public inbox for linuxppc-dev@ozlabs.org 
 help / color / mirror / Atom feed
From: Gabriel Paubert <paubert@iram•es>
To: Stephen N Chivers <schivers@csc•com.au>
Cc: Chris Proctor <cproctor@csc•com.au>, linuxppc-dev@lists•ozlabs.org
Subject: Re: arch/powerpc/math-emu/mtfsf.c - incorrect mask?
Date: Thu, 6 Feb 2014 09:26:37 +0100	[thread overview]
Message-ID: <20140206082635.GA7048@visitor2.iram.es> (raw)
In-Reply-To: <OF784BA598.ACFE9DA4-ONCA257C76.00827472-CA257C77.000BCFAD@csc.com>

On Thu, Feb 06, 2014 at 12:09:00PM +1000, Stephen N Chivers wrote:
> I have a MPC8548e based board and an application that makes
> extensive use of floating point including numerous calls to cos.
> In the same program there is the use of an sqlite database.
> 
> The kernel is derived from 2.6.31 and is compiled with math emulation.
> 
> At some point after the reading of the SQLITE database, the
> return result from cos goes from an in range value to an out
> of range value.
> 
> This is as a result of the FP rounding mode mutating from "round to 
> nearest"
> to "round toward zero".
> 
> The cos in the glibc version being used is known to be sensitive to 
> rounding
> direction and Joseph Myers has previously fixed glibc.
> 
> The failure does not occur on a machine that has a hardware floating
> point unit (a MPC7410 processor).
> 
> I have traced the mutation to the following series of instructions:
> 
>         mffs            f0
>         mtfsb1          4*cr7+so
>         mtfsb0          4*cr7+eq
>         fadd            f13,f1,f2
>         mtfsf           1, f0
> 
> The instructions are part of the stream emitted by gcc for the conversion
> of a 128 bit floating point value into an integer in the sqlite database 
> read.
> 
> Immediately before the execution of the mffs instruction the "rounding
> mode" is "round to nearest".
> 
> On the MPC8548 board, the execution of the mtfsf instruction does not
> restore the rounding mode to "round to nearest".
> 
> I believe that the mask computation in mtfsf.c is incorrect and is 
> reversed.
> 
> In the latest version of the file (linux-3.14-rc1), the mask is computed 
> by:
> 
>                  mask = 0;
>                  if (FM & (1 << 0))
>                         mask |= 0x90000000;
>                  if (FM & (1 << 1))
>                         mask |= 0x0f000000;
>                  if (FM & (1 << 2))
>                         mask |= 0x00f00000;
>                  if (FM & (1 << 3))
>                         mask |= 0x000f0000;
>                  if (FM & (1 << 4))
>                         mask |= 0x0000f000;
>                  if (FM & (1 << 5))
>                         mask |= 0x00000f00;
>                  if (FM & (1 << 6))
>                         mask |= 0x000000f0;
>                  if (FM & (1 << 7))
>                         mask |= 0x0000000f;
> 
> I think it should be:
> 
>                 mask = 0;
>                 if (FM & (1 << 0))
>                         mask |= 0x0000000f;
>                 if (FM & (1 << 1))
>                         mask |= 0x000000f0;
>                 if (FM & (1 << 2))
>                         mask |= 0x00000f00;
>                 if (FM & (1 << 3))
>                         mask |= 0x0000f000;
>                 if (FM & (1 << 4))
>                         mask |= 0x000f0000;
>                 if (FM & (1 << 5))
>                         mask |= 0x00f00000;
>                 if (FM & (1 << 6))
>                         mask |= 0x0f000000;
>                 if (FM & (1 << 7))
>                         mask |= 0x90000000;
> 
> With the above mask computation I get consistent results for both the 
> MPC8548
> and MPC7410 boards.
> 
> Am I missing something subtle?

No I think you are correct. This said, this code may probably be optimized 
to eliminate a lot of the conditional branches. I think that:

mask = (FM & 1);
mask |= (FM << 3) & 0x10;
mask |= (FM << 6) & 0x100;
mask |= (FM << 9) & 0x1000;
mask |= (FM << 12) & 0x10000;
mask |= (FM << 15) & 0x100000;
mask |= (FM << 18) & 0x1000000;
mask |= (FM << 21) & 0x10000000;
mask *= 15;

should do the job, in less code space and without a single branch.

Each one of the "mask |=" lines should be translated into an
rlwinm instruction followed by an "or". Actually it should be possible
to transform each of these lines into a single "rlwimi" instruction
but I don't know how to coerce gcc to reach this level of optimization.

Another way of optomizing this could be:

mask = (FM & 0x0f) | ((FM << 12) & 0x000f0000);
mask = (mask & 0x00030003) | ((mask << 6) & 0x03030303);
mask = (mask & 0x01010101) | ((mask << 3) & 0x10101010);
mask *= 15;

It's not easy to see which of the solutions is faster, the second one
needs to create quite a few constants, but its dependency length is 
lower. It is very likely that the first solution is faster in cache-cold
case and the second in cache-hot. 

Regardless, the original code is rather naïve, larger and slower in all cases,
with timing variation depending on branch mispredictions.

	Regards,
	Gabriel

  reply	other threads:[~2014-02-06  9:09 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-02-06  2:09 arch/powerpc/math-emu/mtfsf.c - incorrect mask? Stephen N Chivers
2014-02-06  8:26 ` Gabriel Paubert [this message]
2014-02-07  1:27   ` Stephen N Chivers
2014-02-07 10:10     ` Gabriel Paubert
2014-02-07 20:49       ` James Yang
2014-02-09 19:42         ` Stephen N Chivers
2014-02-10 16:50           ` James Yang
2014-02-10 11:03         ` Gabriel Paubert
2014-02-10 11:17           ` David Laight
2014-02-10 12:21             ` Gabriel Paubert
2014-02-10 12:32               ` David Laight
2014-02-10 13:00                 ` Gabriel Paubert
2014-02-10 17:03           ` James Yang
2014-02-11  7:26             ` Gabriel Paubert
2014-02-11 20:57               ` Linux-3.14-rc2: Order of serial node compatibles in DTS files Stephen N Chivers
2014-02-11 22:33                 ` Kumar Gala
2014-02-11 22:51                   ` Sebastian Hesselbarth
2014-02-11 23:38                     ` Stephen N Chivers
2014-02-11 23:43                       ` Sebastian Hesselbarth
2014-02-12 11:00                         ` Arnd Bergmann
2014-02-11 23:41                     ` Scott Wood
2014-02-11 23:46                       ` Sebastian Hesselbarth
2014-02-12  0:21                         ` Stephen N Chivers
2014-02-12  5:28                           ` Kevin Hao
2014-02-12  8:30                             ` Sebastian Hesselbarth
2014-02-12 10:31                               ` Kevin Hao
2014-02-12 11:26                                 ` Sebastian Hesselbarth
2014-02-12 11:32                                   ` Kevin Hao
2014-02-12  8:25                           ` Sebastian Hesselbarth
2014-02-12 10:35                             ` Kevin Hao

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=20140206082635.GA7048@visitor2.iram.es \
    --to=paubert@iram$(echo .)es \
    --cc=cproctor@csc$(echo .)com.au \
    --cc=linuxppc-dev@lists$(echo .)ozlabs.org \
    --cc=schivers@csc$(echo .)com.au \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html
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