[PATCH] Remove unnecessary invalidation support in threading
2015-10-29  Jeff Law  <law@redhat.com>

* tree-ssa-scopedtables.h (const_and_copies): Remove invalidate
method.
* tree-ssa-scopedtables.h (const_and_copies::invalidate): Remove.
* tree-ssa-threadedge.c
(record_temporary_equivalences_from_stmts_at_dest): Remove
backedge_seen argument and associated code which invalidated
equivalences based on the value of that argument.
(thread_through_normal_block): Corresponding changes.

From-SVN: r229559

rs6000: Save the PIC reg when needed

The PIC reg (r30) needs to be saved whenever the prologue sets it up,
not just if it is used elsewhere in the function.  Without this patch
the prologue for such a function will modify r30 without saving it
first, leading to disaster back in its callers.  This happened in the
testsuite for -m32 libgfortran and libstdc++, bootstrapped with -mlra,
many hundred times.

2015-10-29  Segher Boessenkool  <segher@kernel.crashing.org>

* config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Move this
function earlier in the file.
(first_reg_to_save): Use rs6000_reg_live_or_pic_offset_p instead of
df_regs_ever_live_p.

From-SVN: r229552

lra: Don't remove the scratch in (mem:BLK (scratch))

LRA wants to replace SCRATCH registers with real registers.  It should
not do that with (mem:BLK (scratch)), which is special, not really a
scratch register.

2015-10-29  Segher Boessenkool  <segher@kernel.crashing.org>

* lra-constraints.c (process_address_1): Handle (mem:BLK (scratch))
by ignoring it.

From-SVN: r229551

Fix target/68124

        PR target/68124
        PR rtl-opt/67609
        * config/i386/i386.c (ix86_cannot_change_mode_class): Tighten
        sse check to the exact conditions of PR 67609.

From-SVN: r229550

Require c99_runtime for builtin-convert-1.c

...which was failing on *-elf.  This test is making sure that we narrow
maths calls, e.g. sinh -> sinhf, but we can only do that if we know that
the target has a full c99 libm or if we have other proof that the function
we want to generate is available.

Tested on aarch64-none-elf (where it's skipped) and aarch64-linux-gnu
(where it isn't).

gcc/testsuite/
* gcc.dg/torture/builtin-convert-1.c: Require c99_runtime.

From-SVN: r229548

rs6000.c (rs6000_init_libfuncs): Split libfunc setup into 3 functions...

[gcc]
2015-10-29  Michael Meissner  <meissner@linux.vnet.ibm.com>

* config/rs6000/rs6000.c (rs6000_init_libfuncs): Split libfunc
setup into 3 functions: init_float128_ibm, init_float128_ieee, and
rs6000_init_libfuncs. If -mfloat128, add IFmode functions for all
of the traditional names that TFmode uses for handling IEEE
extended double. If -mfloat128, add KFmode functions for all of
the emulation functions. If -mabi=ieeelongdouble and -mfloat128,
make TFmode use the same emulation functions as KFmode.
(init_float128_ibm): Likewise.
(init_float128_ieee): Likewise.
(rs6000_generate_compare): For IEEE 128-bit floating point
comparisons, call the unordered comparison function instead of the
ordered comparison function.
(rs6000_expand_float128_convert): Deal with operands that are
memory operands. Restructure the code to use a switch statement on
the mode. Add support for TFmode defaulting to either IBM extended
double or IEEE 128-bit floating point. If the underlying types are
the same, use a move instead of a conversion function.
(TARGET_C_MODE_FOR_SUFFIX): Define 'q' and 'Q' as the suffix to
use for IEEE 128-bit floating point constants with -mfloat128.
(rs6000_c_mode_for_suffix): Likewise.
(TARGET_INVALID_BINARY_OP): Do not allow inter-mixing of IEEE
128-bit floating point with IBM extended double floating point.
(rs6000_invalid_binary_op): Likewise.
(rs6000_gen_le_vsx_permute): On little endian systems generate a
ROTATE insn instead of VEC_SELECT for IEEE 128-bit floating point
types that can go in vector registers.
(chain_contains_only_swaps): Properly swap IEEE 128-bit floating
point types that can go in vector registers on little endian
PowerPC systems.
(mark_swaps_for_removal): Likewise.
(rs6000_analyze_swaps): Likewise.
(rs6000_mangle_type): Use U10__float128 for IEEE 128-bit floating point.

* config/rs6000/rs6000.md (FLOAT128_SFDFTF): Delete iterator,
rework IEEE 128-bit floating point insns to deal with TFmode being
either IBM extended double or IEEE 128-bit floating point.
(IFKF): Likewise.
(IBM128): Update iterator to add condition that the mode is IBM
extended double.
(IEEE128): New iterator for IEEE 128-bit floating point.
(TFIFKF): Rename TFIFKF iterator to FLOAT128.
(FLOAT128): Likewise.
(signbit<mode>2): FLOAT128_IBM_P condition test moved into IBM128
iterator.
(neg<mode>2): Replace TFIFKF iterator with FLOAT128. Add support
for TFmode being IEEE 128-bit floating point. Use IEEE128 iterator
instead of hard coding TFmode or KFmode.
(negtf2_internal): Likewise.
(neg<mode>2_internal): Likewise.
(abs<mode>2): Likewise.
(abstf2_internal): Likewise.
(abs<mode>2_internal): Likewise.
(ieee_128bit_neg<mode>2): Likewise.
(ieee_128bit_neg<mode>2_internal): Likewise.
(ieee_128bit_abs<mode>2): Likewise.
(ieee_128bit_abs<mode>2_internal): Likewise.
(ieee_128bit_nabs<mode>2): Likewise.
(ieee_128bit_nabs<mode>2_internal): Likewise.
(extendiftf2): Add explicit conversions between 128-bit floating
point types. Drop the old conversions that had become unwieldy.
(extend<FLOAT128_SFDFTF:mode><IFKF:mode>2): Likewise.
(extendifkf2): Likewise.
(trunc<IFKF:mode><FLOAT128_SFDFTF:mode>2): Likewise.
(extendtfkf2): Likewise.
(fix_trunc<IFKF:mode><SDI:mode>2): Likewise.
(trunciftf2): Likewise.
(fixuns_trunc<IFKF:mode><SDI:mode>2): Likewise.
(truncifkf2): Likewise.
(float<SDI:mode><IFKF:mode>2): Likewise.
(trunckftf2): Likewise.
(floatuns<SDI:mode><IFKF:mode>2): Likewise.
(trunctfif2): Likewise.
(FP iterator): Allow TFmode to be IEEE 128-bit floating point.
(extenddftf2): Rework 128-bit floating point conversions to
properly handle -mabi=ieeelongdouble. Merge IFmode, TFmode, and
KFmode expanders into one function.
(extenddf<mode>2): Likewise.
(extenddftf2_fprs): Likewise.
(extenddf<mode>2_fprs): Likewise.
(extenddftf2_vsx): Likewise.
(extenddf<mode>2_vsx): Likewise.
(extendsftf2): Likewise.
(extendsf<mode>2): Likewise.
(trunctfdf2): Likewise.
(trunc<mode>df2): Likewise.
(trunctfdf2_internal1): Likewise.
(trunc<mode>df2_internal1): Likewise.
(trunctfdf2_internal2): Likewise.
(trunc<mode>df2_internal2): Likewise.
(trunctfsf2): Likewise.
(trunc<mode>sf2): Likewise.
(trunctfsf2_fprs): Likewise.
(trunc<mode>sf2_fprs): Likewise.
(floatsit2f): Likewise.
(floatsi<mode>2): Likewise.
(fix_trunc_helper): Likewise.
(fix_trunc_helper<mode>): Likewise.
(fix_trunctfsi2): Likewise.
(fix_trunc<mode>si2): Likewise.
(fix_trunctfsi2_fprs): Likewise.
(fix_trunc<mode>si2_fprs): Likewise.
(fix_trunctfsi2_internal): Likewise.
(fix_trunc<mode>si2_internal): Likewise.
(fix_trunctfdi2): Likewise.
(fix_trunc<mode>di2): Likewise.
(fixuns_trunctf<mode>2): Likewise.
(fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
(floatditf2): Likewise.
(floatdi<mode>2): Likewise.
(floatuns<mode>tf2): Likewise.
(floatuns<SDI:mode><IEEE128:mode>): Likewise.
(cmptf_internal1): Use a mode iterator to add support for both
types (IFmode, TFmode) that support IBM extended double.
(cmp<mode>_internal1): Likewise.
(cmptf_internal2): Likewise.
(cmp<mode>_internal2): Likewise.

* doc/extend.texi (Floating Types): Document __ibm128 and
__float128 on PowerPC.

* doc/invoke.texi (RS/6000 and PowerPC Options): Document
-mfloat128 and -mno-float128.

[gcc/testsuite]
2015-10-29  Michael Meissner  <meissner@linux.vnet.ibm.com>

* gcc.target/powerpc/float128-call.c: New test for -mfloat128 on
PowerPC.
* gcc.target/powerpc/float128-mix.c: Likewise.

From-SVN: r229547

compiler, reflect, runtime: remove zero field from type descriptor

    Type descriptors picked up a zero field because the gc map
    implementation used it.  However, it's since been dropped by the gc
    library.  It was never used by gccgo.  Drop it now in preparation for
    upgrading to the Go 1.5 library.

    Reviewed-on: https://go-review.googlesource.com/16486

From-SVN: r229546

rs6000.h (ALTIVEC_VECTOR_MODE): Add IEEE 128-bit floating point modes that can go in vector registers.

2015-10-29  Michael Meissner  <meissner@linux.vnet.ibm.com>

* config/rs6000/rs6000.h (ALTIVEC_VECTOR_MODE): Add IEEE 128-bit
floating point modes that can go in vector registers.
(MODES_TIEABLE_P): Move tests for vector modes before tests for
scalar floating point, so that IEEE 128-bit floating point that
can go in vector registers bind with vectors and not FP.
(struct rs6000_args): Add libcall field.

* config/rs6000/rs6000.opt (-mfloat128-*): Delete -mfloat128-none
and -mfloat128-software switches.  Replace them with a binary
-mfloat128 switch.
(-mfloat128): Likewise.

* config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Allow
128-bit floating point types in GPRs, even if the appropriate
option enabling the type was not used.
(rs6000_debug_reg_global): Remove -mfloat128-{software,none}
debugging.
(rs6000_setup_reg_addr_masks): Do not allow pre-increment and
pre-decrement on IEEE 128-bit floating point values.
(rs6000_init_hard_regno_mode_ok): Change test for whether TFmode
is IEEE 128-bit floating point.
(rs6000_init_hard_regno_mode_ok): Add reload handlers for IEEE
128-bit floating point types that can go in vector registers.
(rs6000_option_override_internal): Change -mfloat128-none and
-mfloat128-software to -mfloat128, and move code to be near other
VSX option handling.
(rs6000_option_override_internal): Disable -mfloat128 if we don't
have the Altivec ABI.
(rs6000_init_builtins): Don't make TFmode use either IFmode or
KFmode floating point nodes. Instead, have three separate nodes.
(rs6000_scalar_mode_supported_p): Add support for IFmode to allow
eventually moving the long double default to IEEE 128-bit floating
point.
(rs6000_opt_masks): Add -mfloat128.
(struct rs6000_opt_var): Fix typo in comment.
(init_cumulative_args): Initialize libcall field in
CUMULATIVE_ARGS.
(rs6000_function_arg): Treat library functions as if they had
prototypes to prevent IEEE 128-bit support functions from passing
arguments in both GPRs and vector registers.
(rs6000_arg_partial_bytes): Likewise.

