Richard Biener [Tue, 15 Sep 2015 12:37:19 +0000 (12:37 +0000)]
re PR middle-end/67563 (verify_flow_info failed)
2015-09-15 Richard Biener <rguenther@suse.de>
PR middle-end/67563
* gimple-fold.c (gimplify_and_update_call_from_tree): Do not
transfer EH info from old to new stmt.
(replace_call_with_value): Likewise.
(replace_call_with_call_and_fold): Likewise.
(gimple_fold_builtin_memory_op): Likewise.
(gimple_fold_builtin_memset): Likewise.
(gimple_fold_builtin_stpcpy): Likewise.
(gimple_fold_call): Likewise.
This patch rewrites the shrink-wrapping algorithm, allowing non-linear
pieces of CFG to be duplicated for use without prologue instead of just
linear pieces.
* shrink-wrap.c (requires_stack_frame_p): Fix formatting.
(dup_block_and_redirect): Delete function.
(can_dup_for_shrink_wrapping): New function.
(fix_fake_fallthrough_edge): New function.
(try_shrink_wrapping): Rewrite function.
(convert_to_simple_return): Call fix_fake_fallthrough_edge.
The flag diagnostic_context::some_warnings_are_errors controls whether to give...
The flag diagnostic_context::some_warnings_are_errors controls whether
to give the message "all warnings being treated as errors". However, when
warnings are buffered and then discarded, this flag is not reset. It turns
out we do not need this flag at all, since we already count explicitly how
many warnings were converted into errors, and this number is kept up to
date for the buffered diagnostics used by Fortran.
gcc/ChangeLog:
2015-09-14 Manuel López-Ibáñez <manu@gcc.gnu.org>
PR fortran/67460
* diagnostic.c (diagnostic_initialize): Do not set
some_warnings_are_errors.
(diagnostic_finish): Use DK_WERROR count instead.
(diagnostic_report_diagnostic): Do not set
some_warnings_are_errors.
* diagnostic.h (struct diagnostic_context): Remove
some_warnings_are_errors.
gcc/testsuite/ChangeLog:
2015-09-14 Manuel López-Ibáñez <manu@gcc.gnu.org>
PR fortran/67460
* gfortran.dg/pr67460.f90: New test.
Ian Lance Taylor [Mon, 14 Sep 2015 17:54:08 +0000 (17:54 +0000)]
compiler: Ignore result context in constant expressions.
When evaluating a constant expression, the gofrontend would incorrectly
force each operand to be represented as the resulting type before
checking if the operation was valid with the untyped constants.
According to the language specification on constant
expressions(http://golang.org/ref/spec#Constant_expressions): "Untyped
boolean, numeric, and string constants may be used as operands wherever
it is legal to use an operand of boolean, numeric, or string type,
respectively."
for each M, so pointer equality with CONSTM1_RTX is enough. Also,
HOST_BITS_PER_WIDE_INT == 32 is doubly dead: HOST_WIDE_INT is always
64 bits now, and we always use const_int rather than const_double
or const_wide_int for all-ones values (or any other value that
fits in a signed HOST_WIDE_INT).
gcc/
* config/sparc/predicates.md (const_all_ones_operand): Use
CONSTM1_RTX to simplify definition.
* lto-wrapper.c (merge_and_complain): Add OPT_fdiagnostics_show_caret,
OPT_fdiagnostics_show_option, OPT_fdiagnostics_show_location_, and
OPT_fshow_column to handled saved option cases.
(append_compiler_options): Do not skip the above added options.
[AArch64] Handle literal pools for functions > 1 MiB in size.
This patch fixes the issue in PR63304 where we have
functions that are > 1MiB. The idea is to use adrp / ldr or adrp / add
instructions to address the literal pools under the use of a command line
option. I would like to turn this on by default on trunk but keep this
disabled by default for the release branches in order to get some
serious testing for this feature while it bakes on trunk.
As a follow-up I would like to try and see if estimate_num_insns or
something else can give us a heuristic to turn this on for "large" functions.
After all the number of incidences of this are quite low in real life,
so may be we should look to restrict this use as much as possible on the
grounds that this code generation implies an extra integer register for
addressing for every floating point and vector constant and I don't think
that's great in code that already may have high register pressure.
Tested on aarch64-none-elf with no regressions. A previous
version was bootstrapped and regression tested.
