public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "macro at orcam dot me.uk" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/112092] New: RISC-V: Wrong RVV code produced for vsetvl-11.c and vsetvlmax-8.c
Date: Thu, 26 Oct 2023 01:00:18 +0000	[thread overview]
Message-ID: <bug-112092-4@http.gcc.gnu.org/bugzilla/> (raw)

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112092

            Bug ID: 112092
           Summary: RISC-V: Wrong RVV code produced for vsetvl-11.c and
                    vsetvlmax-8.c
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: macro at orcam dot me.uk
  Target Milestone: ---
            Target: riscv*-*-*

There is incorrect code produced for
gcc/testsuite/gcc.target/riscv/rvv/vsetvl/vsetvl-11.c, where we have:

  if (cond)
    vl = __riscv_vsetvl_e32m1(avl);
  else
    vl = __riscv_vsetvl_e16mf2(avl);

however the relevant parts of the assembly produced are:

        beq     a5,zero,.L2
        vsetvli zero,a6,e32,m1,tu,ma
.L3:
[...]
        ret
.L2:
        vsetvli zero,a6,e32,m1,tu,ma
        j       .L3

so both VSETVLI instructions are identical (and the whole conditional
redundant) while the former one is AFAICT supposed to be:

        vsetvli zero,a6,e16,mf2,ta,ma

according to the intrinsic used.  Additionally the pass condition of the
test case is too relaxed, making the test pass regardless.

RTL dumps indicate correct code generation up until the "vsetvl" pass,
where:

(insn 20 19 21 4 (set (reg/v:SI 16 a6 [orig:136 vl ] [136])
        (unspec:SI [
                (reg/v:SI 16 a6 [orig:147 avl ] [147])
                (const_int 16 [0x10])
                (const_int 7 [0x7])
                (const_int 2 [0x2]) repeated x2
            ] UNSPEC_VSETVL))
".../gcc/testsuite/gcc.target/riscv/rvv/vsetvl/vsetvl-11.c":12:10 1507
{vsetvlsi_no_side_effects}
     (nil))

is replaced with:

(insn 58 19 21 4 (parallel [
            (set (reg:SI 66 vl)
                (unspec:SI [
                        (reg/v:SI 16 a6 [orig:147 avl ] [147])
                        (const_int 32 [0x20])
                        (const_int 0 [0])
                    ] UNSPEC_VSETVL))
            (set (reg:SI 67 vtype)
                (unspec:SI [
                        (const_int 32 [0x20])
                        (const_int 0 [0]) repeated x2
                        (const_int 1 [0x1])
                    ] UNSPEC_VSETVL))
        ]) ".../gcc/testsuite/gcc.target/riscv/rvv/vsetvl/vsetvl-11.c":12:10
1505 {vsetvl_discard_resultsi}
     (nil))

Similarly with vsetvlmax-8.c.

             reply	other threads:[~2023-10-26  1:00 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-10-26  1:00 macro at orcam dot me.uk [this message]
2023-10-26  1:46 ` [Bug target/112092] " juzhe.zhong at rivai dot ai
2023-10-26  1:57 ` juzhe.zhong at rivai dot ai
2023-10-26  4:01 ` macro at orcam dot me.uk
2023-10-26  6:38 ` kito at gcc dot gnu.org
2023-10-26  6:51 ` juzhe.zhong at rivai dot ai
2023-10-26  7:08 ` juzhe.zhong at rivai dot ai
2023-10-26 23:31 ` macro at orcam dot me.uk
2023-10-27  0:57 ` juzhe.zhong at rivai dot ai
2023-10-27  1:03 ` juzhe.zhong at rivai dot ai
2023-10-31 13:58 ` [Bug target/112092] RISC-V: Suboptimal " macro at orcam dot me.uk
2023-11-08  6:38 ` cvs-commit at gcc dot gnu.org
2023-11-08  6:39 ` juzhe.zhong at rivai dot ai

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-112092-4@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).