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: new port: vn8


Hello Mike Stump,

Thanks for your comment.
I'll remove all of redundant code for this target.

I wonder that what paper is?
Is it an introduction about new feature in our target?

On Wed, Jan 27, 2016 at 5:53 AM, Mike Stump <mikestump@comcast.net> wrote:
> I donât see the point of libgcc/config/vn8/test.patch.  I suspect either you need to apply it to your diffs before you send them out, or remove the file a junk from the diff set.
>
> I saw:
>
>   inflating: libgcc/config/vn8/lib1funcs-fixed.S
>   inflating: libgcc/config/vn8/lib1funcs-fixed.S.old
>   inflating: libgcc/config/vn8/lib1funcs.S
>   inflating: libgcc/config/vn8/lib1funcs.S.old
>
> and suspect the .old files are old, and arenât used by the build.  If so, they can be removed.
>
> I saw:
>
>   inflating: gcc/config/vn8/vn8.md.bak
>   inflating: gcc/config/vn8/vn8.md.bk
>   inflating: gcc/config/vn8/t-vn8.bak
>
> and suspect these arenât needed (or are badly named) and should be removed.
>
> I glanced at the non-port files, and all that code looks fine.
>
>> I used gcc-4.9.0 version for new port.
>
> So, you will want to check out a copy of the compiler from git or svn.  You will want to do this checkout from the trunk.  You then want to apply your patch set to it, and build and test it, fixing any issues that arise, and then contribute that.  The patch set you sent would be a back port to a release branch, if any, and would come after the port is accepted into trunk.
>
>> //          output_addr_const (stderr, addr);
>> //          fprintf(stderr,"\nâ);
>> //          rtx temp_rtx = XEXP(XEXP(x,0),0);
>
> It is now time to figure out what to do with this code.  If you like it and want to keep it, instead do:
>
> #define PORTDEBUGGING 0
>
>   if (PORTDEBUGGING)
>     {
>     }
>
> and turn it into normal code.
>
>> #if 0
>> ADJUST_BYTESIZE (QI, 1);
>> ADJUST_BYTESIZE (HI, 1);
>> ADJUST_BYTESIZE (SI, 2);
>> ADJUST_BYTESIZE (PSI, 3);
>> ADJUST_BYTESIZE (DI, 4);
>> ADJUST_BYTESIZE (TI, 8);
>> #endif // 1
>
> You should just remove the whole block.
>
>
> Iâd recommend fixing all the above and reposting.  Someone else will do a more in depth port review and comment further.
>
> If you donât have your paper work done, you will want to start up that process now.  The port canât go into the compiler without it.
>
>
>
> One last comment, while you wait for the review of the port code, you should consider your FIXMEs and decide if there is any progress you can make on them.
>
> ./gcc/config/vn8/vn8.c:  /* FIXME: For setjmp and in vn8_builtin_setjmp_frame_value we don't know
> ./gcc/config/vn8/vn8.c:  /* FIXME:  Non-generic addresses are not mode-dependent in themselves.
> ./gcc/config/vn8/vn8.c:      /* FIXME: We ship info on failing tail-call in struct machine_function.
> ./gcc/config/vn8/vn8.c:      /* FIXME: For OS_task and OS_main, this might be over-conservative.  */
> ./gcc/config/vn8/vn8.c:  /* FIXME: Ideally, the following test is not needed.
> ./gcc/config/vn8/vn8.c:  /* FIXME: This hook gets called with MODE:REGNO combinations that don't
> ./gcc/config/vn8/vn8.c:     FIXME: Filter out cases where the target object is known to
> ./gcc/config/vn8/vn8.c:  /* FIXME: Register allocator might come up with spill fails if it is left
> ./gcc/config/vn8/vn8.c:  /* FIXME: Register allocator does a bad job and might spill address



-- 
Thanks & Best regards
Nguyán Sinh Ngác
Software Department
IC Design & Research Education Center
Email: ngoc.nguyensinh@icdrec.edu.vn
Cell Phone: +84 (0)974117946


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