[PATCH] D151347: [RISCV] Use vfslide1down for build_vectors of non-constant floats

[Craig Topper via Phabricator via llvm-commits]

craig.topper added inline comments.


================
Comment at: llvm/lib/Target/RISCV/RISCVISelLowering.h:168
+  // Matches the semantics of vfslide1up/vfslide1down. The first operand is
+  // passthru operand, the second is source vector, third is the FLen scalar
+  // value. The fourth and fifth operands are the mask and VL operands.
----------------
It's not an FLen scalar is it? It should be a scalar with the same element type as the vector operand.


================
Comment at: llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td:2461
+  SDTCisVec<0>, SDTCisSameAs<1, 0>, SDTCisSameAs<2, 0>, SDTCisFP<0>,
+  SDTCisFP<3>, SDTCVecEltisVT<4, i1>, SDTCisSameNumEltsAs<0, 4>,
+  SDTCisVT<5, XLenVT>
----------------
I think `SDTCisFP<3>` can be replaced with `SDTCisEltOfVec<3, 0>` for a stronger guarantee.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D151347/new/

https://reviews.llvm.org/D151347