2018-11-29 Michael Ploujnikov <michael.ploujnikov@oracle.com>
There can be at most one .resolver clone per function
* config/rs6000/rs6000.c (make_resolver_func): Generate
resolver symbol with clone_function_name instead of
clone_function_name_numbered.
From-SVN: r266617
2018-11-29 Richard Biener <rguenther@suse.de>
PR tree-optimization/88243
* tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def
type of all pattern-sequence stmts to vect_internal_def.
* gcc.dg/torture/pr88243.c: New testcase.
From-SVN: r266615
Now that __alignof__ and alignof sometimes disagree it matters which one
we use. The standard says that std::alignment_of<T>::value equals
alignof(T), so we need to use that.
Change the only uses of alignment_of to use __alignof__ to avoid a
change in alignment.
PR libstdc++/88119
* include/ext/aligned_buffer.h (__aligned_membuf): Add comment.
(__aligned_buffer): Use __alignof__ instead of std::alignment_of.
* include/std/type_traits (alignment_of): Use alignof instead of
__alignof__.
* testsuite/20_util/alignment_of/value.cc: Fix test to check values
match alignof not __alignof__, as required by the standard.
From-SVN: r266613
gcc/ChangeLog:
2018-11-29 Andre Vieira <andre.simoesdiasvieira@arm.com>
PR target/88224
* config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
(cortex-r7, cortex-r8): Update default and add new configuration.
* doc/invoke.texi (armv7-r): Add two new vfp options.
(nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
support this option.
From-SVN: r266612
This split is disabled for power7 and up, so we don't often see its
bad effects. However, on a powerpc-linux compiler (which defaults
to PPC750 judging from rs6000/sysv4.h) we see
long double ld1 (void) { return 1.0L; }
compiled with -msoft-float -O2 -S resulting in
ld1:
li 3,0
li 4,0
mr 6,4
mr 5,3
li 4,0
lis 3,0x3ff0
blr
Things go awry in init-regs, with the TFmode reg being initialized to
zero on seeing the subreg from the split. (And that initialization
itself is split by rs6000_emit_move!) Later passes apparently don't
clean up the rubbish.
Since the split was added for Darwin (as the comment says), let's get
rid of it on other targets.
* config/rs6000/rs6000.c (rs6000_emit_move): Disable long
double split for targets other than Darwin.
From-SVN: r266610
Once upon a time CONST_DOUBLE was used to hold large integer constants
that didn't fit into a CONST_INT. Now that CONST_INT is always at
least 64 bits, and the rs6000 backend uses CONST_WIDE_INT for larger
integers, there is no need for old code dealing with integers in
CONST_DOUBLE. The rs6000 backend also doesn't create DImode subregs
of TFmode CONST_DOUBLE, as once we did. This patch cleans up a few
leftovers from the past.
* config/rs6000/predicates.md (easy_fp_constant): Remove code
dealing with integers in const_double. Assert on unexpected
modes. Delete superfluous ABI_V4 test.
* config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes.
Don't call easy_fp_constant for Pmode.
(rs6000_hash_constant): Remove code dealing with integers in
const_double.
(rs6000_legitimate_constant_p): Likewise.
(output_toc): Formatting, use CONST_DOUBLE_P.
From-SVN: r266608
The patch fixes two cases where the "G" and "H" constraints were used
incorrectly. Their purpose is calculating insn lengths. Thus it
never makes sense to put "GH" together or with "F" in an insn
alternative.
movdi_internal32 used "GHF" in an alternative so I replaced that with
"F", and added length attributes for the insn, which were missing.
There are some formatting changes too, to make alternatives line up
with attributes.
The FMOVE128 version of mov<mode>_softfloat also had "GHF" in an
alternative, so "GH" is dropped and "F" moved to a separate
alternative in order to get insn lengths correct. Well by correct I
mean we need to choose the maximum insn length otherwise branches
might not reach. Note the very large length for "F". I think it
would be better to force soft-float long double constants to memory in
64-bit mode, but that's a patch for another day.
