In OpenACC 'kernels' decomposition, we're improperly nesting synchronous and
asynchronous data and compute regions, giving rise to data races when the
asynchronicity is actually executed, as is visible in at least on test case
with GCN offloading.
The proper fix is to correctly use the asynchronous interfaces, making the
currently synchronous data regions fully asynchronous (see also
<https://gcc.gnu.org/PR97390> "[OpenACC] 'async' clause on 'data' construct",
which is to share the same implementation), but that's for later; for now add
some more synchronization.
gcc/
* omp-oacc-kernels-decompose.cc (add_wait): New function, split out
of...
(add_async_clauses_and_wait): ...here. Call new outlined function.
(decompose_kernels_region_body): Add wait at the end of
explicitly-asynchronous kernels regions.
libgomp/
* testsuite/libgomp.oacc-c-c++-common/f-asyncwait-1.c: Remove GCN
offloading execution XFAIL.
Co-Authored-By: Thomas Schwinge <thomas@codesourcery.com>
... as otherwise 'gcc/omp-low.c:lower_omp_target' has to create a temporary:
13073 else if (is_gimple_reg (var))
13074 {
13075 gcc_assert (offloaded);
13076 tree avar = create_tmp_var (TREE_TYPE (var));
13077 mark_addressable (avar);
..., which (a) is only implemented for actualy *offloaded* regions (but not
data regions), and (b) the subsequently synthesized code for writing to and
later reading back from the temporary fundamentally conflicts with OpenACC
'async' (as used by OpenACC 'kernels' decomposition). That's all not trivial
to make work, so let's just avoid this case.
gcc/
PR middle-end/100280
* omp-oacc-kernels-decompose.cc (maybe_build_inner_data_region):
Mark variables used in synthesized data clauses as addressable.
gcc/testsuite/
PR middle-end/100280
* c-c++-common/goacc/kernels-decompose-pr100280-1.c: New.
* c-c++-common/goacc/classify-kernels-parloops.c: Likewise.
* c-c++-common/goacc/classify-kernels-unparallelized-parloops.c:
Likewise.
* c-c++-common/goacc/classify-kernels-unparallelized.c: Test
'--param openacc-kernels=decompose'.
* c-c++-common/goacc/classify-kernels.c: Likewise.
* c-c++-common/goacc/kernels-decompose-2.c: Update.
* c-c++-common/goacc/kernels-decompose-ice-1.c: Remove.
* c-c++-common/goacc/kernels-decompose-ice-2.c: Likewise.
* gfortran.dg/goacc/classify-kernels-parloops.f95: New.
* gfortran.dg/goacc/classify-kernels-unparallelized-parloops.f95:
Likewise.
* gfortran.dg/goacc/classify-kernels-unparallelized.f95: Test
'--param openacc-kernels=decompose'.
* gfortran.dg/goacc/classify-kernels.f95: Likewise.
libgomp/
PR middle-end/100280
* testsuite/libgomp.oacc-c-c++-common/declare-vla-kernels-decompose-ice-1.c:
Update.
* testsuite/libgomp.oacc-c-c++-common/f-asyncwait-1.c: Likewise.
* testsuite/libgomp.oacc-c-c++-common/kernels-decompose-1.c:
Likewise.
Suggested-by: Julian Brown <julian@codesourcery.com>
gcc/ChangeLog:
* config/epiphany/epiphany.c (epiphany_handle_interrupt_attribute):
Use %qs format specifier.
(epiphany_override_options): Wrap keyword in %<, %>.
From the perspective of the pipeline, `andn + and + ior` version take
2 cycles(AND and ANDN doesn't have dependence), but xor + and + xor
will take 3 cycles.
- xorl %edi, %esi
andl %edx, %esi
- movl %esi, %eax
- xorl %edi, %eax
+ andn %edi, %edx, %eax
+ orl %esi, %eax
gcc/ChangeLog:
PR target/94790
* config/i386/i386.md (*xor2andn): New define_insn_and_split.
gcc/testsuite/ChangeLog:
PR target/94790
* gcc.target/i386/pr94790-1.c: New test.
* gcc.target/i386/pr94790-2.c: Ditto.
7: r120:V4SI=const_vector
8: r121:V4SI=unspec[r120:V4SI,r120:V4SI,0xc] 260
with r121:v4SI = r120:V4SI when r120 is a vector with same element.
gcc/ChangeLog:
* config/rs6000/altivec.md (sldoi_to_mov<mode>): New.
gcc/testsuite/ChangeLog:
* gcc.target/powerpc/sldoi_to_mov.c: New test.
The testcase is x86 specific, other targets have different costs defined.
2022-01-12 Uroš Bizjak <ubizjak@gmail.com>
gcc/testsuite/ChangeLog:
PR target/103935
* g++.dg/vect/slp-pr98855.cc: Compile only for x86 targets.
Add CC clobber to 32-bit vector mode logic insns to allow variants with
general-purpose registers. Also improve ix86_sse_movcc to emit insn with
CC clobber for narrow vector modes in order to re-enable conditional moves
for 16-bit and 32-bit narrow vector modes with -msse2.
2022-01-12 Uroš Bizjak <ubizjak@gmail.com>
gcc/ChangeLog:
PR target/100637
PR target/103861
* config/i386/i386-expand.c (ix86_emit_vec_binop): New static function.
