrinsic_shift (se, expr, false, false); break; case GFC_ISYM_SHIFTR: gfc_conv_intrinsic_shift (se, expr, true, false); break; case GFC_ISYM_ISHFT: gfc_conv_intrinsic_ishft (se, expr); break; case GFC_ISYM_ISHFTC: gfc_conv_intrinsic_ishftc (se, expr); break; case GFC_ISYM_LEADZ: gfc_conv_intrinsic_leadz (se, expr); break; case GFC_ISYM_TRAILZ: gfc_conv_intrinsic_trailz (se, expr); break; case GFC_ISYM_POPCNT: gfc_conv_intrinsic_popcnt_poppar (se, expr, 0); break; case GFC_ISYM_POPPAR: gfc_conv_intrinsic_popcnt_poppar (se, expr, 1); break; case GFC_ISYM_LBOUND: gfc_conv_intrinsic_bound (se, expr, 0); break; case GFC_ISYM_LCOBOUND: conv_intrinsic_cobound (se, expr); break; case GFC_ISYM_TRANSPOSE: /* The scalarizer has already been set up for reversed dimension access order ; now we just get the argument value normally. */ gfc_conv_expr (se, expr->value.function.actual->expr); break; case GFC_ISYM_LEN: gfc_conv_intrinsic_len (se, expr); break; case GFC_ISYM_LEN_TRIM: gfc_conv_intrinsic_len_trim (se, expr); break; case GFC_ISYM_LGE: gfc_conv_intrinsic_strcmp (se, expr, GE_EXPR); break; case GFC_ISYM_LGT: gfc_conv_intrinsic_strcmp (se, expr, GT_EXPR); break; case GFC_ISYM_LLE: gfc_conv_intrinsic_strcmp (se, expr, LE_EXPR); break; case GFC_ISYM_LLT: gfc_conv_intrinsic_strcmp (se, expr, LT_EXPR); break; case GFC_ISYM_MASKL: gfc_conv_intrinsic_mask (se, expr, 1); break; case GFC_ISYM_MASKR: gfc_conv_intrinsic_mask (se, expr, 0); break; case GFC_ISYM_MAX: if (expr->ts.type == BT_CHARACTER) gfc_conv_intrinsic_minmax_char (se, expr, 1); else gfc_conv_intrinsic_minmax (se, expr, GT_EXPR); break; case GFC_ISYM_MAXLOC: gfc_conv_intrinsic_minmaxloc (se, expr, GT_EXPR); break; case GFC_ISYM_MAXVAL: gfc_conv_intrinsic_minmaxval (se, expr, GT_EXPR); break; case GFC_ISYM_MERGE: gfc_conv_intrinsic_merge (se, expr); break; case GFC_ISYM_MERGE_BITS: gfc_conv_intrinsic_merge_bits (se, expr); break; case GFC_ISYM_MIN: if (expr->ts.type == BT_CHARACTER) gfc_conv_intrinsic_minmax_char (se, expr, -1); else gfc_conv_intrinsic_minmax (se, expr, LT_EXPR); break; case GFC_ISYM_MINLOC: gfc_conv_intrinsic_minmaxloc (se, expr, LT_EXPR); break; case GFC_ISYM_MINVAL: gfc_conv_intrinsic_minmaxval (se, expr, LT_EXPR); break; case GFC_ISYM_NEAREST: gfc_conv_intrinsic_nearest (se, expr); break; case GFC_ISYM_NORM2: gfc_conv_intrinsic_arith (se, expr, PLUS_EXPR, true); break; case GFC_ISYM_NOT: gfc_conv_intrinsic_not (se, expr); break; case GFC_ISYM_OR: gfc_conv_intrinsic_bitop (se, expr, BIT_IOR_EXPR); break; case GFC_ISYM_PARITY: gfc_conv_intrinsic_arith (se, expr, NE_EXPR, false); break; case GFC_ISYM_PRESENT: gfc_conv_intrinsic_present (se, expr); break; case GFC_ISYM_PRODUCT: gfc_conv_intrinsic_arith (se, expr, MULT_EXPR, false); break; case GFC_ISYM_RANK: gfc_conv_intrinsic_rank (se, expr); break; case GFC_ISYM_RRSPACING: gfc_conv_intrinsic_rrspacing (se, expr); break; case GFC_ISYM_SET_EXPONENT: gfc_conv_intrinsic_set_exponent (se, expr); break; case GFC_ISYM_SCALE: gfc_conv_intrinsic_scale (se, expr); break; case GFC_ISYM_SIGN: gfc_conv_intrinsic_sign (se, expr); break; case GFC_ISYM_SIZE: gfc_conv_intrinsic_size (se, expr); break; case GFC_ISYM_SIZEOF: case GFC_ISYM_C_SIZEOF: gfc_conv_intrinsic_sizeof (se, expr); break; case GFC_ISYM_STORAGE_SIZE: gfc_conv_intrinsic_storage_size (se, expr); break; case GFC_ISYM_SPACING: gfc_conv_intrinsic_spacing (se, expr); break; case GFC_ISYM_SUM: gfc_conv_intrinsic_arith (se, expr, PLUS_EXPR, false); break; case GFC_ISYM_TRANSFER: if (se->ss && se->ss->info->useflags) /* Access the previously obtained result. */ gfc_conv_tmp_array_ref (se); else gfc_conv_intrinsic_transfer (se, expr); break; case GFC_ISYM_TTYNAM: gfc_conv_intrinsic_ttynam (se, expr); break; case GFC_ISYM_UBOUND: gfc_conv_intrinsic_bound (se, expr, 1); break; case GFC_ISYM_UCOBOUND: conv_intrinsic_cobound (se, expr); break; case GFC_ISYM_XOR: gfc_conv_intrinsic_bitop (se, expr, BIT_XOR_EXPR); break; case GFC_ISYM_LOC: gfc_conv_intrinsic_loc (se, expr); break; case GFC_ISYM_THIS_IMAGE: /* For num_images() == 1, handle as LCOBOUND. */ if (expr->value.function.actual->expr && gfc_option.coarray == GFC_FCOARRAY_SINGLE) conv_intrinsic_cobound (se, expr); else trans_this_image (se, expr); break; case GFC_ISYM_IMAGE_INDEX: trans_image_index (se, expr); break; case GFC_ISYM_NUM_IMAGES: trans_num_images (se); break; case GFC_ISYM_ACCESS: case GFC_ISYM_CHDIR: case GFC_ISYM_CHMOD: case GFC_ISYM_DTIME: case GFC_ISYM_ETIME: case GFC_ISYM_EXTENDS_TYPE_OF: case GFC_ISYM_FGET: case GFC_ISYM_FGETC: case GFC_ISYM_FNUM: case GFC_ISYM_FPUT: case GFC_ISYM_FPUTC: case GFC_ISYM_FSTAT: case GFC_ISYM_FTELL: case GFC_ISYM_GETCWD: case GFC_ISYM_GETGID: case GFC_ISYM_GETPID: case GFC_ISYM_GETUID: case GFC_ISYM_HOSTNM: case GFC_ISYM_KILL: case GFC_ISYM_IERRNO: case GFC_ISYM_IRAND: case GFC_ISYM_ISATTY: case GFC_ISYM_JN2: case GFC_ISYM_LINK: case GFC_ISYM_LSTAT: case GFC_ISYM_MALLOC: case GFC_ISYM_MATMUL: case GFC_ISYM_MCLOCK: case GFC_ISYM_MCLOCK8: case GFC_ISYM_RAND: case GFC_ISYM_RENAME: case GFC_ISYM_SECOND: case GFC_ISYM_SECNDS: case GFC_ISYM_SIGNAL: case GFC_ISYM_STAT: case GFC_ISYM_SYMLNK: case GFC_ISYM_SYSTEM: case GFC_ISYM_TIME: case GFC_ISYM_TIME8: case GFC_ISYM_UMASK: case GFC_ISYM_UNLINK: case GFC_ISYM_YN2: gfc_conv_intrinsic_funcall (se, expr); break; case GFC_ISYM_EOSHIFT: case GFC_ISYM_PACK: case GFC_ISYM_RESHAPE: /* For those, expr->rank should always be >0 and thus the if above the switch should have matched. */ gcc_unreachable (); break; default: gfc_conv_intrinsic_lib_function (se, expr); break; } } static gfc_ss * walk_inline_intrinsic_transpose (gfc_ss *ss, gfc_expr *expr) { gfc_ss *arg_ss, *tmp_ss; gfc_actual_arglist *arg; arg = expr->value.function.actual; gcc_assert (arg->expr); arg_ss = gfc_walk_subexpr (gfc_ss_terminator, arg->expr); gcc_assert (arg_ss != gfc_ss_terminator); for (tmp_ss = arg_ss; ; tmp_ss = tmp_ss->next) { if (tmp_ss->info->type != GFC_SS_SCALAR && tmp_ss->info->type != GFC_SS_REFERENCE) { int tmp_dim; gcc_assert (tmp_ss->dimen == 2); /* We just invert dimensions. */ tmp_dim = tmp_ss->dim[0]; tmp_ss->dim[0] = tmp_ss->dim[1]; tmp_ss->dim[1] = tmp_dim; } /* Stop when tmp_ss points to the last valid element of the chain... */ if (tmp_ss->next == gfc_ss_terminator) break; } /* ... so