* config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Add -mfloat128 as
an option that can be turned on via -mcpu=<xxx>.

* config/rs6000/rs6000-opts.h (enum float128_type_t): Delete, no
longer used.

* config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
__FLOAT128__ if -mfloat128. Define __LONG_DOUBLE_IEEE128__ if long
double is IEEE 128-bit. Define __LONG_DOUBLE_IBM128__ if long
double is IBM extended double.

* config/rs6000/predicates.md (reg_or_indexed_operand): Allow
SUBREGs.

From-SVN: r229545

re PR fortran/68054 (ICE on using protected attribute in program without program statement)

2015-10-29  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/68054
* decl.c (match_attr_spec): PROTECTED can only be a module.

2015-10-29  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/68054
* gfortran.dg/pr68054.f90: New test.

From-SVN: r229542

compiler, runtime: change type hash/equal to Go funcs

    Change the type descriptor hash and equal functions from C code pointers
    to Go func values.  This permits them to be set to a Go function
    closure.  This is in preparation for the Go 1.5, so that we can use a
    closure for the hash/equal functions returned by the new reflect.ArrayOf
    function.

    Reviewed-on: https://go-review.googlesource.com/16485

From-SVN: r229541

[multiple changes]

2015-10-29  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/67885
* trans-decl.c (generate_local_decl): Mark PARAMETER entities in
BLOCK construct.

2015-10-26  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/67885
* gfortran.dg/pr67885.f90: New test.

From-SVN: r229540

[PATCH 6/9] ENABLE_CHECKING refactoring: generators

        * genautomata.c: Use CHECKING_P instead of ENABLE_CHECKING.
        * genconditions.c: Define CHECKING_P in the generated code.
        * genextract.c: Use flag_checking in insn_extract.
        * gengtype.c (main): Remove conditional compilation.
        * gengtype.h: Likewise.

From-SVN: r229539

[PATCH][PR tree-optimization/67892] Use FSM threader to handle backedges

PR tree-optimization/67892
* tree-ssa-threadedge.c (simplify_controL_stmt_condition): Fix typo
in comment.
(thread_through_normal_block): If we have seen a backedge, then
do nothing.  No longer call find_jump_threads_backwards here.
(thread_across_edge): Use find_jump_threads_backwards to find
jump threads if the old style threader was not successful.
* tree-ssa-threadbackward.c (get_gimple_control_stmt): Use
gsi_last_nondebug_bb.  Return NULL if the block does not end
with a control statement.
(find_jump_threads_backwards): Setup code moved here from
tree-ssa-threadedge.c::thread_through_normal_block.  Accept
single edge argument instead of name & block.
* tree-ssa-threadbackward.h (find_jump_threads_backwards): Update
prototype.

        PR tree-optimization/67892
* gcc.dg/tree-ssa/pr21417: Update expected output.
* gcc.dg/tree-ssa/ssa-dom-thread-2b.c: Likewise.

From-SVN: r229538

Allow CAST_RESTRICT folding

2015-10-29  Tom de Vries  <tom@codesourcery.com>

* fold-const.c (fold_unary_loc): Remove folding inhibition for restrict
types.

From-SVN: r229537

omp-low.c (lower_omp_target): Remove unreachable code & merge ifs.

* omp-low.c (lower_omp_target): Remove unreachable code & merge
ifs.

From-SVN: r229535

Fix X - (X / Y) * Y in match.pd.

2015-10-29  Marc Glisse  <marc.glisse@inria.fr>

* match.pd (X-(X/Y)*Y): Properly handle conversions and commutativity.

From-SVN: r229534

decl.c: Reorder #include's and remove duplicates.

ada
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* gcc-interface/decl.c: Reorder #include's and remove duplicates.
* gcc-interface/misc.c: Likewise.
* gcc-interface/targtyps.c: Likewise.
* gcc-interface/trans.c: Likewise.
* gcc-interface/utils.c: Likewise.

c
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* c-array-notation.c: Reorder #include's and remove duplicates.
* c-aux-info.c: Likewise.
* c-convert.c: Likewise.
* c-decl.c: Likewise.
* c-errors.c: Likewise.
* c-lang.c: Likewise.
* c-objc-common.c: Likewise.
* c-parser.c: Likewise.
* c-typeck.c: Likewise.

c-family
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* array-notation-common.c: Reorder #include's and remove duplicates.
* c-ada-spec.c: Likewise.
* c-cilkplus.c: Likewise.
* c-common.c: Likewise.
* c-cppbuiltin.c: Likewise.
* c-dump.c: Likewise.
* c-format.c: Likewise.
* c-gimplify.c: Likewise.
* c-indentation.c: Likewise.
* c-lex.c: Likewise.
* c-omp.c: Likewise.
* c-opts.c: Likewise.
* c-pch.c: Likewise.
* c-ppoutput.c: Likewise.
* c-pragma.c: Likewise.
* c-pretty-print.c: Likewise.
* c-semantics.c: Likewise.
* c-ubsan.c: Likewise.
* cilk.c: Likewise.
* stub-objc.c: Likewise.

cp
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* call.c: Reorder #include's and remove duplicates.
* class.c: Likewise.
* constexpr.c: Likewise.
* cp-array-notation.c: Likewise.
* cp-cilkplus.c: Likewise.
* cp-gimplify.c: Likewise.
* cp-lang.c: Likewise.
* cp-objcp-common.c: Likewise.
* cp-ubsan.c: Likewise.
* cvt.c: Likewise.
* cxx-pretty-print.c: Likewise.
* decl.c: Likewise.
* decl2.c: Likewise.
* dump.c: Likewise.
* error.c: Likewise.
* except.c: Likewise.
* expr.c: Likewise.
* friend.c: Likewise.
* init.c: Likewise.
* lambda.c: Likewise.
* lex.c: Likewise.
* mangle.c: Likewise.
* method.c: Likewise.
* name-lookup.c: Likewise.
* optimize.c: Likewise.
* parser.c: Likewise.
* pt.c: Likewise.
* ptree.c: Likewise.
* repo.c: Likewise.
* rtti.c: Likewise.
* search.c: Likewise.
* semantics.c: Likewise.
* tree.c: Likewise.
* typeck.c: Likewise.
* typeck2.c: Likewise.
* vtable-class-hierarchy.c: Likewise.

fortran

2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* array.c: Reorder #include's and remove duplicates.
* convert.c: Likewise.
* cpp.c: Likewise.
* decl.c: Likewise.
* f95-lang.c: Likewise.
* frontend-passes.c: Likewise.
* iresolve.c: Likewise.
* match.c: Likewise.
* module.c: Likewise.
* options.c: Likewise.
* parse.c: Likewise.
* resolve.c: Likewise.
* simplify.c: Likewise.
* target-memory.c: Likewise.
* trans-array.c: Likewise.
* trans-common.c: Likewise.
* trans-const.c: Likewise.
* trans-decl.c: Likewise.
* trans-expr.c: Likewise.
* trans-intrinsic.c: Likewise.
* trans-io.c: Likewise.
* trans-openmp.c: Likewise.
* trans-stmt.c: Likewise.
* trans-types.c: Likewise.
* trans.c: Likewise.

go
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* go-backend.c: Reorder #include's and remove duplicates.
* go-lang.c: Likewise.

java
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* boehm.c: Reorder #include's and remove duplicates.
* builtins.c: Likewise.
* class.c: Likewise.
* constants.c: Likewise.
* decl.c: Likewise.
* except.c: Likewise.
* expr.c: Likewise.
* java-gimplify.c: Likewise.
* jcf-dump.c: Likewise.
* jcf-io.c: Likewise.
* jcf-parse.c: Likewise.
* jvgenmain.c: Likewise.
* lang.c: Likewise.
* mangle.c: Likewise.
* mangle_name.c: Likewise.
* resource.c: Likewise.
* typeck.c: Likewise.
* verify-glue.c: Likewise.
* verify-impl.c: Likewise.

jit
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* dummy-frontend.c: Reorder #include's and remove duplicates.
* jit-builtins.c: Likewise.
* jit-playback.c: Likewise.
* jit-recording.c: Likewise.
* libgccjit.c: Likewise.

lto
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* lto-lang.c: Reorder #include's and remove duplicates.
* lto-object.c: Likewise.
* lto-partition.c: Likewise.
* lto-symtab.c: Likewise.
* lto.c: Likewise.

objc
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* objc-lang.c: Reorder #include's and remove duplicates.
* objc-map.c: Likewise.

objcp
2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* objcp-decl.c: Reorder #include's and remove duplicates.
* objcp-lang.c: Likewise.

From-SVN: r229533

Remove redundant variable from tree-call-cdce.c

shrink_wrap_one_built_in_call had both guard_bb and guard_bb0.
It looks like an earlier version of the pass may have updated
one of the variables in the while loop, but now they're just
two names for the same thing.

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* tree-call-cdce.c (shrink_wrap_one_built_in_call): Remove
guard_bb0 and use guard_bb throughout.

From-SVN: r229532

Remove unnecessary label from tree-call-cdce.c

Tested on x86_64-linux-gnu, arm-linux-gnueabi and aarch64-linux-gnu.

gcc/
* tree-call-cdce.c (shrink_wrap_one_built_in_call): Remove
unnecessary label.

From-SVN: r229531

config.host (arm*-*-eabi*, [...]): Include crtfastmath.o.

2015-10-29  Christophe Lyon  <christophe.lyon@linaro.org>

* config.host (arm*-*-eabi*, arm*-*-symbianelf*, arm*-*-rtems*):
Include crtfastmath.o.

From-SVN: r229529

re PR middle-end/68142 (unsafe association of multiplication)

2015-10-29  Richard Biener  <rguenther@suse.de>

PR middle-end/68142
* fold-const.c (extract_muldiv_1): Avoid introducing undefined
overflow.

* c-c++-common/ubsan/pr68142.c: New testcase.

From-SVN: r229528

alias.c: Reorder #include statements and remove duplicates.

