Add testsuite infrastructure for the A extension and use it to require the A
extension for dg-do run and add the add extension for non-A dg-do compile.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/amo-table-a-6-amo-add-1.c: Add A extension to
dg-options for dg-do compile.
* gcc.target/riscv/amo-table-a-6-amo-add-2.c: Ditto.
* gcc.target/riscv/amo-table-a-6-amo-add-3.c: Ditto.
* gcc.target/riscv/amo-table-a-6-amo-add-4.c: Ditto.
* gcc.target/riscv/amo-table-a-6-amo-add-5.c: Ditto.
* gcc.target/riscv/amo-table-a-6-compare-exchange-1.c: Ditto.
* gcc.target/riscv/amo-table-a-6-compare-exchange-2.c: Ditto.
* gcc.target/riscv/amo-table-a-6-compare-exchange-3.c: Ditto.
* gcc.target/riscv/amo-table-a-6-compare-exchange-4.c: Ditto.
* gcc.target/riscv/amo-table-a-6-compare-exchange-5.c: Ditto.
* gcc.target/riscv/amo-table-a-6-compare-exchange-6.c: Ditto.
* gcc.target/riscv/amo-table-a-6-compare-exchange-7.c: Ditto.
* gcc.target/riscv/amo-table-a-6-subword-amo-add-1.c: Ditto.
* gcc.target/riscv/amo-table-a-6-subword-amo-add-2.c: Ditto.
* gcc.target/riscv/amo-table-a-6-subword-amo-add-3.c: Ditto.
* gcc.target/riscv/amo-table-a-6-subword-amo-add-4.c: Ditto.
* gcc.target/riscv/amo-table-a-6-subword-amo-add-5.c: Ditto.
* gcc.target/riscv/inline-atomics-2.c: Ditto.
* gcc.target/riscv/inline-atomics-3.c: Require A extension for dg-do
run.
* gcc.target/riscv/inline-atomics-4.c: Ditto.
* gcc.target/riscv/inline-atomics-5.c: Ditto.
* gcc.target/riscv/inline-atomics-6.c: Ditto.
* gcc.target/riscv/inline-atomics-7.c: Ditto.
* gcc.target/riscv/inline-atomics-8.c: Ditto.
* lib/target-supports.exp: Add testing infrastructure to require the A
extension or add it to an existing -march.
Signed-off-by: Patrick O'Neill <patrick@rivosinc.com>
Non-atomic targets are currently prevented from using the optimized fencing for
seq_cst load/seq_cst store. This patch removes that constraint.
gcc/ChangeLog:
* config/riscv/sync-rvwmo.md (atomic_load_rvwmo<mode>): Remove
TARGET_ATOMIC constraint
(atomic_store_rvwmo<mode>): Ditto.
* config/riscv/sync-ztso.md (atomic_load_ztso<mode>): Ditto.
(atomic_store_ztso<mode>): Ditto.
* config/riscv/sync.md (atomic_load<mode>): Ditto.
(atomic_store<mode>): Ditto.
Signed-off-by: Patrick O'Neill <patrick@rivosinc.com>
The XTheadFMemIdx ISA extension provides additional load and store
instructions for floating-point registers with new addressing modes.
The following memory accesses types are supported:
* load/store: [w,d] (single-precision FP, double-precision FP)
The following addressing modes are supported:
* register offset with additional immediate offset (4 instructions):
flr<type>, fsr<type>
* zero-extended register offset with additional immediate offset
(4 instructions): flur<type>, fsur<type>
These addressing modes are also part of the similar XTheadMemIdx
ISA extension support, whose code is reused and extended to support
floating-point registers.
One challenge that this patch needs to solve are GP registers in FP-mode
(e.g. "(reg:DF a2)"), which cannot be handled by the XTheadFMemIdx
instructions. Such registers are the result of independent
optimizations, which can happen after register allocation.
This patch uses a simple but efficient method to address this:
add a dependency for XTheadMemIdx to XTheadFMemIdx optimizations.
This allows to use the instructions from XTheadMemIdx in case
of such registers.
The added tests ensure that this feature won't regress without notice.
Testing: GCC regression test suite and SPEC CPU 2017 intrate (base&peak).
Signed-off-by: Christoph Müllner <christoph.muellner@vrull.eu>
gcc/ChangeLog:
* config/riscv/riscv.cc (riscv_index_reg_class):
Return GR_REGS for XTheadFMemIdx.
(riscv_regno_ok_for_index_p): Add support for XTheadFMemIdx.
