rguenth [Mon, 26 Jun 2017 07:19:37 +0000 (07:19 +0000)]
2017-06-26 Richard Biener <rguenther@suse.de>
PR tree-optimization/80928
* cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
(copy_bbs): Set BB_DUPLICATED flag early.
(execute_on_growing_pred): Do not execute for BB_DUPLICATED
marked blocks.
(execute_on_shrinking_pred): Likewise.
* tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
BB_DUPLICATED blocks.
* tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
iterate over all PHIs considering removal of *gsi.
tkoenig [Sat, 24 Jun 2017 07:07:56 +0000 (07:07 +0000)]
2017-06-24 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/52473
* Makefile.am: Add i_cshift1a_c. Add rules to generate files
from cshift1a.m4.
* Makefile.in: Regenerated.
* m4/cshift1a.m4: New file.
* m4/cshift.m4 (cshift1): Split up inner loop by removing
condition. Use memcpy where possible. Call helper functions
based on dtype.
* libgfortran.h: Add prototypes for cshift1_16_c10,
cshift1_16_c16, cshift1_16_c4, cshift1_16_c8, cshift1_16_i1,
cshift1_16_i16, cshift1_16_i2, cshift1_16_i4, cshift1_16_i8,
cshift1_16_r10, cshift1_16_r16, cshift1_16_r4, cshift1_16_r8,
cshift1_4_c10, cshift1_4_c16, cshift1_4_c4, cshift1_4_c8,
cshift1_4_i1, cshift1_4_i16, cshift1_4_i2, cshift1_4_i4,
cshift1_4_i8, cshift1_4_r10, cshift1_4_r16, cshift1_4_r4,
cshift1_4_r8, cshift1_8_c10, cshift1_8_c16, cshift1_8_c4,
cshift1_8_c8, cshift1_8_i1, cshift1_8_i16, cshift1_8_i2,
cshift1_8_i4, cshift1_8_i8, cshift1_8_r10, cshift1_8_r16,
cshift1_8_r4 and cshift1_8_r8.
* generated/cshift1_16_c10.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_c16.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_c4.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_c8.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_i1.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_i16.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_i2.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_i4.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_i8.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_r10.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_r16.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_r4.c: New file, generated from cshift1a.m4.
* generated/cshift1_16_r8.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_c10.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_c16.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_c4.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_c8.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_i1.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_i16.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_i2.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_i4.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_i8.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_r10.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_r16.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_r4.c: New file, generated from cshift1a.m4.
* generated/cshift1_4_r8.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_c10.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_c16.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_c4.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_c8.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_i1.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_i16.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_i2.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_i4.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_i8.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_r10.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_r16.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_r4.c: New file, generated from cshift1a.m4.
* generated/cshift1_8_r8.c: New file, generated from cshift1a.m4.
2017-06-24 Thomas Koenig <tkoenig@gcc.gnu.org>
PR fortran/52473
* gfortran.dg/cshift_2.f90: New test.
jason [Fri, 23 Jun 2017 23:29:51 +0000 (23:29 +0000)]
PR c++/79056 - C++17 ICE with invalid template syntax.
* parser.c (cp_parser_simple_type_specifier): Don't assume that type
is a TYPE_DECL.
(cp_parser_check_for_invalid_template_id): Handle TYPE_DECL.
* pt.c (template_placeholder_p): New.
* cp-tree.h: Declare it.
ian [Fri, 23 Jun 2017 20:19:40 +0000 (20:19 +0000)]
runtime: complete defer handling in CgocallBackDone
When C code calls a Go function, it actually calls a function
generated by cgo. That function is written in Go, and, among other
things, it calls the real Go function like this:
CgocallBack()
defer CgocallBackDone()
RealGoFunction()
The deferred CgocallBackDone function enters syscall mode as we return
to C. Typically the C function will then eventually return to Go.
However, in the case where the C function is running on a thread
created in C, it will not return to Go. For that case we will have
allocated an m struct, with an associated g struct, for the duration
of the Go code, and when the Go is complete we will return the m and g
to a free list.
That all works, but we are running in a deferred function, which means
that we have been invoked by deferreturn, and deferreturn expects to
do a bit of cleanup to record that the defer has been completed. Doing
that cleanup while using an m and g that have already been returned to
the free list is clearly a bad idea. It was kind of working because
deferreturn was holding the g pointer in a local variable, but there
were races with some other thread picking up and using the newly freed g.
