bernds [Tue, 17 Nov 2015 21:20:20 +0000 (21:20 +0000)]
PR target/66785
* regrename.c (record_operand_use): Keep track of failed operands
and stop appending if we see any.
* regrename.h (struct operand_rr_info): Add a failed field and shrink
n_chains to short.
dmalcolm [Tue, 17 Nov 2015 20:09:56 +0000 (20:09 +0000)]
Fix uninitialized src_range values for c_expr
gcc/c/ChangeLog:
* c-parser.c (set_c_expr_source_range): Bulletproof both
overloaded implementations against NULL expr->value.
(c_parser_braced_init): Set src_range for "ret" to a sane pair of
values.
(c_parser_unary_expression): Likewise when handling addresses of
labels.
(c_parser_postfix_expression): Likewise for statement expressions,
for __FUNCTION__, __PRETTY_FUNCTION_ and __func__ keywords, for
__builtin_va_arg, and for __builtin_offset_of.
(c_parser_postfix_expression_after_paren_type): Initialize expr's
src_range using the range of the braced initializer.
(c_parser_transaction_expression): Set src_range for "ret" to a
sane pair of values.
gcc/testsuite/ChangeLog:
* gcc.dg/plugin/diagnostic-test-expressions-1.c (vector): New
macro.
(test_braced_init): New function.
(test_statement_expression): New function.
(test_address_of_label): New function.
(test_transaction_expressions): New function.
(test_keywords): New function.
(test_builtin_va_arg): New function.
(test_builtin_offsetof): New function.
* lib/multiline.exp (_build_multiline_regex): Escape braces.
rsandifo [Tue, 17 Nov 2015 18:56:52 +0000 (18:56 +0000)]
Simplify rs6000_builtin_vectorized_function
After the previous patches it's no longer necessary for
TARGET_BUILTIN_VECTORIZED_FUNCTION to return functions that
map to the vector optab of the original operation. We'll use
a vector form of the internal function instead.
Tested on powerpc64-linux-gnu.
gcc/
* config/rs6000/rs6000.c (rs6000_builtin_vectorized_function): Remove
entries that map directly to optabs.
rsandifo [Tue, 17 Nov 2015 18:56:21 +0000 (18:56 +0000)]
Simplify ix86_builtin_vectorized_function
After the previous patches it's no longer necessary for
TARGET_BUILTIN_VECTORIZED_FUNCTION to return functions that
map to the vector optab of the original operation. We'll use
a vector form of the internal function instead.
Tested on x86_64-linux-gnu.
gcc/
* config/i386/i386.c (ix86_builtin_vectorized_function): Remove
entries that map directly to optabs.
rsandifo [Tue, 17 Nov 2015 18:55:55 +0000 (18:55 +0000)]
Vectorize internal functions
This patch tries to vectorize built-in and internal functions as
internal functions first, falling back on the current built-in
target hooks otherwise.
This means that we'll automatically pick up vector versions of optabs
without the target having to implement any special hooks. E.g. we'll
use V4SF sqrt if the target defines a "sqrtv4sf2" optab. As well as
being simpler, it means that the target-independent code has more
idea what the vectorized function does.
Tested on x86_64-linux-gnu, aarch64-linux-gnu, arm-linux-gnu and
powerpc64-linux-gnu.
gcc/
* internal-fn.h (direct_internal_fn_info): Add vectorizable flag.
* internal-fn.c (direct_internal_fn_array): Update accordingly.
* tree-vectorizer.h (vectorizable_function): Delete.
* tree-vect-stmts.c: Include internal-fn.h.
(vectorizable_internal_function): New function.
(vectorizable_function): Inline into...
(vectorizable_call): ...here. Explicitly reject calls that read
from or write to memory. Try using an internal function before
falling back on the old vectorizable_function behavior.
rsandifo [Tue, 17 Nov 2015 18:55:13 +0000 (18:55 +0000)]
Make builtin_vectorized_function take a combined_fn
This patch replaces the fndecl argument to builtin_vectorized_function
with a combined_fn and gets the vectoriser to call it for internal
functions too. The patch also moves vectorisation of machine-specific
built-ins to a new hook, builtin_md_vectorized_function.
Tested on x86_64-linux-gnu, aarch64-linux-gnu, arm-linux-gnu and
powerpc64-linux-gnu.
gcc/
* target.def (builtin_vectorized_function): Take a combined_fn (in
the form of an unsigned int) rather than a function decl.
(builtin_md_vectorized_function): New.
* targhooks.h (default_builtin_vectorized_function): Replace the
fndecl argument with an unsigned int.
(default_builtin_md_vectorized_function): Declare.
* targhooks.c (default_builtin_vectorized_function): Replace the
fndecl argument with an unsigned int.
(default_builtin_md_vectorized_function): New function.
* doc/tm.texi.in (TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION):
New hook.
* doc/tm.texi: Regenerate.
* tree-vect-stmts.c (vectorizable_function): Update call to
builtin_vectorized_function, also passing internal functions.
Call builtin_md_vectorized_function for target-specific builtins.
* config/aarch64/aarch64-protos.h
(aarch64_builtin_vectorized_function): Replace fndecl argument
with an unsigned int.
* config/aarch64/aarch64-builtins.c: Include case-cfn-macros.h.
