[patch, fortran, testsuite, committed] Fix some more coarray stuff
Hi, I have just committed the attached patch to a test case that was failing on the shared coarray branch. Again, the person who wrote the test case depended on only running on a single image :-) Best regards Thomas Correct coarray indices for test case. gcc/testsuite/ChangeLog: * gfortran.dg/coarray/send_char_array_1.f90: Correct coarray indces. diff --git a/gcc/testsuite/gfortran.dg/coarray/send_char_array_1.f90 b/gcc/testsuite/gfortran.dg/coarray/send_char_array_1.f90 index 3e16fa83685..b3caf80b1ad 100644 --- a/gcc/testsuite/gfortran.dg/coarray/send_char_array_1.f90 +++ b/gcc/testsuite/gfortran.dg/coarray/send_char_array_1.f90 @@ -25,28 +25,28 @@ program send_convert_char_array allocate(character(kind=4, len=5)::co_str_k4_arr(4)[*]) ! First check send/copy to self - co_str_k1_scal[1] = str_k1_scal + co_str_k1_scal[this_image()] = str_k1_scal if (co_str_k1_scal /= str_k1_scal // ' ') STOP 1 - co_str_k4_scal[1] = str_k4_scal + co_str_k4_scal[this_image()] = str_k4_scal if (co_str_k4_scal /= str_k4_scal // 4_' ') STOP 2 - co_str_k4_scal[1] = str_k1_scal + co_str_k4_scal[this_image()] = str_k1_scal if (co_str_k4_scal /= str_k4_scal // 4_' ') STOP 3 - co_str_k1_scal[1] = str_k4_scal + co_str_k1_scal[this_image()] = str_k4_scal if (co_str_k1_scal /= str_k1_scal // ' ') STOP 4 - co_str_k1_arr(:)[1] = str_k1_arr + co_str_k1_arr(:)[this_image()] = str_k1_arr if (any(co_str_k1_arr /= ['abc ', 'EFG ', 'klm ', 'NOP '])) STOP 5 - co_str_k4_arr(:)[1] = [4_'abc', 4_'EFG', 4_'klm', 4_'NOP']! str_k4_arr + co_str_k4_arr(:)[this_image()] = [4_'abc', 4_'EFG', 4_'klm', 4_'NOP']! str_k4_arr if (any(co_str_k4_arr /= [4_'abc ', 4_'EFG ', 4_'klm ', 4_'NOP '])) STOP 6 - co_str_k4_arr(:)[1] = str_k1_arr + co_str_k4_arr(:)[this_image()] = str_k1_arr if (any(co_str_k4_arr /= [ 4_'abc ', 4_'EFG ', 4_'klm ', 4_'NOP '])) STOP 7 - co_str_k1_arr(:)[1] = str_k4_arr + co_str_k1_arr(:)[this_image()] = str_k4_arr if (any(co_str_k1_arr /= ['abc ', 'EFG ', 'klm ', 'NOP '])) STOP 8 end program send_convert_char_array
[patch, fortran, committed] Set DECL_ARTIFICIAL on auxiliary variables
Hello world, committed to trunk as obvious and simple. Best regards Thomas Set DECL_ARTIFICIAL on gfortran internal variables. It seems we sometimes use DECL_ARTIFICIAL as choosing between different code paths. In order not to make -fdebug-aux-vars do different things, set DECL_ARTIFICIAL on the variables to avoid these different code paths (and the corresponding regressions). gcc/fortran/ChangeLog: * trans.c (create_var_debug_raw): Set DECL_ARTIFICIAL on variables. diff --git a/gcc/fortran/trans.c b/gcc/fortran/trans.c index fa7fd9d88aa..57a344a63c9 100644 --- a/gcc/fortran/trans.c +++ b/gcc/fortran/trans.c @@ -128,6 +128,9 @@ create_var_debug_raw (tree type, const char *prefix) t = build_decl (input_location, VAR_DECL, get_identifier (name_buf), type); + /* Not setting this causes some regressions. */ + DECL_ARTIFICIAL (t) = 1; + /* We want debug info for it. */ DECL_IGNORED_P (t) = 0; /* It should not be nameless. */
Ping: [PATCH v6] cacheflush.2: Document __builtin___clear_cache() as a more portable alternative
Ping On 12/15/20 2:30 PM, Alejandro Colomar wrote: > Reported-by: Heinrich Schuchardt > Signed-off-by: Alejandro Colomar > Cc: Martin Sebor > Cc: Dave Martin > --- > > v6: > - GCC has always exposed 'void *', as Martin Sebor noted. > It's Clang (and maybe others) that (following GCC's docs) > exposed 'char *'. > > man2/cacheflush.2 | 24 > 1 file changed, 24 insertions(+) > > diff --git a/man2/cacheflush.2 b/man2/cacheflush.2 > index aba625721..7a2eed506 100644 > --- a/man2/cacheflush.2 > +++ b/man2/cacheflush.2 > @@ -86,6 +86,30 @@ On Linux, this call first appeared on the MIPS > architecture, > but nowadays, Linux provides a > .BR cacheflush () > system