re PR target/65614 (PowerPC VSX systems should use XSCPSGNDP to copy scalar fp data to/from Altivec registers)
[gcc]
2015-04-06 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/65614
* config/rs6000/rs6000.c (struct processor_costs): Add cost field
for SF->DF conversions to make FLOAT_EXTEND more expensive, so
that LFD is used to load double constants instead of LFS. Add
defaults for all costs structures. Add comments for missing
initialization fields.
(size32_cost): Likewise.
(size64_cost): Likewise.
(rs64a_cost): Likewise.
(mpccore_cost): Likewise.
(ppc403_cost): Likewise.
(ppc405_cost): Likewise.
(ppc440_cost): Likewise.
(ppc476_cost): Likewise.
(ppc601_cost): Likewise.
(ppc603_cost): Likewise.
(ppc604_cost): Likewise.
(ppc604e_cost): Likewise.
(ppc620_cost): Likewise.
(ppc630_cost): Likewise.
(ppccell_cost): Likewise.
(ppc750_cost): Likewise.
(ppc7450_cost): Likewise.
(ppc8540_cost): Likewise.
(ppce300c2c3_cost): Likewise.
(ppce500mc_cost): Likewise.
(ppce500mc64_cost): Likewise.
(ppce5500_cost): Likewise.
(ppce6500_cost): Likewise.
(titan_cost): Likewise.
(power4_cost): Likewise.
(power6_cost): Likewise.
(power7_cost): Likewise.
(power8_cost): Likewise.
(ppca2_cost): Likewise.
(rs6000_rtx_costs): Make FLOAT_EXTEND use SFDF_convert field.
* config/rs6000/rs6000.md (extendsfdf2_fpr): Generate XSCPSGNDP
instead of XXLOR to copy SFmode to clear out dirty bits created
when SFmode denormals are generated.
(mov<mode>_hardfloat, FMOVE32 case): Likewise.
(truncdfsf2_fpr): Add support for ISA 2.07 XSRSP instruction.
[gcc/testsuite]
2015-04-06 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/65614
* gcc.target/powerpc/compress-float-ppc-pic.c: Run test on power5
to get floating point compression.
* gcc.target/powerpc/compress-foat-ppc.c: Likewise.
gcc/
* config/i386/intelmic-mkoffload.c (generate_host_descr_file): Call
GOMP_offload_unregister from the destructor.
libgomp/
* libgomp-plugin.h (struct mapping_table): Replace with addr_pair.
* libgomp.h (struct gomp_memory_mapping): Remove.
(struct target_mem_desc): Change type of mem_map from
gomp_memory_mapping * to splay_tree_s *.
(struct gomp_device_descr): Remove register_image_func, get_table_func.
Add load_image_func, unload_image_func.
Change type of mem_map from gomp_memory_mapping to splay_tree_s.
Remove offload_regions_registered.
(gomp_init_tables): Remove.
(gomp_free_memmap): Change type of argument from gomp_memory_mapping *
to splay_tree_s *.
* libgomp.map (GOMP_4.0.1): Add GOMP_offload_unregister.
* oacc-host.c (host_dispatch): Do not initialize register_image_func,
get_table_func, mem_map.is_initialized, mem_map.splay_tree.root,
offload_regions_registered.
Initialize load_image_func, unload_image_func, mem_map.root.
(goacc_host_init): Do not initialize host_dispatch.mem_map.lock.
* oacc-init.c (lazy_open): Don't call gomp_init_tables.
(acc_shutdown_1): Use dev's lock and splay_tree instead of mem_map's.
* oacc-mem.c (lookup_host): Get gomp_device_descr *dev instead of
gomp_memory_mapping *. Use dev's lock and splay_tree.
(lookup_dev): Use dev's lock.
(acc_deviceptr): Pass dev to lookup_host instead of mem_map.
(acc_is_present): Likewise.
(acc_map_data): Likewise.
(acc_unmap_data): Likewise. Use dev's lock.
(present_create_copy): Likewise.
(delete_copyout): Pass dev to lookup_host instead of mem_map.
(update_dev_host): Likewise.
(gomp_acc_remove_pointer): Likewise. Use dev's lock.
* oacc-parallel.c (GOACC_parallel): Use dev's lock and splay_tree.
* plugin/plugin-host.c (GOMP_OFFLOAD_register_image): Remove.
(GOMP_OFFLOAD_get_table): Remove
(GOMP_OFFLOAD_load_image): New function.
(GOMP_OFFLOAD_unload_image): New function.
* target.c (register_lock): New mutex for offload image registration.
(num_devices): Do not guard with PLUGIN_SUPPORT.
(gomp_realloc_unlock): New static function.
(gomp_map_vars_existing): Add device descriptor argument. Unlock mutex
before gomp_fatal.
(gomp_map_vars): Use dev's lock and splay_tree instead of mem_map's.
