| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Prior to this fix, if pp(N) tried to print a "base type only" field, and
Base_Type(N) was not yet set, it would raise an exception, which was
confusing. This patch makes it simply ignore such fields. Similarly
for Impl_Base_Type_Only and Root_Type_Only fields.
We do this by having alternative versions of Base_Type,
Implementation_Base_Type, and Root_Type that return Empty
in error cases, and call these alteratives from Treepr.
We don't want to Base_Type and friends to return Empty;
we want them to blow up when called from anywhere but
Treepr.
gcc/ada/ChangeLog:
* atree.ads (Node_To_Fetch_From_If_Set): Alternative to
Node_To_Fetch_From that returns Empty in error cases.
For use only in Treepr.
* treepr.adb (Print_Entity_Field): Avoid printing field
if Node_To_Fetch_From_If_Set returns Empty.
* einfo-utils.ads (Base_Type_If_Set): Alternative to
Base_Type that returns Empty in error cases.
(Implementation_Base_Type_If_Set): Likewise.
(Root_Type_If_Set): Likewise.
(Underlying_Type): Use more accurate result subtype.
* einfo-utils.adb (Base_Type): Add Asserts.
(Implementation_Base_Type): Add Assert; minor cleanup.
(Root_Type): Add Assert; minor cleanup. Remove Assert that
is redundant with predicate.
(Base_Type_If_Set): Body of new function.
(Implementation_Base_Type_If_Set): Body of new function.
(Root_Type_If_Set): Body of new function.
|
|
The QNX Certified Products Defect Notification from February 2025
mentions a potential memory leak when pthread_create is interrupted by a
signal. It recommends to disable signals for this function call.
gcc/ada/ChangeLog:
* adaint.c: Add functions to disable and enable signals on QNX.
* libgnarl/s-taprop__qnx.adb (Create_Task): Disable
signals when calling pthread_create.
|
|
equality
The initial implementation of the warning resulted in unwanted false
positives for types that have a user-defined equality function (in
which case abstract equality on components will typically not ever
be invoked). The conditions for reporting the warning are refined
by this change to exclude checking for presence of abstract component
equality functions in the case where the containing type has a user-defined
equality.
gcc/ada/ChangeLog:
* exp_ch4.adb (Expand_N_Op_Eq): Test for absence of user-defined
equality on type being compared (for both array and record types)
as a condition for checking for abstract equality on component
types. Add a "???" comment about current limitations on issuing
the new warning.
(Warn_On_Abstract_Equality_For_Component): Remove temporary disabling
of the warning. Improve comment on declaration.
|
|
This patch fixes a reference to an Ada RM clause, fixes an occurrence of
"unconstrained" that should have been "indefinite", and removes an
incorrect claim that completing a partial view without discriminants
with a full view with defaulted discriminants is an error situation.
gcc/ada/ChangeLog:
* sem_ch3.adb (Process_Discriminants, Process_Full_View): Fix
comments.
|
|
Unfolding of static expressions is needed when evaluating static bounds, even in
the presence of strict analysis. Otherwise, we may wrongly identify static
predicates as dynamic ones, and thus require unnecessary "others" default case.
gcc/ada/ChangeLog:
* sem_attr.adb (Eval_Attribute): Remove strict analysis condition.
|
|
unit naming
This patch improves the warning message when the actual file name of a child
package with a single letter parent isn't the expected one because it may
collide with a predefined unit name. The warning explain why in this specific
case the expected name is not the standard one using the minus to separate
parents with children.
gcc/ada/ChangeLog:
* par-load.adb (Load): Better warning message.
|
|
When creating the local functions to implement the Valid_Scalars
attribute for array and record types, set a Related_Expression that
points to the original attribute reference (blah'Valid_Scalars).
This allows the Inspector to give a more user-friendly name
when these functions are called and they are known, for example,
to always return True.
gcc/ada/ChangeLog:
* exp_attr.adb
(Build_Array_VS_Func and Build_Record_VS_Func): Pass in the
Attr as the Related_Node parametr when calling
Make_Temporary for the Func_Id for the array and record
Valid_Scalars local functions.
|
|
The meaning of the return value of Find_Type_Name depends greatly on
whether the declaration it's passed is a completion. This patch adds a
description of this to the documentation comment of Find_Type_Name.
gcc/ada/ChangeLog:
* sem_ch3.ads (Find_Type_Name): Improve documentation comment.
|
|
The new warning is spuriously flagged on membership tests in
vss-xml-implementation-html_writer_data.adb, leading to the GNAT CB
failing, so we disable it until that issue can be resolved.
gcc/ada/ChangeLog:
* exp_ch4.adb (Warn_On_Abstract_Equality_For_Component): Temporarily
disable warning.
|
|
ifcvt likes to emit
(set
(if_then_else)
(ge (reg 1) (reg2))
(reg 1)
(reg 2))
which can be recognized as min/max patterns in the backend.
This patch adds such patterns and the respective iterators as well as a
test.
gcc/ChangeLog:
* config/riscv/bitmanip.md (*<bitmanip_minmax_cmp_insn>_cmp_<mode>3):
New min/max ifcvt pattern.
* config/riscv/iterators.md (minu): New iterator.
