[clang] [llvm] [Clang][LLVM] Implement single-single vectors MOP4{A/S} (PR #127797)

[via cfe-commits]

================
@@ -5617,9 +5648,13 @@ class sme2_fp16_quarter_tile_outer_product<bit M, bit N, bit S, string mnemonic,
   let Constraints = "$ZAda = $_ZAda";
 }
 
-multiclass sme2_fmop4as_fp16_non_widening<bit S, string mnemonic> {
+multiclass sme2_fmop4as_fp16_non_widening<bit S, string mnemonic, string op> {
   // Single vectors
-  def _MZZ_H : sme2_fp16_quarter_tile_outer_product<0, 0, S, mnemonic, ZPR16Mul2_Lo, ZPR16Mul2_Hi>;
+  def _MZZ_H : sme2_fp16_quarter_tile_outer_product<0, 0, S, mnemonic, ZPR16Mul2_Lo, ZPR16Mul2_Hi>, SMEPseudo2Instr<NAME # _MZZ_H, 1>;
+
+  def NAME # _MZZ_H # _PSEUDO : sme2_quarter_tile_outer_product_pseudo_single_single<ZPR16Mul2_Lo, ZPR16Mul2_Hi, SMEMatrixTileH>, SMEPseudo2Instr<NAME # _MZZ_H, 0>;
+
+  def : SME2_ZA_Tile_TwoVec_Pat<NAME # _MZZ_H, !cast<SDPatternOperator>(op # "_1x1"), timm32_0_3, nxv8f16>;
----------------
CarolineConcatto wrote:

I believe that here timm32_0_3 should be timm32_0_1.
Half-precision, single vectors" variants: is the name of the ZA tile ZA0-ZA1, encoded in the "ZAda" field.
From:https://developer.arm.com/documentation/ddi0602/2024-12/SME-Instructions/FMOP4A--non-widening---Floating-point-quarter-tile-outer-products--accumulating-?lang=en


https://github.com/llvm/llvm-project/pull/127797