that we can attach the rest of the chain to it. */ tmp_ss->next = ss; return arg_ss; } /* Move the given dimension of the given gfc_ss list to a nested gfc_ss list. This has the side effect of reversing the nested list, so there is no need to call gfc_reverse_ss on it (the given list is assumed not to be reversed yet). */ static gfc_ss * nest_loop_dimension (gfc_ss *ss, int dim) { int ss_dim, i; gfc_ss *new_ss, *prev_ss = gfc_ss_terminator; gfc_loopinfo *new_loop; gcc_assert (ss != gfc_ss_terminator); for (; ss != gfc_ss_terminator; ss = ss->next) { new_ss = gfc_get_ss (); new_ss->next = prev_ss; new_ss->parent = ss; new_ss->info = ss->info; new_ss->info->refcount++; if (ss->dimen != 0) { gcc_assert (ss->info->type != GFC_SS_SCALAR && ss->info->type != GFC_SS_REFERENCE); new_ss->dimen = 1; new_ss->dim[0] = ss->dim[dim]; gcc_assert (dim < ss->dimen); ss_dim = --ss->dimen; for (i = dim; i < ss_dim; i++) ss->dim[i] = ss->dim[i + 1]; ss->dim[ss_dim] = 0; } prev_ss = new_ss; if (ss->nested_ss) { ss->nested_ss->parent = new_ss; new_ss->nested_ss = ss->nested_ss; } ss->nested_ss = new_ss; } new_loop = gfc_get_loopinfo (); gfc_init_loopinfo (new_loop); gcc_assert (prev_ss != NULL); gcc_assert (prev_ss != gfc_ss_terminator); gfc_add_ss_to_loop (new_loop, prev_ss); return new_ss->parent; } /* Create the gfc_ss list for the SUM/PRODUCT arguments when the function is to be inlined. */ static gfc_ss * walk_inline_intrinsic_arith (gfc_ss *ss, gfc_expr *expr) { gfc_ss *tmp_ss, *tail, *array_ss; gfc_actual_arglist *arg1, *arg2, *arg3; int sum_dim; bool scalar_mask = false; /* The rank of the result will be determined later. */ arg1 = expr->value.function.actual; arg2 = arg1->next; arg3 = arg2->next; gcc_assert (arg3 != NULL); if (expr->rank == 0) return ss; tmp_ss = gfc_ss_terminator; if (arg3->expr) { gfc_ss *mask_ss; mask_ss = gfc_walk_subexpr (tmp_ss, arg3->expr); if (mask_ss == tmp_ss) scalar_mask = 1; tmp_ss = mask_ss; } array_ss = gfc_walk_subexpr (tmp_ss, arg1->expr); gcc_assert (array_ss != tmp_ss); /* Odd thing: If the mask is scalar, it is used by the frontend after the array (to make an if around the nested loop). Thus it shall be after array_ss once the gfc_ss list is reversed. */ if (scalar_mask) tmp_ss = gfc_get_scalar_ss (array_ss, arg3->expr); else tmp_ss = array_ss; /* "Hide" the dimension on which we will sum in the first arg's scalarization chain. */ sum_dim = mpz_get_si (arg2->expr->value.integer) - 1; tail = nest_loop_dimension (tmp_ss, sum_dim); tail->next = ss; return tmp_ss; } static gfc_ss * walk_inline_intrinsic_function (gfc_ss * ss, gfc_expr * expr) { switch (expr->value.function.isym->id) { case GFC_ISYM_PRODUCT: case GFC_ISYM_SUM: return walk_inline_intrinsic_arith (ss, expr); case GFC_ISYM_TRANSPOSE: return walk_inline_intrinsic_transpose (ss, expr); default: gcc_unreachable (); } gcc_unreachable (); } /* This generates code to execute before entering the scalarization loop. Currently does nothing. */ void gfc_add_intrinsic_ss_code (gfc_loopinfo * loop ATTRIBUTE_UNUSED, gfc_ss * ss) { switch (ss->info->expr->value.function.isym->id) { case