Jackson Woodruff [Fri, 24 Nov 2017 16:03:13 +0000 (16:03 +0000)]
Factor out division by squares
This patch implements the some of the division optimizations discussed in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71026.
The division reciprocal optimization now handles divisions by squares:
x / (y * y) -> x * (1 / y) * (1 / y)
This requires at least one more division by y before it triggers - the
3 divisions of (1/ y) are then CSEd into a single division. Overall
this changes 1 division into 1 multiply, which is generally much faster.
2017-11-24 Jackson Woodruff <jackson.woodruff@arm.com>
gcc/
PR tree-optimization/71026
* tree-ssa-math-opts (is_division_by_square, is_square_of): New.
(insert_reciprocals): Change to insert reciprocals before a division
by a square and to insert the square of a reciprocal.
(execute_cse_reciprocals_1): Change to consider division by a square.
(register_division_in): Add importance parameter.
testsuite/
PR tree-optimization/71026
* gfortran.dg/extract_recip_1.f: New test.
* gcc.dg/extract_recip_3.c: New test.
* gcc.dg/extract_recip_4.c: New test.
Jakub Jelinek [Fri, 24 Nov 2017 08:34:13 +0000 (09:34 +0100)]
tree-object-size.c (pass_through_call): Use gimple_call_return_flags ERF_RETURN*ARG* for builtins other than...
* tree-object-size.c (pass_through_call): Use gimple_call_return_flags
ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
check for the latter with gimple_call_builtin_p. Do not handle
BUILT_IN_STPNCPY_CHK which is not a pass through call.
* config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
fixes. Declare temp and g variables at the top in order to avoid
{} in most of the cases.
Jakub Jelinek [Thu, 23 Nov 2017 20:18:10 +0000 (21:18 +0100)]
parser.c (cp_parser_omp_declare): Change return type to bool from void, return true for declare simd.
* parser.c (cp_parser_omp_declare): Change return type to bool from
void, return true for declare simd.
(cp_parser_pragma): Return cp_parser_omp_declare returned value
rather than always false.
Charles Baylis [Thu, 23 Nov 2017 18:39:44 +0000 (18:39 +0000)]
This patch adds support for modelling the varying costs of different addressing modes.
This patch adds support for modelling the varying costs of
different addressing modes. The generic cost table treats
all addressing modes as having equal cost.
gcc/ChangeLog:
2017-11-23 Charles Baylis <charles.baylis@linaro.org>
Charles Baylis [Thu, 23 Nov 2017 18:37:18 +0000 (18:37 +0000)]
[ARM] Refactor costs calculation for MEM.
This patch moves the calculation of costs for MEM into a
separate function, and reforms the calculation into two
parts. Firstly any additional cost of the addressing mode
is calculated, and then the cost of the memory access itself
is added.
In this patch, the calculation of the cost of the addressing
mode is omitted, to be added in a subsequent patch.
gcc/ChangeLog:
<date> Charles Baylis <charles.baylis@linaro.org>
* config/arm/arm.c (arm_mem_costs): New function.
(arm_rtx_costs_internal): Use arm_mem_costs.
gcc/testsuite/ChangeLog:
<date> Charles Baylis <charles.baylis@linaro.org>
* gcc.target/arm/addr-modes-float.c: New test.
* gcc.target/arm/addr-modes-int.c: New test.
* gcc.target/arm/addr-modes.h: New header.
Mark Wielaard [Thu, 23 Nov 2017 16:56:05 +0000 (16:56 +0000)]
Fix .debug_rnglists generation with -gdwarf-5 -gsplit-dwarf.
Early debug broke generation of .debug_rnglists when using both -gdwarf5
and -gsplit-dwarf. It introduces a generation for init_sections_and_labels,
but doesn't account for the generation of up to 4 unique ranges labels,
two created in init_sections_and_labels and two in output_rnglists.
Fix this by passing generation to output_rnglists and creating 4 unique
labels per generation.
Without this fix using -gdwarf-5 -gsplit-dwarf could result in:
Error: symbol `.Ldebug_ranges2' is already defined
gcc/ChangeLog:
* dwarf2out.c (init_sections_and_labels): Use generation to create
unique ranges_section_label and ranges_base_label. Return generation.
