From: Uros Bizjak <ubizjak@gmail.com>
To: Jakub Jelinek <jakub@redhat.com>
Cc: "gcc-patches@gcc.gnu.org" <gcc-patches@gcc.gnu.org>,
Richard Biener <richard.guenther@gmail.com>
Subject: Re: [PATCH] i386: Fix -mavx -mno-mavx2 ICE with VEC_COND_EXPR [PR93637]
Date: Mon, 10 Feb 2020 20:16:00 -0000 [thread overview]
Message-ID: <CAFULd4YMFow6NvpdMKqsDEBoNrjvdGBayz+r_oT5FhJ7=3pgkw@mail.gmail.com> (raw)
In-Reply-To: <20200210143247.GQ17695@tucnak>
On Mon, Feb 10, 2020 at 3:33 PM Jakub Jelinek <jakub@redhat.com> wrote:
>
> Hi!
>
> As mentioned in the PR, for -mavx -mno-avx2 the backend does support
> vcondv4div4df and vcondv8siv8sf optabs (while generally 32-byte vectors
> aren't much supported in that case, it is performed using
> vandps/vandnps/vorps). The problem is that after the last generic vector
> lowering (where the VEC_COND_EXPR still compares two V4DF vectors and
> has two V4DI last operands and V4DI result and so is considered ok) fre4
> folds the condition into constant, at which point the middle-end during
> expansion will try vcond_mask_optab and fall back to trying to expand it
> as the constant vector < 0 vcondv4div4di, but neither of them is supported
> for -mavx -mno-avx2 and thus we ICE.
>
> So, the options I see is either what the following patch does, also support
> vcond_mask_v4div4di and vcond_mask_v4siv4si already for TARGET_AVX, or
> require for vcondv4div4df and vcondv8siv8sf TARGET_AVX2 rather than current
> TARGET_AVX.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
> Or do you prefer the disabling of the vcond patterns instead?
>
> 2020-02-10 Jakub Jelinek <jakub@redhat.com>
>
> PR target/93637
> * config/i386/sse.md (VI_256_AVX2): New mode iterator.
> (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
> Change condition from TARGET_AVX2 to TARGET_AVX.
>
> * gcc.target/i386/avx-pr93637.c: New test.
LGTM as an ICE fix for stage4. We can try Richard's suggestion "later".
Thanks,
Uros.
>
> --- gcc/config/i386/sse.md.jj 2020-02-10 13:14:02.970131692 +0100
> +++ gcc/config/i386/sse.md 2020-02-10 13:15:54.343473253 +0100
> @@ -3430,13 +3430,19 @@ (define_expand "vcond_mask_<mode><avx512
> (match_operand:<avx512fmaskmode> 3 "register_operand")))]
> "TARGET_AVX512BW")
>
> +;; As vcondv4div4df and vcondv8siv8sf are enabled already with TARGET_AVX,
> +;; and their condition can be folded late into a constant, we need to
> +;; support vcond_mask_v4div4di and vcond_mask_v8siv8si for TARGET_AVX.
> +(define_mode_iterator VI_256_AVX2 [(V32QI "TARGET_AVX2") (V16HI "TARGET_AVX2")
> + V8SI V4DI])
> +
> (define_expand "vcond_mask_<mode><sseintvecmodelower>"
> - [(set (match_operand:VI_256 0 "register_operand")
> - (vec_merge:VI_256
> - (match_operand:VI_256 1 "nonimmediate_operand")
> - (match_operand:VI_256 2 "nonimm_or_0_operand")
> + [(set (match_operand:VI_256_AVX2 0 "register_operand")
> + (vec_merge:VI_256_AVX2
> + (match_operand:VI_256_AVX2 1 "nonimmediate_operand")
> + (match_operand:VI_256_AVX2 2 "nonimm_or_0_operand")
> (match_operand:<sseintvecmode> 3 "register_operand")))]
> - "TARGET_AVX2"
> + "TARGET_AVX"
> {
> ix86_expand_sse_movcc (operands[0], operands[3],
> operands[1], operands[2]);
> --- gcc/testsuite/gcc.target/i386/avx-pr93637.c.jj 2020-02-10 13:19:18.212437488 +0100
> +++ gcc/testsuite/gcc.target/i386/avx-pr93637.c 2020-02-10 13:18:25.651220171 +0100
> @@ -0,0 +1,17 @@
> +/* PR target/93637 */
> +/* { dg-do compile } */
> +/* { dg-options "-mavx -mno-avx2 -O3 --param sccvn-max-alias-queries-per-access=3" } */
> +
> +double
> +foo (void)
> +{
> + int i;
> + double r = 7.0;
> + double a[] = { 0.0, 0.0, -0.0, 0.0, 0.0, -0.0, 1.0, 0.0, 0.0, -0.0, 1.0, 0.0, 1.0, 1.0 };
> +
> + for (i = 0; i < sizeof (a) / sizeof (a[0]); ++i)
> + if (a[i] == 0.0)
> + r = a[i];
> +
> + return r;
> +}
>
> Jakub
>
prev parent reply other threads:[~2020-02-10 20:16 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-02-10 14:33 Jakub Jelinek
2020-02-10 14:40 ` Richard Biener
2020-02-10 14:47 ` Jakub Jelinek
2020-02-10 20:16 ` Uros Bizjak [this message]
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='CAFULd4YMFow6NvpdMKqsDEBoNrjvdGBayz+r_oT5FhJ7=3pgkw@mail.gmail.com' \
--to=ubizjak@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jakub@redhat.com \
--cc=richard.guenther@gmail.com \
/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).