On Mon, Dec 17 2018 20:15:02 EET Bernd Edlinger wrote:
> out of curiosity I looked at the clobber statement in
> gdb/nat/linux-ptrace.c:
>
>asm volatile ("pushq %0;"
> ".globl linux_ptrace_test_ret_to_nx_instr;"
> "linux_ptrace_test_ret
The attach patch shuffles lines around to eliminate 80
lines of #ifdef...#endif; thereby making the file more
readable. Tested on i586-*-freebsd and x86_64-*-freebsd.
Patch is a pre-requisite to fixing issues and adding
missing functionality. OK to commit?
2018-12-18 Steven G. Kargl
This patch is the analog to r266868-r266870, but for SSSE3.
The SSSE3 tests had been inadvertently made to PASS without actually running
the test code. Actually running the code turned up some previously undetected
issues.
This patch fixes some issues in the implementations, fixes up the tests
to
On Wed, Dec 19, 2018 at 12:00 PM Andi Kleen wrote:
>
> On Wed, Dec 19, 2018 at 10:01:15AM +0800, Bin.Cheng wrote:
> > On Tue, Dec 18, 2018 at 7:15 PM Bin.Cheng wrote:
> > >
> > > On Sun, Dec 16, 2018 at 9:11 AM Andi Kleen wrote:
> > > >
> > > > "bin.cheng" writes:
> > > >
> > > > > Hi,
> > > >
On Wed, Dec 19, 2018 at 10:01:15AM +0800, Bin.Cheng wrote:
> On Tue, Dec 18, 2018 at 7:15 PM Bin.Cheng wrote:
> >
> > On Sun, Dec 16, 2018 at 9:11 AM Andi Kleen wrote:
> > >
> > > "bin.cheng" writes:
> > >
> > > > Hi,
> > > >
> > > > Due to ICE and mal-functional bugs, indirect call value profil
On Wed, Dec 19, 2018 at 09:26:51AM +0800, Bin.Cheng wrote:
> On Wed, Dec 19, 2018 at 5:27 AM Andi Kleen wrote:
> >
> > > Yes, take g++.dg/tree-prof/morefunc.C as an example:
> > > - int i;
> > > - for (i = 0; i < 1000; i++)
> > > + int i, j;
> > > + for (i = 0; i < 100; i++)
> > > +for
On 12/15/18 3:36 AM, Bernd Edlinger wrote:
this patch implements an error message, for non-static initialization of a
flexible array member.
This duplicates the existing error message from the C-FE, to avoid ICE and
wrong code generation
issues, as pointed out in the PR.
It is a bit funny that
On 12/18/18 6:19 PM, Jakub Jelinek wrote:
On Tue, Dec 18, 2018 at 05:40:03PM -0500, Jason Merrill wrote:
On 12/18/18 3:45 PM, Jakub Jelinek wrote:
The following testcase FAILs, because parsing creates a TREE_CONSTANT
CONSTRUCTOR that contains CONST_DECL elts. cp_fold_r can handle that,
but con
Cherry, see
https://gcc.gnu.org/ml/gcc-testresults/2018-12/msg02241.html
https://gcc.gnu.org/ml/gcc-testresults/2018-12/msg02240.html
still shows ~180 test failures on ARM32. Sorry, no access to an ARM32 box until
next year.
Matthias
On 13.12.18 00:27, Ian Lance Taylor wrote:
> On Wed, Dec 12,
On 12/18/18 2:12 PM, H.J. Lu wrote:
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index ac2ee59d92c..47f2fc3f518 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -358,6 +358,7 @@ Objective-C and Objective-C++ Dialects}.
-Wuseless-cast -Wvariadic-macros -Wvector-operation
For HiSilicon's tsv110 cpu core, it supports some v8_4A features, but
some mandatory features are not implemented. Revert to ARMv8.2 that
all mandatory features are supported.