call on some other architectures, but with different arguments. > +.SH NOTES > +Unless you need the finer grained control that this system call provides, > +you probably want to use the GCC built-in function > +.BR __builtin___clear_cache (), > +which provides a portable interface > +across platforms supported by GCC and compatible compilers: > +.PP > +.in +4n > +.EX > +.BI "void __builtin___clear_cache(void *" begin ", void *" end ); > +.EE > +.in > +.PP > +On platforms that don't require instruction cache flushes, > +.BR __builtin___clear_cache () > +has no effect. > +.PP > +.IR Note : > +On some GCC-compatible compilers, > +the prototype for this built-in function uses > +.I char * > +instead of > +.I void * > +for the parameters. > .SH BUGS > Linux kernels older than version 2.6.11 ignore the > .I addr > -- Alejandro Colomar Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/ http://www.alejandro-colomar.es/
Re: [PATCH] store VLA bounds in attribute access as strings (PR 97172)
On 12/19/20 12:48 AM, Richard Biener via Gcc-patches wrote: On December 19, 2020 1:55:02 AM GMT+01:00, Martin Sebor via Gcc-patches wrote: To keep tree expressions stored by the front end in attribute access for nontrivial VLA bounds from getting corrupted during Gimplification and to avoid breaking the preconditions verified by the LTO streamer that no such trees exist in the IL, the attached patch replaces those bounds with a string representation of those expressions (as STRING_CST). It also tweaks the pretty-printer to improve the formatting of the VLA bounds and avoid inserting spurious spaces in some cases. The strings are only used by the front end to verify that redeclarations of the same function match in the form and bounds of their VLA arguments, and they're not needed in the middle end. I considered removing them just before the front end finishes but I couldn't find an efficient way to do that. Is there some data structure that stores all function declarations in a translation unit? If there is, then traversing it and removing the attribute arguments might also be an option, either in addition to this change or in lieu of it. There is the free lang data pass in tree.c which walks all reachable tree nodes. You said in response to Honza's suggestion in pr97172 to do that: The frontend needs to make sure no frontend specific tree codes leak into GENERIC so GENERICization is the place where the FE should clear those. FLD is too late and doing it there would be a hack. Are you now saying you've had a change of heart and that doing it there isn't a hack after all? Martin The patch was tested on x86_64-linux. Martin
[committed] doc: Fix a typo [PR98400]
Hi!
Fix a typo in an option name in documentation.
Committed to trunk as obvious.
2020-12-20 Jakub Jelinek
PR other/98400
* doc/invoke.texi (-mbackchain): Fix a typo - -mmo-backchain ->
-mno-backchain.
--- gcc/doc/invoke.texi
+++ gcc/doc/invoke.texi
@@ -28172,7 +28172,7 @@ the backchain is placed into the topmost word of the
96/160 byte register
save area.
In general, code compiled with @option{-mbackchain} is call-compatible with
-code compiled with @option{-mmo-backchain}; however, use of the backchain
+code compiled with @option{-mno-backchain}; however, use of the backchain
for debugging purposes usually requires that the whole binary is built with
@option{-mbackchain}. Note that the combination of @option{-mbackchain},
@option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
Jakub
[committed] libstdc++: Fix indentation in
libstdc++-v3/ChangeLog:
* include/std/semaphore: Fix indentation.
Tested x86_64-linux. Committed to trunk.
commit 3621a7bb5795ad4e47b30e1a8a06f1abe2b3b036
Author: Jonathan Wakely
Date: Sun Dec 20 18:49:13 2020
libstdc++: Fix indentation in
libstdc++-v3/ChangeLog:
* include/std/semaphore: Fix indentation.
