For the targets supporting low-overhead loops, IVOPTs has to take care of
the loops which will probably be transformed in RTL doloop optimization,
to try to make selected IV candidate set optimal. The process of doloop
support includes:
1) Analyze the current loop will be transformed to doloop or not, find and
mark its compare type IV use as doloop use (iv_group field doloop_p), and
set flag doloop_use_p of ivopts_data to notify subsequent processings on
doloop. See analyze_and_mark_doloop_use and its callees for the details.
The target hook predict_doloop_p can be used for target specific checks.
2) Add one doloop dedicated IV cand {(may_be_zero ? 1 : (niter + 1)), +, -1},
set flag doloop_p of iv_cand, step cost is set as zero and no extra cost
like biv. For cost determination between doloop IV cand and IV use, the
target hooks doloop_cost_for_generic and doloop_cost_for_address are
provided to add on extra costs for generic type and address type IV use.
Zero cost is assigned to the pair between doloop IV cand and doloop IV
use, and bound zero is set for IV elimination.
3) With the cost setting in step 2), the current cost model based IV
selection algorithm will process as usual, pick up doloop dedicated IV if
profitable.
gcc/ChangeLog
2019-09-14 Kewen Lin <linkw@gcc.gnu.org>
PR middle-end/80791
* config/rs6000/rs6000.c (TARGET_HAVE_COUNT_REG_DECR_P): New macro.
(TARGET_DOLOOP_COST_FOR_GENERIC): Likewise.
(TARGET_DOLOOP_COST_FOR_ADDRESS): Likewise.
* target.def (have_count_reg_decr_p): New hook.
(doloop_cost_for_generic): Likewise.
(doloop_cost_for_address): Likewise.
* doc/tm.texi.in (TARGET_HAVE_COUNT_REG_DECR_P): Likewise.
(TARGET_DOLOOP_COST_FOR_GENERIC): Likewise.
(TARGET_DOLOOP_COST_FOR_ADDRESS): Likewise.
* doc/tm.texi: Regenerate.
* tree-ssa-loop-ivopts.c (comp_cost::operator+=): Consider infinite cost
addend.
(record_group): Init doloop_p.
(add_candidate_1): Add optional argument doloop, change the handlings
accordingly.
(add_candidate): Likewise.
(generic_predict_doloop_p): Update attribute.
(force_expr_to_var_cost): Add costing for expressions COND_EXPR/LT_EXPR/
LE_EXPR/GT_EXPR/GE_EXPR/EQ_EXPR/NE_EXPR/UNORDERED_EXPR/ORDERED_EXPR/
UNLT_EXPR/UNLE_EXPR/UNGT_EXPR/UNGE_EXPR/UNEQ_EXPR/LTGT_EXPR/MAX_EXPR/
MIN_EXPR.
(get_computation_cost): Update for doloop IV cand extra cost.
(determine_group_iv_cost_cond): Update for doloop IV cand.
(determine_iv_cost): Likewise.
(ivopts_estimate_reg_pressure): Likewise.
(may_eliminate_iv): Update handlings for doloop IV cand.
(add_iv_candidate_for_doloop): New function.
(find_iv_candidates): Call function add_iv_candidate_for_doloop.
(iv_ca_set_no_cp): Update for doloop IV cand.
(iv_ca_set_cp): Likewise.
(iv_ca_dump): Dump register cost.
(find_doloop_use): New function.
(analyze_and_mark_doloop_use): Likewise.
(tree_ssa_iv_optimize_loop): Call function analyze_and_mark_doloop_use.
gcc/testsuite/ChangeLog
2019-09-14 Kewen Lin <linkw@gcc.gnu.org>
PR middle-end/80791
* gcc.dg/tree-ssa/ivopts-3.c: Adjust for doloop change.
* gcc.dg/tree-ssa/ivopts-lt.c: Likewise.
* gcc.dg/tree-ssa/pr32044.c: Likewise.
