From: Richard Henderson <richard.henderson@linaro.org>
To: "Alex Bennée" <alex.bennee@linaro.org>
Cc: mmorrell@tachyum.com, qemu-devel@nongnu.org
Subject: Re: [PATCH 03/11] softfloat: Introduce float_flag_inorm_denormal
Date: Mon, 7 Jun 2021 13:52:52 -0700 [thread overview]
Message-ID: <2d66f57e-7ff0-4d57-22e1-d5156665e2fb@linaro.org> (raw)
In-Reply-To: <87pmwxmw0i.fsf@linaro.org>
On 6/7/21 10:19 AM, Alex Bennée wrote:
>> If you've got a better ordering of operations for this, do tell.
>
> What I really want is to know which instructions translate into the if
> (s->flush_inputs_to_zero) and verifying that is only checked once. Maybe
> I'm just suspicious of compilers ability to optimise things away...
> Dump of assembler code for function float32_mul:
> 0x0000000000895d60 <+0>: movzbl 0x1(%rdx),%eax
> 0x0000000000895d64 <+4>: test $0x10,%al
> 0x0000000000895d66 <+6>: je 0x895e30 <float32_mul+208>
s->float_exception_flags & float_flag_inexact
> 0x0000000000895d6c <+12>: cmpb $0x0,(%rdx)
> 0x0000000000895d6f <+15>: jne 0x895e30 <float32_mul+208>
s->float_rounding_mode == float_round_nearest_even
> 0x0000000000895d75 <+21>: test $0x7f800000,%edi
> 0x0000000000895d7b <+27>: jne 0x895da0 <float32_mul+64>
> 0x0000000000895d7d <+29>: test $0x7fffffff,%edi
> 0x0000000000895d83 <+35>: je 0x895da0 <float32_mul+64>
float32_is_denormal
> 0x0000000000895d85 <+37>: cmpb $0x0,0x5(%rdx)
> 0x0000000000895d89 <+41>: je 0x895e60 <float32_mul+256>
s->flush_inputs_to_zero
> 0x0000000000895d8f <+47>: or $0x40,%eax
> 0x0000000000895d92 <+50>: and $0x80000000,%edi
> 0x0000000000895d98 <+56>: mov %al,0x1(%rdx)
flush-to-zero and set iflush_denormal
> 0x0000000000895da0 <+64>: test $0x7f800000,%esi
> 0x0000000000895da6 <+70>: jne 0x895dd0 <float32_mul+112>
> 0x0000000000895da8 <+72>: test $0x7fffffff,%esi
> 0x0000000000895dae <+78>: je 0x895dd0 <float32_mul+112>
float32_is_denormal (second operand)
> 0x0000000000895db0 <+80>: cmpb $0x0,0x5(%rdx)
> 0x0000000000895db4 <+84>: movzbl 0x1(%rdx),%eax
> 0x0000000000895db8 <+88>: je 0x895e50 <float32_mul+240>
> 0x0000000000895dbe <+94>: or $0x40,%eax
> 0x0000000000895dc1 <+97>: and $0x80000000,%esi
s->flush_inputs_to_zero,
flush-to-zero,
set iflush_denormal.
...
> 0x0000000000895e50 <+240>: or $0x20,%eax
> 0x0000000000895e53 <+243>: mov %al,0x1(%rdx)
> 0x0000000000895e56 <+246>: jmpq 0x895dd0 <float32_mul+112>
set inorm_denormal (second operand)
> 0x0000000000895e60 <+256>: or $0x20,%eax
> 0x0000000000895e63 <+259>: mov %al,0x1(%rdx)
> 0x0000000000895e66 <+262>: jmpq 0x895da0 <float32_mul+64>
set inorm_denormal (first operand)
There do seem to be 3 reads/writes to exception_flags for float_raise.
r~
next prev parent reply other threads:[~2021-06-07 20:53 UTC|newest]
Thread overview: 38+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-27 4:13 [PATCH 00/11] softfloat: Improve denormal handling Richard Henderson
2021-05-27 4:13 ` [PATCH 01/11] softfloat: Rename float_flag_input_denormal to float_flag_iflush_denormal Richard Henderson
2021-06-07 15:16 ` Alex Bennée
2021-06-19 15:08 ` Philippe Mathieu-Daudé
2021-05-27 4:13 ` [PATCH 02/11] softfloat: Rename float_flag_output_denormal to float_flag_oflush_denormal Richard Henderson
2021-06-07 15:16 ` Alex Bennée
2021-06-19 15:08 ` Philippe Mathieu-Daudé
2021-05-27 4:13 ` [PATCH 03/11] softfloat: Introduce float_flag_inorm_denormal Richard Henderson
2021-05-28 17:41 ` Michael Morrell
2021-05-29 15:21 ` Richard Henderson
2021-07-14 16:44 ` Michael Morrell
2021-07-14 16:57 ` Richard Henderson
2021-07-14 17:50 ` Michael Morrell
2022-01-12 0:02 ` Michael Morrell
2021-06-07 15:35 ` Alex Bennée
2021-06-07 16:28 ` Alex Bennée
2021-06-07 16:41 ` Richard Henderson
2021-06-07 17:19 ` Alex Bennée
2021-06-07 20:52 ` Richard Henderson [this message]
2021-05-27 4:13 ` [PATCH 04/11] softfloat: Introduce float_flag_result_denormal Richard Henderson
2021-06-07 16:30 ` Alex Bennée
2021-06-19 15:10 ` Philippe Mathieu-Daudé
2021-05-27 4:13 ` [PATCH 05/11] target/i386: Use float_flag_inorm_denormal Richard Henderson
2021-06-19 18:41 ` Richard Henderson
2021-05-27 4:14 ` [PATCH 06/11] target/rx: Handle the FPSW.DN bit in helper_set_fpsw Richard Henderson
2021-05-28 15:34 ` Yoshinori Sato
2021-05-27 4:14 ` [PATCH 07/11] target/rx: Use FloatRoundMode " Richard Henderson
2021-05-28 15:35 ` Yoshinori Sato
2021-06-01 3:27 ` Philippe Mathieu-Daudé
2021-05-27 4:14 ` [PATCH 08/11] target/rx: Fix setting of FPSW.CE Richard Henderson
2021-05-28 15:35 ` Yoshinori Sato
2021-05-27 4:14 ` [PATCH 09/11] target/mips: Drop inline markers from msa_helper.c Richard Henderson
2021-06-01 3:27 ` Philippe Mathieu-Daudé
2021-05-27 4:14 ` [PATCH 10/11] target/mips: Do not check MSACSR_FS_MASK in update_msacsr Richard Henderson
2021-06-19 15:15 ` Philippe Mathieu-Daudé
2021-05-27 4:14 ` [PATCH 11/11] target/mips: Drop denormal operand to update_msacsr Richard Henderson
2021-06-01 3:29 ` Philippe Mathieu-Daudé
2021-06-07 16:31 ` [PATCH 00/11] softfloat: Improve denormal handling Alex Bennée
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=2d66f57e-7ff0-4d57-22e1-d5156665e2fb@linaro.org \
--to=richard.henderson@linaro.org \
--cc=alex.bennee@linaro.org \
--cc=mmorrell@tachyum.com \
--cc=qemu-devel@nongnu.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).