* [PATCH] c++: Fix up handling of register ... asm ("...") vars in templates [PR33661, PR98847]
@ 2021-01-28 8:52 Jakub Jelinek
2021-01-28 13:14 ` Jason Merrill
0 siblings, 1 reply; 2+ messages in thread
From: Jakub Jelinek @ 2021-01-28 8:52 UTC (permalink / raw)
To: Jason Merrill; +Cc: gcc-patches
Hi!
As the testcase shows, for vars appearing in templates, we don't attach
the asm spec string to the pattern decls, nor pass it back to cp_finish_decl
during instantiation.
The following patch does that.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2021-01-27 Jakub Jelinek <jakub@redhat.com>
PR c++/33661
PR c++/98847
* decl.c (cp_finish_decl): For register vars with asmspec in templates
call set_user_assembler_name and set DECL_HARD_REGISTER.
* pt.c (tsubst_expr): When instantiating DECL_HARD_REGISTER vars,
pass asmspec_tree to cp_finish_decl.
* g++.target/i386/pr98847.C: New test.
--- gcc/cp/decl.c.jj 2021-01-07 22:54:34.000000000 +0100
+++ gcc/cp/decl.c 2021-01-27 10:48:14.399582310 +0100
@@ -7840,6 +7840,12 @@ cp_finish_decl (tree decl, tree init, bo
retrofit_lang_decl (decl);
SET_DECL_DEPENDENT_INIT_P (decl, true);
}
+
+ if (VAR_P (decl) && DECL_REGISTER (decl) && asmspec)
+ {
+ set_user_assembler_name (decl, asmspec);
+ DECL_HARD_REGISTER (decl) = 1;
+ }
return;
}
--- gcc/cp/pt.c.jj 2021-01-22 10:07:53.643466723 +0100
+++ gcc/cp/pt.c 2021-01-27 10:59:04.109126313 +0100
@@ -18227,6 +18227,7 @@ tsubst_expr (tree t, tree args, tsubst_f
bool const_init = false;
unsigned int cnt = 0;
tree first = NULL_TREE, ndecl = error_mark_node;
+ tree asmspec_tree = NULL_TREE;
maybe_push_decl (decl);
if (VAR_P (decl)
@@ -18250,7 +18251,18 @@ tsubst_expr (tree t, tree args, tsubst_f
now. */
predeclare_vla (decl);
- cp_finish_decl (decl, init, const_init, NULL_TREE, 0);
+ if (VAR_P (decl) && DECL_HARD_REGISTER (pattern_decl))
+ {
+ tree id = DECL_ASSEMBLER_NAME (pattern_decl);
+ const char *asmspec = IDENTIFIER_POINTER (id);
+ gcc_assert (asmspec[0] == '*');
+ asmspec_tree
+ = build_string (IDENTIFIER_LENGTH (id) - 1,
+ asmspec + 1);
+ TREE_TYPE (asmspec_tree) = char_array_type_node;
+ }
+
+ cp_finish_decl (decl, init, const_init, asmspec_tree, 0);
if (ndecl != error_mark_node)
cp_finish_decomp (ndecl, first, cnt);
--- gcc/testsuite/g++.target/i386/pr98847.C.jj 2021-01-27 11:01:55.312161637 +0100
+++ gcc/testsuite/g++.target/i386/pr98847.C 2021-01-27 11:00:26.601179659 +0100
@@ -0,0 +1,20 @@
+// PR c++/98847
+// { dg-do run }
+// { dg-options "-O2 -masm=att" }
+
+template <int N>
+int
+foo ()
+{
+ register int edx asm ("edx");
+ asm ("movl $1234, %%edx" : "=r" (edx));
+ return edx;
+}
+
+int
+main ()
+{
+ if (foo<0> () != 1234)
+ __builtin_abort ();
+ return 0;
+}
Jakub
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] c++: Fix up handling of register ... asm ("...") vars in templates [PR33661, PR98847]
2021-01-28 8:52 [PATCH] c++: Fix up handling of register ... asm ("...") vars in templates [PR33661, PR98847] Jakub Jelinek
@ 2021-01-28 13:14 ` Jason Merrill
0 siblings, 0 replies; 2+ messages in thread
From: Jason Merrill @ 2021-01-28 13:14 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: gcc-patches
On 1/28/21 3:52 AM, Jakub Jelinek wrote:
> Hi!