diff --git a/libstdc++-v3/include/std/semaphore
b/libstdc++-v3/include/std/semaphore
index ff83a56af4b..9648c3da7ce 100644
--- a/libstdc++-v3/include/std/semaphore
+++ b/libstdc++-v3/include/std/semaphore
@@ -87,7 +87,8 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
{ return _M_sem._M_try_acquire_until(__atime); }
};
- using binary_semaphore = std::counting_semaphore<1>;
+ using binary_semaphore = std::counting_semaphore<1>;
+
_GLIBCXX_END_NAMESPACE_VERSION
} // namespace
#endif // __cpp_lib_atomic_wait
Re: [PATCH v6] cacheflush.2: Document __builtin___clear_cache() as a more portable alternative
On 12/15/20 2:30 PM, Alejandro Colomar wrote: > Reported-by: Heinrich Schuchardt > Signed-off-by: Alejandro Colomar > Cc: Martin Sebor > Cc: Dave Martin > --- > > v6: > - GCC has always exposed 'void *', as Martin Sebor noted. > It's Clang (and maybe others) that (following GCC's docs) > exposed 'char *'. Thanks, Alex. Patch applied. Cheers, Michael > man2/cacheflush.2 | 24 > 1 file changed, 24 insertions(+) > > diff --git a/man2/cacheflush.2 b/man2/cacheflush.2 > index aba625721..7a2eed506 100644 > --- a/man2/cacheflush.2 > +++ b/man2/cacheflush.2 > @@ -86,6 +86,30 @@ On Linux, this call first appeared on the MIPS > architecture, > but nowadays, Linux provides a > .BR cacheflush () > system call on some other architectures, but with different arguments. > +.SH NOTES > +Unless you need the finer grained control that this system call provides, > +you probably want to use the GCC built-in function > +.BR __builtin___clear_cache (), > +which provides a portable interface > +across platforms supported by GCC and compatible compilers: > +.PP > +.in +4n > +.EX > +.BI "void __builtin___clear_cache(void *" begin ", void *" end ); > +.EE > +.in > +.PP > +On platforms that don't require instruction cache flushes, > +.BR __builtin___clear_cache () > +has no effect. > +.PP > +.IR Note : > +On some GCC-compatible compilers, > +the prototype for this built-in function uses > +.I char * > +instead of > +.I void * > +for the parameters. > .SH BUGS > Linux kernels older than version 2.6.11 ignore the > .I addr > -- Michael Kerrisk Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/ Linux/UNIX System Programming Training: http://man7.org/training/
Re: [PATCH 3/4] libstdc++: Add floating-point std::to_chars implementation
On Thu, 17 Dec 2020, Patrick Palka via Gcc-patches wrote: > > >libstdc++-v3/ChangeLog: > > > > > > * config/abi/pre/gnu.ver: Add new exports. > > > * include/std/charconv (to_chars): Declare the floating-point > > > overloads for float, double and long double. > > > * src/c++17/Makefile.am (sources): Add floating_to_chars.cc. > > > * src/c++17/Makefile.in: Regenerate. > > > * src/c++17/floating_to_chars.cc: New file. > > > (to_chars): Define for float, double and long double. > > > * testsuite/20_util/to_chars/long_double.cc: New test. > > > > Sorry it took so long to review, this is OK for trunk. > > > > The patch needs some minor changes to rebase it on the current trunk: > > The linker script has additions since you send this patch, so the > > context in the patch is wrong and it doesn't apply, and in > > the first line of context in the patch needs to have 'noexcept' added. > > That rebase should be easy though. > > > > I'll look at adding __float128 support for powerpc64le. > > Thanks a lot. I committed the patch series just now, after rebasing > and retesting on x86_64, aarch64 and ppc64le. This breaks with the `vax-netbsdelf' target in GCC compilation: .../libstdc++-v3/src/c++17/floating_to_chars.cc:126:38: error: static assertion failed 126 | static_assert(__DBL_MANT_DIG__ == 53); | ^ make[5]: *** [Makefile:577: floating_to_chars.lo] Error 1 So what's the fallback/alternative for non-IEEE-754 FP targets? Shouldn't we call into libc (possibly with a warning attribute) for the formats we don't have explicit handcoded support for? Maciej
[COMMITTED] gcc: xtensa: implement bswapsi2, bswapdi2 and helpers
2020-12-20 Max Filippov
gcc/
* config/xtensa/xtensa.md (bswapsi2, bswapdi2): New patterns.
gcc/testsuite/
* gcc.target/xtensa/bswap.c: New test.
libgcc/
* config/xtensa/lib1funcs.S (__bswapsi2, __bswapdi2): New
functions.
* config/xtensa/t-xtensa (LIB1ASMFUNCS): Add _bswapsi2 and
_bswapdi2.
---
gcc/config/xtensa/xtensa.md | 21 +
gcc/testsuite/gcc.target/xtensa/bswap.c | 14 +
libgcc/config/xtensa/lib1funcs.S| 39 +
libgcc/config/xtensa/t-xtensa | 1 +
4 files changed, 75 insertions(+)
create mode 100644 gcc/testsuite/gcc.target/xtensa/bswap.c
diff --git a/gcc/config/xtensa/xtensa.md b/gcc/config/xtensa/xtensa.md
index 462a7244a35d..cfb6a546d922 100644
--- a/gcc/config/xtensa/xtensa.md
+++ b/gcc/config/xtensa/xtensa.md
@@ -469,6 +469,27 @@
})
+;; Byte swap.