GFC_ISYM_UBOUND: case GFC_ISYM_LBOUND: case GFC_ISYM_UCOBOUND: case GFC_ISYM_LCOBOUND: case GFC_ISYM_THIS_IMAGE: break; default: gcc_unreachable (); } } /* The LBOUND, LCOBOUND, UBOUND and UCOBOUND intrinsics with one parameter are expanded into code inside the scalarization loop. */ static gfc_ss * gfc_walk_intrinsic_bound (gfc_ss * ss, gfc_expr * expr) { if (expr->value.function.actual->expr->ts.type == BT_CLASS) gfc_add_class_array_ref (expr->value.function.actual->expr); /* The two argument version returns a scalar. */ if (expr->value.function.actual->next->expr) return ss; return gfc_get_array_ss (ss, expr, 1, GFC_SS_INTRINSIC); } /* Walk an intrinsic array libcall. */ static gfc_ss * gfc_walk_intrinsic_libfunc (gfc_ss * ss, gfc_expr * expr) { gcc_assert (expr->rank > 0); return gfc_get_array_ss (ss, expr, expr->rank, GFC_SS_FUNCTION); } /* Return whether the function call expression EXPR will be expanded inline by gfc_conv_intrinsic_function. */ bool gfc_inline_intrinsic_function_p (gfc_expr *expr) { gfc_actual_arglist *args; if (!expr->value.function.isym) return false; switch (expr->value.function.isym->id) { case GFC_ISYM_PRODUCT: case GFC_ISYM_SUM: /* Disable inline expansion if code size matters. */ if (optimize_size) return false; args = expr->value.function.actual; /* We need to be able to subset the SUM argument at compile-time. */ if (args->next->expr && args->next->expr->expr_type != EXPR_CONSTANT) return false; return true; case GFC_ISYM_TRANSPOSE: return true; default: return false; } } /* Returns nonzero if the specified intrinsic function call maps directly to an external library call. Should only be used for functions that return arrays. */ int gfc_is_intrinsic_libcall (gfc_expr * expr) { gcc_assert (expr->expr_type == EXPR_FUNCTION && expr->value.function.isym); gcc_assert (expr->rank > 0); if (gfc_inline_intrinsic_function_p (expr)) return 0; switch (expr->value.function.isym->id) { case GFC_ISYM_ALL: case GFC_ISYM_ANY: case GFC_ISYM_COUNT: case GFC_ISYM_JN2: case GFC_ISYM_IANY: case GFC_ISYM_IALL: case GFC_ISYM_IPARITY: case GFC_ISYM_MATMUL: case GFC_ISYM_MAXLOC: case GFC_ISYM_MAXVAL: case GFC_ISYM_MINLOC: case GFC_ISYM_MINVAL: case GFC_ISYM_NORM2: case GFC_ISYM_PARITY: case GFC_ISYM_PRODUCT: case GFC_ISYM_SUM: case GFC_ISYM_SHAPE: case GFC_ISYM_SPREAD: case GFC_ISYM_YN2: /* Ignore absent optional parameters. */ return 1; case GFC_ISYM_RESHAPE: case GFC_ISYM_CSHIFT: case GFC_ISYM_EOSHIFT: case GFC_ISYM_PACK: case GFC_ISYM_UNPACK: /* Pass absent optional parameters. */ return 2; default: return 0; } } /* Walk an intrinsic function. */ gfc_ss * gfc_walk_intrinsic_function (gfc_ss * ss, gfc_expr * expr, gfc_intrinsic_sym * isym) { gcc_assert (isym); if (isym->elemental) return gfc_walk_elemental_function_args (ss, expr->value.function.actual, NULL, GFC_SS_SCALAR); if (expr->rank == 0) return ss; if (gfc_inline_intrinsic_function_p (expr)) return walk_inline_intrinsic_function (ss, expr); if (gfc_is_intrinsic_libcall (expr)) return gfc_walk_intrinsic_libfunc (ss, expr); /* Special cases. */ switch (isym->id) { case GFC_ISYM_LBOUND: case GFC_ISYM_LCOBOUND: case