(ix86_expand_sse_movcc): Use ix86_emit_vec_binop instead of gen_rtx_X
when constructing vector logic RTXes.
(expand_vec_perm_pshufb2): Ditto.
* config/i386/mmx.md (negv2qi): Disparage GPR alternative a bit.
(<plusminus:insn>v2qi3): Ditto.
(vcond<mode><mode>): Re-enable for TARGET_SSE2.
(vcondu<mode><mode>): Ditto.
(vcond_mask_<mode><mode>): Ditto.
(one_cmpl<VI_32:mode>2): Remove expander.
(one_cmpl<VI_16_32:mode>2): Rename from one_cmplv2qi.
Use VI_16_32 mode iterator.
(one_cmpl<VI_16_32:mode>2 splitters): Use VI_16_32 mode iterator.
Use lowpart_subreg instead of gen_lowpart to create subreg.
(*andnot<VI_16_32:mode>3): Merge from "*andnot<VI_32:mode>" and
"*andnotv2qi3" insn patterns using VI_16_32 mode iterator.
Disparage GPR alternative a bit. Add CC clobber.
(*andnot<VI_16_32:mode>3 splitters): Use VI_16_32 mode iterator.
Use lowpart_subreg instead of gen_lowpart to create subreg.
(*<any_logic:code><VI_16_32:mode>3): Merge from
"*<any_logic:code><VI_32:mode>" and "*<any_logic:code>v2qi3" insn patterns
using VI_16_32 mode iterator. Disparage GPR alternative a bit.
Add CC clobber.
(*<any_logic:code><VI_16_32:mode>3 splitters):Use VI_16_32 mode
iterator. Use lowpart_subreg instead of gen_lowpart to create subreg.
gcc/testsuite/ChangeLog:
PR target/100637
PR target/103861
* g++.target/i386/pr100637-1b.C (dg-options):
Use -msse2 instead of -msse4.1.
* g++.target/i386/pr100637-1w.C (dg-options): Ditto.
* g++.target/i386/pr103861-1.C (dg-options): Ditto.
* gcc.target/i386/pr100637-4b.c (dg-options): Ditto.
* gcc.target/i386/pr103861-4.c (dg-options): Ditto.
* gcc.target/i386/pr100637-1b.c: Remove scan-assembler
directives for logic instructions.
* gcc.target/i386/pr100637-1w.c: Ditto.
* gcc.target/i386/warn-vect-op-2.c:
Update dg-warning for vector logic operation.
Add support for the compiler using XXSPLTIB reg,255 to load all 1's into a
register on power9 and above instead of using VSPLTI{B,H,W} reg,-1.
gcc/testsuite/
2022-01-12 Michael Meissner <meissner@the-meissners.org>
PR testsuite/102935
* gcc.target/powerpc/pr101384-1.c: Update insn regexp for power9
and power10.
On AIX, stat will store inodes in 32bit even when using LARGE_FILES.
If the inode is larger, it will return -1 in st_ino.
Thus, in incpath.c when comparing include directories, if several
of them have 64bit inodes, they will be considered as duplicated.
gcc/ChangeLog:
2022-01-12 Clément Chigot <clement.chigot@atos.net>
* configure.ac: Check sizeof ino_t and dev_t.
(HOST_STAT_FOR_64BIT_INODES): New AC_DEFINE to provide stat
syscall being able to handle 64bit inodes.
* config.in: Regenerate.
* configure: Regenerate.
* incpath.c (HOST_STAT_FOR_64BIT_INODES): New define.
(remove_duplicates): Use it.
libcpp/ChangeLog:
2022-01-12 Clément Chigot <clement.chigot@atos.net>
* configure.ac: Check sizeof ino_t and dev_t.
* config.in: Regenerate.
* configure: Regenerate.
* include/cpplib.h (INO_T_CPP): Change for AIX.
(DEV_T_CPP): New macro.
(struct cpp_dir): Use it.
This adjusts the CFG cleanup flow back to what it was before the
last change which fixes the observed regression of 541.leela_r with
LTO and FDO.
2022-01-12 Richard Biener <rguenther@suse.de>
PR tree-optimization/103990
* tree-pass.h (tail_merge_optimize): Drop unused argument.
* tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
* tree-ssa-pre.c (pass_pre::execute): Retain TODO_cleanup_cfg
and adjust call to tail_merge_optimize.
GCC 10 gained the "access" function and type attribute, which
optionally can take a size-index param:
https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html
-fanalyzer in trunk (for GCC 12) has gained a -Wanalyzer-tainted-size to
complain about attacker-controlled size values, but this was only being
used deep inside the region-model code when handling the hardcoded known
behavior of certain functions (memset, IIRC).
This patch extends -Wanalyzer-tainted-size to also complain about
unsanitized attacker-controlled values being passed to function
parameters marked as a size via the "access" attribute.
Note that -fanalyzer-checker=taint is currently required in
addition to -fanalyzer to use this warning, due to scaling issues
(see bug 103533).
gcc/analyzer/ChangeLog:
PR analyzer/103940
* engine.cc (impl_sm_context::impl_sm_context): Add
"unknown_side_effects" param and use it to initialize
new m_unknown_side_effects field.
(impl_sm_context::unknown_side_effects_p): New.
(impl_sm_context::m_unknown_side_effects): New.
(exploded_node::on_stmt): Pass unknown_side_effects to sm_ctxt
ctor.
