I noticed that __builtin_assoc_barrier makes a differnce for FMAs formation
but it was not documented. This adds that documentation even with a small
example.
Build the HTML documents to make sure everything looks correct.
gcc/ChangeLog:
PR middle-end/115023
* doc/extend.texi (_
When writing the CodeView type definition for a struct, translate
DW_TAG_inheritance DIEs into LF_BCLASS records, to record which other
structs this one inherits from.
gcc/
* dwarf2codeview.cc (enum cv_leaf_type): Add LF_BCLASS.
(struct codeview_subtype): Add lf_bclass to union.
Follow MSVC in having a special type value, T_HRESULT, for (signed)
longs that have been typedef'd with the name "HRESULT". This is so that
the debugger can display user-friendly constant names when debugging COM
code.
gcc/
* dwarf2codeview.cc (get_type_num_typedef): New function.
Translate DW_TAG_ptr_to_member_type DIEs into special extended
LF_POINTER CodeView types.
gcc/
* dwarf2codeview.cc (struct codeview_custom_type): Add new fields to
lf_pointer struct in union.
(write_lf_pointer): Write containing_class and ptr_to_mem_type if
applicab
Test case pr117245.c is using trampolines, so mark it as such.
With this patch the test is now properly marked as unsupported
for pru-unknown-elf. Test still passes for x86_64-pc-linux-gnu.
Pushed to trunk as obvious.
gcc/testsuite/ChangeLog:
* gcc.dg/pr117245.c: Require effective targ
On Tue, Oct 22, 2024 at 2:34 PM Haochen Jiang wrote:
>
> From: "Hu, Lin1"
>
> gcc/ChangeLog:
>
> * common/config/i386/cpuinfo.h (get_available_features): Detect movrs.
> * common/config/i386/i386-common.cc
> (OPTION_MASK_ISA2_MOVRS_SET): New.
> (OPTION_MASK_ISA2_MO
Test case pr116887.c is passing -fpic, so mark it as such.
With this patch the test is now properly marked as unsupported
for pru-unknown-elf. Test still passes for x86_64-pc-linux-gnu.
Pushed to trunk as obvious.
gcc/testsuite/ChangeLog:
* gcc.dg/pr116887.c: Require effective target f
On Sat, Nov 02, 2024 at 02:47:55PM -0400, Andrew MacLeod wrote:
> Fresh extract and bootstrap, and I dont see those spurious failures I
> mnetioned. And everything is good with the port of these patches too on
> GCC14. so whenever seems reasonable.
Let's commit the backport now then.
Ja
Fresh extract and bootstrap, and I dont see those spurious failures I
mnetioned. And everything is good with the port of these patches too on
GCC14. so whenever seems reasonable.
Andrew
On 11/1/24 17:10, Andrew MacLeod wrote:
It eas an existing branch. I will get a scrtach branch and do
ev
On 11/2/24 8:44 AM, Thomas Koenig wrote:
Ping **(5./7.) ?
MASKR and MASKL are obvious candidates for unsigned, too; in the
previous version of the doc patch, I had promised that these would
take unsigned arguments in the future. What I had in mind was
they could take an unsigned argument and re
On Sat, 2024-11-02 at 16:38 +0800, Xi Ruoyao wrote:
> The patch at [1] changes these builtins to accept unsigned vector
> arguments to be consistent with other LSX/LASX bitwise builtins, but
> there's some concern about the backward compatibility. Make use of the
> existing built-in function overl
Ping **(5./7.) ?
MASKR and MASKL are obvious candidates for unsigned, too; in the
previous version of the doc patch, I had promised that these would
take unsigned arguments in the future. What I had in mind was
they could take an unsigned argument and return an unsigned result.
Thinking about t
For `(a | b) == 0`, we can "assert" on the true edge that
both `a == 0` and `b == 0` but nothing on the false edge.
For `(a | b) != 0`, we can "assert" on the false edge that
both `a == 0` and `b == 0` but nothing on the true edge.
This adds that predicate and allows us to optimize f0, f1,
and f2 i
Sometimes we get back a full ssa name when looking up the comparison of the
GIMPLE_COND
rather than a predicate. We then want to lookup the `val != 0` for the
predicate.
Note this might happen with other boolean assignments and COND_EXPR but I am
not sure
if it is as important; I have not found
After the last patch, we also want to record `(A CMP B) != 0`
as `(A CMP B)` and `(A CMP B) == 0` as `(A CMP B)` with the
true/false edges swapped.