From-SVN: r275713
2019-09-13 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/91566
* simplify.c (gfc_simplify_merge): Need to simplify expression
after insertation of parenthesis.
2019-09-13 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/91566
* gfortran.dg/pr91566.f90:
From-SVN: r275704
According to `gnatmake' documentation:
"Any uppercase or multi-character switch that is not a 'gnatmake' switch
is passed to 'gcc' (e.g., '-O', '-gnato,' etc.)"
however the `--sysroot=' switch is actually rejected:
gnatmake: invalid switch: --sysroot=...
likely because it is one of the very few GCC driver options that have a
leading double dash and therefore we don't have a blanket fall-through
for such switches that would satisfy what our documentation claims.
The option is actually shared between the compiler and the linker, so
pass the switch to both build stages if requested, removing GNAT
testsuite issues like:
gnatmake: invalid switch: --sysroot=.../sysroot
compiler exited with status 1
Executing on host: .../gcc/gnatclean -c -q -n ./abstract_with_anonymous_result (timeout = 300)
spawn -ignore SIGHUP .../gcc/gnatclean -c -q -n ./abstract_with_anonymous_result
PASS: gnat.dg/abstract_with_anonymous_result.adb (test for excess errors)
UNRESOLVED: gnat.dg/abstract_with_anonymous_result.adb compilation failed to produce executable
in a test environment where `--with-build-sysroot=.../sysroot' has been
used to build a cross-compiler. Passing to the compilation stage only
would lead to errors like:
.../bin/riscv64-linux-gnu-ld: cannot find crt1.o: No such file or directory
.../bin/riscv64-linux-gnu-ld: cannot find -lc
collect2: error: ld returned 1 exit status
gnatlink: error when calling .../gcc/xgcc
gnatmake: *** link failed.
compiler exited with status 1
Executing on host: .../gcc/gnatclean -c -q -n ./abstract_with_anonymous_result (timeout = 300)
spawn -ignore SIGHUP .../gcc/gnatclean -c -q -n ./abstract_with_anonymous_result
./abstract_with_anonymous_result.ali
./abstract_with_anonymous_result.o
FAIL: gnat.dg/abstract_with_anonymous_result.adb (test for excess errors)
Excess errors:
.../bin/riscv64-linux-gnu-ld: cannot find crt1.o: No such file or directory
.../bin/riscv64-linux-gnu-ld: cannot find -lc
gnatlink: error when calling .../gcc/xgcc
UNRESOLVED: gnat.dg/abstract_with_anonymous_result.adb compilation failed to produce executable
instead.
gcc/ada/
* make.adb (Scan_Make_Arg): Also accept `--sysroot=' for the
compiler and the linker.
From-SVN: r275702
2019-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
PR fortran/91716
* trans-array.c (gfc_conv_array_initializer): Always assign the
array type of the field to the string constant.
testsuite:
2019-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
PR fortran/91716
* gfortran.dg/pr91716.f90: New test.
From-SVN: r275698
2019-09-13 Paul Thomas <pault@gcc.gnu.org>
PR fortran/91717
* dependency.c (gfc_dep_resolver): Flag identical components
and exit with return value 1 if set and no array refs.
2019-09-13 Paul Thomas <pault@gcc.gnu.org>
PR fortran/91717
* gfortran.dg/dependency_55.f90 : New test.
From-SVN: r275696
2019-09-12 Richard Biener <rguenther@suse.de>
PR tree-optimization/91750
* tree-vect-loop.c (vectorizable_induction): Compute IV increments
in the type of the evolution.
* gcc.dg/vect/pr91750.c: New testcase.
From-SVN: r275685
PR libstdc++/91748
* include/bits/stl_algo.h (for_each_n): Fix random access iterator
case.
* testsuite/25_algorithms/for_each/for_each_n.cc: Test with random
access iterators.
From-SVN: r275683
2019-09-11 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/91553
* simplify.c (gfc_convert_constant): During conversion check if the
constant is enclosed in parenthesis, and simplify expression.