* config/riscv/riscv.cc (riscv_noce_conversion_profitable_p):
Remove riscv-specific adjustment.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/zbb-min-max-04.c: New test.
|
|
Before noce_find_if_block processes a block it sets up an if_info
structure that holds the original costs. At that point the costs of
the then/else blocks have not been added so we only care about the
"if" cost.
The code originally used BRANCH_COST for that but was then changed
to COST_N_INSNS (2) - a compare and a jump.
This patch computes the jump costs via
insn_cost (if_info.jump, ...)
under the assumption that the target takes BRANCH_COST into account
when costing a jump instruction.
In noce_convert_multiple_sets we keep track of the need for the initial
CC comparison. If needed for the generated sequence we add its
cost in default_noce_conversion_profitable_p.
gcc/ChangeLog:
* ifcvt.cc (noce_convert_multiple_sets_1): Add use_cond_earliest
param.
(noce_convert_multiple_sets): Set use_cond_earliest.
(noce_process_if_block): Just use original cost.
(noce_find_if_block): Use insn_cost (jump_insn).
gcc/testsuite/ChangeLog:
* gcc.target/riscv/addsieq.c: Remove xfail and expect conversion
through noce_convert_multiple_sets.
* gcc.target/riscv/addsifeq.c: Ditto.
* gcc.target/riscv/addsifge.c: Ditto.
* gcc.target/riscv/addsifgt.c: Ditto.
* gcc.target/riscv/addsifle.c: Ditto.
* gcc.target/riscv/addsiflt.c: Ditto.
* gcc.target/riscv/addsifne.c: Ditto.
* gcc.target/riscv/addsige.c: Ditto.
* gcc.target/riscv/addsigeu.c: Ditto.
* gcc.target/riscv/addsigt.c: Ditto.
* gcc.target/riscv/addsigtu.c: Ditto.
* gcc.target/riscv/addsile.c: Ditto.
* gcc.target/riscv/addsileu.c: Ditto.
* gcc.target/riscv/addsilt.c: Ditto.
* gcc.target/riscv/addsiltu.c: Ditto.
|
|
Hard register constraints are only supported for LRA and not old reload.
Targets hppa, m68k, pdp11, rx, sh, vax do not default to LRA which is
why these tests fail there. Limit the tests to LRA targets, although,
this means that currently on these targets the tests are skipped by
default. Since the tests are about general logic, the extra coverage we
loose by skipping these targets is negligible.
For hppa, register 0 cannot be used as a general register. Therefore,
use temporary registers r20 and r21 instead. Likewise, for AVR use r20
and r24 instead.
gcc/testsuite/ChangeLog:
* gcc.dg/asm-hard-reg-error-4.c: Limit the test to LRA targets.
Use registers r20 and r21 for hppa. Likewise, for AVR use r20
and r24 instead.
* gcc.dg/asm-hard-reg-error-5.c: Ditto.
|
|
can_find_related_mode_p incorrectly handled VLS (Vector Length Specific)
types by using TARGET_MIN_VLEN directly, which is in bits, instead of
converting it to bytes as required.
This patch fixes the issue by dividing TARGET_MIN_VLEN by 8 to convert
from bits to bytes when calculating the number of units for VLS modes.
The fix enables proper vectorization for several test cases:
- zve32f-1.c: Now correctly finds vector mode for SF mode in foo3,
enabling vectorization of an additional loop.
- zve32f_zvl256b-1.c and zve32x_zvl256b-1.c: Added -mrvv-max-lmul=m2
option to handle V8SI[2] (vector array mode) requirements during
vectorizer analysis, which needs V16SI to pass, and V16SI was enabled
incorrectly before.
Changes since V4:
- Fix testsuite, also triaged why changed.
gcc/ChangeLog:
* config/riscv/riscv-selftests.cc (riscv_run_selftests): Call
run_vectorize_related_mode_selftests.
(test_vectorize_related_mode): New function to test
vectorize_related_mode behavior.
(run_vectorize_related_mode_selftests): New function to run all
vectorize_related_mode tests.
(run_vectorize_related_mode_vla_selftests): New function to test
VLA modes.
(run_vectorize_related_mode_vls_rv64gcv_selftests): New function to
test VLS modes on rv64gcv.
(run_vectorize_related_mode_vls_rv32gc_zve32x_zvl256b_selftests):
New function to test VLS modes on rv32gc_zve32x_zvl256b.
(run_vectorize_related_mode_vls_selftests): New function to run all
VLS mode tests.
* config/riscv/riscv-v.cc (can_find_related_mode_p): Fix VLS type
handling by converting TARGET_MIN_VLEN from bits to bytes.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/zve32f-1.c: Update expected
vectorization count from 2 to 3.
* gcc.target/riscv/rvv/autovec/zve32f_zvl256b-1.c: Add
-mrvv-max-lmul=m2 option.
* gcc.target/riscv/rvv/autovec/zve32x_zvl256b-1.c: Add
-mrvv-max-lmul=m2 option.
|
|
Committing r16-3673 cause the test cases in the list to fail.
Make the test pass by modifying the test case or adding compilation
options.
gcc/testsuite/ChangeLog:
* gcc.target/loongarch/cmodel-extreme-1.c: Add -fPIC.
