Richard Biener [Wed, 8 Jan 2020 14:30:44 +0000 (14:30 +0000)]
re PR tree-optimization/93199 (Compile time hog in sink_clobbers)
2020-01-08 Richard Biener <rguenther@suse.de>
PR middle-end/93199
* tree-eh.c (sink_clobbers): Update virtual operands for
the first and last stmt only. Add a dry-run capability.
(pass_lower_eh_dispatch::execute): Perform clobber sinking
after CFG manipulations and in RPO order to catch all
secondary opportunities reliably.
PR target/93182
* config/avr/avr.opt (-nodevicespecs): New driver option.
* config/avr/driver-avr.c (avr_devicespecs_file): Only issue
"-specs=device-specs/..." if that option is not set.
* doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
arm: Fix rmprofile multilibs when architecture includes +mp or +sec (PR target/93188)
When only the rmprofile multilibs are built, compiling for armv7-a
should select the generic v7 multilibs. This used to work before +sec
and +mp were added to the architecture options but it was broken by
that update. This patch fixes those variants and adds some tests to
ensure that they remain fixed ;-)
PR target/93188
* config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
when only building rm-profile multilibs.
* gcc.target/arm/multilib.exp: Add new tests for rm-profile only.
Thomas Rodgers [Wed, 8 Jan 2020 03:00:40 +0000 (03:00 +0000)]
Rename condition_variable_any wait* methods to match current draft standard
2020-01-07 Thomas Rodgers <trodgers@redhat.com>
* include/std/condition_variable
(condition_variable_any::wait_on): Rename to match current draft
standard.
(condition_variable_any::wait_on_until): Likewise.
(condition_variable_any::wait_on_for): Likewise.
* testsuite/30_threads/condition_variable_any/stop_token/wait_on.cc:
Adjust tests to account for renamed methods.
Feng Xue [Wed, 8 Jan 2020 02:55:00 +0000 (02:55 +0000)]
Find matched aggregate lattice for self-recursive CP (PR ipa/93084)
2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
PR ipa/93084
* ipa-cp.c (self_recursively_generated_p): Find matched aggregate
lattice for a value to check.
(propagate_vals_across_arith_jfunc): Add an assertion to ensure
finite propagation in self-recursive scc.
2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
PR ipa/93084
* gcc.dg/ipa/ipa-clone-3.c: New test.
compiler: fix loopdepth tracking in array slicing expression in escape analysis
In the gc compiler, for slicing an array, its AST has an implicit
address operation node. There isn't such node in the gofrontend
AST. During the escape analysis, we create a fake node to mimic
the gc compiler's behavior. For the fake node, the loopdepth was
not tracked correctly, causing miscompilation. Since this is an
address operation, do the same thing as we do for the address
operator.
This avoids generating a write barrier for code that appears in the
Go1.14beta1 runtime package in (*pageAlloc).sysGrow:
s.summary[l] = s.summary[l][:needIdxLimit]
Andrew Stubbs [Tue, 7 Jan 2020 15:27:50 +0000 (15:27 +0000)]
[amdgcn] Add more modes for vector comparisons
2020-01-07 Andrew Stubbs <ams@codesourcery.com>
gcc/
* config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
(VEC_ALLREG_ALT): New iterator.
(VEC_ALLREG_INT_MODE): New iterator.
(VCMP_MODE): New iterator.
(VCMP_MODE_INT): New iterator.
(vec_cmpu<mode>di): Use VCMP_MODE_INT.
(vec_cmp<u>v64qidi): New define_expand.
(vec_cmp<mode>di_exec): Use VCMP_MODE.
(vec_cmpu<mode>di_exec): New define_expand.
(vec_cmp<u>v64qidi_exec): New define_expand.
(vec_cmp<mode>di_dup): Use VCMP_MODE.
(vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
(vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
(vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
(vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
(vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
(vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
(vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
(vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
(vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
this.
* config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
* config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
Jason Merrill [Tue, 7 Jan 2020 15:05:25 +0000 (10:05 -0500)]
PR c++/47877 - -fvisibility-inlines-hidden and member templates.
DECL_VISIBILITY_SPECIFIED is also true if an enclosing scope has explicit
visibility, and we don't want that to override -fvisibility-inlines-hidden.
So check for the attribute specifically on the function, like we already do
for template argument visibility restriction.
* decl2.c (determine_visibility): -fvisibility-inlines-hidden beats
explicit class visibility for a template.
