platform/upstream/gcc.git
14 years agographite-dependences.c (reduction_dr_1): Remove wrong assert: reduction BBs can have...
Sebastian Pop [Wed, 25 Nov 2009 05:15:47 +0000 (05:15 +0000)]
graphite-dependences.c (reduction_dr_1): Remove wrong assert: reduction BBs can have multiple reduction writes to memory.

2009-10-17  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-dependences.c (reduction_dr_1): Remove wrong
assert: reduction BBs can have multiple reduction writes
to memory.

From-SVN: r154606

14 years agointerchange-3.f90: New.
Sebastian Pop [Wed, 25 Nov 2009 05:15:18 +0000 (05:15 +0000)]
interchange-3.f90: New.

2009-10-17  Sebastian Pop  <sebastian.pop@amd.com>

* gfortran.dg/graphite/interchange-3.f90: New.

From-SVN: r154605

14 years agointerchange-1.c: Un-XFAILed.
Sebastian Pop [Wed, 25 Nov 2009 05:14:51 +0000 (05:14 +0000)]
interchange-1.c: Un-XFAILed.

2009-10-17  Sebastian Pop  <sebastian.pop@amd.com>

* gcc.dg/graphite/interchange-1.c: Un-XFAILed.

From-SVN: r154604

14 years agographite-dependences.c (dependence_polyhedron_1): Add the context to the dependence...
Sebastian Pop [Wed, 25 Nov 2009 05:14:24 +0000 (05:14 +0000)]
graphite-dependences.c (dependence_polyhedron_1): Add the context to the dependence polyhedron.

2009-10-17  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-dependences.c (dependence_polyhedron_1): Add the
context to the dependence polyhedron.

From-SVN: r154603

14 years agographite-poly.c (pbb_remove_duplicate_pdrs): Initialize PBB_PDR_DUPLICATES_REMOVED.
Sebastian Pop [Wed, 25 Nov 2009 05:13:57 +0000 (05:13 +0000)]
graphite-poly.c (pbb_remove_duplicate_pdrs): Initialize PBB_PDR_DUPLICATES_REMOVED.

2009-10-17  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.c (pbb_remove_duplicate_pdrs): Initialize
PBB_PDR_DUPLICATES_REMOVED.
(new_poly_bb): Same.

From-SVN: r154602

14 years agographite-sese-to-poly.c (find_scop_parameters): Initialize SCOP_CONTEXT.
Sebastian Pop [Wed, 25 Nov 2009 05:13:30 +0000 (05:13 +0000)]
graphite-sese-to-poly.c (find_scop_parameters): Initialize SCOP_CONTEXT.

2009-10-17  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (find_scop_parameters): Initialize
SCOP_CONTEXT.
(build_loop_iteration_domains): Extract bounds on parameters
based on the data size.
(build_scop_context): Do not initialize SCOP_CONTEXT;
add new constraints.

From-SVN: r154601

14 years agographite-ppl.c (ppl_min_for_le_polyhedron): New.
Sebastian Pop [Wed, 25 Nov 2009 05:13:03 +0000 (05:13 +0000)]
graphite-ppl.c (ppl_min_for_le_polyhedron): New.

2009-10-17  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-ppl.c (ppl_min_for_le_polyhedron): New.
* graphite-ppl.h (graphite-ppl.h): Declared.

From-SVN: r154600

14 years agographite-ppl.c (ppl_print_linear_expr): New.
Sebastian Pop [Wed, 25 Nov 2009 05:12:35 +0000 (05:12 +0000)]
graphite-ppl.c (ppl_print_linear_expr): New.

2009-10-17  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-ppl.c (ppl_print_linear_expr): New.
(debug_ppl_linear_expr): New.
* graphite-ppl.h (ppl_print_linear_expr): Declared.
(debug_ppl_linear_expr): Declared.

From-SVN: r154599

14 years agographite-interchange.c: Fix comments.
Sebastian Pop [Wed, 25 Nov 2009 05:12:07 +0000 (05:12 +0000)]
graphite-interchange.c: Fix comments.

2009-10-16  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-interchange.c: Fix comments.
* gcc.dg/graphite/block-1.c: Fix comments.

From-SVN: r154598

14 years agointerchange-7.c: Un-XFAILed.
Sebastian Pop [Wed, 25 Nov 2009 05:11:38 +0000 (05:11 +0000)]
interchange-7.c: Un-XFAILed.

2009-10-16  Sebastian Pop  <sebastian.pop@amd.com>

* gcc.dg/graphite/interchange-7.c: Un-XFAILed.
* gcc.dg/graphite/interchange-9.c: Un-XFAILed.

From-SVN: r154597

14 years agoEmpty patch, as it already is applied to trunk.
Sebastian Pop [Wed, 25 Nov 2009 05:11:10 +0000 (05:11 +0000)]
Empty patch, as it already is applied to trunk.

From-SVN: r154596

14 years agographite-interchange.c (memory_stride_in_loop): The depth argument corresponds to...
Sebastian Pop [Wed, 25 Nov 2009 05:10:43 +0000 (05:10 +0000)]
graphite-interchange.c (memory_stride_in_loop): The depth argument corresponds to a dynamic time dimension...

2009-10-15  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-interchange.c (memory_stride_in_loop): The depth
argument corresponds to a dynamic time dimension, so use
psct_dynamic_dim to refer to the transformed scattering dimension.

From-SVN: r154595

14 years agographite-interchange.c (memory_stride_in_loop): Fix comments.
Sebastian Pop [Wed, 25 Nov 2009 05:10:14 +0000 (05:10 +0000)]
graphite-interchange.c (memory_stride_in_loop): Fix comments.

2009-10-15  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-interchange.c (memory_stride_in_loop): Fix comments.

From-SVN: r154594

14 years agographite-blocking.c (lst_do_strip_mine): Avoid strip mining the root of the LST.
Sebastian Pop [Wed, 25 Nov 2009 05:09:36 +0000 (05:09 +0000)]
graphite-blocking.c (lst_do_strip_mine): Avoid strip mining the root of the LST.

2009-10-15  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-blocking.c (lst_do_strip_mine): Avoid strip mining the
root of the LST.
* graphite-interchange.c (lst_do_interchange): Avoid interchanging
the root of the LST.
* graphite-poly.c (scop_to_lst): Fix LST sequence in an outermost
fake loop.
(print_lst): Print the root of LST in a different format.
* graphite-poly.h (lst_depth): Adjust to include the root of the LST.

From-SVN: r154593

14 years agographite-poly.c (print_scop): Fix pretty printing of a SCoP.
Sebastian Pop [Wed, 25 Nov 2009 05:09:07 +0000 (05:09 +0000)]
graphite-poly.c (print_scop): Fix pretty printing of a SCoP.

2009-10-15  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.c (print_scop): Fix pretty printing of a SCoP.

From-SVN: r154592

14 years agopasses.texi: Document the Graphite infrastructure.
Sebastian Pop [Wed, 25 Nov 2009 05:08:37 +0000 (05:08 +0000)]
passes.texi: Document the Graphite infrastructure.

2009-10-15  Sebastian Pop  <sebastian.pop@amd.com>

* doc/passes.texi: Document the Graphite infrastructure.

From-SVN: r154591

14 years agore PR middle-end/41193 (slow compilation with graphite / ice with graphite)
Sebastian Pop [Wed, 25 Nov 2009 05:08:09 +0000 (05:08 +0000)]
re PR middle-end/41193 (slow compilation with graphite / ice with graphite)

2009-10-15  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (insert_copyin): Use gsi_insert_seq_on_edge.
(rewrite_commutative_reductions_out_of_ssa): Call gsi_commit_edge_inserts,
and check the SSA representation.