* config/rs6000/constraints.md (G, H): Comment on purpose of
constraint. Correct mode comments and "G" mode test.
* config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from
alternative handling "F". Add length attr. Formatting.
(mov<mode>_softfloat <FMOVE128>): Delete "GH" from
alternative, and move "F" to separate alternative. Correct
insn lengths.
From-SVN: r266607
easy_fp_constant accepts const_double, easy_vector_constant accepts
const_vector. Calling one from the other therefore will always return
false. I believe the zero_constant call in easy_vector_constant will
cover the 0.0 case of easy_fp_constant.
* config/rs6000/predicates.md (easy_vector_constant): Don't call
easy_fp_constant.
From-SVN: r266606
Finally, the point of the previous patches in this series, support for
inline PLT calls, keyed off -fno-plt. This emits code using new
relocations that tie all insns in the sequence together, so that the
linker can edit the sequence back to a direct call should the call
target turn out to be local. An example of ELFv2 code to call puts is
as follows:
.reloc .,R_PPC64_PLTSEQ,puts
std 2,24(1)
.reloc .,R_PPC64_PLT16_HA,puts
addis 12,2,0
.reloc .,R_PPC64_PLT16_LO_DS,puts
ld 12,0(12)
.reloc .,R_PPC64_PLTSEQ,puts
mtctr 12
.reloc .,R_PPC64_PLTCALL,puts
bctrl
ld 2,24(1)
"addis 12,2,puts@plt@ha" and "ld 12,puts@plt@l(12)" are also supported
by the assembler. gcc instead uses the explicit R_PPC64_PLT16_HA and
R_PPC64_PLT16_LO_DS relocs because when the call is to __tls_get_addr
an extra reloc is emitted at every place where one is shown above, to
specify the __tls_get_addr arg. The linker expects the extra reloc to
come first. .reloc enforces that ordering.
The patch also changes code emitted for longcalls if the assembler
supports the new marker relocs, so that these too can be edited. One
side effect of longcalls using PLT16 relocs is that they can now be
resolved lazily by ld.so.
I don't support lazy inline PLT calls for ELFv1, because ELFv1 would
need barriers to reliably load both the function address and toc
pointer from the PLT. ELFv1 -fno-plt uses the longcall sequence
instead, which isn't edited by GNU ld.
* config.in (HAVE_AS_PLTSEQ): Add.
* config/rs6000/predicates.md (indirect_call_operand): New.
* config/rs6000/rs6000-protos.h (rs6000_pltseq_template),
(rs6000_sibcall_sysv): Declare.
* config/rs6000/rs6000.c (init_cumulative_args): Set cookie
CALL_LONG for -fno-plt.
(print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ.
(rs6000_indirect_call_template_1): Emit .reloc directives for
UNSPEC_PLTSEQ calls.
(rs6000_pltseq_template): New function.
(rs6000_longcall_ref): Add arg parameter. Use PLT16 insns if
relocs supported by assembler. Move SYMBOL_REF test to callers.
(rs6000_call_aix): Adjust rs6000_longcall_ref call. Package
insns in UNSPEC_PLTSEQ, preserving original func_desc.
(rs6000_call_sysv): Likewise.
(rs6000_sibcall_sysv): New function.
* config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default.
* config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA,
UNSPEC_PLT16_LO): New.
(pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New.
(call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero
cookie in constraints. Test explicitly for flags in length attr.
Handle unspec operand 1.
(call_value_indirect_nonlocal_sysv): Likewise.
(call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1.
(call_indirect_elfv2, call_value_indirect_elfv2): Likewise.
(sibcall, sibcall_value): Use rs6000_sibcall_sysv.
(sibcall_indirect_nonlocal_sysv): New pattern.
(sibcall_value_indirect_nonlocal_sysv): Likewise.
(sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect
call alternatives.
* configure.ac: Check for gas plt sequence marker support.
* configure: Regenerate.
