Address review comments

Author: virginia-cangelosi

clang/include/clang/Basic/arm_sme.td

@@ -324,19 +324,19 @@ let SMETargetGuard = "sme2,sme-mop4,sme-b16b16" in {
 ////////////////////////////////////////////////////////////////////////////////
 // SME2 - FP8 FMOP4A, FMOP4S
 
-multiclass MOP4_FP8<string za, string t, list<ImmCheck> checks> {
-  def _1x1 : Inst<"svmop4a" # "[_1x1]" # za # "[_{d}_{d}]", "vidd>", t, MergeNone, "aarch64_sme_fp8_fmop4a" # za # "_1x1", [IsInOutZA, IsStreaming, IsOverloadNone], checks>;
-  def _1x2 : Inst<"svmop4a" # "[_1x2]" # za # "[_{d}_{d}]", "vid2>", t, MergeNone, "aarch64_sme_fp8_fmop4a" # za # "_1x2", [IsInOutZA, IsStreaming, IsOverloadNone], checks>;
-  def _2x1 : Inst<"svmop4a" # "[_2x1]" # za # "[_{d}_{d}]", "vi2d>", t, MergeNone, "aarch64_sme_fp8_fmop4a" # za # "_2x1", [IsInOutZA, IsStreaming, IsOverloadNone], checks>;
-  def _2x2 : Inst<"svmop4a" # "[_2x2]" # za # "[_{d}_{d}]", "vi22>", t, MergeNone, "aarch64_sme_fp8_fmop4a" # za # "_2x2", [IsInOutZA, IsStreaming, IsOverloadNone], checks>;
+multiclass MOP4_FP8<string za, list<ImmCheck> checks> {
+  def _1x1 : Inst<"svmop4a" # "[_1x1]" # za # "[_{d}_{d}]", "vidd>", "m", MergeNone, "aarch64_sme_fp8_fmop4a" # za # "_1x1", [IsInOutZA, IsStreaming, IsOverloadNone], checks>;
+  def _1x2 : Inst<"svmop4a" # "[_1x2]" # za # "[_{d}_{d}]", "vid2>", "m", MergeNone, "aarch64_sme_fp8_fmop4a" # za # "_1x2", [IsInOutZA, IsStreaming, IsOverloadNone], checks>;
+  def _2x1 : Inst<"svmop4a" # "[_2x1]" # za # "[_{d}_{d}]", "vi2d>", "m", MergeNone, "aarch64_sme_fp8_fmop4a" # za # "_2x1", [IsInOutZA, IsStreaming, IsOverloadNone], checks>;
+  def _2x2 : Inst<"svmop4a" # "[_2x2]" # za # "[_{d}_{d}]", "vi22>", "m", MergeNone, "aarch64_sme_fp8_fmop4a" # za # "_2x2", [IsInOutZA, IsStreaming, IsOverloadNone], checks>;
 }
 
 let SMETargetGuard = "sme2,sme-mop4,sme-f8f32" in {
-  defm SVMOP4A_FP8_ZA32 : MOP4_FP8<"_za32", "m", [ImmCheck<0, ImmCheck0_3>]>;
+  defm SVMOP4A_FP8_ZA32 : MOP4_FP8<"_za32", [ImmCheck<0, ImmCheck0_3>]>;
 }
 
 let SMETargetGuard = "sme2,sme-mop4,sme-f8f16" in {
-  defm SVMOP4A_FP8_ZA16 : MOP4_FP8<"_za16", "m", [ImmCheck<0, ImmCheck0_1>]>;
+  defm SVMOP4A_FP8_ZA16 : MOP4_FP8<"_za16", [ImmCheck<0, ImmCheck0_1>]>;
 }
 
 ////////////////////////////////////////////////////////////////////////////////

llvm/include/llvm/IR/IntrinsicsAArch64.td

@@ -3131,14 +3131,13 @@ let TargetPrefix = "aarch64" in {
           llvm_nxv16i8_ty],
           [ImmArg<ArgIndex<0>>, IntrInaccessibleMemOnly, IntrHasSideEffects]>;
 
-  def int_aarch64_sme_fp8_fmop4a_za16_1x1 : SME_FP8_OuterProduct_Intrinsic_Single_Single;
-  def int_aarch64_sme_fp8_fmop4a_za32_1x1 : SME_FP8_OuterProduct_Intrinsic_Single_Single;
-  def int_aarch64_sme_fp8_fmop4a_za16_1x2 : SME_FP8_OuterProduct_Intrinsic_Single_Multi;
-  def int_aarch64_sme_fp8_fmop4a_za32_1x2 : SME_FP8_OuterProduct_Intrinsic_Single_Multi;
-  def int_aarch64_sme_fp8_fmop4a_za16_2x1 : SME_FP8_OuterProduct_Intrinsic_Single_Multi;
-  def int_aarch64_sme_fp8_fmop4a_za32_2x1 : SME_FP8_OuterProduct_Intrinsic_Single_Multi;
-  def int_aarch64_sme_fp8_fmop4a_za16_2x2 : SME_FP8_OuterProduct_Intrinsic_Multi_Multi;
-  def int_aarch64_sme_fp8_fmop4a_za32_2x2 : SME_FP8_OuterProduct_Intrinsic_Multi_Multi;
+  // 16 and 32 bit multi-vector floating point 8 Quarter Tile Quarter Product
+  foreach za = ["za16", "za32"] in {
+    def int_aarch64_sme_fp8_fmop4a_ # za # "_1x1" : SME_FP8_OuterProduct_Intrinsic_Single_Single;
+    def int_aarch64_sme_fp8_fmop4a_ # za # "_1x2" : SME_FP8_OuterProduct_Intrinsic_Single_Multi;
+    def int_aarch64_sme_fp8_fmop4a_ # za # "_2x1" : SME_FP8_OuterProduct_Intrinsic_Single_Multi;
+    def int_aarch64_sme_fp8_fmop4a_ # za # "_2x2" : SME_FP8_OuterProduct_Intrinsic_Multi_Multi;
+  }
 
