On Oct 27, 2024, Jeff Law wrote:
> On 10/25/24 5:52 AM, Alexandre Oliva wrote:
>> Refactor ifcombine_ifandif, moving the common code from the various
>> paths that apply the combined condition to a new function.
>>
>> for gcc/ChangeLog
>> * tree-ssa-ifc
n laxer.
I'm testing an incremental patch that adds simple bounds to the amount
of stmt moving, and flow info resetting and UB avoiding for moved stmts,
and arranges to reset flow info and avoid UB in inner blocks when the
outer cond is relaxed.
--
Alexandre Oliva, happy hackerht
unsigned arithmetic but cannot do so for signed division. */
|| ((ass = dyn_cast (stmt))
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for
On Oct 22, 2024, Richard Biener wrote:
> On Mon, Oct 21, 2024 at 4:30 AM Alexandre Oliva wrote:
>>
>> On Oct 10, 2024, Richard Biener wrote:
>>
>> > As you special-case hard register uses, do we want to ever make
>> > hard-register
>> > uses
,
...
*/
- return ifcombine_ifandif (inner_cond_bb, true, outer_cond_bb, false,
- true);
+ return ifcombine_ifandif (inner_cond_bb, true, outer_cond_bb, false);
}
return false;
--
Alexandre Oliva, happy hacker
useful.
Ok to install?
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
eric_expr (dump_file, t);
+ if (ts)
+ {
+ fprintf (dump_file, " and ");
+ print_generic_expr (dump_file, ts);
+ }
fprintf (dump_file, "\n");
}
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free S
/* Clear range info from all stmts in BB which is now executed
- conditional on a always true/false condition. */
+ conditional on a always true/false condition. ??? When we
+ combine noncontiguous blocks, do we need to adjust the
+ intervenin
er. */
+
+ inner_taken->probability *= profile_probability::even ();
+ inner_not_taken->probability = profile_probability::always ()
+ - inner_taken->probability;
+
+ prob *= inner_taken->probability;
+ outer2->probability += prob;
+ outer_to_inner->probability =
cond_set_condition_from_tree (inner_cond, cond2);
+ }
+ else
+ gimple_cond_set_condition_from_tree (inner_cond,
+inner_inv
+? boolean_false_node
+: boolean_true_node);
+
cond2 = TREE_OPERAND (cond, 1);
+ cond = TREE_OPERAND (cond, 0);
+}
+
bool outer_p = cond2 || (single_pred (gimple_bb (inner_cond))
!= gimple_bb (outer_cond));
bool result_inv = outer_p ? outer_inv : inner_inv;
--
Alexandre Oliva, happy hackerhttps
On Nov 10, 2024, Jakub Jelinek wrote:
> On Sun, Nov 10, 2024 at 01:30:06PM -0300, Alexandre Oliva wrote:
>> I'm surprised the commit-time checker didn't catch them.
> I'm surprised too, but don't want to try to push further broken commits just
> to double
very helpful to avoid typos in filenames.
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving &q
This patch introduces various improvements to the logic that merges
field compares, moving it into ifcombine.
Before the patch, we could merge:
(a.x1 EQNE b.x1) ANDOR (a.y1 EQNE b.y1)
into something like:
(((type *)&a)[Na] & MASK) EQNE (((type *)&b)[Nb] & MASK)
if both of A's fields li
at the best thing to do in case the note doesn't
carry location information, or the referenced block is missing or
however else it could be NULL. Staying in the previous scope is
somewhat sensible, but it amounts to silently dropping debug
information; a visible change of scope might be pr
B and the
@@ -1245,7 +1325,7 @@ tree_ssa_ifcombine_bb (basic_block inner_cond_bb)
of outer, we'd get ([1 ? 0 :] b ? 1 : (a | c) ? 0 : -1), which would
yield 1 rather than 0 when (a). */
if (!changed
- && !same_phi_args_p (outer_cond_bb, inner_
On Nov 22, 2024, Alexandre Oliva wrote:
> - Rework BIT_XOR handling to avoid having to match patterns again.
I goofed here, and only an -O3 profiling bootstrap of gcc-14 caught it.