2015-10-29  Andrew MacLeod  <amacleod@redhat.com>

* alias.c: Reorder #include statements and remove duplicates.
* asan.c: Likewise.
* attribs.c: Likewise.
* auto-inc-dec.c: Likewise.
* auto-profile.c: Likewise.
* bb-reorder.c: Likewise.
* bt-load.c: Likewise.
* builtins.c: Likewise.
* caller-save.c: Likewise.
* calls.c: Likewise.
* ccmp.c: Likewise.
* cfg.c: Likewise.
* cfganal.c: Likewise.
* cfgbuild.c: Likewise.
* cfgcleanup.c: Likewise.
* cfgexpand.c: Likewise.
* cfghooks.c: Likewise.
* cfgloop.c: Likewise.
* cfgloopanal.c: Likewise.
* cfgloopmanip.c: Likewise.
* cfgrtl.c: Likewise.
* cgraph.c: Likewise.
* cgraphbuild.c: Likewise.
* cgraphclones.c: Likewise.
* cgraphunit.c: Likewise.
* cilk-common.c: Likewise.
* combine-stack-adj.c: Likewise.
* combine.c: Likewise.
* compare-elim.c: Likewise.
* convert.c: Likewise.
* coverage.c: Likewise.
* cppbuiltin.c: Likewise.
* cprop.c: Likewise.
* cse.c: Likewise.
* cselib.c: Likewise.
* data-streamer-in.c: Likewise.
* data-streamer-out.c: Likewise.
* data-streamer.c: Likewise.
* dbxout.c: Likewise.
* dce.c: Likewise.
* ddg.c: Likewise.
* debug.c: Likewise.
* df-core.c: Likewise.
* df-problems.c: Likewise.
* df-scan.c: Likewise.
* dfp.c: Likewise.
* dojump.c: Likewise.
* dominance.c: Likewise.
* double-int.c: Likewise.
* dse.c: Likewise.
* dumpfile.c: Likewise.
* dwarf2asm.c: Likewise.
* dwarf2cfi.c: Likewise.
* dwarf2out.c: Likewise.
* emit-rtl.c: Likewise.
* except.c: Likewise.
* explow.c: Likewise.
* expmed.c: Likewise.
* expr.c: Likewise.
* final.c: Likewise.
* fixed-value.c: Likewise.
* fold-const.c: Likewise.
* function.c: Likewise.
* fwprop.c: Likewise.
* gcse.c: Likewise.
* generic-match-head.c: Likewise.
* ggc-common.c: Likewise.
* gimple-builder.c: Likewise.
* gimple-expr.c: Likewise.
* gimple-fold.c: Likewise.
* gimple-iterator.c: Likewise.
* gimple-low.c: Likewise.
* gimple-match-head.c: Likewise.
* gimple-pretty-print.c: Likewise.
* gimple-ssa-isolate-paths.c: Likewise.
* gimple-ssa-strength-reduction.c: Likewise.
* gimple-streamer-in.c: Likewise.
* gimple-streamer-out.c: Likewise.
* gimple-walk.c: Likewise.
* gimple.c: Likewise.
* gimplify-me.c: Likewise.
* gimplify.c: Likewise.
* godump.c: Likewise.
* graph.c: Likewise.
* graphite-poly.c: Likewise.
* haifa-sched.c: Likewise.
* hw-doloop.c: Likewise.
* ifcvt.c: Likewise.
* incpath.c: Likewise.
* init-regs.c: Likewise.
* internal-fn.c: Likewise.
* ipa-chkp.c: Likewise.
* ipa-comdats.c: Likewise.
* ipa-cp.c: Likewise.
* ipa-devirt.c: Likewise.
* ipa-icf-gimple.c: Likewise.
* ipa-icf.c: Likewise.
* ipa-inline-analysis.c: Likewise.
* ipa-inline-transform.c: Likewise.
* ipa-inline.c: Likewise.
* ipa-polymorphic-call.c: Likewise.
* ipa-profile.c: Likewise.
* ipa-prop.c: Likewise.
* ipa-pure-const.c: Likewise.
* ipa-ref.c: Likewise.
* ipa-reference.c: Likewise.
* ipa-split.c: Likewise.
* ipa-utils.c: Likewise.
* ipa-visibility.c: Likewise.
* ipa.c: Likewise.
* ira-build.c: Likewise.
* ira-color.c: Likewise.
* ira-conflicts.c: Likewise.
* ira-costs.c: Likewise.
* ira-emit.c: Likewise.
* ira-lives.c: Likewise.
* ira.c: Likewise.
* jump.c: Likewise.
* langhooks.c: Likewise.
* lcm.c: Likewise.
* lists.c: Likewise.
* loop-doloop.c: Likewise.
* loop-init.c: Likewise.
* loop-invariant.c: Likewise.
* loop-iv.c: Likewise.
* loop-unroll.c: Likewise.
* lower-subreg.c: Likewise.
* lra-assigns.c: Likewise.
* lra-coalesce.c: Likewise.
* lra-constraints.c: Likewise.
* lra-eliminations.c: Likewise.
* lra-lives.c: Likewise.
* lra-remat.c: Likewise.
* lra-spills.c: Likewise.
* lra.c: Likewise.
* lto-cgraph.c: Likewise.
* lto-compress.c: Likewise.
* lto-opts.c: Likewise.
* lto-section-in.c: Likewise.
* lto-section-out.c: Likewise.
* lto-streamer-in.c: Likewise.
* lto-streamer-out.c: Likewise.
* lto-streamer.c: Likewise.
* mode-switching.c: Likewise.
* modulo-sched.c: Likewise.
* omp-low.c: Likewise.
* optabs.c: Likewise.
* opts-global.c: Likewise.
* passes.c: Likewise.
* plugin.c: Likewise.
* postreload-gcse.c: Likewise.
* postreload.c: Likewise.
* predict.c: Likewise.
* print-tree.c: Likewise.
* profile.c: Likewise.
* real.c: Likewise.
* realmpfr.c: Likewise.
* recog.c: Likewise.
* ree.c: Likewise.
* reg-stack.c: Likewise.
* regcprop.c: Likewise.
* reginfo.c: Likewise.
* regrename.c: Likewise.
* regstat.c: Likewise.
* reload.c: Likewise.
* reload1.c: Likewise.
* reorg.c: Likewise.
* resource.c: Likewise.
* rtl-chkp.c: Likewise.
* rtl-error.c: Likewise.
* rtlanal.c: Likewise.
* rtlhooks.c: Likewise.
* sanopt.c: Likewise.
* sched-deps.c: Likewise.
* sched-ebb.c: Likewise.
* sched-rgn.c: Likewise.
* sese.c: Likewise.
* shrink-wrap.c: Likewise.
* simplify-rtx.c: Likewise.
* stack-ptr-mod.c: Likewise.
* statistics.c: Likewise.
* stmt.c: Likewise.
* stor-layout.c: Likewise.
* store-motion.c: Likewise.
* stringpool.c: Likewise.
* symtab.c: Likewise.
* target-globals.c: Likewise.
* targhooks.c: Likewise.
* toplev.c: Likewise.
* tracer.c: Likewise.
* trans-mem.c: Likewise.
* tree-affine.c: Likewise.
* tree-call-cdce.c: Likewise.
* tree-cfg.c: Likewise.
* tree-cfgcleanup.c: Likewise.
* tree-chkp-opt.c: Likewise.
* tree-chkp.c: Likewise.
* tree-chrec.c: Likewise.
* tree-complex.c: Likewise.
* tree-data-ref.c: Likewise.
* tree-dfa.c: Likewise.
* tree-diagnostic.c: Likewise.
* tree-dump.c: Likewise.
* tree-eh.c: Likewise.
* tree-emutls.c: Likewise.
* tree-if-conv.c: Likewise.
* tree-inline.c: Likewise.
* tree-into-ssa.c: Likewise.
* tree-iterator.c: Likewise.
* tree-loop-distribution.c: Likewise.
* tree-nested.c: Likewise.
* tree-nrv.c: Likewise.
* tree-object-size.c: Likewise.
* tree-outof-ssa.c: Likewise.
* tree-parloops.c: Likewise.
* tree-phinodes.c: Likewise.
* tree-predcom.c: Likewise.
* tree-pretty-print.c: Likewise.
* tree-profile.c: Likewise.
* tree-scalar-evolution.c: Likewise.
* tree-sra.c: Likewise.
* tree-ssa-address.c: Likewise.
* tree-ssa-alias.c: Likewise.
* tree-ssa-ccp.c: Likewise.
* tree-ssa-coalesce.c: Likewise.
* tree-ssa-copy.c: Likewise.
* tree-ssa-dce.c: Likewise.
* tree-ssa-dom.c: Likewise.
* tree-ssa-dse.c: Likewise.
* tree-ssa-forwprop.c: Likewise.
* tree-ssa-ifcombine.c: Likewise.
* tree-ssa-live.c: Likewise.
* tree-ssa-loop-ch.c: Likewise.
* tree-ssa-loop-im.c: Likewise.
* tree-ssa-loop-ivcanon.c: Likewise.
* tree-ssa-loop-ivopts.c: Likewise.
* tree-ssa-loop-manip.c: Likewise.
* tree-ssa-loop-niter.c: Likewise.
* tree-ssa-loop-prefetch.c: Likewise.
* tree-ssa-loop-unswitch.c: Likewise.
* tree-ssa-loop.c: Likewise.
* tree-ssa-math-opts.c: Likewise.
* tree-ssa-operands.c: Likewise.
* tree-ssa-phiopt.c: Likewise.
* tree-ssa-phiprop.c: Likewise.
* tree-ssa-pre.c: Likewise.
* tree-ssa-propagate.c: Likewise.
* tree-ssa-reassoc.c: Likewise.
* tree-ssa-sccvn.c: Likewise.
* tree-ssa-scopedtables.c: Likewise.
* tree-ssa-sink.c: Likewise.
* tree-ssa-strlen.c: Likewise.
* tree-ssa-structalias.c: Likewise.
* tree-ssa-tail-merge.c: Likewise.
* tree-ssa-ter.c: Likewise.
* tree-ssa-threadedge.c: Likewise.
* tree-ssa-threadupdate.c: Likewise.
* tree-ssa-uncprop.c: Likewise.
* tree-ssa-uninit.c: Likewise.
* tree-ssa.c: Likewise.
* tree-ssanames.c: Likewise.
* tree-stdarg.c: Likewise.
* tree-streamer-in.c: Likewise.
* tree-streamer-out.c: Likewise.
* tree-streamer.c: Likewise.
* tree-switch-conversion.c: Likewise.
* tree-tailcall.c: Likewise.
* tree-vect-data-refs.c: Likewise.
* tree-vect-generic.c: Likewise.
* tree-vect-loop-manip.c: Likewise.
* tree-vect-loop.c: Likewise.
* tree-vect-patterns.c: Likewise.
* tree-vect-slp.c: Likewise.
* tree-vect-stmts.c: Likewise.
* tree-vectorizer.c: Likewise.
* tree-vrp.c: Likewise.
* tree.c: Likewise.
* tsan.c: Likewise.
* ubsan.c: Likewise.
* value-prof.c: Likewise.
* var-tracking.c: Likewise.
* varasm.c: Likewise.
* varpool.c: Likewise.
* vtable-verify.c: Likewise.
* web.c: Likewise.
* wide-int-print.cc: Likewise.
* wide-int.cc: Likewise.
* xcoffout.c: Likewise.

From-SVN: r229526

re PR c++/67845 (ICE on invalid use of const qualifier on x86_64-linux-gnu in merge_types, at cp/typeck.c:854)

/cp
2015-10-29  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/67845
* decl.c (grokfndecl): In case of erroneous cv-qualified non-member
functions consistently reset TREE_TYPE (decl) too.

/testsuite
2015-10-29  Paolo Carlini  <paolo.carlini@oracle.com>

PR c++/67845
* g++.dg/other/cv_func4.C: New.

