Ilya Leoshkevich [Fri, 15 Nov 2019 12:55:05 +0000 (12:55 +0000)]
Free dominance info at the beginning of pass_jump_after_combine
try_forward_edges does not update dominance info, and merge_blocks
relies on it being up-to-date. In PR92430 stale dominance info makes
merge_blocks produce a loop in the dominator tree, which in turn makes
delete_basic_block loop forever.
Fix by freeing dominance info at the beginning of cleanup_cfg.
Also, since the whole point of this pass is to perform jump threading
(other cleanups are not valuable at this point), skip it completely when
flag_thread_jumps is not set.
gcc/ChangeLog:
2019-11-15 Ilya Leoshkevich <iii@linux.ibm.com>
Backport from mainline
PR rtl-optimization/92430
* cfgcleanup.c (pass_jump_after_combine::gate): New function.
(pass_jump_after_combine::execute): Free
dominance info at the beginning.
gcc/testsuite/ChangeLog:
2019-11-15 Ilya Leoshkevich <iii@linux.ibm.com>
Backport from mainline
PR rtl-optimization/92430
* gcc.dg/pr92430.c: New test (from Arseny Solokha).
This was dormant for quite some time, but it started happening for me
on gcc.c-torture/compile/pr65153.c sometime after r276645 for -mabi=32 linux runs.
The pattern accepts any SMALL_OPERAND constant value while it asserts during the final
that the value is in the mode size range. I this case it happens that combine_and_move_insns
during ira makes a pattern with negative "shift count" which fails at final stage.
This simple fix just truncates the constant operand to mode size the same as shift patterns.
Eric Botcazou [Wed, 13 Nov 2019 11:59:20 +0000 (11:59 +0000)]
c-ada-spec.c (get_underlying_decl): Do not look through typedefs.
* c-ada-spec.c (get_underlying_decl): Do not look through typedefs.
(dump_forward_type): Do not generate a declaration for function types.
(dump_nested_type) <ARRAY_TYPE>: Do not generate a nested declaration
of the component type if it is declared in another file.
Thomas Schwinge [Mon, 11 Nov 2019 08:06:25 +0000 (09:06 +0100)]
[build] Properly track GCC language configure fragments
The 'gcc/configure' script sources all 'gcc/*/config-lang.in' files, but fails
to emit such dependency information into the build machinery. That means,
currently, when something gets changed in a 'gcc/*/config-lang.in' file, this
is not noticed, and doesn't propagate through the build machinery.
Handling of configure fragments is modelled in the same way as it already
exists for Makefile fragments.
gcc/
* Makefile.in (LANG_CONFIGUREFRAGS): Define.
(config.status): Use/depend on it.
* configure.ac (all_lang_configurefrags): Track, 'AC_SUBST'.
* configure: Regenerate.
backport: pa.md (memory_barrier): Revise to use ldcw barriers.
Backport from mainline
2019-11-07 John David Anglin <danglin@gcc.gnu.org>
* config/pa/pa.md (memory_barrier): Revise to use ldcw barriers.
Enhance comment.
(memory_barrier_coherent, memory_barrier_64, memory_barrier_32): New
insn patterns using ldcw instruction.
(memory_barrier): Remove insn pattern using sync instruction.
* config/pa/pa.opt (coherent-ldcw): New option.
(ordered): New option.
Jakub Jelinek [Fri, 8 Nov 2019 19:02:24 +0000 (20:02 +0100)]
re PR middle-end/92384 (Empty class instances have different equal testing result among GCC versions)
PR c++/92384
* function.c (assign_parm_setup_block, assign_parm_setup_stack): Don't
copy TYPE_EMPTY_P arguments from data->entry_parm to data->stack_parm
slot.
(assign_parms): For TREE_ADDRESSABLE parms with TYPE_EMPTY_P type
force creation of a unique data.stack_parm slot.
Jakub Jelinek [Fri, 8 Nov 2019 18:58:14 +0000 (19:58 +0100)]
backport: re PR middle-end/92231 (ICE in gimple_fold_stmt_to_constant_1)
Backported from mainline
2019-10-31 Jakub Jelinek <jakub@redhat.com>
PR middle-end/92231
* tree.h (fndecl_built_in_p): Use fndecl_built_in_p instead of
DECL_BUILT_IN in comment. Remove redundant ()s around return
argument.
* tree.c (free_lang_data_in_decl): Check if var is FUNCTION_DECL
before calling fndecl_built_in_p.
* gimple-fold.c (gimple_fold_stmt_to_constant_1): Check if
TREE_OPERAND (fn, 0) is a FUNCTION_DECL before calling
fndecl_built_in_p on it.
* lto-lang.c (handle_const_attribute): Don't call fndecl_built_in_p
on *node that is not FUNCTION_DECL.