* sm-taint.cc: Include "stringpool.h" and "attribs.h".
(tainted_size::tainted_size): Drop "dir" param.
(tainted_size::get_kind): Drop "FINAL".
(tainted_size::emit): Likewise.
(tainted_size::m_dir): Drop unused field.
(class tainted_access_attrib_size): New subclass.
(taint_state_machine::on_stmt): Call check_for_tainted_size_arg on
external functions with unknown side effects.
(taint_state_machine::check_for_tainted_size_arg): New.
(region_model::check_region_for_taint): Drop "dir" param from
tainted_size ctor.
* sm.h (sm_context::unknown_side_effects_p): New.
gcc/testsuite/ChangeLog:
PR analyzer/103940
* gcc.dg/analyzer/taint-size-access-attr-1.c: New test.
Signed-off-by: David Malcolm <dmalcolm@redhat.com>
Add a few test-cases that test expansion of __atomic_exchange.
Tested on nvptx.
gcc/testsuite/ChangeLog:
2022-01-12 Tom de Vries <tdevries@suse.de>
* gcc.target/nvptx/atomic-exchange-1.c: New test.
* gcc.target/nvptx/atomic-exchange-2.c: New test.
* gcc.target/nvptx/atomic-exchange-3.c: New test.
* gcc.target/nvptx/atomic-exchange-4.c: New test.
Fix a few issues in test-cases gcc.target/nvptx/atomic_fetch-*.c:
- atomic_fetch-1.c uses scan-assembler instead of scan-assembler-times,
which is less accurate
- atomic_fetch-2.c only contains negative testing using
scan-assembler-not
- the test-cases use stack variables to generate generic addresses,
while stack atomics are not natively supported
- the test-cases only test (64-bit) x (generic), instead of
(32-bit, 64-bit) x (generic, global, shared)
- the test-cases use a hardcoded '0' instead of the clearer
MEMMODEL_RELAXED
Tested on nvptx.
gcc/testsuite/ChangeLog:
2022-01-12 Tom de Vries <tdevries@suse.de>
* gcc.target/nvptx/atomic_fetch-1.c: Rewrite.
* gcc.target/nvptx/atomic_fetch-2.c: Rewrite.
We currently crash when the floating-point to_chars overloads are passed
a precision value near INT_MAX, ultimately due to overflow in the bounds
checks that verify the output range is large enough.
The simplest portable fix seems to be to replace bounds checks of the form
A >= B + C (where B + C may overflow) with the otherwise equivalent check
A >= B && A - B >= C, which is the approach this patch takes.
Before we could do this in __floating_to_chars_hex, there we first need
to track the unbounded "excess" precision (i.e. the number of trailing
fractional digits in the output that are guaranteed to be '0') separately
from the bounded "effective" precision (i.e. the number of significant
fractional digits in the output), like we do in __f_t_c_precision.
PR libstdc++/103955
libstdc++-v3/ChangeLog:
* src/c++17/floating_to_chars.cc (__floating_to_chars_hex):
Track the excess precision separately from the effective
precision. Avoid overflow in bounds check by splitting it into
two checks.
(__floating_to_chars_precision): Avoid overflow in bounds checks
similarly.
* testsuite/20_util/to_chars/103955.cc: New test.
Fixes:
In file included from ./tm.h:23,
from gcc/genconfig.c:25:
gcc/config/elfos.h:209: warning: "READONLY_DATA_SECTION_ASM_OP" redefined
209 | #define READONLY_DATA_SECTION_ASM_OP "\t.section\t.rodata"
|
In file included from ./tm.h:21,
from gcc/genconfig.c:25:
gcc/config/epiphany/epiphany.h:671: note: this is the location of the previous definition
671 | #define READONLY_DATA_SECTION_ASM_OP "\t.section .rodata"
gcc/ChangeLog:
* config.gcc: Include elfos.h before ${tm_file}.
-fsignaling-nans is already passed by ieee.exp, so it's not needed.
We must use dg-additional-options instead of dg-options, otherwise we
override flags passed from ieee.exp. And we need to use -w because
some options only make sense for the Fortran source.
gcc/testsuite/ChangeLog:
* gfortran.dg/ieee/signaling_1.f90: Adjust flags.
This is maybe just a shot in the dark, but IMHO we shouldn't be diagnosing
-Wuseless-cast on casts the compiler adds on its own when calling its move
function. We don't seem to warn when user calls std::move either.
We call move on elinit (*NON_LVALUE_EXPR <(struct C[2] &&) &D.2497->b>)[0]
so it is already an xvalue_p and try to static_cast it to struct C &&.
But we don't warn e.g. on std::move (std::move (whatever)).
Fixed by not doing the static cast and just returning expr from move
if expr is already an xvalue.
2022-01-11 Jakub Jelinek <jakub@redhat.com>
Jason Merrill <jason@redhat.com>
PR c++/103480
* tree.c (move): If expr is xvalue_p, just return expr without
build_static_cast.
* g++.dg/warn/Wuseless-cast2.C: New test.
When the __GLIBC_PREREQ macro isn't defined, the
#if ... && defined __GLIBC_PREREQ && __GLIBC_PREREQ (2, 32)
directive has invalid syntax - the __GLIBC_PREREQ in there evaluates
to 0 and is followed by (2, 32).
