GCC generating wrong assembly code?

GCC generating wrong assembly code?

[Dennis Newbold]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1611 bytes --]

Dear Linux-mips list members:

     I'm trying to build gcj (GNU Java ahead-of-time compiler) from
the sources.  It ran for quite awhile, and then on a particular file,
it got about 20 "Error: branch out of range" errors from the gas
assembler.  I'm hoping that someone on this list that understands gcc
and mips assembler better than I can suggest a command-line switch or
some other way to get gcc to generate code which does not result in
this error.

Details:  source file, and assembler output file are in the attached zip
file.  Command line used to invoke gcc is:

gcc -c -g -DIN_GCC -W -Wall -Wwrite-strings -Wstrict-prototypes
-Wmissing-prototypes -Wtraditional -fno-common -DHAVE_CONFIG_H -I.
-I../../gcc/gcc -I../../gcc/gcc/config -I../../gcc/include
../../gcc/gcc/expr.c

Assembly output file was produced using the above command line, replacing
"-c" with "-S".

I'm building it with  gcc 2.95.3 and GNU make vers. 3.75 on a Linux-mips
system. The configure script says that the cpu type is
mipsel-pc-linux-gnu.  In any case, the processor is the QED (now
PMC-Sierra) RM5230.

Also, if anyone has successfully built gcj for a MIPS system, and would
mind sending me a copy of the binary executable(s), that would work too.

Please send any replies directly to me (dennisn@pe.net), as I'm not
subscribed to the mailing list.  Thanks alot.

Dennis Newbold
dennisn@pe.net

---------------------------------------------
|                                           |
|     The way to be happy is to be good     |
|                                           |
---------------------------------------------

[-- Attachment #2: Type: APPLICATION/ZIP, Size: 314171 bytes --]

Re: GCC generating wrong assembly code?

[H. J. Lu]

On Fri, Oct 25, 2002 at 11:30:36AM -0700, Dennis Newbold wrote:
> Dear Linux-mips list members:
> 
>      I'm trying to build gcj (GNU Java ahead-of-time compiler) from
> the sources.  It ran for quite awhile, and then on a particular file,
> it got about 20 "Error: branch out of range" errors from the gas
> assembler.  I'm hoping that someone on this list that understands gcc
> and mips assembler better than I can suggest a command-line switch or
> some other way to get gcc to generate code which does not result in
> this error.
> 
> Details:  source file, and assembler output file are in the attached zip
> file.  Command line used to invoke gcc is:
> 
> gcc -c -g -DIN_GCC -W -Wall -Wwrite-strings -Wstrict-prototypes
> -Wmissing-prototypes -Wtraditional -fno-common -DHAVE_CONFIG_H -I.
> -I../../gcc/gcc -I../../gcc/gcc/config -I../../gcc/include
> ../../gcc/gcc/expr.c
> 
> Assembly output file was produced using the above command line, replacing
> "-c" with "-S".
> 
> I'm building it with  gcc 2.95.3 and GNU make vers. 3.75 on a Linux-mips
> system. The configure script says that the cpu type is
> mipsel-pc-linux-gnu.  In any case, the processor is the QED (now
> PMC-Sierra) RM5230.
> 
> Also, if anyone has successfully built gcj for a MIPS system, and would
> mind sending me a copy of the binary executable(s), that would work too.
> 

The gcc 3.2 mips/mipsel binary rpms for RedHat 7.3 are at

ftp://ftp.linux-mips.org/pub/linux/mips/redhat/7.3/test/

They come from my RedHat 8.0 MIPS port. The simple install instructions
are in INSTALL. I am interested in any feedbacks.

gcj is in there.


H.J.

Re: GCC generating wrong assembly code?

[Ralf Baechle]

On Fri, Oct 25, 2002 at 11:30:36AM -0700, Dennis Newbold wrote:

>      I'm trying to build gcj (GNU Java ahead-of-time compiler) from
> the sources.  It ran for quite awhile, and then on a particular file,
> it got about 20 "Error: branch out of range" errors from the gas
> assembler.  I'm hoping that someone on this list that understands gcc

This is supposedly fixed in the very latest versions.

As a temporary workaround that works for most files enable optimization
-O or even -O2 for those files affected by this problem.  -fno-inline
or -Os may also help - basically everything that reduces the code size.

  Ralf