* gfortran.dg/graphite/id-18.f90: New, reduced from PR41193.

From-SVN: r154590

14 years agopr40157.c: Use -floop-block as reported in the PR.
Sebastian Pop [Wed, 25 Nov 2009 05:07:43 +0000 (05:07 +0000)]
pr40157.c: Use -floop-block as reported in the PR.

2009-10-15  Sebastian Pop  <sebastian.pop@amd.com>

* gcc.dg/graphite/pr40157.c: Use -floop-block as reported in the PR.

From-SVN: r154589

14 years agointerchange-11.c: New.
Sebastian Pop [Wed, 25 Nov 2009 05:07:16 +0000 (05:07 +0000)]
interchange-11.c: New.

2009-10-14  Sebastian Pop  <sebastian.pop@amd.com>

* gcc.dg/graphite/interchange-11.c: New.

From-SVN: r154588

14 years agointerchange-8.c: Enable loop interchange.
Sebastian Pop [Wed, 25 Nov 2009 05:06:49 +0000 (05:06 +0000)]
interchange-8.c: Enable loop interchange.

2009-10-14  Sebastian Pop  <sebastian.pop@amd.com>

* gcc.dg/graphite/interchange-8.c: Enable loop interchange.

From-SVN: r154587

14 years agographite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK): Add -fno-loop-strip-mine and -fno-loop...
Sebastian Pop [Wed, 25 Nov 2009 05:06:21 +0000 (05:06 +0000)]
graphite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK): Add -fno-loop-strip-mine and -fno-loop-interchange.

2009-10-14  Sebastian Pop  <sebastian.pop@amd.com>

* gcc.dg/graphite/graphite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK):
Add -fno-loop-strip-mine and -fno-loop-interchange.
(DEFAULT_FLAGS_GRAPHITE_INTERCHANGE): Add -fno-loop-block and
-fno-loop-strip-mine.
* gfortran.dg/graphite/graphite.exp: Same.

From-SVN: r154586

14 years agographite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK): Renamed DEFAULT_FLAGS_GRAPHITE_BLOCK.
Sebastian Pop [Wed, 25 Nov 2009 05:05:53 +0000 (05:05 +0000)]
graphite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK): Renamed DEFAULT_FLAGS_GRAPHITE_BLOCK.

2009-10-14  Sebastian Pop  <sebastian.pop@amd.com>

* gcc.dg/graphite/graphite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK):
Renamed DEFAULT_FLAGS_GRAPHITE_BLOCK.
(DEFAULT_CFLAGS_GRAPHITE_IDENTITY): Renamed
DEFAULT_FLAGS_GRAPHITE_IDENTITY.
(DEFAULT_CFLAGS_GRAPHITE_INTERCHANGE): Renamed
DEFAULT_FLAGS_GRAPHITE_INTERCHANGE.
(DEFAULT_CFLAGS_GRAPHITE_SCOP): Renamed
DEFAULT_FLAGS_GRAPHITE_SCOP.
(DEFAULT_CFLAGS_RUN_ID): Renamed DEFAULT_FLAGS_RUN_ID.
(DEFAULT_CFLAGS_GRAPHITE): Renamed DEFAULT_FLAGS_GRAPHITE.

From-SVN: r154585

14 years agographite-poly.c (pbb_number_of_iterations_at_time): Cleanup comments.
Sebastian Pop [Wed, 25 Nov 2009 05:05:22 +0000 (05:05 +0000)]
graphite-poly.c (pbb_number_of_iterations_at_time): Cleanup comments.

2009-10-14  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.c (pbb_number_of_iterations_at_time): Cleanup comments.

From-SVN: r154584

14 years agographite-interchange.c (build_linearized_memory_access): Pass an offset as parameter.
Konrad Trifunovic [Wed, 25 Nov 2009 05:04:53 +0000 (05:04 +0000)]
graphite-interchange.c (build_linearized_memory_access): Pass an offset as parameter.

2009-10-14  Konrad Trifunovic  <konrad.trifunovic@gmail.com>
    Sebastian Pop  <sebastian.pop@amd.com>

* graphite-interchange.c (build_linearized_memory_access): Pass an
offset as parameter.
(memory_stride_in_loop): Include transform scattering in the
access functions to be able to querry the strides in the
transformed loops.
(pbb_interchange_loop_depths): Interchange the psct_dynamic_dim,
not the psct_iterator_dim.

Co-Authored-By: Sebastian Pop <sebastian.pop@amd.com>
From-SVN: r154583

14 years agographite-blocking.c (pbb_do_strip_mine): Removed.
Sebastian Pop [Wed, 25 Nov 2009 05:04:26 +0000 (05:04 +0000)]
graphite-blocking.c (pbb_do_strip_mine): Removed.

2009-10-14  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-blocking.c (pbb_do_strip_mine): Removed.
(lst_do_strip_mine_loop): New.
(lst_do_strip_mine): New.
(scop_do_strip_mine): Call lst_do_strip_mine.
* graphite-poly.h (lst_add_loop_under_loop): New.
(lst_find_first_pbb): New.

From-SVN: r154582

14 years agographite-poly.c (loop_to_lst): Fix LST creation.
Sebastian Pop [Wed, 25 Nov 2009 05:03:58 +0000 (05:03 +0000)]
graphite-poly.c (loop_to_lst): Fix LST creation.

2009-10-14  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.c (loop_to_lst): Fix LST creation.

From-SVN: r154581

14 years agoid-15.c: (8 * 8) replaced with sizeof(unsigned long) to make it run on 32bit systems.
Tobias Grosser [Wed, 25 Nov 2009 05:03:27 +0000 (05:03 +0000)]
id-15.c: (8 * 8) replaced with sizeof(unsigned long) to make it run on 32bit systems.

2009-10-14  Tobias Grosser  <grosser@fim.uni-passau.de>

* testsuite/gcc.dg/graphite/id-15.c: (8 * 8) replaced with
sizeof(unsigned long) to make it run on 32bit systems.

From-SVN: r154580

14 years ago* testsuite/gcc.dg/graphite/id-15.c: Adjusted.
Sebastian Pop [Wed, 25 Nov 2009 05:02:58 +0000 (05:02 +0000)]
* testsuite/gcc.dg/graphite/id-15.c: Adjusted.

From-SVN: r154579

14 years agorun-id-1.c: Adjusted size of data to fit on 32-bit machines.
Sebastian Pop [Wed, 25 Nov 2009 05:02:31 +0000 (05:02 +0000)]
run-id-1.c: Adjusted size of data to fit on 32-bit machines.

* testsuite/gcc.dg/graphite/run-id-1.c: Adjusted size of data
to fit on 32-bit machines.

From-SVN: r154578

14 years agographite-sese-to-poly.c (write_alias_graph_to_ascii_dimacs): Fix Comment.
Ramakrishna Upadrasta [Wed, 25 Nov 2009 05:02:04 +0000 (06:02 +0100)]
graphite-sese-to-poly.c (write_alias_graph_to_ascii_dimacs): Fix Comment.