From-SVN: r266605
The current code handling __tls_get_addr calls for powerpc*-linux
generates a call then overwrites the call insn with a special
tls_{gd,ld}_{aix,sysv} pattern. It's done that way to support
!TARGET_TLS_MARKERS, where the arg setup insns need to be emitted
immediately before the branch and link. When TARGET_TLS_MARKERS, the
arg setup insns are split from the actual call, but we then have a
non-standard call pattern that needs to be carried through to output.
This patch changes that scheme, to instead use the standard call
patterns for __tls_get_addr calls, except for the now rare
!TARGET_TLS_MARKERS case. Doing it this way should be better for
maintenance as the !TARGET_TLS_MARKERS code can eventually disappear.
It also makes it possible to support longcalls (and in following
patches, inline plt calls) for __tls_get_addr without introducing yet
more special call patterns.
__tls_get_addr calls do however need to be different to standard
calls, because when TARGET_TLS_MARKERS the calls are decorated with an
argument specifier, eg. "bl __tls_get_addr(thread_var@tlsgd)" that
causes a reloc to be emitted by the assembler tying the call to its
arg setup insns. I chose to smuggle the arg in the currently unused
stack size rtl.
I've also introduced rs6000_call_sysv to generate rtl for sysv calls,
as rs6000_call_aix does for aix and elfv2 calls. This allows
rs6000_longcall_ref to be local to rs6000.c since the calls in the
expanders never did anything for darwin.
* config/rs6000/predicates.md (unspec_tls): New.
* config/rs6000/rs6000-protos.h (rs6000_call_template),
(rs6000_sibcall_template): Update prototype.
(rs6000_longcall_ref): Delete.
(rs6000_call_sysv): Declare.
* config/rs6000/rs6000.c (edit_tls_call_insn): New function.
(global_tlsarg): New variable.
(rs6000_legitimize_tls_address): Rewrite __tls_get_addr call
handling.
(print_operand): Extract UNSPEC_TLSGD address operand.
(rs6000_call_template, rs6000_sibcall_template): Remove arg
parameter, extract from second call operand instead.
(rs6000_longcall_ref): Make static, localize vars.
(rs6000_call_aix): Rename parameter to reflect new usage. Take
tlsarg from global_tlsarg. Don't create unused rtl or nop insns.
(rs6000_sibcall_aix): Rename parameter to reflect new usage. Take
tlsarg from global_tlsarg.
(rs6000_call_sysv): New function.
* config/rs6000/rs6000.md: Adjust rs6000_call_template and
rs6000_sibcall_template throughout.
(tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete.
(tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete.
(tls_gdld_nomark): New insn.
(tls_gd): Swap operand order. Simplify mode selection.
(tls_gd_high, tls_gd_low): Swap operand order.
(tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD.
Simplify mode selection.
(tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD.
(call, call_value): Don't assert for second call operand.
Use rs6000_call_sysv.
From-SVN: r266604
There is really no need to define a TLSmode mode iterator that is
identical (since !TARGET_64BIT == TARGET_32BIT) to the much used P
mode iterator. It's nonsense to think we might ever want to support
32-bit TLS on 64-bit or vice versa! The patch also fixes a minor
error in the call mems. All other direct calls use (call (mem:SI ..)).
* config/rs6000/rs6000.md (TLSmode): Delete mode iterator. Replace
with P throughout except for call mems which should use SI.
(tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode
attributes. Replace with bits, mode and ptrload respectively.
From-SVN: r266602
Like the last patch for external calls, now handle most assembly code
for indirect calls in one place. The patch also merges some insns,
correcting some !rs6000_speculate_indirect_jumps cases branching to
LR, which don't require a speculation barrier.
* config/rs6000/rs6000-protos.h (rs6000_indirect_call_template),
(rs6000_indirect_sibcall_template): Declare.
* config/rs6000/rs6000.c (rs6000_indirect_call_template_1),
(rs6000_indirect_call_template, rs6000_indirect_sibcall_template):
New functions.