This is enough to fix the original issue in `gcc.dg/tree-ssa/pr111456-1.c`
and make sure we don't regress it when enhancing ifcombine.
This adds that
To make it easier to add more predicates in some cases,
factor out the code. Plus it makes the code slightly more
readable since it is not indented as much.
Bootstrapped and tested on x86_64.
gcc/ChangeLog:
* tree-ssa-sccvn.cc (insert_predicates_for_cond): New function,
factored out fro
On Sat, 2 Nov 2024 at 07:11, Jakub Jelinek wrote:
>
> Hi!
>
> These overloads incorrectly cast the result of the float __builtin_*
> to _Float or __gnu_cxx::__bfloat16_t. For std::ilogb that changes
> behavior for the INT_MAX return because that isn't representable in
> either of the floating poi
On Sat, 2 Nov 2024, Sam James wrote:
> Some references to feed discussion which I had in my logs from
> discussing it with someone the other week after that interaction we
> had with alanc:
> * https://github.com/llvm/llvm-project/issues/28790 (not very
> insightful, other than to show it has a
From: Robin Dapp
This patch adds an else operand to vectorized masked load calls.
The current implementation adds else-value arguments to the respective
target-querying functions that is used to supply the vectorizer with the
proper else value.
We query the target for its supported else operand
From: Robin Dapp
This patch adds a zero else operand to masked loads, in particular the
masked gather load builtins that are used for gather vectorization.
gcc/ChangeLog:
* config/i386/i386-expand.cc (ix86_expand_special_args_builtin):
Add else-operand handling.
(ix86_ex
From: Robin Dapp
This adds zero else operands to masked loads and their intrinsics.
I needed to adjust more than initially thought because we rely on
combine for several instructions and a change in a "base" pattern
needs to propagate to all those.
gcc/ChangeLog:
* config/aarch64/aarch6
From: Robin Dapp
This patch adds else operands to masked loads. Currently the default
else operand predicate just accepts "undefined" (i.e. SCRATCH) values.
PR middle-end/115336
PR middle-end/116059
gcc/ChangeLog:
* config/riscv/autovec.md: Add else operand.
*
From: Robin Dapp
When predicating a load we implicitly assume that the else value is
zero. This matters in case the loaded value is padded (like e.g.
a Bool) and we must ensure that the padding bytes are zero on targets
that don't implicitly zero inactive elements.
A former version of this patc
From: Robin Dapp
This patch adds an undefined else operand to the masked loads.
gcc/ChangeLog:
* config/gcn/predicates.md (maskload_else_operand): New
predicate.
* config/gcn/gcn-valu.md: Use new predicate.
---
gcc/config/gcn/gcn-valu.md | 14 +-
gcc/confi
From: Robin Dapp
This patch amends the documentation for masked loads (maskload,
vec_mask_load_lanes, and mask_gather_load as well as their len
counterparts) with an else operand.
gcc/ChangeLog:
* doc/md.texi: Document masked load else operand.
---
gcc/doc/md.texi | 63
From: Robin Dapp
This patch adds else-operand handling to the internal functions.
gcc/ChangeLog:
* internal-fn.cc (add_mask_and_len_args): Rename...
(add_mask_else_and_len_args): ...to this and add else handling.
(expand_partial_load_optab_fn): Use adjusted function.
From: Robin Dapp
No change apart from rebasing against the latest strided load addition
so the CI can pick it up.
Bootstrapped and regtested on x86, aarch64, and power10.
Regtested on rv64gcv.
Robin Dapp (8):
docs: Document maskload else operand and behavior.
ifn: Add else-operand handling.
> >
> > Please reconsider? Why to we need to match LLVM here?<
>
> Matching llvm is not really the goal. I implemented it since it is
> useful optimization for code that builds small objects on heap and
> compiler can optimize away their use. This is relatively common for
> code with higer abstra
Here's the adjusted patch that I'll push once #6/7 and #8/7++ are also
approved. Retested along with them.
Prepare to handle noncontiguous ifcombine, introducing logic to modify
the outer condition when needed. There are two cases worth
mentioning:
- when blocks are noncontiguous, we have to
On Nov 2, 2024, Sam James wrote:
> BTW, forgive the possibly silly question, but I don't see any testcases
> for the series.
They're in the original patch, upthread, that this series was split out
of at reviewers' request. The tests all depended on a feature that was
not to be included in this
It became apparent that conditions could be combined that had deep SSA
dependency trees, that might thus require moving lots of statements.
Set a hard upper bound for now, hopefully to be replaced by a
dynamically computed bound, based on probabilities and costs.
