* tree-data-ref.c (compute_estimated_nb_iterations,
analyze_array_indexes, compute_overlap_steps_for_affine_1_2,
analyze_subscript_affine_affine, find_data_references_in_loop):
Fixed to use chrec_contains_undetermined to test the values of
loop->estimated_nb_iterations.
* tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
Compute the estimation only when loop->estimated_nb_iterations
has not yet been initialized.
(convert_step_widening, scev_probably_wraps_p): Add a call to
estimate_numbers_of_iterations_loop.
* tree-vrp.c (execute_vrp): Don't call estimate_numbers_of_iterations.
From-SVN: r100749
2005-06-07 Uros Bizjak <uros@kss-loka.si>
* config/i386/i386.h (enum ix86_entity): New.
(enum ix86_stack_slot): New.
(OPTIMIZE_MODE_SWITCHING): Redefine to use
ix86_optimize_mode_switching[] array.
(NUM_MODES_FOR_MODE_SWITCHING): Redefine for 4 entities.
(MODE_NEEDED): Use ix86_mode_needed() function.
(EMIT_MODE_SET): Redefine for changed emit_i387_cw_initialization ()
function.
(struct machine_function): Use optimize_mode_switching[] array.
* config/i386/i386.c (MAX_386_STACK_LOCALS): Remove.
(ix86_mode_needed): New function.
(emit_i387_cw_initialization): Cleanup. Use mode to calculate
correct stack positions for stored control words.
(assign_386_stack_local): Use enum ix86_stack_slot. Change assert.
(ix86_expand_builtin) [IX86_BUILTIN_LDMXCSR]: Change constant to
SLOT_TEMP.
[IX86_BUILTIN_STMXCSR]: Same.
* config/i386/i386-protos.h (assign_stack_local): Change prototype.
(emit_i387_cw_initialization): Change prototype.
(ix86_mode_needed): New prototype.
* config/i386/i386.md (i387_cw attribute): Change order of elements.
(truncdfsf2, truncxfsf2, truncxfdf2, fix_trunc<mode>_fisttp_i387_1)
(*fix_trunc<mode>_i387_1, lrint<mode>2, *fist<mode>2_floor_1)
(*fist<mode>2_ceil_1): Change constant in call to
assign_386_stack_local to SLOT_TEMP.
(*fix_trunc<mode>_i387_1): Change constant in call to
assign_386_stack_local to SLOT_CW_STORED and SLOT_CW_TRUNC.
Use new ix86_optimize_mode_switching[] array.
(frndintxf2_floor, *fist<mode>2_floor_1): Change constants in call to
assign_386_stack_local to SLOT_CW_STORED and SLOT_CW_FLOOR.
Use new ix86_optimize_mode_switching[] array.
(frndintxf2_ceil, *fist<mode>2_ceil_1): Change constants in call to
assign_386_stack_local to SLOT_CW_STORED and SLOT_CW_CEIL.
Use new ix86_optimize_mode_switching[] array.
(frndintxf2_trunc): Change constants in call to
assign_386_stack_local to SLOT_CW_STORED and SLOT_CW_TRUNC.
Use new ix86_optimize_mode_switching[] array.
(frndintxf2_mask_pm): Change constants in call to
assign_386_stack_local to SLOT_CW_STORED and SLOT_CW_MASK_PM.
Use new ix86_optimize_mode_switching[] array.
(define_peephole2): Change constant from 17 to FLAGS_REG;
From-SVN: r100732
PR middle-end/21850
* tree.c (get_unwidened): Stop at NOP_EXPR/CONVERT_EXPR that convert
from vector types.
* gcc.c-torture/execute/20050607-1.c: New test.
From-SVN: r100725
* tree-ssa-threadupdate.c (struct thread_stats_d): Declare.
(thread_stats): New local variable.
(redirect_edges): Increment count of threaded edges.
(thread_through_all_blocks): Initialize thread_stats.
Display number of threaded jumps if TDF_STATS is enabled.
From-SVN: r100722
2005-06-07 Thomas Koenig <Thomas.Koenig@online.de>
PR libfortran/21926
* m4/matmul.m4: Correct zeroing of result for non-packed
arrays with lowest stride is one.
* generated/matmul_c4.c: Regenerated.
* generated/matmul_c8.c: Regenerated.
* generated/matmul_i4.c: Regenerated.
* generated/matmul_i8.c: Regenerated.
* generated/matmul_r4.c: Regenerated.
* generated/matmul_r8.c: Regenerated.
2005-06-07 Thomas Koenig <Thomas.Koenig@online.de>
PR libfortran/21926
* gfortran.fortran-torture/execute/intrinsic_matmul.f90:
Test case where the return array has lowest stride one,
but isn't packed.