(output_rnglists): Add generation argument. Use generation to create
unique ranges labels.
(dwarf2out_finish): Get generation from init_sections_and_labels
and pass generation to output_rnglists.
* doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
* cfgloop.h (struct loop): Add unroll field.
* function.h (struct function): Add has_unroll bitfield.
* gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
(gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
* loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
(pass_rtl_unroll_loops::gate): Likewise.
* loop-unroll.c (decide_unrolling): Tweak note message. Skip loops
for which loop->unroll==1.
(decide_unroll_constant_iterations): Use note for consistency and
take loop->unroll into account. Return early if loop->unroll is set.
Fix thinko in existing test.
(decide_unroll_runtime_iterations): Use note for consistency and
take loop->unroll into account.
(decide_unroll_stupid): Likewise.
* lto-streamer-in.c (input_cfg): Read loop->unroll.
* lto-streamer-out.c (output_cfg): Write loop->unroll.
* tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
New case.
(replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
(print_loop): Print loop->unroll if set.
* tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
* tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
* tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
New case.
* tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
loop->unroll is set and smaller than the trip count. Otherwise bypass
entirely the heuristics if loop->unroll is set. Remove dead note.
Fix off-by-one bug in other note.
(try_peel_loop): Bail out if loop->unroll is set. Fix formatting.
(tree_unroll_loops_completely_1): Force unrolling if loop->unroll
is greater than 1.
(tree_unroll_loops_completely): Make static.
(pass_complete_unroll::execute): Use correct type for variable.
(pass_complete_unrolli::execute): Fix formatting.
* tree.def (ANNOTATE_EXPR): Add 3rd operand.
ada/
* gcc-interface/trans.c (gnat_gimplify_stmt) <LOOP_STMT>: Pass 3rd
operand to ANNOTATE_EXPR and also pass unrolling hints.
c/
* c-parser.c (c_parser_while_statement): Pass 3rd operand to
ANNOTATE_EXPR.
(c_parser_do_statement): Likewise.
(c_parser_for_statement): Likewise.
cp/
* pt.c (tsubst_expr) <ANNOTATE_EXPR>: Recurse on 3rd operand.
* semantics.c (finish_while_stmt_cond): Pass 3rd operand to
ANNOTATE_EXPR.
(finish_do_stmt): Likewise.
(finish_for_cond): Likewise.
fortran/
* trans-stmt.c (gfc_trans_forall_loop): Pass 3rd operand to
ANNOTATE_EXPR.
Jakub Jelinek [Thu, 23 Nov 2017 11:26:32 +0000 (12:26 +0100)]
re PR middle-end/82253 (ICE in convert_move, at expr.c:604)
PR middle-end/82253
* expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
bitpos/bitsize covering the whole destination, use store_expr only if
the complex mode is the same. Otherwise, use expand_normal and if
it returns CONCAT, subreg each part separately instead of trying to
subreg the whole result.
Paul Thomas [Thu, 23 Nov 2017 09:52:04 +0000 (09:52 +0000)]
re PR fortran/82814 (ICE from submodule character function)
2017-11-23 Paul Thomas <pault@gcc.gnu.org>
PR fortran/82814
* trans-types.c (gfc_sym_type): If a character function result
is missing the charlen backend_decl, use the one from the name-
space procedure symbol, if present.
2017-11-23 Paul Thomas <pault@gcc.gnu.org>
PR fortran/82814
* gfortran.dg/submodule_31.f08: New test.
Richard Biener [Thu, 23 Nov 2017 09:05:11 +0000 (09:05 +0000)]
re PR tree-optimization/23094 (store ccp, or store copy prop misses an optimization)
2017-11-23 Richard Biener <rguenther@suse.de>
PR tree-optimization/23094
* tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
come here from walking over backedges in the first iteration.
(vn_reference_lookup_3): Skip clobbers that store the same value.
Richard Biener [Thu, 23 Nov 2017 08:30:41 +0000 (08:30 +0000)]
re PR tree-optimization/81403 (wrong code at -O3)
2017-11-23 Richard Biener <rguenther@suse.de>
PR tree-optimization/81403
* tree-ssa-pre.c (get_representative_for): Add parameter specifying
a block we need a leader relative to.
