This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RFC: untangle arm_address_cost


> Richard Earnshaw <rearnsha@arm.com> writes:
> 
> >> This completely untested patch replaces two messy macros in arm.c 
> >> with inline functions.  It's a fair bit more legible this way.
> >> I'm floating this exclusively for comments - I know people have
> >> plans in this area and I'm not sure how this interacts.
> ...
> > This is definitely a worth-while clean-up.  I can't see any reason why it 
> > shouldn't go on the trunk as well.
> 
> Testing revealed what should have been obvious in the first place,
> namely that you can't just blindly call XEXP (x, 0/1) on something
> that may not have subexpressions... like say a REG... so here's a
> revised patch.  I've tested this version against a unified tree
> arm-elf simulator:
> 
>                 === gcc Summary ===
> # of expected passes            23540
> # of unexpected failures        25
> # of unexpected successes       10
> # of expected failures          74
> # of unresolved testcases       5
> # of untested testcases         7
> # of unsupported tests          298
> 

The Changes to Thumb are fairly trivial, but you should also run a thumb 
testsuite run if you are changing thumb code generation.  Use a site.exp 
file containing something like

case "$target_triplet" in {
    { "arm*-*-elf*" } {
      set target_list { "arm-sim/-mthumb" }
      set board_info(arm-sim,gdb,skip_huge_test) 1
      set board_info(arm-sim,sim,options) "-m 4194304"
      set board_info(arm-sim/-mthumb,sim,options) "-m 4194304"
    }
}

R.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]