---
gcc/ChangeLog| 5 +
gcc/config/aarch64/aarch64-cores.def | 6 +++---
2 files changed, 8 i
On Tue, Dec 18, 2018 at 7:15 PM Bin.Cheng wrote:
>
> On Sun, Dec 16, 2018 at 9:11 AM Andi Kleen wrote:
> >
> > "bin.cheng" writes:
> >
> > > Hi,
> > >
> > > Due to ICE and mal-functional bugs, indirect call value profile
> > > transformation
> > > is disabled on GCC-7/8/trunk. This patch resto
On Wed, Dec 19, 2018 at 5:27 AM Andi Kleen wrote:
>
> > Yes, take g++.dg/tree-prof/morefunc.C as an example:
> > - int i;
> > - for (i = 0; i < 1000; i++)
> > + int i, j;
> > + for (i = 0; i < 100; i++)
> > +for (j = 0; j < 50; j++)
> > g += tc->foo();
> > if (g<100) g++;
> >
On 12/18/18 3:58 AM, Alexandre Oliva wrote:
> Jeff, you mentioned you had changes to the VRP overflow test that would
> fix this, but I couldn't figure out whether or not you ever put them in
> and it regressed again later, or what. Anyway, here's my take on it.
No, they're not on the trunk yet.
On Tue, Dec 18, 2018 at 05:40:03PM -0500, Jason Merrill wrote:
> On 12/18/18 3:45 PM, Jakub Jelinek wrote:
> > The following testcase FAILs, because parsing creates a TREE_CONSTANT
> > CONSTRUCTOR that contains CONST_DECL elts. cp_fold_r can handle that,
> > but constexpr evaluation doesn't touch
On 12/18/18 3:45 PM, Jakub Jelinek wrote:
The following testcase FAILs, because parsing creates a TREE_CONSTANT
CONSTRUCTOR that contains CONST_DECL elts. cp_fold_r can handle that,
but constexpr evaluation doesn't touch those CONSTRUCTORs.
Fixed thusly, bootstrapped/regtested on x86_64-linux a
On 12/18/18 3:49 PM, Jakub Jelinek wrote:
cp_parser_parenthesized_expression_list creates expression_list in GC
memory; if it is called when current_function_decl is NULL, there might be
ggc_collect in the middle of the parsing and collect that vector.
Fixed by temporarily bumping function_depth
On 12/18/18 4:12 PM, H.J. Lu wrote:
On Tue, Dec 18, 2018 at 12:36 PM Jason Merrill wrote:
On 12/18/18 9:10 AM, H.J. Lu wrote:
+ switch (TREE_CODE (rhs))
+{
+case ADDR_EXPR:
+ base = TREE_OPERAND (rhs, 0);
+ while (handled_component_p (base))
+ {
+ if (TREE_CODE (b
On 12/18/18 4:50 PM, Jakub Jelinek wrote:
On Tue, Dec 18, 2018 at 04:23:12PM -0500, Vladimir Makarov wrote:
The following patch fixes
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87759
The patch was bootstrapped and tested on x86-64.
Committed as rev. 267244.
The test FAILs on i686-lin
On Tue, Dec 18, 2018 at 04:23:12PM -0500, Vladimir Makarov wrote:
> The following patch fixes
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87759
>
> The patch was bootstrapped and tested on x86-64.
>
> Committed as rev. 267244.
The test FAILs on i686-linux, fixed thusly, committed as ob
On 12/11/18 4:19 PM, Jason Merrill wrote:
On 12/11/18 6:08 PM, Martin Sebor wrote:
On 12/11/18 3:52 PM, Marek Polacek wrote:
On Tue, Dec 11, 2018 at 03:46:37PM -0700, Martin Sebor wrote:
On 12/11/18 1:47 PM, Jakub Jelinek wrote:
On Tue, Dec 11, 2018 at 01:36:58PM -0700, Martin Sebor wrote:
A
> Yes, take g++.dg/tree-prof/morefunc.C as an example:
> - int i;
> - for (i = 0; i < 1000; i++)
> + int i, j;
> + for (i = 0; i < 100; i++)
> +for (j = 0; j < 50; j++)
> g += tc->foo();
> if (g<100) g++;
> }
> @@ -27,8 +28,9 @@ void test1 (A *tc)
> static __attribute__((alw
The following patch fixes
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87759
The patch was bootstrapped and tested on x86-64.