(phi_translate_1): For nary processing require a leader from
get_representative_for given we run expression simplification
using match-and-simplify. Remove previous fix.
Jeff Law [Thu, 23 Nov 2017 00:04:07 +0000 (17:04 -0700)]
gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range): Use new method allocate_value_range rather than accessing the...
* gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
Use new method allocate_value_range rather than accessing the
vrp_value_range_pool data member directly.
* tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
to use extract_range_from_stmt method to avoid need for
extract_range_from_assignment method.
(vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
method rather than setting values_propgated data member directly.
* vr-values.h (class vr_values): Privatize vrp_value_range_pool,
and values propagated data members and extract_range_from_assignment
method. Reorder private data members to conform to standards.
Add new methods set_lattice_propagation_complete and
allocate_value_range.
Eric Botcazou [Wed, 22 Nov 2017 21:43:22 +0000 (21:43 +0000)]
re PR rtl-optimization/83030 (ICE in create_pseudo_cfg, at dwarf2cfi.c:2840)
PR rtl-optimization/83030
* doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
(Insns): Delete entry for REG_CROSSING_JUMP in register notes.
* bb-reorder.c (update_crossing_jump_flags): Do not test whether the
CROSSING_JUMP_P flag is already set before setting it.
* cfgrtl.c (fixup_partition_crossing): Likewise.
* reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P insn
as useless.
Janne Blomqvist [Wed, 22 Nov 2017 20:42:39 +0000 (22:42 +0200)]
Make shift argument to eoshift0 and eoshift2 be of type index_type
Test case that requires 4 GB RAM that exercises eoshift0:
program main
implicit none
integer(1) :: line(2_8**31+10)
line = 42
line = eoshift(line, 2_8**31+7)
print *, line(1:4)
print *, line((ubound(line, 1, kind=8)-4):ubound(line, 1, kind=8))
end program main
Before patch:
❯ ./a.out
0 0 0 0
42 42 42 42 42
With patch:
❯ ./a.out
42 42 42 0
0 0 0 0 0
Regtested on x86_64-pc-linux-gnu.
libgfortran/ChangeLog:
2017-11-22 Janne Blomqvist <jb@gcc.gnu.org>
* intrinsics/eoshift0.c (eoshift0): Make shift an index_type.
* intrinsics/eoshift2.c (eoshift2): Make shift an index_type.
David Malcolm [Wed, 22 Nov 2017 20:37:58 +0000 (20:37 +0000)]
C/C++: fix quoting of "aka" typedef information (PR 62170)
PR 62170 describes a problem with how the quoting in pp_format
interacts with the "aka" information for typedefs in %qT for
the C family of frontends, and also now for %qH and %qI in the
C++ frontend: we print:
'Py_ssize_t* {aka int*}'
^^^^^^^^^^^^^^^^^^^^^^ colorized as "quote"
i.e.
'[START_COLOR]Py_ssize_t* {aka int*}[END_COLOR]'
when we should print:
'Py_ssize_t*' {aka 'int*'}
^^^^^^^^^^^ ^^^^ colorized as "quote"
i.e.
'[START_COLOR]Py_ssize_t*[END_COLOR]' {aka '[START_COLOR]int*[END_COLOR]'}
where the opening and closing quote characters and colorization are
added by the 'q' handling within pp_format.
This patch fixes the quoting by updating the %T handling in C and C++
and the %H/%I handling in C++ to insert the quoting appropriately.
It converts the "quote" param of the pp_format_decoder callback from
bool to bool *, allowing for the %T and %H/%I handlers to write
false back to it, to avoid printing the closing quote for the cases
like the above where the trailing closing quote isn't needed.
It introduces pp_begin_quote/pp_end_quote to simplify this. These
take a "bool show_color", rather than using "pp_show_color (pp)"
since cxx_pp's pp_show_color isn't currently initialized (since
cxx_initialize_diagnostics happens before diagnostic_color_init).
gcc/c/ChangeLog:
PR c++/62170
* c-objc-common.c (c_tree_printer): Convert penultimate param from
bool to bool *. Within '%T' handling, if showing an "aka", use
"quoted" param to add appropriate quoting.
gcc/cp/ChangeLog:
PR c++/62170
* error.c (type_to_string): Add leading comment. Add params
"postprocessed", "quote", and "show_color", using them to fix
quoting of the "aka" for types involving typedefs.
