* [PATCH] c++: Don't shortcut TREE_CONSTANT vector type CONSTRUCTORs in cxx_eval_constant_expression [PR107295]
@ 2022-10-19 7:48 Jakub Jelinek
2022-10-20 14:51 ` Jason Merrill
0 siblings, 1 reply; 4+ messages in thread
From: Jakub Jelinek @ 2022-10-19 7:48 UTC (permalink / raw)
To: Jason Merrill; +Cc: gcc-patches
Hi!
The excess precision support broke building skia (dependency of firefox)
on ia32 (it has something like the a constexpr variable), but as the other
cases show, it is actually a preexisting problem if one uses casts from
constants with wider floating point types.
The problem is that cxx_eval_constant_expression tries to short-cut
processing of TREE_CONSTANT CONSTRUCTORs if they satisfy
reduced_constant_expression_p - instead of calling cxx_eval_bare_aggregate
on them it just verifies flags and if they are TREE_CONSTANT even after
that, just fold.
Now, on the testcase we have a TREE_CONSTANT CONSTRUCTOR containing
TREE_CONSTANT NOP_EXPR of REAL_CST. And, fold, which isn't recursive,
doesn't optimize that into VECTOR_CST, while later on we are only able
to optimize VECTOR_CST arithmetics, not arithmetics with vector
CONSTRUCTORs.
The following patch fixes that by only returning what fold returned
if for vector types it returned VECTOR_CST, otherwise let us
call cxx_eval_bare_aggregate. That function will try to constant
evaluate all the elements and if anything changes, return a CONSTRUCTOR,
in the vector type cases with fold called on it at the end.
Now, just calling cxx_eval_bare_aggregate for vector types doesn't work
either (e.g. constexpr-builtin4.C breaks), because cxx_eval_bare_aggregate
if nothing changes (like all elts are already REAL_CSTs or INTEGER_CSTs)
will return the old CONSTRUCTOR and nothing folds it into a VECTOR_CST.
Also, the reason for the short-cutting is I think trying to avoid
allocating a new CONSTRUCTOR when nothing changes and we just create
GC garbage by it.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2022-10-19 Jakub Jelinek <jakub@redhat.com>
PR c++/107295
* constexpr.cc (cxx_eval_constant_expression) <case CONSTRUCTOR>:
Don't short-cut TREE_CONSTANT vector ctors if fold doesn't turn them
into VECTOR_CST.
* g++.dg/ext/vector42.C: New test.
--- gcc/cp/constexpr.cc.jj 2022-10-17 12:29:33.518016420 +0200
+++ gcc/cp/constexpr.cc 2022-10-19 01:29:28.761935708 +0200
@@ -7391,7 +7391,12 @@ cxx_eval_constant_expression (const cons
VECTOR_CST if applicable. */
verify_constructor_flags (t);
if (TREE_CONSTANT (t))
- return fold (t);
+ {
+ r = fold (t);
+ if (TREE_CODE (TREE_TYPE (t)) != VECTOR_TYPE
+ || TREE_CODE (r) == VECTOR_CST)
+ return r;
+ }
}
r = cxx_eval_bare_aggregate (ctx, t, lval,
non_constant_p, overflow_p);
--- gcc/testsuite/g++.dg/ext/vector42.C.jj 2022-10-18 12:33:42.938510483 +0200
+++ gcc/testsuite/g++.dg/ext/vector42.C 2022-10-18 12:32:27.448544476 +0200
@@ -0,0 +1,12 @@
+// PR c++/107295
+// { dg-do compile { target c++11 } }
+
+template <typename T> struct A {
+ typedef T __attribute__((vector_size (sizeof (int)))) V;
+};
+template <int, typename T> using B = typename A<T>::V;
+template <typename T> using V = B<4, T>;
+using F = V<float>;
+constexpr F a = F () + 0.0f;
+constexpr F b = F () + (float) 0.0;
+constexpr F c = F () + (float) 0.0L;
Jakub
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] c++: Don't shortcut TREE_CONSTANT vector type CONSTRUCTORs in cxx_eval_constant_expression [PR107295]
2022-10-19 7:48 [PATCH] c++: Don't shortcut TREE_CONSTANT vector type CONSTRUCTORs in cxx_eval_constant_expression [PR107295] Jakub Jelinek
@ 2022-10-20 14:51 ` Jason Merrill
2022-10-21 7:30 ` [PATCH] c++, v2: " Jakub Jelinek
0 siblings, 1 reply; 4+ messages in thread
From: Jason Merrill @ 2022-10-20 14:51 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: gcc-patches
On 10/19/22 03:48, Jakub Jelinek wrote:
> Hi!
>
> The excess precision support broke building skia (dependency of firefox)
> on ia32 (it has something like the a constexpr variable), but as the other
> cases show, it is actually a preexisting problem if one uses casts from
> constants with wider floating point types.