>
> As the testcase shows, for vars appearing in templates, we don't attach
> the asm spec string to the pattern decls, nor pass it back to cp_finish_decl
> during instantiation.
>
> The following patch does that.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
OK.
> 2021-01-27 Jakub Jelinek <jakub@redhat.com>
>
> PR c++/33661
> PR c++/98847
> * decl.c (cp_finish_decl): For register vars with asmspec in templates
> call set_user_assembler_name and set DECL_HARD_REGISTER.
> * pt.c (tsubst_expr): When instantiating DECL_HARD_REGISTER vars,
> pass asmspec_tree to cp_finish_decl.
>
> * g++.target/i386/pr98847.C: New test.
>
> --- gcc/cp/decl.c.jj 2021-01-07 22:54:34.000000000 +0100
> +++ gcc/cp/decl.c 2021-01-27 10:48:14.399582310 +0100
> @@ -7840,6 +7840,12 @@ cp_finish_decl (tree decl, tree init, bo
> retrofit_lang_decl (decl);
> SET_DECL_DEPENDENT_INIT_P (decl, true);
> }
> +
> + if (VAR_P (decl) && DECL_REGISTER (decl) && asmspec)
> + {
> + set_user_assembler_name (decl, asmspec);
> + DECL_HARD_REGISTER (decl) = 1;
> + }
> return;
> }
>
> --- gcc/cp/pt.c.jj 2021-01-22 10:07:53.643466723 +0100
> +++ gcc/cp/pt.c 2021-01-27 10:59:04.109126313 +0100
> @@ -18227,6 +18227,7 @@ tsubst_expr (tree t, tree args, tsubst_f
> bool const_init = false;
> unsigned int cnt = 0;
> tree first = NULL_TREE, ndecl = error_mark_node;
> + tree asmspec_tree = NULL_TREE;
> maybe_push_decl (decl);
>
> if (VAR_P (decl)
> @@ -18250,7 +18251,18 @@ tsubst_expr (tree t, tree args, tsubst_f
> now. */
> predeclare_vla (decl);
>
> - cp_finish_decl (decl, init, const_init, NULL_TREE, 0);
> + if (VAR_P (decl) && DECL_HARD_REGISTER (pattern_decl))
> + {
> + tree id = DECL_ASSEMBLER_NAME (pattern_decl);
> + const char *asmspec = IDENTIFIER_POINTER (id);
> + gcc_assert (asmspec[0] == '*');
> + asmspec_tree
> + = build_string (IDENTIFIER_LENGTH (id) - 1,
> + asmspec + 1);
> + TREE_TYPE (asmspec_tree) = char_array_type_node;
> + }
> +
> + cp_finish_decl (decl, init, const_init, asmspec_tree, 0);
>
> if (ndecl != error_mark_node)
> cp_finish_decomp (ndecl, first, cnt);
> --- gcc/testsuite/g++.target/i386/pr98847.C.jj 2021-01-27 11:01:55.312161637 +0100
> +++ gcc/testsuite/g++.target/i386/pr98847.C 2021-01-27 11:00:26.601179659 +0100
> @@ -0,0 +1,20 @@
> +// PR c++/98847
> +// { dg-do run }
> +// { dg-options "-O2 -masm=att" }
> +
> +template <int N>
> +int
> +foo ()
> +{
> + register int edx asm ("edx");
> + asm ("movl $1234, %%edx" : "=r" (edx));
> + return edx;
> +}
> +
> +int
> +main ()
> +{
> + if (foo<0> () != 1234)
> + __builtin_abort ();
> + return 0;
> +}
>
> Jakub
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2021-01-28 13:14 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-28 8:52 [PATCH] c++: Fix up handling of register ... asm ("...") vars in templates [PR33661, PR98847] Jakub Jelinek
2021-01-28 13:14 ` 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).