Re: target register load optimizations

[Andrew Pinski <pinskia at physics dot uc dot edu>]

I think it also causes to fail on powerpc-apple-darwin6.6  
(host-specific hook):
stage1/xgcc -Bstage1/  
-B/Users/pinskia/fsf-clean-nocheck/powerpc-apple-darwin6.6/bin/ -c    
-funsigned-char -mdynamic-no-pic -O2 -g -DGATHER_STATISTICS -DIN_GCC    
-W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes  
-pedantic -Wno-long-long -Werror    -DHAVE_CONFIG_H    -I. -I.  
-I/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc  
-I/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/.  
-I/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/config  
-I/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/../include  
/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/config/darwin.c
stage1/xgcc -Bstage1/  
-B/Users/pinskia/fsf-clean-nocheck/powerpc-apple-darwin6.6/bin/ -c    
-funsigned-char -mdynamic-no-pic -O2 -g -DGATHER_STATISTICS -DIN_GCC    
-W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes  
-pedantic -Wno-long-long -Werror    -DHAVE_CONFIG_H    -I. -I.  
-I/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc  
-I/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/.  
-I/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/config  
-I/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/../include \
         
/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/config/rs6000/ 
host-darwin.c
In file included from  
/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/config/rs6000/ 
host-darwin.c:27:
/Volumes/UFS_Partition/pinskia/src/fsf/gcc-clean/src/gcc/hooks.h:26:  
warning: ISO C forbids forward references to `enum' types

        * hooks.c (hook_reg_class_void_no_regs): New function.  <--  
This part did it.

Thanks,
Andrew Pinski

On Wednesday, Jun 18, 2003, at 17:08 US/Eastern, Kazu Hirata wrote:

> Hi Joern,
>
> > > It turned out that there was a problem with running flow betweeen
> > > reload and epilogue generation.  mark_regs_live_at_end uses
> > > HAVE_epilogue && reload_completed to check if an rtl epilogue
> > > has already been generated.  I've introduced a new variable
> > > epilogue_comleted for this and patched rtl.h, flow.c, toplev.c,
> > > function.c, config/ia64/ia64.c, config/rs6000/rs6000.c,
> > > config/sh/sh.c and config/sparc/sparc.c accordingly.
> >
> > Attached is the new version of the patch.
>
> This patch seems to cause the h8300 port not to build, saying
>
> libbackend.a(bt-load.o)(.text+0x1e13): In function `migrate_btr_def':
> /home/kazu/gnu/gcc/gcc-head/gcc/gcc/bt-load.c:1209: undefined  
> reference to `insn_default_latency'
> libbackend.a(bt-load.o)(.text+0x20ba):/home/kazu/gnu/gcc/gcc-head/gcc/ 
> gcc/bt-load.c:1211: undefined reference to `result_ready_cost'
> collect2: ld returned 1 exit status
>
> Could you take a look at this?
>
> Thanks,
>
> Kazu Hirata