(aarch64_builtin_vectorized_function): Update after above changes.
Use CASE_CFN_*.
* config/arm/arm-protos.h (arm_builtin_vectorized_function): Replace
fndecl argument with an unsigned int.
* config/arm/arm-builtins.c: Include case-cfn-macros.h
(arm_builtin_vectorized_function): Update after above changes.
Use CASE_CFN_*.
* config/i386/i386.c: Include case-cfn-macros.h
(ix86_veclib_handler): Take a combined_fn rather than a
built_in_function.
(ix86_veclibabi_svml, ix86_veclibabi_acml): Likewise. Use
mathfn_built_in rather than calling builtin_decl_implicit directly.
(ix86_builtin_vectorized_function) Update after above changes.
Use CASE_CFN_*.
* config/rs6000/rs6000.c: Include case-cfn-macros.h
(rs6000_builtin_vectorized_libmass): Replace fndecl argument
with a combined_fn. Use CASE_CFN_*. Use mathfn_built_in rather
than calling builtin_decl_implicit directly.
(rs6000_builtin_vectorized_function): Update after above changes.
Use CASE_CFN_*. Move BUILT_IN_MD to...
(rs6000_builtin_md_vectorized_function): ...this new function.
(TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION): Define.
rsandifo [Tue, 17 Nov 2015 18:54:36 +0000 (18:54 +0000)]
Use IFN_SQRT in tree-vect-patterns.c
In practice all targets that can vectorise sqrt define the appropriate
sqrt<mode>2 optab. The only case where this isn't immediately obvious
is the libmass support in rs6000.c, but Mike Meissner said that it shouldn't
be exercised for sqrt.
This patch therefore uses the internal function interface instead of
going via the target hook.
Tested on x86_64-linux-gnu, aarch64-linux-gnu, arm-linux-gnueabi and
powerpc64-linux-gnu.
gcc/
* tree-vect-patterns.c: Include internal-fn.h.
(vect_recog_pow_pattern): Use IFN_SQRT instead of BUILT_IN_SQRT*.
rsandifo [Tue, 17 Nov 2015 18:51:55 +0000 (18:51 +0000)]
Extend tree-call-cdce to calls whose result is used
For -fmath-errno, builtins.c currently expands calls to sqrt to:
y = sqrt_optab (x);
if (y != y)
[ sqrt (x); or errno = EDOM; ]
The drawbacks of this are:
- the call to sqrt is protected by the result of the optab rather
than the input. It would be better to check __builtin_isless (x, 0),
like tree-call-cdce.c does.
- the branch isn't exposed at the gimple level and so gets little
high-level optimisation.
- we do this for log too, but for log a zero input produces
-inf rather than a NaN, and sets errno to ERANGE rather than EDOM.
This patch moves the code to tree-call-cdce.c instead, with the optab
operation being represented as an internal function. This means that
we can use the existing argument-based range checks rather than the
result-based checks and that we get more gimple optimisation of
the branch.
Previously the pass was only enabled by default at -O2 or above,
but the old builtins.c code was enabled at -O. The patch therefore
enables the pass at -O as well.
The previous patch to cfgexpand.c handled cases where functions
don't (or are assumed not to) set errno, so this patch makes
the builtins.c code dead.
Tested on x86_64-linux-gnu, aarch64-linux-gnu, arm-linux-gnueabi
and visium-elf (for the EDOM stuff).
gcc/
* builtins.c (expand_errno_check, expand_builtin_mathfn)
(expand_builtin_mathfn_2): Delete.
(expand_builtin): Remove handling of functions with
internal function equivalents.
* internal-fn.def (SET_EDOM): New internal function.
* internal-fn.h (set_edom_supported_p): Declare.
* internal-fn.c (expand_SET_EDOM): New function.
(set_edom_supported_p): Likewise.
* tree-call-cdce.c: Include builtins.h and internal-fn.h.
Rewrite comment at head of file.
(is_call_dce_candidate): Rename to...
(can_test_argument_range): ...this. Don't check gimple_call_lhs
or gimple_call_builtin_p here.
(edom_only_function): New function.
(shrink_wrap_one_built_in_call_with_conds): New function, split out
from...
(shrink_wrap_one_built_in_call): ...here.
(can_use_internal_fn, use_internal_fn): New functions.
(shrink_wrap_conditional_dead_built_in_calls): Call use_internal_fn
for calls that have an lhs.
(pass_call_cdce::gate): Remove optimize_function_for_speed_p check.
(pass_call_cdce::execute): Skip blocks that are optimized for size.
Check gimple_call_builtin_p here. Use can_use_internal_fn for
calls with an lhs.
* opts.c (default_options_table): Enable -ftree-builtin-call-cdce
at -O and above.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* Makefile.in (MOSTLYCLEANFILES): Add cfn-operators.pd.
(generated_files): Likewise.
(s-cfn-operators, cfn-operators.pd): New rules.
(s-match): Depend on cfn-operators.pd.
* gencfn-macros.c: Expand comment to describe -o behavior.
(print_define_operator_list): New function.
(main): Accept -o. Call print_define_operator_list.
* genmatch.c (main): Add the current directory to the include path.
