Uros Bizjak [Sat, 13 Oct 2012 21:39:29 +0000 (23:39 +0200)]
alpha.md (I24MODE): New mode iterator.
* config/alpha/alpha.md (I24MODE): New mode iterator.
(any_divmod): New code iterator.
(<code>si3): Macroize expander from {div,mod,udiv,umod}si3 using
any_divmod code iterator.
(<code>si3): Macroize expander from {div,mod,udiv,umod}di3 using
any_divmod code iterator.
(extendqi<mode>2): Macroize insn from extendqi{hi,si}2 using
I24MODE mode iterator.
(unaligned_store<mode>): Macroize expander from unaligned_store{qi,hi}
using I12MODE mode iterator.
(mov<mode>): Macroize expander from mov{qi,hi} using
I12MODE mode iterator.
Eric Botcazou [Sat, 13 Oct 2012 20:22:07 +0000 (20:22 +0000)]
re PR rtl-optimization/54871 (gfortran.dg/vector_subscript_1.f90 FAILs)
PR rtl-optimization/54871
* loop-iv.c (simplify_using_initial_values): When scanning previous
basic blocks, prune the recorded conditions if the current insn was
not used to make a replacement.
* loop-unroll.c (decide_unroll_constant_iterations): Clean up message.
(unroll_loop_constant_iterations): Clarify head comment.
(decide_unroll_runtime_iterations): Clean up message.
(unroll_loop_runtime_iterations): Clarify head comment.
(decide_peel_simple): Clean up message.
(peel_loop_simple): Clarify head comment.
(decide_unroll_stupid): Clean up message.
(unroll_loop_stupid): Clarify head comment.
Uros Bizjak [Sat, 13 Oct 2012 08:36:21 +0000 (10:36 +0200)]
alpha.md (FMODE): New mode iterator.
* config/alpha/alpha.md (FMODE): New mode iterator.
(modesuffix): Handle SF and DF modes.
(opmode): New mode attribute.
(abs<mode>2): Macroize insn from abs{sf,df}2 using FMODE mode iterator.
(*nabs<mode>2): Macroize insn from *nabs{sf,df}2 using
FMODE mode iterator.
(neg<mode>2): Macroize insn from neg{sf,df}2 using FMODE mode iterator.
(copysign<mode>3): Macroize insn from copysign{sf,df}3 using
FMODE mode iterator.
(*ncopysign<mode>3): Macroize insn from *ncopysign{sf,df}3 using
FMODE mode iterator.
(*add<mode>3_ieee): Macroize insn from *add{sf,df}_ieee using
FMODE mode iterator.
(add<mode>3): Macroize insn from add{sf,df}3 using FMODE mode iterator.
(*sub<mode>3_ieee): Macroize insn from *sub{sf,df}3_ieee using
FMODE mode iterator.
(sub<mode>3): Macroize insn from sub{sf,df}3 using FMODE mode iterator.
(*mul<mode>3_ieee): Macroize insn from *mul{sf,df}3_ieee using
FMODE mode iterator.
(mul<mode>3): Macroize insn from mul{sf,df}3 using FMODE mode iterator.
(*div<mode>3_ieee): Macroize insn from *div{sf,df}3_ieee using
FMODE mode iterator.
(div<mode>3): Macroize insn from div{sf,df}3 using FMODE mode iterator.
(*sqrt<mode>2_ieee): Macroize insn from *sqrt{sf,df}2_ieee using
FMODE mode iterator.
(sqrt<mode>2): Macroize insn from sqrt{sf,df}2
using FMODE mode iterator.
(*mov<mode>cc_internal): Macroize insn from *mov{sf,df}cc_internal
using FMODE mode iterator.
(mov<mode>cc): Macroize expander from mov{sf,df}cc
using FMODE mode iterator.
Jakub Jelinek [Fri, 12 Oct 2012 18:23:03 +0000 (20:23 +0200)]
re PR c/54381 (-Wsizeof-pointer-memaccess refers to "destination" for strncmp)
PR c/54381
* c-common.h (sizeof_pointer_memaccess_warning): Adjust prototype.
* c-common.c (sizeof_pointer_memaccess_warning): Take array of 3
locs and array of 3 trees instead of just single loc and single
sizeof_arg tree. Handle __builtin___*_chk builtins too, and
also stpncpy, bcopy, bcmp, bzero, snprintf and vsnprintf builtins.
For *cmp* builtins that take two sources strings report warnings
about first and second source, not about destination and source.
