[PATCH] D88750: [GlobalISel] Add G_VECREDUCE_* opcodes for vector reductions

[Amara Emerson via Phabricator via llvm-commits]

aemerson added inline comments.


================
Comment at: llvm/docs/GlobalISel/GenericOpcode.rst:553
+
+G_VECREDUCE_STRICT_FADD, G_VECREDUCE_STRICT_FMUL
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
----------------
nikic wrote:
> aemerson wrote:
> > arsenm wrote:
> > > This name is extremely confusing since these aren't strict FP operations
> > What other name do you suggest? If we change it then we should also change the SDAG node too.
> > 
> > VECREDUCE_STRICT_ORDER_FADD? VECREDUCE_ORDERED_FADD? VECREDUCE_SEQ_FADD?
> I've also found the naming of these SDAG nodes rather confusing. A rename to something like VECREDUCE_ORDERED_FADD would be great.
I'm leaning towards VECREDUCE_SEQ_FADD myself since the term "ordered" is already used in FP contexts.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88750/new/

https://reviews.llvm.org/D88750