Eric Botcazou [Fri, 8 Nov 2019 12:33:48 +0000 (12:33 +0000)]
re PR target/92095 (internal error with -O1 -mcpu=niagara2 -fPIE)
PR target/92095
* config/sparc/sparc-protos.h (output_load_pcrel_sym): Declare.
* config/sparc/sparc.c (sparc_cannot_force_const_mem): Revert latest
change.
(got_helper_needed): New static variable.
(output_load_pcrel_sym): New function.
(get_pc_thunk_name): Remove after inlining...
(load_got_register): ...here. Rework the initialization of the GOT
register and of the GOT helper.
(save_local_or_in_reg_p): Test the REGNO of the GOT register.
(sparc_file_end): Test got_helper_needed to decide whether the GOT
helper must be emitted. Use output_asm_insn instead of fprintf.
(sparc_init_pic_reg): In PIC mode, always initialize the PIC register
if optimization is enabled.
* config/sparc/sparc.md (load_pcrel_sym<P:mode>): Emit the assembly
by calling output_load_pcrel_sym.
S/390: Fix failing RTL check in s390_canonicalize_comparison
The new sigfpe-eh.c fails with
internal compiler error: RTL check: expected elt 0 type 'e' or 'u', have 'w' (rtx const_int)
This is most likely due to a typo: XEXP (*op1, 0) was used, when
XEXP (*op0, 1) was intended. This did not cause any user-visible
problems, because reversed_comparison_code_parts ignores the
respective argument, and the release compiler is built without RTL
checks.
gcc/ChangeLog:
2019-11-06 Ilya Leoshkevich <iii@linux.ibm.com>
* config/s390/s390.c (s390_canonicalize_comparison): Use XEXP
(*op0, 1) instead of XEXP (*op1, 0).
Backport from trunk
2019-10-26 Segher Boessenkool <segher@kernel.crashing.org>
PR target/91289
* config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't add an
immediate to r0; use r11 instead. Save and restore r11 to r0 around
this.
Andreas Krebbel [Tue, 5 Nov 2019 15:27:59 +0000 (15:27 +0000)]
IBM Z: gen-vect-26/28: Vectorizing without peeling is ok for Z
These tests check if loop peeling has been applied to avoid
having to vectorize unaligned loops. On Z we do not have any
alignment requirements for vectorization so we also don't need want
the loop peeling here.
2019-11-05 Andreas Krebbel <krebbel@linux.ibm.com>
Backport from mainline
2019-11-05 Andreas Krebbel <krebbel@linux.ibm.com>
* gcc.dg/tree-ssa/gen-vect-26.c: Disable loop peeling check for
IBM Z.
* gcc.dg/tree-ssa/gen-vect-28.c: Likewise.
Andreas Krebbel [Tue, 5 Nov 2019 15:26:19 +0000 (15:26 +0000)]
IBM Z: gen-vect-11/32: Set min-vect-loop-bound param back to default
In the Z backend we still set min-vect-loop-bound to 2 to work around
corner cases where awkward epilogue code gets generated in the
vectorizer. This has a particular bad impact when vectorizing loops
with a low iteration count. Due to this we do not vectorize the loop
in gen-vect-11/32 - what actually is a pity.
The patch sets min-vect-loop-bound back to the default value of 0 in
order to enable vectorization.
2019-11-05 Andreas Krebbel <krebbel@linux.ibm.com>
Backport from mainline
2019-11-05 Andreas Krebbel <krebbel@linux.ibm.com>
* gcc.dg/tree-ssa/gen-vect-11.c: Add --param min-vect-loop-bound=0
for IBM Z.
* gcc.dg/tree-ssa/gen-vect-11.c: Likewise.
Andreas Krebbel [Tue, 5 Nov 2019 15:25:26 +0000 (15:25 +0000)]
IBM Z: Fix testsuite useable_hw check
This fixes various issues with the useable_hw check in s390.exp. The
check is supposed to verify whether a testcase can be run on the
current hardware.
- the test never returned true for -m31 because vzero is not available
in ESA mode and -m31 defaults to -mesa
- the missing v0 clobber on the vzero instruction made the check fail
if the stack pointer got saved in f0
- the lcbb instruction used for checking whether we are on a z13
also requires vx. Replace it with an instruction from the generic
instruction set extensions.
- no support for z14 and z15 so far
gcc/testsuite/ChangeLog:
2019-11-05 Andreas Krebbel <krebbel@linux.ibm.com>
Backport from mainline
2019-11-05 Andreas Krebbel <krebbel@linux.ibm.com>
* gcc.target/s390/s390.exp
(check_effective_target_s390_useable_hw): Add inline asm for z14
and z15. Replace instruction for z13 with lochiz. Add register
clobbers. Check also for __zarch__ when doing the __VX__ test.
Iain Sandoe [Mon, 4 Nov 2019 23:06:21 +0000 (23:06 +0000)]
[objective-c/c++, testsuite] Fix stubify tests for -fnext-runtime.