Mark Wielaard [Mon, 14 Sep 2015 09:49:47 +0000 (09:49 +0000)]
PR28901 -Wunused-variable ignores unused const initialised variables in C
12 years ago it was decided that -Wunused-variable shouldn't warn about
static const variables because some code used const static char rcsid[]
strings which were never used but wanted in the code anyway. But as the
bug points out this hides some real bugs. These days the usage of rcsids
is not very popular anymore. So this patch changes the default to warn
about unused static const variables in C with -Wunused-variable. And it
adds a new option -Wno-unused-const-variable to turn this warning off.
For C++ this new warning is off by default, since const variables can be
used as #defines in C++. New testcases for the new defaults in C and C++
are included testing the new warning and suppressing it with an unused
attribute or using -Wno-unused-const-variable.
gcc/ChangeLog
PR c/28901
* toplev.c (check_global_declaration): Check and use
warn_unused_const_variable.
* doc/invoke.texi (Warning Options): Add -Wunused-const-variable.
(-Wunused-variable): Remove non-constant. For C implies
-Wunused-const-variable.
(-Wunused-const-variable): New.
gcc/c-family/ChangeLog
PR c/28901
* c.opt (Wunused-variable): Option from common.opt.
(Wunused-const-variable): New option.
* config/i386/haswell.md: New file describing Haswell pipeline.
* config/i386/i386.c (processor_alias_table): Use CPU_HASWELL for
haswell-like processors.
(ix86_reassociation_width): Increase reassociation width for 64-bit
Haswell processor family.
* config/i386/i386.md: Introduce haswell cpu and include new md file.
This macro was converted to the TARGET_STRUCT_VALUE_RTX hook many years
ago, however there are still some lingering definitions, and a use in
libobjc. All the remaining definitions define the macro to 0, which
libobjc treats the same as undefined, so it won't break anything else to
have libobjc stop checking the macro. However it may be that this part
of libobjc has been broken for a long time on targets that only define
the hook, but that is a separate issue.
re PR libfortran/67527 (io.h sanitizer complains on 1 << 31)
PR libfortran/67527
PR libfortran/67535
PR libfortran/67536
* io/io.h: Use unsigned values for 31-bit left shifts.
* io/unix.c (buf_read): Do not call memcpy() with NULL pointer arg.
* io/write.c (nml_write_obj): Likewise.
David S. Miller [Sat, 12 Sep 2015 00:13:23 +0000 (00:13 +0000)]
Convert SPARC backend over to LRA.
gcc/
* config/sparc/constraints.md: Make "U" constraint a real register
constraint.
* config/sparc/sparc.c (TARGET_LRA_P): Define.
(D_MODES, DF_MODES): Add missing cast.
(TF_MODES, TF_MODES_NO_S): Include T_MODE.
(OF_MODES, OF_MODES_NO_S): Include O_MODE.
(sparc_register_move_cost): Decrease Niagara/UltrsSPARC memory
cost to 8.
* config/sparc/sparc.h (PROMOTE_MODE): Define.
* config/sparc/sparc.md (*movsi_lo_sum, *movsi_high): Do not
provide these insn when flag_pic.
Jeff Law [Fri, 11 Sep 2015 21:32:38 +0000 (15:32 -0600)]
[PATCH] Another small cleanup to the const_and_copies stack
2015-09-11 Jeff Law <law@redhat.com>
PR tree-optimization/47679
* tree-ssa-dom.c (struct cond_equivalence): Update comment.
* tree-ssa-scopedtables.h (class const_and_copies): Prefix data
member with m_. Update inline member functions as necessary. Add
toplevel comment.
* tree-ssa-scopedtables.c: Update const_and_copies's member
functions to use m_ prefix to access the stack.
The variable `static bool enable_polly_vector' is always assigned to false.
This results in dead code in optimize-isl.c. No functional change intended.
Passes bootstrap and regtest.
gcc/ChangeLog:
2015-09-11 Aditya Kumar <aditya.k7@samsung.com>
* graphite-optimize-isl.c (disable_tiling): Remove.
(get_schedule_for_band): Do not use disable_tiling.
(get_prevector_map): Delete function.
(enable_polly_vector): Remove.
(get_schedule_for_band_list): Remove dead code.
Refactor graphite-optimize-isl.c. Renamed function name, variable names etc.,
and indented the source according to gcc style guidelines. Modified comments
accordingly. No functional change intended.
re PR bootstrap/67363 (r227188 breaks build for mingw-w64)
PR bootstrap/67363
* configure.ac: Check if setenv and unsetenv are declared.
* configure: Rebuild.
* config.in: Rebuild.
* system.h: Declare setenv and unsetenv if not declared.