2009-10-14  Ramakrishna Upadrasta <Ramakrishna.Upadrasta@inria.fr>

* graphite-sese-to-poly.c (write_alias_graph_to_ascii_dimacs): Fix Comment.
(write_alias_graph_to_ascii_dot): New.
(write_alias_graph_to_ascii_ecc): Ditto.
(partition_drs_to_sets): Add testing of optimality of current method which
assigns alias numbers according to DFS Comopnent number.
used as heuristic for the upcoming ECC algorithm.
(build_scop_drs): Write to file also with the ecc and dot format.

From-SVN: r154577

14 years agointerchange-1.f: XFail.
Sebastian Pop [Wed, 25 Nov 2009 05:01:36 +0000 (05:01 +0000)]
interchange-1.f: XFail.

2009-10-13  Sebastian Pop  <sebastian.pop@amd.com>

* gfortran.dg/graphite/interchange-1.f: XFail.

From-SVN: r154576

14 years agographite.c (graphite_initialize): Do not handle functions with more than 100 basic...
Sebastian Pop [Wed, 25 Nov 2009 05:01:06 +0000 (05:01 +0000)]
graphite.c (graphite_initialize): Do not handle functions with more than 100 basic blocks.

2009-10-13  Sebastian Pop  <sebastian.pop@amd.com>

* graphite.c (graphite_initialize): Do not handle functions with
more than 100 basic blocks.

From-SVN: r154575

14 years agographite-blocking.c (pbb_do_strip_mine): Use PARAM_LOOP_BLOCK_TILE_SIZE.
Sebastian Pop [Wed, 25 Nov 2009 05:00:38 +0000 (05:00 +0000)]
graphite-blocking.c (pbb_do_strip_mine): Use PARAM_LOOP_BLOCK_TILE_SIZE.

2009-10-13  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-blocking.c (pbb_do_strip_mine): Use
PARAM_LOOP_BLOCK_TILE_SIZE.
* params.def (PARAM_LOOP_BLOCK_TILE_SIZE): Declared.

From-SVN: r154574

14 years agographite-poly.c (dot_lst_1): New.
Sebastian Pop [Wed, 25 Nov 2009 05:00:08 +0000 (05:00 +0000)]
graphite-poly.c (dot_lst_1): New.

2009-10-13  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.c (dot_lst_1): New.
(dot_lst): New.
* graphite-poly.h (dot_lst): Declared.

From-SVN: r154573

14 years agographite-dependences.c (dot_deps_stmt_1): New.
Sebastian Pop [Wed, 25 Nov 2009 04:59:39 +0000 (04:59 +0000)]
graphite-dependences.c (dot_deps_stmt_1): New.

2009-10-13  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-dependences.c (dot_deps_stmt_1): New.
(dot_deps_stmt): New.
* graphite-dependences.h (dot_deps_stmt): Declared.

From-SVN: r154572

14 years agoChangeLog.graphite: New.
Sebastian Pop [Wed, 25 Nov 2009 04:59:10 +0000 (04:59 +0000)]
ChangeLog.graphite: New.

2009-10-09  Sebastian Pop  <sebastian.pop@amd.com>

* ChangeLog.graphite: New.
* testsuite/libgomp.graphite/graphite.exp (PARALLEL_CFLAGS): Add
-fno-loop-block.

From-SVN: r154571

14 years agographite-sese-to-poly.c (rewrite_cross_bb_scalar_dependence): Do not replace cross...
Sebastian Pop [Wed, 25 Nov 2009 04:58:45 +0000 (04:58 +0000)]
graphite-sese-to-poly.c (rewrite_cross_bb_scalar_dependence): Do not replace cross BB scalar dependences ending on PHI nodes.

2009-10-09  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (rewrite_cross_bb_scalar_dependence): Do
not replace cross BB scalar dependences ending on PHI nodes.
(rewrite_cross_bb_scalar_deps): Filter out GIMPLE_PHI nodes.

From-SVN: r154570

14 years agographite-poly.c (apply_poly_transforms): Implement the high level driver for flag_loo...
Sebastian Pop [Wed, 25 Nov 2009 04:58:16 +0000 (04:58 +0000)]
graphite-poly.c (apply_poly_transforms): Implement the high level driver for flag_loop_block.

2009-10-09  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.c (apply_poly_transforms): Implement the high
level driver for flag_loop_block.
* tree-ssa-loop.c (gate_graphite_transforms): Remove call to sorry for
flag_loop_block.

* gcc.dg/graphite/graphite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK): Add
-floop-block.
* gfortran.dg/graphite/graphite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK):
Add -floop-block.

From-SVN: r154569

14 years agographite-poly.h (lst_find_pbb): New.
Sebastian Pop [Wed, 25 Nov 2009 04:57:45 +0000 (04:57 +0000)]
graphite-poly.h (lst_find_pbb): New.

2009-10-09  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.h (lst_find_pbb): New.
(find_lst_loop): New.

From-SVN: r154568

14 years agographite-poly.c (new_scop): Remove init of SCOP_DEP_GRAPH.
Sebastian Pop [Wed, 25 Nov 2009 04:57:17 +0000 (04:57 +0000)]
graphite-poly.c (new_scop): Remove init of SCOP_DEP_GRAPH.

2009-10-09  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.c (new_scop): Remove init of SCOP_DEP_GRAPH.
* graphite-poly.h (struct scop): Remove dep_graph field.
(SCOP_DEP_GRAPH): Removed.

From-SVN: r154567

14 years agographite-poly.h (copy_lst): Do full copy of LST.
Sebastian Pop [Wed, 25 Nov 2009 04:56:49 +0000 (04:56 +0000)]
graphite-poly.h (copy_lst): Do full copy of LST.

2009-10-09  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.h (copy_lst): Do full copy of LST.

From-SVN: r154566

14 years agographite-sese-to-poly.c (reduction_phi_p): Call remove_invariant_phi when the loop...
Sebastian Pop [Wed, 25 Nov 2009 04:56:21 +0000 (04:56 +0000)]
graphite-sese-to-poly.c (reduction_phi_p): Call remove_invariant_phi when the loop stride is zero.

2009-10-07  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (reduction_phi_p): Call remove_invariant_phi
when the loop stride is zero.

* gcc.dg/graphite/id-16.c: New.

From-SVN: r154565

14 years agographite-dependences.c (reduction_dr_1): New.
Sebastian Pop [Wed, 25 Nov 2009 04:55:53 +0000 (04:55 +0000)]
graphite-dependences.c (reduction_dr_1): New.

2009-10-06  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-dependences.c (reduction_dr_1): New.
(reduction_dr_p): New.
(graphite_legal_transform_dr): Call reduction_dr_p.
(reduction_ddr): Renamed reduction_ddr_p.
* graphite-poly.h (same_pdr_p): New.
(number_of_write_pdrs): New.
* graphite-sese-to-poly.c (nb_data_writes_in_bb): New.
(split_reduction_stmt): Do not split reduction statements
when there are no writes to memory.
(translate_scalar_reduction_to_array_for_stmt): Insert the
memory reduction statement just after the scalar reduction statement.

* gcc.dg/graphite/interchange-10.c: Updated to differ from interchange-4.c.
Un-XFAIL-ed.
* gcc.dg/graphite/interchange-3.c: Un-XFAIL-ed.
* gcc.dg/graphite/interchange-4.c: Un-XFAIL-ed.

From-SVN: r154564

14 years agographite-interchange.c (lst_apply_interchange): New.
Sebastian Pop [Wed, 25 Nov 2009 04:55:25 +0000 (04:55 +0000)]
graphite-interchange.c (lst_apply_interchange): New.