* config/riscv/riscv.h (HARDFP_REG_P): New macro.
* config/riscv/thead.cc (is_fmemidx_mode): New function.
(th_memidx_classify_address_index): Add support for XTheadFMemIdx.
(th_fmemidx_output_index): New function.
(th_output_move): Add support for XTheadFMemIdx.
* config/riscv/thead.md (TH_M_ANYF): New mode iterator.
(TH_M_NOEXTF): Likewise.
(*th_fmemidx_movsf_hardfloat): New INSN.
(*th_fmemidx_movdf_hardfloat_rv64): Likewise.
(*th_fmemidx_I_a): Likewise.
(*th_fmemidx_I_c): Likewise.
(*th_fmemidx_US_a): Likewise.
(*th_fmemidx_US_c): Likewise.
(*th_fmemidx_UZ_a): Likewise.
(*th_fmemidx_UZ_c): Likewise.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/xtheadfmemidx-index-update.c: New test.
* gcc.target/riscv/xtheadfmemidx-index-xtheadbb-update.c: New test.
* gcc.target/riscv/xtheadfmemidx-index-xtheadbb.c: New test.
* gcc.target/riscv/xtheadfmemidx-index.c: New test.
* gcc.target/riscv/xtheadfmemidx-uindex-update.c: New test.
* gcc.target/riscv/xtheadfmemidx-uindex-xtheadbb-update.c: New test.
* gcc.target/riscv/xtheadfmemidx-uindex-xtheadbb.c: New test.
* gcc.target/riscv/xtheadfmemidx-uindex.c: New test.
Signed-off-by: Christoph Müllner <christoph.muellner@vrull.eu>
The XTheadMemIdx ISA extension provides a additional load and store
instructions with new addressing modes.
The following memory accesses types are supported:
* load: b,bu,h,hu,w,wu,d
* store: b,h,w,d
The following addressing modes are supported:
* immediate offset with PRE_MODIFY or POST_MODIFY (22 instructions):
l<ltype>.ia, l<ltype>.ib, s<stype>.ia, s<stype>.ib
* register offset with additional immediate offset (11 instructions):
lr<ltype>, sr<stype>
* zero-extended register offset with additional immediate offset
(11 instructions): lur<ltype>, sur<stype>
The RISC-V base ISA does not support index registers, so the changes
are kept separate from the RISC-V standard support as much as possible.
To combine the shift/multiply instructions into the memory access
instructions, this patch comes with a few insn_and_split optimizations
that allow the combiner to do this task.
Handling the different cases of extensions results in a couple of INSNs
that look redundant on first view, but they are just the equivalence
of what we already have for Zbb as well. The only difference is, that
we have much more load instructions.
We already have a constraint with the name 'th_f_fmv', therefore,
the new constraints follow this pattern and have the same length
as required ('th_m_mia', 'th_m_mib', 'th_m_mir', 'th_m_miu').
The added tests ensure that this feature won't regress without notice.
Testing: GCC regression test suite, GCC bootstrap build, and
SPEC CPU 2017 intrate (base&peak) on C920.
Signed-off-by: Christoph Müllner <christoph.muellner@vrull.eu>
gcc/ChangeLog:
* config/riscv/constraints.md (th_m_mia): New constraint.
(th_m_mib): Likewise.
(th_m_mir): Likewise.
(th_m_miu): Likewise.
* config/riscv/riscv-protos.h (enum riscv_address_type):
Add new address types ADDRESS_REG_REG, ADDRESS_REG_UREG,
and ADDRESS_REG_WB and their documentation.
(struct riscv_address_info): Add new field 'shift' and
document the field usage for the new address types.
(riscv_valid_base_register_p): New prototype.
(th_memidx_legitimate_modify_p): Likewise.
(th_memidx_legitimate_index_p): Likewise.
(th_classify_address): Likewise.
(th_output_move): Likewise.
(th_print_operand_address): Likewise.
* config/riscv/riscv.cc (riscv_index_reg_class):
Return GR_REGS for XTheadMemIdx.
(riscv_regno_ok_for_index_p): Add support for XTheadMemIdx.
(riscv_classify_address): Call th_classify_address() on top.
(riscv_output_move): Call th_output_move() on top.
(riscv_print_operand_address): Call th_print_operand_address()
on top.
* config/riscv/riscv.h (HAVE_POST_MODIFY_DISP): New macro.
(HAVE_PRE_MODIFY_DISP): Likewise.
* config/riscv/riscv.md (zero_extendqi<SUPERQI:mode>2): Disable
for XTheadMemIdx.