From-SVN: r229523

Add BUILD_NO_PIE_CFLAGS and BUILD_NO_PIE_FLAG

We shouldn't use NO_PIE_CFLAGS and NO_PIE_FLAG with CXX_FOR_BUILD
when CXX_FOR_BUILD != CXX.  This patch adds BUILD_NO_PIE_CFLAGS
and BUILD_NO_PIE_FLAG to use with CXX_FOR_BUILD.  They are set to
NO_PIE_CFLAGS and NO_PIE_FLAG when build machine == host machine.
Otherwise, they are set to NO_PIE_CFLAGS_FOR_BUILD and
NO_PIE_FLAG_FOR_BUILD.

* Makefile.in (NO_PIE_CFLAGS): New.
(NO_PIE_FLAG): Likewise.
(NO_PIE_CFLAGS_FOR_BUILD): Likewise.
(NO_PIE_FLAG_FOR_BUILD): Likewise.
(BUILD_NO_PIE_CFLAGS): Likewise.
(BUILD_NO_PIE_FLAG): Likewise.
(COMPILER): Replace @NO_PIE_CFLAGS@ with $(NO_PIE_CFLAGS).
(LINKER): Replace @NO_PIE_FLAG@ with $(NO_PIE_FLAG).
(BUILD_CFLAGS): Replace @NO_PIE_CFLAGS@ with
$(BUILD_NO_PIE_CFLAGS).
(BUILD_CXXFLAGS): Likewise.
(BUILD_LDFLAGS ): Replace @NO_PIE_FLAG@ with
$(BUILD_NO_PIE_FLAG).
* configure.ac (BUILD_NO_PIE_CFLAGS): New.  AC_SUBST.
(BUILD_NO_PIE_FLAG): Likewise.
(NO_PIE_CFLAGS_FOR_BUILD): Likewise.
(NO_PIE_FLAG_FOR_BUILD): Likewise.
* configure: Regenerated.

From-SVN: r229522

Improve filenames for test cases of OpenACC combined directives

libgomp/
* testsuite/libgomp.oacc-c-c++-common/combdir-1.c: Rename to...
* testsuite/libgomp.oacc-c-c++-common/combined-directives-1.c:
... this.  Add a description of the test at the top of the file.
* testsuite/libgomp.oacc-fortran/combdir-1.f90: Rename file to...
* testsuite/libgomp.oacc-fortran/combined-directives-1.f90:
... this.  Add a description of the test at the top of the file.

Co-Authored-By: Cesar Philippidis <cesar@codesourcery.com>
From-SVN: r229518

re PR middle-end/56956 (ftrapv traps on valid abs-like code)

2015-10-29  Richard Biener  <rguenther@suse.de>

PR middle-end/56956
* fold-const.c (fold_cond_expr_with_comparison): Do not fold
unsigned conditonal negation to ABS_EXPR.

* c-c++-common/ubsan/pr56956.c: New testcase.

From-SVN: r229517

gimple-match-head.c (gimple_simplify): Remove premature checking of builtin_decl_implicit of function calls we simplify.

2015-10-29  Richard Biener  <rguenther@suse.de>

* gimple-match-head.c (gimple_simplify): Remove premature checking
of builtin_decl_implicit of function calls we simplify.

fortran/
* f95-lang.c: Properly build variadic types for classification
and comparison builtins.

From-SVN: r229515

tree-ssa-loop-ivopts.c (split_address_cost): Check depends_on.

* tree-ssa-loop-ivopts.c (split_address_cost): Check depends_on.
(get_computation_cost_at): Ditto.
(determine_use_iv_cost_address): Pass NULL for arguments depends_on
and inv_expr_id.

From-SVN: r229514

Daily bump.

From-SVN: r229511

* gnat.dg/discr44.adb: New test.

From-SVN: r229508

Remove superfluous code in intra_create_variable_infos

2015-10-28  Tom de Vries  <tom@codesourcery.com>

* tree-ssa-structalias.c (intra_create_variable_infos): Remove
superfluous code.

From-SVN: r229506

Fix for ICE with -g on testcase with incomplete types.

gcc/c/
PR debug/66068
* c-typeck.c (c_build_qualified_type): Clear C_TYPE_INCOMPLETE_VARS
after calling build_qualified_type.
gcc/testsuite/
PR debug/66068
* gcc.dg/debug/pr66068.c: New test.

From-SVN: r229505

Make-lang.in (c.tags): Also include libcpp TAGS.

gcc/c/
* Make-lang.in (c.tags): Also include libcpp TAGS.
gcc/cp/
* Make-lang.in (c++.tags): Also include libcpp TAGS.

From-SVN: r229504

re PR fortran/67933 (ICE for array of a derived type with allocatable class in derived type object)

2015-01-28  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/67933
* gfortran.dg/allocate_with_source_15.f03: Correct trivial
array bounds error at line 25.

From-SVN: r229503

DR 1518

DR 1518
* class.c (type_has_user_provided_or_explicit_constructor): New.
(check_bases_and_members): Use it.
* cp-tree.h: Declare it.

From-SVN: r229501

* c-common.c (pointer_int_sum): Fold the MULT_EXPR.

From-SVN: r229500

omp-low.c (MASK_GANG, [...]): Delete.

* omp-low.c (MASK_GANG, MASK_WORKER, MASK_VECTOR): Delete.
(extract_omp_for_data): Remove OpenACC special handling of
chunking.

From-SVN: r229499

nvptx.c (nvptx_print_operand): Remove 'd' case.

* config/nvptx/nvptx.c (nvptx_print_operand): Remove 'd' case.
(struct parallel): Update comment.
(nvptx_reorg): Likewise.
(nvptx_neuter): Cleanup whitespace.

From-SVN: r229498

* tree-eh.c (mark_reachable_handlers): Fix typo in assert.

From-SVN: r229497

oddspreg-3.c: Disable for MIPS16.

2015-10-28  Catherine Moore  <clm@codesourcery.com>

        * gcc.target/mips/oddspreg-3.c: Disable for MIPS16.
        * gcc.target/mips/oddspreg-6.c: Likewise.
        * gcc.target/mips/oddspreg-1.c: Likewise.
        * gcc.target/mips/oddspreg-2.c: Likewise.

From-SVN: r229496

re PR target/67839 (Bit addressable instructions generated for invalid memory address)

gcc/ChangeLog

PR target/67839
* config/avr/predicates.md (low_io_address_operand): Don't
consider MODE when computing upper bound.
(io_address_operand): Likewise.

gcc/testsuite/ChangeLog

PR target/67839
* gcc.target/avr/pr67839.c: New test.

From-SVN: r229495

fold-const.c (operand_equal_p): Do not verify that types are compatible for MEM_REFs.

* fold-const.c (operand_equal_p): Do not verify that types are
compatible for MEM_REFs.

From-SVN: r229494

gimple-match-head.c (gimple_simplify): Allow VECTOR_CSTs in simplifying VEC_COND_EXPR conditions.

2015-10-28  Richard Biener  <rguenther@suse.de>

* gimple-match-head.c (gimple_simplify): Allow VECTOR_CSTs
in simplifying VEC_COND_EXPR conditions.

* gcc.dg/tree-ssa/operand-equal-2.c: Adjust.

From-SVN: r229492

Generate constraints only once in intra_create_variable_infos

2015-10-28  Tom de Vries  <tom@codesourcery.com>

* tree-ssa-structalias.c (intra_create_variable_infos): Don't iterate
into vi_next of a full_var.

From-SVN: r229490

Improve alias dump info

2015-10-28  Tom de Vries  <tom@codesourcery.com>

* tree-ssa-structalias.c (new_var_info, make_heapvar)
(make_constraint_from_restrict, make_constraint_from_global_restrict)
(create_function_info_for, create_variable_info_for_1)
(create_variable_info_for): Add and handle add_id parameter.
(get_call_vi, new_scalar_tmp_constraint_exp, handle_rhs_call)
(init_base_vars): Add extra argument to calls to new_var_info.
(get_vi_for_tree): Add extra argument to call to
create_variable_info_for.
(process_constraint, do_deref, process_all_all_constraints): Add extra
argument to calls to new_scalar_tmp_constraint_exp.
(handle_lhs_call, find_func_aliases_for_builtin_call): Add extra
argument to calls to make_heapvar.
(make_restrict_var_constraints): Add extra argument to call to
make_constraint_from_global_restrict.
(intra_create_variable_infos): Add extra argument to call to
create_variable_info_for_1.
(ipa_pta_execute): Add extra argument to call to
create_function_info_for.

* gcc.dg/tree-ssa/pta-callused.c: Update to scan for CALLUSED(id).

From-SVN: r229489

Handle noplt in call expanders for AArch64.

2015-10-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>

* config/aarch64/aarch64.md (call, call_value, sibcall): Handle noplt.
(sibcall_value): Likewise.

From-SVN: r229488

loop-g-1.c: New.

* testsuite/libgomp.oacc-c-c++-common/loop-g-1.c: New.
* testsuite/libgomp.oacc-c-c++-common/loop-g-2.c: New.
* testsuite/libgomp.oacc-c-c++-common/loop-gwv-1.c: New.
* testsuite/libgomp.oacc-c-c++-common/loop-v-1.c: New.
* testsuite/libgomp.oacc-c-c++-common/loop-w-1.c: New.
* testsuite/libgomp.oacc-c-c++-common/loop-wv-1.c: New.

From-SVN: r229487

nvptx.h (struct machine_function): Add axis_predicate.

* config/nvptx/nvptx.h (struct machine_function): Add
axis_predicate.
* config/nvptx/nvptx-protos.h (nvptx_expand_oacc_fork,
nvptx_expand_oacc_join): Declare.
* config/nvptx/nvptx.md (UNSPEC_NTID, UNSPEC_TID): Delete.
(UNSPEC_DIM_SIZE, UNSPEC_SHARED_DATA, UNSPEC_BIT_CONV,
UNSPEC_SHUFFLE, UNSPEC_BR_UNIFIED): New.
(UNSPECV_BARSYNC, UNSPECV_DIM_POS, UNSPECV_FORK, UNSPECV_FORKED,
UNSPECV_JOINING, UNSPECV_JOIN): New.
(BITS, BITD): New mode iterators.
(br_true_uni, br_false_uni): New.
(*oacc_ntid_insn, oacc_ntid, *oacc_tid_insn, oacc_tid): Delete.
(oacc_dim_size, oacc_dim_pos): New.
(nvptx_fork, nvptx_forked, nvptx_joining, nvptx_join): New.
(oacc_fork, oacc_join): New.
(nvptx_shuffle<mode>, unpack<mode>si2, packsi<mode>2): New.
(worker_load<mode>, worker_store<mode>): New.
(nvptx_barsync): New.
* config/nvptx/nvptx.c: Include gimple.h & dumpfile.h.
(SHUFFLE_UP, SHUFFLE_DOWN, SHUFFLE_BFLY, SHUFFLE_IDX): Define.
(worker_bcast_hwm, worker_bcast_align, worker_bcast_name,
worker_bcast_sym): New.
(nvptx_option_override): Initialize worker broadcast buffer.
(nvptx_emit_forking, nvptx_emit_joining): New.
(nvptx_init_axis_predicate): New.
(nvptx_declare_function_name): Init axis predicates.
(nvptx_expand_call): Add fork/join markers around routine call.
(nvptx_expand_oacc_fork, nvptx_expand_oacc_join): New.
(nvptx_gen_unpack, nvptx_gen_pack, nvptx_gen_shuffle): New.
(nvptx_gen_vcast): New.
(struct wcast_data_t): New.
(enum propagate_mask): New.
(nvptx_gen_wcast): New.
(nvptx_print_operand): Add 'S' case.
(struct parallel): New.
(parallel::parallel, parallel::~parallel): New.
(bb_insn_map_t, insn_bb_t, insn_bb_vec_t): New typedefs.
(nvptx_split_blocks, nvptx_discover_pre, nvptx_dump_pars,
nvptx_find_par, nvptx_discover_pars): New.
(nvptx_propagate): New.
(vprop_gen, nvptx_vpropagate): New.
(wprop_gen, nvptx_wpropagate): New.
(nvptx_wsync): New.
(nvptx_single, nvptx_skip_par): New.
(nvptx_process_pars, nvptx_neuter_pars): New.
(ntptx_reorg): Split blocks, generate parallel structure, apply
neutering.
(nvptx_cannot_copy_insn_p): New.
(nvptx_file_end): Emit worker broadcast decl.
(nvptx_goacc_fork_join): New.
(TARGET_CANNOT_COPY_INSN_P): Override.
(TARGET_GOACC_FORK_JOIN): Override.