Committed as rev. 267244.
Index: ChangeLog
===
--- ChangeLog (revision 267243)
+++ ChangeLog (workin
On Tue, Dec 18, 2018 at 12:36 PM Jason Merrill wrote:
>
> On 12/18/18 9:10 AM, H.J. Lu wrote:
> > + switch (TREE_CODE (rhs))
> > +{
> > +case ADDR_EXPR:
> > + base = TREE_OPERAND (rhs, 0);
> > + while (handled_component_p (base))
> > + {
> > + if (TREE_CODE (base) == C
Hi Chung-Lin!
On Tue, 18 Dec 2018 23:06:38 +0800, Chung-Lin Tang
wrote:
> this part includes some of the lookup_goacc_asyncqueue fixes we talked about.
> I am still thinking about how the queue lock problem should really be solved,
> so regard
> this patch as just fixing some of the problems.
Hi!
As mentioned in the PR, there is a VEX encoded vpclmulqdq instruction
with ymm arguments that needs VPCLMULQDQ ISA, and then EVEX encoded
vpclmulqdq with zmm arguments that needs VPCLMULQDQ + AVX512F ISAs and
vpclmulqdq with xmm or ymm arguments that needs VPCLMULQDQ + AVX512VL ISAs.
So, _mm2
Hi!
cp_parser_parenthesized_expression_list creates expression_list in GC
memory; if it is called when current_function_decl is NULL, there might be
ggc_collect in the middle of the parsing and collect that vector.
Fixed by temporarily bumping function_depth. Or should that be done in some
other
Hi!
The following testcase FAILs, because parsing creates a TREE_CONSTANT
CONSTRUCTOR that contains CONST_DECL elts. cp_fold_r can handle that,
but constexpr evaluation doesn't touch those CONSTRUCTORs.
Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
trunk?
2018-12-1
On 12/10/18 5:47 PM, Segher Boessenkool wrote:
Not all qualifiers are asm qualifiers. We can talk about that in a
nicer way than just giving a generic parser error.
This also adds two testcases for C++, that previously were for C only.
2018-12-10 Segher Boessenkool
c/
* c-parser.c
On 12/10/18 5:47 PM, Segher Boessenkool wrote:
Previously, "volatile" was allowed. Changing this simplifies the code,
makes things more regular, and makes the C and C++ frontends handle
this the same way.
2018-12-10 Segher Boessenkool
cp/
* parser.c (cp_parser_asm_definition): Do n
On 12/10/18 5:47 PM, Segher Boessenkool wrote:
As suggested by Jason.
Segher
2018-12-10 Segher Boessenkool
c/
* c-parser.c (c_parser_asm_statement): Rewrite the loop to work without
"done" boolean variable.
cp/
* parser.c (cp_parser_asm_definition): Rewrite the lo
On 12/18/18 4:22 PM, David Malcolm wrote:
On Mon, 2018-12-17 at 18:30 -0500, David Malcolm wrote:
On Mon, 2018-12-17 at 14:33 -0500, Jason Merrill wrote:
On 12/14/18 7:17 PM, David Malcolm wrote:
+ /* Since default args are effectively part of the function
type,
+strip location wr
On 12/18/18 9:10 AM, H.J. Lu wrote:
+ switch (TREE_CODE (rhs))
+{
+case ADDR_EXPR:
+ base = TREE_OPERAND (rhs, 0);
+ while (handled_component_p (base))
+ {
+ if (TREE_CODE (base) == COMPONENT_REF)
+ break;
+ base = TREE_OPERAND (base, 0);
+
On Mon, 2018-12-17 at 18:30 -0500, David Malcolm wrote:
> On Mon, 2018-12-17 at 14:33 -0500, Jason Merrill wrote:
> > On 12/14/18 7:17 PM, David Malcolm wrote:
> > > + /* Since default args are effectively part of the function
> > > type,
> > > + strip location wrappers here, since otherwise
On 12/17/18 6:30 PM, David Malcolm wrote:
On Mon, 2018-12-17 at 14:33 -0500, Jason Merrill wrote:
On 12/14/18 7:17 PM, David Malcolm wrote:
+ /* Since default args are effectively part of the function
type,
+strip location wrappers here, since otherwise the
location of
+one
On Tue, Dec 18, 2018 at 3:57 PM Jakub Jelinek wrote:
>
> Hi!