(*zero_extendqi<SUPERQI:mode>2_internal): Convert to expand,
create INSN with same name and disable it for XTheadMemIdx.
(extendsidi2): Likewise.
(*extendsidi2_internal): Disable for XTheadMemIdx.
* config/riscv/thead.cc (valid_signed_immediate): New helper
function.
(th_memidx_classify_address_modify): New function.
(th_memidx_legitimate_modify_p): Likewise.
(th_memidx_output_modify): Likewise.
(is_memidx_mode): Likewise.
(th_memidx_classify_address_index): Likewise.
(th_memidx_legitimate_index_p): Likewise.
(th_memidx_output_index): Likewise.
(th_classify_address): Likewise.
(th_output_move): Likewise.
(th_print_operand_address): Likewise.
* config/riscv/thead.md (*th_memidx_operand): New splitter.
(*th_memidx_zero_extendqi<SUPERQI:mode>2): New INSN.
(*th_memidx_extendsidi2): Likewise.
(*th_memidx_zero_extendsidi2): Likewise.
(*th_memidx_zero_extendhi<GPR:mode>2): Likewise.
(*th_memidx_extend<SHORT:mode><SUPERQI:mode>2): Likewise.
(*th_memidx_bb_zero_extendsidi2): Likewise.
(*th_memidx_bb_zero_extendhi<GPR:mode>2): Likewise.
(*th_memidx_bb_extendhi<GPR:mode>2): Likewise.
(*th_memidx_bb_extendqi<SUPERQI:mode>2): Likewise.
(TH_M_ANYI): New mode iterator.
(TH_M_NOEXTI): Likewise.
(*th_memidx_I_a): New combiner optimization.
(*th_memidx_I_b): Likewise.
(*th_memidx_I_c): Likewise.
(*th_memidx_US_a): Likewise.
(*th_memidx_US_b): Likewise.
(*th_memidx_US_c): Likewise.
(*th_memidx_UZ_a): Likewise.
(*th_memidx_UZ_b): Likewise.
(*th_memidx_UZ_c): Likewise.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/xtheadmemidx-helpers.h: New test.
* gcc.target/riscv/xtheadmemidx-index-update.c: New test.
* gcc.target/riscv/xtheadmemidx-index-xtheadbb-update.c: New test.
* gcc.target/riscv/xtheadmemidx-index-xtheadbb.c: New test.
* gcc.target/riscv/xtheadmemidx-index.c: New test.
* gcc.target/riscv/xtheadmemidx-modify-xtheadbb.c: New test.
* gcc.target/riscv/xtheadmemidx-modify.c: New test.
* gcc.target/riscv/xtheadmemidx-uindex-update.c: New test.
* gcc.target/riscv/xtheadmemidx-uindex-xtheadbb-update.c: New test.
* gcc.target/riscv/xtheadmemidx-uindex-xtheadbb.c: New test.
* gcc.target/riscv/xtheadmemidx-uindex.c: New test.
Currently we have the documentation for __builtin_vec_bcdsub_{eq,gt,lt} but
not for __builtin_bcdsub_{gl}e, this patch is to supplement the descriptions
for them. Although they are mainly for __builtin_bcdcmp{ge,le}, we already
have some testing coverage for __builtin_vec_bcdsub_{eq,gt,lt}, this patch
adds the corresponding explicit test cases as well.
gcc/ChangeLog:
* doc/extend.texi (__builtin_bcdsub_le, __builtin_bcdsub_ge): Add
documentation for the builti-ins.
gcc/testsuite/ChangeLog:
* gcc.target/powerpc/bcd-3.c (do_sub_ge, do_suble): Add functions
to test builtins __builtin_bcdsub_ge and __builtin_bcdsub_le.
The MICROBLAZE_VERSION_COMPARE was incorrectly using strcasecmp
instead of strverscmp to check the mcpu version against feature
options. By simply changing the define to use strverscmp,
the new version 10.0 is treated correctly as a higher version
than previous versions.
Fix incorrect warning with -mcpu=10.0:
warning: '-mxl-multiply-high' can be used only with
'-mcpu=v6.00.a' or greater
Signed-off-by: Neal Frager <neal.frager@amd.com>
Signed-off-by: Michael J. Eager <eager@eagercon.com>
PR111971 test case uses a multi-reg variable containing a fixed reg. LRA
rejects such multi-reg because of this when matching the constraint for
an asm insn. The rejection results in LRA cycling. The patch fixes this issue.
gcc/ChangeLog:
PR rtl-optimization/111971
* lra-constraints.cc: (process_alt_operands): Don't check start
hard regs for regs originated from register variables.
gcc/testsuite/ChangeLog:
PR rtl-optimization/111971
* gcc.target/powerpc/pr111971.c: New test.