Please consider this incremental patchlet as if included as part of the
previous one. Bootstrapped on top of
-1,5 +1,5 @@
/* { dg-do run } */
-/* { dg-options "-O" } */
+/* { dg-options "-O -fdump-tree-ifcombine-details" } */
/* Check that conversions and selections of similar bit ranges across different
types don't prevent combination. */
@@ -25,7 +25,7 @@ void f (void) {
This patch introduces various improvements to the logic that merges
field compares, while moving it into ifcombine.
Before the patch, we could merge:
(a.x1 EQNE b.x1) ANDOR (a.y1 EQNE b.y1)
into something like:
(((type *)&a)[Na] & MASK) EQNE (((type *)&b)[Nb] & MASK)
if both of A's fie
@1 @2))
-/* These are used by ifcombine fold_truth_andor. */
-(match (any_convert @0)
- (convert @0))
-(match (any_convert @0)
- (view_convert @0))
-(match (bit_and_cst @0 @1)
- (bit_and @0 uniform_integer_cst_p@1))
-(match (rshift_cst @0 @1)
- (rshift @0 uniform_integer_cst_p@1)
- (if (tree_f
le } */
/* { dg-timeout-factor 2.0 } */
-/* { dg-options "-O1 -fno-reorder-blocks -fno-tree-cselim --save-temps" } */
+/* { dg-options "-O1 -fno-reorder-blocks -fno-tree-cselim -fno-tree-ifcombine
--save-temps" } */
__attribute__((noinline, noclone)) int
--
Alexa
This patch introduces various improvements to the logic that merges
field compares, while moving it into ifcombine.
Before the patch, we could merge:
(a.x1 EQNE b.x1) ANDOR (a.y1 EQNE b.y1)
into something like:
(((type *)&a)[Na] & MASK) EQNE (((type *)&b)[Nb] & MASK)
if both of A's field
On Dec 22, 2024, Alexandre Oliva wrote:
> for gcc/testsuite/ChangeLog
> PR testsuite/118025
> * gcc.dg/field-merge-1.c: Convert constants to desired types.
> * gcc.dg/field-merge-3.c: Likewise.
> * gcc.dg/field-merge-4.c: Likewise.
> * gcc.
On Dec 21, 2024, Alexandre Oliva wrote:
> On Dec 20, 2024, Jakub Jelinek wrote:
>> On Wed, Dec 18, 2024 at 12:59:11AM -0300, Alexandre Oliva wrote:
>>> * gcc.dg/field-merge-16.c: New.
>> Note the test FAILs on i686-linux or on x86_64-linux with -m32.
> Indeed, t
file -mcmodel=large" } */
+/* { dg-require-profiling "-fprofile" } */
/* { dg-skip-if "PR90698" { *-*-darwin* } } */
/* { dg-skip-if "PR113909" { *-*-solaris2* } } */
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activi
-target vect_float } */
-/* { dg-additional-options "-Ofast -mavx" { target avx_runtime } } */
+/* { dg-additional-options "-Ofast" { target i?86-*-* x86_64-*-* } } */
+/* { dg-additional-options "-mavx" { target avx_runtime } } */
int A0[4] = {36,39,42,45};
--
b/gcc/testsuite/gcc.target/sparc/tls-ld-uint8.c
index 0dcff66eb15e4..7a7492197f952 100644
--- a/gcc/testsuite/gcc.target/sparc/tls-ld-uint8.c
+++ b/gcc/testsuite/gcc.target/sparc/tls-ld-uint8.c
@@ -1,6 +1,7 @@
/* { dg-do compile } */
/* { dg-options "-O2" }
/* { dg-add-options tls } */
On Jan 31, 2025, Richard Biener wrote:
> On Thu, Jan 30, 2025 at 8:49 PM Alexandre Oliva wrote:
>> Split -Ofast out of the avx conditional, so that it is passed on the
>> same targets that expect the transformation.
> It effectively would require -Ofast on all targets (it&#x
* We expect threadfull1 to eliminate the call to foo(), but not all targets
+ manage that at that point. Calling conventions (mandatory promotion) play a
+ role, but there's more than that. */
+/* { dg-final { scan-tree-dump-not "foo" "optimized" } } */
--
Alexandre Ol
That's because the assert went in as part of an earlier patch. I take
it it should be backed out along with the to-be-split-out bits above,
right?