+
+(define_insn "bswapsi2"
+ [(set (match_operand:SI 0 "register_operand" "=&a")
+ (bswap:SI (match_operand:SI 1 "register_operand" "r")))]
+ "!optimize_size"
+ "ssai\t8\;srli\t%0, %1, 16\;src\t%0, %0, %1\;src\t%0, %0, %0\;src\t%0, %1,
%0"
+ [(set_attr "type""arith")
+ (set_attr "mode""SI")
+ (set_attr "length" "15")])
+
+(define_insn "bswapdi2"
+ [(set (match_operand:DI 0 "register_operand" "=&a")
+ (bswap:DI (match_operand:DI 1 "register_operand" "r")))]
+ "!optimize_size"
+ "ssai\t8\;srli\t%0, %D1, 16\;src\t%0, %0, %D1\;src\t%0, %0, %0\;src\t%0,
%D1, %0\;srli\t%D0, %1, 16\;src\t%D0, %D0, %1\;src\t%D0, %D0, %D0\;src\t%D0,
%1, %D0"
+ [(set_attr "type""arith")
+ (set_attr "mode""DI")
+ (set_attr "length" "27")])
+
+
;; Negation and one's complement.
(define_insn "negsi2"
diff --git a/gcc/testsuite/gcc.target/xtensa/bswap.c
b/gcc/testsuite/gcc.target/xtensa/bswap.c
new file mode 100644
index ..057a3569703f
--- /dev/null
+++ b/gcc/testsuite/gcc.target/xtensa/bswap.c
@@ -0,0 +1,14 @@
+/* { dg-do compile } */
+/* { dg-options "-O1" } */
+
+unsigned long f32(unsigned long v)
+{
+ return __builtin_bswap32(v);
+}
+
+unsigned long long f64(unsigned long long v)
+{
+ return __builtin_bswap64(v);
+}
+
+/* { dg-final { scan-assembler-times "ssai" 2 } } */
diff --git a/libgcc/config/xtensa/lib1funcs.S b/libgcc/config/xtensa/lib1funcs.S
index 77e78885d84b..b19deae14483 100644
--- a/libgcc/config/xtensa/lib1funcs.S
+++ b/libgcc/config/xtensa/lib1funcs.S
@@ -840,5 +840,44 @@ __lshrdi3:
#endif /* L_lshrdi3 */
+#ifdef L_bswapsi2
+ .align 4
+ .global __bswapsi2
+ .type __bswapsi2, @function
+__bswapsi2:
+ leaf_entry sp, 16
+ ssai8
+ srlia3, a2, 16
+ src a3, a3, a2
+ src a3, a3, a3
+ src a2, a2, a3
+ leaf_return
+ .size __bswapsi2, . - __bswapsi2
+
+#endif /* L_bswapsi2 */
+
+
+#ifdef L_bswapdi2
+ .align 4
+ .global __bswapdi2
+ .type __bswapdi2, @function
+__bswapdi2:
+ leaf_entry sp, 16
+ ssai8
+ srlia4, a2, 16
+ src a4, a4, a2
+ src a4, a4, a4
+ src a4, a2, a4
+ srlia2, a3, 16
+ src a2, a2, a3
+ src a2, a2, a2
+ src a2, a3, a2
+ mov a3, a4
+ leaf_return
+ .size __bswapdi2, . - __bswapdi2
+
+#endif /* L_bswapdi2 */
+
+
#include "ieee754-df.S"
#include "ieee754-sf.S"
diff --git a/libgcc/config/xtensa/t-xtensa b/libgcc/config/xtensa/t-xtensa
index 90df5f15daa9..9836c96aefc6 100644
--- a/libgcc/config/xtensa/t-xtensa
+++ b/libgcc/config/xtensa/t-xtensa
@@ -2,6 +2,7 @@ LIB1ASMSRC = xtensa/lib1funcs.S
LIB1ASMFUNCS = _mulsi3 _divsi3 _modsi3 _udivsi3 _umodsi3 \
_umulsidi3 _clz _clzsi2 _ctzsi2 _ffssi2 \
_ashldi3 _ashrdi3 _lshrdi3 \
+ _bswapsi2 _bswapdi2 \
_negsf2 _addsubsf3 _mulsf3 _divsf3 _cmpsf2 _fixsfsi _fixsfdi \
_fixunssfsi _fixunssfdi _floatsisf _floatunsisf \
_sqrtf _recipsf2 _rsqrtsf2 \
--
2.20.1
[PATCH] PR fortran/93685 - [9/10/11 Regression] ICE in gfc_constructor_append_expr, at fortran/constructor.c:135
Dear all,
the patch is almost self-explaining. We mishandled the case of initializing
pointer components to derived types of type CHARACTER via DATA, which screwed
up both legal and illegal uses. The attached patch fixes that, and checks
that the legal cases work.