* match.pd (DEFINE_MATH_FN): Delete. Include cfn-operators.pd
instead.
rsandifo [Tue, 17 Nov 2015 18:48:23 +0000 (18:48 +0000)]
Add null identifiers to genmatch
This patch adds a null identifier that can never match anything and
can never be generated. It is only valid in operator lists and fors.
Later patches will add uses of it.
The idea is to allow operator lists for maths functions that have
four entries:
- float built-in
- double built-in
- long double built-in
- internal function
Not all maths functions have an associated internal function,
and for those the final operator will be "null". Any simplification
that tries to use a null substitution will be skipped.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* doc/match-and-simplify.texi: Document the "null" identifier.
* genmatch.c (id_base::NULL_ID): New kind.
(null_id): New variable.
(get_operator): Add a parameter that says whether null identifiers
are allowed.
(contains_id): New function.
(lower_for): Skip substitutions that would have a null_id in
either the match or the result.
(parser::parse_for): Allow the null identifier to be used.
(parser::parse_operator_list): Likewise.
(main): Initialize null_id.
rsandifo [Tue, 17 Nov 2015 18:47:44 +0000 (18:47 +0000)]
Add genmatch support for internal functions
This patch makes genmatch match calls based on combined_fn rather
than built_in_function and extends the matching to internal functions.
It also uses fold_const_call to fold the calls to a constant, rather
than going through fold_builtin_n.
In order to slightly simplify the code and remove potential
ambiguity, the patch enforces lower case for tree codes
(foo->FOO_EXPR), caps for functions (no built_in_hypot->BUILT_IN_HYPOT)
and requires an exact match for user-defined identifiers. The first two
were already met in practice but there were a couple of cases where
operator lists were defined in one case and used in another.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* match.pd: Use HYPOT and COS rather than hypot and cos.
Use CASE_CFN_* macros. Guard log/exp folds with
SCALAR_FLOAT_TYPE_P.
* genmatch.c (internal_fn): New enum.
(fn_id::fn): Change to an unsigned int.
(fn_id::fn_id): Accept internal_fn too.
(add_builtin): Rename to...
(add_function): ...this and turn into a template.
(get_operator): Only try one variation if the original name fails.
Only add _EXPR if the original name was all lower case.
Try converting internal and built-in function names to their
CFN equivalents.
(expr::gen_transform): Use maybe_build_call_expr_loc for generic.
(dt_simplify::gen_1): Likewise.
(dt_node::gen_kids_1): Use gimple_call_combined_fn for gimple
and get_call_combined_fn for generic.
(dt_simplify::gen): Use combined_fn as the type of fn_ids.
(decision_tree::gen): Likewise.
(main): Use lower case in the strings for {VIEW_,}CONVERT[012].
Use add_function rather than add_builtin. Register internal
functions too.
* generic-match-head.c: Include case-cfn-macros.h.
* gimple-fold.c (replace_stmt_with_simplification): Use
gimple_call_combined_fn to test whether we can keep an
existing call.
* gimple-match.h (code_helper): Replace built_in_function
with combined_fn.
* gimple-match-head.c: Include fold-const-call.h, internal-fn.h
and case-fn-macros.h.
(gimple_resimplify1): Use fold_const_call.
(gimple_resimplify2, gimple_resimplify3): Likewise.
(build_call_internal, build_call): New functions.
(maybe_push_res_to_seq): Use them.
(gimple_simplify): Use fold_const_call. Set *rcode to a combined_fn
rather than a built-in function.
* tree.h (build_call_expr_internal_loc): Declare.
(maybe_build_call_expr_loc): Likewise.
* tree.c (build_call_expr_internal_loc_array): New function.
(maybe_build_call_expr_loc): Likewise.
rsandifo [Tue, 17 Nov 2015 18:47:02 +0000 (18:47 +0000)]
Extend mathfn_built_in to handle combined_fn
This patch extends mathfn_built_in to handle combined_fn, but keeps the
old built_in_function interface around since it's a common case.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* builtins.h (mathfn_built_in): Add a variant that takes
a combined_fn.
* builtins.c: Include case-cfn-macros.h.
(CASE_MATHFN): Use CASE_CFN_*.
(CASE_MATHFN_REENT): Use CFN_ codes.
(mathfn_built_in_2, mathfn_built_in_1): Replace built_in_function
argument with a combined_fn.
(mathfn_built_in): Add a variant that takes a combined_fn.
(expand_builtin_int_roundingfn_2): Update callers accordingly.
(fold_builtin_sincos, fold_builtin_classify): Likewise.
rsandifo [Tue, 17 Nov 2015 18:46:22 +0000 (18:46 +0000)]
Use combined_fn in tree-vect-patterns.c
Another patch to extend uses of built_in_function to combined_fn,
this time in tree-vect-patterns.c. The old code didn't handle the
long double pow variants, but I think that's because noone had a target
that would benefit rather than because the code would mishandle them.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* tree-vect-patterns.c: Include case-cfn-macros.h.
(vect_recog_pow_pattern): Use combined_fn instead of built-in codes.
rsandifo [Tue, 17 Nov 2015 18:45:41 +0000 (18:45 +0000)]
Use combined_fn in tree-ssa-math-opts.c
Another patch to extend uses of built_in_function to combined_fn, this time
in tree-ssa-math-opts.c.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* tree-ssa-math-opts.c: Include case-cfn-macros.h.
