Add LE/BE SHA1H patterns with a V2SI input. This avoids unnecessary
DUPs when using intrinsics like vsha1h_u32 (vgetq_lane_u32 (x, 0)).
gcc/
* config/aarch64/aarch64-simd.md (aarch64_crypto_sha1hv4si):
New pattern.
(aarch64_be_crypto_sha1hv4si): New pattern.
From-SVN: r241791
The add expander still contains some expansion code that was required for the
previous prolog/epilog code, but which is no longer needed. I also noticed
that the current version splits off immediates from frame addressing
instructions, which doesn't seem a good idea. Avoiding this resulted in small
codesize improvements.
gcc/
* config/aarch64/aarch64.md (add<mode>3): Remove
redundant code. Don't split frame based additions.
From-SVN: r241790
2016-11-02 Richard Biener <rguenther@suse.de>
* gimple-ssa-store-merging.c (struct store_immediate_info): Remove
redundant val and dest members.
(store_immediate_info::store_immediate_info): Adjust.
(merged_store_group::merged_store_group): Adjust.
(merged_store_group::apply_stores): Likewise.
(struct imm_store_chain_info): Add base_addr field.
(imm_store_chain_info::imm_store_chain_info): New constructor.
(imm_store_chain_info::terminate_and_process_chain): Do not pass base.
(imm_store_chain_info::output_merged_store): Likewise. Use
addr_base which is already the address.
(imm_store_chain_info::output_merged_stores): Likewise.
(pass_tree_store_merging::terminate_all_aliasing_chains): Take
imm_store_chain_info instead of base. Fix alias check.
(pass_tree_store_merging::terminate_and_release_chain): Likewise.
(imm_store_chain_info::coalesce_immediate_stores): Adjust.
(pass_store_merging::execute): Refuse to operate on TARGET_MEM_REF.
use the address of the base and adjust for other changes.
From-SVN: r241789
* fold-const-call.c (host_size_t_cst_p): Test whether
t is convertible to size_t.
* gcc.dg/tree-ssa/builtins-folding-gimple-ub.c (main): Add
test case.
From-SVN: r241780
PR tree-optimization/78170
* gimple-ssa-store-merging.c (encode_tree_to_bitpos): Truncate padding
introduced by native_encode_expr on little-endian as well.
* gcc.c-torture/execute/pr78170.c: New test.
From-SVN: r241779
PR tree-optimization/78162
* gimple-ssa-store-merging.c (execute): Mark stores with bitpos < 0
as invalid.
* gcc.c-torture/compile/pr78162.c: New test.
From-SVN: r241778
2016-11-02 Richard Biener <rguenther@suse.de>
PR tree-optimization/78047
* tree-ssa-structalias.c (push_fields_onto_fieldstack): Initialize
fake field at offset zero conservatively regarding to may_have_pointers.
From-SVN: r241775
gcc/c-family/
* c.opt (-fnew-inheriting-ctors): New.
* c-opts.c: Default to on for ABI 11+.
gcc/cp/
* call.c (enum rejection_reason_code): Add rr_inherited_ctor.
(inherited_ctor_rejection): New.
(add_function_candidate): Reject inherited ctors for copying.
(enforce_access): Use strip_inheriting_ctors.
(print_z_candidate): Likewise. Handle rr_inherited_ctor.
(convert_like_real): Avoid copying inheriting ctor parameters.
(build_over_call): Likewise. A base ctor inheriting from vbase
has no parms. Sorry about varargs.
(joust): A local constructor beats inherited with the same convs.
* class.c (add_method): Handle hiding inheriting ctors.
(one_inherited_ctor): Handle new semantics.
(add_implicitly_declared_members): Pass using_decl down.
(build_clone): A base ctor inheriting from vbase has no parms.
* cp-tree.h (DECL_INHERITED_CTOR): Store this instead of the base.
(SET_DECL_INHERITED_CTOR): Likewise.
(DECL_INHERITED_CTOR_BASE): Adjust.
* constexpr.c: Adjust.
* error.c (dump_function_decl): Decorate inheriting ctors.
* init.c (emit_mem_initializers): Suppress access control in
inheriting ctor.
* mangle.c (write_special_name_constructor): Handle new inheriting
ctor mangling.
* method.c (strip_inheriting_ctors, inherited_ctor_binfo)
(ctor_omit_inherited_parms, binfo_inherited_from): New.
(synthesized_method_walk): Use binfo_inherited_from. Suppress
access control in inheriting ctor.
(deduce_inheriting_ctor): Deleted if ambiguous ctor inheritance.
(maybe_explain_implicit_delete): Explain ambigous ctor inheritance.
(add_one_base_init, do_build_copy_constructor): Adjust.
(locate_fn_flags, explain_implicit_non_constexpr): Adjust.
(implicitly_declare_fn): Adjust.
(get_inherited_ctor): Remove.
* name-lookup.c (do_class_using_decl): Check for indirect ctor
inheritance.