2019-09-11 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/91553
* gfortran.dg/pr91553.f90: New test.
From-SVN: r275657
2019-09-11 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/91642
* io.c (gfc_match_inquire): null() cannot be in an iolength inquire
list.
2019-09-11 Steven G. Kargl <kargl@gcc.gnu.org>
PR fortran/91642
* gfortran.dg/pr91642.f90: New test.
From-SVN: r275655
Change Named_object::get_backend to ignore aliases when creating type
descriptors for types, to be consistent with
Type::needs_specific_type_functions and the Specific_type_functions
traversal class. For example, when compiling a package that creates an
alias to an an externally defined type, e.g.
import "foo"
type MyFoo = foo.Foo
it makes sense to skip the alias (not try to generate type specific
functions for it) and let the normal mechanisms take care of the alias
target, depending on whether the target is defined locally or defined
elsewhere.
Testcase for this problen can be found in CL 193261.
Fixesgolang/go#33866.
Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/191961
From-SVN: r275651
When compiling the x_test package, force the test package to be
imported first. That ensures that we will see the types defined in
the test package before the types defined in the non-test version of
the package. This matters if the types differ in some way, such as by
adding a new method.
This avoids a failure in internal/poll on Solaris, in which the test
package adds a method to a type (FD.EOFError). I think it was Solaris-
specific because files are sorted in a different order by default.
The go tool handles this kind of thing correctly, by rebuilding
dependent packages. This is just a hack sufficient to run the libgo
testsuite without using the go tool.
Fixes https://gcc.gnu.org/PR91712
Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/194637
From-SVN: r275648
* testsuite/20_util/result_of/sfinae_friendly_1.cc: Add -Wno-volatile
for C++2a and up. Define HAS_52748_FIXED and fix incorrect tests.
* testsuite/tr1/3_function_objects/result_of.cc: Add -Wno-volatile
for C++2a and up.
From-SVN: r275643
https://gcc.gnu.org/ml/gcc-patches/2019-09/msg00697.html
gcc/
* tree.h (MARK_TS_TYPE_NON_COMMON): New.
* tree.c (tree_node_structure_for_code): Reformat and alphabetize.
gcc/cp/
* c-objcp-common.c (cp-objcp-common.c): Alphababetize and
correctly mark all C++ nodes.
From-SVN: r275641
PR rtl-optimization/89795
* rtlanal.c (nonzero_bits1) <SUBREG>: Do not propagate results from
inner REGs to paradoxical SUBREGs if WORD_REGISTER_OPERATIONS is set.
From-SVN: r275635
PR tree-optimization/91723
* tree-vect-stmts.c (vectorizable_call): Use types_compatible_p check
instead of pointer equality when checking if argument vectypes are
the same.
* gcc.dg/vect/vect-fma-3.c: New test.
From-SVN: r275634
PR middle-end/91725
* match.pd ((A / (1 << B)) -> (A >> B)): Call tree_nonzero_bits instead
of get_nonzero_bits, only call it for integral types.
* gcc.c-torture/compile/pr91725.c: New test.
From-SVN: r275633
2019-09-11 Richard Biener <rguenther@suse.de>
Revert
2019-09-09 Barnaby Wilks <barnaby.wilks@arm.com>
* match.pd: Add flag_unsafe_math_optimizations check
before deciding on the widest type in a binary math operation.
* gcc.dg/fold-binary-math-casts.c: New test.
From-SVN: r275632
* parser.c (CP_PARSER_FLAGS_DELAY_NOEXCEPT): New parser flag.
(cp_parser_lambda_declarator_opt): Pass CP_PARSER_FLAGS_NONE to
cp_parser_exception_specification_opt.
(cp_parser_direct_declarator): Adjust a call to
cp_parser_exception_specification_opt.
(cp_parser_member_declaration): Pass CP_PARSER_FLAGS_DELAY_NOEXCEPT
to cp_parser_declarator if not processing a friend or typedef
declaration.