Pass devicep to gomp_map_vars_existing. Unlock mutex before gomp_fatal.
(gomp_copy_from_async): Use dev's lock and splay_tree instead of
mem_map's.
(gomp_unmap_vars): Likewise.
(gomp_update): Remove gomp_memory_mapping argument. Use dev's lock and
splay_tree instead of mm's. Unlock mutex before gomp_fatal.
(gomp_offload_image_to_device): New static function.
(GOMP_offload_register): Add mutex lock.
Call gomp_offload_image_to_device for all initialized devices.
Replace gomp_realloc with gomp_realloc_unlock.
(GOMP_offload_unregister): New function.
(gomp_init_tables): Replace with gomp_init_device. Replace a call to
get_table_func from the plugin with calls to init_device_func and
gomp_offload_image_to_device.
(gomp_free_memmap): Change type of argument from gomp_memory_mapping *
to splay_tree_s *.
(GOMP_target): Do not call gomp_init_tables. Use dev's lock and
splay_tree instead of mem_map's. Unlock mutex before gomp_fatal.
(GOMP_target_data): Do not call gomp_init_tables.
(GOMP_target_update): Likewise. Remove argument from gomp_update.
(gomp_load_plugin_for_device): Replace register_image and get_table
with load_image and unload_image in DLSYM ().
(gomp_register_images_for_device): Remove function.
(gomp_target_init): Do not initialize current_device.mem_map.*,
current_device.offload_regions_registered.
Remove call to gomp_register_images_for_device.
Do not free offload_images and num_offload_images.
liboffloadmic/
* plugin/libgomp-plugin-intelmic.cpp: Include map.
(AddrVect, DevAddrVect, ImgDevAddrMap): New typedefs.
(num_devices, num_images, address_table): New static vars.
(num_libraries, lib_descrs): Remove static vars.
(set_mic_lib_path): Rename to ...
(init): ... this. Allocate address_table and get num_devices.
(GOMP_OFFLOAD_get_num_devices): return num_devices.
(load_lib_and_get_table): Remove static function.
(offload_image): New static function.
(GOMP_OFFLOAD_get_table): Remove function.
(GOMP_OFFLOAD_load_image, GOMP_OFFLOAD_unload_image): New functions.
Martin Liska [Sun, 5 Apr 2015 17:17:29 +0000 (19:17 +0200)]
re PR middle-end/65665 (g++.dg/torture/pr64378.C -O2 -flto -fno-use-linker-plugin -flto-partition=none)
Fix PR ipa/65665
PR ipa/65665
* ipa-icf.c (sem_function::equals_wpa): Verify that IPA CP
has computed data structure.
(sem_item_optimizer::update_hash_by_addr_refs): Likewise.
Richard Biener [Sat, 4 Apr 2015 10:47:08 +0000 (10:47 +0000)]
re PR tree-optimization/64909 (Missed vectorization with bdver1)
2015-04-04 Richard Biener <rguenther@suse.de>
PR tree-optimization/64909
PR tree-optimization/65660
* tree-vectorizer.h (vect_get_known_peeling_cost): Adjust
to take a cost vector for scalar iteration cost.
(vect_get_single_scalar_iteration_cost): Likewise.
* tree-vect-loop.c (vect_get_single_scalar_iteration_cost):
Compute the scalar iteration cost into a cost vector.
(vect_get_known_peeling_cost): Use the scalar cost vector to
account for the cost of the peeled iterations.
(vect_estimate_min_profitable_iters): Likewise.
* tree-vect-data-refs.c (vect_peeling_hash_get_lowest_cost):
Likewise.
Jan Hubicka [Fri, 3 Apr 2015 18:19:53 +0000 (20:19 +0200)]
re PR ipa/65655 (ICE in speculative_call_info, at cgraph.c:1151)
PR ipa/65655
* ipa-inline-analysis.c (edge_set_predicate): Do not redirect
speculative indirect edges to avoid ordering issue.
* g++.dg/torture/pr65655.C: New testcase.
* config/i386/sync.md (UNSPEC_MOVA): Remove.
(atomic_load<mode>): Change operand 0 predicate to
nonimmediate_operand and fix up the destination when needed.
Use UNSPEC_LDA.
(atomic_loaddi_fpu): Use UNSPEC_LDA.
(atomic_store<mode>): Change operand 1 predicate to
nonimmendate_operand and move the source to register when needed.
Use UNSPEC_STA.
(atomic_store<mode>_1): Use UNSPEC_STA.
(atomic_storedi_fpu): Change operand 1 to nonimmediate_operand.
Fix moves from memory operand. Use UNSPEC_STA.
* expmed.c (strict_volatile_bitfield_p): Check that the access will
not cross a MODESIZE boundary.
(store_bit_field, extract_bit_field): Added assertions in the
strict volatile bitfields code path.