On riscv insn-emit.cc has grown to over 1.2 mio lines of code and
compiling it takes considerable time.
Therefore, this patch adjust genemit to create several partitions
(insn-emit-1.cc to insn-emit-n.cc). The available patterns are
written to the given files in a sequential fashion.
Similar to match.pd a configure option --with-emitinsn-partitions=num
is introduced that makes the number of partition configurable.
gcc/ChangeLog:
PR bootstrap/84402
PR target/111600
* Makefile.in: Handle split insn-emit.cc.
* configure: Regenerate.
* configure.ac: Add --with-insnemit-partitions.
* genemit.cc (output_peephole2_scratches): Print to file instead
of stdout.
(print_code): Ditto.
(gen_rtx_scratch): Ditto.
(gen_exp): Ditto.
(gen_emit_seq): Ditto.
(emit_c_code): Ditto.
(gen_insn): Ditto.
(gen_expand): Ditto.
(gen_split): Ditto.
(output_add_clobbers): Ditto.
(output_added_clobbers_hard_reg_p): Ditto.
(print_overload_arguments): Ditto.
(print_overload_test): Ditto.
(handle_overloaded_code_for): Ditto.
(handle_overloaded_gen): Ditto.
(print_header): New function.
(handle_arg): New function.
(main): Split output into 10 files.
* gensupport.cc (count_patterns): New function.
* gensupport.h (count_patterns): Define.
* read-md.cc (md_reader::print_md_ptr_loc): Add file argument.
* read-md.h (class md_reader): Change definition.
Control flow redundancy may choose abnormal edges for early checking,
but that breaks because we can't insert checks on such edges.
Introduce conditional checking on the dest block of abnormal edges,
and leave it for the optimizer to drop the conditional.
for gcc/ChangeLog
PR tree-optimization/111943
* gimple-harden-control-flow.cc: Adjust copyright year.
(rt_bb_visited): Add vfalse and vtrue data members.
Zero-initialize them in the ctor.
(rt_bb_visited::insert_exit_check_on_edge): Upon encountering
abnormal edges, insert initializers for vfalse and vtrue on
entry, and insert the check sequence guarded by a conditional
in the dest block.
for libgcc/ChangeLog
* hardcfr.c: Adjust copyright year.
for gcc/testsuite/ChangeLog
PR tree-optimization/111943
* gcc.dg/harden-cfr-pr111943.c: New.
The following adjusts final value replacement to also rewrite the
replacement to defined overflow behavior if there's conditionally
evaluated stmts (with possibly undefined overflow), not only when
we "folded casts". The patch hooks into expression_expensive for
this.
PR tree-optimization/112305
* tree-scalar-evolution.h (expression_expensive): Adjust.
* tree-scalar-evolution.cc (expression_expensive): Record
when we see a COND_EXPR.
(final_value_replacement_loop): When the replacement contains
a COND_EXPR, rewrite it to defined overflow.
* tree-ssa-loop-ivopts.cc (may_eliminate_iv): Adjust.
* gcc.dg/torture/pr112305.c: New testcase.
The lowering done for invoking `new' on a single dimension array was
moved from the code generator to the front-end semantic pass in
r14-4996. This removes the detritus left behind in the code generator
from that deletion.
gcc/d/ChangeLog:
* expr.cc (ExprVisitor::visit (NewExp *)): Remove unused assignments.
Now loongarch.md uses HAVE_AS_TLS, we need this to fix the failure
building a cross compiler if the cross assembler is not installed yet.
gcc/ChangeLog:
PR target/112299
* config/loongarch/loongarch-opts.h (HAVE_AS_TLS): Define to 0
if not defined yet.
gcc/ChangeLog:
* config/i386/avx512cdintrin.h (target): Push evex512 for
avx512cd.
* config/i386/avx512vlintrin.h (target): Split avx512cdvl part
out from avx512vl.
* config/i386/i386-builtin.def (BDESC): Do not check evex512
for builtins not needed.
Hi,
This patch let the INT64 to FP16 convert split to two small converts
(INT64 -> FP32 and FP32 -> FP16) when expanding instead of dealy the
split to split1 pass. This change could make it possible to combine
the FP32 to FP16 and vcond patterns and so we don't need to add an
combine pattern for INT64 to FP16 and vcond patterns.