(cp_parser_late_noexcept_specifier): Adjust a call to
cp_parser_noexcept_specification_opt.
(cp_parser_noexcept_specification_opt): New parameter for parser flags,
drop the FRIEND_P parameter. Use the new parameter.
(cp_parser_exception_specification_opt): Likewise.
(cp_parser_transaction): Adjust a call to
cp_parser_noexcept_specification_opt.
(cp_parser_transaction_expression): Likewise.
* g++.dg/cpp1z/using7.C: New test.
* g++.dg/cpp1z/using8.C: New test.
From-SVN: r275617
Restore Solaris compatibility fixes lost when internal/x/net/lif moved
to golang.org/x/net/lif. Also fix the Makefile for x/net/lif and
x/net/route.
Change x/sys/cpu to get the cache line size from goarch.sh as the
gofrontend version of internal/cpu does.
Partially based on work by Rainer Orth.
Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/194438
From-SVN: r275611
Backport of https://golang.org/cl/194440. Original description:
If an embedded field refers to a type via a pointer, the parser needs
to know the name of the embedded field. It is possible that the
pointer type is not yet resolved. This CL fixes the parser to handle
that case by setting the pointer element type to the unresolved named
type while the pointer is being resolved.
Updates golang/go#34182
Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/194562
From-SVN: r275606
After previous patches, it's now possible for call_used_regs to be
the "real" set of call-clobbered registers, without any special
handling for fixed registers. This patch therefore removes the
separate call_really_used_regs and updates the targets that define
CALL_REALLY_USED_REGISTERS so that they handle call_used_regs in
the same way that they used to handle call_really_used_regs.
With this change, it's no longer necessary for targets that define
CALL_REALLY_USED_REGISTERS to define CALL_USED_REGISTER as well.
2019-09-10 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* doc/tm.texi.in: Document that exactly one of CALL_USED_REGISTERS
and CALL_REALLY_USED_REGISTERS must be defined, and that
CALL_REALLY_USED_REGISTERS is preferred.
* doc/tm.texi: Regenerate.
* hard-reg-set.h (target_hard_regs::x_call_really_used_regs): Delete.
(call_really_used_regs): Likewise.
* reginfo.c: Raise an #error if both CALL_USED_REGISTERS and
CALL_REALLY_USED_REGISTERS are defined.
(initial_call_used_regs): Use CALL_REALLY_USED_REGISTERS as the
initial value if defined.
(initial_call_really_used_regs): Delete.
(saved_call_really_used_regs): Likewise.
(CALL_REALLY_USED_REGNO_P): Likewise.
(init_reg_sets): Remove handling of call_really_used_regs.
(save_register_info, restore_register_info, globalize_reg): Likewise.
(init_reg_sets_1): Likewise. Use call_used_regs instead of
CALL_REALLY_USED_REGNO_P. Don't set call_used_regs for registers
outside operand_reg_set.
(fix_register): Don't change call_used_regs if
CALL_REALLY_USED_REGISTERS is defined.
* config/csky/csky.h (CALL_USED_REGISTERS): Delete.
* config/csky/csky.c (get_csky_live_regs): Use call_used_regs
instead of call_really_used_regs.
(csky_conditional_register_usage): Remove the old handling of
call_used_regs and change the handling of call_really_used_regs
to use call_used_regs instead.
* config/ia64/ia64.h (CALL_USED_REGISTERS): Delete.
* config/ia64/ia64.c (fix_range): Don't set call_used_regs when
making a register fixed.
* config/m32r/m32r.h (CALL_USED_REGISTERS): Delete.
* config/m32r/m32r.c (MUST_SAVE_REGISTER): Use call_used_regs
instead of call_really_used_regs.
(m32r_conditional_register_usage): Don't set call_used_regs when
making a register fixed.
* config/mips/mips.h (CALL_USED_REGISTERS): Delete.
* config/mips/mips.c (mips_global_pointer): Use call_used_regs
instead of call_really_used_regs.