Also reset flow sensitive info a
On Fri, Nov 01, 2024 at 12:08:45PM -0400, Jason Merrill wrote:
> On 10/31/24 6:09 AM, Nathaniel Shead wrote:
> > Bootstrapped and regtested on x86_64-pc-linux-gnu, OK for trunk?
> >
> > -- >8 --
> >
> > In cases like the linked PR we sometimes get mutually recursive
> > dependencies that both rel
On Fri, Nov 01, 2024 at 11:23:41AM -0400, Jason Merrill wrote:
> On 8/22/24 8:02 PM, Nathaniel Shead wrote:
> > On Thu, Aug 22, 2024 at 02:20:14PM -0400, Patrick Palka wrote:
> > > On Mon, 12 Aug 2024, Nathaniel Shead wrote:
> > >
> > > > Bootstrapped and regtested on x86_64-pc-linux-gnu, OK for t
The recent patch to improve detection of variably modified types
also fixed some issues related to unspecified sizes.
https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;h=9eae9268e41463927c9383004e58708048ec379f
There are some loose ends / possible cleanups which I missed and
which are addressed by
On Sat, 2024-11-02 at 09:52 +0800, Lulu Cheng wrote:
>
> 在 2024/11/2 上午1:36, Xi Ruoyao 写道:
> > Without optimization, GCC does not emit a jump table for the test case.
> >
> > I'm not sure if the test case has been wrong in the first place or
> > something has changed in these months...
>
> It wa
This patch adds optimization of the following patterns:
(zero_extend:M (subreg:N (not:O==M (X:Q==M ->
(xor:M (zero_extend:M (subreg:N (X:M)), 0x))
... where mask takes 0x bits of N mode bitsize
For the cases when X:M doesn't have any non-zero bits outside of mode N,
(zero_e
> On 31 Oct 2024, at 7:29 pm, Jakub Jelinek wrote:
>
> External email: Use caution opening links or attachments
>
>
> On Thu, Oct 31, 2024 at 08:21:09AM +, Kugan Vivekanandarajah wrote:
>>
>>
>>> On 31 Oct 2024, at 6:18 pm, Jakub Jelinek wrote:
>>>
>>> External email: Use caution opening li
On Sat, 2024-11-02 at 14:26 +0800, Xi Ruoyao wrote:
> On Sat, 2024-11-02 at 14:13 +0800, Lulu Cheng wrote:
> >
> > 在 2024/11/2 上午1:10, Xi Ruoyao 写道:
> > > On Thu, 2024-10-31 at 23:58 +0800, Xi Ruoyao wrote:
> > >
> > > /* snip */
> > >
> > > > ---
> > > >
> > > > Now running bootstrap & regtest
The patch at [1] changes these builtins to accept unsigned vector
arguments to be consistent with other LSX/LASX bitwise builtins, but
there's some concern about the backward compatibility. Make use of the
existing built-in function overloading mechanism to provide the backward
compatibility.
Dep
On Oct 30, 2024, Richard Biener wrote:
> I think since you make the outer condition the short-circuiting one what's in
> the inner block isn't executed when it wasn't before the transform? So in
> fact you shouldn't need to process stmts in BB even? Only when the
> outer condition is now uncond
[x]vldi.{b/h/w/d} is not implemented in LoongArch.
Use the macro [x]vrepli.{b/h/w/d} to replace.
gcc/ChangeLog:
* config/loongarch/lasx.md: Fixed.
* config/loongarch/lsx.md: Fixed.
---
gcc/config/loongarch/lasx.md | 2 +-
gcc/config/loongarch/lsx.md | 2 +-
2 files changed, 2 in
> Am 01.11.2024 um 11:13 schrieb Jakub Jelinek :
>
> Hi!
>
> Apparently tree_output_constant_def doesn't strictly guarantee that the
> returned VAR_DECL will have the same or uselessly convertible type as
> the type of the constant passed to it, compare_constants says:
> /* For arrays, check
Hi!
Here is a new version of the toplevel asm PR41045 patch, not yet
with the new constraints for symbol definitions (I intend to
deal with that in a follow-up).
Compared to the last patch, this one has some testcases added
and some ICEs related to "m"/"=m" fixed; for LTO streaming I've
added a so
Hi!
These overloads incorrectly cast the result of the float __builtin_*
to _Float or __gnu_cxx::__bfloat16_t. For std::ilogb that changes
behavior for the INT_MAX return because that isn't representable in
either of the floating point formats, for the others it is I think
just a very inefficient
44 matches
Mail list logo