2022-01-12 Jakub Jelinek <jakub@redhat.com>
* libgfortran.h (POWER_IEEE128): Use __GLIBC_PREREQ in a separate
#if directive inside of #if ... && defined __GLIBC_PREREQ.
This fixes:
FAIL: c-c++-common/builtin-shufflevector-3.c -Wc++-compat (test for excess errors)
Excess errors:
.../gcc/testsuite/c-c++-common/builtin-shufflevector-3.c:6:1: warning: SSE vector argument without SSE enabled changes the ABI [-Wpsabi]
2022-01-12 Jakub Jelinek <jakub@redhat.com>
PR middle-end/101530
* c-c++-common/builtin-shufflevector-3.c: Add -Wno-psabi to
dg-options.
This adds a testcase for the fixed PR, VN now gets us the transform
via IV equality plus predication.
2022-01-12 Richard Biener <rguenther@suse.de>
PR tree-optimization/76174
* gcc.dg/tree-ssa/pr76174.c: New testcase.
These tweaks are installed to avoid build-warnings for
config/cris/cris.c, like:
x/gcc/config/cris/cris.c: In function 'const char* cris_op_str(rtx)':
x/gcc/config/cris/cris.c:728:23: warning: unquoted identifier or keyword \
'cris_op_str' in format [-Wformat-diag]
728 | internal_error ("MULT case in cris_op_str");
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from ./tm.h:20,
from x/gcc/backend.h:28,
from x/gcc/config/cris/cris.c:26:
x/gcc/config/cris/cris.c: In function 'void cris_expand_return(bool)':
x/gcc/config/cris/cris.h:42:33: warning: unquoted operator '->' in \
format [-Wformat-diag]
42 | do { if (!(x)) internal_error ("CRIS-port assertion failed: " #x); \
} while (0)
x/gcc/config/cris/cris.c:1862:3: note: in expansion of macro 'CRIS_ASSERT'
1862 | CRIS_ASSERT (cfun->machine->return_type != CRIS_RETINSN_RET \
|| !on_stack);
| ^~~~~~~~~~~
x/gcc/config/cris/cris.c: In function 'void cris_option_override()':
x/gcc/config/cris/cris.c:2298:9: warning: space followed by punctuation \
character ':' [-Wformat-diag]
2298 | error ("unknown CRIS version specification in %<-march=%> or "
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2299 | "%<-mcpu=%> : %s", cris_cpu_str);
| ~~~~~~~~~~~~~~~~~
x/gcc/config/cris/cris.c:2334:9: warning: space followed by punctuation \
character ':' [-Wformat-diag]
2334 | error ("unknown CRIS cpu version specification in %<-mtune=%> : %s",
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from ./tm.h:20,
from x/gcc/backend.h:28,
from x/gcc/config/cris/cris.c:26:
x/gcc/config/cris/cris.c: In function 'rtx_def* cris_split_movdx(rtx_def**)':
x/gcc/config/cris/cris.h:42:33: warning: unquoted identifier or keyword \
'GET_CODE' in format [-Wformat-diag]
42 | do { if (!(x)) internal_error ("CRIS-port assertion failed: " #x); \
} while (0)
x/gcc/config/cris/cris.c:2457:3: note: in expansion of macro 'CRIS_ASSERT'
2457 | CRIS_ASSERT (GET_CODE (dest) != SUBREG && GET_CODE (src) != SUBREG);
| ^~~~~~~~~~~
Not that I therefore agree that operators, identifiers and keywords
should have to be dressed up like this for internal error messages;
they were more readable without these garments, if only slightly so.
2022-01-11 Hans-Peter Nilsson <hp@axis.com>
* config/cris/cris.c: Quote identifiers in parameters to error
and internal_error, and remove extraneous spaces with punctuation.
* config/cris/cris.h (CRIS_ASSERT): When passing on stringified
expression to internal_error, pass it as a parameter instead of
appending it to the format part.
Noted by Richard Sandiford in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103974#c7 (thanks!)
Mea culpa: I wrongly thought the default development-level value
("yes,extra") would include everything interesting to normal target
hacking (i.e. as opposed to hacking stuff like GC). I see
rtl-checking is marked as "expensive" and presumably therefore left
out. Maybe it could be split into rtl-static (cheap; catching type
errors including this kind of foulups) and rtl-dynamic (the expensive
parts). I suppose that's for whomever feels a strong enough itch.
A quick (error-prone) grep-and-eyeball in config/ shows this was the
only file missing the parenthesis. This lets cris-elf configured with
--enable-checking=yes,extra,rtl survive make all-gcc.
2022-01-11 Hans-Peter Nilsson <hp@axis.com>
* config/cris/cris.c (cris_postdbr_cmpelim): Parenthesize
parameter to as_a.
Currently, the 3rd parameter of function .DEFERRED_INIT is IS_VLA, which is
not needed at all;
In this patch, we change the 3rd parameter from IS_VLA to the name of the var
decl for the following purposes:
1. Fix (or work around) PR103720:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103720
As confirmed in PR103720, with the current definition of .DEFERRED_INIT,
Dom transformed:
c$a$0_6 = .DEFERRED_INIT (8, 2, 0);
_1 = .DEFERRED_INIT (8, 2, 0);
into:
c$a$0_6 = .DEFERRED_INIT (8, 2, 0);
_1 = c$a$0_6;
which is incorrectly done due to Dom treating the two calls to const function
.DEFERRED_INIT as the same call since all actual parameters are the same.