Fixed to work for x86 as well as powerpc (use the final assembler output
instead of the jump rtl which varies between X86 and PPC). Amended the
options to reflect this change.
gcc/testsuite/
Backport from mainline.
2019-06-20 Iain Sandoe <iain@sandoe.co.uk>
Iain Sandoe [Wed, 30 Oct 2019 20:54:16 +0000 (20:54 +0000)]
[testsuite] Make the Wnonnull test independent of system headers.
This backports two patches that fix failure of Wnonnull test on some targets.
277280 To avoid the result of this test depending on the implementation of
the system 'string.h', provide prototypes for the two functions used
in the test. This generalises the fix from 277202.
277202 was the initial fix for Wnonnull on Darwin.
gcc/testsuite/
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline.
2019-10-22 Iain Sandoe <iain@sandoe.co.uk>
* gcc.dg/Wnonnull.c: Provide prototypes for strlen and memcpy.
Use __SIZE_TYPE__ instead of size_t.
Backport from mainline.
2019-10-19 Iain Sandoe <iain@sandoe.co.uk>
* gcc.dg/Wnonnull.c: Add attributed function declarations for
memcpy and strlen for Darwin.
Iain Sandoe [Wed, 30 Oct 2019 20:48:44 +0000 (20:48 +0000)]
[Darwin, machopic] Back out part of PR71767 fix.
This is part of making our codegen match what the system tools produce.
We applied a conservative, but fairly large, hammer to fix 71767.
However, ideally, we want minimise the number of symbols visible to
ld64 and to match the cases emitted by clang (since that's what ld64
is expecting). Now we've improved the handling of indirections, we
can make the indirection symbols local when they are in the regular
non-lazy symbol pointers section. We will continue to make any
indirections in the data section visible (since right now we have no
way to track if a given symbol follows a weak global).
This change makes no difference to handling of labels for constants
(to be revised in a future patch).
There's a mechanical change to a number of tests (allowing 'l' or 'L'
as the indirection symbol prefix).
gcc/
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-10-13 Iain Sandoe <iain@sandoe.co.uk>
* config/darwin.c (machopic_indirection_name): Rework the
function to emit linker-visible symbols only for indirections
in the data section. Clean up the code and update comments.
gcc/testsuite/
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline.
2019-10-13 Iain Sandoe <iain@sandoe.co.uk>
Iain Sandoe [Wed, 30 Oct 2019 20:38:47 +0000 (20:38 +0000)]
[Darwin] Some TLC for older Darwin versions.
This is part 3 of build fixes for older Darwin versions (and will also
fix testsuite complaints about linkage with objects that are built for
an newer OS version than is being built, when used with a modern linker).
The library handling and some of the options for creating the crts for
the older PPC Darwin versions had bit-rotted somewhat. This adjusts the
build criteria for the crts to avoid newer ld64 versions warnings about
mismatches in build and object versions.
Added to some of the comments that it is documented why the specs are as
they are.
gcc/
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-07-03 Iain Sandoe <iain@sandoe.co.uk>
* config/darwin.h (REAL_LIBGCC_SPEC): Adjust for earlier Darwin.
(STARTFILE_SPEC): Split crt3 into a separate spec.
(DARWIN_EXTRA_SPECS): Add crt2 and crt3 spec.
(DARWIN_CRT2_SPEC): New.
(DARWIN_CRT3_SPEC): New.
(MIN_LD64_OMIT_STUBS): Revise to 62.1.
* config/rs6000/darwin.h (DARWIN_CRT2_SPEC): Revise conditions.
(DARWIN_CRT3_SPEC): New.
libgcc/
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline.
2019-07-03 Iain Sandoe <iain@sandoe.co.uk>
* config.host (powerpc-*-darwin*,powerpc64-*-darwin*): Revise crt
list.
* config/rs6000/t-darwin: Build crt3_2 for older systems. Revise
mmacosx-version-min for crts to run across all system versions.
* config/rs6000/t-darwin64 (LIB2ADD): Remove.
* config/t-darwin: Revise mmacosx-version-min for crts to run across
system versions >= 10.4.
Iain Sandoe [Wed, 30 Oct 2019 20:25:33 +0000 (20:25 +0000)]
[Darwin, PPC] Move the out of line register save/restore to an endfile.
Part 2 of fixes to build and link conditions for older Darwin versions.
We have been including these endfiles in libgcc, which means that we have to
append -lgcc even when using shared libgcc. In preparation for revision of
libgcc split this into an endfile. Build the new version.
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline.
2019-06-25 Iain Sandoe <iain@sandoe.co.uk>
* config.host: Add libef_ppc.a to the extra files for powerpc-darwin.
* config/rs6000/t-darwin: (PPC_ENDFILE_SRC, PPC_ENDFILE_OBJS): New.