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Enginee
ilp32 } */
-/* { dg-options "-O1 -mvis3" } */
+/* { dg-skip-if "prevents fsmuld" { *-*-* } { "-mfix-ut699" } { "" } } */
+/* { dg-options "-O1 -mvis3 -mcpu=ultrasparc" } */
float fnegs (float a)
{
--
Alexandre Oliva, happy hackerhttp
align-functions -mcpu=niagara7" } */
+/* { dg-options "-falign-functions -mtune=niagara7" } */
/* { dg-final { scan-assembler "\.align 64" } } */
void foo(void) {}
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist
On Feb 6, 2025, Sam James wrote:
> Richard Biener writes:
>> On Thu, Feb 6, 2025 at 2:41 PM Alexandre Oliva wrote:
>>>
>>> On Jan 27, 2025, Richard Biener wrote:
>>> > (I see the assert is no longer in the patch).
>>>
>>> That
ts expected asm" { *-*-* } { "-mfix-ut699" } { "" } } */
/* { dg-require-effective-target ilp32 } */
/* { dg-options "-mcpu=ultrasparc -O" } */
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist
On Dec 10, 2024, Richard Biener wrote:
> On Mon, Dec 9, 2024 at 8:55 PM Alexandre Oliva wrote:
>> Regstrapped on x86_64-linux-gnu. The aarch64 CI had reported a
>> regression with v3, so I'm also running a regstrap on aarch64-linux-gnu,
>> but it will still be a w
On Dec 11, 2024, Richard Biener wrote:
> Am 11.12.2024 um 14:50 schrieb Alexandre Oliva :
>> if (get_best_mode (end_bit - first_bit, first_bit, 0, 0,
> I think These 0, 0 args are supposed to indicate Maximum extent of the
> resulting Access - but Widening to the largest Mode fi
lr_align, BITS_PER_WORD, volatilep, &rnmode)
+ || !get_best_mode (end_bit - boundary, boundary, 0, lr_end_region,
+ lr_align, BITS_PER_WORD, volatilep, &rnmode2))
return 0;
r_split_load = tru
|| q.c.b != p.c.b
+ || p.d.b != q.d.b
+ || q.d.a != p.d.a
+ )
+__builtin_abort ();
+ return;
+}
+
+int main () {
+ f ();
+ g ();
+ h ();
+ return 0;
+}
+
+/* { dg-final { scan-tree-dump-times "optimizing" 9 "ifcombine" } } */
--
Alexandre Oliva, happy hac
+ return 0;
+}
+
+int main()
+{
+ unsigned char t = 0x81;
+ int x = t, y;
+ int tt = f(&x, &y);
+ if (!tt)
+__builtin_abort();
+}
+
+/* { dg-final { scan-tree-dump-not "optimizing" "ifcombine" } } */
--
Alexandre Oliva, happy hackerhttps://FSFLA.org
? -2 << (__CHAR_BIT__ - 1)
-: -1 & ((1 << (__CHAR_BIT__ - 1) << 1) - 1)), 42 };
+struct s2 q = { -2 << (__CHAR_BIT__ - 1), 42 };
void f (void) {
if (0
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software A
name (data->used, *t, data->outer);
return NULL;
}
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
On Dec 17, 2024, Alexandre Oliva wrote:
> diff --git a/gcc/gimple-fold.cc b/gcc/gimple-fold.cc
> index 92f02ddd77408..06913d57f8aec 100644
> --- a/gcc/gimple-fold.cc
> +++ b/gcc/gimple-fold.cc
> @@ -7554,13 +7554,10 @@ decode_field_reference (tree *pexp, HOST_WIDE_I
On Dec 17, 2024, Alexandre Oliva wrote:
> On Dec 17, 2024, Sam James wrote:
>>> +/* { dg-options "-O3 -fno-tree-copy-prop -fno-tree-vrp" */
>> Missing closing '}'.
> Thanks, I've fixed that, and retested.