(execute_cse_sincos_1): Use combined_fn instead of built-in codes.
(pass_cse_sincos::execute): Likewise.
rsandifo [Tue, 17 Nov 2015 18:44:22 +0000 (18:44 +0000)]
Use combined_fn in tree-vrp.c
Another patch to extend uses of built_in_function to combined_fn, this time
in tree-vrp.c.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* tree-vrp.c: Include case-cfn-macros.h.
(extract_range_basic): Switch on combined_fn rather than handling
built-in functions and internal functions separately.
rsandifo [Tue, 17 Nov 2015 18:43:42 +0000 (18:43 +0000)]
Make more use of combined_fn
This patch generalises fold-const.[hc] routines to use combined_fn
instead of built_in_function. It also updates gimple-ssa-backprop,c
since the update is simple and it avoids churn on the call to
negate_mathfn_p.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* fold-const.h (negate_mathfn_p): Take a combined_fn rather
than a built_in_function.
(tree_call_nonnegative_warnv_p): Take a combined_fn rather than
a function decl.
(integer_valued_real_call_p): Likewise.
* fold-const.c: Include case-cfn-macros.h
(negate_mathfn_p): Take a combined_fn rather than a built_in_function.
(negate_expr_p): Update accordingly.
(tree_call_nonnegative_warnv_p): Take a combined_fn rather than
a function decl.
(integer_valued_real_call_p): Likewise.
(tree_invalid_nonnegative_warnv_p): Update accordingly.
(integer_valued_real_p): Likewise.
* gimple-fold.c (gimple_call_nonnegative_warnv_p): Update call
to tree_call_nonnegative_warnv_p.
(gimple_call_integer_valued_real_p): Likewise
integer_valued_real_call_p.
* gimple-ssa-backprop.c: Include case-cfn-macros.h.
(backprop::process_builtin_call_use): Extend to combined_fn.
(strip_sign_op_1): Likewise.
(backprop::process_use): Don't check for built-in calls here.
(backprop::execute): Likewise.
(backprop::optimize_builtin_call): Update call to negate_mathfn_p.
rsandifo [Tue, 17 Nov 2015 18:41:55 +0000 (18:41 +0000)]
Add gencfn-macros.c
This patch automatically generates case macros such as:
CASE_CFN_SQRT
for each {F,,L} floating-point built-in function and each {,L,LL,IMAX}
integer built-in function. The macros match the same built-in
functions as CASE_FLT_FN and CASE_INT_FN but in addition include
the associated internal function, if any.
The idea is to make sure that users of combined_fn don't need to know
which built-in functions have internal-function equivalents. If we add
a new function to internal-fn.def, all combined_fn users should pick it
up automatically.
The generator wants to use "hash_set <nofree_string_hash>",
so the patch follows hash_map in using the types given by the
traits as the key. This is a no-op for current users of hash_set.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* Makefile.in (HASH_TABLE_H): Add GGC_H.
(MOSTLYCLEANFILES, generated_files): Add case-fn-macros.h.
(s-case-cfn-macros, case-cfn-macros.h, build/gencfn-macros.o)
(build/gencfn-macros$(build_exeext): New rules.
(genprogerr): Add cfn-macros.
* hash-set.h (hash_set): Use the traits value_type as the key.
* gencfn-macros.c: New file.
rsandifo [Tue, 17 Nov 2015 18:40:31 +0000 (18:40 +0000)]
Add internal bitcount functions
This patch adds internal function equivalents of all the INT_FN functions.
Unlike the math functions, these functions never set errno and the internal
functions should be exactly equivalent to the built-in ones. The reason
for defining the internal functions is so that we can extend the
functionality to other modes, in particular vector modes.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* internal-fn.def (DEF_INTERNAL_INT_FN): New macro.
(CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): New functions.
* builtins.c (associated_internal_fn): Handle them.
rsandifo [Tue, 17 Nov 2015 18:39:02 +0000 (18:39 +0000)]
Add internal math functions
This patch adds internal functions for simple FLT_FN built-in functions,
in cases where an associated optab already exists. Unlike some of the
built-in functions, these internal functions never set errno.
LDEXP is an odd-one out in that its second operand is an integer.
All the others operate on uniform types.
The patch also adds a function to query the internal function associated
with a built-in function (if any), and another to test whether a given
gcall could be replaced by a call to an internal function on the current
target (as long as the caller deals with errno appropriately).
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* builtins.h (associated_internal_fn): Declare.
(replacement_internal_fn): Likewise.
* builtins.c: Include internal-fn.h
(associated_internal_fn, replacement_internal_fn): New functions.
* internal-fn.def (DEF_INTERNAL_FLT_FN): New macro.
(ACOS, ASIN, ATAN, COS, EXP, EXP10, EXP2, EXPM1, LOG, LOG10, LOG1P)
(LOG2, LOGB, SIGNIFICAND, SIN, SQRT, TAN, CEIL, FLOOR, NEARBYINT)
(RINT, ROUND, TRUNC, ATAN2, COPYSIGN, FMOD, POW, REMAINDER, SCALB)
(LDEXP): New functions.
* internal-fn.c: Include recog.h.
(unary_direct, binary_direct): New macros.
(expand_direct_optab_fn): New function.
(expand_unary_optab_fn): New macro.
(expand_binary_optab_fn): Likewise.
