These can be generated by vec_perm_const now. We could keep the
patterns around, as technically it may be less work, but this
exercises a code path needed by less primary platforms.
* config/i386/sse.md (VEC_EXTRACT_EVENODD_MODE): Remove.
(vec_extract_even<mode>, vec_extract_odd<mode>): Remove.
From-SVN: r180453
The ones that expand to VPERM can be handled by generic code.
The even v4si and v4sf expanders remain until vector.md can be
updated to not invoke them directly.
* config/rs6000/altivec.md (vec_extract_evenv8hi,
vec_extract_evenv16qi, vec_extract_oddv4si,
vec_extract_oddv4sf): Remove.
From-SVN: r180452
The can_vec_perm_p interface changed to use a C integer array. This
allows easy re-use from the rtl level and the gimple level within
the vectorizer. It allows both to determine if a given permutation
is (un-)supported without having to create tree/rtl garbage.
The expand_vec_perm interface changed to use rtl. This allows easy
re-use from the rtl level, so that expand_vec_perm can be used in the
fallback implementation of other optabs.
* target.def (vec_perm_const_ok): Change parameters to mode and
array of indicies.
* doc/tm.texi: Rebuild.
* config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
parameters to mode and array of indicies.
* expr.c (expand_expr_real_2) [VEC_PERM_EXPR]: Expand operands here.
* optabs.c (can_vec_perm_p): Rename from can_vec_perm_expr_p.
Change parameters to mode and array of indicies.
(expand_vec_perm_1): Rename from expand_vec_perm_expr_1.
(expand_vec_perm): Rename from expand_vec_perm_expr. Change
parameters to mode and rtx inputs. Try lowering to QImode
vec_perm_const before trying fully variable permutation.
* optabs.h: Update decls.
* tree-vect-generic.c (lower_vec_perm): Extract array of indices from
VECTOR_CST to pass to can_vec_perm_p.
* tree-vect-slp.c (vect_get_mask_element): Change mask parameter type
from int pointer to unsigned char pointer.
(vect_transform_slp_perm_load): Update for change to can_vec_perm_p.
* tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
From-SVN: r180449
PR libstdc++/50862
* include/std/condition_variable (condition_variable_any::wait): Fix
deadlock and ensure _Lock::lock() is called on exit.
(condition_variable_any::native_handle): Remove, as per LWG 1500.
* testsuite/30_threads/condition_variable_any/50862.cc: New.
From-SVN: r180446
libcpp/
* include/line-map.h (linemap_expand_location): Take a line table
parameter. Update comment.
(linemap_resolve_location): Update comment.
(linemap_expand_location_full): Remove.
* line-map.c (linemap_resolve_location): Handle reserved
locations; return a NULL map in those cases.
(linemap_expand_location): If location is reserved, return a
zeroed expanded location. Update comment. Take a line table to
assert that the function takes non-virtual locations only.
(linemap_expand_location_full): remove.
(linemap_dump_location): Handle the fact that
linemap_resolve_location can return NULL line maps when the
location resolves to a reserved location.
gcc/
* input.c (expand_location): Rewrite using
linemap_resolve_location and linemap_expand_location. Add a
comment.
From-SVN: r180426
PR tree-optimization/50596
* tree-vect-stmts.c (vect_mark_relevant): Only use
FOR_EACH_IMM_USE_FAST if lhs is SSA_NAME.
(vectorizable_store): If is_pattern_stmt_p look through
VIEW_CONVERT_EXPR on lhs.
* tree-vect-patterns.c (check_bool_pattern, adjust_bool_pattern):
Use unsigned type instead of signed.
(vect_recog_bool_pattern): Optimize also stores into bool memory in
addition to casts from bool to integral types.
(vect_mark_pattern_stmts): If pattern_stmt already has vinfo
created, don't create it again.
* gcc.dg/vect/vect-cond-10.c: New test.