From-SVN: r100721
* Makefile.in (tree-chrec.o): Depend on CFGLOOP_H and TREE_FLOW_H.
* tree-chrec.c: Include cfgloop.h and tree-flow.h.
(evolution_function_is_invariant_rec_p,
evolution_function_is_invariant_p): New.
(chrec_convert): Use an extra parameter AT_STMT for refining the
information that is passed down to convert_step. Integrate the
code that was in count_ev_in_wider_type.
* tree-chrec.h (count_ev_in_wider_type): Removed.
(chrec_convert): Modify its declaration.
(evolution_function_is_invariant_p): Declared.
(evolution_function_is_affine_p): Use evolution_function_is_invariant_p.
* tree-flow.h (can_count_iv_in_wider_type): Renamed convert_step.
(scev_probably_wraps_p): Declared.
* tree-scalar-evolution.c (count_ev_in_wider_type): Removed.
(follow_ssa_edge_in_rhs, interpret_rhs_modify_expr):
Use an extra parameter AT_STMT for refining the information that is
passed down to convert_step.
(follow_ssa_edge_inner_loop_phi, follow_ssa_edge,
analyze_scalar_evolution_1): Initialize AT_STMT with the current
analyzed statement.
(instantiate_parameters_1): Don't know yet how to initialize AT_STMT.
* tree-ssa-loop-ivopts.c (idx_find_step): Update the use of
can_count_iv_in_wider_type to use convert_step.
* tree-ssa-loop-niter.c (can_count_iv_in_wider_type_bound): Move
code that is independent of the loop over the known iteration
bounds to convert_step_widening, the rest is moved to
proved_non_wrapping_p.
(scev_probably_wraps_p): New.
(can_count_iv_in_wider_type): Renamed convert_step.
* tree-vrp.c (adjust_range_with_scev): Take an extra AT_STMT parameter.
Use scev_probably_wraps_p for computing init_is_max.
(vrp_visit_assignment): Pass the current analyzed statement to
adjust_range_with_scev.
(execute_vrp): Call estimate_numbers_of_iterations for refining the
information provided by scev analyzer.
testsuite:
* testsuite/gcc.dg/vect/vect-77.c: Remove xfail from lp64.
* testsuite/gcc.dg/vect/vect-78.c: Same.
From-SVN: r100718
2005-06-07 Eric Christopher <echristo@redhat.com>
* config/mips/predicates.md (sleu_operand): Use
IN_RANGE to specify range of operand.
From-SVN: r100716
PR tree-optimization/21847
* tree-ssa-dce.c (mark_stmt_if_obviously_necessary): With
-fnon-call-exceptions, also mark statements inherently
necessary if they may throw.
From-SVN: r100714
2005-06-07 Benjamin Kosnik <bkoz@redhat.com>
* acinclude.m4 (GLIBCXX_ENABLE_C99): Use C++ compiler for complex
tests.
* configure: Regenerated.
From-SVN: r100713
2005-06-07 Adrian Straetling <straetling@de.ibm.com>
gcc/ChangeLog:
* config/s390/s390.md: ("UNSPECV_MB", "UNSPECV_CAS"): New constants.
("type"): Add "sem" to 'type' attribute.
("memory_barrier", "*memory_barrier", "sync_compare_and_swapdi",
"sync_compare_and_swapsi", "sync_compare_and_swap_ccdi",
"sync_compare_and_swap_ccsi", "*sync_compare_and_swap_ccdi",
"*sync_compare_and_swap_ccsi"): New patterns.
* config/s390/2064.md: ("z_sem"): New insn_reservation.
* config/s390/2084.md: ("x_sem"): New insn_reservation.
* config/s390/s390.c: (s390_compare_emitted): New global variable.
(s390_emit_compare): Do not emit comparison again after cas.
* config/s390/s390.h (s390_compare_emitted): Declare.
libstdc++/ChangeLog:
* config/cpu/s390/atomicity.h: (__exchange_and_add,
__atomic_add): Use the builtins for atomic memory operations.
From-SVN: r100711
PR debug/21946
* dwarf2out.c (add_loc_descr_op_piece): New function.
(multiple_reg_loc_descriptor, concat_loc_descriptor,
loc_descriptor): Use it.
* var-tracking.c: Include regs.h and expr.h.
(emit_note_insn_var_location): Skip over pieces where offset
is smaller than previous offset plus previous piece mode size.
Optimize adjacent hard registers or memory locations.
* Makefile.in (var-tracking.o): Depend on $(REGS_H) and $(EXPR_H).
From-SVN: r100706
2005-06-07 Richard Guenther <rguenth@gcc.gnu.org>
* c-typeck.c (c_finish_if_stmt): Use void_type_node as type
for COND_EXPR.
* gimplify.c (gimplify_cond_expr): No need to fix up the
type of COND_EXPRs.
From-SVN: r100705