* gcc.target/loongarch/cmodel-extreme-2.c: Likewise.
* gcc.target/loongarch/tls-gd-noplt.c: Likewise.
* gcc.target/loongarch/tls-extreme-macro.c: Likewise.
* gcc.target/loongarch/func-call-medium-2.c: Modify.
* gcc.target/loongarch/func-call-medium-3.c: Modify.
* gcc.target/loongarch/func-call-medium-4.c: Removed.
|
|
|
|
C2Y removes various instances of undefined behavior, typically making
them either constraint violations or implementation-defined.
In many but not all such cases, GCC's existing behavior is compatible
with the C2Y changes. For an initial batch of such cases, add
explicit tests for how GCC behaves in C2Y mode; more such cases will
need tests added in future patches, and there are also some such
changes that will need changes to how GCC behaves, not just new tests.
(Some of the individual examples in these tests may have been
constraint violations even before C2Y.)
Tested for x86_64-pc-linux-gnu.
* gcc.dg/c2y-function-qual-1.c, gcc.dg/c2y-incomplete-1.c,
gcc.dg/c2y-inline-1.c, gcc.dg/c2y-pointer-1.c,
gcc.dg/c2y-register-array-1.c, gcc.dg/c2y-storage-class-1.c,
gcc.dg/c2y-struct-empty-1.c: New tests.
|
|
There are at least two cases where tree-switch-conversion leads
to unpleasant resource allocation:
PR49857
The lookup table lives in RAM. This is the case for all
devices that locate .rodata in RAM, which is for almost
all AVR devices.
PR81540
Code is bloated for 64-bit inputs.
As far as PR49857 is concerned, a target hook that may add an
address-space qualifier to the lookup table is the obvious
solution, though a respective patch has always been rejected by
global maintainers for non-technical reasons.
gcc/
PR target/81540
PR target/49857
* common/config/avr/avr-common.cc: Disable -ftree-switch-conversion.
|
|
That was introduced in commit 42db504c2fb2b460e24ca0e73b8559fc33b3cf33,
over 15 years ago!
gcc/ChangeLog:
* config/xtensa/xtensa.h (TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P):
Change "Xtrnase" in the comment to "Xtensa".
|
|
PR121878 shows a typo in the tt-ascalon-d8's pipeline description that
leads to an ICE. The problem is that the vector define_insn_reservation
patterns test for scalar modes rather than vector modes, meaning the
insns don't get handled correctly. We could correct the modes, but given
we could have multiple VLEN values, the number of modes we'd have to check
can be large and mode iterators are not allowed in the mode attribute check.
Instead, I've removed the mode check and replaced it with a test of the
Selected Elenent Width (SEW).
2025-09-09 Peter Bergner <bergner@tenstorrent.com>
gcc/
PR target/121878
* config/riscv/tt-ascalon-d8.md (tt_ascalon_d8_vec_idiv_half): Test the
Selected Element Width (SEW) rather than the mode.
(tt_ascalon_d8_vec_idiv_single): Likewise.
(tt_ascalon_d8_vec_idiv_double): Likewise.
(tt_ascalon_d8_vec_float_divsqrt_half): Likewise.
(tt_ascalon_d8_vec_float_divsqrt_single): Likewise.
(tt_ascalon_d8_vec_float_divsqrt_double): Likewise.
gcc/testsuite/
PR target/121878
* gcc.target/riscv/pr121878.c: New test.
Signed-off-by: Peter Bergner <bergner@tenstorrent.com>
|
|
The recent addition of Pragma_Unsigned_Base_Range to the enumeration type
Pragma_Id has made it exceed 256 enumeration values and thus overflow the
unsigned 8-bit type used for it on the C side. This should have stopped
the build, except that a glitch in the Snames machinery causes one value
to be dropped on the C side, leaving it at just 256 enumeration values.
This fixes both issues, i.e. ensures that the number of enumeration values
is the same on both sides and bumps the size of the C type to 16 bits.
gcc/ada/
PR ada/121885
* snames.ads-tmpl (Pragma_Id): Rename Unknown_Pragma to
Pragma_Unknown for the sake of XSnamesT.
* snames.adb-tmpl (Get_Pragma_Id): Adjust to above renaming.
* snames.h-tmpl (Attribute_Id): Change underlying type to Byte.
(Get_Attribute_Id): Use Byte as return value.
(Pragma_Id): Change underlying type to Word.
(Get_Pragma_Id): Use Word as return value.
* types.h (Word): New typedef.
* exp_prag.adb (Expand_N_Pragma): Remove useless comment.
* par-prag.adb (Prag): Adjust to above renaming.
* sem_prag.adb (Analyze_Pragma): Likewise.
(Sig_Flags): Likewise.
|
|
The length returned by vect_get_loop_len is REALLEN + BIAS, but was
assumed to be REALLEN - BIAS. If BIAS is -1, this leads to wrong
code.
gcc/ChangeLog:
* tree-vect-loop.cc (vectorizable_live_operation_1): Fix
load/store bias handling.
gcc/testsuite/ChangeLog:
* gcc.dg/vect/nodump-extractlast-1.c: New test.