Here's the consolidated patch what I
nt even if otherwise unused
+ if (c < 0 && (b&1))
+return 1;
+ return 0;
+}
+
+int main()
+{
+ unsigned char t = 0x81;
+ int x = t, y;
+ int tt = f(&x, &y);
+ if (!tt)
+__builtin_abort();
+}
+
+/* { dg-final { scan-tree-dump-not "optimizing" "
2 };
-struct s2 q = { (le
-? -2 << (__CHAR_BIT__ - 1)
-: -1 & ((1 << (__CHAR_BIT__ - 1) << 1) - 1)), 42 };
+struct s2 q = { -2 << (__CHAR_BIT__ - 1), 42 };
void f (void) {
if (0
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs
+
+void f ()
+{
+ if (p.a || p.b || p.c || p.d)
+return;
+ __builtin_abort ();
+}
+
+void g ()
+{
+ if (q.a || q.b || q.c || q.d)
+__builtin_abort ();
+}
+
+int main () {
+ f ();
+ g ();
+ return 0;
+}
+
+/* { dg-final { scan-tree-dump-times "optimizing" 6 "ifcombine&quo
On Dec 17, 2024, Sam James wrote:
>> +/* { dg-options "-O3 -fno-tree-copy-prop -fno-tree-vrp" */
> Missing closing '}'.
Thanks, I've fixed that, and retested.
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activi
/* As we strip non-widening conversions in finding a common
name that is tested make sure to end up with an integral
@@ -944,7 +944,8 @@ ifcombine_ifandif (basic_block inner_cond_bb, bool
inner_inv,
}
/* See if we have two comparisons that we can merge into one. */
- else if (TRE
ncerns you might have, I'd be
happy to double-check, and hopefully alleviate them, or at least follow
leads to other problems I haven't considered.
Thanks,
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain E
=1 -Os -fno-tree-coalesce-vars -finline-stringops"
} */
+
+typedef unsigned v32su __attribute__((vector_size (32)));
+
+v32su foo_v32su_4;
+
+unsigned
+foo (v32su v32su_2)
+{
+ v32su_2 *= v32su_2;
+ if (foo_v32su_4[3])
+v32su_2 &= (v32su){};
+ return v32su_2[1];
+}
--
Alexandre
file mode 100644
index 0..6c24cad652968
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/strub-pr118007.c
@@ -0,0 +1,5 @@
+/* { dg-require-effective-target strub } */
+/* { dg-do compile } */
+/* { dg-options "-fstrub=all -O2" } */
+
+void rb_ec_error_print(struct rb_execution_context_st
abort();
+}
+
+static int privileged_traced_syscall() {
+ return _raw_syscall(0, 0, 0);
+}
+
+void privileged_traced_raise() {
+ privileged_traced_syscall();
+ __builtin_unreachable ();
+}
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GN
tion entirely, we need not
commit to an exit block yet, it's as if we'd merged the blocks and
were starting afresh. This is sound as long as we never replace
@@ -1369,6 +1373,15 @@ public:
{}
/* opt_pass methods: */
+ bool gate (function *) final override {
ous
blocks is needed, I'll be happy to introduce a param for that.
Thanks,
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Exclu
On Dec 20, 2024, Jakub Jelinek wrote:
> On Wed, Dec 18, 2024 at 12:59:11AM -0300, Alexandre Oliva wrote:
>> * gcc.dg/field-merge-16.c: New.
> Note the test FAILs on i686-linux or on x86_64-linux with -m32.
Indeed, thanks. Here's a fix.
On 32-bit hosts, data types with
This patch introduces various improvements to the logic that merges
field compares, while moving it into ifcombine.