2009-10-06  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-interchange.c (lst_apply_interchange): New.
(lst_interchange_profitable_p): New.
(lst_try_interchange_loops): New.
(lst_try_interchange): New.
(lst_do_interchange): New.
(pbb_do_interchange): Removed.
(scop_do_interchange): Call lst_do_interchange.

From-SVN: r154563

14 years agographite-poly.c (print_scop): Print SCOP_ORIGINAL_SCHEDULE and SCOP_TRANSFORMED_SCHEDULE.
Sebastian Pop [Wed, 25 Nov 2009 04:54:59 +0000 (04:54 +0000)]
graphite-poly.c (print_scop): Print SCOP_ORIGINAL_SCHEDULE and SCOP_TRANSFORMED_SCHEDULE.

2009-10-06  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-poly.c (print_scop): Print SCOP_ORIGINAL_SCHEDULE and
SCOP_TRANSFORMED_SCHEDULE.
(loop_to_lst): New.
(scop_to_lst): New.
(print_lst): New.
(debug_lst): New.
* graphite-poly.h (lst_p): New.
(struct lst): New.
(LST_LOOP_P): New.
(LST_LOOP_FATHER): New.
(LST_PBB): New.
(LST_SEQ): New.
(scop_to_lst): Declared.
(print_lst): Declared.
(debug_lst): Declared.
(new_lst_loop): New.
(new_lst_stmt): New.
(copy_lst): New.
(lst_depth): New.
(lst_dewey_number): New.
(struct scop): Add original_schedule and transformed_schedule fields.
(SCOP_ORIGINAL_SCHEDULE): New.
(SCOP_TRANSFORMED_SCHEDULE): New.
* graphite-sese-to-poly.c (build_poly_scop): Call scop_to_lst.

From-SVN: r154562

14 years agographite-dependences.c (reduction_ddr): New.
Sebastian Pop [Wed, 25 Nov 2009 04:54:30 +0000 (04:54 +0000)]
graphite-dependences.c (reduction_ddr): New.

2009-10-05  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-dependences.c (reduction_ddr): New.
(graphite_legal_transform_bb): Call reduction_ddr.
* graphite-poly.c (new_poly_bb): Pass a new bool parameter.
Initialize PBB_IS_REDUCTION.
* graphite-poly.h (struct poly_bb): New bool field is_reduction.
(PBB_IS_REDUCTION): New.
(new_poly_bb): Update declaration.
* graphite-scop-detection.h (build_scop_bbs): Removed.
(nb_reductions_in_loop): Removed.
* graphite-sese-to-poly.c (try_generate_gimple_bb): Pass a sbitmap
parameter for reductions.
(try_generate_gimple_bb): Update call to new_poly_bb.
(build_scop_bbs_1): Same.
(build_scop_bbs): Same.
(gsi_for_phi_node): New.
(scalar_close_phi_node_p): Remove gcc_assert.
(split_reduction_stmt): New.
(is_reduction_operation_p): New.
(phi_contains_arg): New.
(follow_ssa_with_commutative_ops): New.
(detect_commutative_reduction_arg): New.
(detect_commutative_reduction_assign): New.
(follow_inital_value_to_phi): New.
(edge_initial_value_for_loop_phi): New.
(initial_value_for_loop_phi): New.
(detect_commutative_reduction): New.
(translate_scalar_reduction_to_array_for_stmt): New.
(insert_copyout): New.
(insert_copyin): New.
(translate_scalar_reduction_to_array): New.
(rewrite_commutative_reductions_out_of_ssa_close_phi): New.
(rewrite_commutative_reductions_out_of_ssa_loop): New.
(rewrite_commutative_reductions_out_of_ssa): New.
(build_poly_scop): Call rewrite_commutative_reductions_out_of_ssa.
* sese.h (split_region_for_bb): New.

* gcc.dg/graphite/graphite.exp (DEFAULT_CFLAGS_GRAPHITE_BLOCK): Add -ffast-math.
* gcc.dg/graphite/interchange-1.c: Fix format.
* gcc.dg/graphite/interchange-10.c: New.

From-SVN: r154561

14 years agographite-sese-to-poly.c (insert_out_of_ssa_copy): Remove buggy assert.
Sebastian Pop [Wed, 25 Nov 2009 04:54:00 +0000 (04:54 +0000)]
graphite-sese-to-poly.c (insert_out_of_ssa_copy): Remove buggy assert.

2009-10-05  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (insert_out_of_ssa_copy): Remove
buggy assert.

From-SVN: r154560

14 years agographite-sese-to-poly.c (scev_analyzable_p): New.
Sebastian Pop [Wed, 25 Nov 2009 04:53:33 +0000 (04:53 +0000)]
graphite-sese-to-poly.c (scev_analyzable_p): New.

2009-10-05  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (scev_analyzable_p): New.
(rewrite_cross_bb_scalar_dependence): New.
(rewrite_cross_bb_scalar_deps): New.
(rewrite_reductions_out_of_ssa): Use bb_in_sese_p.
Call rewrite_cross_bb_scalar_deps.

From-SVN: r154559

14 years agographite-sese-to-poly.c (gsi_for_ssa_name_def): Removed.
Sebastian Pop [Wed, 25 Nov 2009 04:53:05 +0000 (04:53 +0000)]
graphite-sese-to-poly.c (gsi_for_ssa_name_def): Removed.

2009-10-05  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (gsi_for_ssa_name_def): Removed.
(insert_out_of_ssa_copy): Directly use gsi_after_labels and
gsi_for_stmt.

From-SVN: r154558

14 years agographite-clast-to-gimple.c (build_cloog_prog): Use pbb_index.
Sebastian Pop [Wed, 25 Nov 2009 04:52:38 +0000 (04:52 +0000)]
graphite-clast-to-gimple.c (build_cloog_prog): Use pbb_index.

2009-10-05  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-clast-to-gimple.c (build_cloog_prog): Use pbb_index.
* graphite-interchange.c (pbb_do_interchange): Same.
* graphite-poly.c (print_scattering_function): Same.
(debug_pdrs): Same.
* graphite-poly.h (pbb_loop): New.
* graphite-sese-to-poly.c (create_linear_expr_from_tree): Use pbb_loop.

From-SVN: r154557

14 years agographite-scop-detection.c (limit_scops): Do not build poly_bbs.
Sebastian Pop [Wed, 25 Nov 2009 04:52:08 +0000 (04:52 +0000)]
graphite-scop-detection.c (limit_scops): Do not build poly_bbs.

2009-09-27  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-scop-detection.c (limit_scops): Do not build poly_bbs.

From-SVN: r154556

14 years agographite-sese-to-poly.c (partition_drs_to_sets): Drs is not modified, so don't pass...
Sebastian Pop [Wed, 25 Nov 2009 04:51:41 +0000 (04:51 +0000)]
graphite-sese-to-poly.c (partition_drs_to_sets): Drs is not modified, so don't pass a pointer to it.

2009-09-27  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (partition_drs_to_sets): Drs is not modified,
so don't pass a pointer to it.
(build_alias_set_for_drs): Same.
(build_base_obj_set_for_drs): Same.
(build_scop_drs): Same.

From-SVN: r154555

14 years agographite-sese-to-poly.c (build_scop_drs): Disable writing to file the alias graph...
Li Feng [Wed, 25 Nov 2009 04:51:14 +0000 (04:51 +0000)]
graphite-sese-to-poly.c (build_scop_drs): Disable writing to file the alias graph of data references.