Signed-off-by: Juergen Christ <jchrist@linux.ibm.com>
|
|
This patch adds a mechanism to keep track whether nested functions require
non-local context because they reference a variable of an enclosing
scope. This is used for extending the existing -Wreturn-address warning
to also warn for such nested functions. Certain exceptions are implemented
for functions that do not requite a non-local context, because they reference
only static variables, named constants, non-local variables in unevaluated
sizeof, typeof or countof operators, or typedef. The logic is based on the
existing infrastructure for determining use of undeclared static functions.
Finally, To make sure that no trampolines are generated even when not using
optimization, we mark the exempt functions with TREE_NO_TRAMPOLINE.
gcc/c/ChangeLog:
* c-tree.h: Add new macro C_DECL_NONLOCAL_CONTEXT and prototype
for mark_decl_used.
* c-typeck.cc (mark_decl_used): New function.
(function_to_pointer_conversion): Mark exempt functions.
(record_maybe_used_decl): Add `address' parameter.
(build_external_ref): Change to mark_decl_used.
(pop_maybe_used): Call mark_decl_used.
(c_mark_addressable): Ditto.
(c_finish_goto_label): Ditto.
(c_finish_return): Add warning.
* c-decl.cc (declspecs_add_type): Ditto.
(grokdeclarator): Don't set C_DECL_REGISTER on function
declarators.
gcc/testsuite/ChangeLog:
* gcc.dg/Wreturn-nested-1.c: New test.
* gcc.dg/Wreturn-nested-2.c: New test.
* gcc.dg/Wtrampolines-2.c: New test.
* gcc.dg/Wtrampolines-3.c: New test.
|
|
This pattern enables the combine pass (or late-combine, depending on the case)
to merge a float_extend'ed vec_duplicate into a minus RTL instruction. The other
minus operand is already wide.
Before this patch, we have four instructions, e.g.:
fcvt.d.s fa0,fa0
vsetvli a5,zero,e64,m1,ta,ma
vfmv.v.f v2,fa0
vfsub.vv v1,v1,v2
After, we get only one:
vfwsub.wf v1,v1,fa0
gcc/ChangeLog:
* config/riscv/autovec-opt.md (*vfwsub_wf_<mode>): New pattern to
combine float_extend + vec_duplicate + vfsub.vv into vfwsub.wf.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/vx_vf/vf-1-f16.c: Add vfwsub.wf.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-1-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-2-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-2-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-3-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-3-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-4-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-4-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_binop.h
(DEF_VF_BINOP_WIDEN_CASE_2, DEF_VF_BINOP_WIDEN_CASE_3): Swap operands.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_vfwsub-run-2-f16.c: New test.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_vfwsub-run-2-f32.c: New test.
|
|
A static member template doesn't always have a DECL_INITIAL, as in the
below testcase, and so checking its TREE_CODE performs a null-pointer
dereference. I don't think we need to specially detect this case as
traits we care about are unlikely to be members, so this just adds the
missing null check.
PR c++/121859
gcc/cp/ChangeLog:
* constraint.cc (maybe_diagnose_standard_trait): Check if expr
is non-null.
gcc/testsuite/ChangeLog:
* g++.dg/cpp2a/concepts-traits5.C: New test.
Signed-off-by: Nathaniel Shead <nathanieloshead@gmail.com>
|
|
This is something that has bothered me for a few years but I've only found
time for it now.
The glob used for finding *_1.* etc. counterparts to the *_0.* tests is too
broad, so if one has say next to *_1.c file also *_1.c~ or *_1.c.~1~
or *_1.c.orig or *_1.c.bak etc. files, lto.exp will report a warning and the
test will fail.
So, e.g. in rpm build if some backported commit in patch form adds some
gcc/testsuite/*.dg/lto/ test and one uses -b option to patch, if one doesn't
remove the backup files, the test will fail.
Looking through all the *.dg/lto/ directories, I only see c, C, ii, f, f90
and d extensions used right now for the *_1.* files (and higher), while for
the *_0.* files also m, mm and f03 extensions are used.
So, the following patch only searches for those (plus for Fortran uses the
extensions searched by the gfortran.dg/lto/ driver, i.e. \[fF\]{,90,95,03,08}
, not just f, f90 and f03).
Tested on x86_64-linux and verified I got exactly the same number of
grep '^Executing.on.host.*_[1-9]\.' testsuite/*/*.log
before/after this patch when doing make check RUNTESTFLAGS=lto.exp
except for 2 new ones which were previously failed because I had backup
files for 2 tests.
2025-09-10 Jakub Jelinek <jakub@redhat.com>
* lib/lto.exp (lto-execute-1): Search for _1.* etc. files
only with a list of known extensions.
|
|
THe lowering of .{ADD,SUB,MUL}_OVERFLOW ifns is optimized, so that we don't
in the common cases uselessly don't create a large _Complex _BitInt
temporary with the first (real) part being the result and second (imag) part
just being a huge 0 or 1, although we still do that if it can't be done.