Regtests cleanly on x86_64-pc-linux-gnu.
OK for master? As it is a regression, I intend to backport as seems applicable.
(As one part of the patch partially reverts sth. that was part of
commit eb401400f59e4d1f28bbdc788c3234e0968081d7
Author: Andre Vehreschild
Date: Wed Dec 14 12:52:09 2016 +0100
re PR fortran/78672 (Gfortran test suite failures with a sanitized compiler)
I'd be glad if somebody else cross-checks on a different platform.)
Thanks,
Harald
PR93685 - ICE in gfc_constructor_append_expr, at fortran/constructor.c:135
Fix handling of F2018 enhancements to DATA statements that allows
initialization of pointer components to derived types, and adjust error
handling for the CHARACTER case.
gcc/fortran/ChangeLog:
* data.c (gfc_assign_data_value): Restrict use of
create_character_initializer to constant initializers.
* trans-expr.c (gfc_conv_initializer): Ensure that character
initializer is constant, otherwise fall through to get the same
error handling as for non-character cases.
gcc/testsuite/ChangeLog:
* gfortran.dg/pr93685_1.f90: New test.
* gfortran.dg/pr93685_2.f90: New test.
diff --git a/gcc/fortran/data.c b/gcc/fortran/data.c
index 3e52a5717b5..76ddd9dab7f 100644
--- a/gcc/fortran/data.c
+++ b/gcc/fortran/data.c
@@ -546,12 +546,11 @@ gfc_assign_data_value (gfc_expr *lvalue, gfc_expr *rvalue, mpz_t index,
return false;
}
- if (ref || last_ts->type == BT_CHARACTER)
+ if (ref || (last_ts->type == BT_CHARACTER
+ && rvalue->expr_type == EXPR_CONSTANT))
{
/* An initializer has to be constant. */
- if (rvalue->expr_type != EXPR_CONSTANT
- || (lvalue->ts.u.cl->length == NULL
- && !(ref && ref->u.ss.length != NULL)))
+ if (lvalue->ts.u.cl->length == NULL && !(ref && ref->u.ss.length != NULL))
return false;
expr = create_character_initializer (init, last_ts, ref, rvalue);
}
diff --git a/gcc/fortran/trans-expr.c b/gcc/fortran/trans-expr.c
index bfe08be2a94..f66afab85d1 100644
--- a/gcc/fortran/trans-expr.c
+++ b/gcc/fortran/trans-expr.c
@@ -7877,12 +7877,14 @@ gfc_conv_initializer (gfc_expr * expr, gfc_typespec * ts, tree type,
return se.expr;
case BT_CHARACTER:
- {
- tree ctor = gfc_conv_string_init (ts->u.cl->backend_decl,expr);
- TREE_STATIC (ctor) = 1;
- return ctor;
- }
+ if (expr->expr_type == EXPR_CONSTANT)
+ {
+ tree ctor = gfc_conv_string_init (ts->u.cl->backend_decl, expr);
+ TREE_STATIC (ctor) = 1;
+ return ctor;
+ }
+ /* Fallthrough. */
default:
gfc_init_se (&se, NULL);
gfc_conv_constant (&se, expr);
diff --git a/gcc/testsuite/gfortran.dg/pr93685_1.f90 b/gcc/testsuite/gfortran.dg/pr93685_1.f90
new file mode 100644
index 000..34d6e2cf355
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/pr93685_1.f90
@@ -0,0 +1,20 @@
+! { dg-do run }
+! PR93685 - ICE in gfc_constructor_append_expr, at fortran/constructor.c:135
+
+program p
+ implicit none
+ type t
+ character, pointer :: a
+ end type t
+ type u
+ integer, pointer :: i
+ end type u
+ type(t) :: x
+ type(u) :: y
+ character, target :: c = 'c'
+ integer , target :: i = 10
+ data x%a /c/
+ data y%i /i/
+ if (x% a /= "c") stop 1
+ if (y% i /= 10) stop 2
+end
diff --git a/gcc/testsuite/gfortran.dg/pr93685_2.f90 b/gcc/testsuite/gfortran.dg/pr93685_2.f90
new file mode 100644
index 000..2e3ef5a62fd
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/pr93685_2.f90
@@ -0,0 +1,18 @@
+! { dg-do compile }
+! PR93685 - ICE in gfc_constructor_append_expr, at fortran/constructor.c:135
+
+program p
+ implicit none
+ type t
+ character :: a
+ end type t
+ type u
+ integer :: i
+ end type u
+ type(t) :: x
+ type(u) :: y
+ character, target :: c = 'c'
+ integer , target :: i = 10
+ data x%a /c/ ! { dg-error "non-constant initialization expression"}
+ data y%i /i/ ! { dg-error "non-constant initialization expression"}
+end
Re: [PATCH] Hurd: Enable ifunc by default
Ping?