Consider this code:
void
foo (_Float16 *__restrict r, int64_t *__restrict a, _FLoat16 *__restrict b,
int64_t *__restrict pred, int n)
{
for (int i = 0; i < n; i += 1)
{
r[i] = pred[i] ? (_Float16) a[i] : b[i];
}
}
Before this patch:
...
vfncvt.f.f.w v2,v2
vmerge.vvm v1,v1,v2,v0
vse16.v v1,0(a0)
...
After this patch:
...
vfncvt.f.f.w v1,v2,v0.t
vse16.v v1,0(a0)
...
gcc/ChangeLog:
* config/riscv/autovec.md (<float_cvt><mode><vnnconvert>2):
Change to define_expand.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/cond/cond_convert_int2float-rv32-1.c:
Add vfncvt.f.f.w assert.
* gcc.target/riscv/rvv/autovec/cond/cond_convert_int2float-rv32-2.c:
Ditto.
* gcc.target/riscv/rvv/autovec/cond/cond_convert_int2float-rv64-1.c:
Ditto.
* gcc.target/riscv/rvv/autovec/cond/cond_convert_int2float-rv64-2.c:
Ditto.
This moves a few more value_replacements simplifications to match.
/* a == 1 ? b : a * b -> a * b */
/* a == 1 ? b : b / a -> b / a */
/* a == -1 ? b : a & b -> a & b */
Also adds a testcase to show can we catch these where value_replacement would not
(but other passes would).
Bootstrapped and tested on x86_64-linux-gnu with no regressions.
gcc/ChangeLog:
* match.pd (`a == 1 ? b : a OP b`): New pattern.
(`a == -1 ? b : a & b`): New pattern.
gcc/testsuite/ChangeLog:
* gcc.dg/tree-ssa/phi-opt-value-4.c: New test.
This moves a few simple patterns that are done in value replacement
in phiopt over to match.pd. Just the simple ones which might show up
in other code.
This allows some optimizations to happen even without depending
on sinking from happening and in some cases where phiopt is not
invoked (cond-1.c is an example there).
Changes since v1:
* v2: Add an extra testcase to showcase improvements at -O1.
Bootstrapped and tested on x86_64-linux-gnu with no regressions.
gcc/ChangeLog:
* match.pd: (`a == 0 ? b : b + a`,
`a == 0 ? b : b - a`): New patterns.
gcc/testsuite/ChangeLog:
* gcc.dg/tree-ssa/cond-1.c: New test.
* gcc.dg/tree-ssa/phi-opt-value-1.c: New test.
* gcc.dg/tree-ssa/phi-opt-value-1a.c: New test.
* gcc.dg/tree-ssa/phi-opt-value-2.c: New test.
PR 111157 shows that IPA-modref and IPA-CP (when plugged into value
numbering) can optimize out a store both before a call (because the
call will overwrite it) and in the call (because the store is of the
same value) and by eliminating both create miscompilation.
This patch fixes that by pruning any constants from the list of IPA-CP
aggregate value constants that it knows the contents of the memory can
be "killed." Unfortunately, doing so is tricky. First, IPA-modref
loads override kills and so only stores not loaded are truly not
necessary. Looking stuff up there means doing what most of what
modref_may_alias may do but doing exactly what it does is tricky
because it takes also aliasing into account and has bail-out counters.
To err on the side of caution in order to avoid this miscompilation we
have to prune a constant when in doubt. However, pruning can
interfere with the mechanism of how clone materialization
distinguishes between the cases when a parameter was entirely removed
and when it was both IPA-CPed and IPA-SRAed (in order to make up for
the removal in debug info, which can bump into an assert when
compiling g++.dg/torture/pr103669.C when we are not careful).
Therefore this patch:
1) marks constants that IPA-modref has in its kill list with a new
"killed" flag, and
2) prunes the list from entries with this flag after materialization
and IPA-CP transformation is done using the template introduced in
the previous patch
It does not try to look up anything in the load lists, this will be
done as a follow-up in order to ease review.
gcc/ChangeLog:
2023-10-27 Martin Jambor <mjambor@suse.cz>
PR ipa/111157
* ipa-prop.h (struct ipa_argagg_value): Newf flag killed.
* ipa-modref.cc (ipcp_argagg_and_kill_overlap_p): New function.
(update_signature): Mark any any IPA-CP aggregate constants at
positions known to be killed as killed. Move check that there is
clone_info after this pruning.
* ipa-cp.cc (ipa_argagg_value_list::dump): Dump the killed flag.