It was also kind of working because of a special check in freedefer;
that check is no longer necessary.
This patch changes the special case of releasing the m and g to do the
defer cleanup in CgocallBackDone itself.
This patch also checks for the special case of a panic through
CgocallBackDone. In that special case, we don't want to release the m
and g. Since we are returning to C code that was not called by Go
code, we know that the panic is not going to be caught and we are
going to exit the program. So for that special case we keep the m and
g structs so that the rest of the panic code can use them.
* config/rs6000/rs6000-string.c: (expand_block_clear,
do_load_for_compare, select_block_compare_mode,
compute_current_alignment, expand_block_compare,
expand_strncmp_align_check, expand_strn_compare,
expand_block_move, rs6000_output_load_multiple)
Move functions related to string/block move/compare
to a separate file.
* config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
* config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
for this function which is now used in two files.
* config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
* config.gcc: Add rs6000-string.o to extra_objs for
targets powerpc*-*-* and rs6000*-*-*.
meissner [Fri, 23 Jun 2017 18:25:10 +0000 (18:25 +0000)]
[gcc]
2017-06-23 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/80510
* config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
32-bit, since indexed is not valid for DImode.
(mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
(define_peephole2 for Altivec d-form load): Add 32-bit support.
(define_peephole2 for Altivec d-form store): Likewise.
[gcc/testsuite]
2017-06-23 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/80510
* gcc.target/powerpc/pr80510-1.c: Allow test to run on 32-bit.
* gcc.target/powerpc/pr80510-2.c: Likewise.
meissner [Fri, 23 Jun 2017 18:08:31 +0000 (18:08 +0000)]
2017-06-23 Michael Meissner <meissner@linux.vnet.ibm.com>
PR ipa/81185
* multiple_target.c (create_dispatcher_calls): Only create the
dispatcher call if the function is the default clone of a
versioned function.
segher [Fri, 23 Jun 2017 17:37:27 +0000 (17:37 +0000)]
Fix expand_builtin_atomic_fetch_op for pre-op (PR80902)
__atomic_add_fetch adds a value to some memory, and returns the result.
If there is no direct support for this, expand_builtin_atomic_fetch_op
is asked to implement this as __atomic_fetch_add (which returns the
original value of the mem), followed by the addition. Now, the
__atomic_add_fetch could have been a tail call, but we shouldn't
perform the __atomic_fetch_add as a tail call: following code would
not be executed, and in fact thrown away because there is a barrier
after tail calls.
This fixes it.
PR middle-end/80902
* builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
a call, force the call to not be a tail call.
ian [Fri, 23 Jun 2017 17:10:18 +0000 (17:10 +0000)]
os: align siginfo argument to waitid
Backport https://golang.org/cl/46511 from gc trunk, as it may fix a
bug reported for gccgo running on MIPS
(https://groups.google.com/d/msg/golang-dev/sDg-t1_DPw0/-AJmLxgPBQAJ).
ian [Fri, 23 Jun 2017 16:05:44 +0000 (16:05 +0000)]
runtime: don't crash if no p in kickoff
The kickoff function for g0 can be invoked without a p, for example
from mcall(exitsyscall0) in exitsyscall after exitsyscall has cleared
the p field. The assignment gp.param = nil will invoke a write barrier.
If gp.param is not already nil, this will require a p. Avoid the problem
for a specific case that is known to be OK: when the value in gp.param
is a *g.
ian [Fri, 23 Jun 2017 16:03:49 +0000 (16:03 +0000)]
compiler: add go:notinheap magic comment
Implement go:notinheap as the gc compiler does. A type marked as
go:notinheap may not live in the heap, and does not require a write
barrier. Struct and array types that incorporate notinheap types are
themselves notinheap. Allocating a value of a notinheap type on the
heap is an error.