* optimize.c (cdtor_comdat_group): Adjust for new mangling.
(maybe_clone_body): Handle omitted parms in base clone.
(maybe_thunk_body): Don't thunk if base clone omits parms.
* pt.c (tsubst_decl): Adjust.
(instantiate_template_1): Suppress access control in inheriting
ctor.
(fn_type_unification): Do deduction with inherited ctor.
* tree.c (special_function_p): Adjust.
gcc/
* tree-inline.c (copy_tree_body_r): Only copy the taken branch of
a COND_EXPR with constant condition.
libiberty/
* cp-demangle.c (d_ctor_dtor_name): Handle inheriting constructor.
From-SVN: r241765
When a symbol cannot be demangled in ada_demangle a new demangled VEC
will be allocated without deleting the demangled VEC already in use.
Running testsuite/test-demangle under valgrind will show the leak for
this entry in testsuite/demangle-expected:
# Elaborated flag (not demangled)
--format=gnat
x_E
<x_E>
11 bytes in 1 blocks are definitely lost in loss record 1 of 1
at 0x4C27BE3: malloc (vg_replace_malloc.c:299)
by 0x413FE7: xmalloc (xmalloc.c:148)
by 0x4025EC: ada_demangle (cplus-dem.c:930)
by 0x402C59: cplus_demangle (cplus-dem.c:892)
by 0x400FEC: main (test-demangle.c:317)
libiberty/ChangeLog:
* cplus-dem.c (ada_demangle): Initialize demangled to NULL and
XDELETEVEC demangled when unknown.
From-SVN: r241760
2016-11-01 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/78178
* match.c (match_simple_where): Fill in locus for assigment
in simple WHERE statement.
2016-11-01 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/69544
* gfortran.dg/where_6.f90: New test.
From-SVN: r241756
* dwarf2out.c (add_name_and_src_coords_attributes): Add NO_LINKAGE_NAME
argument, don't call add_linkage_name if it is true.
(gen_variable_die): For C++ inline static data members, consider the
initial call when old_die is NULL to be declaration and call
add_name_and_src_coords_attributes in that case with true as
NO_LINKAGE_NAME. Add DW_AT_inline attribute if needed.
(gen_member_die): For C++ inline static data members, emit a
definition DIE right away in DW_TAG_compile_unit context.
cp/
* cp-objcp-common.c (cp_decl_dwarf_attribute): Handle DW_AT_inline.
testsuite/
* g++.dg/debug/dwarf2/inline-var-1.C: New test.
From-SVN: r241753
It started failing after the following commit: 32e90dc6a0cda45 ("PR
rtl-optimization/61047").
The change that made xtensa backend go ICE looks completely unrelated,
and indeed, the issue is caused by the side effect of
compute_frame_size() function call hidden in the
INITIAL_ELIMINATION_OFFSET macro. This call updates the value of the
xtensa_current_frame_size static variable, used in "return" instruction
predicate. Prior to the change the value of xtensa_current_frame_size was
set to 0 after the end of epilogue generation, which enabled the "return"
instruction for the CALL0 ABI, but after the change the additional
INITIAL_ELIMINATION_OFFSET calls make xtensa_current_frame_size non-zero
and "return" pattern unavailable.
Get rid of the global xtensa_current_frame_size and
xtensa_callee_save_size variables by moving them into the
machine_function structure. Implement predicate for the "return" pattern
as a function. Don't communicate completion of epilogue generation
through zeroing of xtensa_current_frame_size, add explicit epilogue_done
variable to the machine_function structure. Don't update stack frame
layout after the completion of reload.
2016-11-01 Max Filippov <jcmvbkbc@gmail.com>
gcc/
* config/xtensa/xtensa-protos.h
(xtensa_use_return_instruction_p): New prototype.
* config/xtensa/xtensa.c (xtensa_current_frame_size,
xtensa_callee_save_size): Remove.
(struct machine_function): Add new fields: current_frame_size,
callee_save_size, frame_laid_out and epilogue_done.
(compute_frame_size, xtensa_expand_prologue,
xtensa_expand_epilogue): Replace xtensa_callee_save_size with
cfun->machine->callee_save_size and xtensa_current_frame_size
with cfun->machine->current_frame_size.
(compute_frame_size): Update cfun->machine->frame_laid_out and
don't update frame layout after reload completion.
(xtensa_expand_epilogue): Set cfun->machine->epilogue_done
instead of zeroing xtensa_current_frame_size.
(xtensa_use_return_instruction_p): New function.
* config/xtensa/xtensa.h (xtensa_current_frame_size): Remove
declaration.
(INITIAL_ELIMINATION_OFFSET): Use return value of
compute_frame_size instead of xtensa_current_frame_size value.
* config/xtensa/xtensa.md ("return" pattern): Use new predicate
function xtensa_use_return_instruction_p instead of inline code.
From-SVN: r241748
* tree.h (BLOCK_IN_COLD_SECTION_P): Define.
* final.c (final_scan_insn): Set BLOCK_IN_COLD_SECTION_P.
