[clang] [llvm] [AArch64] Add intrinsics for SME FP8 FVDOT, FVDOTB and FVDOTT intrinsics (PR #119922)

[via llvm-commits]

================
@@ -5882,3 +5882,14 @@ multiclass sme2_fp8_fdot_index_za32_vg1x4<string mnemonic,
 
   def : SME2_ZA_TwoOp_VG4_Multi_Index_Pat<NAME, intrinsic, sme_elm_idx0_7, ZPR4b8, nxv16i8, VectorIndexS32b_timm, tileslice16>;
 }
+
+multiclass sme2_fp8_fdotv_index_za32_vg1x4<string mnemonic, bit T, SDPatternOperator intrinsic> {
+  def NAME : sme2_fp8_multi_vec_array_vg4_index<mnemonic, T>,
+                                            SMEPseudo2Instr<NAME, 1> {
+    let Uses=[FPMR, FPCR];
+  }
+
+  def _PSEUDO : sme2_za_array_2op_multi_index_pseudo<NAME, sme_elm_idx0_7, ZZ_b_mul_r, ZPR4b8, VectorIndexS32b_timm, SMEMatrixArray>;
+
+  def : SME2_ZA_TwoOp_VG2_Multi_Index_Pat<NAME, intrinsic, sme_elm_idx0_7, ZPR4b8, nxv16i8, VectorIndexH32b_timm, tileslice16>;
----------------
SpencerAbson wrote:

I think the appropriate `imm_ty` argument to this pattern is`VectorInedxS32b_timm`.

https://github.com/llvm/llvm-project/pull/119922