(direct_unary_optab_supported_p): Likewise.
(direct_binary_optab_supported_p): Likewise.
rsandifo [Tue, 17 Nov 2015 18:37:45 +0000 (18:37 +0000)]
Add basic support for direct_optab internal functions
This patch adds a concept of internal functions that map directly to an
optab (here called "direct internal functions"). The function can only
be used if the associated optab can be used.
so the patch converts them to the new infrastructure. These four
all need different types of optabs, but future patches will add
regular unary and binary ones.
In general we need one or two modes to decide whether an optab is
supported, depending on whether it's a convert_optab or not.
This in turn means that we need up to two types to decide whether
an internal function is supported. The patch records which types
are needed for each internal function, using -1 if the return type
should be used and N>=0 if the type of argument N should be used.
(LOAD_LANES and STORE_LANES are unusual in that both optab modes
come from the same array type.)
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* coretypes.h (tree_pair): New type.
* internal-fn.def (DEF_INTERNAL_OPTAB_FN): New macro. Use it
for MASK_LOAD, LOAD_LANES, MASK_STORE and STORE_LANES.
* internal-fn.h (direct_internal_fn_info): New structure.
(direct_internal_fn_array): Declare.
(direct_internal_fn_p, direct_internal_fn): New functions.
(direct_internal_fn_types, direct_internal_fn_supported_p): Declare.
* internal-fn.c (not_direct, mask_load_direct, load_lanes_direct)
(mask_store_direct, store_lanes_direct): New macros.
(direct_internal_fn_array) New array.
(get_multi_vector_move): Return the optab handler without asserting
that it is available.
(expand_LOAD_LANES): Rename to...
(expand_load_lanes_optab_fn): ...this and add an optab argument.
(expand_STORE_LANES): Rename to...
(expand_store_lanes_optab_fn): ...this and add an optab argument.
(expand_MASK_LOAD): Rename to...
(expand_mask_load_optab_fn): ...this and add an optab argument.
(expand_MASK_STORE): Rename to...
(expand_mask_store_optab_fn): ...this and add an optab argument.
(direct_internal_fn_types, direct_optab_supported_p)
(multi_vector_optab_supported_p, direct_internal_fn_supported_p)
(direct_internal_fn_supported_p): New functions.
(direct_mask_load_optab_supported_p): New macro.
(direct_load_lanes_optab_supported_p): Likewise.
(direct_mask_store_optab_supported_p): Likewise.
(direct_store_lanes_optab_supported_p): Likewise.
rsandifo [Tue, 17 Nov 2015 18:34:47 +0000 (18:34 +0000)]
Add a combined_fn enum
I'm working on a patch series that needs to be able to treat built-in
functions and internal functions in a similar way. This patch adds a
new enum, combined_fn, that combines the two together. It also adds
utility functions for seeing which combined_fn (if any) is called by
a given CALL_EXPR or gcall.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
gcc/
* tree-core.h (internal_fn): Move immediately after the definition
of built_in_function.
(combined_fn): New enum.
* tree.h (as_combined_fn, builtin_fn_p, as_builtin_fn)
(internal_fn_p, as_internal_fn): New functions.
(get_call_combined_fn, combined_fn_name): Declare.
* tree.c (get_call_combined_fn): New function.
(combined_fn_name): Likewise.
* gimple.h (gimple_call_combined_fn): Declare.
* gimple.c (gimple_call_combined_fn): New function.
PR target/56036
* doc/invoke.texi (Option Summary): Add -mms-bitfields to x86
option list.
(x86 Options): Add -mms-bitfields and -mno-ms-bitfields. Move
discussion of the Microsoft structure layout details here from
its former home in extend.texi.
* doc/extend.texi (x86 Variable Attributes): Replace detailed
discussion with pointer to its new location. Add cross-reference
to corresponding type attributes.
(x86 Type Attributes): Add cross-references to command-line options
and variable attributes.
ktkachov [Tue, 17 Nov 2015 13:20:08 +0000 (13:20 +0000)]
[ARM] PR 68143 Properly update memory offsets when expanding setmem
PR target/68143
* config/arm/arm.c (arm_block_set_unaligned_vect): Keep track of
offset from dstbase and use it appropriately in
adjust_automodify_address.
(arm_block_set_aligned_vect): Likewise.
* tree-if-conv.c: Include varasm.h
(ref_DR_map): Define.
(baseref_DR_map): Like wise
(struct ifc_dr): Add new tree predicate field.
(hash_memrefs_baserefs_and_store_DRs_read_written_info): New function.
(memrefs_read_or_written_unconditionally): Remove.
(write_memrefs_written_at_least_once): Remove.
(ifcvt_memrefs_wont_trap): Use hash maps to query
unconditional read/written information.
(if_convertible_loop_p_1): Initialize hash maps and predicates
before hashing data references and delete hashmaps at the end.
meissner [Mon, 16 Nov 2015 22:13:21 +0000 (22:13 +0000)]
2015-11-16 Michael Meissner <meissner@linux.vnet.ibm.com>
* config/rs6000/vsx.md (VSX_L): Do not include IBM extended double
128-bit types, just types that fit in a single vector.