   class SME_AddVectorToTile_Intrinsic
       : DefaultAttrsIntrinsic<[],

llvm/lib/Target/AArch64/SMEInstrFormats.td

@@ -5783,7 +5783,7 @@ multiclass sme2_fmop4a_fp8_fp32_4way<string mnemonic, string op> {
 
   def NAME # _MZZ_BtoS_PSEUDO : sme2_quarter_tile_outer_product_pseudo<ZPR8Mul2_Lo, ZPR8Mul2_Hi, SMEMatrixTileS>, SMEPseudo2Instr<NAME # _MZZ_BtoS, 0>;
 
-  def : SME2_ZA_Tile_TwoVec_Pat<NAME # _MZZ_BtoS, !cast<SDPatternOperator>(op # "_1x1"), timm32_0_3, nxv16i8>;
+  def : SME2_ZA_Tile_Vec_Single_Single_Pat<NAME # _MZZ_BtoS, !cast<SDPatternOperator>(op # "_1x1"), timm32_0_3, nxv16i8>;
 
   // Multiple and single vectors
   def _M2ZZ_BtoS : sme2_fp8_fp32_quarter_tile_outer_product<0, 1, mnemonic, ZZ_b_mul_r_Lo, ZPR8Mul2_Hi>, SMEPseudo2Instr<NAME # _M2ZZ_BtoS, 1>;
@@ -5797,7 +5797,7 @@ multiclass sme2_fmop4a_fp8_fp32_4way<string mnemonic, string op> {
 
   def NAME # _MZ2Z_BtoS_PSEUDO : sme2_quarter_tile_outer_product_pseudo<ZPR8Mul2_Lo, ZZ_b_mul_r_Hi, SMEMatrixTileS>, SMEPseudo2Instr<NAME # _MZ2Z_BtoS, 0>;
 
-  def : SME2_ZA_Tile_Vec_Single_Multi_Pat<NAME # _MZ2Z_BtoS, !cast<SDPatternOperator>(op # "_1x2"), timm32_0_3, nxv16i8>;
+  def : SME2_ZA_Tile_Vec_Multi_Pat<NAME # _MZ2Z_BtoS, !cast<SDPatternOperator>(op # "_1x2"), timm32_0_3, nxv16i8>;
 
   // Multiple vectors
   def _M2Z2Z_BtoS : sme2_fp8_fp32_quarter_tile_outer_product<1, 1, mnemonic, ZZ_b_mul_r_Lo, ZZ_b_mul_r_Hi>, SMEPseudo2Instr<NAME # _M2Z2Z_BtoS, 1>;
@@ -6048,7 +6048,7 @@ multiclass sme2_fmop4a_fp8_fp16_2way<string mnemonic, string op> {
 
   def NAME # _MZZ_BtoH_PSEUDO : sme2_quarter_tile_outer_product_pseudo<ZPR8Mul2_Lo, ZPR8Mul2_Hi, SMEMatrixTileH>, SMEPseudo2Instr<NAME # _MZZ_BtoH, 0>;
 
-  def : SME2_ZA_Tile_TwoVec_Pat<NAME # _MZZ_BtoH, !cast<SDPatternOperator>(op # "_1x1"), timm32_0_1, nxv16i8>;
+  def : SME2_ZA_Tile_Vec_Single_Single_Pat<NAME # _MZZ_BtoH, !cast<SDPatternOperator>(op # "_1x1"), timm32_0_1, nxv16i8>;
 
   // Multiple and single vectors
   def _M2ZZ_BtoH : sme2_fp8_fp16_quarter_tile_outer_product<0b0, 0b1, mnemonic, ZZ_b_mul_r_Lo, ZPR8Mul2_Hi>, SMEPseudo2Instr<NAME # _M2ZZ_BtoH, 1>;
@@ -6062,7 +6062,7 @@ multiclass sme2_fmop4a_fp8_fp16_2way<string mnemonic, string op> {
 
   def NAME # _MZ2Z_BtoH_PSEUDO : sme2_quarter_tile_outer_product_pseudo<ZPR8Mul2_Lo, ZZ_b_mul_r_Hi, SMEMatrixTileH>, SMEPseudo2Instr<NAME # _MZ2Z_BtoH, 0>;
 
-  def : SME2_ZA_Tile_Vec_Single_Multi_Pat<NAME # _MZ2Z_BtoH, !cast<SDPatternOperator>(op # "_1x2"), timm32_0_1, nxv16i8>;
+  def : SME2_ZA_Tile_Vec_Multi_Pat<NAME # _MZ2Z_BtoH, !cast<SDPatternOperator>(op # "_1x2"), timm32_0_1, nxv16i8>;
 
   // Multiple vectors
   def _M2Z2Z_BtoH : sme2_fp8_fp16_quarter_tile_outer_product<0b1, 0b1, mnemonic, ZZ_b_mul_r_Lo, ZZ_b_mul_r_Hi>, SMEPseudo2Instr<NAME # _M2Z2Z_BtoH, 1>;