The same issue has been exposed in PR102608 due to a different optimization VN,
the fix for PR102608 is to specially handle call to .DEFERRED_INIT in VN to
exclude it from CSE.
To fix PR103720, we could do the same as the fix to PR102608 to specially
handle call to .DEFERRED_INIT in Dom to exclude it from being optimized.
However, in addition to Dom and VN, there should be other optimizations that
have the same issue as PR103720 or PR102608 (As I built Linux kernel with
-ftrivial-auto-var-init=zero -Werror, I noticed a bunch of bugos warnings).
Other than identifying all the optimizations and specially handling call to
.DEFERRED_INIT in all these optimizations, changing the 3rd parameter of the
function .DEFERRED_INIT from IS_VLA to the name string of the var decl might
be a better workaround (or a fix). After this change, since the 3rd actual
parameter is the name string of the variable, different calls for different
variables will have different name strings as the 3rd actual, As a result, the
optimization that previously treated the different calls to .DEFERRED_INIT as
the same will be prevented.
2. Prepare for enabling -Wuninitialized + -ftrivail-auto-var-init for address
taken variables.
As discussion in the following thread:
https://gcc.gnu.org/pipermail/gcc-patches/2021-August/577431.html
With the current implemenation of -ftrivial-auto-var-init and uninitialized
warning analysis, the uninitialized warning for an address taken auto variable
might be missed since the variable is completely eliminated by optimization and
replaced with a temporary variable in all the uses.
In order to improve such situation, changing the 3rd parameter of the function
.DEFERRED_INIT to the name string of the variable will provide necessary
information to uninitialized warning analysis to make the missing warning
possible.
gcc/ChangeLog:
2022-01-11 qing zhao <qing.zhao@oracle.com>
* gimplify.c (gimple_add_init_for_auto_var): Delete the 3rd argument.
Change the 3rd argument of function .DEFERRED_INIT to the name of the
decl.
(gimplify_decl_expr): Delete the 3rd argument when call
gimple_add_init_for_auto_var.
* internal-fn.c (expand_DEFERRED_INIT): Update comments to reflect
the 3rd argument change of function .DEFERRED_INIT.
* tree-cfg.c (verify_gimple_call): Update comments and verification
to reflect the 3rd argument change of function .DEFERRED_INIT.
* tree-sra.c (generate_subtree_deferred_init): Delete the 3rd argument.
(sra_modify_deferred_init): Change the 3rd argument of function
.DEFERRED_INIT to the name of the decl.
gcc/testsuite/ChangeLog:
2022-01-11 qing zhao <qing.zhao@oracle.com>
* c-c++-common/auto-init-1.c: Adjust testcase to reflect the 3rd
argument change of function .DEFERRED_INIT.
* c-c++-common/auto-init-10.c: Likewise.
* c-c++-common/auto-init-11.c: Likewise.
* c-c++-common/auto-init-12.c: Likewise.
* c-c++-common/auto-init-13.c: Likewise.
* c-c++-common/auto-init-14.c: Likewise.
* c-c++-common/auto-init-15.c: Likewise.
* c-c++-common/auto-init-16.c: Likewise.
* c-c++-common/auto-init-2.c: Likewise.
* c-c++-common/auto-init-3.c: Likewise.
* c-c++-common/auto-init-4.c: Likewise.
* c-c++-common/auto-init-5.c: Likewise.
* c-c++-common/auto-init-6.c: Likewise.
* c-c++-common/auto-init-7.c: Likewise.
* c-c++-common/auto-init-8.c: Likewise.
* c-c++-common/auto-init-9.c: Likewise.
* c-c++-common/auto-init-esra.c: Likewise.
* c-c++-common/auto-init-padding-1.c: Likewise.
* gcc.target/aarch64/auto-init-2.c: Likewise.
Here is a patch to fix up the ppc64be vs. ppc64le byteswapping
of IBM extended real(kind=16) and complex(kind=16).
Similarly to the BT_COMPLEX case it halves size and doubles nelems
for the bswap_array calls. Of course for r16_ibm and r16_ieee conversions
one needs to make sure it is only done when the on file data is in that
format and not in IEEE quad.
2022-01-11 Jakub Jelinek <jakub@redhat.com>
* io/transfer.c (unformatted_read, unformatted_write): When
byteswapping IBM extended real(kind=16), handle it as byteswapping
two real(kind=8) values.
This patch handles the environment variables for the REAL(KIND=16)
variables like for the little/big-endian routines, so users without
who have no access to the source or are unwilling to recompile
can use this.
Syntax is, for example
GFORTRAN_CONVERT_UNIT="r16_ieee:10;little_endian:10" ./a.out
libgfortran/ChangeLog:
* runtime/environ.c (R16_IEEE): New macro.
(R16_IBM): New macro.
(next_token): Handle IBM R16 conversion cases.
(push_token): Likewise.
(mark_single): Likewise.
(do_parse): Likewise, initialize endian.
This patch, based on Jakub's work, implements the CONVERT
specifier for the power-ieee128 brach. It allows specifying
the conversion as r16_ieee,big_endian and the other way around,
based on a table. Setting the conversion via environment
variable and via program option does not yet work.
gcc/ChangeLog:
* flag-types.h (enum gfc_convert): Add flags for
conversion.
gcc/fortran/ChangeLog:
* libgfortran.h (unit_convert): Add flags.
libgfortran/ChangeLog:
* Makefile.in: Regenerate.