* config/rs6000/rs6000.md (call_indirect_nonlocal_sysv),
(call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv),
(call_indirect_aix, call_value_indirect_aix): Use
rs6000_indirect_call_template and rs6000_indirect_sibcall_template.
call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and
handle both speculation and non-speculation cases.
(call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete.
(call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete.
From-SVN: r266601
This is a first step in tidying rs6000 call patterns, in preparation
to support inline plt calls.
* config/rs6000/rs6000-protos.h (rs6000_call_template): Declare.
(rs6000_sibcall_template): Declare.
(macho_call_template): Rename from output_call.
* config/rs6000/rs6000.c (rs6000_call_template_1): New function.
(rs6000_call_template, rs6000_sibcall_template): Likewise.
(macho_call_template): Rename from output_call.
* config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv),
(tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv),
(tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv),
(call_nonlocal_sysv_secure, call_value_nonlocal_sysv),
(call_value_nonlocal_sysv_secure, call_nonlocal_aix),
(call_value_nonlocal_aix): Use rs6000_call_template and update
occurrences of output_call to macho_call_template.
(sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix),
(sibcall_value_aix): Use rs6000_sibcall_template.
From-SVN: r266600
2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
* config/rs6000/rs6000-string.c (expand_block_clear): Change how
we determine if unaligned vsx is ok.
From-SVN: r266599
Implement the proposed semantics from P1164R0, which reverts the changes
of LWG 2935. This means that failure to create a directory because a
non-directory already exists with that name will be reported as an
error.
While rewriting the function, also fix PR 87846, which is a result of
the C++17 changes to how a trailing slash on a path affects the last
component of a path.
PR libstdc++/86910
PR libstdc++/87846
* src/filesystem/ops.cc (experimental::create_directories): Report
an error when the path resolves to an existing non-directory (P1164).
* src/filesystem/std-ops.cc (create_directories): Likewise. Handle
empty filenames due to trailing slashes.
* testsuite/27_io/filesystem/operations/create_directories.cc: Test
when some component of the path exists and is not a directory. Test
trailing slashes.
* testsuite/experimental/filesystem/operations/create_directories.cc:
Likewise.
From-SVN: r266598
PR c/88065 - ICE in -Wsizeof-pointer-memaccess on an invalid strncpy
PR c/87297 - ICE on strncpy with an undeclared argument
gcc/c-family/ChangeLog:
PR c/88065
PR c/87297
* c-warn.c (sizeof_pointer_memaccess_warning): Bail if source
or destination is an error.
gcc/testsuite/ChangeLog:
PR c/88065
PR c/87297
* c-c++-common/Wsizeof-pointer-memaccess4.c: New test.
From-SVN: r266594
This sets a precedence in upstream for merging architecture agnostic
C bindings into one block, rather than separate duplicated blocks.
A nice side-effect is it almost completes the C bindings for
s390-linux-gnu and sparc-linux-gnu, and fixes a bug on MIPS32 where
O_SYNC had the wrong value.
Backported from upstream druntime master.
Reviewed-on: https://github.com/dlang/druntime/pull/2357
From-SVN: r266593
* predict.c (determine_unlikely_bbs): Forward declare; also determine
edges that are always known to be taken because it is only likely
edge out of given BB.
(tree_estimate_probability): Call before profile guessing to get bit
of extra precision.
* gcc.dg/predict-13.c: Update template.
* gcc.dg/predict-13b.c: New testcase.
* gcc.dg/predict-22.c: New testcase.
* gcc.dg/tree-ssa/ipa-split-4.c: Change abort to other function to
get sane profile.
From-SVN: r266587
2018-11-28 Vladimir Makarov <vmakarov@redhat.com>
PR target/88207
* ira-costs.c (scan_one_insn): Process subregs when updating costs
for pseudos and allocnos from insn.
From-SVN: r266582
Support inlining functions that contain only assignments and return
statements, with expressions of either constants or parameters.
Functions that contain other kinds of statements or expressions are
not yet inlined. With this change, about 100 functions in the
standard library are inlinable.