The optimizable_arith_overflow function checks when that is possible, like
whether the ifn result is used at most twice, once in REALPART_EXPR and once
in IMAGPART_EXPR in the same bb, etc. For IMAGPART_EXPR it then checks
if it has a single use which is a cast to some integral non-bitint type
(usually bool or int etc.). The final check is whether that cast stmt
appears after the REALPART_EXPR (the usual case), in that case it is
optimizable, otherwise it is not (because the lowering for optimizable
ifns of this kind is done at the location of the REALPART_EXPR and it
tweaks the IMAGPART_EXPR cast location at that point, so otherwise it
would be set after use.
Now, we also have an optimization for the REALPART_EXPR lhs being used
in a single stmt - store in the same bb, in that case we don't have to
store the real part result in a temporary but it can go directly into
memory.
Except that nothing checks for the IMAGPART_EXPR cast being before or after
the store in this case, so the following testcase ICEs because we have
a use before a def stmt.
In bar (the function handled right already before this patch) we have
_6 = .SUB_OVERFLOW (y_4(D), x_5(D));
_1 = REALPART_EXPR <_6>;
_2 = IMAGPART_EXPR <_6>;
a = _1;
_3 = (int) _2;
baz (_3);
before the lowering, so we can just store the limbs of the .SUB_OVERFLOW
into the limbs of a variable and while doing that compute the value we
eventually store into _3 instead of the former a = _1; stmt.
In foo we have
_5 = .SUB_OVERFLOW (y_3(D), x_4(D));
_1 = REALPART_EXPR <_5>;
_2 = IMAGPART_EXPR <_5>;
t_6 = (int) _2;
baz (t_6);
a = _1;
and we can't do that because the lowering would be at the a = _1; stmt
and would try to set t_6 to the overflow flag at that point. We don't
need to punt completely and mark _5 as _Complex _BitInt VAR_DECL though
in this case, all we need is not merge the a = _1; store with the
.SUB_OVERFLOW and REALPART_EXPR/IMAGPART_EXPR lowering. So, add _1
to m_names and lower the first 3 stmts at the _1 = REALPART_EXPR <_5>;
location, optimizable_arith_overflow returned non-zero and so the
cast after IMAGPART_EXPR was after it and then a = _1; will copy from
the temporary VAR_DECL to memory.
2025-09-10 Jakub Jelinek <jakub@redhat.com>
PR middle-end/121828
* gimple-lower-bitint.cc (gimple_lower_bitint): For REALPART_EXPR
consumed by store in the same bb and with REALPART_EXPR from
optimizable_arith_overflow, don't add REALPART_EXPR lhs to
the m_names bitmap only if the cast from IMAGPART_EXPR doesn't
appear in between the REALPART_EXPR and the store.
* gcc.dg/bitint-126.c: New test.
|
|
I thought this wouldn't be necessary because RAW_DATA_CST can only appear
inside of (array) CONSTRUCTORs within DECL_INITIAL of TREE_STATIC vars,
so there shouldn't be a need to expand it. Except that we have an
optimization when reading ARRAY_REF from such a CONSTRUCTOR which will
try to expand the constructor if it either can be stored by pieces
(I think that is just fine) or if it is mostly zeros (which is at least
75% of the initializer zeros). Now the second case is I think in some
cases desirable (say 256MB initializer and just 20 elements out of that
non-zero, so clear everything and store 20 elements must be fastest and
short), but could be really bad as well (say 40GB initializer with
10GB non-zero in it, especially if it doesn't result in the original
variable being optimized away). Maybe it would help if expand_constructor
and store_constructor* etc. had some optional argument with addresses into
the original VAR_DECL so that it could be copying larger amounts of data
like larger RAW_DATA_CSTs from there instead of pushing those into new
.rodata again. And another problem is that we apparently expand the
initializes twice, expand_constructor in store_constructor can expand
the stores one and if expand_constructor returns non-NULL, we then
expand_expr the CONSTRUCTOR again. to the same location.
This patch doesn't address either of those issues, just adds RAW_DATA_CST
support to store_constructor for now. For the can_store_by_pieces
cases it stores those by pieces using a new callback very similar to
string_cst_read_str, for the rest (unfortunately) forces it into a memory
and copies from there.
2025-09-10 Jakub Jelinek <jakub@redhat.com>
PR middle-end/121831
* expr.cc (raw_data_cst_read_str): New function.
(store_constructor) <case ARRAY_TYPE>: Handle RAW_DATA_CST.
* g++.dg/lto/pr121831_0.C: New test.
* g++.dg/lto/pr121831_1.C: New test.
|
|
To properly implement __builtin_ffs for SI mode, implement clz and
(for >= z17) ctz for SI mode. Otherwise, gcc falls back to a libcall
which causes problems for Linux kernel code.
Also adjust the C?Z_DEFINED_VALUE_AT_ZERO macros to return 2. Since
the optabs now return exactly the value set by these macros, return
value 2 is more appropriate and leads to better code.
gcc/ChangeLog:
* config/s390/s390.h (CLZ_DEFINED_VALUE_AT_ZERO): Adjust and
return 2.
(CTZ_DEFINED_VALUE_AT_ZERO): Return 2.
* config/s390/s390.md (clzsi2): Implement.
(ctzsi2): Implement.
gcc/testsuite/ChangeLog:
* gcc.dg/vect/pr109011-2.c: Fix expected outcome.