From-SVN: r229486

re PR ipa/67600 (Segfault when assigning only one char to ostreambuf_iterator<char> compiled with -O2 or -O3)

2015-10-28  Richard Biener  <rguenther@suse.de>

PR ipa/67600
* g++.dg/torture/pr67600.C: Remove preprocessed headers in favor
of #includes.

From-SVN: r229485

fold-const.c (negate_expr_p): Adjust the division case to properly avoid introducing undefined overflow.

2015-10-28  Richard Biener  <rguenther@suse.de>

* fold-const.c (negate_expr_p): Adjust the division case to
properly avoid introducing undefined overflow.
(fold_negate_expr): Likewise.

From-SVN: r229484

re PR middle-end/65962 (Missed vectorization of strided stores)

2015-10-28  Richard Biener  <rguenther@suse.de>

PR tree-optimization/65962
* tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
Avoid creating loop carried dependences also for outer loops
of the loop a use to replace is in.

* gcc.dg/vect/vect-62.c: Adjust.

From-SVN: r229481

common.opt (fchecking): New flag controlling flag_checking.

2015-10-28  Richard Biener  <rguenther@suse.de>

* common.opt (fchecking): New flag controlling flag_checking.
* passes.c (verify_curr_properties): Drop DEBUG_FUNCTION.
* timevar.c (timer::print): Adjust output.
* doc/invoke.texi (fchecking): Document.

From-SVN: r229480

re PR middle-end/68067 (Wrong constant folding)

2015-10-28  Richard Biener  <rguenther@suse.de>

PR middle-end/68067
* fold-const.c (negate_expr_p): We cannot negate plus or minus
if overflow is not wrapping.  Likewise multiplication unless
one operand is constant and not power of two.
(fold_negate_expr): Adjust accordingly.

* gcc.dg/torture/pr68067-1.c: New testcase.
* gcc.dg/torture/pr68067-2.c: Likewise.

From-SVN: r229479

rs6000: p8vector-builtin-8.c test requires int128

For 32-bit targets p8vector_ok does not imply we have int128.

2015-10-28  Segher Boessenkool  <segher@kernel.crashing.org>

gcc/testsuite/
* gcc.target/powerpc/p8vector-builtin-8.c: Add "target int128".

From-SVN: r229474

rs6000: Fix tests for xvmadd and xvnmsub

The patterns involved can create vmadd resp. vnmsub instructions instead.
This patch changes the testcases to allow those.

2015-10-26  Segher Boessenkool  <segher@kernel.crashing.org>

gcc/testsuite/
* gcc.target/powerpc/vsx-builtin-2.c: Allow vmadd and vnmsub as well
as xvmadd and xvnmsub.
* gcc.target/powerpc/vsx-vector-2.c: Allow vmadd as well as xvmadd.

From-SVN: r229473

omp-low.c (struct omp_context): Remove gwv_below, gwv_this fields.

* omp-low.c (struct omp_context): Remove gwv_below, gwv_this
fields.
(is_oacc_parallel, is_oacc_kernels): New.
(enclosing_target_ctx): May return NULL.
(ctx_in_oacc_kernels_region): New.
(check_oacc_kernel_gwv): New.
(oacc_loop_or_target_p): Delete.
(scan_omp_for): Don't calculate gwv mask.  Check parallel clause
operands.  Strip reductions fro kernels.
(scan_omp_target): Don't calculate gwv mask.
(lower_oacc_head_mark, lower_oacc_loop_marker,
lower_oacc_head_tail): New.
(struct oacc_collapse): New.
(expand_oacc_collapse_init, expand_oacc_collapse_vars): New.
(expand_omp_for_static_nochunk, expand_omp_for_static_chunk):
Remove OpenACC handling.
(expand_oacc_for): New.
(expand_omp_for): Call expand_oacc_for.
(lower_omp_for): Call lower_oacc_head_tail.

From-SVN: r229472

plugin-nvptx.c (nvptx_exec): Remove check on compute dimensions.

* plugin/plugin-nvptx.c (nvptx_exec): Remove check on compute
dimensions.

From-SVN: r229471

[PATCH 7/9] ENABLE_CHECKING refactoring: middle-end, LTO FE

[PATCH 7/9] ENABLE_CHECKING refactoring: middle-end, LTO FE
gcc/lto/ChangeLog:
2015-10-27  Mikhail Maltsev  <maltsevm@gmail.com>

* lto.c (unify_scc): Use flag_checking and remove ENABLE_CHECKING
conditionals.
(lto_fixup_state): Likewise.
(do_whole_program_analysis): Use
symtab_node::checking_verify_symtab_nodes and remove ENABLE_CHECKING
conditionals.

gcc/ChangeLog:

2015-10-27  Mikhail Maltsev  <maltsevm@gmail.com>

* attribs.c (check_attribute_tables): New function, broken out from...
(init_attributes): Use it.
* cfgcleanup.c (try_optimize_cfg): Use flag_checking, CHECKING_P
gcc_checking_assert and checking_* functions to eliminate
ENABLE_CHECKING conditionals.
* cfgexpand.c (expand_goto, expand_debug_expr): Likewise.
(pass_expand::execute): Likewise.
* cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
* cgraphunit.c (mark_functions_to_output): Likewise.
(cgraph_node::expand_thunk): Likewise.
(symbol_table::compile): Likewise.
* ddg.c (add_cross_iteration_register_deps): Likewise.
(create_ddg_all_sccs): Likewise.
* df-core.c (df_finish_pass, df_analyze): Likewise.
* diagnostic-core.h: Likewise.
* diagnostic.c (diagnostic_report_diagnostic): Likewise.
* dominance.c (calculate_dominance_info): Likewise.
* dwarf2out.c (add_AT_die_ref): Likewise.
(const_ok_for_output_1, mem_loc_descriptor): Likewise.
(loc_list_from_tree, gen_lexical_block_die): Likewise.
gen_type_die_with_usage, gen_type_die): Likewise.
(dwarf2out_decl): Likewise.
* emit-rtl.c (verify_rtx_sharing, reorder_insns_nobb): Likewise.
* except.c (duplicate_eh_regions): Likewise.
* fwprop.c (register_active_defs, update_df_init): Likewise.
(fwprop_init, fwprop_done): Likewise.
(update_uses): Likewise.
* ggc-page.c (ggc_grow): Likewise.
* gimplify.c (gimplify_body): Likewise.
(gimplify_hasher::equal): Likewise.
* graphite-isl-ast-to-gimple.c (graphite_verify): Likewise.
* graphite-scop-detection.c (canonicalize_loop_closed_ssa_form):
Likewise.
* graphite-sese-to-poly.c (rewrite_reductions_out_of_ssa): Likewise.
(rewrite_cross_bb_scalar_deps_out_of_ssa): Likwise.
* hash-table.h (::find_empty_slot_for_expand): Likewise.
* ifcvt.c (if_convert): Likewise.
* ipa-cp.c (ipcp_propagate_stage): Likewise.
* ipa-devirt.c (type_in_anonymous_namespace_p): Likewise.
(odr_type_p, odr_types_equivalent_p): Likewise.
(add_type_duplicate, get_odr_type): Likewise.
* ipa-icf.c (sem_item_optimizer::execute): Likewise.
(sem_item_optimizer::subdivide_classes_by_equality): Likewise.
(sem_item_optimizer::verify_classes): Likewise.
(sem_item_optimizer::traverse_congruence_split): Likewise.
(sem_item_optimizer::checking_verify_classes): New.
* ipa-icf.h (sem_item_optimizer::checking_verify_classes): Add new
method.
* cfgrtl.c (commit_edge_insertions): Likewise.
(fixup_reorder_chain, cfg_layout_finalize): Likewise.
(rtl_flow_call_edges_add): Likewise.
* cgraph.c (symbol_table::create_edge): Likewise.
(cgraph_edge::redirect_call_stmt_to_callee): Likewise.
* cgraph.h (symtab_node): Likewise.
(symtab_node::checking_verify_symtab_nodes): Define.
(cgraph_node::checking_verify_cgraph_nodes): Define.
* cfghooks.h (checking_verify_flow_info): Define.
* cfgloop.h (checking_verify_loop_structure): Define.
* dominance.h (checking_verify_dominators): Define.
* et-forest.c: Fix comment.
* ipa-inline-analysis.c (compute_inline_parameters): Use flag_checking,
CHECKING_P gcc_checking_assert and checking_* functions to eliminate
ENABLE_CHECKING conditionals.
* ipa-inline-transform.c (save_inline_function_body): Likewise.
* ipa-inline.c (inline_small_functions): Likewise.
(early_inliner): Likewise.
* ipa-inline.h (estimate_edge_growth): Likewise.
* ipa-visibility.c (function_and_variable_visibility): Likewise.
* ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
(ipa_single_use): Likewise.
* ira-int.h: Likewise.
* ira.c (ira): Likewise.
* loop-doloop.c (doloop_optimize_loops): Likewise.
* loop-init.c (loop_optimizer_init, fix_loop_structure): Likewise.
* loop-invariant.c (move_loop_invariants): Likewise.
* lra-assigns.c (lra_assign): Likewise.
* lra-constraints.c (lra_constraints): Likewise.
* lra-eliminations.c (lra_eliminate): Likewise.
* lra-int.h (struct lra_reg): Likewise.
* lra-lives.c (check_pseudos_live_through_calls): Likewise.
(lra_create_live_ranges_1): Likewise.
* lra-remat.c (create_remat_bb_data): Likewise.
* lra.c (lra_update_insn_recog_data, restore_scratches): Likewise.
(lra): Likewise.
(check_rtl): Always define. Remove incorrect guard around
extract_constrain_insn call.
* lto-cgraph.c (input_cgraph_1: Use flag_checking,
CHECKING_P gcc_checking_assert and checking_* functions to eliminate
ENABLE_CHECKING conditionals.
* lto-streamer-out.c (DFS::DFS): Likewise.
(lto_output): Likewise.
* lto-streamer.c (lto_streamer_init): Likewise.
* omp-low.c (scan_omp_target, expand_omp_taskreg): Likewise.
expand_omp_target, execute_expand_omp): Likewise.
(lower_omp_target): Likewise.
* passes.c (execute_function_todo): Likewise.
(execute_todo, execute_one_pass): Likewise.
(verify_curr_properties): Always define.
* predict.c (tree_estimate_probability: Use flag_checking,
CHECKING_P gcc_checking_assert and checking_* functions to eliminate
ENABLE_CHECKING conditionals.
(propagate_freq): Likewise.
* pretty-print.c (pp_format): Likewise.
* real.c (real_to_decimal_for_mode): Likewise.
* recog.c (split_all_insns): Likewise.
* regcprop.c (kill_value_one_regno): Likewise.
(copy_value): Likewise.
(validate_value_data): Define unconditionally.
* reload.c: Fix comment.
* timevar.c: Include options.h
* tree-ssa.h (checking_verify_ssa): Define.
* tree-ssa-loop-manip.h (checking_verify_loop_closed_ssa): Define.
* sched-deps.c (CHECK): Remove unused macro.
(add_or_update_dep_1, sd_add_dep: Use flag_checking, CHECKING_P
gcc_checking_assert and checking_* functions to eliminate
ENABLE_CHECKING conditionals.
* sel-sched-ir.c (free_regset_pool, tidy_control_flow): Likewise.
* sel-sched.c (struct moveop_static_params): Likewise.
(find_best_reg_for_expr, move_cond_jump): Likewise.
(move_op_orig_expr_not_found): Likewise.
(code_motion_process_successors, move_op): Likewise.
* ssa-iterators.h (first_readonly_imm_use): Likewise.
(next_readonly_imm_use): Likewise.
* store-motion.c (compute_store_table): Likewise.
* symbol-summary.h (function_summary::function_summary): Likewise.
* target.h (cumulative_args_t): Likewise.
(get_cumulative_args, pack_cumulative_args): Likewise.
* timevar.c: (timer::print): Likewise.
* trans-mem.c (ipa_tm_execute): Likewise.
* tree-cfg.c (move_stmt_op): Likewise.
(move_sese_region_to_fn): Likewise.
(gimple_flow_call_edges_add): Likewise.
* tree-cfgcleanup.c (cleanup_tree_cfg_noloop, repair_loop_structures):
Likewise.
* tree-eh.c (remove_unreachable_handlers): Likewise.
* tree-if-conv.c (pass_if_conversion::execute): Likewise.
* tree-inline.c (expand_call_inline, optimize_inline_calls): Likewise.
* tree-into-ssa.c (update_ssa): Likewise.
* tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
* tree-outof-ssa.c (eliminate_useless_phis, rewrite_trees): Likewise.
* tree-parloops.c (pass_parallelize_loops::execute): Likewise.
* tree-predcom.c (suitable_component_p): Likewise.
* tree-profile.c (gimple_gen_const_delta_profiler): Likewise.
* tree-ssa-alias.c (refs_may_alias_p_1): Likewise.
* tree-ssa-live.c (verify_live_on_entry): Likewise.
* tree-ssa-live.h (register_ssa_partition): Likewise.
* tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely): Likewise.
* tree-ssa-loop-manip.c (add_exit_phi): Likewise.
(tree_transform_and_unroll_loop): Likewise.
* tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
* tree-ssa-operands.c (get_expr_operands): Likewise.
* tree-ssa-propagate.c (replace_exp_1): Likewise.
* tree-ssa-structalias.c (rewrite_constraints): Likewise.
* tree-ssa-ter.c (free_temp_expr_table): Likewise.
* tree-ssa-threadupdate.c (duplicate_thread_path): Likewise.
* tree-ssanames.c (release_ssa_name_fn): Likewise.
* tree-stdarg.c (expand_ifn_va_arg): Likewise.
* tree-vect-loop-manip.c
(slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
(slpeel_checking_verify_cfg_after_peeling): Likewise.
(vect_do_peeling_for_loop_bound): Likewise.
(vect_do_peeling_for_alignment): Likewise.
* tree-vrp.c (supports_overflow_infinity): Likewise.
(set_value_range): Likewise.
* tree.c (free_lang_data_in_cgraph): Likewise.
* value-prof.c (gimple_remove_histogram_value): Likewise.
(free_hist): Likewise.
* var-tracking.c (canonicalize_values_star): Likewise.
(compute_bb_dataflow, vt_find_locations, vt_emit_notes): Likewise.