This is not just an optimization. There is code where a write barrier
may not occur that was getting a write barrier with gccgo but not gc,
because the types in question were notinheap. The case I found was
setting the mcache field in exitsyscallfast.
willschm [Fri, 23 Jun 2017 14:40:26 +0000 (14:40 +0000)]
[gcc]
2017-06-23 Will Schmidt <will_schmidt@vnet.ibm.com>
* config/rs6000/rs6000.c: Add include of ssa-propagate.h for
update_call_from_tree(). (rs6000_gimple_fold_builtin): Add
handling for early expansion of vector shifts (sl,sr,sra,rl).
(builtin_function_type): Add vector shift right instructions
to the unsigned argument list.
[gcc/testsuite]
2017-06-23 Will Schmidt <will_schmidt@vnet.ibm.com>
ian [Fri, 23 Jun 2017 13:45:36 +0000 (13:45 +0000)]
runtime: improve handling of panic during deferred function
When a panic occurs while processing a deferred function that
recovered an earlier panic, we shouldn't report the recovered panic
in the panic stack trace. Stop doing so by keeping track of the panic
that triggered a defer, marking it as aborted if we see the defer again,
and discarding aborted panics when a panic is recovered. This is what
the gc runtime does.
The test for this is TestRecursivePanic in runtime/crash_test.go.
We don't run that test yet, but we will soon.
rtl-optimizatoin/79286
* ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
* rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
trap. PIC register plus a const unspec without offset can never trap.
glisse [Fri, 23 Jun 2017 09:05:47 +0000 (09:05 +0000)]
Refactor fileptr_type_node handling
2017-06-23 Marc Glisse <marc.glisse@inria.fr>
gcc/
* tree.h (builtin_structptr_type): New type.
(builtin_structptr_types): Declare new array.
* tree.c (builtin_structptr_types): New array.
(free_lang_data, build_common_tree_nodes): Use it.
gcc/c-family/
* c-common.c (c_common_nodes_and_builtins): Use builtin_structptr_types.
gcc/cp/
* decl.c (duplicate_decls): Use builtin_structptr_types.
gcc/lto/
* lto-lang.c (lto_init): Use builtin_structptr_types.
ian [Thu, 22 Jun 2017 22:08:50 +0000 (22:08 +0000)]
runtime: export getm function
Use go:linkname to export the getm function. This makes it visible to
runtime/testdata/testprogcgo/dropm_stub.go, which uses it as part of
the TestEnsureDropM test in runtime/crash_cgo_test.go. That test is
not run today, but it will be soon.
ian [Thu, 22 Jun 2017 15:46:47 +0000 (15:46 +0000)]
runtime: make NumGoroutine wait for system goroutines to register
In libgo system goroutines register themselves after they start.
That means that there is a small race between the goroutine being
seen by the scheduler and the scheduler knowing that the goroutine
is a system goroutine. That in turn means that runtime.NumGoroutines
can overestimate the number of goroutines at times.
This patch fixes the overestimate by counting the number of system
goroutines waiting to start, and pausing NumGoroutines until those
goroutines have all registered.
This is kind of a lot of mechanism for this not very important
problem, but I couldn't think of a better approach.
The test for this is TestNumGoroutine in runtime/proc_test.go.
The test is not currently run, but it will be soon.
ian [Thu, 22 Jun 2017 14:46:12 +0000 (14:46 +0000)]
runtime: don't assume that _ = *s will panic if s is nil
With the gc toolchain apparently
var s *string
_ = *s
is enough to panic with a nil pointer dereference. The gccgo compiler
will simply discard the dereference, which I think is a reasonable and
acceptable optimization. Change the tests to use an exported variable
instead. The tests are not currently run, but they will be with a
later patch to gotools.
ian [Thu, 22 Jun 2017 14:44:30 +0000 (14:44 +0000)]
runtime: change some stack fields to uintptr
Because of how gccgo implements cgo calls, the code in dropm may not
have any write barriers. As a step toward implementing that, change
the gcstack, gcnextsegment, and gcnextsp fields of the g struct to
uintptr, so that assignments to them do not require write barriers.
The gcinitialsp field remains unsafe.Pointer, as on 32-bit systems
that do not support split stack it points to a heap allocated space
used for the goroutine stack.
The test for this is runtime tests like TestCgoCallbackGC, which are
not run today but will be run with a future gotools patch.
ian [Thu, 22 Jun 2017 14:33:13 +0000 (14:33 +0000)]
runtime: avoid write barriers when calling deferred function
Calling a deferred function currently requires changing from a uintptr
to the function code to a Go function value. That is done by setting
the value of a func local variable using unsafe.Pointer. The local
variable will always be on the stack. Adjust the code that sets the
local variable to avoid generating a write barrier.