* c-parser.c (struct c_tree_loc_pair): Removed.
(c_parser_expr_list): Remove struct c_tree_loc_pair * argument,
add location_t * and tree * arguments, fill in array of 3
sizeof_arg trees and corresponding locs.
(c_parser_attributes, c_parser_objc_keywordexpr): Adjust
c_parser_expr_list callers.
(c_parser_postfix_expression_after_primary): Likewise. Pass
array of 3 sizeof_arg trees and locs (corresponding to first
3 arguments) to sizeof_pointer_memaccess_warning.
* semantics.c (finish_call_expr): Pass array of 3 sizeof_arg
trees and locs (corresponding to first 3 arguments) to
sizeof_pointer_memaccess_warning.
* c-c++-common/Wsizeof-pointer-memaccess1.c: New test.
* c-c++-common/Wsizeof-pointer-memaccess2.c: New test.
* gcc.dg/Wsizeof-pointer-memaccess1.c: New test.
* gcc.dg/torture/Wsizeof-pointer-memaccess1.c: Test also stpncpy.
Adjust expected wording of warnings for *cmp* builtins.
* g++.dg/torture/Wsizeof-pointer-memaccess1.C: Likewise.
* g++.dg/torture/Wsizeof-pointer-memaccess2.C: Likewise.
Diego Novillo [Fri, 12 Oct 2012 15:17:17 +0000 (11:17 -0400)]
Add more C++ support in gengtype.
This patch combines the changes from
http://gcc.gnu.org/ml/gcc-patches/2012-08/msg02016.html with other
additions to support C++ inside GTY'd structures.
The main changes wrt Aaron's original patch are:
- Support for function declarations inside classes.
- Support scoping in identifiers. This does not mean that gengtype
supports scopes, it just knows that 'Foo::id' is a single entity.
- Explicit non-support for typedef and enum inside class/struct.
Since gengtype does not really know about scopes, it cannot
understand these types, but it knows enough to recognize and reject
them. GTY'd struct/class that need to typedef their own types
should use GTY((user)).
- Documentation on what is and is not supported.
There is one check I needed to remove that gave me some trouble.
When a ctor is detected, we have already parsed the name of the
ctor as a type, which is then registered in the list of structures.
We go on to recognize it as a ctor *after* the type has been
registered. We reject the field in declarator() and it is never
added to the list of fields for the class.
However, when we reach the end of the class, we find that the
type we created while parsing the ctor has line number
information in it (the line where the ctor was) and gengtype
thinks that it is a duplicate structure definition.
I took out this check for two reasons: (a) It is actually
unnecessary because if there were really duplicate definitions of
this structure, the code would not compile, and (b) all the other
alternatives required making the parser much more convoluted and
I'm trying hard not to make gengtype parser too smart.
2012-10-12 Aaron Gray <aaronngray.lists@gmail.com>
Diego Novillo <dnovillo@google.com>
* gengtype-lex.l: Support for C++ single line comments.
Support for classes.
(CXX_KEYWORD): New. Support C++ keywords inline, public,
protected, private, template, operator, friend, &, ~.
(TYPEDEF): New. Support typedef.
* gengtype-parser.c: updated 'token_names[]'
(direct_declarator): Add support for parsing functions
and ctors.
2012-10-12 Diego Novillo <dnovillo@google.com>
* doc/gty.texi: Document C++ limitations in gengtype.
* gengtype-lex.l (CID): Rename from ID.
(ID): Include scoping '::' as part of the identifier name.
* gengtype-parse.c (token_names): Update.
(token_value_format): Update.
(consume_until_eos): Rename from consume_until_semi.
Remove unused argument IMMEDIATE. Update all callers.
Also consider '}' as a finalizer.
(consume_until_comma_or_eos): Rename from
consume_until_comma_or_semi.
Remove unused argument IMMEDIATE. Update all callers.
Also consider '}' as a finalizer.
(direct_declarator): Add documentation on ctor support.
Add argument IN_STRUCT.
If the token following ID is a '(', consider ID a
function and return NULL.
If the token following '(' is not a '*', and IN_STRUCT is
true, conclude that this is a ctor and return NULL.
If the token is IGNORABLE_CXX_KEYWORD, return NULL.
(inner_declarator): Add argument IN_STRUCT.
Update all callers.
(declarator): Add argument IN_STRUCT with default value
false. Update all callers.
(type): Document argument NESTED.