Before the patch, we could merge:
(a.x1 EQNE b.x1) ANDOR (a.y1 EQNE b.y1)
into something like:
(((type *)&a)[Na] & MASK) EQNE (((type *)&b)[Nb] & MASK)
if both of A's fie
nd_bb)
+ return false;
tree t1, t2;
bool logical_op_non_short_circuit = LOGICAL_OP_NON_SHORT_CIRCUIT;
if (param_logical_op_non_short_circuit != -1)
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Act
On Dec 3, 2024, Richard Biener wrote:
> On Fri, Nov 22, 2024 at 10:22 PM Alexandre Oliva wrote:
>> + /* Identify the load, if there is one. */
>> + if (TREE_CODE (exp) == SSA_NAME
>> + && !SSA_NAME_IS_DEFAULT_DEF (exp))
>> +{
>> +
Hello, Dimitar,
On Dec 22, 2024, Dimitar Dimitrov wrote:
> On Sat, Dec 21, 2024 at 02:28:33AM -0300, Alexandre Oliva wrote:
>> On Dec 20, 2024, Jakub Jelinek wrote:
>>
>> > On Wed, Dec 18, 2024 at 12:59:11AM -0300, Alexandre Oliva wrote:
>> >> * gcc.dg/fiel
;rl_load, rl_loc);
rr_inner = decode_field_reference (&rr_arg, &rr_bitsize, &rr_bitpos,
&rr_unsignedp, &rr_reversep, &volatilep,
- &rr_and_mask, &rr_signbit, &r_xor, 0,
+
(lr_inner
+ && tree_fits_poly_uint64_p (TYPE_SIZE (TREE_TYPE (lr_inner
lr_end_region = tree_to_poly_uint64 (TYPE_SIZE (TREE_TYPE (lr_inner)));
if (!get_best_mode (end_bit - first_bit, first_bit, 0, lr_end_region,
lr_align, BITS_PER_WORD, volatilep, &rnmode)
id y() {
+ for (;;) {
+b[w(8, *p)] = h;
+for (; a + k; j = o)
+ i &= c = x(6) < 0;
+ }
+}
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key
c.dg/dfp/int128-4.c
index b91800989f947..b83e7e4221cec 100644
--- a/gcc/testsuite/gcc.dg/dfp/int128-4.c
+++ b/gcc/testsuite/gcc.dg/dfp/int128-4.c
@@ -1,5 +1,6 @@
/* PR libgcc/65833 */
-/* { dg-do run { target { int128 && bitint } } } */
+/* { dg-require-effective-target int128 } */
+/* {
ECL_P (TREE_OPERAND (expr, 0)) && !bit_field_ref_in_bounds_p (expr))
+ return true;
+ /* Fall through. */
+
case COMPONENT_REF:
case REALPART_EXPR:
case IMAGPART_EXPR:
- case BIT_FIELD_REF:
case VIEW_CONVERT_EXPR:
case WITH_SIZE_EXPR:
expr = TREE_
On Jan 22, 2025, Alexandre Oliva wrote:
> I have another patch coming up that doesn't raise concerns for me, so
> I'll hold off from installing the above, in case you also prefer the
> other one.
Unlike other access patterns, BIT_FIELD_REFs aren't regarded as
On Jan 22, 2025, Alexandre Oliva wrote:
> I have another patch coming up that doesn't raise concerns for me, so
> I'll hold off from installing the above, in case you also prefer the
> other one.
And here's an unrelated bit that came to mind while working on this, but
{
+return (((0x80 & (c&-1)) != 0) && ((0x1248c0 & (c&-1)) == 0x124880));
+}
+
+__attribute__((noipa))
+int test4(signed char c)
+{
+return (((0x400 & (c&-1)) == 0) && ((0x40 & (c&-1)) == 0x40));
+}
+
+int main() {
+ if (test(0x80) == 0 ||
you mentioned, it's also needed to satisfy the
assert, because we could pick the wrong load to replace, and then we'd
get a mismatch.
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
ern bool operation_could_trap_p (enum tree_code, bool, bool, tree);
+extern bool access_in_bounds_of_type_p (tree, poly_uint64, poly_uint64);
extern bool tree_could_trap_p (tree);
extern tree rewrite_to_non_trapping_overflow (tree);
extern bool stmt_could_throw_p (function *, gimple *);
--
Al
|| !c) {
+signed char e = b[10];
+d = e < 0 || b[10] > 1;
+if (d)
+ __builtin_abort ();
+ }
+ return 0;
+}
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
M
d))
> & !tree_could_trap_p (inner)
> (&& btw) looks like bad style.
Oops, yeah, typo (keyboard woes), thanks.
> Did you consider not applying this ifcombine to tree_could_trap_p original
> refs
> at all?
Yeah. It works. But then I figured we could take a safe step f
alignment and size seem to guarantee we won't
combine accesses with different trapping properties (though one could
presumably be marked as non-trapping if it is dominated by another
access to the same word), but I should probably guard against this.
Will do.