2009-09-26  Li Feng  <nemokingdom@gmail.com>

* graphite-sese-to-poly.c (build_scop_drs): Disable writing to file
the alias graph of data references.

From-SVN: r154554

14 years agographite-sese-to-poly.c (write_alias_graph_to_ascii_dimacs): New.
Li Feng [Wed, 25 Nov 2009 04:50:45 +0000 (04:50 +0000)]
graphite-sese-to-poly.c (write_alias_graph_to_ascii_dimacs): New.

2009-09-25  Li Feng  <nemokingdom@gmail.com>

* graphite-sese-to-poly.c (write_alias_graph_to_ascii_dimacs): New.
(build_scop_drs): When debugging, write the alias graph to file,
otherwise, should be disabled.

From-SVN: r154553

14 years agographite-scop-detection.c (stmt_simple_memref_p): Removed.
Sebastian Pop [Wed, 25 Nov 2009 04:50:18 +0000 (04:50 +0000)]
graphite-scop-detection.c (stmt_simple_memref_p): Removed.

2009-09-17  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-scop-detection.c (stmt_simple_memref_p): Removed.
(is_simple_operand): Remove call to stmt_simple_memref_p.
(stmt_simple_for_scop_p): Update call to is_simple_operand.

From-SVN: r154552

14 years agointerchange-mvt.c: New File.
David Edelsohn [Wed, 25 Nov 2009 04:49:50 +0000 (04:49 +0000)]
interchange-mvt.c: New File.

2009-09-17  David Edelsohn  <edelsohn@gnu.org>

* testsuite/gcc.dg/graphite/interchange-mvt.c: New File.

From-SVN: r154551

14 years agoscop-dsyrk.c: New File.
David Edelsohn [Wed, 25 Nov 2009 04:49:22 +0000 (04:49 +0000)]
scop-dsyrk.c: New File.

2009-09-17  David Edelsohn  <edelsohn@gnu.org>

* testsuite/gcc.dg/graphite/scop-dsyrk.c: New File.
* testsuite/gcc.dg/graphite/scop-dsyr2k.c: New File.
* testsuite/gcc.dg/graphite/scop-mvt.c: New File.
* testsuite/gcc.dg/graphite/scop-sor.c: New File.

From-SVN: r154550

14 years agore PR tree-optimization/41118 (Wrong dependence analysis in graphite for unrestricted...
Li Feng [Wed, 25 Nov 2009 04:48:51 +0000 (04:48 +0000)]
re PR tree-optimization/41118 (Wrong dependence analysis in graphite for unrestricted pointers)

2009-09-17 Li Feng  <nemokingdom@gmail.com>

PR middle-end/41118
* graphite-dependences.c (poly_drs_may_alias_p): Adjust definition.
(pddr_original_scattering): Make sure 2 pdr2 in the same base object set.
(graphite_carried_dependence_level_k): Ditto.
* graphite-poly.c (new_poly_dr): Add init of PDR_BASE_OBJECT_SET.
* graphite-poly.h (struct poly_dr): Add member dr_base_object_set.
(new_poly_dr): Adjust declaration.
* graphite-sese-to-poly.c (free_data_refs_aux): New.
(free_gimple_bb): Added free_data_refs_aux.
(build_poly_dr): Add dr_base_object_set.
(partition_drs_to_sets): New.
(dr_same_base_object_p): New.
(build_alias_set_for_drs): New.
(build_base_object_set_for_drs): New.
(build_scop_drs): Add build_base_obj_set_for_drs.
* graphite-sese-to-poly.h: Added #define for alias set number index and
base object set index.
* libgomp/testsuite/libgomp.graphite/force-parallel-6.c: Refine tests.
* libgomp/testsuite/libgomp.graphite/pr4118.c: New.

From-SVN: r154549

14 years agoEmpty patch as it has already been committed to trunk.
Sebastian Pop [Wed, 25 Nov 2009 04:48:22 +0000 (04:48 +0000)]
Empty patch as it has already been committed to trunk.

From-SVN: r154548

14 years agographite-sese-to-poly.c (struct irp_data): Removed.
Sebastian Pop [Wed, 25 Nov 2009 04:47:54 +0000 (04:47 +0000)]
graphite-sese-to-poly.c (struct irp_data): Removed.

2009-09-14  Sebastian Pop  <sebastian.pop@amd.com>

* graphite-sese-to-poly.c (struct irp_data): Removed.
(idx_record_params): Removed.
(find_params_in_bb): Scan for parameters the access functions; don't
parse the tree representation.

From-SVN: r154547

14 years ago* functions.texi: Rebuild.
Ben Elliston [Wed, 25 Nov 2009 04:45:26 +0000 (04:45 +0000)]
* functions.texi: Rebuild.

From-SVN: r154546

14 years agoREADME: Mention changes to Makefile.in and functions.texi.
Manuel López-Ibáñez [Wed, 25 Nov 2009 04:34:10 +0000 (15:34 +1100)]
README: Mention changes to Makefile.in and functions.texi.

* README: Mention changes to Makefile.in and functions.texi.
* gather-docs: Mention 'make stamp-functions' in the header.

Co-Authored-By: Ben Elliston <bje@au.ibm.com>
From-SVN: r154545

14 years agoEmpty patch as it was reverted.
Sebastian Pop [Wed, 25 Nov 2009 04:34:05 +0000 (04:34 +0000)]
Empty patch as it was reverted.

From-SVN: r154544

14 years agoEmpty patch as it was reverted.
Sebastian Pop [Wed, 25 Nov 2009 04:33:35 +0000 (04:33 +0000)]
Empty patch as it was reverted.

From-SVN: r154543

14 years agoPartially revert the previous patch, except the following.
Sebastian Pop [Wed, 25 Nov 2009 04:33:03 +0000 (04:33 +0000)]
Partially revert the previous patch, except the following.

2009-09-02  Sebastian Pop  <sebastian.pop@amd.com>

Partially revert the previous patch, except the following.
* tree-scalar-evolution.c (instantiate_scev_not): Adapted to pass
as parameters the operands of the not expression.

From-SVN: r154542

14 years agotree-scalar-evolution.c (instantiate_scev_assign): New.
Sebastian Pop [Wed, 25 Nov 2009 04:32:35 +0000 (04:32 +0000)]
tree-scalar-evolution.c (instantiate_scev_assign): New.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_assign): New.
Do not call analyze_scalar_evolution on assignments.
(instantiate_scev_phi): Call analyze_scalar_evolution.
(instantiate_scev_name): Call instantiate_scev_assign and
instantiate_scev_phi.
(instantiate_scev_not): Adapted to pass as parameters the operands
of the not expression.

From-SVN: r154541

14 years agotree-scalar-evolution.c (instantiate_scev_binary): Adapted to pass as parameters...
Sebastian Pop [Wed, 25 Nov 2009 04:32:07 +0000 (04:32 +0000)]
tree-scalar-evolution.c (instantiate_scev_binary): Adapted to pass as parameters the operands of the binary expression.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_binary): Adapted
to pass as parameters the operands of the binary expression.

From-SVN: r154540

14 years agotree-scalar-evolution.c (instantiate_scev_bitnot): Renamed instantiate_scev_not.
Sebastian Pop [Wed, 25 Nov 2009 04:31:40 +0000 (04:31 +0000)]
tree-scalar-evolution.c (instantiate_scev_bitnot): Renamed instantiate_scev_not.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_bitnot): Renamed
instantiate_scev_not.  Handle NEGATE_EXPR.
(instantiate_scev_r): Handle NEGATE_EXPR.