(mips_interrupt_extra_call_saved_reg_p): Likewise.
(mips_cfun_call_saved_reg_p): Likewise.
(mips_swap_registers): Remove the old handling of call_used_regs
and change the handling of call_really_used_regs to use call_used_regs
instead.
(mips_conditional_register_usage): Likewise.
* config/mn10300/mn10300.h (CALL_USED_REGISTERS): Delete.
* config/mn10300/mn10300.c (fp_regs_to_save): Use call_used_regs
instead of call_really_used_regs.
(mn10300_get_live_callee_saved_regs): Likewise.
(mn10300_expand_prologue, mn10300_expand_epilogue): Likewise.
(mn10300_conditional_register_usage): Don't set call_used_regs when
making a register fixed.
* config/rs6000/rs6000.h (CALL_USED_REGISTERS): Delete.
* config/rs6000/rs6000.c (rs6000_conditional_register_usage):
Remove the old handling of call_used_regs and change the handling
of call_really_used_regs to use call_used_regs instead.
* config/s390/s390.h (CALL_USED_REGISTERS): Delete.
* config/s390/s390.c (s390_regs_ever_clobbered): Use call_used_regs
instead of call_really_used_regs.
(s390_register_info_gprtofpr, s390_register_info): Likewise.
(s390_hard_regno_rename_ok, s390_hard_regno_scratch_ok): Likewise.
(s390_emit_prologue, s300_set_up_by_prologue): Likewise.
(s390_can_use_return_insn, s390_optimize_prologue): Likewise.
(s390_conditional_register_usage): Remove the old handling of
call_used_regs and change the handling of call_really_used_regs
to use call_used_regs instead.
* config/sh/sh.h (CALL_USED_REGISTERS): Delete.
* config/sh/sh.c (output_stack_adjust, calc_live_regs): Likewise.
(sh_fix_range, reg_unused_after): Likewise.
(sh_conditional_register_usage): Remove the old handling of
call_used_regs and change the handling of call_really_used_regs
to use call_used_regs instead.
* config/sparc/sparc.h (CALL_USED_REGISTERS): Delete.
* config/sparc/sparc.c (sparc_conditional_register_usage): Don't set
call_used_regs when making a register fixed.
* config/tilegx/tilegx.h (CALL_USED_REGISTERS): Delete.
* config/tilegx/tilegx.c (tilegx_conditional_register_usage): Don't set
call_used_regs when making a register fixed.
* config/tilepro/tilepro.h (CALL_USED_REGISTERS): Delete.
* config/tilepro/tilepro.c (tilepro_conditional_register_usage): Don't
set call_used_regs when making a register fixed.
* config/visium/visium.h (CALL_USED_REGISTERS): Delete.
* config/visium/visium.c (visium_conditional_register_usage): Remove
the old handling of call_used_regs and change the handling of
call_really_used_regs to use call_used_regs instead.
From-SVN: r275605
Now that tests of call_used_regs go through call_used_or_fixed_reg_p,
we can hide call_used_regs from target-independent code. (It still
needs to be available to targets for the conditional register usage
hooks.)
2019-09-10 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* hard-reg-set.h (call_used_regs): Only define if IN_TARGET_CODE.
(call_used_or_fixed_reg_p): Expand definition of call_used_regs.
* reginfo.c (call_used_regs): New macro.
From-SVN: r275604
This cleans up a couple of places in which the previous patch had:
call_used_or_fixed_regs & ~fixed_reg_set
In that context, regs_invalidated_by_call is IMO more obvious.
2019-09-10 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* config/frv/frv.c (frv_ifcvt_modify_tests): Use
regs_invalidated_by_call & ~fixed_reg_set instead of
call_used_or_fixed_regs & ~fixed_reg_set.
* config/sh/sh.c (output_stack_adjust): Likewise.
From-SVN: r275601
CALL_USED_REGISTERS and call_used_regs infamously contain all fixed
registers (hence the need for CALL_REALLY_USED_REGISTERS etc.).