These came up while building 1.14beta1 while the code was still invalid.
The policy is to not bother committing invalid test cases that cause
compiler crashes.
Normally we only create SVE ACLE functions when arm_sve.h is included.
But for LTO we need to do it at start-up, so that the functions are
already defined when streaming in the LTO objects.
One hitch with doing that is that LTO doesn't yet implement the
simulate_enum_decl langhook. This patch adds a simple default
implementation that it can use.
2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* langhooks-def.h (lhd_simulate_enum_decl): Declare.
(LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
* langhooks.c: Include stor-layout.h.
(lhd_simulate_enum_decl): New function.
* config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
handle_arm_sve_h for the LTO frontend.
(register_vector_type): Cope with null returns from pushdecl.
gcc/testsuite/
* gcc.target/aarch64/sve/pcs/asm_4.c: New test.
[AArch64] Use type attributes to mark types that use the SVE PCS
The SVE port needs to maintain a different type identity for
GNU vectors and "SVE vectors", since the types use different ABIs.
Until now we've done that using pointer equality between the
TYPE_MAIN_VARIANT and the built-in SVE type.
However, as Richard B noted, that doesn't work well for LTO,
where we stream both GNU and SVE types from a file instead of
creating them directly. We need a mechanism for distinguishing
the types using streamed type information.
This patch does that using a new type attribute. This attribute
is only meant to be used for the built-in SVE types and shouldn't
be user-visible. The patch tries to ensure this by including a space
in the attribute name, like we already do for things like "fn spec"
and "omp declare simd".
2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
(aarch64_sve::nvectors_if_data_type): Replace with...
(aarch64_sve::builtin_type_p): ...this.
* config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
(find_vector_type): Delete.
(add_sve_type_attribute): New function.
(lookup_sve_type_attribute): Likewise.
(register_builtin_types): Add an "SVE type" attribute to each type.
(register_tuple_type): Likewise.
(svbool_type_p, nvectors_if_data_type): Delete.
(mangle_builtin_type): Use lookup_sve_type_attribute.
(builtin_type_p): Likewise. Add an overload that returns the
number of constituent vector and predicate registers.
* config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
(aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
instead of aarch64_sve_argument_p.
(aarch64_takes_arguments_in_sve_regs_p): Likewise.
(aarch64_pass_by_reference): Likewise.
(aarch64_function_value_1): Likewise.
(aarch64_return_in_memory): Likewise.
(aarch64_layout_arg): Likewise.
Don't mangle attributes that have a space in their name
The SVE port needs to maintain a different type identity for
GNU vectors and "SVE vectors" even during LTO, since the types
use different ABIs. The easiest way of doing that seemed to be
to use type attributes. However, these type attributes shouldn't
be user-facing; they're just a convenient way of representing the
types internally in GCC.
There are already several internal-only attributes, such as "fn spec"
and "omp declare simd". They're distinguished from normal user-facing
attributes by having a space in their name, which means that it isn't
possible to write them directly in C or C++.
Taking the same approach mostly works well for SVE. The only snag
I've hit so far is that the new attribute needs to (and only exists to)
affect type identity. This means that it would normally get included
in mangled names, to distinguish it from types without the attribute.
However, the SVE ABI specifies a separate mangling for SVE vector types,
rather than using an attribute mangling + a normal vector mangling.
So we need some way of suppressing the attribute mangling for this case.
There are currently no other target-independent or target-specific
internal-only attributes that affect type identity, so this patch goes
for the simplest fix of skipping mangling for attributes whose names
contain a space (which usually wouldn't give a valid symbol anyway).
Other options I thought about were:
(1) Also make sure that targetm.mangled_type returns nonnull.
(2) Check directly for the target-specific name.
(3) Add a new target hook.
(4) Add new information to attribute_spec. This would be very invasive
at this stage, but maybe we should consider replacing all the boolean
fields with flags? That should make the tables slightly easier to
read and would make adding new flags much simpler in future.
2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
gcc/cp/
* mangle.c (mangle_type_attribute_p): New function, split out from...
(write_CV_qualifiers_for_type): ...here. Don't mangle attributes
that contain a space.
Martin Liska [Tue, 7 Jan 2020 09:15:38 +0000 (10:15 +0100)]
Make warn_inline Optimization option.
2020-01-07 Martin Liska <mliska@suse.cz>
PR tree-optimization/92860
* common.opt: Make in Optimization option
as it is affected by -O0, which is an Optimization
option.