From-SVN: r154539

14 years agotree-scalar-evolution.c (instantiate_scev_1): Renamed instantiate_scev_r.
Sebastian Pop [Wed, 25 Nov 2009 04:31:10 +0000 (04:31 +0000)]
tree-scalar-evolution.c (instantiate_scev_1): Renamed instantiate_scev_r.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_1): Renamed
instantiate_scev_r.
(instantiate_scev_r): Move code in instantiate_scev_2
and instantiate_scev_1.

From-SVN: r154538

14 years agotree-scalar-evolution.c (instantiate_scev_3): New.
Sebastian Pop [Wed, 25 Nov 2009 04:30:41 +0000 (04:30 +0000)]
tree-scalar-evolution.c (instantiate_scev_3): New.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_3): New.
(instantiate_scev_1): Move code in instantiate_scev_3.

From-SVN: r154537

14 years agotree-scalar-evolution.c (instantiate_scev_poly): New.
Sebastian Pop [Wed, 25 Nov 2009 04:30:10 +0000 (04:30 +0000)]
tree-scalar-evolution.c (instantiate_scev_poly): New.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_poly): New.
(instantiate_scev_1): Move code in instantiate_scev_poly.

From-SVN: r154536

14 years agotree-scalar-evolution.c (instantiate_scev_bitnot): New.
Sebastian Pop [Wed, 25 Nov 2009 04:29:43 +0000 (04:29 +0000)]
tree-scalar-evolution.c (instantiate_scev_bitnot): New.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_bitnot): New.
(instantiate_scev_1): Move code in instantiate_scev_bitnot.

From-SVN: r154535

14 years agotree-scalar-evolution.c (instantiate_scev_convert): New.
Sebastian Pop [Wed, 25 Nov 2009 04:29:12 +0000 (04:29 +0000)]
tree-scalar-evolution.c (instantiate_scev_convert): New.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_convert): New.
(instantiate_scev_1): Move code in instantiate_scev_convert.

From-SVN: r154534

14 years agotree-scalar-evolution.c (instantiate_scev_binary): New.
Sebastian Pop [Wed, 25 Nov 2009 04:28:44 +0000 (04:28 +0000)]
tree-scalar-evolution.c (instantiate_scev_binary): New.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_binary): New.
(instantiate_scev_1): Move code in instantiate_scev_binary.

From-SVN: r154533

14 years agotree-scalar-evolution.c (instantiate_scev_name): New.
Sebastian Pop [Wed, 25 Nov 2009 04:28:15 +0000 (04:28 +0000)]
tree-scalar-evolution.c (instantiate_scev_name): New.

2009-09-01  Sebastian Pop  <sebastian.pop@amd.com>

* tree-scalar-evolution.c (instantiate_scev_name): New.
(instantiate_scev_1): Move code in instantiate_scev_name.

From-SVN: r154532

14 years agore PR fortran/42008 (Wrongly rejected derived types with default initializers in...
Jerry DeLisle [Wed, 25 Nov 2009 02:42:22 +0000 (02:42 +0000)]
re PR fortran/42008 (Wrongly rejected derived types with default initializers in PURE procedures)

2009-11-24  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR fortran/42008
* decl.c (variable_decl): Do not error on initialization within a
derived type specification of a pure procedure.

From-SVN: r154531

14 years agore PR fortran/42008 (Wrongly rejected derived types with default initializers in...
Jerry DeLisle [Wed, 25 Nov 2009 02:41:20 +0000 (02:41 +0000)]
re PR fortran/42008 (Wrongly rejected derived types with default initializers in PURE procedures)

2009-11-24  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR fortran/42008
* gfortran.dg/pure_initializer_2.f90: New test.

From-SVN: r154530

14 years agore PR fortran/41807 (data statement with nested type constructors)
Jerry DeLisle [Wed, 25 Nov 2009 02:37:57 +0000 (02:37 +0000)]
re PR fortran/41807 (data statement with nested type constructors)

2009-11-24  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

PR fortran/41807
* decl.c (variable_decl): Do not error on initialization within a
derived type specification of a pure procedure.

From-SVN: r154529

14 years agopa.c (output_call): Only use sr4 for long interspace calls if call binds local and...
John David Anglin [Wed, 25 Nov 2009 00:31:11 +0000 (00:31 +0000)]
pa.c (output_call): Only use sr4 for long interspace calls if call binds local and generating non...

* pa.c (output_call): Only use sr4 for long interspace calls if
call binds local and generating non PIC code.
(attr_length_call): Adjust length calculation for above.

From-SVN: r154527

14 years agoDaily bump.
GCC Administrator [Wed, 25 Nov 2009 00:16:44 +0000 (00:16 +0000)]
Daily bump.

From-SVN: r154526

14 years agoi386.md (plusminus_carry_mnemonic): New code attribute.
Uros Bizjak [Tue, 24 Nov 2009 23:06:47 +0000 (00:06 +0100)]
i386.md (plusminus_carry_mnemonic): New code attribute.

* config/i386/i386.md (plusminus_carry_mnemonic): New code attribute.
(add<mode>3_carry): Change operand order of plus RTX.
(*add<mode>3_carry): Ditto.
(*addsi3_carry_zext): Ditto.
(*add<dwi>3_doubleword): Update for changed operand order.
(*negti2_1 splitter): Ditto.
(*negdi2_1 splitter): Ditto.
(<plusminus_insn><mode>3_carry): Macroize expander from
{add,sub}<mode>3_carry patterns using plusminus code iterator.
(*<plusminus_insn><mode>3_carry): Macroize insn from
*{add,sub}<mode>3_carry patetrns using plusminus code iterator.
(neg<mode>2): Macroize expander from neg{qi,hi,si,di,ti}2
patterns using SDWIM mode iterator.
(*neg<dwi>2_doubleword): New insn_and_split pattern.  Macroize
pattern from *neg{di,ti}2_1 patterns and corresponding splitters
using DWIH mode iterator.
(*neg<mode>2_1): Macroize insn from neg{qi,hi,si,di}2_1 patterns
using SWI mode iterator.
(*neg<mode>2_cmpz): Ditto from neg{qi,hi,si,di}2_cmpz patterns.
(one_cmpl<mode>2): Macroize expander from one_cmpl{qi,hi,si,di}2
patterns using SWIM mode iterator.
(*one_cmpl<mode>2_1): Macroize insn from one_cmpl{hi,si,di}2_1
patterns using SWI248 mode iterator.
(*one_cmpl<mode>2_2): Macroize insn from one_cmpl{qi,hi,si,di}2_2
patterns using SWI mode iterator.
(*one_cmpl<mode>2_2 splitter): Macroize splitter from
one_cmpl{qi,hi,si,di}2_2 splitter patterns using SWI mode iterator.

From-SVN: r154522

14 years ago* zlib.m4: New file.
Joel Brobecker [Tue, 24 Nov 2009 22:40:46 +0000 (22:40 +0000)]
* zlib.m4: New file.

From-SVN: r154521

14 years agore PR c++/42137 (error: expected constructor, destructor, or type conversion before...
Jason Merrill [Tue, 24 Nov 2009 20:55:24 +0000 (15:55 -0500)]
re PR c++/42137 (error: expected constructor, destructor, or type conversion before ‘{’ token)

PR c++/42137
* parser.c (cp_parser_mem_initializer_id): Pass typename_type to
cp_parser_class_name.
(cp_parser_unqualified_id): Same, rather than class_type.