>
> We weren't vectorizing with unconditional or masked scatters when
> -mprefered-vector-width={128,256}. While for DI index and DF/DI
> stores or SI index and SF/SI stores we even have the builtins,
> for the remaining combinations I
* include/bits/fs_dir.h (operator<<): Overload for directory_entry,
as per LWG 3171.
* testsuite/27_io/filesystem/directory_entry/lwg3171.cc: New test.
Tested x86_64-linux, committed to trunk.
commit 0d24038c0b565dbcd5e7729423398da281245c41
Author: Jonathan Wakely
Date:
On Tue, Dec 18, 2018 at 04:33:48PM +, Uecker, Martin wrote:
> > Yes, something like this. If the trampolines are pre-allocated, this could
> > even avoid the need to clear the cache on archs where this is needed.
>
> And if we can make the trampolines be all the same (and it somehow derived
>
On 18/12/18 15:52 +, Jonathan Wakely wrote:
Now that path::operator/=(basic_string_view) works directly
from the string argument, instead of constructing a temporary path from
the string, it's potentially more efficient to do 'path(x) /= s' instead
of 'x / s'. This changes the only relevant p
Am Dienstag, den 18.12.2018, 17:29 +0100 schrieb Martin Uecker:
> Am Dienstag, den 18.12.2018, 17:24 +0100 schrieb Jakub Jelinek:
> > On Tue, Dec 18, 2018 at 09:03:41AM -0700, Jeff Law wrote:
> > > Right. This is the classic example and highlights the ABI concerns. If
> > > we use the low bit to
Am Dienstag, den 18.12.2018, 17:24 +0100 schrieb Jakub Jelinek:
> On Tue, Dec 18, 2018 at 09:03:41AM -0700, Jeff Law wrote:
> > Right. This is the classic example and highlights the ABI concerns. If
> > we use the low bit to distinguish between a normal function pointer and
> > a pointer to a des
Am Dienstag, den 18.12.2018, 09:03 -0700 schrieb Jeff Law:
> On 12/18/18 8:32 AM, Jakub Jelinek wrote:
> > On Tue, Dec 18, 2018 at 10:23:46AM -0500, Paul Koning wrote:
> > >
> > >
> > > > On Dec 17, 2018, at 2:23 PM, Szabolcs Nagy
> > > > wrote:
> > > >
> > > > On 17/12/2018 18:22, Uecker, Mar
On Tue, Dec 18, 2018 at 09:03:41AM -0700, Jeff Law wrote:
> Right. This is the classic example and highlights the ABI concerns. If
> we use the low bit to distinguish between a normal function pointer and
> a pointer to a descriptor and qsort doesn't know about it, then we lose.
>
> One way arou
On Tue, Dec 18, 2018 at 11:18:03PM +1030, Alan Modra wrote:
> On Tue, Dec 18, 2018 at 03:20:02AM -0600, Segher Boessenkool wrote:
> > Hi Alan,
> >
> > On Tue, Dec 18, 2018 at 10:39:27AM +1030, Alan Modra wrote:
> > > On Mon, Dec 17, 2018 at 11:05:57AM -0600, Segher Boessenkool wrote:
> > > > On Mo
On 12/18/18 8:32 AM, Jakub Jelinek wrote:
> On Tue, Dec 18, 2018 at 10:23:46AM -0500, Paul Koning wrote:
>>
>>
>>> On Dec 17, 2018, at 2:23 PM, Szabolcs Nagy wrote:
>>>
>>> On 17/12/2018 18:22, Uecker, Martin wrote:
>
> ...
