public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/114676] [12/13/14 Regression] DSE removes assignment that is used later
Date: Thu, 11 Apr 2024 12:35:54 +0000	[thread overview]
Message-ID: <bug-114676-4-BNHyD4W2Dl@http.gcc.gnu.org/bugzilla/> (raw)
In-Reply-To: <bug-114676-4@http.gcc.gnu.org/bugzilla/>

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114676

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
It depends on what the vec_xl*/vec_xst* documentation requires/user expect.
If the expectation is that the loads/stores should alias the scalar pointee of
the pointer type passed to those intrinsics, then
--- gcc/config/s390/s390-c.cc.jj        2024-01-03 11:51:54.847407588 +0100
+++ gcc/config/s390/s390-c.cc   2024-04-10 16:30:31.896197832 +0200
@@ -498,8 +498,8 @@ s390_expand_overloaded_builtin (location
        /* Build a vector type with the alignment of the source
           location in order to enable correct alignment hints to be
           generated for vl.  */
-       tree mem_type = build_aligned_type (return_type,
-                                           TYPE_ALIGN (TREE_TYPE (TREE_TYPE
((*arglist)[1]))));
+       unsigned align = TYPE_ALIGN (TREE_TYPE (TREE_TYPE ((*arglist)[1])));
+       tree mem_type = build_aligned_type (return_type, align);
        return build2 (MEM_REF, mem_type,
                       fold_build_pointer_plus ((*arglist)[1], (*arglist)[0]),
                       build_int_cst (TREE_TYPE ((*arglist)[1]), 0));
@@ -511,11 +511,13 @@ s390_expand_overloaded_builtin (location
        /* Build a vector type with the alignment of the target
           location in order to enable correct alignment hints to be
           generated for vst.  */
-       tree mem_type = build_aligned_type (TREE_TYPE((*arglist)[0]),
-                                           TYPE_ALIGN (TREE_TYPE (TREE_TYPE
((*arglist)[2]))));
+       unsigned align = TYPE_ALIGN (TREE_TYPE (TREE_TYPE ((*arglist)[2])));
+       tree mem_type = build_aligned_type (TREE_TYPE ((*arglist)[0]), align);
        return build2 (MODIFY_EXPR, mem_type,
-                      build1 (INDIRECT_REF, mem_type,
-                              fold_build_pointer_plus ((*arglist)[2],
(*arglist)[1])),
+                      build2 (MEM_REF, mem_type,
+                              fold_build_pointer_plus ((*arglist)[2],
+                                                       (*arglist)[1]),
+                              build_int_cst (TREE_TYPE ((*arglist)[2]), 0)),
                       (*arglist)[0]);
       }
     case S390_OVERLOADED_BUILTIN_s390_vec_load_pair:
might be all that is needed (if it is needed at all).
If the expectation is that one can read what has been written by those
intrinsics or write
what will be read by those intrinsics though unrelated effective pointers, then
it should alias everything,
which could be say just using ptr_type_node instead of the current types in
both of the build_int_cst (..., 0) calls above.

  parent reply	other threads:[~2024-04-11 12:35 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-04-10 10:52 [Bug tree-optimization/114676] New: " aleksei.nikiforov at linux dot ibm.com
2024-04-10 11:06 ` [Bug target/114676] " pinskia at gcc dot gnu.org
2024-04-10 11:22 ` aleksei.nikiforov at linux dot ibm.com
2024-04-10 11:25 ` pinskia at gcc dot gnu.org
2024-04-10 11:28 ` aleksei.nikiforov at linux dot ibm.com
2024-04-10 11:33 ` pinskia at gcc dot gnu.org
2024-04-10 13:40 ` jakub at gcc dot gnu.org
2024-04-10 16:47 ` rguenth at gcc dot gnu.org
2024-04-11 12:30 ` krebbel at gcc dot gnu.org
2024-04-11 12:35 ` jakub at gcc dot gnu.org [this message]
2024-04-11 12:36 ` jakub at gcc dot gnu.org
2024-04-11 15:48 ` krebbel at gcc dot gnu.org
2024-04-12 13:32 ` law at gcc dot gnu.org
2024-04-12 18:36 ` sjames at gcc dot gnu.org
2024-04-17 17:01 ` krebbel at gcc dot gnu.org
2024-04-17 17:34 ` jakub at gcc dot gnu.org
2024-04-18 14:29 ` aleksei.nikiforov at linux dot ibm.com
2024-04-22  9:37 ` krebbel at gcc dot gnu.org
2024-04-23  8:17 ` cvs-commit at gcc dot gnu.org

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bug-114676-4-BNHyD4W2Dl@http.gcc.gnu.org/bugzilla/ \
    --to=gcc-bugzilla@gcc.gnu.org \
    --cc=gcc-bugs@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).