A write barrier is never needed here. Also, for deferreturn, we must
avoid write barriers entirely when called from a cgo function; that
requires more than just this, but this is a start.
The test for this is runtime tests that use the go tool; these are not
currently run, but they will be in the future.
ian [Thu, 22 Jun 2017 14:31:16 +0000 (14:31 +0000)]
runtime: rename _defer struct _panic field to panicStack
The gc version of the _defer struct has a _panic field that has a
completely different meaning. We are going to want that bring that new
meaning into the gofrontend to improve panic reports with nested
panic calls. Simplify that by first renaming the existing _panic field.
ian [Thu, 22 Jun 2017 14:28:57 +0000 (14:28 +0000)]
runtime: adjust tests for gofrontend
- don't run tests that depend on SetCgoTraceback
- don't expect a '(' after the function name in a traceback
- change the expected name of nested functions in a traceback
These tests are not currently run, but they will be soon.
ian [Thu, 22 Jun 2017 14:27:09 +0000 (14:27 +0000)]
runtime: don't build test files that use SetCgoTraceback for gccgo
The gofrontend doesn't support the runtime.SetCgoTraceback function,
which is specifically for handling mixed Go and C tracebacks.
Use a build tag to avoid compiling the runtime/testdata/testprogcgo
files that refer to SetCgoTraceback. These files are not currently
compiled anyhow, but they will be with a future gotools patch.
ian [Thu, 22 Jun 2017 14:23:54 +0000 (14:23 +0000)]
runtime: build testprogcgo with -pthread
Building this test with gccgo requires an explicit -pthread option to
be passed to the C compiler, so that it links against -lpthread.
This test is not built today, but it will be soon with a future patch.
ian [Wed, 21 Jun 2017 22:29:58 +0000 (22:29 +0000)]
compiler: fix missing case in Array_type::get_value_pointer
Update the code in Array_type::get_value_pointer that handles
"lvalue" context to look for both regular var expressions
and temp var expressions, since both can appear in array/slice
index expressions on the left hand side of assignments.
ian [Wed, 21 Jun 2017 22:29:36 +0000 (22:29 +0000)]
Forgot to remove this file in the last libgo commit.
libgo: remove old MIPS architecture names
This removes the old names for the 3 main MIPS ABIs: mipso32, mipsn32
and mipsn64. It also removes the mipso64 ABI which has no equivalent
architecture name in go. This ABI has been dead for sometime and I doubt
anyone will miss it.
ian [Wed, 21 Jun 2017 22:11:04 +0000 (22:11 +0000)]
libgo: remove old MIPS architecture names
This removes the old names for the 3 main MIPS ABIs: mipso32, mipsn32
and mipsn64. It also removes the mipso64 ABI which has no equivalent
architecture name in go. This ABI has been dead for sometime and I doubt
anyone will miss it.
ian [Wed, 21 Jun 2017 21:46:03 +0000 (21:46 +0000)]
libgo, syscall: fix ptrace implementation on MIPS
On MIPS, the correct structure for PtraceRegs is 'struct pt_regs' which
is declared in linux/ptrace.h. Previously no PtraceRegs structure was
created on MIPS because 'struct user_regs_struct' doesn't exist there.
Fallback to using pt_regs when the PtraceRegs structure is generated in
mksysinfo.sh, then adjust syscall_linux_mipsx.go to read the program
counter from the correct field.
In addition, implement PtraceGetRegs and PtraceSetRegs on all 3 ABI
variants.
syscall_linux_mips64x.go can now be removed since the ptrace code on
all 3 ABIs is identical.
ian [Wed, 21 Jun 2017 21:42:41 +0000 (21:42 +0000)]
libgo, syscall: fix ptrace implementation on MIPS
On MIPS, the correct structure for PtraceRegs is 'struct pt_regs' which
is declared in linux/ptrace.h. Previously no PtraceRegs structure was
created on MIPS because 'struct user_regs_struct' doesn't exist there.
