[Bug optimization/14110] New: altivec code misoptimized with any -O

[Bug optimization/14110] New: altivec code misoptimized with any -O

[lu_zero at gentoo dot org]

that problem happens while building libavicodec's idct_altivec.c.
I restricted the issue inq the idct_add_altivec() function.

the code is usually built with 
"-O4 -mcpu=7450 -mtune=7450 -maltivec -mabi=altivec -pipe -ffast-math
-fomit-frame-pointer" and with gcc-3.2.3 produces correct code.

if I change it removing -O4 and not putting any other -O (tested -O2 and -O1)
the result is a fully working code.

-- 
           Summary: altivec code misoptimized with any -O
           Product: gcc
           Version: 3.3.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: lu_zero at gentoo dot org
                CC: gcc-bugs at gcc dot gnu dot org
  GCC host triplet: powerpc-unknown-linux-gnu
GCC target triplet: powerpc-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug optimization/14110] altivec code misoptimized with any -O

[lu_zero at gentoo dot org]

------- Additional Comments From lu_zero at gentoo dot org  2004-02-11 14:25 -------
Created an attachment (id=5724)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=5724&action=view)
Preprocessed source


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug optimization/14110] altivec code misoptimized with any -O

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-02-11 16:46 -------
Can you provide the non-preprocessed source as looking to the preprocessed source for altivec is a 
pain?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug optimization/14110] altivec code misoptimized with any -O

[lu_zero at gentoo dot org]

------- Additional Comments From lu_zero at gentoo dot org  2004-02-11 17:15 -------
Created an attachment (id=5725)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=5725&action=view)
source


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug optimization/14110] altivec code misoptimized with any -O

[khalek at linuxgamers dot net]

------- Additional Comments From khalek at linuxgamers dot net  2004-03-15 12:10 -------
I have something similiar when attempting to build mpegvideo_altivec.c from ffmpeg. 

The following command always gives an ICE:
gcc -Wall  -maltivec -mabi=altivec -g -O3 -DHAVE_AV_CONFIG_H -I..
-D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE -c -o
ppc/mpegvideo_altivec.o ppc/mpegvideo_altivec.c

Only when dropping -O entirely does it compile.

Output when adding -v and -save-tempts is as follows:

Reading specs from /usr/lib/gcc-lib/powerpc-linux/3.3.3/specs
Configured with: ../src/configure -v
--enable-languages=c,c++,java,f77,pascal,objc,ada --prefix=/usr
--mandir=/usr/share/man --infodir=/usr/share/info
--with-gxx-include-dir=/usr/include/c++/3.3 --enable-shared --with-system-zlib
--enable-nls --without-included-gettext --enable-__cxa_atexit
--enable-clocale=gnu --enable-java-gc=boehm --enable-java-awt=xlib
--enable-objc-gc --disable-multilib powerpc-linux
Thread model: posix
gcc version 3.3.3 (Debian 20040306)
 /usr/lib/gcc-lib/powerpc-linux/3.3.3/cc1 -E -quiet -v -I.. -D__GNUC__=3
-D__GNUC_MINOR__=3 -D__GNUC_PATCHLEVEL__=3 -D__unix__ -D__gnu_linux__
-D__linux__ -Dunix -D__unix -Dlinux -D__linux -Asystem=unix -Asystem=posix
-DHAVE_AV_CONFIG_H -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_GNU_SOURCE
ppc/mpegvideo_altivec.c -maltivec -mabi=altivec -Wall -O3 mpegvideo_altivec.i
ignoring nonexistent directory "/usr/powerpc-linux/include"
#include "..." search starts here:
#include <...> search starts here:
 ..
 /usr/local/include
 /usr/lib/gcc-lib/powerpc-linux/3.3.3/include
 /usr/include
End of search list.
 /usr/lib/gcc-lib/powerpc-linux/3.3.3/cc1 -fpreprocessed mpegvideo_altivec.i
-quiet -dumpbase mpegvideo_altivec.c -maltivec -mabi=altivec -auxbase-strip
ppc/mpegvideo_altivec.o -g -O3 -Wall -version -o mpegvideo_altivec.s
GNU C version 3.3.3 (Debian 20040306) (powerpc-linux)
        compiled by GNU C version 3.3.3 (Debian 20040306).
GGC heuristics: --param ggc-min-expand=72 --param ggc-min-heapsize=80312
ppc/mpegvideo_altivec.c: In function `dct_quantize_altivec':
ppc/mpegvideo_altivec.c:516: internal compiler error: in
add_location_or_const_value_attribute, at dwarf2out.c:9496
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug optimization/14110] altivec code misoptimized with any -O

[khalek at linuxgamers dot net]

------- Additional Comments From khalek at linuxgamers dot net  2004-03-15 12:14 -------
Created an attachment (id=5919)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=5919&action=view)
Preprocessed source


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug optimization/14110] altivec code misoptimized with any -O

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-03-15 14:19 -------
Jonathan Gray,
Your bug is that 3.3.3 did not have a patch to fix that ICE, it has been fixed already in 3.3.4.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |khalek at linuxgamers dot
                   |                            |net


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug rtl-optimization/14110] altivec code misoptimized with any -O

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2004-10-11 10:45 -------
Lu Zero, I'm revisiting this bug report. The testcase you provided cannot be 
executed to check for the miscompilation, which makes it very difficult for us 
to check for the problem.

Would you please provide a testcase with a main() invoking idc_add_altivec() 
with some valid parameters, and aborting if the result is incorrect (assuming 
it does not crash right way because of the miscompilation)?

Thanks

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |giovannibajo at libero dot
                   |                            |it
             Status|UNCONFIRMED                 |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug rtl-optimization/14110] altivec code misoptimized with any -O

[lu_zero at gentoo dot org]

------- Additional Comments From lu_zero at gentoo dot org  2004-10-11 17:16 -------
Fixed at least on gcc-3.4.x and in the hammer branch.
gcc-3.3.4 should have other problems (I didn't check yet)

the best way to check about the problem works is just use mplayer or ffplay...

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110

[Bug rtl-optimization/14110] altivec code misoptimized with any -O

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-11 17:20 -------
Fixed so closing.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |3.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14110