--
Alexandre Oliva, happy ha
on/118456 */
+/* Check that compares with constants take into account sign/zero extension of
+ both the bitfield and of the shifting type. */
+
+#define shift (__CHAR_BIT__ - 4)
+
+struct S {
+ signed char a : shift + 2;
+ signed char b : shift + 2;
+ short ignore[0];
+} s;
+
+__attribute__(
pa))
+int f3 ()
+{
+ if (c.d == a.d
+ && c.e == a.e)
+return 0;
+ return -1;
+}
+
+__attribute__((noipa))
+int f4 ()
+{
+ if (c.d != a.d
+ || c.e != a.e)
+return -1;
+ return 0;
+}
+
+int main() {
+ if (f1 () < 0
+ || f2 () < 0
+ || f3 () < 0
+ |
dth
and bit position. */
if (l_const.get_precision ())
@@ -8550,10 +8559,6 @@ fold_truth_andor_for_ifcombine (enum tree_code code,
tree truth_type,
return constant_boolean_node (wanted_code == NE_EXPR, truth_type);
}
- /* When we set l_const, we also set r_const, so we need
24-December/672161.html
https://gcc.gnu.org/pipermail/gcc-patches/2024-December/672162.html
They're meant to go together, but each one should suffice.
Apologies for the noise
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist
On Dec 20, 2024, Jakub Jelinek wrote:
> On Wed, Dec 18, 2024 at 12:59:11AM -0300, Alexandre Oliva wrote:
>> * gcc.dg/field-merge-16.c: New.
> Note the test FAILs on i686-linux or on x86_64-linux with -m32.
Also fixed herein.
A number of tests that check for specific ifcombine tra
On Jan 10, 2025, Alexandre Oliva wrote:
> dfp.exp sets the default to compile when dfprt is not available, but
> some dfp bitint tests override the default without that requirement,
> and try to run even when dfprt is not available.
> Instead of overriding the default, rewrite the
/* { dg-options "-O2" } */
__attribute__((noipa)) int
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
}
+// { dg-skip-if "requires hosted libstdc++ for string" { ! hostedlib } }
#include
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and
dg-final { scan-assembler "umlal" } } */
+/* After zero-extending both to SImode, either signed- or unsigned-widening
+ multiply will do. */
+/* { dg-final { scan-assembler {[us]mlal} } } */
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activ
+{
+ unsigned short a = 0x78 - 0x80 - 0x80;
+ if (foo (&a) != 0 || bar (&a) != (a > 0xff))
+__builtin_abort ();
+ unsigned short b = 0x88;
+ if (foo (&b) != 1 || bar (&b) != 1)
+ __builtin_abort ();
+ unsigned short c = 8;
+ if (foo (&c) != 0 || bar (&c) != 0)
+
ltilib flags override anything we can apply to a test, so
# skip if any of the above options are set there.
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for in
/* PR tree-optimization/118805 */
+
+/* Test that ifcombine doesn't get confused by tests for the sign bit of
+ extended values that would normally be folded before. */
+
+unsigned char a = 255;
+int b;
+int main() {
+ int c = 0;
+ if (c > a && a >= 255)
+__builtin_abor
] == "N" } {
+ return
+}
+
+# Evaluate selector, return if it does not match.
+switch [dg-process-target-1 [lindex $args 1]] {
+ "N" { return }
+ "P" { return }
+}
+
+eval dg-do $args
+}
+
# Skip the test (report it as UNS
ap-trampoline, 0 otherwise.
proc check_effective_target_heap_trampoline {} {
if { [istarget aarch64*-*-linux*]
-|| [istarget i?86-*-darwin*]
-|| [istarget x86_64-*-darwin*]
-|| [istarget i?86-*-linux*]
-|| [istarget x86_64-*-linux*] } {
+|| { [check_effective_target_x86]
+ && { [istarget *-*-darwin*]
+ || [istarget *-*-linux*] } } } {
return 1
}
return 0
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity
Excluding neuro-others for not behaving ""normal"" is *not* inclusive
k-function-bodies PREFIX TERMINATOR[ OPTION[ SELECTOR
[MATCHED]]] } }
#
# See sourcebuild.texi for details.
--
Alexandre Oliva, happy hackerhttps://FSFLA.org/blogs/lxo/
Free Software Activist GNU Toolchain Engineer
More tolerance and less prejudice are key for
et riscv*-*-*] } {
return 1
@@ -14093,10 +14098,9 @@ proc dg-require-python-h { args } {
# Return 1 if the target supports heap-trampoline, 0 otherwise.
proc check_effective_target_heap_trampoline {} {
if { [istarget aarch64*-*-linux*]
-|| [istarget i?86-*-darwin*]
-
On Feb 24, 2025, Mike Stump wrote:
> I thought I saw one more needing review.