From-SVN: r229470

Daily bump.

From-SVN: r229469

internal-fn.def (IFN_GOACC_DIM_SIZE, [...]): New.

* internal-fn.def (IFN_GOACC_DIM_SIZE, IFN_GOACC_DIM_POS,
IFN_GOACC_LOOP): New.
* internal-fn.h (enum ifn_unique_kind): Add IFN_UNIQUE_OACC_FORK,
IFN_UNIQUE_OACC_JOIN, IFN_UNIQUE_OACC_HEAD_MARK,
IFN_UNIQUE_OACC_TAIL_MARK.
(enum ifn_goacc_loop_kind): New.
* internal-fn.c (expand_UNIQUE): Add IFN_UNIQUE_OACC_FORK,
IFN_UNIQUE_OACC_JOIN cases.
(expand_GOACC_DIM_SIZE, expand_GOACC_DIM_POS): New.
(expand_GOACC_LOOP): New.
* target-insns.def (oacc_dim_pos, oacc_dim_size): New.
* omp-low.c: Include gimple-pretty-print.h.
(struct oacc_loop): New.
(enum oacc_loop_flags): New.
(oacc_thread_numbers): New.
(oacc_xform_loop): New.
(new_oacc_loop_raw, new_oacc_loop_outer, new_oacc_loop,
new_oacc_loop_routine, finish_oacc_loop, free_oacc_loop): New,
(dump_oacc_loop_part, dump_oacc_loop, debug_oacc_loop): New,
(oacc_loop_discover_walk, oacc_loop_sibling_nrevers,
oacc_loop_discovery): New.
(oacc_loop_xform_head_tail, oacc_loop_xform_loop,
oacc_loop_process): New.
(oacc_loop_fixed_partitions, oacc_loop_partition): New.
(execute_oacc_device_lower): Discover & process loops.  Process
internal fns.
* target.def (goacc.fork_join): Change sense of hook, clarify
documentation.
* doc/tm.texi: Regenerated.

From-SVN: r229466

target-insns.def (oacc_fork, oacc_join): Define.

* target-insns.def (oacc_fork, oacc_join): Define.
* target.def (goacc.validate_dims): Adjust doc to avoid warning.
(goacc.fork_join): New GOACC hook.
* targhooks.h (default_goacc_fork_join): Declare.
* omp-low.c (default_goacc_forkjoin): New.
* doc/tm.texi.in (TARGET_GOACC_FORK_JOIN): Add.
* doc/tm.texi: Regenerate.

From-SVN: r229465

omp-low.c (oacc_init_rediction_array): New.

* omp-low.c (oacc_init_rediction_array): New.
(oacc_initialize_reduction_data): Initialize array.

From-SVN: r229464

* omp-low.c (pass_oacc_device_lower::execute): Ignore errors.

From-SVN: r229462

c-parser.c (c_parser_oacc_shape_clause): New.

gcc/c/
* c-parser.c (c_parser_oacc_shape_clause): New.
(c_parser_oacc_simple_clause): New.
(c_parser_oacc_all_clauses): Add auto, gang, seq, vector, worker.
(OACC_LOOP_CLAUSE_MASK): Add gang, worker, vector, auto, seq.

gcc/testsuite/
* c-c++-common/goacc/loop-shape.c: New test.

Co-Authored-By: Bernd Schmidt <bernds@redhat.com>
Co-Authored-By: James Norris <jnorris@codesourcery.com>
Co-Authored-By: Joseph Myers <joseph@codesourcery.com>
Co-Authored-By: Julian Brown <julian@codesourcery.com>
Co-Authored-By: Thomas Schwinge <thomas@codesourcery.com>
From-SVN: r229461

parser.c (cp_parser_omp_clause_name): Add auto, gang, seq, vector, worker.

gcc/cp/
* parser.c (cp_parser_omp_clause_name): Add auto, gang, seq,
vector, worker.
(cp_parser_oacc_simple_clause): New.
(cp_parser_oacc_shape_clause): New.
(cp_parser_oacc_all_clauses): Add auto, gang, seq, vector, worker.
(OACC_LOOP_CLAUSE_MASK): Likewise.
* semantics.c (finish_omp_clauses): Add auto, gang, seq, vector,
worker. Unify the handling of teams, tasks and vector_length with
the other loop shape clauses.

gcc/testsuite/
* g++.dg/g++.dg/gomp/pr33372-1.C: Adjust diagnostic.
* gcc/testsuite/g++.dg/gomp/pr33372-3.C: Likewise.

From-SVN: r229460

internal-fn.c (expand_UNIQUE): New.

* internal-fn.c (expand_UNIQUE): New.
* internal-fn.h (enum ifn_unique_kind): New.
* internal-fn.def (IFN_UNIQUE): New.
* target-insns.def (unique): Define.
* gimple.h (gimple_call_internal_unique_p): New.
* gimple.c (gimple_call_same_target_p): Check internal fn
uniqueness.
* tracer.c (ignore_bb_p): Check for IFN_UNIQUE call.
* tree-ssa-threadedge.c
(record_temporary_equivalences_from_stmts): Likewise.
* tree-cfg.c (gmple_call_initialize_ctrl_altering): Likewise.

From-SVN: r229459

re PR rtl-optimization/67609 (Generates wrong code for SSE2 _mm_load_pd)

PR rtl-opt/67609

* config/i386/i386.c (ix86_cannot_change_mode_class): Disallow
narrowing subregs on SSE and MMX registers.
* doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Clarify when subregs that
appear to be sub-words of multi-register pseudos must be rejected.
* doc/tm.texi: Regenerate.
testsuite/
* gcc.target/i386/pr67609-2.c: New test.

From-SVN: r229458

[AArch64] PR 68102: Check that operand is REG before checking the REGNO in mov-immediate splitters

PR target/68102
* config/aarch64/aarch64.md (*movsi_aarch64): Check that
operands[0] is a reg before taking its REGNO in split condition.
(*movdi_aarch64): Likewise.

* gcc.target/aarch64/pr68102_1.c: New test.

From-SVN: r229455

[AArch64] Handle vector float modes properly in aarch64_output_simd_mov_immediate

* config/aarch64/aarch64.c (aarch64_output_simd_mov_immediate):
Handle floating point inner modes properly.