(ipa_argagg_value_list::push_adjusted_values): Clear the new flag.
(push_agg_values_from_plats): Likewise.
(ipa_push_agg_values_from_jfunc): Likewise.
(estimate_local_effects): Likewise.
(push_agg_values_for_index_from_edge): Likewise.
* ipa-prop.cc (write_ipcp_transformation_info): Stream the killed
flag.
(read_ipcp_transformation_info): Likewise.
(ipcp_get_aggregate_const): Update comment, assert that encountered
record does not have killed flag set.
(ipcp_transform_function): Prune all aggregate constants with killed
set.
gcc/testsuite/ChangeLog:
2023-09-18 Martin Jambor <mjambor@suse.cz>
PR ipa/111157
* gcc.dg/lto/pr111157_0.c: New test.
* gcc.dg/lto/pr111157_1.c: Second file of the same new test.
PR 111157 points to another place where IPA-CP collected aggregate
compile-time constants need to be filtered, in addition to the one
place that already does this in ipa-sra. In order to re-use code,
this patch turns the common bit into a template.
The functionality is still covered by testcase gcc.dg/ipa/pr108959.c.
gcc/ChangeLog:
2023-09-13 Martin Jambor <mjambor@suse.cz>
PR ipa/111157
* ipa-prop.h (ipcp_transformation): New member function template
remove_argaggs_if.
* ipa-sra.cc (zap_useless_ipcp_results): Use remove_argaggs_if to
filter aggreagate constants.
GCC recently changed its register allocator which causes this
testcase to fail.
This patch updates the regex to be more robust to change by accepting
any s register in the range of 1-9 for cm.push and cm.popret insns.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rv32i_zcmp.c: Accept any register in the
range of 1-9 for cm.push and cm.popret insns.
Signed-off-by: Patrick O'Neill <patrick@rivosinc.com>
This patch optimizes PR middle-end/101955 for the ARC backend. On ARC
CPUs with a barrel shifter, using two shifts is optimal as:
asl_s r0,r0,31
asr_s r0,r0,31
but without a barrel shifter, GCC -O2 -mcpu=em currently generates:
and r2,r0,1
ror r2,r2
add.f 0,r2,r2
sbc r0,r0,r0
with this patch, we now generate the smaller, faster and non-flags
clobbering:
bmsk_s r0,r0,0
neg_s r0,r0
2023-10-30 Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog
PR middle-end/101955
* config/arc/arc.md (*extvsi_1_0): New define_insn_and_split
to convert sign extract of the least significant bit into an
AND $1 then a NEG when !TARGET_BARREL_SHIFTER.
gcc/testsuite/ChangeLog
PR middle-end/101955
* gcc.target/arc/pr101955.c: New test case.
This patch overhauls the ARC backend's insn_cost target hook, and makes
some related improvements to rtx_costs, BRANCH_COST, etc. The primary
goal is to allow the backend to indicate that shifts and rotates are
slow (discouraged) when the CPU doesn't have a barrel shifter. I should
also acknowledge Richard Sandiford for inspiring the use of set_cost
in this rewrite of arc_insn_cost; this implementation borrows heavily
for the target hooks for AArch64 and ARM.
The motivating example is derived from PR rtl-optimization/110717.
struct S { int a : 5; };
unsigned int foo (struct S *p) {
return p->a;
}
With a barrel shifter, GCC -O2 generates the reasonable:
foo: ldb_s r0,[r0]
asl_s r0,r0,27
j_s.d [blink]
asr_s r0,r0,27
What's interesting is that during combine, the middle-end actually
has two shifts by three bits, and a sign-extension from QI to SI.
Trying 8, 9 -> 11:
8: r158:SI=r157:QI#0<<0x3
REG_DEAD r157:QI
9: r159:SI=sign_extend(r158:SI#0)
REG_DEAD r158:SI
11: r155:SI=r159:SI>>0x3
REG_DEAD r159:SI
Whilst it's reasonable to simplify this to two shifts by 27 bits when
the CPU has a barrel shifter, it's actually a significant pessimization
when these shifts are implemented by loops. This combination can be
prevented if the backend provides accurate-ish estimates for insn_cost.
Previously, without a barrel shifter, GCC -O2 -mcpu=em generates:
foo: ldb_s r0,[r0]
mov lp_count,27
lp 2f
add r0,r0,r0
nop
2: # end single insn loop
mov lp_count,27
lp 2f
asr r0,r0
nop
2: # end single insn loop
j_s [blink]
which contains two loops and requires about ~113 cycles to execute.