(arg_to_string): Update for new params to type_to_string.
(cxx_format_postprocessor::handle): Likewise.
(cp_printer): Convert penultimate param from bool to bool *.
Update call to type_to_string and calls to
defer_phase_2_of_type_diff.
gcc/fortran/ChangeLog:
PR c++/62170
* error.c (gfc_notify_std): Convert "quoted" param from bool to
bool *.
gcc/ChangeLog:
PR c++/62170
* pretty-print.c (pp_format): Move quoting implementation to
pp_begin_quote and pp_end_quote. Update pp_format_decoder call
to pass address of "quote" local.
(pp_begin_quote): New function.
(pp_end_quote): New function.
* pretty-print.h (printer_fn): Convert penultimate param from bool
to bool *.
(pp_begin_quote): New decl.
(pp_end_quote): New decl.
* tree-diagnostic.c (default_tree_printer): Convert penultimate
param from bool to bool *.
* tree-diagnostic.h (default_tree_printer): Likewise.
gcc/testsuite/ChangeLog:
PR c++/62170
* g++.dg/diagnostic/aka1.C: Update expected error messages to
reflect fixes to quoting.
* g++.dg/diagnostic/aka2.C: New test case.
* g++.dg/parse/error55.C: Update expected error messages to
reflect fixes to quoting.
* gcc.dg/diag-aka-1.c: Likewise.
* gcc.dg/diag-aka-2.c: New test case.
* gcc.dg/pr13804-1.c: Update expected error messages to reflect
fixes to quoting.
* gcc.dg/pr56980.c: Likewise.
* gcc.dg/pr65050.c: Likewise.
* gcc.dg/redecl-14.c: Likewise.
* gcc.dg/utf16-4.c Likewise.
* gcc.target/i386/sse-vect-types.c (__m128d): Likewise.
* obj-c++.dg/invalid-type-1.mm: Likewise.
* objc.dg/proto-lossage-4.m: Likewise.
Jeff Law [Wed, 22 Nov 2017 20:31:14 +0000 (13:31 -0700)]
* gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
Initialize vr_values.
(evrp_range_analyzer::try_find_new_range): Call methods attached to
vr_values via vr_values class instance rather than delegators.
(evrp_range_analyzer::record_ranges_from_phis): Likewise.
(evrp_range_analyzer::record_ranges_from_stmt): Likewise.
(evrp_range_analyzer::push_value_range): Likewise.
(evrp_range_analyzer::pop_value_range): Likewise.
* gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
most delegators. Those remaining are exposed as public interfaces.
Make vr_values a pointer and private.
(evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
vr_values.
(evrp_range_analyzer::get_vr_value): New method.
* gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
(evrp_folder::evrp_folder): New ctor to initialize vr_values.
(class evrp_dom_walker): Attach evrp_folder class, initialize
it in the ctor. Remove temporary delegators.
(evrp_dom_walker::before_dom_children): Call methods in attached
evrp_range_analyzer class via class instance pointer. Use
free value_range_constant_singleton to remove need for
op_with_constant_singleton_value delegator method. Do not
create a vrp_prop class instance for every call! Narrow
scope of a couple variables.
(evrp_dom_walker::cleanup): Call methods in attached
evrp_range_analyzer class via class instance pointer.
* vr-values.h (class vr_values): Privatize many methods and
data members.
This patch replaces the REDUC_*_EXPR tree codes with internal functions.
This is needed so that the upcoming in-order reductions can also use
internal functions without too much complication.
2017-11-22 Richard Sandiford <richard.sandiford@linaro.org>
gcc/
* tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
* cfgexpand.c (expand_debug_expr): Remove handling for them.
* expr.c (expand_expr_real_2): Likewise.
* fold-const.c (const_unop): Likewise.
* optabs-tree.c (optab_for_tree_code): Likewise.
* tree-cfg.c (verify_gimple_assign_unary): Likewise.