From-SVN: r229454

re PR fortran/67933 (ICE for array of a derived type with allocatable class in derived type object)

2015-01-27  Paul Thomas  <pault@gcc.gnu.org>

PR fortran/67933
* gfortran.dg/allocate_with_source_15.f03: New test

From-SVN: r229452

[PR fortran/63865] OpenACC cache directive: match Fortran support with C/C++

gcc/fortran/
PR fortran/63865
* openmp.c (resolve_oacc_cache): Remove function.
(gfc_match_oacc_cache): Enable array sections.
(resolve_omp_clauses, gfc_resolve_oacc_directive): Change
accordingly.
* trans-openmp.c (gfc_trans_omp_clauses): Likewise.
gcc/testsuite/
PR fortran/63865
* gfortran.dg/goacc/coarray.f95: Expect the OpenACC cache
directive to work.
* gfortran.dg/goacc/loop-1.f95: Likewise.
* gfortran.dg/goacc/cache-1.f95: Likewise, and extend testing.
* gfortran.dg/goacc/cray.f95: Likewise.
* gfortran.dg/goacc/parameter.f95: Likewise.

Co-Authored-By: James Norris <jnorris@codesourcery.com>
From-SVN: r229448

2015-10-27  Alan Hayward  <alan.hayward@arm.com>

* tree-vect-looop.c
(vectorizable_live_operation): Change iterator.

From-SVN: r229447

re PR fortran/68108 (erroneous error message 'scalar integer expression expected')

2015-10-27  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/68108
* decl.c (char_len_param_value): Check for REF_ARRAY.

2015-10-27  Steven G. Kargl  <kargl@gcc.gnu.org>

PR fortran/68108
* gfortran.dg/pr67805_2.f90: New test.

From-SVN: r229446

isl schedule tree

Use isl_schedule_node instead of isl_band_list for isl-0.15.
Passes regtest and bootstrap for isl-0.15 and isl-0.12.2 on x86_64-linux.

gcc/ChangeLog:

        * graphite-optimize-isl.c (get_schedule_for_node_st): New callback
          function to schedule based on isl_schedule_node.
        (get_schedule_map_st): New schedule optimizer based on isl_schedule_node.
        (scop_get_domains): New. Return the isl_union_set containing the domains of all the pbbs.
        (optimize_isl): Call the new function get_schedule_map_st for isl-0.15

gcc/testsuite/ChangeLog:

        * gcc.dg/graphite/block-0.c: Changed to match pattern.
        * gcc.dg/graphite/interchange-1.c: Same.
        * gcc.dg/graphite/interchange-10.c: Same.
        * gcc.dg/graphite/interchange-11.c: Same.
        * gcc.dg/graphite/interchange-13.c: Same.
        * gcc.dg/graphite/interchange-3.c: Same.
        * gcc.dg/graphite/interchange-4.c: Same.
        * gcc.dg/graphite/interchange-7.c: Same.
        * gcc.dg/graphite/interchange-9.c: Same.
        * gcc.dg/graphite/uns-interchange-9.c: Same.
        * gfortran.dg/graphite/interchange-3.f90: Same.

Co-Authored-By: Aditya Kumar <aditya.k7@samsung.com>
From-SVN: r229445

Properly handle -fno-plt in ix86_expand_call

prepare_call_address in calls.c is the wrong place to handle -fno-plt.
We shoudn't force function address into register and hope that load
function address via GOT and indirect call via register will be folded
into indirect call via GOT, which doesn't always happen.  Also non-PIC
case can only be handled in backend.  Instead, backend should expand
external function call into indirect call via GOT for -fno-plt.

This patch reverts -fno-plt in prepare_call_address and handles it in
ix86_expand_call.  Other backends may need similar changes to support
-fno-plt.  Alternately, we can introduce a target hook to indicate
whether an external function should be called via register for -fno-plt
so that i386 backend can disable it in prepare_call_address.

gcc/

PR target/67215
* calls.c (prepare_call_address): Don't handle -fno-plt here.
* config/i386/i386.c (ix86_expand_call): Generate indirect call
via GOT for -fno-plt.  Support indirect call via GOT for x32.
* config/i386/predicates.md (sibcall_memory_operand): Allow
GOT memory operand.

gcc/testsuite/

PR target/67215
* gcc.target/i386/pr67215-1.c: New test.
* gcc.target/i386/pr67215-2.c: Likewise.
* gcc.target/i386/pr67215-3.c: Likewise.

From-SVN: r229444

re PR tree-optimization/68104 (ice in vect_update_misalignment_for_peel with -O3)

2015-10-27  Richard Biener  <rguenther@suse.de>

PR tree-optimization/68104
* tree-vect-data-refs.c (vect_compute_data_ref_alignment): Move
strided access check ...
(vect_compute_data_refs_alignment): ... here.

* gcc.dg/torture/pr68104.c: New testcase.

From-SVN: r229440

config.gcc: Handle --enable-fdpic.

gcc/ChangeLog
* config.gcc: Handle --enable-fdpic.
* config/sh/constraints.md (Ccl): New constraint.
* config/sh/linux.h (SUBTARGET_LINK_EMUL_SUFFIX): Handle -mfdpic.
* config/sh/sh-c.c (sh_cpu_cpp_builtins): Add __FDPIC__ and
__SH_FDPIC__.
* config/sh/sh-mem.cc (expand_block_move): Support FDPIC for calls to
library functions.
* config/sh/sh-protos.h (function_symbol_result): New struct.
(function_symbol): Return function_symbol_result.
(sh_get_fdpic_reg_initial_val, sh_load_function_descriptor): New
declarations.
* config/sh/sh.c (TARGET_ASM_INTEGER, sh_assemble_integer): Implement
target hook.
(TARGET_CANNOT_FORCE_CONST_MEM, sh_cannot_force_const_mem_p): Likewise.
(sh_option_override): Force -fPIC if FDPIC is in effect.
(sh_asm_output_addr_const_extra): Add UNSPEC_GOTFUNCDESC and
UNSPEC_GOTOFFFUNCDESC cases.
(prepare_move_operands): Use FDPIC initial GOT register for
TLS-related GOT access; inhibit cross-section address offset constants
for FDPIC.
(sh_assemble_integer): New function.
(sh_cannot_copy_insn_p): Inhibit copying insns that are FDPIC
PC-relative call sites.
(expand_ashiftrt): Adapt invocation of function_symbol.
(sh_expand_prologue): Inhibit PC-relative GOT address load for FDPIC.
(nonpic_symbol_mentioned_p): Add cases for UNSPEC_GOTFUNCDESC and
UNSPEC_GOTOFFFUNCDESC.
(legitimize_pic_address): Resolve function symbols to function
descriptors for FDPIC.  Do not use GOT-relative addressing for local
data that may be read-only on FDPIC.
(sh_emit_storesi, sh_emit_storehi): New functions.
(sh_trampoline_init): Generate FDPIC trampolines.
(sh_function_ok_for_sibcall): Add TARGET_FDPIC check.
(sh_expand_sym_label2reg): Don't assume sibcalls are local.
(sh_output_mi_thunk): Generate FDPIC call.
(function_symbol): Return function_symbol_result.  For SFUNC_STATIC on
FDPIC, generate call site labels to use PC-relative addressing rather
than GOT-relative addressing.
(sh_conditional_register_usage): Make PIC register fixed and call used
when FDPIC is in effect.
(sh_legitimate_constant_p): Impose FDPIC constant constraints.
(sh_cannot_force_const_mem_p, sh_load_function_descriptor,
sh_get_fdpic_reg_initial_val): New functions.
* config/sh/sh.h (SUBTARGET_ASM_SPEC, SUBTARGET_LINK_EMUL_SUFFIX):
Handle -mfdpic.
(FDPIC_SELF_SPECS, SUBTARGET_DRIVER_SELF_SPECS,
PIC_OFFSET_TABLE_REG_CALL_CLOBBERED,
SH_OFFSETS_MUST_BE_WITHIN_SECTIONS_P): New macros.
(DRIVER_SELF_SPECS): Add SUBTARGET_DRIVER_SELF_SPECS and
FDPIC_SELF_SPECS.
(TRAMPOLINE_SIZE): Select trampoline size for FDPIC.
(ASM_PREFERRED_EH_DATA_FORMAT): Add EH format constraints for FDPIC.
(ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): Handle FDPIC case.
* config/sh/sh.md (UNSPEC_GOTFUNCDESC, UNSPEC_GOTOFFFUNCDESC): New
constants.
(calli_fdpic, call_valuei_fdpic, sibcalli_fdpic, sibcalli_pcrel_fdpic,
sibcall_pcrel_fdpic, sibcall_valuei_fdpic, sibcall_valuei_pcrel_fdpic,
sibcall_value_pcrel_fdpic, sym2GOTFUNCDESC, symGOTFUNCDESC2reg,
sym2GOTOFFFUNCDESC, symGOTOFFFUNCDESC2reg): New patterns.
(udivsi3_i1, udivsi3_i4, udivsi3_i4_single, udivsi3,
*divsi_inv_call_combine, divsi3_i4, divsi3_i4_single, divsi3, ashlsi3,
ashlsi3_d_call, ashrsi3_n, lshrsi3, lshrsi3_d_call, calli, call_valuei,
call, call_value, sibcalli, sibcalli_pcrel, sibcall_pcrel, sibcall,
sibcall_valuei, sibcall_valuei_pcrel, sibcall_value_pcrel,
sibcall_value, GOTaddr2picreg, symGOT_load, symGOTOFF2reg,
block_move_real, block_lump_real, block_move_real_i4,
block_lump_real_i4): Add support for FDPIC calls.
(mulsi3, ic_invalidate_line, initialize_trampoline, call_pop,
call_value_pop): Adjust for new function_symbol signature.
* config/sh/sh.opt (-mfdpic): New option.
* doc/install.texi (Options specification): Document --enable-fdpic.
* doc/invoke.texi (SH Options): Document -mfdpic.

include/ChangeLog:
* longlong.h (udiv_qrnnd): Add FDPIC compatible version for SH.

libitm/ChangeLog:
* config/sh/sjlj.S (_ITM_beginTransaction): Bypass PLT calling
GTM_begin_transaction for compatibility with FDPIC.

Co-Authored-By: Andrew Stubbs <ams@codesourcery.com>
Co-Authored-By: Joseph Myers <joseph@codesourcery.com>
Co-Authored-By: Mark Shinwell <shinwell@codesourcery.com>
Co-Authored-By: Rich Felker <dalias@libc.org>
From-SVN: r229438

tree-scalar-evolution.c: Handle LSHIFT by constant

gcc/:

tree-scalar-evolution.c (interpret_rhs_expr): Handle some LSHIFT_EXPRs
as equivalent MULT_EXPRs.

gcc/testsuite/:

gcc.dg/vect/vect-strided-shift.c: New.

From-SVN: r229437

[ARM] PR target/67929 Tighten vfp3_const_double_for_bits checks

PR target/67929
* config/arm/arm.c (vfp3_const_double_for_bits): Rewrite.
* config/arm/constraints.md (Dp): Update callsite.
* config/arm/predicates.md (const_double_vcvt_power_of_two): Likewise.

* gcc.target/arm/pr67929_1.c: New test.