* io/file_pos.c (unformatted_backspace): Mask off
R16 parts for convert.
* io/inquire.c (inquire_via_unit): Add cases for
R16 parts.
* io/open.c (st_open): Add cases for R16 conversion.
* io/transfer.c (unformatted_read): Adjust for R16 conversions.
(unformatted_write): Likewise.
(us_read): Mask of R16 bits.
(data_transfer_init): Likewiese.
(write_us_marker): Likewise.
I've just tried to build libgfortran on an old glibc system
(gcc112.fsffrance.org) and unfortunately we still have work to do:
[jakub@gcc2-power8 obj38]$ LD_PRELOAD=/home/jakub/gcc/obj38/powerpc64le-unknown-linux-gnu/libgfortran/.libs/libgfortran.so.5.0.0 /bin/true
[jakub@gcc2-power8 obj38]$ LD_BIND_NOW=1 LD_PRELOAD=/home/jakub/gcc/obj38/powerpc64le-unknown-linux-gnu/libgfortran/.libs/libgfortran.so.5.0.0 /bin/true
/bin/true: symbol lookup error: /home/jakub/gcc/obj38/powerpc64le-unknown-linux-gnu/libgfortran/.libs/libgfortran.so.5.0.0: undefined symbol: __atan2ieee128
While we do use some libquadmath APIs:
readelf -Wr /home/jakub/gcc/obj38/powerpc64le-unknown-linux-gnu/libgfortran/.libs/libgfortran.so.5.0.0 | grep QUADMATH
0000000000251268 000005e400000026 R_PPC64_ADDR64 0000000000000000 quadmath_snprintf@QUADMATH_1.0 + 0
0000000000251270 0000030600000026 R_PPC64_ADDR64 0000000000000000 strtoflt128@QUADMATH_1.0 + 0
00000000002502e0 0000011600000015 R_PPC64_JMP_SLOT 0000000000000000 ynq@QUADMATH_1.0 + 0
0000000000250390 0000016000000015 R_PPC64_JMP_SLOT 0000000000000000 sqrtq@QUADMATH_1.0 + 0
0000000000250508 000001fa00000015 R_PPC64_JMP_SLOT 0000000000000000 fmaq@QUADMATH_1.0 + 0
0000000000250530 0000021200000015 R_PPC64_JMP_SLOT 0000000000000000 fabsq@QUADMATH_1.0 + 0
0000000000250760 0000030600000015 R_PPC64_JMP_SLOT 0000000000000000 strtoflt128@QUADMATH_1.0 + 0
0000000000250990 000003df00000015 R_PPC64_JMP_SLOT 0000000000000000 cosq@QUADMATH_1.0 + 0
00000000002509f0 0000040a00000015 R_PPC64_JMP_SLOT 0000000000000000 expq@QUADMATH_1.0 + 0
0000000000250a88 0000045100000015 R_PPC64_JMP_SLOT 0000000000000000 erfcq@QUADMATH_1.0 + 0
0000000000250a98 0000045e00000015 R_PPC64_JMP_SLOT 0000000000000000 jnq@QUADMATH_1.0 + 0
0000000000250ac8 0000047e00000015 R_PPC64_JMP_SLOT 0000000000000000 sinq@QUADMATH_1.0 + 0
0000000000250e38 000005db00000015 R_PPC64_JMP_SLOT 0000000000000000 fmodq@QUADMATH_1.0 + 0
0000000000250e48 000005e000000015 R_PPC64_JMP_SLOT 0000000000000000 tanq@QUADMATH_1.0 + 0
0000000000250e58 000005e400000015 R_PPC64_JMP_SLOT 0000000000000000 quadmath_snprintf@QUADMATH_1.0 + 0
0000000000250f20 0000062900000015 R_PPC64_JMP_SLOT 0000000000000000 copysignq@QUADMATH_1.0 + 0
we don't do it consistently:
readelf -Wr /home/jakub/gcc/obj38/powerpc64le-unknown-linux-gnu/libgfortran/.libs/libgfortran.so.5.0.0 | grep ieee128
0000000000250310 0000012800000015 R_PPC64_JMP_SLOT 0000000000000000 __atan2ieee128 + 0
0000000000250340 0000014200000015 R_PPC64_JMP_SLOT 0000000000000000 __clogieee128 + 0
0000000000250438 000001a300000015 R_PPC64_JMP_SLOT 0000000000000000 __acoshieee128 + 0
00000000002504b8 000001cc00000015 R_PPC64_JMP_SLOT 0000000000000000 __csinieee128 + 0
0000000000250500 000001f300000015 R_PPC64_JMP_SLOT 0000000000000000 __sinhieee128 + 0
0000000000250570 0000022a00000015 R_PPC64_JMP_SLOT 0000000000000000 __asinieee128 + 0
0000000000250580 0000022d00000015 R_PPC64_JMP_SLOT 0000000000000000 __roundieee128 + 0
00000000002505a0 0000023e00000015 R_PPC64_JMP_SLOT 0000000000000000 __logieee128 + 0
00000000002505c8 0000024900000015 R_PPC64_JMP_SLOT 0000000000000000 __tanieee128 + 0
0000000000250630 0000027500000015 R_PPC64_JMP_SLOT 0000000000000000 __ccosieee128 + 0
0000000000250670 0000028a00000015 R_PPC64_JMP_SLOT 0000000000000000 __log10ieee128 + 0
00000000002506c8 000002bd00000015 R_PPC64_JMP_SLOT 0000000000000000 __cexpieee128 + 0
00000000002506d8 000002c800000015 R_PPC64_JMP_SLOT 0000000000000000 __coshieee128 + 0
00000000002509b0 000003ef00000015 R_PPC64_JMP_SLOT 0000000000000000 __truncieee128 + 0
0000000000250af8 000004a600000015 R_PPC64_JMP_SLOT 0000000000000000 __expieee128 + 0