From-SVN: r180424
gcc/testsuite/
* gcc.target/sparc/vis3move-1.c: New test.
* gcc.target/sparc/vis3move-2.c: New test.
* gcc.target/sparc/vis3move-3.c: New test.
From-SVN: r180418
PR libstdc++/49894
* include/std/mutex (__mutex_base,__recursive_mutex_base): Define new
base classes to manage construction/destruction of native mutexes,
using NSDMI when INIT macros are defined.
(mutex,recursive_mutex,timed_mutex,recursive_timed_mutex): Derive from
new base classes.
* include/std/condition_variable (condition_variable): Use NSDMI when
INIT macro is defined. Use noexcept.
* src/condition_variable.cc (condition_variable): Explicitly-default
constructor/destructor when using NSDMI. Use noexcept.
(condition_variable_any): Likewise.
From-SVN: r180411
* config/sparc/sparc.md (cpu_feature, enabled): New attributes.
(*movsi_insn_novis3, *movsi_insn_vis3): Consolidate into one pattern
called *movsi_insn.
(*movdi_insn_sp32_v9_novis3, *movdi_insn_sp32_v9_vis3): Consolidate
into *movdi_insn_sp32.
(*movdi_insn_sp64_novis3, *movdi_insn_sp64_vis3): Consolidate into
one pattern called *movdi_insn_sp64.
(*movsf_insn_novis3, *movsf_insn_vis3, *movsf_insn_no_fpu):
Consolidate into one pattern called *movsf_insn.
(*movdf_insn_sp32_no_fpu, *movdf_insn_sp32_v9_novis3,
*movdf_insn_sp32_v9_vis3, *movdf_insn_sp32_v9_no_fpu): Consolidate
into *movdf_insn_sp32.
(*movdf_insn_sp64_novis3, *movdf_insn_sp64_vis3,
*movdf_insn_sp64_no_fpu): Consolidate into one pattern called
*movdf_insn_sp64.
(*zero_extendsidi2_insn_sp64_novis3,
*zero_extendsidi2_insn_sp64_vis3): Consolidate into one pattern
called *zero_extendsidi2_insn_sp64.
(*sign_extendsidi2_insn_novis3, *sign_extendsidi2_insn_vis3):
Consolidate into one pattern named *sign_extendsidi2_insn.
(*mov<VM32:mode>_insn_novis3, *mov<VM32:mode>_insn_vis3):
Consolidate into one pattern named *mov<VM32:mode>_insn.
(*mov<VM64:mode>_insn_sp64_novis3,
*mov<VM64:mode>_insn_sp64_novis3): Consolidate into one pattern
named *mov<VM64:mode>_insn_sp64.
(*mov<VM64:mode>_insn_sp32_novis3,
*mov<VM64:mode>_insn_sp32_vis3): Consolidate into one pattern
named *mov<VM64:mode>_insn_sp32.
From-SVN: r180407
2011-10-24 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
* tree-ssa-strlen.c (get_string_length): Change assertion to STPCPY.
(zero_length_string): Change assertion to accept strinfo without
length but with stmt instead.
Set the endptr pointer also if starting a new chain.
(adjust_related_strinfos): Ignore strinfos marked for delayed
length computation.
(handle_builtin_strcpy): Mark earlier strinfo elements also for
delayed length computation.
2011-10-24 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
* gcc.dg/strlenopt-22.c: New testcase.
* gcc.dg/strlenopt-4.c: Change scan value for s390(x).
From-SVN: r180402
PR target/50820
Port from 4.6 branch r180379
* doc/invoke.texi (AVR Options): New subsubsection to explain EIND
handling and indirect jump/calls on devices > 128k.
From-SVN: r180388
2011-10-24 Richard Guenther <rguenther@suse.de>
PR tree-optimization/50838
* tree-data-ref.c (dr_analyze_indices): Properly canonicalize
a MEM_REF base if we change it.
* gcc.dg/torture/pr50838.c: New testcase.
From-SVN: r180381