Build objects for the out of line save/restore register functions
so that they can be used for any supported Darwin version.
* config/t-darwin: Default the build Darwin version to Darwin8
(MacOS 10.4).
Iain Sandoe [Wed, 30 Oct 2019 20:12:17 +0000 (20:12 +0000)]
[Darwin, PPC] Move the out of line register save/restore to an endfile.
Part 1 of fixes to build and link conditions for older Darwin versions.
We have been including these endfiles in libgcc, which means that we have to
append -lgcc even when using shared libgcc. In preparation for revision of
libgcc split this into an endfile.
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-06-27 Iain Sandoe <iain@sandoe.co.uk>
* config/rs6000/darwin.h (ENDFILE_SPEC): Correct whitespace in the
spec.
Backport from mainline
2019-06-25 Iain Sandoe <iain@sandoe.co.uk>
Iain Sandoe [Wed, 30 Oct 2019 20:05:32 +0000 (20:05 +0000)]
[Darwin] The need for FDE symbols is dependent on linker used, not OS rev.
For very old toolchains, the compiler generated extra symbols that mark the
start of each FDE. We no longer need this (since xcode 3 era) - so, for
compatibility with newer linkers, omit this when it is not required.
Since we have detection of the linker version, we can use that directly to
determine if support is needed.
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-06-18 Iain Sandoe <iain@sandoe.co.uk>
* config/darwin.c (darwin_emit_unwind_label): New default to false.
(darwin_override_options): Set darwin_emit_unwind_label as needed.
Iain Sandoe [Wed, 30 Oct 2019 19:56:41 +0000 (19:56 +0000)]
[Darwin] Use of symbol stubs should depend on the linker in use.
This is a combination of two patches that prevents a bootstrap fail when
a modern assembler/linker pair are used that don't support the JBSR reloc
(but don't require it for long calls).
272356 The need for picsym stubs is dependent on linker used, not OS rev.
For very old toolchains, the compiler generated pic symbol stubs that
provide the necessary indirections. We no longer need this (since xcode
3 era) and it's more efficient for the linker to make one stub when it
knows a symbol is needed that for us to emit them speculatively in every
object.
Our current codegen is making the assumption that a specific OS version
uses a specific linker version - and therefore the presence of support
could be based on the target OS rev. Of course, that's way too simplistic
(most likely bogus for cross-toolchains) and we want to make things explict.
Since we have detection of the linker version, we can use that directly
(A config test for support for stub-less linking might also be feasible
but much more involved).
Finally, should the user wish to generate code that caters for export to
use in an environment with an older toolchain, the generation of stubs
can be forced from the command line.
In addition to the points above, branch islanding and a long branch opt
for PowerPC Darwin has become conflated with the emission of these stubs
(which is confusing, and will be rectified in a following patch).
274397 There is no need to distinguish PIC/non-PIC symbol stubs.
So we can use a single flag for both.
2019-10-30 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-08-13 Iain Sandoe <iain@sandoe.co.uk>
Thomas Schwinge [Wed, 30 Oct 2019 11:51:15 +0000 (12:51 +0100)]
[LIBPHOBOS] Fix multi-lib RUNTESTFLAGS handling
Testing libphobos fails for multi-lib configs:
$ make check-target-libphobos RUNTESTFLAGS="--target_board=unix\{-m32,\}"
fails for every 32bit execution, because the host libgcc_s.so is used which
is not the correct version:
spawn [open ...]
./test_aa.exe: /lib/i386-linux-gnu/libgcc_s.so.1: version `GCC_7.0.0' not found (required by ./test_aa.exe)
FAIL: libphobos.aa/test_aa.d execution test
This can be fixed by adding a few lines from libstdc++/testsuite/lib/libstdc++.exp
to libphobos/testsuite/lib/libphobos.exp
Thomas Schwinge [Wed, 30 Oct 2019 10:50:58 +0000 (11:50 +0100)]
Avoid '-Wmaybe-uninitialized' diagnostic in 'gcc/dwarf2out.c'
With '--enable-checking=yes,extra,rtl':
In file included from [...]/source-gcc/gcc/coretypes.h:433,
from [...]/source-gcc/gcc/dwarf2out.c:60:
[...]/source-gcc/gcc/machmode.h: In function 'dw_loc_descr_node* mem_loc_descriptor(rtx, machine_mode, machine_mode, var_init_status)':
[...]/source-gcc/gcc/machmode.h:520:42: error: 'int_mode' may be used uninitialized in this function [-Werror=maybe-uninitialized]
520 | ? mode_size_inline (mode) : mode_size[mode]);
| ^~~~
[...]/source-gcc/gcc/dwarf2out.c:15464:19: note: 'int_mode' was declared here
15464 | scalar_int_mode int_mode, inner_mode, op1_mode;
| ^~~~~~~~
cc1plus: all warnings being treated as errors
make[3]: *** [dwarf2out.o] Error 1
Paul Thomas [Mon, 28 Oct 2019 18:28:48 +0000 (18:28 +0000)]
re PR fortran/91926 (assumed rank optional)
2019-10-28 Paul Thomas <pault@gcc.gnu.org>
Backport from trunk
PR fortran/91926
* trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Correct the
assignment of the attribute field to account correctly for an
assumed shape dummy. Assign separately to the gfc and cfi
descriptors since the atribute can be different. Add branch to
correctly handle missing optional dummies.