* tree-inline.c (tree_inlinable_function_p):
Use opt_for_fn for warn_inline.
(expand_call_inline): Likewise.
2020-01-07 Martin Liska <mliska@suse.cz>
PR tree-optimization/92860
* gcc.dg/pr92860-2.c: New test.
Jakub Jelinek [Tue, 7 Jan 2020 07:14:41 +0000 (08:14 +0100)]
re PR fortran/93162 (gcc/fortran/trans-openmp.c:2469:50: runtime error: load of value 145992800, which is not a valid value for type 'ar_type' since r279628)
PR fortran/93162
* trans-openmp.c (gfc_trans_omp_clauses): Check for REF_ARRAY type
before testing u.ar.type == AR_FULL.
Jakub Jelinek [Tue, 7 Jan 2020 07:13:50 +0000 (08:13 +0100)]
re PR c++/91369 (Implement P0784R7: constexpr new)
PR c++/91369
* constexpr.c (struct constexpr_global_ctx): Add heap_alloc_count
member, initialize it to zero in ctor.
(cxx_eval_call_expression): Bump heap_dealloc_count when deleting
a heap object. Don't cache calls to functions which allocate some
heap objects and don't deallocate them or deallocate some heap
objects they didn't allocate.
* g++.dg/cpp1y/constexpr-new.C: Expect an error explaining why
static_assert failed for C++2a.
* g++.dg/cpp2a/constexpr-new9.C: New test.
Luo Xiong Hu [Tue, 7 Jan 2020 02:38:16 +0000 (02:38 +0000)]
ipa-inline: Adjust condition for caller_growth_limits
Inline should return failure either (newsize > param_large_function_insns)
OR (newsize > limit). Sometimes newsize is larger than
param_large_function_insns, but smaller than limit, inline doesn't return
failure even if the new function is a large function.
Therefore, param_large_function_insns and param_large_function_growth should be
OR instead of AND, otherwise --param large-function-growth won't
work correctly with --param large-function-insns.
gcc/ChangeLog:
2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
* ipa-inline-analysis.c (estimate_growth): Fix typo.
* ipa-inline.c (caller_growth_limits): Use OR instead of AND.
2020-01-06 Michael Meissner <meissner@linux.ibm.com>
* config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
helper function to return the valid addressing formats for a given
hard register and mode.
(rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
Fix bad code of vector extract of PC-relative address with variable element #.
2020-01-06 Michael Meissner <meissner@linux.ibm.com>
* config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
Use 'Q' for doing vector extract from memory.
(vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
memory.
(vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
doing vector extract from memory.
(vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
extract from memory.
Jason Merrill [Tue, 7 Jan 2020 01:07:59 +0000 (20:07 -0500)]
PR c++/92552 - ICE with inherited constrained default ctor.
We set TYPE_HAS_USER_CONSTRUCTOR on the template type in lookup_using_decl,
but we didn't copy it to the instantiation. Setting it in
one_inherited_ctor is too late, as that gets called after we decide whether
to set CLASSTYPE_LAZY_DEFAULT_CTOR. This change affects other testcases as
well; the changes are fixes for the other inherited constructor tests as
well.
* pt.c (instantiate_class_template_1): Copy
TYPE_HAS_USER_CONSTRUCTOR.
* class.c (one_inherited_ctor): Don't set it here.
pa.md: Revert change to use ordered_comparison_operator instead of...
* config/pa/pa.md: Revert change to use ordered_comparison_operator
instead of cmpib_comparison_operator in cmpib patterns.
* config/pa/predicates.md (cmpib_comparison_operator): Revert removal
of cmpib_comparison_operator. Revise comment.
IFN_DIV_POW2 currently requires all elements to be shifted by the
same amount, in a similar way as for WIDEN_LSHIFT_EXPR. This patch
enforces that when building the SLP tree.
If in future targets want to support IFN_DIV_POW2 without this
restriction, we'll probably need the kind of vector-vector/
vector-scalar split that we already have for normal shifts.
2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
in an IFN_DIV_POW2 node to be equal.
Check mask argument's type when vectorising conditional functions
We can't yet vectorise conditional internal functions whose boolean
condition is fed by a data access (or more generally, by a tree of logic
ops in which all the leaves are data accesses). Although we should add
that eventually, we'd need further work to generate good-quality code.
Unlike vectorizable_load and vectorizalbe_store, vectorizable_call
wasn't checking whether the mask had a suitable type, leading to an
ICE on the testcases.