0000000000250b50 000004c600000015 R_PPC64_JMP_SLOT 0000000000000000 __fmodieee128 + 0
0000000000250bb0 000004e700000015 R_PPC64_JMP_SLOT 0000000000000000 __tanhieee128 + 0
0000000000250c38 0000051300000015 R_PPC64_JMP_SLOT 0000000000000000 __acosieee128 + 0
0000000000250ce0 0000055400000015 R_PPC64_JMP_SLOT 0000000000000000 __sinieee128 + 0
0000000000250d60 0000057e00000015 R_PPC64_JMP_SLOT 0000000000000000 __atanieee128 + 0
0000000000250dd8 000005b100000015 R_PPC64_JMP_SLOT 0000000000000000 __sqrtieee128 + 0
0000000000250e98 0000060200000015 R_PPC64_JMP_SLOT 0000000000000000 __cosieee128 + 0
0000000000250eb0 0000060a00000015 R_PPC64_JMP_SLOT 0000000000000000 __atanhieee128 + 0
0000000000250ef0 0000062000000015 R_PPC64_JMP_SLOT 0000000000000000 __asinhieee128 + 0
0000000000250fd8 0000067f00000015 R_PPC64_JMP_SLOT 0000000000000000 __csqrtieee128 + 0
0000000000251038 000006ad00000015 R_PPC64_JMP_SLOT 0000000000000000 __cabsieee128 + 0
All these should for POWER_IEEE128 use atan2q@QUADMATH_1.0 etc.
It seems all these come from f951 compiled sources.
For user code, I think the agreement was if you want to use successfully
-mabi=ieeelongdouble, you need glibc 2.32 or later, which is why the Fortran
FE doesn't conditionalize on whether glibc 2.32 is available or not and just
emits __WHATEVERieee128 entrypoints.
But for Fortran compiled sources in libgfortran, we need to use
__WHATEVERieee128 only if glibc 2.32 or later and WHATEVERq (from
libquadmath) otherwise.
The following patch implements that, adds -fbuilding-libgfortran option
similar to e.g. -fbuilding-libgcc used when building libgcc and if
that option is set and the TARGET_GLIBC_{MAJOR,MINOR} macros indicate
no glibc or glibc older than 2.32, it will use the libquadmath APIs
rather than glibc 2.32 APIs.
2022-01-07 Jakub Jelinek <jakub@redhat.com>
gcc/fortran/
* trans-types.c (gfc_init_kinds): When setting abi_kind to 17, if not
targetting glibc 2.32 or later and -fbuilding-libgfortran, set
gfc_real16_is_float128 and c_float128 in gfc_real_kinds.
(gfc_build_real_type): Don't set c_long_double if c_float128 is
already set.
* trans-intrinsic.c (builtin_decl_for_precision): Don't use
long_double_built_in if gfc_real16_is_float128 and
long_double_type_node == gfc_float128_type_node.
* lang.opt (fbuilding-libgfortran): New undocumented option.
libgfortran/
* Makefile.am (AM_FCFLAGS): Add -fbuilding-libgfortran after
-fallow-leading-underscore.
* Makefile.in: Regenerated.
With Fortran adding support for changing the long double format, this
patch removes the code that only allowed C/C++ to change the long double
format for GLIBC 2.32 and later without a warning.
gcc/
2022-01-05 Michael Meissner <meissner@the-meissners.org>
* config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
checks for only C/C++ front ends before allowing the long double
format to change without a warning.
This test FAILs because
f951: Error: '-mabi=ieeelongdouble' requires full ISA 2.06 support
compiler exited with status 1
FAIL: gfortran.dg/pr47614.f -O0 (test for excess errors)
As powerpc64le* only supports -mcpu=power8 and newer, I think we shouldn't
be testing with that option.
2022-01-04 Jakub Jelinek <jakub@redhat.com>
* gfortran.dg/pr47614.f: Don't use -mcpu=power4 for
powerpc64le*-*-linux*.
Testing found that we also need libquadmath to be built with
-mno-gnu-attribute, otherwise -mabi=ieeelongdouble programs don't link.
2022-01-03 Jakub Jelinek <jakub@redhat.com>
* configure.ac: Set XCFLAGS to -mno-gnu-attribute on
powerpc64le*-linux*.
* configure: Regenerated.
This brings the library to compile with all specific functions.
It also corrects the patsubst patterns so the right files
get the flags.
It was necessary to manually add -D__powerpc64__ because apparently
this is not set for Fortran.
libgfortran/ChangeLog:
* Makefile.am: Correct files for compilation flags. Add
-D__powerpc64__ for Fortran sources. Get kinds.inc from
grep of kinds.h and kinds-override.h.
* Makefile.in: Regenerate.
* config.h.in: Regenerate.
* configure: Regenerate.