PR c++/11764
* parser.c (cp_parser_expression_statement): Give helpful error
for constructor name used as type.

From-SVN: r154519

14 years agopt.c (determine_specialization): Give helpful error about missing "template<>".
Jason Merrill [Tue, 24 Nov 2009 20:55:17 +0000 (15:55 -0500)]
pt.c (determine_specialization): Give helpful error about missing "template<>".

* pt.c (determine_specialization): Give helpful error about missing
"template<>".

From-SVN: r154518

14 years agosem_util.adb (Set_Debug_Info_Needed): For an E_Class_Wide_Subtype, also set the flag...
Eric Botcazou [Tue, 24 Nov 2009 20:25:58 +0000 (20:25 +0000)]
sem_util.adb (Set_Debug_Info_Needed): For an E_Class_Wide_Subtype, also set the flag on the Equivalent_Type.

* sem_util.adb (Set_Debug_Info_Needed): For an E_Class_Wide_Subtype,
also set the flag on the Equivalent_Type.
* gcc-interface/utils.c (finish_record_type): Replace DO_NOT_FINALIZE
parameter with DEBUG_INFO_P.  Rename FIELDLIST into FIELD_LIST.
(rest_of_record_type_compilation): Rename FIELDLIST into FIELD_LIST.
(build_vms_descriptor32): Adjust call to finish_record_type.
(build_vms_descriptor): Likewise.
(build_unc_object_type): Likewise.
* decl.c (gnat_to_gnu_entity): Adjust calls to finish_record_type and
components_to_record.
(make_packable_type): Adjust call to finish_record_type.
(maybe_pad_type): Likewise.  Tweak condition.
(components_to_record): Likewise.  Replace DO_NOT_FINALIZE parameter
with MAYBE_UNUSED.  Adjust recursive call.
(create_variant_part_from): Adjust call to finish_record_type.  Do not
call rest_of_record_type_compilation on the new record types.
* trans.c (gigi): Adjust call to finish_record_type.
* gigi.h (finish_record_type): Adjust prototype and comment.
(rest_of_record_type_compilation): Adjust comment.

From-SVN: r154515

14 years agoexp_util.adb (Make_CW_Equivalent_Type): Do not mark the type as frozen for targets...
Eric Botcazou [Tue, 24 Nov 2009 20:02:40 +0000 (20:02 +0000)]
exp_util.adb (Make_CW_Equivalent_Type): Do not mark the type as frozen for targets that do not require front-end layout.

* exp_util.adb (Make_CW_Equivalent_Type): Do not mark the type as
frozen for targets that do not require front-end layout.
(New_Class_Wide_Subtype): Always reset the freezing status to False.
* exp_ch8.adb: Do not 'with' Targparm.
(Expand_N_Object_Renaming_Declaration): Always freeze a class-wide
subtype that has been built from the expression.
* exp_intr.adb (Expand_Unc_Deallocation): If the designated type is
class wide, freeze the implicit type that has been built from the
expression at the dereference point.
* freeze.adb (Freeze_Entity): Adjust comment.
* gcc-interface/decl.c (Gigi_Equivalent_Type) <E_Class_Wide_Type>:
Remove useless test.
* gcc-interface/trans.c (process_freeze_entity): Do not special-case
class-wide subtypes.

* s-osinte-aix.adb (clock_gettime): Fix comment.
* s-osinte-darwin.adb (clock_gettime): Likewise.

From-SVN: r154514

14 years agoaltivec-32.c: Don't use -m64.
Janis Johnson [Tue, 24 Nov 2009 19:09:26 +0000 (19:09 +0000)]
altivec-32.c: Don't use -m64.

* gcc.target/powerpc/altivec-32.c: Don't use -m64.
* gcc.target/powerpc/const-compare.c: Require lp64, don't use -m64.
* gcc.target/powerpc/darwin-bool-1.c: Require ilp32, remove dg-skip-if.
* gcc.target/powerpc/paired-1.c: Require ilp32, don't use -m32.
* gcc.target/powerpc/paired-2.c: Ditto.
* gcc.target/powerpc/paired-3.c: Ditto.
* gcc.target/powerpc/paired-4.c: Ditto.
* gcc.target/powerpc/paired-5.c: Ditto.
* gcc.target/powerpc/paired-6.c: Ditto.
* gcc.target/powerpc/paired-7.c: Ditto.
* gcc.target/powerpc/paired-8.c: Ditto.
* gcc.target/powerpc/paired-9.c: Ditto.
* gcc.target/powerpc/paired-10.c: Ditto.
* gcc.target/powerpc/popcount-2.c: Don't use -m32.
* gcc.target/powerpc/popcount-3.c: Don't use -m64.
* gcc.target/powerpc/ppc-paired.c: Require ilp32, don't use -m32.
* gcc.target/powerpc/pr39457.c: Don't use -m64.
* gcc.target/powerpc/vsx-vector-1.c: Ditto.
* gcc.target/powerpc/vsx-vector-2.c: Ditto.
* gcc.target/powerpc/vsx-vector-3.c: Ditto.
* gcc.target/powerpc/vsx-vector-4.c: Ditto.

From-SVN: r154512

14 years ago[multiple changes]
Paolo Carlini [Tue, 24 Nov 2009 16:53:17 +0000 (16:53 +0000)]
[multiple changes]

2009-11-24  Michael Matz  <matz@suse.de>

* expr.c (set_storage_via_libcall): Fix build_call_expr call.

2009-11-24  David Binderman  <dcb314@hotmail.com>

* expr.c (store_field): Remove set but not used local variable
width_mask.
(expand_expr_real_2): Remove treeop2.
* gcse.c (update_ld_motion_stores): Remove new_rtx.
* haifa-sched.c (max_issue): Remove points.
(sched_create_recovery_edges): Remove e.
* ira-costs.c (setup_allocno_cover_class_and_costs): Remove mode.

From-SVN: r154505

14 years agov850.c (function_arg): Fix handling of zero-length function arguments.
Nick Clifton [Tue, 24 Nov 2009 15:38:39 +0000 (15:38 +0000)]
v850.c (function_arg): Fix handling of zero-length function arguments.

        * config/v850/v850.c (function_arg): Fix handling of zero-length
        function arguments.

From-SVN: r154503

14 years agotree.h (union tree_ann_d): Don't declare.
Michael Matz [Tue, 24 Nov 2009 15:37:32 +0000 (15:37 +0000)]
tree.h (union tree_ann_d): Don't declare.

* tree.h (union tree_ann_d): Don't declare.
(tree_base): Remove ann field.
(struct var_ann_d): Declare forward.
(tree_result_decl, tree_parm_decl, tree_var_decl): Add ann field.
(DECL_VAR_ANN_PTR): New macro.
* tree-dfa.c (create_var_ann): Rewrite in terms of above macro,
accept only VAR, PARM or RESULT decls.
(create_tree_common_ann): Remove.
(remove_referenced_var): Use DECL_VAR_ANN_PTR.
* tree-eh.c (lookup_expr_eh_lp): Remove.
* tree-flow-inline.h (var_ann, get_var_ann): Rewrite in terms of
DECL_VAR_ANN_PTR.
(ann_type, tree_common_ann, get_tree_common_ann): Remove.
* tree-flow.h (enum tree_ann_type): Remove.
(struct tree_ann_common_d): Remove.
(struct var_ann_d): Remove common field.
(union tree_ann_d): Remove.
(tree_ann_t, tree_ann_common_t): Remove typedefs.
(tree_common_ann, get_tree_common_ann, ann_type,
create_tree_common_ann, lookup_expr_eh_lp): Don't declare.
* tree-ssa.c (delete_tree_ssa): Use DECL_VAR_ANN_PTR.
* tree.c (copy_node_stat): Use DECL_VAR_ANN_PTR.
* builtins.c (expand_builtin_memcpy): Use
currently_expanding_gimple_stmt instead of tree annotation.
(expand_builtin_memset_args): Ditto.
* cfgexpand.c (currently_expanding_gimple_stmt): Add global variable.
(expand_call_stmt): Don't set tree annotation.
(expand_gimple_basic_block): Set currently_expanding_gimple_stmt.
* expr.c (expand_expr_real): Don't call lookup_expr_eh_lp.
* gimple.h (currently_expanding_gimple_stmt): Declare.

