public inbox for gcc-help@gcc.gnu.org
 help / color / mirror / Atom feed
From: Uros Bizjak <ubizjak@gmail.com>
To: gcc-help@gcc.gnu.org
Cc: Reid M <reid3759@gmail.com>
Subject: Re: build libgcc for i486 with msoft-float failed
Date: Tue, 05 Nov 2013 18:41:00 -0000	[thread overview]
Message-ID: <CAFULd4Z9VgBUQBVGXDSS0iKqFMVD2Zq-LBr98iV14R2mtmkPyQ@mail.gmail.com> (raw)

Hello!

> I was trying to build cross gcc and libgcc based on gcc-4.8 for target
> i486 without floating point support.
>
> I build libgcc with option "-D_SOFT_FLOAT -march=i486 -msoft-float
> -mno-sse -Wa,-march=i486" but I met following errors:
>
> 1.  compilation error in file "crtprec.c" : used inline assembly which
> contains float-point instruction
> ../../../gcc-4.8/libgcc/config/i386/crtprec.c: Assembler messages:
> ../../../gcc-4.8/libgcc/config/i386/crtprec.c:40: Error: `fstcw' is
> not supported on `i486'
> ../../../gcc-4.8/libgcc/config/i386/crtprec.c:45: Error: `fldcw' is
> not supported on `i486'
>
> 2. After I comment out the inline assembly to avoid error 1, I could
> get a libgcc for i486 version. There's some undefined soft-float
> function as the following. The source code of corresponding function
> are existed in libgcc but they aren't included in libgcc.a.
> out/target/product/obj/STATIC_LIBRARIES/libc_common_intermediates/libc_common.a(time.o):
> In function `clock_now':
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__floatsidf'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__floatsidf'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__muldf3'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__adddf3'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__fixdfsi'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__floatsidf'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__floatsidf'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__muldf3'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__adddf3'
> /home/reid/android/android-4.2.1_r1/bionic/libc/unistd/time.c:53:
> undefined reference to `__fixdfsi

> My gcc version is 4.8 and configuration is
> ============================================
> ./gcc-4.8/configure --target=i486-linux-android --disable-multilib
> --disable-nls --disable-shared --disable-threads --enable-visibility
> --disable-decimal-float --disable-bootstrap  --enable-languages=c,c++
> ============================================
>
> Could anybody shed some light on how to build libgcc for target i486
> without floating point support?  Thanks a lot.

The soft-float support has just been committed to mainline SVN (to
become gcc-4.9).

1. gcc for x86 will automatically define _SOFT_FLOAT when -msoft-float
(aka -mno-80387) option is specified, so all you need is to build
libgcc with options "-march=i486 -Wa,-march=i486 -msoft-float". The
library has got #ifndef _SOFT_FLOAT in correct places. I also don't
think you need -mno-sse, 32bit i486 has no SSE enabled by default.

2. The gcc library is missing compiled-in float and double soft-fp
support. You will need following patch:

--cut here--
Index: libgcc/config.host
===================================================================
--- libgcc/config.host  (revision 204405)
+++ libgcc/config.host  (working copy)
@@ -1228,7 +1228,7 @@
   i[34567]86-*-openbsd* | x86_64-*-openbsd*)
        tmake_file="${tmake_file} t-softfp-tf"
        if test "${host_address}" = 32; then
-               tmake_file="${tmake_file} i386/${host_address}/t-softfp"
+               tmake_file="${tmake_file} t-softfp-sfdf
i386/${host_address}/t-softfp"
        fi
        tmake_file="${tmake_file} i386/t-softfp t-softfp"
        ;;
--cut here--

Android is by default compiled with 64bit long double, so this should be enough.

Uros.

             reply	other threads:[~2013-11-05 18:41 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-11-05 18:41 Uros Bizjak [this message]
  -- strict thread matches above, loose matches on Subject: below --
2013-10-06  6:02 Reid M
2013-10-06 17:33 ` Kai Ruottu
2013-10-07 15:54 ` Ian Lance Taylor
2013-10-07 18:43   ` Reid M

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=CAFULd4Z9VgBUQBVGXDSS0iKqFMVD2Zq-LBr98iV14R2mtmkPyQ@mail.gmail.com \
    --to=ubizjak@gmail.com \
    --cc=gcc-help@gcc.gnu.org \
    --cc=reid3759@gmail.com \
    /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 read-only IMAP folder(s) and NNTP newsgroup(s).