Thanks, Richard Sandiford reviewed it.
https://gcc.gnu.org/pipermail/gcc-patches/2025-February/676031.html
--
Alexandre Oliva, happy hackerhttps://blog.lx.oliva.nom.br/
Free Software Activist FS
On Feb 13, 2025, Alexandre Oliva wrote:
> @@ -14108,10 +14113,9 @@ proc dg-require-python-h { args } {
> # Return 1 if the target supports heap-trampoline, 0 otherwise.
> proc check_effective_target_heap_trampoline {} {
> if { [istarget aarch64*-*-linux*]
> - ||
o please go ahead with your patch.
Sorry about the delay.
--
Alexandre Oliva, happy hackerhttps://blog.lx.oliva.nom.br/
Free Software Activist FSFLA co-founder GNU Toolchain Engineer
Learn the truth about Richard Stallman at https://stallmansupport.org/
the trunk, only gcc-14.
Patch withdrawn.
H-P, perhaps it's time to drop the XFAIL on cris in the trunk?
--
Alexandre Oliva, happy hackerhttps://blog.lx.oliva.nom.br/
Free Software Activist FSFLA co-founder GNU Toolchain Engineer
More tolerance and less prejudice
dg-do compile { target { ! riscv_abi_e } } } */
+/* { dg-do compile { target { { ! riscv_abi_e } && rv64 } } } */
/* { dg-add-options riscv_v } */
/* { dg-additional-options "-std=gnu99 -O3 -fno-schedule-insns
-fno-schedule-insns2" } */
--
Alexandre Oliva, happy
On Mar 31, 2025, Jeff Law wrote:
> On 3/31/25 1:05 PM, Alexandre Oliva wrote:
>> The desired vw{add,sub}.wx instructions don't come up on rv32 for
>> the
>> first two functions, we get v{add,sub}.vx instead.
>> I suppose this is an oversight, and something a
ion/112822 */
/* { dg-do compile { target c++17 } } */
/* { dg-options "-w -O2" } */
+/* Required for altivec double vectors. */
+/* { dg-additional-options "-mvsx" { target powerpc*-*-* } } */
/* Verify we do not ICE on the following noisy creduced test case. */
--
|| [istarget alpha*-*-*]
|| [istarget aarch64*-*-*]
|| [is-effective-target arm_neon]
--
Alexandre Oliva, happy hackerhttps://blog.lx.oliva.nom.br/
Free Software Activist FSFLA co-founder GNU Toolchain Engineer
More tolerance and less prejudice are key for
powerpc-ibm-aix*
powerpc-wrs-vxworks* } } */
+/* { dg-additional-options "-Wno-psabi" { target powerpc-ibm-aix*
powerpc-wrs-vxworks* powerpc-*-elf } } */
typedef int __attribute__((__vector_size__(16))) vectype;
--
Alexandre Oliva, happy hackerhttps://blog.lx.oliva.nom
cc.target/powerpc/bfp/scalar-test-neg-5.c
@@ -1,6 +1,8 @@
/* { dg-do compile { target { powerpc*-*-* } } } */
-/* { dg-options "-mdejagnu-cpu=power8 -mvsx" } */
+/* { dg-options "-mdejagnu-cpu=power8 -mvsx -mfloat128" } */
+/* { dg-require-effective-target ppc_ieee128_ok }
ieee } */
/* { dg-require-effective-target fenv_exceptions_long_double } */
/* { dg-skip-if "fenv" { powerpc-ibm-aix* } } */
--
Alexandre Oliva, happy hackerhttps://blog.lx.oliva.nom.br/
Free Software Activist FSFLA co-founder GNU Toolchain Engineer
More toleranc
ns are correctly expanded and match the
expected result.
--
Alexandre Oliva, happy hackerhttps://blog.lx.oliva.nom.br/
Free Software Activist FSFLA co-founder GNU Toolchain Engineer
More tolerance and less prejudice are key for inclusion and diversity.
Excluding neuro-others
1401 - 1500 of 1959 matches
Mail list logo