* config/rs6000/rs6000.md (FMOVE128_GPR): Likewise.
kargl [Mon, 16 Nov 2015 19:15:25 +0000 (19:15 +0000)]
2015-11-16 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/58027
PR fortran/60993
* expr.c (gfc_check_init_expr): Prevent a redundant check when a
__convert_* function was inserted into an array constructor.
(gfc_check_assign_symbol): Check for an initialization expression
when a __convert_* was inserted.
2015-11-16 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/58027
PR fortran/60993
* gfortran.dg/pr58027.f90: New test.
segher [Mon, 16 Nov 2015 15:51:33 +0000 (15:51 +0000)]
simplify-rtx: Simplify sign_extend of lshiftrt to zero_extend (PR68330)
Since r230164, in PR68330 combine ends up with a sign_extend of an
lshiftrt by some constant, and it does not know to morph that into a
zero_extract (the extend will always extend with zeroes). I think
it is best to let simplify-rtx always replace such a sign_extend by
a zero_extend, after which everything works as expected.
rguenth [Mon, 16 Nov 2015 15:04:00 +0000 (15:04 +0000)]
2015-11-16 Richard Biener <rguenther@suse.de>
PR tree-optimization/68306
* tree-vect-data-refs.c (vect_verify_datarefs_alignment): Fix
bogus copying from verify_data_ref_alignment and use continue
instead of return.
olegendo [Mon, 16 Nov 2015 14:11:50 +0000 (14:11 +0000)]
gcc/
PR target/68277
* config/sh/sh.md (addsi3_scr): Handle reg overlap of operands[0] and
operands[2].
(*addsi3): Add another insn_and_split variant for reload.
vries [Mon, 16 Nov 2015 12:40:41 +0000 (12:40 +0000)]
Remove first_pass_instance from pass_ccp
2015-11-16 Tom de Vries <tom@codesourcery.com>
* passes.def: Add arg to pass_ccp pass instantiation.
* tree-ssa-ccp.c (ccp_finalize): Add param nonzero_p. Use nonzero_p
instead of first_pass_instance.
(do_ssa_ccp): Add and handle param nonzero_p.
(pass_ccp::pass_ccp): Initialize nonzero_p.
(pass_ccp::set_pass_param): New member function. Set nonzero_p.
(pass_ccp::execute): Call do_ssa_ccp with extra arg.
(pass_ccp::nonzero_p): New private member.
vries [Mon, 16 Nov 2015 12:40:33 +0000 (12:40 +0000)]
Remove first_pass_instance from pass_object_sizes
2015-11-16 Tom de Vries <tom@codesourcery.com>
* passes.def: Add arg to pass_object_sizes pass instantiation.
* tree-object-size.c (pass_object_sizes::pass_object_sizes): Initialize
insert_min_max_p.
(pass_object_sizes::set_pass_param): New member function. Set
insert_min_max_p.
(pass_object_sizes::insert_min_max_p): New private member.
(pass_object_sizes::execute): Use insert_min_max_p instead of
first_pass_instance.
vries [Mon, 16 Nov 2015 12:40:24 +0000 (12:40 +0000)]
Remove first_pass_instance from pass_dominator
2015-11-16 Tom de Vries <tom@codesourcery.com>
* passes.def: Add arg to pass_dominator pass instantiation.
* tree-pass.h (first_pass_instance): Remove pass_dominator-related bit
of comment.
* tree-ssa-dom.c (pass_dominator::pass_dominator): Initialize
may_peel_loop_headers_p.
(pass_dominator::set_pass_param): New member function. Set
may_peel_loop_headers_p.
(pass_dominator::may_peel_loop_headers_p): New private member.
(pass_dominator::execute): Use may_peel_loop_headers_p instead of
first_pass_instance.
vries [Mon, 16 Nov 2015 12:40:14 +0000 (12:40 +0000)]
Remove first_pass_instance from pass_reassoc
2015-11-16 Tom de Vries <tom@codesourcery.com>
* passes.def: Add arg to pass_reassoc pass instantiation.
* tree-ssa-reassoc.c (reassoc_insert_powi_p): New static variable.
(acceptable_pow_call, reassociate_bb): Use reassoc_insert_powi_p instead
of first_pass_instance.
(execute_reassoc): Add and handle insert_powi_p parameter.
(pass_reassoc::insert_powi_p): New private member.
(pass_reassoc::pass_reassoc): Initialize insert_powi_p.
(pass_reassoc::set_pass_param): New member function. Set insert_powi_p.
(pass_reassoc::execute): Call execute_reassoc with extra arg.
ebotcazou [Mon, 16 Nov 2015 12:16:54 +0000 (12:16 +0000)]
* config/i386/i386.c (ix86_adjust_stack_and_probe): Adjust and use
an lea instruction when possible.
(output_adjust_stack_and_probe): Rotate the loop and simplify.
(ix86_emit_probe_stack_range): Adjust.
(output_probe_stack_range): Rotate the loop and simplify.
tbsaunde [Mon, 16 Nov 2015 02:28:15 +0000 (02:28 +0000)]
PR 68366 - include emit-rtl.h in sdbout.c
Some of the pa target macros rely on macros in emit-rtl.h and sdbout.c
uses some of those macros, which means that sdbout.c needs to include
emit-rtl.h.
pault [Sun, 15 Nov 2015 14:07:52 +0000 (14:07 +0000)]
2015-11-15 Paul Thomas <pault@gcc.gnu.org>
PR fortran/50221
PR fortran/68216
PR fortran/63932
PR fortran/66408
* trans_array.c (gfc_conv_scalarized_array_ref): Pass the
symbol decl for deferred character length array references.