Alex Velenko [Tue, 31 Mar 2015 19:00:52 +0000 (19:00 +0000)]
pr45701-1.c (history_expand_line_internal): Add an extra variable to force stack alignment.
2015-03-31 Alex Velenko <Alex.Velenko@arm.com>
* gcc.target/arm/pr45701-1.c (history_expand_line_internal): Add an
extra variable to force stack alignment.
* gcc.target/arm/pr45701-2.c (history_expand_line_internal): Add an
extra variable to force stack alignment.
Ian Lance Taylor [Tue, 31 Mar 2015 17:54:07 +0000 (17:54 +0000)]
go/cmd/go: always link external test packages first
When linking complex packages that use both internal and external tests as well as many dependencies it is critical that the link order be external test package, internal test package, everything else.
This change is a back (forward?) port of the same change that canonical have been maintaining on their fork of the go tool for gccgo. Now that gccgo uses the go tool from upstream, this patch should be applied both to the gofrontend and golang/go repos.
Uros Bizjak [Tue, 31 Mar 2015 17:06:37 +0000 (19:06 +0200)]
re PR target/58945 (Improve atomic_compare_and_swap*_doubleword pattern)
PR target/58945
* config/i386/sync.md (atomic_compare_and_swap<dwi>_doubleword):
Do not split operands 0 and operands 2 to halfmode.
(atomic_compare_and_swap<mode>): Update for
atomic_compare_and_swap<dwi>_doubleword changes.
Jan Hubicka [Tue, 31 Mar 2015 17:03:16 +0000 (19:03 +0200)]
tree.c (need_assembler_name_p): Artificial types have no ODR names.
* tree.c (need_assembler_name_p): Artificial types have no ODR
names.
* ipa-devirt.c (warn_odr): Do not try to apply ODR cache when
no caching is done.
* lto.c (lto_read_decls): Move code registering odr types out
of TYPE_CANONICAL conditional and also register polymorphic types.
Martin Liska [Tue, 31 Mar 2015 16:03:03 +0000 (18:03 +0200)]
re PR ipa/65557 (ICE: SIGSEGV in hash_table<>::find_slot_with_hash() with -fdevirtualize -fipa-cp -fipa-icf-functions)
Fix PR65557.
* ipa-icf.c (sem_function::equals_wpa): Check if IPA CP
has already filled up function summary.
(sem_item_optimizer::update_hash_by_addr_refs): Likewise.
* g++.dg/ipa/pr65557.C: New test.
* config/s390/s390.c (s390_function_num_hotpatch_hw): Allow hotpatching
nested functions.
(s390_reorg): Adapt to new signature of s390_function_num_hotpatch_hw.
(s390_asm_output_function_label): Adapt to new signature of
s390_function_num_hotpatch_hw
Optimise the code generating assembler output.
Add comments to assembler file.
Richard Biener [Tue, 31 Mar 2015 09:34:08 +0000 (09:34 +0000)]
re PR c++/65626 (ICE in fixup_noreturn_call called by tree-ssa-forwprop.c:2492)
2015-03-31 Richard Biener <rguenther@suse.de>
PR middle-end/65626
* tree-cfgcleanup.c (fixup_noreturn_call): Only split the block
of the noreturn call so it is last and cleanup_control_flow_bb
can do the CFG part.
Ilya Enkovich [Tue, 31 Mar 2015 08:29:28 +0000 (08:29 +0000)]
re PR target/65531 (ICE: symtab_node::verify failed: Two symbols with same comdat_group are not linked by the same_comdat_group list. with -fcheck-pointer-bounds -mmpx)
gcc/
PR target/65531
* ipa-chkp.c (chkp_maybe_create_clone): Don't set
same_comdat_group for external symbols.
* symtab.c (symtab_node::verify_symtab_nodes): Avoid
infinite same_comdat_group traversal loop.
Jakub Jelinek [Mon, 30 Mar 2015 21:56:02 +0000 (23:56 +0200)]
re PR ipa/65610 (Compare debug failure with -g3 -fsanitize=undefined -fno-sanitize=vptr -O3)
PR ipa/65610
* ipa-utils.h (inlined_polymorphic_ctor_dtor_block_p): Declare.
* ipa-polymorphic-call.c (inlined_polymorphic_ctor_dtor_block_p): New
function.
(decl_maybe_in_construction_p, noncall_stmt_may_be_vtbl_ptr_store):
Use it.
* ipa-prop.c (param_type_may_change_p): Likewise.
* tree-ssa-live.c: Include ipa-utils.h and its dependencies.
(remove_unused_scope_block_p): Add in_ctor_dtor_block
argument. Before inlining, preserve
inlined_polymorphic_ctor_dtor_block_p blocks and the outermost block
with FUNCTION_DECL BLOCK_ABSTRACT_ORIGIN inside of them. Adjust
recursive calls.