We try to recover from this to some extent with:
/* Contains 1 for registers that are set or clobbered by calls. */
/* ??? Ideally, this would be just call_used_regs plus global_regs, but
for someone's bright idea to have call_used_regs strictly include
fixed_regs. Which leaves us guessing as to the set of fixed_regs
that are actually preserved. We know for sure that those associated
with the local stack frame are safe, but scant others. */
HARD_REG_SET x_regs_invalidated_by_call;
Since global registers are added to fixed_reg_set and call_used_reg_set
too, it's always the case that:
call_used_reg_set == regs_invalidated_by_call | fixed_reg_set
This patch replaces all uses of call_used_reg_set with a new macro
call_used_or_fixed_regs to make this clearer.
This is part of a series that allows call_used_regs to be what is
now call_really_used_regs. It's a purely mechanical replacement;
later patches clean up obvious oddities like
"call_used_or_fixed_regs & ~fixed_regs".
2019-09-10 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* hard-reg-set.h (target_hard_regs::x_call_used_reg_set): Delete.
(call_used_reg_set): Delete.
(call_used_or_fixed_regs): New macro.
* reginfo.c (init_reg_sets_1, globalize_reg): Remove initialization
of call_used_reg_set.
* caller-save.c (setup_save_areas): Use call_used_or_fixed_regs
instead of call_used_regs.
(save_call_clobbered_regs): Likewise.
* cfgcleanup.c (old_insns_match_p): Likewise.
* config/c6x/c6x.c (c6x_call_saved_register_used): Likewise.
* config/epiphany/epiphany.c (epiphany_conditional_register_usage):
Likewise.
* config/frv/frv.c (frv_ifcvt_modify_tests): Likewise.
* config/sh/sh.c (output_stack_adjust): Likewise.
* final.c (collect_fn_hard_reg_usage): Likewise.
* ira-build.c (ira_build): Likewise.
* ira-color.c (calculate_saved_nregs): Likewise.
(allocno_reload_assign, calculate_spill_cost): Likewise.
* ira-conflicts.c (ira_build_conflicts): Likewise.
* ira-costs.c (ira_tune_allocno_costs): Likewise.
* ira-lives.c (process_bb_node_lives): Likewise.
* ira.c (setup_reg_renumber): Likewise.
* lra-assigns.c (find_hard_regno_for_1, lra_assign): Likewise.
* lra-constraints.c (need_for_call_save_p): Likewise.
(need_for_split_p, inherit_in_ebb): Likewise.
* lra-lives.c (process_bb_lives): Likewise.
* lra-remat.c (call_used_input_regno_present_p): Likewise.
* postreload.c (reload_combine): Likewise.
* regrename.c (find_rename_reg): Likewise.
* reload1.c (reload_as_needed): Likewise.
* rtlanal.c (find_all_hard_reg_sets): Likewise.
* sel-sched.c (mark_unavailable_hard_regs): Likewise.
* shrink-wrap.c (requires_stack_frame_p): Likewise.
From-SVN: r275600
Reusing the reasoning from the call_fixed_reg_set patch:
for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
for (j = 1; j <= MOVE_MAX_WORDS; j++)
if (reg_save_code (i,regno_save_mode[i][j]) == -1)
should be true whenever regno_save_mode[i][j] == VOIDmode,
which it is for:
~call_used_reg_set | ~have_save_mode
So this condition amounts to:
~call_used_reg_set | ~have_save_mode | ~have_save_insn
== ~call_used_reg_set | ~savable_regs
no_caller_save_reg_set is then set if call_used_regs[i], so
no_caller_save_reg_set is:
call_used_reg_set & (~call_used_reg_set | ~savable_regs)
== call_used_reg_set & ~savable_regs
This patch expands its single user accordingly.
Note that ~savable_regs is always empty on LRA targets.
2019-09-10 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* hard-reg-set.h (target_hard_regs::x_no_caller_save_reg_set): Delete.
(no_caller_save_reg_set): Delete.