* tree-inline.c (estimate_operator_cost): Likewise.
* tree-pretty-print.c (dump_generic_node): Likewise.
(op_code_prio): Likewise.
(op_symbol_code): Likewise.
* internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
(IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
* internal-fn.c (direct_internal_fn_optab): New function.
(direct_internal_fn_array, direct_internal_fn_supported_p
(internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
* fold-const-call.c (fold_const_reduction): New function.
(fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
CFN_REDUC_MIN.
* tree-vect-loop.c: Include internal-fn.h.
(reduction_code_for_scalar_code): Rename to...
(reduction_fn_for_scalar_code): ...this and return an internal
function.
(vect_model_reduction_cost): Take an internal_fn rather than
a tree_code.
(vect_create_epilog_for_reduction): Likewise. Build calls rather
than assignments.
(vectorizable_reduction): Use internal functions rather than tree
codes for the reduction operation. Update calls to the functions
above.
* config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
Use calls to internal functions rather than REDUC tree codes.
* config/aarch64/aarch64-simd.md: Update comment accordingly.
Thomas Koenig [Wed, 22 Nov 2017 18:08:07 +0000 (18:08 +0000)]
re PR fortran/36313 ([F03] {MIN,MAX}{LOC,VAL} should accept character arguments)
2017-11-22 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/36313
* Makefile.am: Add i_maxloc0s_c, i_maxloc1s_c, i_maxloc2s_c,
i_minloc0s_c, i_minloc1s_c and i_minloc2s_c.
* Makefile.in: Regenerated.
* generated/maxloc0_16_s1.c: New file.
* generated/maxloc0_16_s4.c: New file.
* generated/maxloc0_4_s1.c: New file.
* generated/maxloc0_4_s4.c: New file.
* generated/maxloc0_8_s1.c: New file.
* generated/maxloc0_8_s4.c: New file.
* generated/maxloc1_16_s1.c: New file.
* generated/maxloc1_16_s4.c: New file.
* generated/maxloc1_4_s1.c: New file.
* generated/maxloc1_4_s4.c: New file.
* generated/maxloc1_8_s1.c: New file.
* generated/maxloc1_8_s4.c: New file.
* generated/maxloc2_16_s1.c: New file.
* generated/maxloc2_16_s4.c: New file.
* generated/maxloc2_4_s1.c: New file.
* generated/maxloc2_4_s4.c: New file.
* generated/maxloc2_8_s1.c: New file.
* generated/maxloc2_8_s4.c: New file.
* generated/minloc0_16_s1.c: New file.
* generated/minloc0_16_s4.c: New file.
* generated/minloc0_4_s1.c: New file.
* generated/minloc0_4_s4.c: New file.
* generated/minloc0_8_s1.c: New file.
* generated/minloc0_8_s4.c: New file.
* generated/minloc1_16_s1.c: New file.
* generated/minloc1_16_s4.c: New file.
* generated/minloc1_4_s1.c: New file.
* generated/minloc1_4_s4.c: New file.
* generated/minloc1_8_s1.c: New file.
* generated/minloc1_8_s4.c: New file.
* generated/minloc2_16_s1.c: New file.
* generated/minloc2_16_s4.c: New file.
* generated/minloc2_4_s1.c: New file.
* generated/minloc2_4_s4.c: New file.
* generated/minloc2_8_s1.c: New file.
* generated/minloc2_8_s4.c: New file.
* m4/iforeach-s.m4: New file.
* m4/ifunction-s.m4: New file.
* m4/maxloc0s.m4: New file.
* m4/maxloc1s.m4: New file.
* m4/maxloc2s.m4: New file.
* m4/minloc0s.m4: New file.
* m4/minloc1s.m4: New file.
* m4/minloc2s.m4: New file.
* gfortran.map: Add new functions.
* libgfortran.h: Add gfc_array_s1 and gfc_array_s4.
2017-11-22 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/36313
* check.c (int_or_real_or_char_check_f2003): New function.
* iresolve.c (gfc_resolve_maxloc): Add number "2" for
character arguments and rank-zero return value.
(gfc_resolve_minloc): Likewise.