(remove_unused_locals): Adjust remove_unused_scope_block_p caller.
Jakub Jelinek [Mon, 30 Mar 2015 17:54:05 +0000 (19:54 +0200)]
re PR middle-end/65597 (ICE in build_outer_var_ref, at omp-low.c:1043)
PR fortran/65597
* trans-openmp.c (gfc_trans_omp_do): For !simple simd with explicit
linear clause for the iterator set OMP_CLAUSE_LINEAR_NO_COPYIN.
For implcitly added !simple OMP_CLAUSE_LINEAR set it too. Use step 1
instead of the original step on the new iterator - count.
* testsuite/libgomp.fortran/pr65597.f90: New test.
Martin Jambor [Sun, 29 Mar 2015 15:38:52 +0000 (17:38 +0200)]
re PR ipa/65478 (crafty performance regression)
PR ipa/65478
* params.def (PARAM_IPA_CP_RECURSION_PENALTY) : New.
(PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
* ipa-prop.h (ipa_node_params): New flags node_within_scc and
node_calling_single_call.
* ipa-cp.c (count_callers): New function.
(set_single_call_flag): Likewise.
(initialize_node_lattices): Count callers and set single_flag_call if
necessary.
(incorporate_penalties): New function.
(good_cloning_opportunity_p): Use it, dump new flags.
(propagate_constants_topo): Set node_within_scc flag if appropriate.
* doc/invoke.texi (ipa-cp-recursion-penalty,
ipa-cp-single-call-pentalty): Document.
Jerry DeLisle [Sat, 28 Mar 2015 13:27:58 +0000 (13:27 +0000)]
re PR fortran/65596 (NAMELIST bug with f2003: reads too far)
2015-03-28 Jerry DeLisle <jvdelisle@gcc.gnu.org>
PR libgfortran/65596
* io/transfer.c (data_transfer_init): If in namelist mode and
delimiter is not specified, set it to DELIM_QUOTE, independent
of -std.
* io/unit.c (init_units): Set flags.delim to the correct
initial value of DELIM_UNSPECIFIED.
Jan Hubicka [Fri, 27 Mar 2015 21:35:51 +0000 (22:35 +0100)]
re PR ipa/65588 (lto1: internal compiler error: Segmentation fault)
PR ipa/65588
* symtab.c (symtab_node::get_partitioning_class): Register vars
are duplicated.
* varpool.c (symbol_table::output_variables) Do not assemble unefined
decls for non-symbols.
H.J. Lu [Fri, 27 Mar 2015 18:11:00 +0000 (18:11 +0000)]
Add default_binds_local_p_2 and use it for x86
Protected data symbol means that it can't be pre-emptied. It doesn't mean
its address won't be external. This is true for pointer to protected
function. With copy relocation, address of protected data defined in the
shared library may also be external. We only know that for sure at
run-time. TARGET_BINDS_LOCAL_P should return false on protected data
symbol.
gcc/
PR target/65248
* output.h (default_binds_local_p_2): New.
* varasm.c (default_binds_local_p_2): Renamed to ...
(default_binds_local_p_3): This. Don't return true on protected
data symbol if protected data may be external.
(default_binds_local_p): Use default_binds_local_p_3.
(default_binds_local_p_1): Likewise.
(default_binds_local_p_2): New.
* config/i386/i386.c (TARGET_BINDS_LOCAL_P): Set to
default_binds_local_p_2 if TARGET_MACHO is undefined.
Jakub Jelinek [Fri, 27 Mar 2015 17:35:14 +0000 (18:35 +0100)]
re PR target/65593 (internal compiler error: in extract_insn, at recog.c:2343)
PR target/65593
* config/i386/i386.c (legitimize_pic_address): If base
is SYMBOL_REF or LABEL_REF using %rip addressing, force
it to reg to avoid PLUS of SYMBOL_REF/LABEL_REF and register.
Jan Hubicka [Fri, 27 Mar 2015 15:19:35 +0000 (16:19 +0100)]
re PR target/65531 (ICE: symtab_node::verify failed: Two symbols with same comdat_group are not linked by the same_comdat_group list. with -fcheck-pointer-bounds -mmpx)
PR target/65531
* symtab.c (symtab_node::verify_symtab_nodes): Fix verification of
comdat groups.
Jan Hubicka [Fri, 27 Mar 2015 15:13:54 +0000 (16:13 +0100)]
re PR ipa/65600 (bost testsuite failure: ICE: Segmentation fault)
PR ipa/65600
* cgraph.c (cgraph_update_edges_for_call_stmt_node): Fix the case
of optimized out indirect call.
(redirect_to_unreachable): Always build symbol table node for
BUILT_IN_UNREACHABLE
* g++.dg/torture/pr65600.C: New testcase.
projects / thirdparty / gcc.git / log