* caller-save.c (init_caller_save): Don't initialize it.
* ira-conflicts.c (ira_build_conflicts): Calculate
no_caller_save_reg_set locally from call_used_reg_set and savable_regs.
From-SVN: r275599
On targets that use reload, call_fixed_reg_set is structurally:
fixed_reg_set -- reginfo.c
| (call_used_reg_set & ~have_save_mode) -- first loop in init_caller_save
| ~have_save_insn -- final loop in init_caller_save
(where "have_save_mode" and "have_save_insn" are just my names).
But the final loop in init_caller_save does:
for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
for (j = 1; j <= MOVE_MAX_WORDS; j++)
if (reg_save_code (i,regno_save_mode[i][j]) == -1)
This last condition ought to be true whenever:
regno_save_mode[i][j] == VOIDmode
since either targetm.hard_regno_mode_ok (i, VOIDmode) should be
false or the VOIDmode save & restore shouldn't match any move insn.
And after the first loop, regno_save_mode[i][j] == VOIDmode whenever
!call_used_regs[i]. So the above is actually:
fixed_reg_set
| (call_used_reg_set & ~have_save_mode)
| (~call_used_reg_set | ~have_save_insn)
which simplifies to:
fixed_reg_set -- reginfo.c
| ~have_save_mode -- first loop in init_caller_save
| ~have_save_insn -- final loop in init_caller_save
| ~call_used_reg_set -- final loop in init_caller_save
So:
~call_fixed_reg_set == (~fixed_reg_set
& have_save_mode
& have_save_insn
& call_used_reg_set) [A]
All users have the form:
(call_used_reg_set or some subset) & ~(call_fixed_reg_set | ...)
i.e.:
(call_used_reg_set or some subset) & ~call_fixed_reg_set & ~(...)
We can therefore drop the "& call_used_reg_set" from [A], leaving:
~fixed_reg_set & have_save_mode & have_save_insn
This patch combines have_save_mode & have_save_insn into a single
condition "a save is possible", represented as savable_regs.
We can then substitute:
~call_fixed_reg_set --> ~fixed_reg_set & savable_regs
(registers we can actually save around calls)
The patch also sets regno_save_mode[i][j] for all registers,
in case non-default ABIs require a save when the default ABI
doesn't. This ensures that savable_regs (like fixed_reg_set but
unlike call_fixed_reg_set) isn't affected by the ABI. This only
becomes significant with later patches and at this point is just
a simplification.
Since init_caller_save is only called for reload targets,
the default assumption for LRA is that all registers are savable,
just like the default assumption before the patch was that
(~)call_fixed_reg_set == (~)fixed_reg_set.
2019-09-10 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* hard-reg-set.h (target_hard_regs::x_call_fixed_reg_set): Delete.
(target_hard_regs::x_savable_regs): New field.
(call_fixed_reg_set): Delete.
(savable_regs): New macro,
* reginfo.c (globalize_reg): Don't set call_fixed_reg_set.
(init_reg_sets_1): Likewise. Initialize savable_regs.
* caller-save.c (init_caller_save): Invoke HARD_REGNO_CALLER_SAVE_MODE
for all registers. Set savable_regs instead of call_fixed_reg_set.
(setup_save_areas, save_call_clobbered_regs): Replace uses of
~call_fixed_reg_set with ~fixed_reg_set & savable_regs.
* config/sh/sh.c (output_stack_adjust): Likewise.
From-SVN: r275598
Kewen Lin
GCC Administrator
Steven G. Kargl
Tobias Burnus
Maciej W. Rozycki
Bernd Edlinger
Ian Lance Taylor
Ian Lance Taylor
Sam Tebbs
Paul Thomas
Uros Bizjak
Richard Biener
Jonathan Wakely
Yuliang Wang
Sandra Loosemore
Michael Meissner
Jakub Jelinek
Nathan Sidwell
Eric Botcazou
Tobias Burnus
Marek Polacek
Richard Sandiford