With this patch to rtx_cost/insn_cost, GCC -O2 -mcpu=em generates:
foo: ldb_s r0,[r0]
mov_s r2,0 ;3
add3 r0,r2,r0
sexb_s r0,r0
asr_s r0,r0
asr_s r0,r0
j_s.d [blink]
asr_s r0,r0
which requires only ~6 cycles, for the shorter shifts by 3 and sign
extension.
2023-10-30 Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog
* config/arc/arc.cc (arc_rtx_costs): Improve cost estimates.
Provide reasonable values for SHIFTS and ROTATES by constant
bit counts depending upon TARGET_BARREL_SHIFTER.
(arc_insn_cost): Use insn attributes if the instruction is
recognized. Avoid calling get_attr_length for type "multi",
i.e. define_insn_and_split patterns without explicit type.
Fall-back to set_rtx_cost for single_set and pattern_cost
otherwise.
* config/arc/arc.h (COSTS_N_BYTES): Define helper macro.
(BRANCH_COST): Improve/correct definition.
(LOGICAL_OP_NON_SHORT_CIRCUIT): Preserve previous behavior.
This patch improves the code generated by the ARC back-end for CPUs
without a barrel shifter but with -mswap. The -mswap option provides
a SWAP instruction that implements SImode rotations by 16, but also
logical shift instructions (left and right) by 16 bits. Clearly these
are also useful building blocks for implementing shifts by 17, 18, etc.
which would otherwise require a loop.
As a representative example:
int shl20 (int x) { return x << 20; }
GCC with -O2 -mcpu=em -mswap would previously generate:
shl20: mov lp_count,10
lp 2f
add r0,r0,r0
add r0,r0,r0
2: # end single insn loop
j_s [blink]
with this patch we now generate:
shl20: mov_s r2,0 ;3
lsl16 r0,r0
add3 r0,r2,r0
j_s.d [blink]
asl_s r0,r0
Although both are four instructions (excluding the j_s),
the original takes ~22 cycles, and replacement ~4 cycles.
2023-10-30 Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog
* config/arc/arc.cc (arc_split_ashl): Use lsl16 on TARGET_SWAP.
(arc_split_ashr): Use swap and sign-extend on TARGET_SWAP.
(arc_split_lshr): Use lsr16 on TARGET_SWAP.
(arc_split_rotl): Use swap on TARGET_SWAP.
(arc_split_rotr): Likewise.
* config/arc/arc.md (ANY_ROTATE): New code iterator.
(<ANY_ROTATE>si2_cnt16): New define_insn for alternate form of
swap instruction on TARGET_SWAP.
(ashlsi2_cnt16): Rename from *ashlsi16_cnt16 and move earlier.
(lshrsi2_cnt16): New define_insn for LSR16 instruction.
(*ashlsi2_cnt16): See above.
gcc/testsuite/ChangeLog
* gcc.target/arc/lsl16-1.c: New test case.
* gcc.target/arc/lsr16-1.c: Likewise.
* gcc.target/arc/swap-1.c: Likewise.
* gcc.target/arc/swap-2.c: Likewise.
Follow up patch to arm: Use deltas for Arm switch tables
This patch moves the switch tables for Arm from the .text section
into the .rodata section.
gcc/ChangeLog:
* config/arm/aout.h: Change to use the Lrtx label.
* config/arm/arm.h (CASE_VECTOR_PC_RELATIVE): Remove arm targets
from (!target_pure_code) condition.
(ADDR_VEC_ALIGN): Add align for tables in rodata section.
* config/arm/arm.cc (arm_output_casesi): Alter the function to include
.Lrtx label and remove adr instructions.
* config/arm/arm.md
(arm_casesi_internal): Use force_reg to generate ldr instructions that
would otherwise be out of range, and change rtl to accommodate force reg.
Additionally remove unnecessary register temp.
(casesi): Remove pure code check for Arm.
* config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Remove arm
targets from JUMP_TABLES_IN_TEXT_SECTION definition.
gcc/testsuite/ChangeLog:
* gcc.target/arm/arm-switchstatement.c: Alter the tests to
change adr instruction to ldr.
Heap-based trampolines are enabled on darwin20 and later,
meaning that no warning is emitted.
gcc/testsuite/ChangeLog:
* gcc.dg/Wtrampolines.c: Skip on darwin20 and later.
The test currently fails on Darwin, where the default arch is core2.
gcc/testsuite/ChangeLog:
PR target/112287
* gcc.target/i386/pr111698.c: Pass -march=sandybridge.
