[PATCH] [mips] Add support for branch-likely pseudo-instructions
Daniel Sanders
daniel.sanders at imgtec.com
Mon Jun 22 07:24:24 PDT 2015
> New diff contains fix for error reporting when branch likely instructions are used on mips32r6/mips64r6.
I should have thought of that :-). Could you add test cases for this?
> Please, take a look at AsmToken::Identifier case, since I am not sure this is the right way to parse a label.
You shouldn't need any of the code you added for this. We can already parse labels so specifying 'brtarget' should be sufficient. I'll take a quick look.
================
Comment at: lib/Target/Mips/AsmParser/MipsAsmParser.cpp:2383-2385
@@ -2354,5 +2382,5 @@
}
- // If both registers are $0 and the pseudo-branch does not accept
- // equality, it will never be taken, so we don't have to emit anything.
- return false;
+ // If both registers are $0 and the pseudo-branch does not accept
+ // equality, it will never be taken, so we don't have to emit anything.
+ return false;
}
----------------
Nit: You seem to have added some extra indentation since the last version of your patch. It was correct in the previous patch
================
Comment at: lib/Target/Mips/MipsInstrInfo.td:1698
@@ -1697,3 +1697,3 @@
brtarget:$offset),
- !strconcat(instr_asm, "\t$rs, $rt, $offset")>;
+ !strconcat(instr_asm, "\t$rs, $rt, $offset")>, PredicateControl;
}
----------------
PredicateControl belongs on MipsAsmPseudoInst rather than CondBranchPseudo
http://reviews.llvm.org/D10537
EMAIL PREFERENCES
http://reviews.llvm.org/settings/panel/emailpreferences/
More information about the llvm-commits
mailing list