From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([2001:4830:134:3::10]:35754)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <rth7680@gmail.com>) id 1aaolB-00067F-4G
	for qemu-devel@nongnu.org; Tue, 01 Mar 2016 13:11:02 -0500
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <rth7680@gmail.com>) id 1aaol6-0008Ld-1Z
	for qemu-devel@nongnu.org; Tue, 01 Mar 2016 13:11:01 -0500
Received: from mail-qg0-x243.google.com ([2607:f8b0:400d:c04::243]:33394)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <rth7680@gmail.com>) id 1aaol5-0008LV-TS
	for qemu-devel@nongnu.org; Tue, 01 Mar 2016 13:10:55 -0500
Received: by mail-qg0-x243.google.com with SMTP id y89so13053228qge.0
	for <qemu-devel@nongnu.org>; Tue, 01 Mar 2016 10:10:55 -0800 (PST)
Sender: Richard Henderson <rth7680@gmail.com>
References: <1456849468-30217-1-git-send-email-kbastian@mail.uni-paderborn.de>
	<1456849468-30217-4-git-send-email-kbastian@mail.uni-paderborn.de>
From: Richard Henderson <rth@twiddle.net>
Message-ID: <56D5DB2C.4010801@twiddle.net>
Date: Tue, 1 Mar 2016 10:10:52 -0800
MIME-Version: 1.0
In-Reply-To: <1456849468-30217-4-git-send-email-kbastian@mail.uni-paderborn.de>
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Subject: Re: [Qemu-devel] [PATCH 3/7] target-tricore: add add.f/sub.f
	instructions
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>, qemu-devel@nongnu.org

> +    f_set_flags(env);                                                          \

You shouldn't need to set the flags every instruction.
You ought to be able to limit the changes to reset and
stores to the PSW.

> +    arg1 = float32_squash_input_denormal(arg1, &env->fp_status);               \
> +    arg2 = float32_squash_input_denormal(arg2, &env->fp_status);               \
> +                                                                               \
> +    if (float32_is_any_nan(arg1) || float32_is_any_nan(arg2)) {                \
> +        f_result = QUIET_NAN;                                                  \
> +        if (float32_is_signaling_nan(arg1) ||                                  \
> +            float32_is_signaling_nan(arg2)) {                                  \
> +            env->fp_status.float_exception_flags |= float_flag_invalid;        \
> +        }                                                                      \
> +    } else if (f_is_pos_inf(arg1) && f_is_neg_inf(arg2)) {                     \
> +        f_result = ADD_NAN;                                                    \
> +    } else if (f_is_pos_inf(arg2) && f_is_neg_inf(arg1)) {                     \
> +        f_result = ADD_NAN;                                                    \
> +    } else {                                                                   \
> +        f_result = float32_##name(arg1, arg2 , &env->fp_status);               \
> +    }                                                                          \

If we assume that exceptional situations are, well, exceptional, then we can
re-order this to

  f_result = float32_op(arg1, arg2, &env->fp_status);
  flags = env->fp_status.float_exception_flags;
  if (flags) {
      /* If the output is a NaN, but the inputs aren't,
         we return a unique value.  */
      if ((flags & float_flag_invalid)
          && !float32_is_any_nan(arg1)
          && !float32_is_any_nan(arg2)) {
          f_result = ADD_NAN;
      }
      f_update_psw_flags(env, flags, false);
  }

This does assume that fp_status.default_nan_mode = 1, so that
float32_default_nan is returned.  Which means that first patch should touch
fpu/softfloat-specialize.h to add tricore to the list of those defaulting to
0x7fc00000.


r~