2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
(vect_check_scalar_mask): ...this.
(vectorizable_store, vectorizable_load): Update call accordingly.
(vectorizable_call): Use vect_check_scalar_mask to check the mask
argument in calls to conditional internal functions.
gcc/testsuite/
* gcc.dg/vect/vect-cond-arith-8.c: New test.
* gcc.target/aarch64/sve/cond_fmul_5.c: Likewise.
If aarch64-tune.md was older than gentune.sh or aarch64-cores.def,
we'd try to overwrite it even if the current contents were correct.
This could cause problems with read-only source directories and
could cause spurious copying for rsync --archive.
2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
Depend on...
(s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
to a temporary file and use move-if-change to update the real
file where necessary.
The constraints for CPY /M allowed p0-p15 instead of the intended p0-p7.
This looks like a pasto from the preceding constant pattern, where p0-p15
is allowed.
2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
rather than Upa for CPY /M.
gcc/testsuite/
* gcc.target/aarch64/sve/acle/general/cpy_1.c: New test.
The contents of the <compare> header are not complete unless concepts
are supported, so the feature test macro should depend on the macro for
concepts.
As a result, the std::lexicographical_compare_three_way function will
not be defined unless concepts are supported, so there is no need to
check __cpp_lib_concepts before using concepts in those functions.
* include/bits/stl_algobase.h (__is_byte_iter, __min_cmp)
(lexicographical_compare_three_way): Do not depend on
__cpp_lib_concepts.
* include/std/version (__cpp_lib_three_way_comparison): Only define
when __cpp_lib_concepts is defined.
* libsupc++/compare (__cpp_lib_three_way_comparison): Likewise.
Andrew Pinski [Mon, 6 Jan 2020 01:08:21 +0000 (01:08 +0000)]
New bitfield testcases.
2020-01-05 Andrew Pinski <apinski@marvell.com>
* gcc.c-torture/compile/20200105-1.c: New testcase.
* gcc.c-torture/compile/20200105-2.c: New testcase.
* gcc.c-torture/compile/20200105-3.c: New testcase.
Recent platform linkers will no longer accept linking for a target
OS version less than 10.4. Recent SDKs no longer have the libgcc_s
shims used for 10.4 and 10.5. So we need to adjust tests that expect
these.
gcc/testsuite/ChangeLog:
2020-01-05 Iain Sandoe <iain@sandoe.co.uk>
* gcc.dg/darwin-version-1.c: Adjust test to use different
options for Darwin4-9 and Darwin10+.
Jakub Jelinek [Sun, 5 Jan 2020 12:52:24 +0000 (13:52 +0100)]
re PR target/93141 (Missed optimization : Use of adc when checking overflow)
PR target/93141
* config/i386/i386.md (SWIDWI): New mode iterator.
(DWI, dwi): Add TImode variants.
(addv<mode>4): Use SWIDWI iterator instead of SWI. Use
<general_hilo_operand> instead of <general_operand>. Use
CONST_SCALAR_INT_P instead of CONST_INT_P.
(*addv<mode>4_1): Rename to ...
(addv<mode>4_1): ... this.
(QWI): New mode attribute.
(*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
define_insn_and_split patterns.
(*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
patterns.
(uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
<general_hilo_operand> instead of <general_operand>.
(*addcarry<mode>_1): New define_insn.
(*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
* gcc.target/i386/pr93141-1.c: New test.
* gcc.dg/pr67089-6.c: Expect 16 ADD_OVERFLOW calls even on ia32.
Jakub Jelinek [Sun, 5 Jan 2020 12:50:40 +0000 (13:50 +0100)]
re PR c++/93138 (elaborated type specifier visibility check problem)
PR c++/93138
* parser.c (cp_parser_check_class_key): Disable access checks for the
simple name lookup.
(cp_parser_maybe_warn_enum_key): Likewise. Return early if
!warn_redundant_tags.
Jason Merrill [Fri, 3 Jan 2020 22:10:56 +0000 (17:10 -0500)]
PR c++/93033 - incorrect tree node sharing with array init.
The split_nonconstant_init piece is the only one necessary to fix the
testcase, but it occurred to me that we might as well not split when
-fno-exceptions.
* typeck2.c (split_nonconstant_init): Unshare non-decl.
* cp-gimplify.c (cp_gimplify_init_expr): Only split if -fexceptions.
gdbinit.in: call a function with "call", not "set"
Calling a function foo in gdb as "set foo()" results in a warning.