* trans-intrinsic.c (gfc_conv_intrinsic_minmaxloc): Handle case of
character arguments and rank-zero return value by removing
unneeded arguments and calling the library function.
2017-11-22 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/36313
* gfortran.dg/maxloc_string_1.f90: New test.
* gfortran.dg/minloc_string_1.f90: New test.
Marek Polacek [Wed, 22 Nov 2017 16:06:18 +0000 (16:06 +0000)]
re PR c++/60336 (empty struct value is passed differently in C and C++)
PR c++/60336
PR middle-end/67239
PR target/68355
* c-decl.c (grokdeclarator): Set DECL_PADDING_P on unnamed bit-fields.
* class.c (layout_class_type): Set DECL_PADDING_P on padding.
* decl.c (cxx_init_decl_processing): Set TRANSLATION_UNIT_WARN_EMPTY_P.
(grokdeclarator): Set DECL_PADDING_P on unnamed bit-fields.
* lto.c (compare_tree_sccs_1): Compare TYPE_EMPTY_P and DECL_PADDING_P.
* calls.c (initialize_argument_information): Call
warn_parameter_passing_abi target hook.
(store_one_arg): Use 0 for empty record size. Don't push 0 size
argument onto stack.
(must_pass_in_stack_var_size_or_pad): Return false for empty types.
* common.opt: Update -fabi-version description.
* config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
(ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
int_size_in_bytes.
(ix86_is_empty_record): New function.
(ix86_warn_parameter_passing_abi): New function.
(TARGET_EMPTY_RECORD_P): Redefine.
(TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
* config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
* doc/tm.texi: Regenerated.
* doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
TARGET_WARN_PARAMETER_PASSING_ABI): Add.
* dwarf2out.c (get_ultimate_context): Move to tree.c.
* explow.c (hard_function_value): Call arg_int_size_in_bytes
instead of int_size_in_bytes.
* expr.c (copy_blkmode_to_reg): Likewise.
* function.c (aggregate_value_p): Return 0 for empty types.
(assign_parm_find_entry_rtl): Call warn_parameter_passing_abi target hook.
(locate_and_pad_parm): Call arg size_in_bytes instead
size_in_bytes.
* lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
* stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
* target.def (empty_record_p, warn_parameter_passing_abi): New target
hooks.
* targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
(std_gimplify_va_arg_expr): Skip empty records. Call
arg_size_in_bytes instead size_in_bytes.
* targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
* tree-core.h (tree_type_common): Add empty_flag.
(tree_decl_common): Update comments.
* tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
DECL_PADDING_P.
(unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
* tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
DECL_PADDING_P.
(pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
* tree.c (default_is_empty_type): New function.
(default_is_empty_record): New function.
(arg_int_size_in_bytes): New function.
(arg_size_in_bytes): New function.
(get_ultimate_context): New function.
* tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
TRANSLATION_UNIT_WARN_EMPTY_P.
(default_is_empty_record, arg_int_size_in_bytes,
arg_size_in_bytes, get_ultimate_context): Declare.
* g++.dg/abi/empty12.C: New test.
* g++.dg/abi/empty12.h: New test.
* g++.dg/abi/empty12a.c: New test.
* g++.dg/abi/empty13.C: New test.
* g++.dg/abi/empty13.h: New test.
* g++.dg/abi/empty13a.c: New test.
* g++.dg/abi/empty14.C: New test.
* g++.dg/abi/empty14.h: New test.
* g++.dg/abi/empty14a.c: New test.
* g++.dg/abi/empty15.C: New test.
* g++.dg/abi/empty15.h: New test.
* g++.dg/abi/empty15a.c: New test.
* g++.dg/abi/empty16.C: New test.
* g++.dg/abi/empty16.h: New test.
* g++.dg/abi/empty16a.c: New test.
* g++.dg/abi/empty17.C: New test.
* g++.dg/abi/empty17.h: New test.
* g++.dg/abi/empty17a.c: New test.
* g++.dg/abi/empty18.C: New test.
* g++.dg/abi/empty18.h: New test.
* g++.dg/abi/empty18a.c: New test.
* g++.dg/abi/empty19.C: New test.
* g++.dg/abi/empty19.h: New test.