Fallback to using pt_regs when the PtraceRegs structure is generated in
mksysinfo.sh, then adjust syscall_linux_mipsx.go to read the program
counter from the correct field.
In addition, implement PtraceGetRegs and PtraceSetRegs on all 3 ABI
variants.
syscall_linux_mips64x.go can now be removed since the ptrace code on
all 3 ABIs is identical.
meissner [Wed, 21 Jun 2017 21:08:40 +0000 (21:08 +0000)]
2017-06-21 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/80510
* gcc.target/powerpc/pr80510-1.c: Restrict test to 64-bit until
32-bit support is added. Change ITYPE size to 64-bit integer.
* gcc.target/powerpc/pr80510-2.c: Likewise.
ville [Wed, 21 Jun 2017 19:53:26 +0000 (19:53 +0000)]
PR libstdc++/80675, PR libstdc++/80940
* include/std/istream:
(__is_convertible_to_basic_istream_test(basic_istream<_Ch, _Up>*)): New.
(__do_is_convertible_to_basic_istream_impl): Likewise.
(__is_convertible_to_basic_istream_impl): Likewise.
(__is_convertible_to_basic_istream): Use the new base.
(__rvalue_istream_type): New.
(operator>>(_Istream&&, _Tp&&)): Use the new helper alias
for the SFINAE check, convert to the helper alias type before
doing the actual extraction.
* include/std/ostream:
(__is_convertible_to_basic_ostream_test(basic_ostream<_Ch, _Up>*)): New.
(__do_is_convertible_to_basic_ostream_impl): Likewise.
(__is_convertible_to_basic_ostream_impl): Likewise.
(__is_convertible_to_basic_ostream): Use the new base.
(__rvalue_ostream_type): New.
(operator<<(_Ostream&&, const _Tp&)): Use the new helper alias
for the SFINAE check, convert to the helper alias type before
doing the actual insertion.
* testsuite/27_io/rvalue_streams-2.cc: Add new tests.
jakub [Wed, 21 Jun 2017 18:30:32 +0000 (18:30 +0000)]
PR c++/81154
* semantics.c (handle_omp_array_sections_1, finish_omp_clauses):
Complain about t not being a variable if t is OVERLOAD even
when processing_template_decl.
dmalcolm [Wed, 21 Jun 2017 16:11:36 +0000 (16:11 +0000)]
C++: Add fix-it hints for -Wold-style-cast
gcc/cp/ChangeLog:
* parser.c (get_cast_suggestion): New function.
(maybe_add_cast_fixit): New function.
(cp_parser_cast_expression): Capture the location of the closing
parenthesis. Call maybe_add_cast_fixit when emitting warnings
about old-style casts.
gcc/testsuite/ChangeLog:
* g++.dg/other/old-style-cast-fixits.C: New test case.
pinskia [Wed, 21 Jun 2017 15:58:12 +0000 (15:58 +0000)]
2017-06-21 Andrew Pinski <apinski@cavium.com>
* config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
Increment Arith_shift and Arith_shift_reg by 1.
* config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
New tuning flag.
* config/aarch64/aarch64.c (thunderx_tunings): Enable
AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
(aarch64_strip_extend): Add new argument and test for it.
(aarch64_cheap_mult_shift_p): New function.
(aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
add a cost if it is true.
Update calls to aarch64_strip_extend.
(aarch64_rtx_costs): Update calls to aarch64_strip_extend.
pinskia [Wed, 21 Jun 2017 15:35:14 +0000 (15:35 +0000)]
2017-06-21 Andrew Pinski <apinski@cavium.com>
* config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
tunings.
(thunderxt88): Likewise.
* config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
(thunderx_prefetch_tune): New variable.
(thunderx2t99_prefetch_tune): Update for the correct values.
(thunderxt88_tunings): New variable.
(thunderx_tunings): Use thunderx_prefetch_tune instead of
generic_prefetch_tune.
(thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
ktkachov [Wed, 21 Jun 2017 15:26:21 +0000 (15:26 +0000)]
[AArch64] Fix atomic_cmp_exchange_zero_reg_1.c with +lse
* config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
(aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
(aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
(aarch64_atomic_cas<mode>, GPI): Likewise.
marxin [Wed, 21 Jun 2017 12:52:14 +0000 (12:52 +0000)]
Rework cold and hot label attributes in predict.c.