So a thread_local static variable for storing the static
Now that path::operator/=(basic_string_view) works directly
from the string argument, instead of constructing a temporary path from
the string, it's potentially more efficient to do 'path(x) /= s' instead
of 'x / s'. This changes the only relevant place in the library.
* src/filesystem/st
The resolution for LWG 2936 defines the comparison more precisely, which
this patch implements. The patch also defines comparisons with strings
to work without constructing a temporary path object (so avoids any
memory allocations).
* include/bits/fs_path.h (path::compare(const string_typ
* include/std/string_view [__cplusplus > 201703L]
(basic_string_view::starts_with(basic_string_view)): Implement
proposed resolution of LWG 3040 to avoid redundant length check.
(basic_string_view::starts_with(_CharT)): Implement proposed
resolution of LWG 3
On Tue, Dec 18, 2018 at 10:23:46AM -0500, Paul Koning wrote:
>
>
> > On Dec 17, 2018, at 2:23 PM, Szabolcs Nagy wrote:
> >
> > On 17/12/2018 18:22, Uecker, Martin wrote:
> >>>
> >>> ...
> >>
> >> So a thread_local static variable for storing the static
> >> chain?
> >
> > something like that
> On Dec 17, 2018, at 2:23 PM, Szabolcs Nagy wrote:
>
> On 17/12/2018 18:22, Uecker, Martin wrote:
>>>
>>> ...
>>
>> So a thread_local static variable for storing the static
>> chain?
>
> something like that, but the more i think about it the
> harder it seems: the call site of the nested f
On 12/18/18 3:16 PM, Bernd Edlinger wrote:
> Hi,
>
> while I looked closely at the asm statement in the gdb,
> I realized that the SP clobber forces the function to use
> the frame pointer, and prevents the red zone. That
> makes the push / pop sequence in the asm statement safe
> to use, as long
On 2018/12/11 9:50 PM, Chung-Lin Tang wrote:
On 2018/12/10 6:02 PM, Chung-Lin Tang wrote:
On 2018/12/7 04:57 AM, Thomas Schwinge wrote>> ---
a/libgomp/plugin/plugin-nvptx.c
+++ b/libgomp/plugin/plugin-nvptx.c
+struct goacc_asyncqueue *
+GOMP_OFFLOAD_openacc_async_construct (void)
+{
+ stru
On 2018/9/25 9:10 PM, Chung-Lin Tang wrote:
Hi Thomas,
These are the OpenACC specific changes, mostly the re-implementation of
async-related acc_* runtime
library API functions to use the new backend plugin interfaces, in a non-target
specific way.
Hi Thomas,
this part includes some of the l
On 2018/12/15 1:52 AM, Thomas Schwinge wrote:
As for the following changes, will you please make sure that there is one
common order for these, used in "libgomp/libgomp-plugin.h" function
prototypes, "libgomp/libgomp.h:acc_dispatch_t",
"libgomp/target.c:gomp_load_plugin_for_device", "libgomp/oacc
Hi!
We weren't vectorizing with unconditional or masked scatters when
-mprefered-vector-width={128,256}. While for DI index and DF/DI
stores or SI index and SF/SI stores we even have the builtins,
for the remaining combinations I had to add a few alt builtins (with spaces
in names as in other cas
On Sat, Nov 10, 2018 at 09:11:19AM -0800, Julian Brown wrote:
> This patch, created while trying to figure out the open-coded linked-list
> handling in gimplify_scan_omp_clauses, factors out four somewhat
> repetitive portions of that function into two new outlined functions.