* trans-stmt.c (gfc_trans_allocate): Keep the string lengths
to update deferred length character string lengths.
* trans-types.c (gfc_get_dtype_rank_type); Use the string
length of deferred character types for the dtype size.
* trans.c (gfc_build_array_ref): For references to deferred
character arrays, use the domain max value, if it is a variable
to set the 'span' and use pointer arithmetic for acces to the
element.
(trans_code): Set gfc_current_locus for diagnostic purposes.
PR fortran/67674
* trans-expr.c (gfc_conv_procedure_call): Do not fix deferred
string lengths of components.
PR fortran/49954
* resolve.c (deferred_op_assign): New function.
(gfc_resolve_code): Call it.
* trans-array.c (concat_str_length): New function.
(gfc_alloc_allocatable_for_assignment): Jump directly to alloc/
realloc blocks for deferred character length arrays because the
string length might change, even if the shape is the same. Call
concat_str_length to obtain the string length for concatenation
since it is needed to compute the lhs string length.
Set the descriptor dtype appropriately for the new string
length.
* trans-expr.c (gfc_trans_assignment_1): Use the rse string
length for all characters, other than deferred types. For
concatenation operators, push the rse.pre block to the inner
most loop so that the temporary pointer and the assignments
are properly placed.
2015-11-15 Paul Thomas <pault@gcc.gnu.org>
PR fortran/50221
* gfortran.dg/deferred_character_1.f90: New test.
* gfortran.dg/deferred_character_4.f90: New test for comment
#4 of the PR.
PR fortran/68216
* gfortran.dg/deferred_character_2.f90: New test.
PR fortran/67674
* gfortran.dg/deferred_character_3.f90: New test.
PR fortran/63932
* gfortran.dg/deferred_character_5.f90: New test.
PR fortran/66408
* gfortran.dg/deferred_character_6.f90: New test.
PR fortran/49954
* gfortran.dg/deferred_character_7.f90: New test.
andreast [Sat, 14 Nov 2015 21:17:24 +0000 (21:17 +0000)]
2015-11-14 Andreas Tobler <andreast@gcc.gnu.org>
* acinclude.m4 (GLIBCXX_ENABLE_CLOCALE): Change locale implementation
from darwin to DragonFly.
* configure: Regenerate.
* config/os/bsd/freebsd/ctype_configure_char.cc: Improve locale
support, do it the same as DragonFly.
* config/os/bsd/freebsd/os_defines.h: Add fine grained C99 defines.
jakub [Sat, 14 Nov 2015 18:42:13 +0000 (18:42 +0000)]
gcc/
2015-11-14 Jakub Jelinek <jakub@redhat.com>
* omp-low.c (lower_omp_ordered): Add argument to GOMP_SMD_ORDERED_*
internal calls - 0 if ordered simd and 1 for ordered threads simd.
* tree-vectorizer.c (adjust_simduid_builtins): If GOMP_SIMD_ORDERED_*
argument is 1, replace it with GOMP_ordered_* call instead of removing
it.
gcc/c/
2015-11-14 Jakub Jelinek <jakub@redhat.com>
* c-typeck.c (c_finish_omp_clauses): Don't mark
GOMP_MAP_FIRSTPRIVATE_POINTER decls addressable.
gcc/cp/
2015-11-14 Jakub Jelinek <jakub@redhat.com>
* semantics.c (finish_omp_clauses): Don't mark
GOMP_MAP_FIRSTPRIVATE_POINTER decls addressable.
libgomp/
2015-11-14 Jakub Jelinek <jakub@redhat.com>
Aldy Hernandez <aldyh@redhat.com>
Ilya Verbin <ilya.verbin@intel.com>
* ordered.c (gomp_doacross_init, GOMP_doacross_post,
GOMP_doacross_wait, gomp_doacross_ull_init, GOMP_doacross_ull_post,
GOMP_doacross_ull_wait): For GFS_GUIDED don't divide number of
iterators or IV by chunk size.
* parallel.c (gomp_resolve_num_threads): Don't assume that
if thr->ts.team is non-NULL, then pool must be non-NULL.
* libgomp-plugin.h (GOMP_PLUGIN_target_task_completion): Declare.
* libgomp.map (GOMP_PLUGIN_1.1): New symbol version, export
GOMP_PLUGIN_target_task_completion.
* Makefile.am (libgomp_la_SOURCES): Add priority_queue.c.
* Makefile.in: Regenerate.
* libgomp.h: Shuffle prototypes and forward definitions around so
priority queues can be defined.
(enum gomp_task_kind): Add GOMP_TASK_ASYNC_RUNNING.
(enum gomp_target_task_state): New enum.
(struct gomp_target_task): Add state, tgt, task and team fields.
(gomp_create_target_task): Change return type to bool, add
state argument.
(gomp_target_task_fn): Change return type to bool.
(struct gomp_device_descr): Add async_run_func.
(struct gomp_task): Remove children, next_child, prev_child,
next_queue, prev_queue, next_taskgroup, prev_taskgroup.