* gcc.dg/vect/pr109011-4.c: Fix expected outcome.
* gcc.target/s390/ffs-1.c: New test.
Signed-off-by: Juergen Christ <jchrist@linux.ibm.com>
|
|
On Fri, Sep 05, 2025 at 09:57:06PM +0200, Matthias Kretz wrote:
> > Hmm, would this fail for x86 long double, which is 80 bits? OK, just
> > checked. It's mangled as 12/16 bytes on i686/x86_64.
>
> It seems that Clang and GCC disagree on mangling 80-Bit long double:
>
> I like Clang's interpretation of https://itanium-cxx-abi.github.io/cxx-abi/
> abi.html#mangle.float better.
This patch changes the mangling of 80-bit long double literals from
24 or 32 digits (on m68k with padding bits in the middle, on x86
at the start) to just 20 hex digits.
2025-09-10 Jakub Jelinek <jakub@redhat.com>
* mangle.cc (write_real_cst): Mangle Intel/Motorola extended
80-bit formats using 20 hex digits instead of 24 or 32.
* g++.target/i386/mangle-ldbl-1.C: New test.
* g++.target/i386/mangle-ldbl-2.C: New test.
* g++.target/m68k/mangle-ldbl-1.C: New test.
* g++.target/m68k/mangle-ldbl-2.C: New test.
|
|
Because of a wrong define_insn for vec_extract_plus a vector access wasn't
combined with a preceeding plus operation which set the offset at which
to perform the vector access even though the instruction offers that capability.
Bootstrapped and regtested on s390x.
gcc/ChangeLog:
* config/s390/vector.md (*vec_extract<mode>_plus_zero_extend):
Fix define insn.
gcc/testsuite/ChangeLog:
* gcc.target/s390/vector/vec-extract-3.c: New test.
Signed-off-by: Maximilian Immanuel Brandtner <maxbr@linux.ibm.com>
|
|
|
|
This patch fixes some issues with import handling.
The main functional change is to allow importing a module's interface
file into its implementation file indirectly. [module.import] p9
forbids an explicit 'import M;' declaration, but there's no provision
against having 'import X;' where module X itself imports M, so this
patch splits up the detection of circular imports to handle this better.
I also updated the errors to be closer to the standard wording.
A second issue I found while testing this is that we don't properly
handle name visibility when a partition implementation unit imports its
primary module interface (g++.dg/modules/part-10). This is resolved by
setting 'module_p' on the primary interface when it gets imported.
Solving this I incidentally removed the assertion that PR121808 was
failing on, which was never valid: we can enter import_module for a
module previously seen as a module-declaration if parsing bails before
declare_module is called. I experimented with guaranteeing that
declare_module always gets called for a module_state generated during
preprocessing if at all possible, but the resulting errors didn't seem a
lot better so I've left it as-is for now.
I did make a small adjustment so that a direct import of a module
doesn't overwrite the location of a module-declaration from earlier in
the file; this is important because preprocessing (and thus the
assigning of these locations) seems to happen for the whole file before
parsing begins, which can otherwise cause confusing locations referring
to later in the file than parsing would otherwise indicate.
PR c++/99682
PR c++/121808
gcc/cp/ChangeLog:
* module.cc (class module_state): Add comment to 'parent'.
(module_state::check_not_purview): Rename to...
(module_state::check_circular_import): ...this. Only handle
interface dependencies, adjust diagnostic message.
(module_state::read_imports): Use new function. Pass location
of import as from_loc instead of the module location.
(module_state::do_import): Set module_p when importing the
primary interface for the current module.
(import_module): Split out check for imports in own unit.
Remove incorrect assertion.
(preprocess_module): Don't overwrite module-decl location with
later import.
gcc/testsuite/ChangeLog:
* g++.dg/modules/circ-1_c.C: Adjust diagnostic.
* g++.dg/modules/mod-decl-1.C: Likewise.
* g++.dg/modules/mod-decl-2_b.C: Likewise.
* g++.dg/modules/pr99174.H: Likewise.
* g++.dg/modules/import-3_a.C: New test.
* g++.dg/modules/import-3_b.C: New test.
* g++.dg/modules/import-3_c.C: New test.
* g++.dg/modules/mod-decl-9.C: New test.
* g++.dg/modules/part-10_a.C: New test.
* g++.dg/modules/part-10_b.C: New test.
* g++.dg/modules/part-10_c.C: New test.
* g++.dg/modules/part-10_d.C: New test.
Signed-off-by: Nathaniel Shead <nathanieloshead@gmail.com>
Reviewed-by: Jason Merrill <jason@redhat.com>
|
|
A lot of places use (*modules)[0] to refer to the module state for the
current TU. This is a bit awkward to type and not particularly clear
what it represents; make a helper function to clarify it.
gcc/cp/ChangeLog:
* module.cc (this_module): New function.
(import_entity_module): Use it.
(trees_out::decl_node): Likewise.
(get_module): Likewise.
(module_state::check_not_purview): Likewise.
(module_state::read_imports): Likewise.
(module_state::read_using_directives): Likewise.
(module_state::read_initial): Likewise.
(get_import_bitmap): Likewise.
(module_may_redeclare): Likewise.
(direct_import): Likewise.