* g++.dg/abi/empty19a.c: New test.
* g++.dg/abi/empty20.C: New test.
* g++.dg/abi/empty21.C: New test.
* g++.dg/abi/empty22.C: New test.
* g++.dg/abi/empty22.h: New test.
* g++.dg/abi/empty22a.c: New test.
* g++.dg/abi/empty23.C: New test.
* g++.dg/abi/empty24.C: New test.
* g++.dg/abi/empty25.C: New test.
* g++.dg/abi/empty25.h: New test.
* g++.dg/abi/empty25a.c: New test.
* g++.dg/abi/empty26.C: New test.
* g++.dg/abi/empty26.h: New test.
* g++.dg/abi/empty26a.c: New test.
* g++.dg/abi/empty27.C: New test.
* g++.dg/abi/empty28.C: New test.
* g++.dg/abi/pr60336-1.C: New test.
* g++.dg/abi/pr60336-10.C: New test.
* g++.dg/abi/pr60336-11.C: New test.
* g++.dg/abi/pr60336-12.C: New test.
* g++.dg/abi/pr60336-2.C: New test.
* g++.dg/abi/pr60336-3.C: New test.
* g++.dg/abi/pr60336-4.C: New test.
* g++.dg/abi/pr60336-5.C: New test.
* g++.dg/abi/pr60336-6.C: New test.
* g++.dg/abi/pr60336-7.C: New test.
* g++.dg/abi/pr60336-8.C: New test.
* g++.dg/abi/pr60336-9.C: New test.
* g++.dg/abi/pr68355.C: New test.
* g++.dg/lto/pr60336_0.C: New test.
Co-Authored-By: H.J. Lu <hongjiu.lu@intel.com> Co-Authored-By: Jason Merrill <jason@redhat.com>
From-SVN: r255066
Functions cmse_nonsecure_call_clear_caller_saved and
cmse_nonsecure_entry_clear_before_return both contain very similar code
to clear registers. What's worse, they differ slightly at times so if a
bug is found in one careful thoughts is needed to decide whether the
other function needs fixing too.
This commit addresses the situation by factoring the two pieces of code
into a new function. In doing so the code generated to clear VFP
registers in cmse_nonsecure_call now uses the same sequence as
cmse_nonsecure_entry functions. Tests expectation are thus updated
accordingly.
2017-11-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
gcc/
* config/arm/arm.c (cmse_clear_registers): New function.
(cmse_nonsecure_call_clear_caller_saved): Replace register clearing
code by call to cmse_clear_registers.
(cmse_nonsecure_entry_clear_before_return): Likewise.
PR82547: Undetected overflow for UNSIGNED wide_ints
wi::add_large and wi::sub_large weren't setting the overflow bit
correctly for unsigned operations if the result needed fewer HWIs
than the precision.
2017-11-22 Richard Sandiford <richard.sandiford@linaro.org>
gcc/
PR middle-end/82547
* wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
for unsigned values with fewer HWIs than the precision.
(test_overflow): New function.
(wide_int_cc_tests): Call it.
arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of padding_bits_to_clear_ptr.
2017-11-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
gcc/
* config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
padding_bits_to_clear_ptr.
(cmse_nonsecure_entry_clear_before_return): Likewise.
[ARM] Use bitmap to control cmse_nonsecure_call register clearing
As part of r253256, cmse_nonsecure_entry_clear_before_return has been
rewritten to use auto_sbitmap instead of an integer bitfield to control
which register needs to be cleared. This commit continues this work in
cmse_nonsecure_call_clear_caller_saved.
2017-11-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
gcc/
* config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
auto_sbitap instead of integer bitfield to control register needing
clearing.
Jakub Jelinek [Wed, 22 Nov 2017 12:35:26 +0000 (13:35 +0100)]
re PR tree-optimization/83044 (ice in contains_struct_check)
PR tree-optimization/83044
* tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
that need the upper bound. Subtract offset from
get_addr_base_and_unit_offset only if positive and subtract it
before division by eltsize rather than after it.
* gcc.dg/pr83044.c: New test.
* c-c++-common/Warray-bounds.c (fb): Fix up MAX value.