2019-10-28 Paul Thomas <pault@gcc.gnu.org>
Backport from trunk
PR fortran/91926
* gfortran.dg/ISO_Fortran_binding_13.f90 : New test.
* gfortran.dg/ISO_Fortran_binding_13.c : Additional source.
* gfortran.dg/ISO_Fortran_binding_14.f90 : New test.
Ilya Leoshkevich [Mon, 28 Oct 2019 13:09:54 +0000 (13:09 +0000)]
Move jump threading before reload
r266734 has introduced a new instance of jump threading pass in order to
take advantage of opportunities that combine opens up. It was perceived
back then that it was beneficial to delay it after reload, since that
might produce even more such opportunities.
Unfortunately jump threading interferes with hot/cold partitioning. In
the code from PR92007, it converts the following
+---------------------- 2/HOT ------------------+
| |
v v
3/HOT --> 8/HOT --> 11/COLD --> 6/COLD --EH--> 16/HOT
This makes hot bb 6 dominated by cold bb 11, and because of this
fixup_partitions makes bb 6 cold as well, which in turn makes EH edge
6->16 a crossing one. Not only can't we have crossing EH edges, we are
also not allowed to introduce new crossing edges after reload in
general, since it might require extra registers on some targets.
Therefore, move the jump threading pass between combine and hot/cold
partitioning. Building SPEC 2006 and SPEC 2017 with the old and the new
code indicates that:
* When doing jump threading right after reload, 3889 edges are threaded.
* When doing jump threading right after combine, 3918 edges are
threaded.
This means this change will not introduce performance regressions.
gcc/ChangeLog:
2019-10-28 Ilya Leoshkevich <iii@linux.ibm.com>
Backport from mainline
PR rtl-optimization/92007
* cfgcleanup.c (thread_jump): Add an assertion that we don't
call it after reload if hot/cold partitioning has been done.
(class pass_postreload_jump): Rename to
pass_jump_after_combine.
(make_pass_postreload_jump): Rename to
make_pass_jump_after_combine.
* passes.def(pass_postreload_jump): Move before reload, rename
to pass_jump_after_combine.
* tree-pass.h (make_pass_postreload_jump): Rename to
make_pass_jump_after_combine.
gcc/testsuite/ChangeLog:
2019-10-28 Ilya Leoshkevich <iii@linux.ibm.com>
Backport from mainline
PR rtl-optimization/92007
* g++.dg/opt/pr92007.C: New test (from Arseny Solokha).
Peter Bergner [Mon, 28 Oct 2019 09:46:15 +0000 (09:46 +0000)]
[rs6000] PR70010, avoid no-vsx function to be inlined to vsx function
In PR70010, a function is marked with target(no-vsx) to disable VSX code
generation. To avoid VSX code generation, this function should not be
inlined into VSX function. To fix the bug, in the current logic when
checking whether the caller's ISA flags supports the callee's ISA flags, we
just need to add a test that enforces that the caller's ISA flags match
exactly the callee's flags, for those flags that were explicitly set in the
callee. If caller without target attribute then using options from command
line.
Jiufu
BR
gcc/
2019-10-16 Peter Bergner <bergner@linux.ibm.com>
Jiufu Guo <guojiufu@linux.ibm.com>
PR target/70010
* config/rs6000/rs6000.c (rs6000_can_inline_p): Prohibit inlining if
the callee explicitly disables some isa_flags the caller is using.
gcc.testsuite/
2019-10-16 Peter Bergner <bergner@linux.ibm.com>
Jiufu Guo <guojiufu@linux.ibm.com>
PR target/70010
* gcc.target/powerpc/pr70010.c: New test.
* gcc.target/powerpc/pr70010-1.c: New test.
* gcc.target/powerpc/pr70010-2.c: New test.
* gcc.target/powerpc/pr70010-3.c: New test.
* gcc.target/powerpc/pr70010-4.c: New test.
Tobias Burnus [Mon, 28 Oct 2019 09:36:27 +0000 (09:36 +0000)]
Fortran] PR91863 - fix call to bind(C) with array descriptor
Backport from mainline
2019-10-28 Tobias Burnus <tobias@codesourcery.com>
PR fortran/91863
* trans-expr.c (gfc_conv_gfc_desc_to_cfi_desc): Don't free data
memory as that's done on the Fortran side.
