[llvm] r208934 - Implement global merge optimization for global variables.

Rafael EspĂ­ndola rafael.espindola at gmail.com
Mon Jun 2 20:48:18 PDT 2014


> Thanks. I will post an update version of your patch as soon as that one is in.

And that took a long time, sorry about that :-(

Attached is my attempt to forward port the patch. I didn't change the
profitability check or the use of getGlobalMergeAlignment.

One issue is that the ARM64 -> AArch64 transition happened and it
seems that for one test we are not produce as good case as before. At
the IR level the constant is merged, be we still compute the address
twice, not sure why,  sorry.

Would you mind taking a look and posting this again? Do CC myself and
Nick. If I remember correctly the only remaining concern was why
getGlobalMergeAlignment was needed and the command line options.  A
better profitability check would be nice, but can be just a FIXME for
now.

Thanks,
Rafael
-------------- next part --------------
A non-text attachment was scrubbed...
Name: t.patch
Type: text/x-patch
Size: 24866 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140602/023751d3/attachment.bin>


More information about the llvm-commits mailing list