PR106907 has a few warnings spotted from cppcheck. These warnings
are related to the need of precedence clarification. Instead of using xor,
it has been changed to equality check, which achieves the same result.
Additionally, comment indentation has been fixed.
2023-10-11 Jeevitha Palanisamy <jeevitha@linux.ibm.com>
gcc/
PR target/106907
* config/rs6000/rs6000.cc (altivec_expand_vec_perm_const): Change bitwise
xor to an equality and fix comment indentation.
2023-10-30 Paul Thomas <pault@gcc.gnu.org>
gcc/fortran
PR fortran/104555
* resolve.cc (resolve_select_type): If the selector expression
has no class component references and the expression is a
derived type, copy the typespec of the symbol to that of the
expression.
gcc/testsuite/
PR fortran/104555
* gfortran.dg/pr104555.f90: New test.
When 2 vectors are equal, kmask is allones and kortest will set CF,
else CF will be cleared.
So CF bit can be used to check for the result of the comparison.
Before:
vmovdqu (%rsi), %ymm0
vpxorq (%rdi), %ymm0, %ymm0
vptest %ymm0, %ymm0
jne .L2
vmovdqu 32(%rsi), %ymm0
vpxorq 32(%rdi), %ymm0, %ymm0
vptest %ymm0, %ymm0
je .L5
.L2:
movl $1, %eax
xorl $1, %eax
vzeroupper
ret
After:
vmovdqu64 (%rsi), %zmm0
xorl %eax, %eax
vpcmpeqd (%rdi), %zmm0, %k0
kortestw %k0, %k0
setc %al
vzeroupper
ret
gcc/ChangeLog:
PR target/104610
* config/i386/i386-expand.cc (ix86_expand_branch): Handle
512-bit vector with vpcmpeq + kortest.
* config/i386/i386.md (cbranchxi4): New expander.
* config/i386/sse.md: (cbranch<mode>4): Extend to V16SImode
and V8DImode.
gcc/testsuite/ChangeLog:
* gcc.target/i386/pr104610-2.c: New test.
The former patch (f08ca5903c) examines the scalar modes by target
hook scalar_mode_supported_p. It causes some i386 regression cases
as XImode and OImode are not enabled in i386 target function. This
patch examines the scalar mode by checking if the corresponding optabs
are available for the mode.
gcc/
PR target/111449
* expr.cc (qi_vector_mode_supported_p): Rename to...
(by_pieces_mode_supported_p): ...this, and extends it to do
the checking for both scalar and vector mode.
(widest_fixed_size_mode_for_size): Call
by_pieces_mode_supported_p to examine the mode.
(op_by_pieces_d::smallest_fixed_size_mode_for_size): Likewise.
libstdc++-v3/ChangeLog:
* src/experimental/contract.cc
[_GLIBCXX_INLINE_VERSION](handle_contract_violation): Provide symbol
without version namespace decoration for gcc.
Static arrays in D are passed around by value, rather than decaying to a
pointer. On x86_64 __builtin_va_list is an exception to this rule, but
semantically it's still treated as a static array.
This makes certain assignment operations fail due a mismatch in types.
As all examples in the test program are rejected by C/C++ front-ends,
these are now errors in D too to be consistent.
PR d/110712
gcc/d/ChangeLog:
* d-codegen.cc (d_build_call): Update call to convert_for_argument.
* d-convert.cc (is_valist_parameter_type): New function.
(check_valist_conversion): New function.
(convert_for_assignment): Update signature. Add check whether
assigning va_list is permissible.
(convert_for_argument): Likewise.
* d-tree.h (convert_for_assignment): Update signature.
(convert_for_argument): Likewise.
* expr.cc (ExprVisitor::visit (AssignExp *)): Update call to
convert_for_assignment.
gcc/testsuite/ChangeLog:
* gdc.dg/pr110712.d: New test.
The large model is not implemented so far for Darwin (and the
codegen will be different when it is).
gcc/testsuite/ChangeLog:
* gcc.target/i386/large-data.c: Skip for Darwin.
Signed-off-by: Iain Sandoe <iain@sandoe.co.uk>
Darwin platforms do not currently emit .cfi_xxx instructions so that these
tests do not work there.
gcc/testsuite/ChangeLog:
* gcc.target/i386/apx-interrupt-1.c: Skip for Darwin.
* gcc.target/i386/apx-push2pop2-1.c: Likewise.
* gcc.target/i386/apx-push2pop2_force_drap-1.c: Likewise.
Signed-off-by: Iain Sandoe <iain@sandoe.co.uk>