(declare_module): Likewise.
(name_pending_imports): Likewise.
(preprocess_module): Likewise.
(finish_module_processing): Likewise.
(late_finish_module): Likewise.
Signed-off-by: Nathaniel Shead <nathanieloshead@gmail.com>
|
|
PR testsuite/121872
gcc/testsuite/ChangeLog:
* g++.dg/modules/default-arg-4_b.C: Change run to compile.
Signed-off-by: Nathaniel Shead <nathanieloshead@gmail.com>
|
|
gcc/fortran/ChangeLog:
* trans-intrinsic.cc (conv_intrinsic_fstat_lstat_stat_sub): Init
some variables.
|
|
Apparently an explicitly defined operator<=> isn't required to return
std::foo_ordering, so build_min_non_dep_op_overload needs to be able
to rebuild forms of (x <=> y) @ 0 where the @ resolved to a built-in and
in turn isn't expressed as a function call.
PR c++/121779
gcc/cp/ChangeLog:
* tree.cc (build_min_non_dep_op_overload): Handle comparison
operator expressions rewritten from a <=> that returns a
non-class type.
gcc/testsuite/ChangeLog:
* g++.dg/lookup/operator-8.C: Remove outdated comment about
this test failing.
* g++.dg/lookup/operator-8a.C: New test.
Reviewed-by: Jason Merrill <jason@redhat.com>
|
|
The previous definition had all the GFX11 register counts doubled to fix a bug
that was encountered in early testing. This seems to have been a
misunderstanding of the problem (which is no longer reproducible).
gcc/ChangeLog:
* config/gcn/gcn-devices.def: Correct the Max ISA VGPRs counts for
GFX10 and GFX11 devices.
* config/gcn/gcn.cc (gcn_hsa_declare_function_name): Remove the wave64
VGPR count fudge.
|
|
Fix an unrecognised insn ICE that only shows while using builtins at -O0.
gcc/ChangeLog:
* config/gcn/gcn.cc (gcn_expand_builtin_1): Enable the "mode" parameter
and ensure that "target" is a register for most of the builtins.
|
|
PR fortran/82480
gcc/fortran/ChangeLog:
* check.cc (error_unsupported_kind): Helper function to report an
unsupported kind of an argument.
(check_minrange4): Helper function to report if an integer variable
does not have a decimal range of at least four.
(gfc_check_fstat): Adjust checks for generalization of instrinsic
function FSTAT.
(gfc_check_fstat_sub): Likewise for subroutine FSTAT.
(gfc_check_stat): Likewise for functio STAT.
(gfc_check_stat_sub): Likewise for subroutine STAT.
* intrinsic.texi: Document generalized versions of intrinsics
STAT/LSTAT/FSTAT.
* iresolve.cc (gfc_resolve_stat): STAT function result shall have
the same kind as the VALUES argument.
(gfc_resolve_lstat): Likewise for LSTAT.
(gfc_resolve_fstat): Likewise for FSTAT.
(gfc_resolve_stat_sub): Resolve proper library subroutine for STAT.
(gfc_resolve_lstat_sub): Likewise for LSTAT.
* trans-decl.cc (gfc_build_intrinsic_function_decls): Declare
fndecls for required subroutines in runtine library.
* trans-intrinsic.cc (conv_intrinsic_fstat_lstat_stat_sub): Emit
runtime wrapper code for the library functions, taking care of
possible kind conversion of the optional STATUS argument of the
subroutine versions of the intrinsics.
(gfc_conv_intrinsic_subroutine): Use it.
* trans.h (GTY): Declare prototypes.
gcc/testsuite/ChangeLog:
* gfortran.dg/stat_3.f90: Extend argument checking.
* gfortran.dg/stat_4.f90: New test.
|
|
DWARF to CTF translation for type bit sizes was using uint32_t, and for
member offsets was inadvertently using unsigned int via get_AT_unsigned.
For very large struct types, at least one of these could be truncated
causing incorrect encoding of the struct type and member offsets.
Use HOST_WIDE_INT to avoid these truncation issues and fix the encoding
for large structs.
PR debug/121411
gcc/
* dwarf2ctf.cc (ctf_get_AT_data_member_location) Use AT_unsigned
when fetching AT_bit_offset and AT_data_member_location. Simplify.
(ctf_die_bitsize): Return unsigned HOST_WIDE_INT instead of
uint32_t.
(gen_ctf_base_type, gen_ctf_sou_type, gen_ctf_enumeration_type):
Adapt accordingly.
gcc/testsuite/
* gcc.dg/debug/ctf/ctf-struct-3.c: New test.
|
|
This pattern enables the combine pass (or late-combine, depending on the case)
to merge a float_extend'ed vec_duplicate into a minus RTL instruction. Both
minus operands are widened.