Samuel Thibault, le dim. 08 nov. 2020 23:52:51 +0100, a ecrit:
> The binutils bugs seem to have been fixed.
>
> 2020-11-08 Samuel Thibault
>
> gcc/
> * config.gcc: Enable default_gnu_indirect_function in *-*-gnu*
> target (but not *-*-kfreebsd*-gnu | *-*-kopensolaris*-gnu).
> ---
> gcc/config.gcc | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/gcc/config.gcc b/gcc/config.gcc
> index b42ebc4e5be..a347c2cec7c 100644
> --- a/gcc/config.gcc
> +++ b/gcc/config.gcc
> @@ -3538,7 +3538,9 @@ esac
> case ${target} in
> *-*-linux*android*|*-*-linux*uclibc*|*-*-linux*musl*)
> ;;
> -*-*-linux*)
> +*-*-kfreebsd*-gnu | *-*-kopensolaris*-gnu)
> +;;
> +*-*-linux* | *-*-gnu*)
> case ${target} in
> aarch64*-* | arm*-* | i[34567]86-* | powerpc*-* | s390*-* | sparc*-* |
> x86_64-*)
> default_gnu_indirect_function=yes
> --
> 2.20.1
>
[PATCH, fix bootstrap] libcody: Include as needed
libcody as originally added to GCC fails to build on *-unknown-freebsd11.4 and presumably others: c++ -std=c++11 -g -include config.h -I ... -MMD -MP -MF netclient.d -c -o netclient.o .../libcody/netclient.cc:114:3: error: unknown type sockaddr_in6 addr; ^~~~ sockaddr_in6 is declared in /usr/include/netinet6/in6.h, which is included by /usr/include/netinet/in.h. Indeed, per https://man7.org/linux/man-pages/man7/ipv6.7.html #include #include should be included, and our own gcc/ada/gsocket.h also has #if !(defined (VMS) || defined (__MINGW32__)) #include #include #include : #endif Okay to push? Gerald libcody/ChangeLog: 2020-12-21 Gerald Pfeifer * netclient.cc: Include . * netserver.cc: Ditto. --- libcody/netclient.cc | 1 + libcody/netserver.cc | 1 + 2 files changed, 2 insertions(+) diff --git a/libcody/netclient.cc b/libcody/netclient.cc index 8cccface71c..c97ed4ca126 100644 --- a/libcody/netclient.cc +++ b/libcody/netclient.cc @@ -14,6 +14,7 @@ #include #include #include +#include #ifndef AI_NUMERICSERV #define AI_NUMERICSERV 0 diff --git a/libcody/netserver.cc b/libcody/netserver.cc index 7e43eb033f4..4e55089161e 100644 --- a/libcody/netserver.cc +++ b/libcody/netserver.cc @@ -13,6 +13,7 @@ #include #include #include +#include #ifndef AI_NUMERICSERV #define AI_NUMERICSERV 0 -- 2.29.2
Re: C++ 20 modules
On Thu, 17 Dec 2020, Nathan Sidwell wrote: > As yesterday, several issues fixed: Here is a new one, *-unknown-freebsd11.4; with my previous patch applied, libcody builds and we now fail in gcc/cp: c++ -std=c++11 -fno-PIE -c -DIN_GCC_FRONTEND -g -DIN_GCC -fno-strict-aliasing -fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrow ing -Wwrite-strings -Wcast-qual -Wno-error=format-diag -Wno-format -Wmissing-fo rmat-attribute -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macr os -Wno-overlength-strings -fno-common -DHAVE_CONFIG_H -I. -Icp -I/scratch/tmp /gerald/GCC-HEAD/gcc -I/scratch/tmp/gerald/GCC-HEAD/gcc/cp -I/scratch/tmp/geral d/GCC-HEAD/gcc/../include -I./../intl -I/scratch/tmp/gerald/GCC-HEAD/gcc/../lib cpp/include -I/scratch/tmp/gerald/GCC-HEAD/gcc/../libcody -I/home/gerald/11-i38 6/include -I/scratch/tmp/gerald/GCC-HEAD/gcc/../libdecnumber -I/scratch/tmp/ge rald/GCC-HEAD/gcc/../libdecnumber/dpd -I../libdecnumber -I/scratch/tmp/gerald/G CC-HEAD/gcc/../libbacktrace -o cp/mapper-client.o -MT cp/mapper-client.o -MMD -MP -MF cp/.deps/mapper-client.TPo /scratch/tmp/gerald/GCC-HEAD/gcc/cp/mapper-client.cc In file included from /scratch/tmp/gerald/GCC-HEAD/gcc/cp/mapper-client.cc:31: In file included from /scratch/tmp/gerald/GCC-HEAD/gcc/cp/mapper-client.h:26: In file included from /scratch/tmp/gerald/GCC-HEAD/gcc/../libcody/cody.hh:24: In file included from /usr/include/c++/v1/memory:653: /usr/include/c++/v1/typeinfo:346:5: error: no member named 'fancy_abort' in namespace 'std::__1'; did you mean simply 'fancy_abort'? _VSTD::abort(); ^~~ /usr/include/c++/v1/__config:782:15: note: expanded from macro '_VSTD' #define _VSTD std::_LIBCPP_ABI_NAMESPACE ^ Compiler is clang version 10.0.1. Gerald
Re: [PATCH] Hurd: Enable ifunc by default
What do you need exactly? How can we test It?
El lun, 21 dic 2020 a las 0:11, Samuel Thibault ()
escribió:
> Ping?
>
> Samuel Thibault, le dim. 08 nov. 2020 23:52:51 +0100, a ecrit:
> > The binutils bugs seem to have been fixed.
> >
> > 2020-11-08 Samuel Thibault
> >
> > gcc/
> > * config.gcc: Enable default_gnu_indirect_function in *-*-gnu*
> > target (but not *-*-kfreebsd*-gnu | *-*-kopensolaris*-gnu).
> > ---
> > gcc/config.gcc | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/gcc/config.gcc b/gcc/config.gcc
> > index b42ebc4e5be..a347c2cec7c 100644
> > --- a/gcc/config.gcc
> > +++ b/gcc/config.gcc
> > @@ -3538,7 +3538,9 @@ esac
> > case ${target} in
> > *-*-linux*android*|*-*-linux*uclibc*|*-*-linux*musl*)
> > ;;
> > -*-*-linux*)
> > +*-*-kfreebsd*-gnu | *-*-kopensolaris*-gnu)
> > +;;
> > +*-*-linux* | *-*-gnu*)
> > case ${target} in
> > aarch64*-* | arm*-* | i[34567]86-* | powerpc*-* | s390*-* |
> sparc*-* | x86_64-*)
> > default_gnu_indirect_function=yes
> > --
> > 2.20.1
> >
>
>
Re: [PATCH] Hurd: Enable ifunc by default
Almudena Garcia, le lun. 21 déc. 2020 01:05:29 +0100, a ecrit: > What do you need exactly? I need gcc people to apply it. > How can we test It? The answer would be lengthy. Basically recompile gcc with it, write or find a program that uses it, and check that it works. But I have already done that simply by letting glibc use ifunc, and it does work, so there is nothing more to be done on the Hurd developers' side. Samuel
Re: [gcc r11-5968] middle-end: Refactor and expose some vectorizer helper functions.