Add pnode field.
(struct gomp_taskgroup): Remove children.
Add taskgroup_queue.
(struct gomp_team): Change task_queue type to a priority queue.
(splay_compare): Define inline.
(priority_queue_offset): New.
(priority_node_to_task): New.
(task_to_priority_node): New.
* oacc-mem.c: Do not include splay-tree.h.
* priority_queue.c: New file.
* priority_queue.h: New file.
* splay-tree.c: Do not include splay-tree.h.
(splay_tree_foreach_internal): New.
(splay_tree_foreach): New.
* splay-tree.h: Become re-entrant if splay_tree_prefix is defined.
(splay_tree_callback): Define typedef.
* target.c (splay_compare): Move to libgomp.h.
(GOMP_target): Don't adjust *thr in any way around running offloaded
task.
(GOMP_target_ext): Likewise. Handle target nowait.
(GOMP_target_update_ext, GOMP_target_enter_exit_data): Check
return value from gomp_create_target_task, if false, fallthrough
as if no dependencies exist.
(gomp_target_task_fn): Change return type to bool, return true
if the task should have another part scheduled later. Handle
target nowait.
(gomp_load_plugin_for_device): Initialize async_run.
* task.c (gomp_init_task): Initialize children_queue.
(gomp_clear_parent_in_list): New.
(gomp_clear_parent_in_tree): New.
(gomp_clear_parent): Handle priorities.
(GOMP_task): Likewise.
(priority_queue_move_task_first,
gomp_target_task_completion, GOMP_PLUGIN_target_task_completion):
New functions.
(gomp_create_target_task): Use priority queues. Change return type
to bool, add state argument, return false if for async
{{enter,exit} data,update} constructs no dependencies need to be
waited for, handle target nowait. Set task->fn to NULL instead of
gomp_target_task_fn.
(verify_children_queue): Remove.
(priority_list_upgrade_task): New.
(priority_queue_upgrade_task): New.
(verify_task_queue): Remove.
(priority_list_downgrade_task): New.
(priority_queue_downgrade_task): New.
(gomp_task_run_pre): Use priority queues.
Abstract code out to priority_queue_downgrade_task.
(gomp_task_run_post_handle_dependers): Use priority queues.
(gomp_task_run_post_remove_parent): Likewise.
(gomp_task_run_post_remove_taskgroup): Likewise.
(gomp_barrier_handle_tasks): Likewise. Handle target nowait target
tasks specially.
(GOMP_taskwait): Likewise.
(gomp_task_maybe_wait_for_dependencies): Likewise. Abstract code to
priority-queue_upgrade_task.
(GOMP_taskgroup_start): Use priority queues.
(GOMP_taskgroup_end): Likewise. Handle target nowait target tasks
specially. If taskgroup is NULL, and thr->ts.level is 0, act as a
barrier.
* taskloop.c (GOMP_taskloop): Handle priorities.
* team.c (gomp_new_team): Call priority_queue_init.
(free_team): Call priority_queue_free.
(gomp_free_thread): Call gomp_team_end if thr->ts.team is artificial
team created for target nowait in implicit parallel region.
(gomp_team_start): For nested check, test thr->ts.level instead of
thr->ts.team != NULL.
* testsuite/libgomp.c/doacross-3.c: New test.
* testsuite/libgomp.c/ordered-5.c: New test.
* testsuite/libgomp.c/priority.c: New test.
* testsuite/libgomp.c/target-31.c: New test.
* testsuite/libgomp.c/target-32.c: New test.
* testsuite/libgomp.c/target-33.c: New test.
* testsuite/libgomp.c/target-34.c: New test.
liboffloadmic/
2015-11-14 Ilya Verbin <ilya.verbin@intel.com>
* runtime/offload_host.cpp (task_completion_callback): New
variable.
(offload_proxy_task_completed_ooo): Call task_completion_callback.
(__offload_register_task_callback): New function.
* runtime/offload_host.h (__offload_register_task_callback): New
declaration.
* plugin/libgomp-plugin-intelmic.cpp (offload): Add async_data
argument, handle async offloading.
(register_main_image): Call register_main_image.
(GOMP_OFFLOAD_init_device, get_target_table, GOMP_OFFLOAD_alloc,
GOMP_OFFLOAD_free, GOMP_OFFLOAD_host2dev, GOMP_OFFLOAD_dev2host,
GOMP_OFFLOAD_dev2dev) Adjust offload callers.
(GOMP_OFFLOAD_async_run): New function.
(GOMP_OFFLOAD_run): Implement using GOMP_OFFLOAD_async_run.
kargl [Sat, 14 Nov 2015 17:31:16 +0000 (17:31 +0000)]
2015-11-14 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/67803
* array.c (gfc_match_array_constructor): If array constructor included
a CHARACTER typespec, check array elements for compatible type.
2015-11-14 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/67803
* gfortran.dg/pr67803.f90: New test.
redi [Sat, 14 Nov 2015 17:24:42 +0000 (17:24 +0000)]
Fix std::wstring capacity test for short wchar_t
* testsuite/21_strings/basic_string/capacity/char/18654.cc: Use
real minimum capacity.
* testsuite/21_strings/basic_string/capacity/wchar_t/18654.cc:
Likewise.
projects / thirdparty / gcc.git / log
