qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Aurelien Jarno <aurelien@aurel32.net>
To: Richard Henderson <rth@twiddle.net>
Cc: qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH 4/5] tcg: Optimize movcond for constant comparisons
Date: Fri, 21 Sep 2012 00:04:19 +0200	[thread overview]
Message-ID: <20120920220419.GC4457@ohm.aurel32.net> (raw)
In-Reply-To: <20120918211148.GA20933@ohm.aurel32.net>

On Tue, Sep 18, 2012 at 11:11:48PM +0200, Aurelien Jarno wrote:
> On Tue, Sep 18, 2012 at 07:23:59AM -0700, Richard Henderson wrote:
> > Signed-off-by: Richard Henderson <rth@twiddle.net>
> > ---
> >  tcg/optimize.c | 29 +++++++++++++++++++++++++++++
> >  1 file changed, 29 insertions(+)
> > 
> > diff --git a/tcg/optimize.c b/tcg/optimize.c
> > index fba0ed9..5b0a8ce 100644
> > --- a/tcg/optimize.c
> > +++ b/tcg/optimize.c
> > @@ -397,6 +397,14 @@ static TCGArg *tcg_constant_folding(TCGContext *s, uint16_t *tcg_opc_ptr,
> >                  args[3] = tcg_swap_cond(args[3]);
> >              }
> >              break;
> > +        CASE_OP_32_64(movcond):
> > +            if (temps[args[1]].state == TCG_TEMP_CONST
> > +                && temps[args[2]].state != TCG_TEMP_CONST) {
> > +                tmp = args[1];
> > +                args[1] = args[2];
> > +                args[2] = tmp;
> > +                args[5] = tcg_swap_cond(args[5]);
> > +            }
> >          default:
> >              break;
> >          }
> > @@ -617,6 +625,27 @@ static TCGArg *tcg_constant_folding(TCGContext *s, uint16_t *tcg_opc_ptr,
> >              }
> >              args += 4;
> >              break;
> > +        CASE_OP_33_64(movcond):
> > +            if (temps[args[1]].state == TCG_TEMP_CONST
> > +                && temps[args[2]].state == TCG_TEMP_CONST) {
> > +                gen_opc_buf[op_index] = op_to_mov(op);
> > +                tmp = do_constant_folding_cond(op, temps[args[1]].val,
> > +                                               temps[args[2]].val, args[5]);
> > +                tcg_opt_gen_mov(s, gen_args, args[0], args[4-tmp],
> > +                                nb_temps, nb_globals);
> > +                gen_args += 2;
> > +            } else {
> > +                reset_temp(args[0], nb_temps, nb_globals);
> > +                gen_args[0] = args[0];
> > +                gen_args[1] = args[1];
> > +                gen_args[2] = args[2];
> > +                gen_args[3] = args[3];
> > +                gen_args[4] = args[4];
> > +                gen_args[5] = args[5];
> > +                gen_args += 6;
> > +            }
> > +            args += 6;
> > +            break;
> >          case INDEX_op_call:
> >              nb_call_args = (args[0] >> 16) + (args[0] & 0xffff);
> >              if (!(args[nb_call_args + 1] & (TCG_CALL_CONST | TCG_CALL_PURE))) {
> > -- 
> > 1.7.11.4
> > 
> 
> Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

Playing a bit more with this patch, I have found it fails when v1 or v2
or constants or copies. The additional patch below fixes the problem.


diff --git a/tcg/optimize.c b/tcg/optimize.c
index f95c511..b7e235e 100644
--- a/tcg/optimize.c
+++ b/tcg/optimize.c
@@ -641,12 +641,23 @@ static TCGArg *tcg_constant_folding(TCGContext *s, uint16_t *tcg_opc_ptr,
         CASE_OP_32_64(movcond):
             if (temps[args[1]].state == TCG_TEMP_CONST
                 && temps[args[2]].state == TCG_TEMP_CONST) {
-                gen_opc_buf[op_index] = op_to_mov(op);
                 tmp = do_constant_folding_cond(op, temps[args[1]].val,
                                                temps[args[2]].val, args[5]);
-                tcg_opt_gen_mov(gen_args, args[0], args[4-tmp],
-                                nb_temps, nb_globals);
-                gen_args += 2;
+                if ((temps[args[4-tmp]].state == TCG_TEMP_COPY
+                    && temps[args[4-tmp]].val == args[0])
+                    || args[0] == args[4-tmp]) {
+                    gen_opc_buf[op_index] = INDEX_op_nop;
+                } else if (tmp && temps[args[4-tmp]].state == TCG_TEMP_CONST) {
+                    gen_opc_buf[op_index] = op_to_movi(op);
+                    tcg_opt_gen_movi(gen_args, args[0], temps[args[4-tmp]].val,
+                                     nb_temps, nb_globals);
+                    gen_args += 2;
+                } else {
+                    gen_opc_buf[op_index] = op_to_mov(op);
+                    tcg_opt_gen_mov(gen_args, args[0], args[4-tmp],
+                                    nb_temps, nb_globals);
+                    gen_args += 2;
+                }
             } else {
                 reset_temp(args[0], nb_temps, nb_globals);
                 gen_args[0] = args[0];

-- 
Aurelien Jarno                          GPG: 1024D/F1BCDB73
aurelien@aurel32.net                 http://www.aurel32.net

  reply	other threads:[~2012-09-20 22:04 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-09-18 14:23 [Qemu-devel] [PATCH 0/5] tcg: movcond Richard Henderson
2012-09-18 14:23 ` [Qemu-devel] [PATCH 1/5] tcg: Introduce movcond Richard Henderson
2012-09-18 21:11   ` Aurelien Jarno
2012-09-20 22:06   ` Aurelien Jarno
2012-09-20 22:47     ` Richard Henderson
2012-09-18 14:23 ` [Qemu-devel] [PATCH 2/5] target-alpha: Use movcond Richard Henderson
2012-09-18 14:23 ` [Qemu-devel] [PATCH 3/5] tcg-i386: Implement movcond Richard Henderson
2012-09-18 21:11   ` Aurelien Jarno
2012-09-18 14:23 ` [Qemu-devel] [PATCH 4/5] tcg: Optimize movcond for constant comparisons Richard Henderson
2012-09-18 21:11   ` Aurelien Jarno
2012-09-20 22:04     ` Aurelien Jarno [this message]
2012-09-21 14:33       ` Richard Henderson
2012-09-21 14:52         ` Aurelien Jarno
2012-09-18 14:24 ` [Qemu-devel] [PATCH 5/5] tcg: Optimize two-address commutative operations Richard Henderson
2012-09-18 21:11   ` Aurelien Jarno
2012-09-18 21:11 ` [Qemu-devel] [PATCH 0/5] tcg: movcond Aurelien Jarno
2012-09-19 13:03   ` malc
2012-09-19 14:26     ` Richard Henderson
2012-09-19 19:07   ` Blue Swirl
2012-09-19 19:14     ` 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=20120920220419.GC4457@ohm.aurel32.net \
    --to=aurelien@aurel32.net \
    --cc=qemu-devel@nongnu.org \
    --cc=rth@twiddle.net \
    /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).