Before this patch, we have six instructions, e.g.:
fcvt.d.s fa0,fa0
vsetvli a5,zero,e64,m1,ta,ma
vfmv.v.f v3,fa0
vfwcvt.f.f.v v1,v2
vsetvli zero,zero,e64,m1,ta,ma
vfsub.vv v1,v1,v3
After, we get only one:
vfwsub.vf v1,v2,fa0
gcc/ChangeLog:
* config/riscv/autovec-opt.md (*vfwsub_vf_<mode>): New pattern to
combine float_extend + vec_duplicate + vfwsub.vv into vfwsub.vf.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/vx_vf/vf-1-f16.c: Add vfwsub.vf.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-1-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-2-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-2-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-3-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-3-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-4-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-4-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_binop.h
(DEF_VF_BINOP_WIDEN_CASE_0, DEF_VF_BINOP_WIDEN_CASE_1): Swap operands.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_binop_widen_run.h: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_vfwsub-run-1-f16.c: New test.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_vfwsub-run-1-f32.c: New test.
|
|
When looking at constexpr references, I've noticed staticp handles
COMPONENT_REFs and ARRAY_REFs (the latter if the index is INTEGER_CST),
but not {REAL,IMAG}PART_EXPR. I think that is incorrect and causes
rejection of constexpr (for C++) or static const (for C) addresses
of __real__ or __imag__ parts of static vars.
2025-09-09 Jakub Jelinek <jakub@redhat.com>
PR c++/121678
* tree.cc (staticp): Handle REALPART_EXPR and IMAGPART_EXPR.
* g++.dg/ext/pr121678.C: New test.
* gcc.dg/pr121678.c: New test.
|
|
Say that -Wunterminated-string-initialization is for C and ObjC only
when it appears in the -Wextra documentation.
gcc/ChangeLog:
* doc/invoke.texi (Warning Options): Note that an option
enabled by -Wextra is for C and ObjC only.
|
|
This pattern enables the combine pass (or late-combine, depending on the case)
to merge a float_extend'ed vec_duplicate into a plus RTL instruction. The other
plus operand is already wide.
Before this patch, we have four instructions, e.g.:
fcvt.d.s fa0,fa0
vsetvli a5,zero,e64,m1,ta,ma
vfmv.v.f v2,fa0
vfadd.vv v1,v1,v2
After, we get only one:
vfwadd.wf v1,v1,fa0
gcc/ChangeLog:
* config/riscv/autovec-opt.md (*vfwadd_wf_<mode>): New pattern to
combine float_extend + vec_duplicate + vfadd.vv into vfwadd.wf.
* config/riscv/vector.md
(@pred_single_widen_<plus_minus:optab><mode>_scalar): Swap and reorder
operands to match the RTL emitted by expand.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/vx_vf/vf-1-f16.c: Add vfwadd.wf.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-1-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-2-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-2-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-3-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-3-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-4-f16.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf-4-f32.c: Likewise.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_binop.h: Add support for single
widening variants.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_binop_widen_run.h: Add support
for single widening variants.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_vfwadd-run-2-f16.c: New test.
* gcc.target/riscv/rvv/autovec/vx_vf/vf_vfwadd-run-2-f32.c: New test.
|
|
This reverts commit 1b7bcac0327ccd84f1966c748f4d1aedef64a9c5.
PR target/121785
gcc/
* config/aarch64/aarch64-simd.md (*bcaxqdi4): Delete.
gcc/testsuite/
* gcc.target/aarch64/simd/bcax_d.c: Remove tests for DImode arguments.
|
|
This reverts commit a632becefad29206a980cc080eee74ed808f9cd3.
|
|
When equality is tested for a composite type that has any record
components whose type has an abstract equality function that will
be called as part of the enclosing type's equality, Program_Error
will be raised. We now issue a warning on the equality test,
mentioning the component type whose abstract equality function
will trigger the exception. Note that this is currently only
done for top-level components of the composite type. Another
limitation is that the warning is not issued when the outer
composite type is tagged.
gcc/ada/ChangeLog:
* exp_ch4.adb (Expand_N_Op_Eq): Check for warning about call to
the abstract equality function of a component type, for both array
and record enclosing types.
(Warn_On_Abstract_Equality_For_Component): New procedure to issue
a warning when an abstract equality function of a component type
will be called and result in Program_Error.
|
|
The recent addition of a language extension for the "continue"
nonreserved keyword caused spurious warnings about unreferenced
subprograms for procedures named "Continue", because the call
statements that are generated when identifiers are determined not to
refer to the keyword were considered to come from expansion.
This patch fixes this by marking those generated calls as coming from
source instead.
gcc/ada/ChangeLog:
* sem_ch5.adb (Make_Call): Mark generated nodes as coming from source.
|
|
gcc/ada/ChangeLog:
* sem_ch3.adb (Analyze_Incomplete_Type_Decl): Remove incorrect
comment.
|
|
Some pragmas where the ghostness is based on the argument of
the pragma triggered ghost policy errors because we had not marked
the pragmas as ghost. However we could only do so once the argument
of the pragma was analyzed.
We can safely ignore the policy checks for those pragmas since
if the argument was ghost then the pragma also had to be ghost.
Marking the pragma afterwards as ghost is only for the cleanup of
ignored ghost purposes.
gcc/ada/ChangeLog:
* ghost.adb (Is_OK_Pragma): Use the
Suppressed_Ghost_Policy_Check_Pragma list for ignoring certain
pragmas.
* sem_prag.ads (Suppressed_Ghost_Policy_Check_Pragma): New variable
to store the pragmas that could be ignored when checking for
consitant ghost policy.
|