* dwarf2out.c (rnglist_idx): New variable.
(struct dw_ranges): Add label, idx and maybe_new_sec fields.
(DEBUG_RNGLISTS_SECTION): Define.
(ranges_base_label): New variable.
(size_of_die) <case dw_val_class_range_list>: If using
DW_FORM_rnglistx, count size of uleb128 of range list index.
(value_format) <case dw_val_class_range_list>: For
-gdwarf-5 -gsplit-dwarf return DW_FORM_rnglistx.
(output_range_list_offset): Handle -gdwarf-5 .debug_rnglists
offsets. Multiply dwarf < 5 offsets by 2 * DWARF_ADDR_SIZE.
(add_ranges_num): Remove useless prototype. Don't multiply
by 2 * DWARF2_ADDR_SIZE. Add maybe_new_sec argument, adjust
for new fields added to dw_ranges struct.
(add_ranges): Add maybe_new_sec argument and pass it
through to add_ranges_num.
(note_rnglist_head): New function.
(add_ranges_by_labels): Pass true as maybe_new_sec to
add_ranges_num, call note_rnglist_head on the head of the list.
(output_ranges): Add function comment. Switch to
.debug_ranges section here and emit .Ldebug_ranges0 label.
(index_rnglists, output_rnglists): New functions.
(gen_subprogram_die): Formatting fixes.
(add_high_low_attributes): Don't divide offsets
by 2 * DWARF2_ADDR_SIZE. Call note_rnglist_head on the
first list element or when pointing into the middle of
a list. Pass true as second argument to add_ranges on the
first block fragment after cold/hot section switch.
(init_sections_and_labels): For -gdwarf-5 use .debug_rnglists
section instead of .debug_ranges. Initialize
ranges_base_label if -gdwarf-5 -gsplit-dwarf.
(dwarf2out_finish): For -gdwarf-5 -gsplit-dwarf call
index_rnglists and add DW_AT_rnglists_base attr. Don't switch
to dwarf_ranges_section here or emit .Ldebug_ranges0 label.
Call output_rnglists for -gdwarf-5.
(dwarf2out_c_finalize): Clear rnglist_idx.
From-SVN: r241747
2016-11-01 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/69544
* match.c (gfc_match_where): Fill in locus for assigment
in simple WHERE statement.
2016-11-01 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/69544
* gfortran.dg/where_5.f90: New test.
From-SVN: r241745
When using cgo -gccgo calls to C.GoString, C.GoStringN, and C.GoBytes
are turned into calls to __go_byte_array_to_string and
__go_string_to_byte_array. Those functions were removed when the string
code was copied from Go 1.7, but we still need them for cgo. While cgo
should be updated, old versions will exist for some time.
Reviewed-on: https://go-review.googlesource.com/32474
From-SVN: r241743
Use mvifdiff and stamp files to track whether a .gox file has changed.
When package A depends on package B, and we rebuild package B, only
rebuild package A if package B's .gox file changes. This is safe
because when package A imports package B it only reads package B's .gox
file. This means that changes that do not affect export information
will not cause dependent packages to be recompiled.
Reviewed-on: https://go-review.googlesource.com/32476
From-SVN: r241742
PR go/78145
compiler: don't put print/println constants into temporaries
It's not necessary, and it breaks setting their type to int64/uint64
when appropriate.
This fixes GCC PR 78145.
Reviewed-on: https://go-review.googlesource.com/32475
From-SVN: r241740
[gcc]
2016-10-31 Michael Meissner <meissner@linux.vnet.ibm.com>
* config/rs6000/vsx.md (VSX_EXTRACT_FL): New iterator for all
binary floating point types supported by the hardware except for
double.
(vsx_xvcvsxwdp_df): Provide scalar result alternative to the
vector instruction for optimizing extracting a SImode from a
V4SImode vector and converting it to floating point.
(vsx_xvcvuxwdp_df): Likewise.
(vsx_extract_si): On ISA 3.0, allow extract target and temporary
registers to be any VSX register. Move stores to the end of the
constraints.
(vsx_extract_si_<uns>float_df): New combiner pattern and splitter
to optimize extracting a SImode from a V4SImode vector and
converting it to a binary floating point type supported by the
hardware. Use the vector converts instead of extracting the
element, sign extending it, and then converting it to double.
Other floating point types than double first convert to double,
then the double is converted to that type.
(vsx_extract_si_<uns>float_<mode>): Likewise.
[gcc/testsuite]
2016-10-31 Michael Meissner <meissner@linux.vnet.ibm.com>
* gcc.target/powerpc/vsx-extract-4.c: New test.
* gcc.target/powerpc/vsx-extract-5.c: Likewise.
From-SVN: r241731
2016-10-31 Andrew Pinski <apinski@cavium.com>
* config/aarch64/driver-aarch64.c (host_detect_local_cpu):
Rewrite handling of part num to handle the case where
multiple implementers share the same part num.
From-SVN: r241726