> The problem is that cxx_eval_constant_expression tries to short-cut
> processing of TREE_CONSTANT CONSTRUCTORs if they satisfy
> reduced_constant_expression_p - instead of calling cxx_eval_bare_aggregate
> on them it just verifies flags and if they are TREE_CONSTANT even after
> that, just fold.
> Now, on the testcase we have a TREE_CONSTANT CONSTRUCTOR containing
> TREE_CONSTANT NOP_EXPR of REAL_CST. And, fold, which isn't recursive,
> doesn't optimize that into VECTOR_CST, while later on we are only able
> to optimize VECTOR_CST arithmetics, not arithmetics with vector
> CONSTRUCTORs.
> The following patch fixes that by only returning what fold returned
> if for vector types it returned VECTOR_CST, otherwise let us
> call cxx_eval_bare_aggregate. That function will try to constant
> evaluate all the elements and if anything changes, return a CONSTRUCTOR,
> in the vector type cases with fold called on it at the end.
> Now, just calling cxx_eval_bare_aggregate for vector types doesn't work
> either (e.g. constexpr-builtin4.C breaks), because cxx_eval_bare_aggregate
> if nothing changes (like all elts are already REAL_CSTs or INTEGER_CSTs)
> will return the old CONSTRUCTOR and nothing folds it into a VECTOR_CST.
That seems like a bug; for VECTOR_TYPE we should fold even if !changed.
> Also, the reason for the short-cutting is I think trying to avoid
> allocating a new CONSTRUCTOR when nothing changes and we just create
> GC garbage by it.
We might limit the shortcut to non-vector types by hoisting the vector
check in reduced_constant_expression_p out of the
CONSTRUCTOR_NO_CLEARING condition:
> if (CONSTRUCTOR_NO_CLEARING (t))
> {
> if (TREE_CODE (TREE_TYPE (t)) == VECTOR_TYPE)
> /* An initialized vector would have a VECTOR_CST. */
> return false;
then we could remove the fold in the shortcut.
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
>
> 2022-10-19 Jakub Jelinek <jakub@redhat.com>
>
> PR c++/107295
> * constexpr.cc (cxx_eval_constant_expression) <case CONSTRUCTOR>:
> Don't short-cut TREE_CONSTANT vector ctors if fold doesn't turn them
> into VECTOR_CST.
>
> * g++.dg/ext/vector42.C: New test.
>
> --- gcc/cp/constexpr.cc.jj 2022-10-17 12:29:33.518016420 +0200
> +++ gcc/cp/constexpr.cc 2022-10-19 01:29:28.761935708 +0200
> @@ -7391,7 +7391,12 @@ cxx_eval_constant_expression (const cons
> VECTOR_CST if applicable. */
> verify_constructor_flags (t);
> if (TREE_CONSTANT (t))
> - return fold (t);
> + {
> + r = fold (t);
> + if (TREE_CODE (TREE_TYPE (t)) != VECTOR_TYPE
> + || TREE_CODE (r) == VECTOR_CST)
> + return r;
> + }
> }
> r = cxx_eval_bare_aggregate (ctx, t, lval,
> non_constant_p, overflow_p);
> --- gcc/testsuite/g++.dg/ext/vector42.C.jj 2022-10-18 12:33:42.938510483 +0200
> +++ gcc/testsuite/g++.dg/ext/vector42.C 2022-10-18 12:32:27.448544476 +0200
> @@ -0,0 +1,12 @@
> +// PR c++/107295
> +// { dg-do compile { target c++11 } }
> +
> +template <typename T> struct A {
> + typedef T __attribute__((vector_size (sizeof (int)))) V;
> +};
> +template <int, typename T> using B = typename A<T>::V;
> +template <typename T> using V = B<4, T>;
> +using F = V<float>;
> +constexpr F a = F () + 0.0f;
> +constexpr F b = F () + (float) 0.0;
> +constexpr F c = F () + (float) 0.0L;
>
> Jakub
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH] c++, v2: Don't shortcut TREE_CONSTANT vector type CONSTRUCTORs in cxx_eval_constant_expression [PR107295]
2022-10-20 14:51 ` Jason Merrill
@ 2022-10-21 7:30 ` Jakub Jelinek
2022-10-21 13:19 ` Jason Merrill
0 siblings, 1 reply; 4+ messages in thread
From: Jakub Jelinek @ 2022-10-21 7:30 UTC (permalink / raw)
To: Jason Merrill; +Cc: gcc-patches
On Thu, Oct 20, 2022 at 10:51:14AM -0400, Jason Merrill wrote:
> That seems like a bug; for VECTOR_TYPE we should fold even if !changed.
>
> > Also, the reason for the short-cutting is I think trying to avoid
> > allocating a new CONSTRUCTOR when nothing changes and we just create
> > GC garbage by it.