2017-06-21 Martin Liska <mliska@suse.cz>
* gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
statements on cold and hot labels.
* predict.c (tree_estimate_probability_bb): Remove the
prediction from this place.
2017-06-21 Martin Liska <mliska@suse.cz>
marxin [Wed, 21 Jun 2017 12:51:46 +0000 (12:51 +0000)]
Make early return predictor more precise.
2017-06-21 Martin Liska <mliska@suse.cz>
PR tree-optimization/79489
* gimplify.c (maybe_add_early_return_predict_stmt): New
function.
(gimplify_return_expr): Call the function.
* predict.c (tree_estimate_probability_bb): Remove handling
of early return.
* predict.def: Update comment about early return predictor.
* gimple-predict.h (is_gimple_predict): New function.
* predict.def: Change default value of early return to 66.
* tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
statements.
* passes.def: Put pass_strip_predict_hints to the beginning of
IPA passes.
pmderodat [Wed, 21 Jun 2017 11:24:51 +0000 (11:24 +0000)]
DWARF: make it possible to emit debug info for declarations only
The DWARF back-end used to systematically ignore file-scope function and
variable declarations. While this is justified in language like C/C++,
where such declarations can appear in several translation units and thus
bloat uselessly the debug info, this behavior is counter-productive in
languages with a well-defined module system. Specifically, it prevents
the description of imported entities, that belong to foreign languages,
making them unavailable from debuggers.
Take for instance:
package C_Binding is
function My_C_Function (I : Integer) return Integer;
pragma Import (C, My_C_Function, "my_c_function");
end C_Binding;
This makes available for Ada programs the C function "my_c_function"
under the following name: C_Binding.My_C_Function. When GCC compiles
it, though, it is represented as a FUNCTION_DECL node with DECL_EXTERNAL
set and a null DECL_INITIAL, which used to be discarded unconditionally
in the DWARF back-end.
This patch moves such filter from the DWARF back-end to the relevant
callers: passes.c:rest_of_decl_compilation and
godump.c:go_early_global_decl. It also This patch also updates the Ada
front-end to call debug hooks for functions such as in the above
example, so that we do generate debugging information for them.
gcc/
* dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
FUNCTION_DECL declarations.
(dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
declarations.
(dwaf2out_decl): Likewise.
* godump.c (go_early_global_decl): Skip call to the real debug hook
for FUNCTION_DECL declarations.
* passes.c (rest_of_decl_compilation): Skip call to the
early_global_decl debug hook for FUNCTION_DECL declarations, unless
-fdump-go-spec is passed.
gcc/ada/
* gcc-interface/ada-tree.h (DECL_FUNCTION_IS_DEF): Update copyright
notice. New macro.
* gcc-interface/trans.c (Subprogram_Body_to_gnu): Tag the subprogram
as a definition.
(Compilation_Unit_to_gnu): Tag the elaboration procedure as a
definition.
* gcc-interface/decl.c (gnat_to_gnu_entity): Tag declarations of
imported subprograms for the current compilation unit as
definitions. Disable debug info for references to variables.
* gcc-interface/gigi.h (create_subprog_decl): Update declaration.
* gcc-interface/utils.c (gnat_pushdecl): Add external DECLs that are
not built-in functions to their binding scope.
(create_subprog_decl): Add a DEFINITION parameter. If it is true, tag
the function as a definition. Update all callers.
(gnat_write_global_declarations): Emit debug info for imported
functions. Filter out external variables for which debug info
is disabled.
gcc/testsuite/
* gnat.dg/debug11_pkg.adb, gnat.dg/debug11_pkg.ads,
gnat.dg/debug11_pkg2.ads: New testcase.
glisse [Wed, 21 Jun 2017 11:20:41 +0000 (11:20 +0000)]
[i386] __builtin_ia32_stmxcsr could be pure
2017-06-21 Marc Glisse <marc.glisse@inria.fr>
gcc/
* config/i386/i386.c (struct builtin_isa): New field pure_p.
Reorder for compactness.
(def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
(def_builtin_pure, def_builtin_pure2): New functions.
(ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
gcc/testsuite/
* gcc.target/i386/getround.c: New file.
jakub [Wed, 21 Jun 2017 10:59:12 +0000 (10:59 +0000)]
* line-map.c (location_adhoc_data_update): Perform addition in
uintptr_t type rather than char * type. Read *data using
ptrdiff_t type instead of int64_t.
(get_combined_adhoc_loc): Change offset type to ptrdiff_t from
int64_t.
wilco [Wed, 21 Jun 2017 10:48:51 +0000 (10:48 +0000)]
Emit SIMD moves as mov
SIMD moves are currently emitted as ORR. Change this to use the MOV
pseudo instruction just like integer moves (the ARM-ARM states MOV is the
preferred disassembly), improving readability of -S output.
gcc/
* config/aarch64/aarch64.md (movti_aarch64):
Emit mov rather than orr.
(movtf_aarch64): Likewise.
* config/aarch64/aarch64-simd.md (aarch64_simd_mov):
Emit mov rather than orr.
wilco [Wed, 21 Jun 2017 10:46:02 +0000 (10:46 +0000)]
Improve dup pattern
Improve the dup pattern to prefer vector registers. When doing a dup
after a load, the register allocator thinks the costs are identical
and chooses an integer load. However a dup from an integer register
includes an int->fp transfer which is not modelled. Adding a '?' to
the integer variant means the cost is increased slightly so we prefer
using a vector register. This improves the following example:
wilco [Wed, 21 Jun 2017 10:40:21 +0000 (10:40 +0000)]
Mark symbols as constant
Aarch64_legitimate_constant_p currently returns false for symbols,
eventhough they are always valid constants. This means LOSYM isn't
CSEd correctly. If we return true CSE works better, resulting in
smaller/faster code (0.3% smaller code on SPEC2006). Avoid this
for TLS symbols since their sequence is complex.
gcc/
* config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
Return true for non-tls symbols.
jgreenhalgh [Wed, 21 Jun 2017 09:58:57 +0000 (09:58 +0000)]
[Patch AArch64] Add initial tuning support for Cortex-A55 and Cortex-A75
This patch adds support for the ARM Cortex-A75 and
Cortex-A55 processors through the -mcpu/-mtune values cortex-a55 and
cortex-a75, and an ARM DynamIQ big.LITTLE configuration of these two
processors through the -mcpu/-mtune value cortex-a75.cortex-a55
The ARM Cortex-A75 is ARM's latest and highest performance applications
processor. For the initial tuning provided in this patch, I have chosen to
share the tuning structure with its predecessor, the Cortex-A73.
The ARM Cortex-A55 delivers the best combination of power efficiency
and performance in its class. For the initial tuning provided in this patch,
I have chosen to share the tuning structure with its predecessor, the
Cortex-A53.
Both Cortex-A55 and Cortex-A75 support ARMv8-A with the ARM8.1-A and
ARMv8.2-A extensions, along with the cryptography extension, and
the RCPC extensions from ARMv8.3-A. This is reflected in the patch,
-mcpu=cortex-a75 is treated as equivalent to passing -mtune=cortex-a75
-march=armv8.2-a+rcpc .
2017-06-21 James Greenhalgh <james.greenhalgh@arm.com>
* config/aarch64/aarch64-cores.def (cortex-a55): New.
(cortex-a75): Likewise.
(cortex-a75.cortex-a55): Likewise.
* config/aarch64/aarch64-tune.md: Regenerate.
* doc/invoke.texi (-mtune): Document new values for -mtune.
vries [Wed, 21 Jun 2017 09:10:16 +0000 (09:10 +0000)]
Add dg-add-options feature stack_size
2017-06-21 Tom de Vries <tom@codesourcery.com>
* doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
stack_size feature.
(Effective-Target Keywords, Other attributes): Suggest using
dg-add-options stack_size feature to get stack limit in stack_size
effective target documentation.
rguenth [Wed, 21 Jun 2017 07:01:34 +0000 (07:01 +0000)]
2017-06-21 Richard Biener <rguenther@suse.de>
PR gcov-profile/81080
* configure.ac: Add AC_SYS_LARGEFILE.
* libgcov.h: Include auto-target.h before tsystem.h to pick
up _FILE_OFFSET_BITS which might differ for multilibs.
* config.in: Regenerate.
* configure: Likewise.
projects / thirdparty / gcc.git / log
