qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Philippe Mathieu-Daudé" <philmd@linaro.org>
To: Richard Henderson <richard.henderson@linaro.org>, qemu-devel@nongnu.org
Cc: pbonzini@redhat.com
Subject: Re: [PATCH 4/4] NOTFORMERGE tcg/i386: Assert sub of immediate has been folded
Date: Mon, 6 Nov 2023 15:49:29 +0100	[thread overview]
Message-ID: <ffe5e4b3-4c9a-9cee-f997-f53be1656b21@linaro.org> (raw)
In-Reply-To: <20231026013945.1152174-5-richard.henderson@linaro.org>

On 26/10/23 03:39, Richard Henderson wrote:
> A release build should simply accept and emit the subtract.
> I'm not even sure if this is reasonable to keep for debug.
> 
> Not-Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   tcg/tcg.c                 | 49 ++++++++++++++++++++++++++-------------
>   tcg/i386/tcg-target.c.inc | 13 ++++++++---
>   2 files changed, 43 insertions(+), 19 deletions(-)
> 
> diff --git a/tcg/tcg.c b/tcg/tcg.c
> index a507c111cf..408647af7e 100644
> --- a/tcg/tcg.c
> +++ b/tcg/tcg.c
> @@ -3618,23 +3618,40 @@ liveness_pass_1(TCGContext *s)
>           do_addsub2:
>               nb_iargs = 4;
>               nb_oargs = 2;
> -            /* Test if the high part of the operation is dead, but not
> -               the low part.  The result can be optimized to a simple
> -               add or sub.  This happens often for x86_64 guest when the
> -               cpu mode is set to 32 bit.  */
> -            if (arg_temp(op->args[1])->state == TS_DEAD) {
> -                if (arg_temp(op->args[0])->state == TS_DEAD) {
> -                    goto do_remove;
> -                }
> -                /* Replace the opcode and adjust the args in place,
> -                   leaving 3 unused args at the end.  */
> -                op->opc = opc = opc_new;
> -                op->args[1] = op->args[2];
> -                op->args[2] = op->args[4];
> -                /* Fall through and mark the single-word operation live.  */
> -                nb_iargs = 2;
> -                nb_oargs = 1;
> +            /*
> +             * Test if the high part of the operation is dead, but the low
> +             * part is still live.  The result can be optimized to a simple
> +             * add or sub.
> +             */
> +            if (arg_temp(op->args[1])->state != TS_DEAD) {
> +                goto do_not_remove;
>               }
> +            if (arg_temp(op->args[0])->state == TS_DEAD) {
> +                goto do_remove;
> +            }
> +            /*
> +             * Replace the opcode and adjust the args in place, leaving 3
> +             * unused args at the end.  Canonicalize subi to andi.

Typo s/andi/addi/.

> +             */
> +            op->args[1] = op->args[2];
> +            {
> +                TCGTemp *src2 = arg_temp(op->args[4]);
> +                if (src2->kind == TEMP_CONST) {
> +                    if (opc_new == INDEX_op_sub_i32) {
> +                        src2 = tcg_constant_internal(TCG_TYPE_I32,
> +                                                     (int32_t)-src2->val);
> +                        opc_new = INDEX_op_add_i32;
> +                    } else if (opc_new == INDEX_op_sub_i64) {
> +                        src2 = tcg_constant_internal(TCG_TYPE_I64, -src2->val);
> +                        opc_new = INDEX_op_add_i64;
> +                    }
> +                }
> +                op->args[2] = temp_arg(src2);
> +            }
> +            op->opc = opc = opc_new;
> +            /* Mark the single-word operation live.  */
> +            nb_iargs = 2;
> +            nb_oargs = 1;
>               goto do_not_remove;

For tcg/tcg.c with your S-o-b and a reworded description:

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>



  parent reply	other threads:[~2023-11-06 14:50 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-10-26  1:39 [PATCH 0/4] tcg: Canonicalize SUBI to ANDI Richard Henderson
2023-10-26  1:39 ` [PATCH 1/4] tcg: Canonicalize subi to addi during opcode generation Richard Henderson
2023-11-06 14:37   ` Philippe Mathieu-Daudé
2023-10-26  1:39 ` [PATCH 2/4] tcg/optimize: Canonicalize subi to addi during optimization Richard Henderson
2023-11-06 14:42   ` Philippe Mathieu-Daudé
2023-10-26  1:39 ` [PATCH 3/4] tcg/optimize: Canonicalize sub2 with constants to add2 Richard Henderson
2023-11-06 14:43   ` Philippe Mathieu-Daudé
2023-10-26  1:39 ` [PATCH 4/4] NOTFORMERGE tcg/i386: Assert sub of immediate has been folded Richard Henderson
2023-10-26  1:53   ` Richard Henderson
2023-11-06 14:49   ` Philippe Mathieu-Daudé [this message]
2023-10-31 22:20 ` [PATCH 0/4] tcg: Canonicalize SUBI to ANDI Richard Henderson

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=ffe5e4b3-4c9a-9cee-f997-f53be1656b21@linaro.org \
    --to=philmd@linaro.org \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=richard.henderson@linaro.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 NNTP newsgroup(s).