> This was done largel
Hi,
while I looked closely at the asm statement in the gdb,
I realized that the SP clobber forces the function to use
the frame pointer, and prevents the red zone. That
makes the push / pop sequence in the asm statement safe
to use, as long as the stack is restored to the original
value. That ca
On Sat, 10 Nov 2018 09:11:19 -0800
Julian Brown wrote:
> This patch, created while trying to figure out the open-coded
> linked-list handling in gimplify_scan_omp_clauses, factors out four
> somewhat repetitive portions of that function into two new outlined
> functions. This was done largely mec
On Mon, Dec 17, 2018 at 08:53:32AM -0500, Jason Merrill wrote:
> On 12/17/18 7:42 AM, H.J. Lu wrote:
> > On Mon, Dec 17, 2018 at 1:39 AM Richard Biener
> > wrote:
> > >
> > > On Fri, Dec 14, 2018 at 11:48 PM H.J. Lu wrote:
> > > >
> > > > On Fri, Dec 14, 2018 at 2:10 PM Jason Merrill wrote:
>
On 2018/12/18 8:36 PM, Jakub Jelinek wrote:
On Fri, Dec 14, 2018 at 06:52:20PM +0100, Thomas Schwinge wrote:
--- a/include/gomp-constants.h
+++ b/include/gomp-constants.h
@@ -199,7 +200,7 @@ enum gomp_map_kind
/* Versions of libgomp and device-specific plugins. GOMP_VERSION
should be
On 12/18/18 7:43 AM, Ulrich Weigand wrote:
> Bill Schmidt wrote:
>
>> +@item
>> +When using @code{vector} in keyword-and-predefine mode; for example,
>> +
>> +@smallexample
>> +typedef signed short int16;
>> +vector int16 data;
>> +@end smallexample
>> +
>> +Note that keyword-and-predefine mode is
Bill Schmidt wrote:
> +@item
> +When using @code{vector} in keyword-and-predefine mode; for example,
> +
> +@smallexample
> +typedef signed short int16;
> +vector int16 data;
> +@end smallexample
> +
> +Note that keyword-and-predefine mode is enabled by disabling GNU
> +extensions (e.g., by using
Hi All,
This patch makes the feature detection code for AArch64 GCC not add features
automatically when the feature had no hwcaps string to match against.
This means that -mcpu=native no longer adds feature flags such as +profile.
The behavior wasn't noticed before because at the time +profile wa
TARGET_VTABLE_ENTRY_ALIGN defaults to POINTER_SIZE, which is 20 for
msp430-elf -mlarge.
g++.dg/torture/pr41257.C ICEs after the invalid alignment of 20 is set.
> during GIMPLE pass: slp
> gcc/testsuite/g++.dg/torture/pr41257.C: In function 'void bar()':
> gcc/testsuite/g++.dg/torture/pr41257.C:17
On 12/18/18 4:33 AM, Ulrich Weigand wrote:
> Bill Schmidt wrote:
>
>> +@item
>> +When using vector in keyword-and-predefine mode; for example,
>> +
>> +@smallexample
>> +/* With -maltivec only: */
> This is a bit confusing (at least to me). What does "with -maltivec only"
> mean here? Just addin
On Thu, Dec 13, 2018 at 02:11:31PM +, Julian Brown wrote:
> > Any reason for the above (ditto in C), rather than just adding
> > && ort != C_ORT_ACC to the while loop condition for CPP_OPEN_SQUARE?
> > (, . or * after id-expression is like any other unhandled
> > characters...