(gfc_conv_procedure_call): Handle void* pointers from
gfc_conv_gfc_desc_to_cfi_desc.
Backport from mainline
2019-10-28 Tobias Burnus <tobias@codesourcery.com>
Iain Sandoe [Sun, 27 Oct 2019 21:25:57 +0000 (21:25 +0000)]
[Darwin, PPC] Fix PR 65342.
The current Darwin load/store lo_sum patterns have neither predicate nor
constraint. This means that most parts of the backend, which rely on
recog() to validate the rtx, can produce invalid combinations/selections.
For 32bit cases this isn't a problem since we can load/store to unaligned
addresses using D-mode insns.
Conversely, for 64bit instructions that use DS mode, this can manifest as
assemble errors (for an assembler that checks the LO14 relocations), or as
crashes caused by wrong offsets (or worse, wrong content for the two LSBs).
What we want to check for Y on Darwin is:
- that the alignment of the Symbols' target is sufficient for DS mode
- that the offset is suitable for DS mode.
(while looking through the Mach-O PIC unspecs).
So, the patch removes the Darwin-specific lo_sum patterns (we begin using
the movdi_internal64 patterns). We also we need to extend the handling of the
mem_operand_gpr constraint to allow looking through Mach-O PIC UNSPECs in
the lo_sum cases.
2019-10-27 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-10-17 Iain Sandoe <iain@sandoe.co.uk>
Iain Sandoe [Sun, 27 Oct 2019 21:21:49 +0000 (21:21 +0000)]
[Darwin, machopic] Fix for 67183
When we're using the LLVM-based assembler (the default on modern Darwin)
the ordering of stubs and non-lazy symbol pointers is important.
Interleaving the output (current GCC behaviour) leads to crashes which
prevents us from building code with symbol stubs.
To resolve this, we order the output of stubs and symbol indirections:
1. Any indirections in the data section
2. Symbol stubs.
3. Non-lazy symbol pointers.
At present, we still emit LTO sections after these.
2019-10-27 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-10-12 Iain Sandoe <iain@sandoe.co.uk>
PR target/67183
* config/darwin.c (machopic_indirection): New field to flag
non-lazy-symbol-pointers in the data section.
(machopic_indirection_name): Compute if an indirection should
appear in the data section.
(machopic_output_data_section_indirection): New callback split
from machopic_output_indirection.
(machopic_output_stub_indirection): Likewise.
(machopic_output_indirection): Retain the code for non-lazy
symbol pointers in their regular section.
(machopic_finish): Use the new callbacks to order the indirection
output.
Iain Sandoe [Sun, 27 Oct 2019 21:17:33 +0000 (21:17 +0000)]
[Darwin, machopic] Preparatory patches.
This is a combined set of 5 patches that allows us to fix
PR 65342 and 67183.
276924 Make machopic_finish() static.
276767 Set a SYMBOL flag for indirections.
We are able to treat these specially where needed in legitimate address
tests (specifically, they are guaranteed to be pointer-aligned).
276708 Compute and cache indirection rules.
This caches a check for the requirement to indirect a symbol in the Darwin
ABI, and uses it where needed. We also ensure that we place the indirection
pointers into the non-lazy symbol pointers section. Other placements have
occurred with various platform toolchains - but these seem to have been
unintentional so we match current platform toolchains.
276675 Consider visibility in indirections.
For weak, hidden vars the indirection should just be as normal, that
is that the indirections for such symbols should appear in the non-lazy
symbol pointers table, not in the .data section.
276674 Initial tidy of Mach-O symbol handling.
We want to improve the detection and caching of symbol-properties
so that (a) we can make the compiler's output match the platform
norms (b) we can improve efficiency by checking flags instead of
inspecting strings. (c) The fix for PR71767 was a largish hammer
and we want to reduce the number of symbols that are made linker-
visible.
This first patch is largely typographical changes with no functional
difference intended:
- Tries to ensure that there's no overlap between the symbols used in
the Mach-O case and those declared in the i386 or rs6000 port trees.
- Some improvement to comments.
- Makes the naming of the symbol flags consistent with other uses.
- Provides a predicate macro for each use.
2019-10-27 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-10-12 Iain Sandoe <iain@sandoe.co.uk>
* config/darwin-protos.h (machopic_finish): Delete.
* config/darwin.c (machopic_finish): Make static.
Backport from mainline
2019-10-09 Iain Sandoe <iain@sandoe.co.uk>
* config/darwin.c (machopic_indirect_data_reference): Set flag to
indicate that the new symbol is an indirection.
(machopic_indirect_call_target): Likewise.
* config/darwin.h (MACHO_SYMBOL_FLAG_INDIRECTION): New.
(MACHO_SYMBOL_INDIRECTION_P): New.
(MACHO_SYMBOL_FLAG_STATIC): Adjust bit number.