Skip over C++ inheritance specifiers.
If a token TYPEDEF is found, emit an error.
If an enum is found inside a class/structure, emit an
error.
(typedefs, structures, param_structs, variables): Initialize.
(new_structure): Do not complain about duplicate
structures if S has a line location set.
* gengtype-state.c (write_state_type): Remove default
handler. Add handler for TYPE_NONE.
(read_state_scalar_char_type):
* gengtype.c: Fix spacing.
* gengtype.h (enum gty_token): Add name. Add token
IGNORABLE_CXX_KEYWORD.
Richard Biener [Fri, 12 Oct 2012 12:14:48 +0000 (12:14 +0000)]
tree-streamer-out.c (pack_ts_target_option): Rename from ...
2012-10-12 Richard Biener <rguenther@suse.de>
* tree-streamer-out.c (pack_ts_target_option): Rename from ...
(write_ts_target_option): ... this.
(pack_ts_optimization): Rename from ...
(write_ts_optimization): ... this.
(streamer_pack_tree_bitfields): Pack them in the bitfield section ...
(streamer_write_tree_body): ... not here.
* tree-streamer-in.c (unpack_ts_target_option): Rename from ...
(lto_input_ts_target_option): ... this.
(unpack_ts_optimization): Rename from ...
(lto_input_ts_optimization): ... this.
(unpack_value_fields): Unpack them from the bitfield section ...
(streamer_read_tree_body): ... not from here.
Uros Bizjak [Fri, 12 Oct 2012 11:32:02 +0000 (13:32 +0200)]
alpha.md (vecmodesuffix): New mode attribute.
* config/alpha/alpha.md (vecmodesuffix): New mode attribute.
(modesuffix): Handle V8QI and V4HI modes.
(any_maxmin): New code iterator.
(maxmin): New code attribute.
(<code><mode>3): Macroize insn from {smax,smin,umax,umin}{qi,hi}3
using any_maxmin code iterator and I12MODE mode iterator.
(<code><mode>3): Macroize insn from {smax,smin,umax,umin}{v8qi,v4hi}3
using any_maxmin code iterator and VEC12 mode iterator.
Marc Glisse [Fri, 12 Oct 2012 09:20:00 +0000 (11:20 +0200)]
optabs.c (vector_compare_rtx): Change prototype.
2012-10-12 Marc Glisse <marc.glisse@inria.fr>
* optabs.c (vector_compare_rtx): Change prototype.
(expand_vec_cond_expr): Handle VEC_COND_EXPR whose first operand
is not a comparison.
* gimplify.c (gimplify_expr): Handle VEC_COND_EXPR.
Oleg Endo [Fri, 12 Oct 2012 00:41:23 +0000 (00:41 +0000)]
re PR target/51244 ([SH] Inefficient conditional branch and code around T bit)
PR target/51244
* config/sh/sh.md (negsi_cond, negdi_cond, stack_protect_test): Remove
get_t_reg_rtx when invoking gen_branch_true or gen_branch_false.
(*zero_extend<mode>si2_compact): Convert to insn_and_split. Convert
zero extensions of T bit stores to reg moves in splitter. Remove
obsolete unnamed peephole2 that caught zero extensions after negc T bit
stores.
(*branch_true_eq, *branch_false_ne): Delete.
(branch_true, branch_false): Convert insn to expander. Move actual
insn logic to...
(*cbranch_t): ...this new insn_and_split. Try to find preceding
redundant T bit stores and tests and combine them with the conditional
branch if possible in the splitter.
(movrt_xor, *movt_movrt): New insn_and_split.
* config/sh/predicates.md (cbranch_treg_value): New predicate.
* config/sh/sh-protos.h (sh_eval_treg_value): Forward declare...
* config/sh/sh.c (sh_eval_treg_value): ...this new function.
(expand_cbranchsi4, expand_cbranchdi4): Remove get_t_reg_rtx
when invoking gen_branch_true or gen_branch_false.
Uros Bizjak [Thu, 11 Oct 2012 20:44:30 +0000 (22:44 +0200)]
alpha.md (IMODE): New mode iterator.
* config/alpha/alpha.md (IMODE): New mode iterator.
(I124MODE): Ditto.
(I248MODE): Ditto.
(modesuffix): Handle QI and HI modes.
(zero_extendqi<mode>2): Macroize insn from zero_extendqi{hi,si,di}2
using I248MODE mode iterator.