>
> I think the
On 12/18/2018 09:32 AM, Mihail Ionescu wrote:
Hi All,
In Thumb mode when the function prologue gets expanded, in case of a
multiple register push, additional mov instructions are generated to
save the high registers which result in lr getting overwritten before
it's value can be used to ret
On Thu, Dec 13, 2018 at 10:52:32PM +0800, Chung-Lin Tang wrote:
> --- gcc/omp-low.c (revision 267050)
> +++ gcc/omp-low.c (working copy)
> @@ -60,6 +60,7 @@ along with GCC; see the file COPYING3. If not see
> #include "hsa-common.h"
> #include "stringpool.h"
> #include "attribs.h"
> +#i
On Tue, Dec 18, 2018 at 03:20:02AM -0600, Segher Boessenkool wrote:
> Hi Alan,
>
> On Tue, Dec 18, 2018 at 10:39:27AM +1030, Alan Modra wrote:
> > On Mon, Dec 17, 2018 at 11:05:57AM -0600, Segher Boessenkool wrote:
> > > On Mon, Dec 17, 2018 at 10:40:01AM +1030, Alan Modra wrote:
> > > > Since I b
On Thu, Dec 13, 2018 at 03:44:25PM +, Julian Brown wrote:
> +static tree
> +convert_to_firstprivate_int (tree var, gimple_seq *gs)
> +{
> + tree type = TREE_TYPE (var), new_type = NULL_TREE;
> + tree tmp = NULL_TREE;
> +
> + if (omp_is_reference (var))
> +type = TREE_TYPE (type);
> +
> +
On Tue, Dec 18, 2018 at 12:40 PM Kyrill Tkachov
wrote:
>
> Hi all,
>
> In GCC 9 the aarch64 port learned how to do V16QImode SAD operations on
> signed and unsigned chars.
> But I had missed enabling the effective target for that.
> This patch enables that target for non-SVE aarch64.
> Two new te
On Fri, Dec 14, 2018 at 06:52:20PM +0100, Thomas Schwinge wrote:
> > --- a/include/gomp-constants.h
> > +++ b/include/gomp-constants.h
>
> > @@ -199,7 +200,7 @@ enum gomp_map_kind
> > /* Versions of libgomp and device-specific plugins. GOMP_VERSION
> > should be incremented whenever an ABI-i
On Fri, Dec 14, 2018 at 07:00:30PM +, Julian Brown wrote:
> OpenACC 2.6 manual deep copy support (attach/detach)
>
> gcc/c-family/
> * c-pragma.h (pragma_omp_clause): Add PRAGMA_OACC_CLAUSE_ATTACH,
> PRAGMA_OACC_CLAUSE_DETACH.
and instead of , ?
> --- a/gcc/c-family/
Hi Chung-Lin!
On Tue, 18 Dec 2018 18:02:54 +0800, Chung-Lin Tang
wrote:
> On 2018/12/17 10:32 PM, Thomas Schwinge wrote:
> >> The reason there are deadlocks from inside the plugin on
> >> GOMP_PLUGIN_fatal() is when we hold the
> >> struct gomp_device_descr's*device* lock, which is also acquir
Hi all,
In GCC 9 the aarch64 port learned how to do V16QImode SAD operations on signed
and unsigned chars.
But I had missed enabling the effective target for that.
This patch enables that target for non-SVE aarch64.
Two new tests now PASS on aarch64:
gcc.dg/vect/slp-reduc-sad.c
gcc.dg/vect/vect-
On Sun, Dec 16, 2018 at 9:11 AM Andi Kleen wrote:
>
> "bin.cheng" writes:
>
> > Hi,
> >
> > Due to ICE and mal-functional bugs, indirect call value profile
> > transformation
> > is disabled on GCC-7/8/trunk. This patch restores the transformation. The
> > main issue is AutoFDO should store cg
On Mon, 17 Dec 2018, Jakub Jelinek wrote:
> Hi!
>
> Some of the following testcases ICE, because I was assuming that
> VEC_UNPACK_{LO,HI}_EXPR and VEC_PACK_TRUNC_EXPR just work on the
> VECTOR_BOOLEAN_TYPE_P mask types that AVX512* has (with scalar modes),
> but they really only work if the wider
Jeff, you mentioned you had changes to the VRP overflow test that would
fix this, but I couldn't figure out whether or not you ever put them in
and it regressed again later, or what. Anyway, here's my take on it.
PR 86153 was originally filed when changes to the C++11's
implementation of vector
On Tue, 18 Dec 2018 03:08:51 -0600
Segher Boessenkool wrote:
> Hi!