GFC_ISYM_UBOUND: case GFC_ISYM_UCOBOUND: case GFC_ISYM_THIS_IMAGE: return gfc_walk_intrinsic_bound (ss, expr); case GFC_ISYM_TRANSFER: return gfc_walk_intrinsic_libfunc (ss, expr); default: /* This probably meant someone forgot to add an intrinsic to the above list(s) when they implemented it, or something's gone horribly wrong. */ gcc_unreachable (); } } static tree conv_intrinsic_atomic_def (gfc_code *code) { gfc_se atom, value; stmtblock_t block; gfc_init_se (&atom, NULL); gfc_init_se (&value, NULL); gfc_conv_expr (&atom, code->ext.actual->expr); gfc_conv_expr (&value, code->ext.actual->next->expr); gfc_init_block (&block); gfc_add_modify (&block, atom.expr, fold_convert (TREE_TYPE (atom.expr), value.expr)); return gfc_finish_block (&block); } static tree conv_intrinsic_atomic_ref (gfc_code *code) { gfc_se atom, value; stmtblock_t block; gfc_init_se (&atom, NULL); gfc_init_se (&value, NULL); gfc_conv_expr (&value, code->ext.actual->expr); gfc_conv_expr (&atom, code->ext.actual->next->expr); gfc_init_block (&block); gfc_add_modify (&block, value.expr, fold_convert (TREE_TYPE (value.expr), atom.expr)); return gfc_finish_block (&block); } static tree conv_intrinsic_move_alloc (gfc_code *code) { stmtblock_t block; gfc_expr *from_expr, *to_expr; gfc_expr *to_expr2, *from_expr2 = NULL; gfc_se from_se, to_se; tree tmp; bool coarray; gfc_start_block (&block); from_expr = code->ext.actual->expr; to_expr = code->ext.actual->next->expr; gfc_init_se (&from_se, NULL); gfc_init_se (&to_se, NULL); gcc_assert (from_expr->ts.type != BT_CLASS || to_expr->ts.type == BT_CLASS); coarray = gfc_get_corank (from_expr) != 0; if (from_expr->rank == 0 && !coarray) { if (from_expr->ts.type != BT_CLASS) from_expr2 = from_expr; else { from_expr2 = gfc_copy_expr (from_expr); gfc_add_data_component (from_expr2); } if (to_expr->ts.type != BT_CLASS) to_expr2 = to_expr; else { to_expr2 = gfc_copy_expr (to_expr); gfc_add_data_component (to_expr2); } from_se.want_pointer = 1; to_se.want_pointer = 1; gfc_conv_expr (&from_se, from_expr2); gfc_conv_expr (&to_se, to_expr2); gfc_add_block_to_block (&block, &from_se.pre); gfc_add_block_to_block (&block, &to_se.pre); /* Deallocate "to". */ tmp = gfc_deallocate_scalar_with_status (to_se.expr, NULL_TREE, true, to_expr2, to_expr->ts); gfc_add_expr_to_block (&block, tmp); /* Assign (_data) pointers. */ gfc_add_modify_loc (input_location, &block, to_se.expr, fold_convert (TREE_TYPE (to_se.expr), from_se.expr)); /* Set "from" to NULL. */ gfc_add_modify_loc (input_location, &block, from_se.expr, fold_convert (TREE_TYPE (from_se.expr), null_pointer_node)); gfc_add_block_to_block (&block, &from_se.post); gfc_add_block_to_block (&block, &to_se.post); /* Set _vptr. */ if (to_expr->ts.type == BT_CLASS) { gfc_free_expr (to_expr2); gfc_init_se (&to_se, NULL); to_se.want_pointer = 1; gfc_add_vptr_component (to_expr); gfc_conv_expr (&to_se, to_expr); if (from_expr->ts.type == BT_CLASS) { gfc_free_expr (from_expr2); gfc_init_se (&from_se, NULL); from_se.want_pointer = 1; gfc_add_vptr_component (from_expr); gfc_conv_expr (&from_se, from_expr); tmp = from_se.expr; } else { gfc_symbol *vtab; vtab = gfc_find_derived_vtab (from_expr->ts.u.derived); gcc_assert (vtab); tmp = gfc_build_addr_expr (NULL_TREE, gfc_get_symbol_decl (vtab)); } gfc_add_modify_loc (input_location, &block, to_se.expr, fold_convert (TREE_TYPE (to_se.expr), tmp)); } return gfc_finish_block (&block); } /* Update _vptr component. */ if (to_expr->ts.type == BT_CLASS) { to_se.want_pointer = 1; to_expr2 = gfc_copy_expr (to_expr); gfc_add_vptr_component (to_expr2); gfc_conv_expr (&to_se, to_expr2); if (from_expr->ts.type == BT_CLASS) { from_se.want_pointer = 1; from_expr2 = gfc_copy_expr (from_expr); gfc_add_vptr_component (from_expr2); gfc_conv_expr (&from_se, from_expr2); tmp = from_se.expr; } else { gfc_symbol *vtab; vtab = gfc_find_derived_vtab (from_expr->ts.u.derived); gcc_assert (vtab); tmp = gfc_build_addr_expr (NULL_TREE, gfc_get_symbol_decl (vtab)); } gfc_add_modify_loc (input_location, &block, to_se.expr, fold_convert (TREE_TYPE (to_se.expr), tmp)); gfc_free_expr (to_expr2); gfc_init_se (&to_se, NULL); if (from_expr->ts.type == BT_CLASS) { gfc_free_expr (from_expr2); gfc_init_se (&from_se, NULL); } } /* Deallocate "to". */ if (from_expr->rank == 0) { to_se.want_coarray = 1; from_se.want_coarray = 1; } gfc_conv_expr_descriptor (&to_se, to_expr); gfc_conv_expr_descriptor (&from_se, from_expr); /* For coarrays, call SYNC ALL if TO is already deallocated as MOVE_ALLOC is an image control "statement", cf. IR F08/0040 in 12-006A. */ if (coarray && gfc_option.coarray == GFC_FCOARRAY_LIB) { tree cond; tmp = gfc_deallocate_with_status (to_se.expr, NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE, true, to_expr, true); gfc_add_expr_to_block (&block, tmp); tmp = gfc_conv_descriptor_data_get (to_se.expr); cond = fold_build2_loc (input_location, EQ_EXPR, boolean_type_node, tmp, fold_convert (TREE_TYPE (tmp), null_pointer_node)); tmp = build_call_expr_loc (input_location, gfor_fndecl_caf_sync_all, 3, null_pointer_node, null_pointer_node, build_int_cst (integer_type_node, 0)); tmp = fold_build3_loc (input_location, COND_EXPR, void_type_node, cond, tmp, build_empty_stmt (input_location)); gfc_add_expr_to_block (&block, tmp); } else { tmp = gfc_conv_descriptor_data_get (to_se.expr); tmp = gfc_deallocate_with_status (tmp, NULL_TREE, NULL_TREE, NULL_TREE, NULL_TREE, true, to_expr, false); gfc_add_expr_to_block (&block, tmp); } /* Move the pointer and update the array descriptor data. */ gfc_add_modify_loc (input_location, &block, to_se.expr, from_se.expr); /* Set "to" to NULL. */ tmp = gfc_conv_descriptor_data_get (from_se.expr); gfc_add_modify_loc (input_location, &block, tmp, fold_convert (TREE_TYPE (tmp), null_pointer_node)); return gfc_finish_block (&block); } tree gfc_conv_intrinsic_subroutine (gfc_code *code) { tree res; gcc_assert (code->resolved_isym); switch (code->resolved_isym->id) { case GFC_ISYM_MOVE_ALLOC: res = conv_intrinsic_move_alloc (code); break; case GFC_ISYM_ATOMIC_DEF: res = conv_intrinsic_atomic_def (code); break; case GFC_ISYM_ATOMIC_REF: res = conv_intrinsic_atomic_ref (code); break; default: res = NULL_TREE; break; } return res; } #include "gt-fortran-trans-intrinsic.h"