(zero_extendhi<mode>2): Macroize insn from zero_extendhi{si,di}2
using I48MODE mode iterator.
(andnot<mode>3): Macroize insn from andnot{si,di}3 using
I48MODE mode iterator.
(ins<modesuffix>l_const): Macroize insn from ins{b,w,l}l_const
using I248MODE mode iterator.
(ins<modesuffix>l): Macroize insn from ins{b,w,l}l
using I248MODE mode iterator.
(*mov<mode>cc_internal): Macroize insn from
*mov{qi,hi,si,di}cc_internal using IMODE mode iterator.
(*mov<mode>cc_lbc): Macroize insn from
*mov{qi,hi,si,di}cc_lbc using IMODE mode iterator.
(*mov<mode>cc_lbs): Macroize insn from
*mov{qi,hi,si,di}cc_lbs using IMODE mode iterator.
(mov<mode>cc): Macroize insn from mov{si,di}cc_lbs
using I48MODE mode iterator.
Steven Bosscher [Thu, 11 Oct 2012 18:54:47 +0000 (18:54 +0000)]
ira-build.c (ira_loop_tree_body_rev_postorder): New function.
* ira-build.c (ira_loop_tree_body_rev_postorder): New function.
(ira_traverse_loop_tree): Traverse a loop's basic blocks in
reverse post-order of the reversed control-flow direction.
* ira-conflicts.c (ira_build_conflicts): Pass add_copies as
the pre-order function to ira_traverse_loop_tree to preserve
the existing semantics.
* ira-lives.c (remove_some_program_points_and_update_live_ranges):
Squeeze out live range chain elements if their program points are
connected.
Uros Bizjak [Thu, 11 Oct 2012 15:44:54 +0000 (17:44 +0200)]
alpha.md (DWI): New mode attribute.
* config/alpha/alpha.md (DWI): New mode attribute.
(*sadd<modesuffix>): Macroize insn from *saddl and *saddq using
I48MODE mode iterator.
(addv<mode>3): Macroize insn from addvsi3 and addvdi3 using
I48MODE mode iterator.
(neg<mode>2): Macroize insn from negsi2 and negdi2 using
I48MODE mode iterator.
(negv<mode>2): Macroize insn from negvsi2 and negvdi2 using
I48MODE mode iterator.
(sub<mode>3): Macroize insn from subsi3 and subdi3 using
I48MODE mode iterator.
(*ssub<modesuffix>): Macroize insn from *ssubl and *ssubq using
I48MODE mode iterator.
(subv<mode>3): Macroize insn from subvsi3 and subvdi3 using
I48MODE mode iterator.
(mul<mode>3): Macroize insn from mulsi3 and muldi3 using
I48MODE mode iterator.
(mulv<mode>3): Macroize insn from mulvsi3 and mulvdi3 using
I48MODE mode iterator.
(*iornot<mode>3): Macroize insn from *iornotsi3 and *iornotdi3 using
I48MODE mode iterator.
(*xornot<mode>3): Macroize insn from *xornotsi3 and *xornotdi3 using
I48MODE mode iterator.
Terry Guo [Thu, 11 Oct 2012 02:22:48 +0000 (02:22 +0000)]
arm.c (arm_arch6m): New variable to denote armv6-m architecture.
2012-10-11 Terry Guo <terry.guo@arm.com>
* config/arm/arm.c (arm_arch6m): New variable to denote armv6-m
architecture.
* config/arm/arm.h (TARGET_HAVE_DMB): The armv6-m also has DMB
instruction.
acinclude.m4 (_gcc_COMPUTE_GAS_VERSION): Allow a single character to quote the VERSION= contents.
* acinclude.m4 (_gcc_COMPUTE_GAS_VERSION): Allow a single
character to quote the VERSION= contents. Sanity-check contents.
* configure.ac ("what linker to use" ld version extraction): Ditto.
* configure: Regenerate.
Mark Kettenis [Wed, 10 Oct 2012 23:11:54 +0000 (23:11 +0000)]
2012-10-10 Mark Kettenis <kettenis@openbsd.org>
* configure.host (*-*-openbsd*) Set cpu_include_dir.
* config/os/bsd/openbsd/ctype_base.h: New file.
* config/os/bsd/openbsd/ctype_configure_char.cc: New file.
* config/os/bsd/openbsd/ctype_inline.h: New file.
* config/os/bsd/openbsd/os_defines.h: New file.