>
> On Fri, Dec 14, 2018 at 03:22:13PM +, Jozef Lawrynowicz wrote:
> > 2018-12-14 Jozef Lawrynowicz
> >
> > gcc/ChangeLog:
> > * combine.c (update_rsp_from_reg_equal): Only look for the nonzero bits
> > of src
Bill Schmidt wrote:
> +@item
> +When using vector in keyword-and-predefine mode; for example,
> +
> +@smallexample
> +/* With -maltivec only: */
This is a bit confusing (at least to me). What does "with -maltivec only"
mean here? Just adding -maltivec will *not* switch to keyword-and-
predefine
On Tue, Dec 18, 2018 at 10:59:20AM +0100, Thomas Schwinge wrote:
> OK for trunk?
>
> commit 20d3cbd6e27b10ae1cd352cc177d7697a4a57db0
> Author: Thomas Schwinge
> Date: Mon Dec 17 18:26:29 2018 +0100
>
> Cleanup libgomp's coalesce chunk data structures
>
> libgomp/
>
On 2018/12/17 10:32 PM, Thomas Schwinge wrote:
The reason there are deadlocks from inside the plugin on GOMP_PLUGIN_fatal() is
when we hold the
struct gomp_device_descr's*device* lock, which is also acquired when we
execute atexit device shutdown handlers, hence the deadlock.
I don't think th
Hi Jakub!
Julian had a look at this, and now I too (and just fixed some
formatting):
On Wed, 2 May 2018 13:02:09 -0700, Cesar Philippidis
wrote:
> Libgomp's usage of struct gomp_coalesce_buf is a little confusing. The
> member chunks is an array where the even elements represent the starting
>
On 2018/12/17 9:52 PM, Thomas Schwinge wrote:
Hi Chung-Lin!
On Fri, 14 Dec 2018 22:52:44 +0800, Chung-Lin Tang
wrote:
On 2018/12/14 10:17 PM, Thomas Schwinge wrote:
On Tue, 25 Sep 2018 21:10:47 +0800, Chung-Lin Tang
wrote:
--- a/libgomp/oacc-async.c
+++ b/libgomp/oacc-async.c
+attribut
Hi All,
In Thumb mode when the function prologue gets expanded, in case of a
multiple register push, additional mov instructions are generated to
save the high registers which result in lr getting overwritten before
it's value can be used to retrieve the return address.
The fix consists of d
Hi Alan,
On Tue, Dec 18, 2018 at 10:39:27AM +1030, Alan Modra wrote:
> On Mon, Dec 17, 2018 at 11:05:57AM -0600, Segher Boessenkool wrote:
> > On Mon, Dec 17, 2018 at 10:40:01AM +1030, Alan Modra wrote:
> > > Since I broke powerpc*-freebsd and the other non-linux powerpc
> > > targets, I guess I o
Hi Bill,
On Mon, Dec 17, 2018 at 03:54:23PM -0600, Bill Schmidt wrote:
> We recently discovered some incorrect documentation about this topic and
> agreed it should be changed.
> This is my attempt to clarify it. Built and verified on
> powerpc64le-linux-gnu. Is this ok for trunk?
Okay for tr
Hi!
On Fri, Dec 14, 2018 at 03:22:13PM +, Jozef Lawrynowicz wrote:
> 2018-12-14 Jozef Lawrynowicz
>
> gcc/ChangeLog:
> * combine.c (update_rsp_from_reg_equal): Only look for the nonzero bits
> of src in nonzero_bits_mode if the mode of src is MODE_INT and
> HWI_COMP
On Tue, Dec 18, 2018 at 08:25:37AM +0100, Uros Bizjak wrote:
> > : Check for non-VECTOR_MODE_P
> > rather than VOIDmode.
>
> This entry doesn't match the change, you are checking for
> VECTOR_MODE_P. On a related note, should similar
Ok, I'll write: Check for VECTOR_MODE_P rather
89 matches
Mail list logo