From-SVN: r154502

14 years agodwarf2out.c (add_location_or_const_value_attribute): Remove useless NOTE_VAR_LOCATION...
Paolo Carlini [Tue, 24 Nov 2009 15:30:01 +0000 (15:30 +0000)]
dwarf2out.c (add_location_or_const_value_attribute): Remove useless NOTE_VAR_LOCATION_STATUS call.

2009-11-24  Paolo Carlini  <paolo.carlini@oracle.com>

* dwarf2out.c (add_location_or_const_value_attribute): Remove
useless NOTE_VAR_LOCATION_STATUS call.
* cfgrtl.c (commit_one_edge_insertion): Remove useless loop, tidy.

From-SVN: r154501

14 years agolto-wrapper.c (lto_wrapper_exit): Don't try to delete files if being called recursively.
Rafael Avila de Espindola [Tue, 24 Nov 2009 15:00:15 +0000 (15:00 +0000)]
lto-wrapper.c (lto_wrapper_exit): Don't try to delete files if being called recursively.

2009-11-24  Rafael Avila de Espindola  <espindola@google.com>

       * lto-wrapper.c (lto_wrapper_exit): Don't try to delete files if
       being called recursively.

From-SVN: r154500

14 years agoMakefile.in (PLUGIN_HEADERS): Added files...
Basile Starynkevitch [Tue, 24 Nov 2009 14:43:23 +0000 (14:43 +0000)]
Makefile.in (PLUGIN_HEADERS): Added files...

2009-11-24  Basile Starynkevitch  <basile@starynkevitch.net>

        * gcc/Makefile.in (PLUGIN_HEADERS): Added files: cppdefault.h flags.h
$(MD5_H) params.def params.h prefix.h tree-inline.h.

--This lifilne, and those below, will be ignored--

M    gcc/ChangeLog
M    gcc/Makefile.in

From-SVN: r154499

14 years agoarm.c (arm_compute_save_reg0_reg12_mask): Fix comment typo.
Paul Brook [Tue, 24 Nov 2009 14:04:50 +0000 (14:04 +0000)]
arm.c (arm_compute_save_reg0_reg12_mask): Fix comment typo.

2009-11-24  Paul Brook  <paul@codesourcery.com>

gcc/
* gcc/config/arm/arm.c (arm_compute_save_reg0_reg12_mask): Fix comment
typo.

From-SVN: r154498

14 years agoarm.c (TARGET_DWARF_REGISTER_SPAN): Define.
Julian Brown [Tue, 24 Nov 2009 13:35:26 +0000 (13:35 +0000)]
arm.c (TARGET_DWARF_REGISTER_SPAN): Define.

gcc/
* config/arm/arm.c (TARGET_DWARF_REGISTER_SPAN): Define.
(arm_dwarf_register_span): New function.
(arm_dbx_register_number): Add VFPv3 dwarf numbering.

Co-Authored-By: Paul Brook <paul@codesourcery.com>
From-SVN: r154497

14 years agocfgrtl.c (commit_one_edge_insertion): Remove set but not used local variable bb_note.
David Binderman [Tue, 24 Nov 2009 12:59:00 +0000 (12:59 +0000)]
cfgrtl.c (commit_one_edge_insertion): Remove set but not used local variable bb_note.

2009-11-24  David Binderman <dcb314@hotmail.com>

* cfgrtl.c (commit_one_edge_insertion): Remove set but not used
local variable bb_note.
        * dominance.c (get_dominated_by): Likewise local var n.
        (output_file_names): Likewise local var idx.
        (add_location_or_const_value_attribute): Likewise local var status.
        * dwarf2out.c (gen_variable_die): Likewise local var field.
        * emit-rtl.c (no_line_numbers): Remove.
(init_emit_once): Remove line_numbers parameter.
        * rtl.h (init_emit_once): Adjust prototype.
        * toplev.c (backend_init): Adjust init_emit_once call.

From-SVN: r154496

14 years agoanonname1.C: Escape backslash in scan-assembler-not regular expression.
Ulrich Weigand [Tue, 24 Nov 2009 12:55:53 +0000 (12:55 +0000)]
anonname1.C: Escape backslash in scan-assembler-not regular expression.

* g++.dg/debug/dwarf2/anonname1.C: Escape backslash in
scan-assembler-not regular expression.

From-SVN: r154495

14 years agore PR tree-optimization/42142 (DCE miscompiles a certain quicksort implementation...
Richard Guenther [Tue, 24 Nov 2009 11:57:26 +0000 (11:57 +0000)]
re PR tree-optimization/42142 (DCE miscompiles a certain quicksort implementation when optimizing with -O1 or higher)

2009-11-24  Richard Guenther  <rguenther@suse.de>

PR tree-optimization/42142
* tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1):
Handle iv-dependent (non-)kills properly.
(mark_aliased_reaching_defs_necessary): Pass the basic-block
of the reference statement to mark_aliased_reaching_defs_necessary_1.

* gcc.c-torture/execute/pr42142.c: New testcase.

From-SVN: r154494

14 years agore PR tree-optimization/42154 (Wrong code from (early) SRA)
Martin Jambor [Tue, 24 Nov 2009 10:56:14 +0000 (11:56 +0100)]
re PR tree-optimization/42154 (Wrong code from (early) SRA)

2009-11-24  Martin Jambor  <mjambor@suse.cz>

PR tree-optimization/42154
* tree-sra.c (struct access): Added comments.
(sra_modify_expr): Build references to the old aggregate with
build_ref_for_offset instead of reusing access->expr.
(load_assign_lhs_subreplacements): Likewise.

* testsuite/gcc.c-torture/execute/pr42154.c: New test.

From-SVN: r154493

14 years agore PR fortran/42045 ([F03] passing a procedure pointer component to a procedure point...
Janus Weil [Tue, 24 Nov 2009 08:16:32 +0000 (09:16 +0100)]
re PR fortran/42045 ([F03] passing a procedure pointer component to a procedure pointer dummy)

2009-11-24  Janus Weil  <janus@gcc.gnu.org>

PR fortran/42045
* resolve.c (resolve_actual_arglist): Make sure procedure pointer
actual arguments are resolved correctly.
(resolve_function): An EXPR_FUNCTION which is a procedure pointer
component, has already been resolved.
(resolve_fl_derived): Procedure pointer components should not be
implicitly typed.

2009-11-24  Janus Weil  <janus@gcc.gnu.org>

PR fortran/42045
* gfortran.dg/proc_ptr_comp_2.f90: Correct invalid test case.
* gfortran.dg/proc_ptr_comp_3.f90: Extended test case.
* gfortran.dg/proc_ptr_comp_24.f90: New.

From-SVN: r154492