[PATCH] D69653: [GlobalISel] Match table opt: fix a bug in matching num of operands

Quentin Colombet via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Nov 1 09:29:31 PDT 2019


qcolombet added a comment.

Thanks for the test case!



================
Comment at: llvm/utils/TableGen/GlobalISelEmitter.cpp:2276
   if (Stash.back().get() == &OpcMatcher) {
-    if (NumOperandsCheck && OpcMatcher.getNumOperands() < getNumOperands())
+    if (NumOperandsCheck && OpcMatcher.isVariadicNumOperands())
       Stash.emplace_back(
----------------
rtereshin wrote:
> qcolombet wrote:
> > Stupid question: what complains if an instruction does not have the `isVariadic` attribute, but we write patterns with different number of operands?
> > (Just wondering if we need an assert that !isVariadic => getNumOperands() == OpcMatcher.getNumOperands()).
> > Stupid question: what complains if an instruction does not have the isVariadic attribute, but we write patterns with different number of operands?
> This does: https://github.com/llvm/llvm-project/blob/master/llvm/utils/TableGen/CodeGenDAGPatterns.cpp#L2414-L2420
> 
> > (Just wondering if we need an assert that !isVariadic => getNumOperands() == OpcMatcher.getNumOperands()).
> Considered the same, decided redundant.
Sounds good.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69653/new/

https://reviews.llvm.org/D69653





More information about the llvm-commits mailing list