Reviewed-on: https://go-review.googlesource.com/c/150073
From-SVN: r266573
* config/h8300/constraints.md: Add "C" constraint for call insns.
* config/h8300/h8300.md (call, call_value): Turn into a define_expand
and define_insn pair. Move invalid call targets into a register in
the expander and fix constraints in the matching pattern.
* config/h8300/predicates.md (call_expander_operand): Renamed from
call_insn_operand. Reject things we shouldn't be trying to handle.
(call_insn_operand): New predicate for use by the call/call_value
insns.
(small_call_insn_operand): Update appropriately.
From-SVN: r266571
2018-11-28 Sam Tebbs <sam.tebbs@arm.com>
* config/aarch64/aarch64.c (aarch64_process_target_attr): Replace
calls to strtok with strtok_r.
From-SVN: r266570
The recent patch for PR 83306 removed the fs_err_concat functions that
were used by the experimental::filesystem::filesystem_error class as
well. This fixes it by doing the string generation directly in
filesystem_error::_M_gen_what() instead of using the removed function.
PR libstdc++/83306
* src/filesystem/path.cc (filesystem_error::_M_gen_what()): Create
string directly, instead of calling fs_err_concat.
From-SVN: r266569
Add fix for "path::lexically_relative is confused by trailing slashes".
* doc/xml/manual/intro.xml: Document LWG 3096 change.
* src/filesystem/std-path.cc (path::lexically_relative(const path&)):
Treat a final empty element equivalently to a final dot element.
* testsuite/27_io/filesystem/path/generation/relative.cc: Add checks
for the examples in the DR.
From-SVN: r266566
The class API provides no way to modify the members, so we can share
them between copies of the same object. Copying becomes a simple
reference count update, which doesn't throw.
Also adjust the what() string to allow distinguishing between an empty
path passed to the constructor, and no path.
PR libstdc++/83306
* include/bits/fs_path.h (filesystem_error): Move data members into
pimpl class owned by shared_ptr. Remove inline definitions of member
functions.
* src/filesystem/std-path.cc (filesystem_error::_Impl): Define.
(filesystem_error): Define member functions.
* testsuite/27_io/filesystem/filesystem_error/cons.cc: New test.
* testsuite/27_io/filesystem/filesystem_error/copy.cc: New test.
From-SVN: r266565
If the allocation of abbrevs->abbrevs in read_abbrevs fails, then
abbrevs->num_abbrevs remains nonzero, and consequently free_abbrevs will
segfault when accessing abbrevs->abbrevs.
Fix this by setting abbrevs->num_abbrevs only after abbrevs->abbrevs
allocation has succeeded.
Bootstrapped and reg-tested on x86_64.
2018-11-28 Tom de Vries <tdevries@suse.de>
* dwarf.c (read_abbrevs): Fix handling of abbrevs->abbrevs allocation
failure.
From-SVN: r266562
2018-11-28 Richard Biener <rguenther@suse.de>
PR tree-optimization/88223
* tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
over a stored-same value may-alias store make sure to consider
partial overlaps which are valid when TBAA reasonings do not
apply and byte-granular overlaps are possible at all.
* gcc.dg/torture/pr88223.c: New testcase.
From-SVN: r266560
2018-11-28 Richard Biener <rguenther@suse.de>
PR tree-optimization/88217
* vr-values.c (vr_values::extract_range_from_phi_node): Make
sure to handle results > +INF and < -INF correctly when
trying to drop down to +INF - 1 or -INF + 1.
* g++.dg/pr88217.C: New testcase.
From-SVN: r266557
git commit 41f70262f (svn rev 264868) exposed a signed/unsigned
comparison. Fixed by matching the type of the local var to that of
the tree field.
* xcoffout.c (do_block): Signed/unsigned warning fix.
From-SVN: r266555
PR testsuite/85368
* gcc.dg/tree-ssa/phi-opt-11.c: For branch_cost targets, expect 0 ifs
rather than 0 or 2 depending on logical_op_short_circuit.
From-SVN: r266552