* configure.ac: Add -mno-gnu-attribute to compile flags.
* generated/_abs_c17.F90: Regenerate.
* generated/_abs_r17.F90: Regenerate.
* generated/_acos_r17.F90: Regenerate.
* generated/_acosh_r17.F90: Regenerate.
* generated/_aimag_c17.F90: Regenerate.
* generated/_aint_r17.F90: Regenerate.
* generated/_anint_r17.F90: Regenerate.
* generated/_asin_r17.F90: Regenerate.
* generated/_asinh_r17.F90: Regenerate.
* generated/_atan2_r17.F90: Regenerate.
* generated/_atan_r17.F90: Regenerate.
* generated/_atanh_r17.F90: Regenerate.
* generated/_conjg_c17.F90: Regenerate.
* generated/_cos_c17.F90: Regenerate.
* generated/_cos_r17.F90: Regenerate.
* generated/_cosh_r17.F90: Regenerate.
* generated/_dim_r17.F90: Regenerate.
* generated/_exp_c17.F90: Regenerate.
* generated/_exp_r17.F90: Regenerate.
* generated/_log10_r17.F90: Regenerate.
* generated/_log_c17.F90: Regenerate.
* generated/_log_r17.F90: Regenerate.
* generated/_mod_r17.F90: Regenerate.
* generated/_sign_r17.F90: Regenerate.
* generated/_sin_c17.F90: Regenerate.
* generated/_sin_r17.F90: Regenerate.
* generated/_sinh_r17.F90: Regenerate.
* generated/_sqrt_c17.F90: Regenerate.
* generated/_sqrt_r17.F90: Regenerate.
* generated/_tan_r17.F90: Regenerate.
* generated/_tanh_r17.F90: Regenerate.
* kinds-override.h: Adjust to trunk.
Change condition to single line so it can be grepped.
* m4/specific.m4: Make sure that real=kind16 is used
for _r17.F90 and _c17.F90 files.
* m4/specific2.m4: Likewise.
The following patch detects the powerpc64le-linux kind == 16 cases
and for the -mabi=ieeelongdouble case (no matter whether it is the
configured in default or just option used on the command line) uses
_r17 or _c17 instead of _r16 or _c17 in the library API names.
From what I can see, e.g. calls to sin on real(kind = 16) works fine
with or without this patch (we call __builtin_sinl and the backend
uses rs6000_mangle_decl_assembler_name which ensures __sinieee128
is called).
What is clearly still broken is IO, where for
real(kind=16) a
a = 1.0
print *, a
end
we call
_gfortran_transfer_real_write (&dt_parm.0, &a, 16);
for both -mabi=ibmlongdouble and -mabi=ieeelongdouble
I don't remember what was the agreement, do we want
_gfortran_transfer_real_write (&dt_parm.0, &a, 17);
for the ieeelongdouble case, or some new entrypoint for
the abi_kind == 17 real/complex IO?
Also, what about kind stored in array descriptors? Shall we use
there the abi_kind or kind?
I guess at least before the IO case is solved there is no point
in checking the testsuite, too many things will be majorly broken...
2021-12-31 Jakub Jelinek <jakub@redhat.com>
* gfortran.h (gfc_real_info): Add abi_kind member.
(gfc_type_abi_kind): Declare.
* trans-types.c (gfc_init_kinds): Initialize abi_kind.
* intrinsic.c (gfc_type_abi_kind): New function.
(conv_name): Use it.
* iresolve.c (resolve_transformational, gfc_resolve_abs,
gfc_resolve_char_achar, gfc_resolve_acos, gfc_resolve_acosh,
gfc_resolve_aimag, gfc_resolve_and, gfc_resolve_aint, gfc_resolve_all,
gfc_resolve_anint, gfc_resolve_any, gfc_resolve_asin,
gfc_resolve_asinh, gfc_resolve_atan, gfc_resolve_atanh,
gfc_resolve_atan2, gfc_resolve_bessel_n2, gfc_resolve_ceiling,
gfc_resolve_cmplx, gfc_resolve_complex, gfc_resolve_cos,
gfc_resolve_cosh, gfc_resolve_count, gfc_resolve_dble,
gfc_resolve_dim, gfc_resolve_dot_product, gfc_resolve_dprod,
gfc_resolve_exp, gfc_resolve_floor, gfc_resolve_hypot,
gfc_resolve_int, gfc_resolve_int2, gfc_resolve_int8, gfc_resolve_long,
gfc_resolve_log, gfc_resolve_log10, gfc_resolve_logical,
gfc_resolve_matmul, gfc_resolve_minmax, gfc_resolve_maxloc,
gfc_resolve_findloc, gfc_resolve_maxval, gfc_resolve_merge,
gfc_resolve_minloc, gfc_resolve_minval, gfc_resolve_mod,
gfc_resolve_modulo, gfc_resolve_nearest, gfc_resolve_or,
gfc_resolve_real, gfc_resolve_realpart, gfc_resolve_reshape,
gfc_resolve_sign, gfc_resolve_sin, gfc_resolve_sinh, gfc_resolve_sqrt,
gfc_resolve_tan, gfc_resolve_tanh, gfc_resolve_transpose,
gfc_resolve_trigd, gfc_resolve_xor, gfc_resolve_random_number):
Likewise.
* trans-decl.c (gfc_build_intrinsic_function_decls): Likewise.