Ian Lance Taylor [Thu, 10 Sep 2015 21:17:00 +0000 (21:17 +0000)]
libgo: Use stat_atim.go on Solaris 12+
From Rainer Orth.
Solaris 12 changes the stat_[amc]tim members of struct stat from
timestruc_t to timespec_t for XPG7 compatiblity, thus breaking the libgo
build. The following patch checks for this change and uses the common
stat_atim.go if appropriate.
2015-09-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
* config/rs6000/rs6000.c (swap_web_entry): Update preceding
commentary to simplify permute mask adjustment equation.
(special_handling_values): Add SH_VPERM.
(const_load_sequence_p): New function.
(insn_is_swappable_p): Add logic to recognize an UNSPEC_VPERM with
the mask loaded from the constant pool.
(adjust_vperm): New function.
(handle_special_swappables): Call adjust_vperm.
(dump_swap_insn_table): Handle SH_VPERM.
[gcc/testsuite]
2015-09-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
* gcc.target/powerpc/swaps-p8-20.c: New test.
* gcc.target/powerpc/swaps-p8-21.c: New test.
Some host-target combinations get bogus warnings about orig_dup
being used uninitialized. I tried to coerce the current uninit
pass to handle this case, but the patch I ended up with only worked
by accident because of the strange way in which the pass handles
limit cases. (If we have more than MAX_NUM_CHAINS chains, it silently
drops the excess chains and continues regardless, so it's quite easy
to come up with cases where the predicates for either the definition
or the use consider an arbitrary subset of the actual conditions.)
For now this patch turns -Wmaybe-uninitialized into a warning for
the affacted function. It will be a warning even if someone turns
off warnings on the command line, but I don't think that's important.
Bootstrapped and regression-tested on x86_64-linux-gnu. Also tested
with a cross-compiler to sparc-linux-gnu (which also triggered the
warning for me).
gcc/
* reload1.c (elimination_costs_in_insn): Locally turn
-Wmaybe-uninitialized into a warning.
Chris Manghane [Thu, 10 Sep 2015 18:24:28 +0000 (18:24 +0000)]
compiler: Report errors from very large types.
The gcc backend throws an internal error when trying to get the size
of a type which is larger than the amount of address space on the
machine. This patch catches this error and reports it in a user
friendly way.
Mark Wielaard [Thu, 10 Sep 2015 15:29:44 +0000 (15:29 +0000)]
Warn when comparing nonnull arguments to NULL in a function.
If an argument is marked as nonnull then passing in a NULL argument
will produce bad results even if the code checks against NULL. GCC
might optimize such checks away so warn the user when the function
contains such comparisons.
nn.c: In function ‘foo’:
nn.c:6:27: warning: nonnull argument ‘bar’ compared to NULL [-Wnonnull]
void foo(void *bar) { if (!bar) abort(); }
^
gcc/ChangeLog
* doc/invoke.texi (Wnonnull): Also warns when comparing against NULL.
gcc/c/ChangeLog
* c-typeck.c (build_binary_op): Check and warn when nonnull arg
parm against NULL.
gcc/cp/ChangeLog
* typeck.c (cp_build_binary_op): Check and warn when nonnull arg
parm against NULL.
Paul Thomas [Thu, 10 Sep 2015 15:22:20 +0000 (15:22 +0000)]
re PR fortran/66993 (Spurious ambiguous symbol error with submodules)
2015-09-10 Paul Thomas <pault@gcc.gnu.org>
PR fortran/66993
* module.c (read_module): If a symtree exists and the symbol has
been associated in a submodule from a parent (sub)module, attach
the symbol to a 'unique symtree' and the new symbol to the
existing symtree.
2015-09-10 Paul Thomas <pault@gcc.gnu.org>
PR fortran/66993
* gfortran.dg/submodule_11.f08: New test.
Andreas Krebbel [Thu, 10 Sep 2015 14:04:28 +0000 (14:04 +0000)]
S/390: Don't use vgm for v1ti and v1tf.
gcc/ChangeLog:
2015-09-10 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* config/s390/s390.c (s390_contiguous_bitmask_vector_p): Reject if
the vector element is bigger than 64 bit.
gcc/testsuite/ChangeLog:
2015-09-10 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* gcc.target/s390/vector/vec-genbytemask-1.c: Add check for V1TI
initialization with a byte mask. No change expected here.
* gcc.target/s390/vector/vec-genmask-1.c: Fix whitespace.
* gcc.target/s390/vector/vec-genmask-2.c: Add check for V1TI
initialization with contigious bitmask. Literal pool is expectd
to be used here.
projects / thirdparty / gcc.git / log