Disregarding, it looks wrong to call a function with "set". Let's use
"call" instead.
2019-11-14 Konstantin Kharlamov <Hi-Angel@yandex.ru>
* gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
Use "call" instead of "set".
Jonathan Wakely [Fri, 3 Jan 2020 14:44:39 +0000 (14:44 +0000)]
libstdc++: Only use std::compare_three_way when concepts are supported
Clang now supports three-way comparisons. That causes both overloads of
std::lexicographical_compare_three_way to be defined, but the second one
uses std::compare_three_way which depends on concepts. Clang does not
yet support concepts, so the second overload should also depend on
__cpp_lib_concepts.
* include/bits/stl_algobase.h (lexicographical_compare_three_way):
Only define four-argument overload when __cpp_lib_concepts is defined.
Tobias Burnus [Fri, 3 Jan 2020 12:56:46 +0000 (12:56 +0000)]
Fortran] OpenMP/OpenACC – fix more issues with OPTIONAL
gcc/fortran/
* trans-openmp.c (gfc_omp_check_optional_argument): Always return a
Boolean expression; handle unallocated/disassociated actual arguments
as absent if passed to nonallocatable/nonpointer dummy array arguments.
(gfc_build_cond_assign): Change to assume a Boolean expr not a pointer.
(gfc_omp_finish_clause, gfc_trans_omp_clauses): Assign NULL to generated
array-data variable if the argument is absent. Simplify code as
'present' is now a Boolean expression.
libgomp/
* testsuite/libgomp.fortran/optional-map.f90: Add test for
unallocated/disassociated actual arguments to nonallocatable/nonpointer
dummy arguments; those are/shall be regarded as absent arguments.
* testsuite/libgomp.fortran/use_device_ptr-optional-2.f90: Ditto.
* testsuite/libgomp.fortran/use_device_ptr-optional-3.f90: New.
Jakub Jelinek [Fri, 3 Jan 2020 10:14:03 +0000 (11:14 +0100)]
re PR target/93089 (Force mprefer-vector-width=512 in 'e' simd clones)
PR target/93089
* config/i386/i386-options.c (ix86_simd_clone_adjust): If
TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
for 'e' simd clones.
* gcc.target/i386/pr93089-2.c: New test.
* gcc.target/i386/pr93089-3.c: New test.
Jakub Jelinek [Fri, 3 Jan 2020 10:11:17 +0000 (11:11 +0100)]
re PR target/93110 (grub-2.04/grub-core/lib/division.c:28:1: internal compiler error: in extract_insn, at recog.c:2294)
PR target/93110
* config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
instead of gen_int_shift_amount + convert_modes.
Jakub Jelinek [Fri, 3 Jan 2020 09:10:13 +0000 (10:10 +0100)]
re PR rtl-optimization/93088 (Compile time hog on gcc/testsuite/gcc.target/i386/pr56348.c w/ -O3 -funroll-loops -fno-tree-dominator-opts -fno-tree-vrp)
PR rtl-optimization/93088
* loop-iv.c (find_single_def_src): Punt after looking through
128 reg copies for regs with single definitions. Move definitions
to first uses.
compiler, runtime, reflect: generate hash functions only for map keys
Right now we generate hash functions for all types, just in case they
are used as map keys. That's a lot of wasted effort and binary size
for types which will never be used as a map key. Instead, generate
hash functions only for types that we know are map keys.
Just doing that is a bit too simple, since maps with an interface type
as a key might have to hash any concrete key type that implements that
interface. So for that case, implement hashing of such types at
runtime (instead of with generated code). It will be slower, but only
for maps with interface types as keys, and maybe only a bit slower as
the aeshash time probably dominates the dispatch time.
Reorg where we keep the equals and hash functions. Move the hash function
from the key type to the map type, saving a field in every non-map type.
That leaves only one function in the alg structure, so get rid of that and
just keep the equal function in the type descriptor itself.
While we're here, reorganize the rtype struct to more closely match
the gc version.
This is the gofrontend version of https://golang.org/cl/191198.
compiler: split writing of equal and hash functions for types
Separate the generation of type equality and hash functions, rather
than doing them in a single operation.
This is almost entirely a pure refactoring in preparation for
generating hash functions only for types that are map keys. The only
change in generated code is that for types that are the size of
numeric types, but not aligned like numeric types, such as [8]byte,
now use standard hash functions. They previously used special-purpose
hash functions because they required special-purpose equal functions.
projects / thirdparty / gcc.git / log