Backport from mainline
2019-10-08 Iain Sandoe <iain@sandoe.co.uk>
* config/darwin.c (machopic_indirect_data_reference): Check for
required indirections before making direct access to defined
values.
(machopic_output_indirection): Place the indirected pointes for
required indirections into the non-lazy symbol pointers section.
(darwin_encode_section_info):
* config/darwin.h (MACHO_SYMBOL_FLAG_MUST_INDIRECT): New.
(MACHO_SYMBOL_MUST_INDIRECT_P): New.
Backport from mainline
2019-10-07 Iain Sandoe <iain@sandoe.co.uk>
* config/darwin.c (machopic_output_indirection): Don't put
hidden symbol indirections into the .data section, use the
non-lazy symbol pointers section as normal.
(darwin_encode_section_info): Record if a symbol is hidden.
* config/darwin.h (MACHO_SYMBOL_FLAG_HIDDEN_VIS): New.
(MACHO_SYMBOL_HIDDEN_VIS_P): New.
Backport from mainline
2019-10-07 Iain Sandoe <iain@sandoe.co.uk>
* config/darwin.c (machopic_symbol_defined_p): Use symbol flag
predicates instead of accessing bits directly.
(machopic_indirect_call_target): Likewise.
(machopic_output_indirection): Likewise.
(darwin_encode_section_info): Improve description. Use renamed
symbol flags. Use predicate macros for variables and functions.
* config/darwin.h:
Rename MACHO_SYMBOL_VARIABLE to MACHO_SYMBOL_FLAG_VARIABLE.
Rename MACHO_SYMBOL_DEFINED to MACHO_SYMBOL_FLAG_DEFINED.
Rename MACHO_SYMBOL_STATIC to MACHO_SYMBOL_FLAG_STATIC.
(MACHO_SYMBOL_VARIABLE_P): New.
(MACHO_SYMBOL_DEFINED_P):New.
(MACHO_SYMBOL_STATIC_P): New.
* config/i386/darwin.h (MACHO_SYMBOL_FLAG_VARIABLE): Delete.
(SYMBOL_FLAG_SUBT_DEP): New.
* config/rs6000/darwin.h (SYMBOL_FLAG_SUBT_DEP): New.
Paul Thomas [Sun, 27 Oct 2019 18:28:52 +0000 (18:28 +0000)]
backport: re PR fortran/86248 (LEN_TRIM in specification expression causes link failure)
2019-10-27 Paul Thomas <pault@gcc.gnu.org>
Backport from mainline
PR fortran/86248
* resolve.c (flag_fn_result_spec): Correct a typo before the
function declaration.
* trans-decl.c (gfc_sym_identifier): Boost the length of 'name'
to allow for all variants. Simplify the code by using a pointer
to the symbol's proc_name and taking the return out of each of
the conditional branches. Allow symbols with fn_result_spec set
that do not come from a procedure namespace and have a module
name to go through the non-fn_result_spec branch.
2019-10-27 Paul Thomas <pault@gcc.gnu.org>
Backport from mainline
PR fortran/86248
* gfortran.dg/char_result_19.f90 : New test.
* gfortran.dg/char_result_mod_19.f90 : Module for the new test.
Jan Hubicka [Sat, 26 Oct 2019 16:57:16 +0000 (18:57 +0200)]
Avoid gimple body modifications while streaming.
Backport from mainline
2019-10-11 Jan Hubicka <hubicka@ucw.cz>
* gimple-streamer-out.c (output_gimple_stmt): Add explicit function
parameter.
* lto-streamer-out.c: Include tree-dfa.h.
(output_cfg): Do not use cfun.
(lto_prepare_function_for_streaming): New.
(output_function): Do not push cfun; do not initialize loop optimizer.
* lto-streamer.h (lto_prepare_function_for_streaming): Declare.
* passes.c (ipa_write_summaries): Use it.
(ipa_write_optimization_summaries): Do not modify bodies.
* tree-dfa.c (renumber_gimple_stmt_uids): Add function parameter.
* tree.dfa.h (renumber_gimple_stmt_uids): Update prototype.
* tree-ssa-dse.c (pass_dse::execute): Update use of
renumber_gimple_stmt_uids.
* tree-ssa-math-opts.c (pass_optimize_widening_mul::execute): Likewise.
2019-10-12 Jan Hubicka <hubicka@ucw.cz>
* lto-streamer-out.c (collect_block_tree_leafs): Renumber statements
so non-virutal are before virutals.
(output_function): Avoid body modifications.
This patch fixes a problem with the thumb1 prologue code where the link
register could be unconditionally used as a scratch register even if the
return value was still live at the end of the prologue.
Additionally, the patch improves the code generated when we are not
using many low call-saved registers to make use of any unused call
clobbered registers to help with the saving of high registers that
cannot be pushed directly (quite rare in normal code as the register
allocator correctly prefers low registers).