Jakub Jelinek [Wed, 22 Nov 2017 10:02:19 +0000 (11:02 +0100)]
re PR debug/83084 (-fcompare-debug failure on ppc64le)
PR debug/83084
* valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
(dead_debug_insert_temp): Likewise, but also ignore even non-volatile
asm.
Jakub Jelinek [Wed, 22 Nov 2017 09:57:28 +0000 (10:57 +0100)]
re PR middle-end/82875 (ICE at -Os on valid code on x86_64-linux-gnu: in find_widening_optab_handler_and_mode, at optabs-query.c:414)
PR middle-end/82875
* optabs.c (expand_doubleword_mult, expand_binop): Before calling
expand_binop with *mul_widen_optab, make sure at least one of the
operands doesn't have VOIDmode.
* gcc.dg/pr82875.c: New test.
* gcc.c-torture/compile/pr82875.c: New test.
Jakub Jelinek [Wed, 22 Nov 2017 09:08:23 +0000 (10:08 +0100)]
re PR rtl-optimization/82044 (runtime signed integer overflow in check_mem_read_rtx() and all_positions_needed_p() in dse.c)
PR rtl-optimization/82044
PR tree-optimization/82042
* dse.c (record_store): Check for overflow.
(check_mem_read_rtx): Properly check for overflow if width == -1, call
add_wild_read instead of clear_rhs_from_active_local_stores on
overflow and log it into dump_file.
Richard Biener [Wed, 22 Nov 2017 09:05:53 +0000 (09:05 +0000)]
gimple-iterator.c (gimple_find_edge_insert_loc): Ignore fake edges to exit when looking for a place to insert.
2017-11-22 Richard Biener <rguenther@suse.de>
* gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
fake edges to exit when looking for a place to insert.
* tree-ssa-pre.c (clear_expression_ids): Inline into callers
and remove.
(insert_into_preds_of_block): Commit edge insertion immediately,
assert that doesn't require new BBs.
(fini_pre): Release expressions.
(pass_pre::execute): Shuffle things around a bit, if the fn
is too large do not compute AVAIL either as this is really the
quadratic bit.
David Malcolm [Tue, 21 Nov 2017 21:59:53 +0000 (21:59 +0000)]
C: don't suggest names that came from earlier failures (PR c/83056)
PR c/83056 reports an issue affecting trunk and gcc-7 in which
the C frontend's implementation of lookup_name_fuzzy uses undeclared
identifiers as suggestions when encountering subsequent undeclared
identifiers.
The fix is to filter out the names bound to error_mark_node
in lookup_name_fuzzy.
The C++ frontend is unaffected, as it already does this.
gcc/c/ChangeLog:
PR c/83056
* c-decl.c (lookup_name_fuzzy): Don't suggest names that came from
earlier failed lookups.
gcc/testsuite/ChangeLog:
PR c/83056
* gcc.dg/spellcheck-pr83056.c: New test case.
Sergey Shalnov [Tue, 21 Nov 2017 19:52:01 +0000 (19:52 +0000)]
i386-opts.h (enum prefer_vector_width): Added new enum for the new option -mprefer-vector-width=[none|128|256|512].
* config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
for the new option -mprefer-vector-width=[none|128|256|512].
* config/i386/i386.c (ix86_target_string): remove old style options
-mprefer-avx256 and make -mprefer-avx128 as alias.
(ix86_option_override_internal): Apply defaults for the
-mprefer-vector-width=[128|256] option.
* config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
Implement macros to work with -mprefer-vector-width=.
* config/i386/i386.opt: Implemented option
-mprefer-vector-width=[none|128|256|512].
* doc/invoke.texi: Documentation for
-mprefer-vector-width=[none|128|256|512].
gcc/testsuite/
* g++.dg/ext/pr57362.C (__attribute__): Test
prefer-vector-width=[128|256] target attribute.
* gcc.target/i386/avx512f-constant-float-return.c (dg-optioins):
Use -mprefer-vector-width=256 instead of -mprefer-avx256.
* gcc.target/i386/avx512f-prefer.c: Ditto.
* gcc.target/i386/pr82460-2.c: Ditto.
projects / thirdparty / gcc.git / log