* acinclude.m4 (GLIBCXX_ENABLE_CLOCALE): Use newlib locale model
for OpenBSD.
* configure: Regenerated.
Dodji Seketeli [Wed, 10 Oct 2012 10:43:53 +0000 (10:43 +0000)]
PR c++/53540 - using fails to be equivalent to typedef
In the example of this problem report, during the substituting of int
into 'function', tsubst_aggr_type fails for the alias ctxt1. This is
because TYPE_TEMPLATE_INFO looks for the TEMPLATE_INFO of the ctxt1
alias at the wrong place and was wrongly finding it to be NULL.
Namely, it was looking for it in the DECL_TEMPLATE_INFO of the
declaration of the type -- as if ctxt1 was an alias template
specialization -- rather than looking for it in its
CLASSTYPE_TEMPLATE_INFO.
Fixed thus. The other hunks of the patch are a cleanup to make a
better use of alias_template_specialization_p.
Bootstrapped and tested on x86_64-unknown-linux-gnu against trunk.
gcc/cp
* cp-tree.h (TYPE_TEMPLATE_INFO): For an alias that is not an
instance of alias template, don't look for its TEMPLATE_INFO in
its declaration.
(alias_template_specialization_p): Take const_tree.
* pt.c (alias_template_specialization_p): Take a const_tree.
Don't call primary_template_instantiation_p.
(primary_template_instantiation_p): Call
alias_template_specialization_p.
Dodji Seketeli [Wed, 10 Oct 2012 10:25:03 +0000 (10:25 +0000)]
PR middle-end/54860 - Make sure attributes hash table is created
On targets cris-elf, alpha and sparc (for instance) it can happen that
the attribute_tables variable is empty for fortran. Thus
register_scoped_attributes (called by init_attributes) won't call
register_scoped_attributes, so the hash table member of
scoped_attributes is not created.
Later when we try to e.g, lookup an attribute by calling
lookup_scoped_attribute_spec, that NULL member hash table comes to
byte us as htab_find_with_hash crashes.
This patch fixes this by ensuring in register_scoped_attributes that
the hash table is created.
Tested on cris-elf, x86_64-unknown-linux-gnu against trunk and some
commenters on the bug bootstrapped it on alpha and sparc.
gcc/
* attribs.c (register_scoped_attributes): Ensure the attribute
hash table is created.
Dodji Seketeli [Wed, 10 Oct 2012 10:24:50 +0000 (10:24 +0000)]
Update g++.dg/cpp0x/gen-attrs-{8,36,37}.C as c++11 attributes to types are ignored
The current implementation of C++11 attributes forbids them from being
applied to a type unless the type is being declared. I forgot to
adjust g++.dg/cpp0x/gen-attrs-{8,36,37}.C that was being run only on
ia32.
Fixed thus, tested on i386-unknown-linux-gnu and
x86_64-unknown-linux-gnu against trunk.
gcc/testsuite/
* g++.dg/cpp0x/gen-attrs-8.C: Update the test to reflect the fact
that c++11 attributes to types are ignored for now.
* g++.dg/cpp0x/gen-attrs-36.C: Likewise.
* g++.dg/cpp0x/gen-attrs-37.C: Likewise
Dodji Seketeli [Wed, 10 Oct 2012 10:24:35 +0000 (10:24 +0000)]
Disambiguate nested objc-message-expressions and c++11 attributes
A couple of obj-c++ tests were failing[1] because the tokens '[[' can
either be the beginning of a c++11 attribute (that is itself at the
beginning of a statement), or the beginning of a nested
objc-message-expression. This patch resolves the ambiguity by
tentatively parsing the c++11 attribute and if it fails, then consider
the objc-message-expression.
I missed this initially because it didn't occur to me that
--enable-languages=all,ada does not include obj-c++. Shame on me. I
have now updated my compile farm scripts to use
--enable-language=all,ada,obj-c++,go and I
[1]:
FAIL: obj-c++.dg/syntax-error-6.mm -fgnu-runtime (test for errors, line 11)
FAIL: obj-c++.dg/syntax-error-6.mm -fgnu-runtime (test for excess errors)
FAIL: obj-c++.dg/template-8.mm -fgnu-runtime (test for excess errors)
Tested on x86_64-unknown-linux-gnu against trunk.
gcc/cp/
* parser (cp_parser_statement): Parse c++11 attributes tentatively.
(cp_parser_std_attribute_spec_seq): Do not warn too early about
using c++11 attributes in non c++11 mode.