>
> We might limit the shortcut to non-vector types by hoisting the vector check
> in reduced_constant_expression_p out of the CONSTRUCTOR_NO_CLEARING
> condition:
>
> > if (CONSTRUCTOR_NO_CLEARING (t))
> > {
> > if (TREE_CODE (TREE_TYPE (t)) == VECTOR_TYPE)
> > /* An initialized vector would have a VECTOR_CST. */
> > return false;
>
> then we could remove the fold in the shortcut.
Ok, so like this?
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2022-10-21 Jakub Jelinek <jakub@redhat.com>
PR c++/107295
* constexpr.cc (reduced_constant_expression_p) <case CONSTRUCTOR>:
Return false for VECTOR_TYPE CONSTRUCTORs even without
CONSTRUCTOR_NO_CLEARING set on them.
(cxx_eval_bare_aggregate): If constant but !changed, fold before
returning VECTOR_TYPE_P CONSTRUCTOR.
(cxx_eval_constant_expression) <case CONSTRUCTOR>: Don't fold
TREE_CONSTANT CONSTRUCTOR, just return it.
* g++.dg/ext/vector42.C: New test.
--- gcc/cp/constexpr.cc.jj 2022-10-19 11:20:28.960225787 +0200
+++ gcc/cp/constexpr.cc 2022-10-20 18:43:42.952440364 +0200
@@ -3104,12 +3104,12 @@ reduced_constant_expression_p (tree t)
case CONSTRUCTOR:
/* And we need to handle PTRMEM_CST wrapped in a CONSTRUCTOR. */
tree field;
+ if (TREE_CODE (TREE_TYPE (t)) == VECTOR_TYPE)
+ /* An initialized vector would have a VECTOR_CST. */
+ return false;
if (CONSTRUCTOR_NO_CLEARING (t))
{
- if (TREE_CODE (TREE_TYPE (t)) == VECTOR_TYPE)
- /* An initialized vector would have a VECTOR_CST. */
- return false;
- else if (TREE_CODE (TREE_TYPE (t)) == ARRAY_TYPE)
+ if (TREE_CODE (TREE_TYPE (t)) == ARRAY_TYPE)
{
/* There must be a valid constant initializer at every array
index. */
@@ -4956,8 +4956,14 @@ cxx_eval_bare_aggregate (const constexpr
TREE_SIDE_EFFECTS (ctx->ctor) = side_effects_p;
}
}
- if (*non_constant_p || !changed)
+ if (*non_constant_p)
return t;
+ if (!changed)
+ {
+ if (VECTOR_TYPE_P (type))
+ t = fold (t);
+ return t;
+ }
t = ctx->ctor;
if (!t)
t = build_constructor (type, NULL);
@@ -7387,11 +7393,10 @@ cxx_eval_constant_expression (const cons
case CONSTRUCTOR:
if (TREE_CONSTANT (t) && reduced_constant_expression_p (t))
{
- /* Don't re-process a constant CONSTRUCTOR, but do fold it to
- VECTOR_CST if applicable. */
+ /* Don't re-process a constant CONSTRUCTOR. */
verify_constructor_flags (t);
if (TREE_CONSTANT (t))
- return fold (t);
+ return t;
}
r = cxx_eval_bare_aggregate (ctx, t, lval,
non_constant_p, overflow_p);
--- gcc/testsuite/g++.dg/ext/vector42.C.jj 2022-10-20 17:57:42.767848544 +0200
+++ gcc/testsuite/g++.dg/ext/vector42.C 2022-10-20 17:57:42.767848544 +0200
@@ -0,0 +1,12 @@
+// PR c++/107295
+// { dg-do compile { target c++11 } }
+
+template <typename T> struct A {
+ typedef T __attribute__((vector_size (sizeof (int)))) V;
+};
+template <int, typename T> using B = typename A<T>::V;
+template <typename T> using V = B<4, T>;
+using F = V<float>;
+constexpr F a = F () + 0.0f;
+constexpr F b = F () + (float) 0.0;
+constexpr F c = F () + (float) 0.0L;
Jakub
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] c++, v2: Don't shortcut TREE_CONSTANT vector type CONSTRUCTORs in cxx_eval_constant_expression [PR107295]
2022-10-21 7:30 ` [PATCH] c++, v2: " Jakub Jelinek
@ 2022-10-21 13:19 ` Jason Merrill
0 siblings, 0 replies; 4+ messages in thread
From: Jason Merrill @ 2022-10-21 13:19 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: gcc-patches
On 10/21/22 03:30, Jakub Jelinek wrote:
> On Thu, Oct 20, 2022 at 10:51:14AM -0400, Jason Merrill wrote:
>> That seems like a bug; for VECTOR_TYPE we should fold even if !changed.