On Sun, 13 Dec 2020, Tamar Christina via Gcc-cvs wrote: > https://gcc.gnu.org/g:10bbba9145700e2c3b4709c1041c66d8ff85c7f1 > > commit r11-5968-g10bbba9145700e2c3b4709c1041c66d8ff85c7f1 > Author: Tamar Christina > Date: Sun Dec 13 13:53:48 2020 + > > middle-end: Refactor and expose some vectorizer helper functions. Umm, this commit does not build: .../gcc/tree-vect-slp.c:197:1: error: '_slp_tree* vect_create_new_slp_node(vec<_stmt_vec_info*>, unsigned int)' was declared 'extern' and later 'static' [-fpermissive] 197 | vect_create_new_slp_node (vec scalar_stmts, unsigned nops) | ^~~~ In file included from .../gcc/tree-vect-slp.c:39: .../gcc/tree-vectorizer.h:2011:17: note: previous declaration of '_slp_tree* vect_create_new_slp_node(vec<_stmt_vec_info*>, unsigned int)' 2011 | extern slp_tree vect_create_new_slp_node (vec, unsigned); | ^~~~ make[2]: *** [Makefile:1122: tree-vect-slp.o] Error 1 due to swapped hunks with a subsequent change. This disturbs bisection, requiring a manual intervention to proceed. Please be more careful and always compile individual changes in a patch series. Maciej
Re: [PATCH] match.pd: Add ~(X - Y) -> ~X + Y simplification [PR96685]
On Sat, 12 Dec 2020, Jakub Jelinek via Gcc-patches wrote: > This patch adds the ~(X - Y) -> ~X + Y simplification requested > in the PR (plus also ~(X + C) -> ~X + (-C) for constants C that can > be safely negated. This regresses VAX code produced by the cmpelim-eq-notsi.c test case (and its similar counterparts) with the `vax-netbsdelf' target. Previously this assembly: .text .align 1 .globl eq_notsi .type eq_notsi, @function eq_notsi: .word 0 # 35[c=0] procedure_entry_mask subl2 $4,%sp# 46[c=32] *addsi3 mcoml 4(%ap),%r0# 32[c=16] *one_cmplsi2_ccz jeql .L1# 34[c=26] *branch_ccz addl2 $2,%r0# 31[c=32] *addsi3 .L1: ret # 40[c=0] return .size eq_notsi, .-eq_notsi was produced. Now this: .text .align 1 .globl eq_notsi .type eq_notsi, @function eq_notsi: .word 0 # 36[c=0] procedure_entry_mask subl2 $4,%sp# 48[c=32] *addsi3 movl 4(%ap),%r0 # 33[c=16] *movsi_2 cmpl %r0,$-1# 34[c=8] *cmpsi_ccz/1 jeql .L3# 35[c=26] *branch_ccz subl3 %r0,$1,%r0# 32[c=32] *subsi3/1 ret # 27[c=0] return .L3: clrl %r0# 31[c=2] *movsi_2 ret # 41[c=0] return .size eq_notsi, .-eq_notsi is, which is clearly worse, both in terms of performance and size. The key here is that the cost of constant 0, here used with a comparison operation eliminated after MCOML in the former assembly sequence, is lower (as per `vax_rtx_costs') in the VAX ISA than the cost of constant -1, used with CMPL in the latter sequence. Not only constant 0 is an implied operand with some machine instructions saving cycles and space otherwise used for an explicitly encoded operand, but if used with a comparison operation it can usually be eliminated, so it should be preferred over all other constants. With the example you gave with the PR I can see progression with f3, f4, f7, f8, regression with f1, f2, and no change in operation cost with f5, f6. Shouldn't a transformation like this respect target-specific expression costs somehow then? Depending on the individual case one form or the other might be cheaper, and somehow we assume here both are equivalent in terms of performance and/or code size (as applicable for the optimisation mode chosen). Maciej
Re: [PATCH] Correct -fdump-go-spec's handling of incomplete types
On Thu, Dec 17, 2020 at 10:22 AM Nikhil Benesch wrote: > > On 12/17/20 1:05 PM, Ian Lance Taylor wrote: > > Thanks for looking into this. I've gotten confused now, though. > > Which patch fixes the problem on Solaris? Thanks. > > > > Ian > > The patch I submitted upstream is all that is needed to fix the > compilation failures on Solaris: > > https://go-review.googlesource.com/c/gofrontend/+/278672 > > That is the same patch as "solaris-godump.patch" in this thread. > > That patch does not fix the fishiness around dummy types that you > pointed out earlier. That is what the other patch was trying (but > failing) to address. But we don't need to fix that fishiness to fix > Solaris. The code in mk[r]sysinfo.sh that is getting confused by the > u?pad128_t dummy type turns out to be unnecessary. So the patch I sent > upstream just removes that code. Thanks. Committed to mainline. Ian
Re: [PATCH] match.pd: Add ~(X - Y) -> ~X + Y simplification [PR96685]
On Mon, Dec 21, 2020 at 02:50:24AM +, Maciej W. Rozycki wrote: > On Sat, 12 Dec 2020, Jakub Jelinek via Gcc-patches wrote: > > > This patch adds the ~(X - Y) -> ~X + Y simplification requested > > in the PR (plus also ~(X + C) -> ~X + (-C) for constants C that can > > be safely negated. > > This regresses VAX code produced by the cmpelim-eq-notsi.c test case (and > its similar counterparts) with the `vax-netbsdelf' target. The point of the match.pd changes is to canonicalize GIMPLE on some form when there are several from GIMPLE POV equivalent or better forms of writing the same thing. The advantage of having one canonical way is that ICF, SCCVN etc. optimizations can then understand the different forms are equivalent. If another form is then better for a particular machine, it should be done either during expansion (being able to produce both RTLs and computing their costs), or during combine with either combine splitters or define_insn_and_split in the backend, or, if it can't be done in RTL, during the isel pass. Jakub