2019-05-08 Mihail Ionescu <mihail.ionescu@arm.com>
Richard Earnshaw <rearnsha@arm.com>
gcc:
PR target/88167
* config/arm/arm.c (thumb1_prologue_unused_call_clobbered_lo_regs): New
function.
(thumb1_epilogue_unused_call_clobbered_lo_regs): New function.
(thumb1_compute_save_core_reg_mask): Don't force a spare work
register if both the epilogue and prologue can use call-clobbered
regs.
(thumb1_unexpanded_epilogue): Use
thumb1_epilogue_unused_call_clobbered_lo_regs. Reverse the logic for
picking temporaries for restoring high regs to match that of the
prologue where possible.
(thumb1_expand_prologue): Add any usable call-clobbered low registers to
the list of work registers. Detect if the return address is still live
at the end of the prologue and avoid using it for a work register if so.
If the return address is not live, add LR to the list of pushable regs
after the first pass.
gcc/testsuite:
PR target/88167
* gcc.target/arm/pr88167-1.c: New test.
* gcc.target/arm/pr88167-2.c: New test.
Jan Hubicka [Fri, 25 Oct 2019 11:17:38 +0000 (13:17 +0200)]
Backport ggc_trim
Backport from mainline
2019-10-18 Jakub Jelinek <jakub@redhat.com>
PR middle-end/92153
* ggc-page.c (release_pages): Read g->alloc_size before free rather
than after it.
2019-10-11 Jan Hubicka <hubicka@ucw.cz>
* ggc-page.c (release_pages): Output statistics when !quiet_flag.
(ggc_collect): Dump later to not interfere with release_page dump.
(ggc_trim): New function.
* ggc-none.c (ggc_trim): New.
* ggc.h (ggc_trim): Declare.
Iain Sandoe [Thu, 24 Oct 2019 14:53:28 +0000 (14:53 +0000)]
[Darwin, PPC] Check for out of range asm values.
There are some cases in which the value for the max skip to a p2align
directive can be negative. The older assembler (and GAS) just ignores
these cases but newer tools produce an error. To preserve behaviour,
we avoid emitting out of range values.
gcc/ChangeLog:
2019-10-24 Iain Sandoe <iain@sandoe.co.uk>
Backport from mainline
2019-10-23 Iain Sandoe <iain@sandoe.co.uk>
* config/rs6000/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Guard
against out of range max skip or log values.
Jonathan Wakely [Thu, 24 Oct 2019 12:55:27 +0000 (13:55 +0100)]
PR libstdc++/90682 allow set_terminate(0) and set_unexpected(0)
Make these functions restore the default handlers when passed a null
pointer. This is consistent with std::pmr::set_default_resource(0), and
also matches the current behaviour of libc++.
In order to avoid duplicating the preprocessor condition from
eh_term_handler.cc more that into a new eh_term_handler.h header and
define a macro that can be used in both eh_term_handler.cc and
eh_terminate.cc.
Backport from mainline
2019-05-31 Jonathan Wakely <jwakely@redhat.com>
PR libstdc++/90682
* libsupc++/eh_term_handler.cc: Include eh_term_handler.h to get
definition of _GLIBCXX_DEFAULT_TERM_HANDLER.
* libsupc++/eh_term_handler.h: New header defining
_GLIBCXX_DEFAULT_TERM_HANDLER.
* libsupc++/eh_terminate.cc: Include eh_term_handler.h.
(set_terminate): Restore default handler when argument is null.
(set_unexpected): Likewise.
* testsuite/18_support/set_terminate.cc: New test.
* testsuite/18_support/set_unexpected.cc: New test.
Jonathan Wakely [Thu, 24 Oct 2019 12:55:21 +0000 (13:55 +0100)]
PR libstdc++/61761 fix std::proj for targets without C99 cproj
The current generic implementation of __complex_proj used when cproj is
not available calculates the wrong projection, giving a different result
than given by C99's cproj.
When C99 cproj is not available but isinf and copysign are, use those to
give correct results for float, double and long double. Otherwise, and
for other specializations of std::complex, just use a generic version
that returns its argument, and so doesn't support infinities.
We might want to consider adding additional overloads of __complex_proj
to support extended types such as _Float64x, _Float128 etc.
Backport from mainline
2019-05-01 Jonathan Wakely <jwakely@redhat.com>
PR libstdc++/61761
* include/std/complex (__complex_proj): Return parameter unchanged.
[_GLIBCXX_USE_C99_COMPLEX] (__complex_proj): Change overloads for
floating-point types to take std::complex arguments.
[_GLIBCXX_USE_C99_MATH_TR1] (__complex_proj): Add overloads for
floating-point types.
* testsuite/26_numerics/complex/proj.cc: New test.
projects / thirdparty / gcc.git / log
