[PATCH] D114713: [AArch64][SVE][NEON] Add NEON-SVE-Bridge intrinsics

[Matt Devereau via Phabricator via llvm-commits]

MattDevereau added inline comments.


================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:1325
       setOperationAction(ISD::MLOAD, VT, Custom);
+      setOperationAction(ISD::INSERT_SUBVECTOR, VT, Custom);
     }
----------------
paulwalker-arm wrote:
> Can you extract this into its own patch as it's really not relevant to the rest of the patch and is currently missing tests.  Presumably `llvm/test/CodeGen/AArch64/sve-insert-vector.ll` needs updating?
i've been adding some tests to assert this block of code. i've got tests for `insert(vscale x n x bfloat, n x bfloat, idx)` and `insert(vscale x n x bfloat, vscale x n x bfloat, idx)`.
the n = 4 and n = 8 tests are fine, but n = 2 for `insert(vscale x 2 x bfloat, 2 x bfloat, idx)`  fails an assertion. i've had a quick poke around but haven't seen an obvious reason why its failing, should I worry about this and spend more time on it or just submit the tests i've already got for `4bf16` and `8bf16`?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D114713/new/

https://reviews.llvm.org/D114713