From-SVN: r229436

Move ldexp, scalbn and scalbln folds to match.pd

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* builtins.c (fold_builtin_load_exponent): Rename to...
(fold_const_builtin_load_exponent): ...this and only handle
constant arguments.
(fold_builtin_2): Update accordingly.
* match.pd: Add rules previously handled by fold_builtin_load_exponent.

gcc/testsuite/
* gcc.dg/torture/builtin-ldexp-1.c: Skip at -O9,

From-SVN: r229434

Rename logb and significand folds

fold_builtin_logb and fold_builtin_significand now only handle
constant arguments, so this patch renames them to fold_const...,
to match fold_const_builtin_pow.  The idea is to differentiate
constant-only folds so that they can be moved to a const_binop-like
function in future.

The functions also had some unnecessary calls to STRIP_NOPS, which
I think are left over from code that has already moved to match.pd.

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* builtins.c (fold_builtin_logb): Rename to...
(fold_const_builtin_logb): ...this and remove STRIP_NOPS call.
(fold_builtin_significand): Rename to...
(fold_const_builtin_significand): ...this and remove STRIP_NOPS call.
(fold_builtin_1): Update accordingly.

From-SVN: r229433

Move fmin and fmax folds to match.pd

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* builtins.c (fold_builtin_fmin_fmax): Delete.
(fold_builtin_2): Handle constant fmin and fmax arguments here.
* match.pd: Add rules previously handled by fold_builtin_fmin_fmax.

From-SVN: r229432

[PATCH] [AArch64] Distinct costs for sign and zero extension

gcc/

* config/aarch64/aarch64-protos.h (cpu_addrcost_table): Split member
for register extension into sign and zero register extension.
* config/aarch64/aarch64.c (generic_addrcost_table): Infer values
for sign and zero register extension.
(cortexa57_addrcost_table): Likewise.
(xgene1_addrcost_table): Likewise.

From-SVN: r229431

Move min(max...) and max(min...) folds to match.pd

This handles both integer and floating-point arguments.  It's needed
for the follow-on patch to move fmin and fmax to match.pd.

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* fold-const.c (fold_minmax): Delete.
(fold_binary_loc): Don't call it.
* match.pd: Add rules previously handled by fold_minmax.

From-SVN: r229430

[multiple changes]

2015-10-27  Tristan Gingold  <gingold@adacore.com>

* bindgen.adb (System_BB_CPU_Primitives_Multiprocessors_Used):
New variable.
(Gen_Adainit): Call Start_All_CPUs if the above
variable is set to true.

2015-10-27  Emmanuel Briot  <briot@adacore.com>

* adaint.c, s-os_lib.adb, s-os_lib.ads (Copy_File_Attributes): New
subprogram.

From-SVN: r229429

Split constant handling out of fold_builtin_fma

Just makes an upcoming patch a bit clearer.

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* builtins.c (fold_builtin_fma): Remove constant handling.
(fold_builtin_3): Handle constant fma arguments here.

From-SVN: r229428

Remove constant handling from fold_builtin_{,f}abs

fold_builtin_fabs and fold_builtin_abs had code to handle constant
arguments, but this simply duplicated what the following fold_build1_loc
would do for ABS_EXPR.

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* builtins.c (fold_builtin_fabs): Remove constant handling.
(fold_builtin_abs): Likewise.

From-SVN: r229427

namet.adb, namet.ads: Minor reformatting.

2015-10-27  Hristian Kirtchev  <kirtchev@adacore.com>

* namet.adb, namet.ads: Minor reformatting.

From-SVN: r229426

Move copysign folds to match.pd

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* builtins.c (fold_builtin_copysign): Delete.
(fold_builtin_2): Handle constant copysign arguments here.
* match.pd: Add rules previously handled by fold_builtin_copysign.

From-SVN: r229425

[multiple changes]

2015-10-27  Ed Schonberg  <schonberg@adacore.com>

* sem_ch4.adb (Analyze_Allocator): Do not perform legality check
on allocators for limited objects in a qualified expression,
because expression has not been resolved.
* sem_res.adb (Resolve_Allocator): Perform check on legality of
limited objects after resolution.  Add sem_ch3.adb to context.

2015-10-27  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_prag.adb (Analyze_Refined_Global_In_Decl_Part): Add variable
States.
(Check_Refined_Global_Item): An object or state acts as a
constituent only when the corresponding encapsulating state
appears in pragma Global.
(Collect_Global_Item): Add a state with non-null visible refinement to
list States.

2015-10-27  Gary Dismukes  <dismukes@adacore.com>

* sem_util.ads, par.adb, sem_ch6.adb: Minor reformatting and a few
typo corrections.

2015-10-27  Pierre-Marie de Rodat  <derodat@adacore.com>

* namet.ads, namet.adb (Name_Equals): New function.
* namet.h (Name_Equals): New macro.

2015-10-27  Arnaud Charlet  <charlet@adacore.com>

* exp_ch6.adb (Build_Procedure_Form): Use 'RESULT' for the extra
parameter, to avoid ambiguity when generating tmps using _xxx which
might end up reusing _result.

From-SVN: r229424

Move signbit folds to match.pd

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* builtins.c (fold_builtin_signbit): Delete.
(fold_builtin_2): Handle constant signbit arguments here.
* match.pd: Add rules previously handled by fold_builtin_signbit.

gcc/testsuite/
PR tree-optimization/68097
* gcc.dg/torture/builtin-nonneg-1.c: Skip at -O0.  Add
--param max-ssa-name-query-depth=3 to dg-options.

From-SVN: r229423

Fold comparisons between sqrt and zero

The expression:

      signbit(sqrt(x))

is always 0 for -ffast-math.  The signbit fold first converts it to:

      sqrt(x) < 0

and whether we realise that this is false depends on a race between two
folders: the sqrt comparison folder, which wants to convert it to

      x < 0*0

and the generic tree_expr_nonnegative_p rule for ... < 0, which would
give the hoped-for 0.

The sqrt code already handles comparisons with negative values specially,
so this patch simply extends that idea to comparisons with zero.

Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.

gcc/
* match.pd: Handle sqrt(x) cmp 0 specially.

gcc/testsuite/
* gcc.dg/torture/builtin-sqrt-cmp-1.c: New test.

From-SVN: r229422

[multiple changes]

2015-10-27  Javier Miranda  <miranda@adacore.com>

* sem_util.ads, sem_util.adb (Defining_Identifier): Adding a formal to
indicate the needed behavior in case of nodes with errors.

2015-10-27  Ed Schonberg  <schonberg@adacore.com>

* sem_attr.adb (Eval_Attribute): If the prefix of attribute
Enum_Rep is an object that is a generated loop variable for an
element iterator, no folding is possible.
* sem_res.adb (Resolve_Entity_Name): Do not check for a missing
initialization in the case of a constant that is an object
renaming.
* exp_attr.adb (Expand_N_Attribute_Reference, case Enum_Rep):
If the prefix is a constant that renames an expression there is
nothing to evaluate statically.

2015-10-27  Vincent Celier  <celier@adacore.com>

* gnatlink.adb: Always delete the response file, even when the
invocation of gcc to link failed.

2015-10-27  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_ch6.adb (Analyze_Subprogram_Body_Helper):
Do not inherit the SPARK_Mode from the context if it has been
set already.
(Build_Subprogram_Declaration): Relocate relevant
pragmas from the subprogram body to the generated corresponding
spec. Do not copy aspect SPARK_Mode as this leads to circularity
in Copy_Separate_Tree. Inherit the attributes that describe
pragmas Ghost and SPARK_Mode.
(Move_Pragmas): New routine.

From-SVN: r229421

inline.adb (Is_Expression_Function): Removed.

2015-10-27  Hristian Kirtchev  <kirtchev@adacore.com>

* inline.adb (Is_Expression_Function): Removed.
* sem_ch6.adb (Analyze_Subprogram_Body_Helper): An internally
generated subprogram body that completes an expression function
inherits the SPARK_Mode from the spec.
* sem_res.adb (Resolve_Call): Update all calls to
Is_Expression_Function.
* sem_util.ads, sem_util.adb (Is_Expression_Function): Reimplemented.
(Is_Expression_Function_Or_Completion): New routine.

From-SVN: r229420

lib-xref-spark_specific.adb, [...]: Minor reformatting.

2015-10-27  Hristian Kirtchev  <kirtchev@adacore.com>

* lib-xref-spark_specific.adb, a-dirval-mingw.adb, exp_ch6.adb,
sem_ch8.adb, s-os_lib.adb: Minor reformatting.

From-SVN: r229419

[multiple changes]

2015-10-27  Yannick Moy  <moy@adacore.com>

* lib-xref-spark_specific.adb
(Enclosing_Subprogram_Or_Library_Package): detect library-level
subprograms and handle entries as subprograms, i.e. now both
library-level subprograms and entry bodies act as enclosing
scopes.
(Traverse_Declarations_Or_Statements): process entry bodies just
like subprogram bodies.
* lib-xref.ads (Enclosing_Subprogram_Or_Library_Package): comment
simplified while keeping its the meaning and minor typo fixed
("of" -> "or").
* spark_xrefs.ads (Xref Section): fix in description of the ALI
line for subprogram calls; such lines start with captial "F"
followed by a space.

2015-10-27  Ed Schonberg  <schonberg@adacore.com>

* sem_ch8.adb (Find_Direct_Name): A parameter association is
a legal context for applying an implicit dereference.
(Analyze_Expanded_Name): Minor code cleanup.

From-SVN: r229418

sinput.ads, [...]: Fix typos.

2015-10-27  Arnaud Charlet  <charlet@adacore.com>

* sinput.ads, spark_xrefs.ads, lib-xref.adb: Fix typos.

From-SVN: r229417

[multiple changes]

2015-10-27  Pascal Obry  <obry@adacore.com>

* a-dirval-mingw.adb: Remove some characters from Invalid_Character set.

2015-10-27  Ed Schonberg  <schonberg@adacore.com>

* exp_ch6.adb (Rewrite_Function_Call_For_C): Use a named
associaion for the added actual in the call because previous
actuals may also have been given by explicit associations.
* lib-xref.adb: Minor style fixes.

From-SVN: r229416

[multiple changes]

2015-10-27  Arnaud Charlet  <charlet@adacore.com>

* gnat1drv.adb: Remove hard coded restrictions.

2015-10-27  Pascal Obry  <obry@adacore.com>

* a-dirval-mingw.adb: Minor reformatting.

From-SVN: r229415

[multiple changes]

2015-10-27  Javier Miranda  <miranda@adacore.com>

* sem_ch6.adb (Analyze_Subprogram_Body_Helper): Improve previous patch.

2015-10-27  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_ch12.adb (Analyze_Formal_Package_Declaration): Code cleanup. Set
and restore the value of global flag Ignore_Pragma_SPARK_Mode. A
formal package declaration acts as a package instantation with
respect to SPARK_Mode legality.

2015-10-27  Hristian Kirtchev  <kirtchev@adacore.com>

* sem_prag.adb (Check_Constituent_Usage): Use
logical operators rather than short circuit operators. Emit an
error when a state with visible refinement is not refined.
* snames.ads-tmpl: Add names for detecting
predefined potentially blocking subprograms.

2015-10-27  Arnaud Charlet  <charlet@adacore.com>

* exp_aggr.adb (Aggr_Assignment_OK_For_Backend): Revert previous
change.
(Expand_Array_Aggregate): Rewrite previous change here
as done for other non GCC back-ends.
(Build_Record_Aggr_Code): Add special case.

From-SVN: r229414