>>
>>> Also, the reason for the short-cutting is I think trying to avoid
>>> allocating a new CONSTRUCTOR when nothing changes and we just create
>>> GC garbage by it.
>>
>> We might limit the shortcut to non-vector types by hoisting the vector check
>> in reduced_constant_expression_p out of the CONSTRUCTOR_NO_CLEARING
>> condition:
>>
>>> if (CONSTRUCTOR_NO_CLEARING (t))
>>> {
>>> if (TREE_CODE (TREE_TYPE (t)) == VECTOR_TYPE)
>>> /* An initialized vector would have a VECTOR_CST. */
>>> return false;
>>
>> then we could remove the fold in the shortcut.
>
> Ok, so like this?
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
OK.
> 2022-10-21 Jakub Jelinek <jakub@redhat.com>
>
> PR c++/107295
> * constexpr.cc (reduced_constant_expression_p) <case CONSTRUCTOR>:
> Return false for VECTOR_TYPE CONSTRUCTORs even without
> CONSTRUCTOR_NO_CLEARING set on them.
> (cxx_eval_bare_aggregate): If constant but !changed, fold before
> returning VECTOR_TYPE_P CONSTRUCTOR.
> (cxx_eval_constant_expression) <case CONSTRUCTOR>: Don't fold
> TREE_CONSTANT CONSTRUCTOR, just return it.
>
> * g++.dg/ext/vector42.C: New test.
>
> --- gcc/cp/constexpr.cc.jj 2022-10-19 11:20:28.960225787 +0200
> +++ gcc/cp/constexpr.cc 2022-10-20 18:43:42.952440364 +0200
> @@ -3104,12 +3104,12 @@ reduced_constant_expression_p (tree t)
> case CONSTRUCTOR:
> /* And we need to handle PTRMEM_CST wrapped in a CONSTRUCTOR. */
> tree field;
> + if (TREE_CODE (TREE_TYPE (t)) == VECTOR_TYPE)
> + /* An initialized vector would have a VECTOR_CST. */
> + return false;
> if (CONSTRUCTOR_NO_CLEARING (t))
> {
> - if (TREE_CODE (TREE_TYPE (t)) == VECTOR_TYPE)
> - /* An initialized vector would have a VECTOR_CST. */
> - return false;
> - else if (TREE_CODE (TREE_TYPE (t)) == ARRAY_TYPE)
> + if (TREE_CODE (TREE_TYPE (t)) == ARRAY_TYPE)
> {
> /* There must be a valid constant initializer at every array
> index. */
> @@ -4956,8 +4956,14 @@ cxx_eval_bare_aggregate (const constexpr
> TREE_SIDE_EFFECTS (ctx->ctor) = side_effects_p;
> }
> }
> - if (*non_constant_p || !changed)
> + if (*non_constant_p)
> return t;
> + if (!changed)
> + {
> + if (VECTOR_TYPE_P (type))
> + t = fold (t);
> + return t;
> + }
> t = ctx->ctor;
> if (!t)
> t = build_constructor (type, NULL);
> @@ -7387,11 +7393,10 @@ cxx_eval_constant_expression (const cons
> case CONSTRUCTOR:
> if (TREE_CONSTANT (t) && reduced_constant_expression_p (t))
> {
> - /* Don't re-process a constant CONSTRUCTOR, but do fold it to
> - VECTOR_CST if applicable. */
> + /* Don't re-process a constant CONSTRUCTOR. */
> verify_constructor_flags (t);
> if (TREE_CONSTANT (t))
> - return fold (t);
> + return t;
> }
> r = cxx_eval_bare_aggregate (ctx, t, lval,
> non_constant_p, overflow_p);
> --- gcc/testsuite/g++.dg/ext/vector42.C.jj 2022-10-20 17:57:42.767848544 +0200
> +++ gcc/testsuite/g++.dg/ext/vector42.C 2022-10-20 17:57:42.767848544 +0200
> @@ -0,0 +1,12 @@
> +// PR c++/107295
> +// { dg-do compile { target c++11 } }
> +
> +template <typename T> struct A {
> + typedef T __attribute__((vector_size (sizeof (int)))) V;
> +};
> +template <int, typename T> using B = typename A<T>::V;
> +template <typename T> using V = B<4, T>;
> +using F = V<float>;
> +constexpr F a = F () + 0.0f;
> +constexpr F b = F () + (float) 0.0;
> +constexpr F c = F () + (float) 0.0L;
>
>
> Jakub
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2022-10-21 13:19 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-10-19 7:48 [PATCH] c++: Don't shortcut TREE_CONSTANT vector type CONSTRUCTORs in cxx_eval_constant_expression [PR107295] Jakub Jelinek
2022-10-20 14:51 ` Jason Merrill
2022-10-21 7:30 ` [PATCH] c++, v2: " Jakub Jelinek
2022-10-21 13:19 ` Jason Merrill
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).