[Clang][LLVM] Implement single-multi vectors MOP4{A/S} (#128854)

Implement all single-multi {BF/F/S/U/SU/US}MOP4{A/S} instructions in
clang and llvm following the acle in
https://github.com/ARM-software/acle/pull/381/files.

This PR depends on #127797

This patch updates the semantics of template arguments in intrinsic
names for clarity and ease of use. Previously, template argument numbers
indicated which character in the prototype string determined the final
type suffix, which was confusing—especially for intrinsics using
multiple prototype modifiers per operand (e.g., intrinsics operating on
arrays of vectors). The number had to reference the correct character in
the prototype (e.g., the ‘u’ in “2.u”), making the system cumbersome and
error-prone.
With this patch, template argument numbers now refer to the operand
number that determines the final type suffix, providing a more intuitive
and consistent approach.

Author: virginia-cangelosi

clang/include/clang/Basic/arm_sme.td

@@ -294,6 +294,7 @@ defm SVMOPS : ZAFPOuterProd<"mops">;
 
 multiclass MOP4<string mode, string za, string t, string i, list<ImmCheck> checks> {
   def _1x1 : Inst<"svmop4" # mode # "[_1x1]" # za # "[_{d}_{d}]", "vidd", t, MergeNone, i # "_1x1", [IsInOutZA, IsStreaming], checks>;
+  def _1x2 : Inst<"svmop4" # mode # "[_1x2]" # za # "[_{d}_{d}]", "vid2", t, MergeNone, i # "_1x2", [IsInOutZA, IsStreaming], checks>;
 }
 
 let SMETargetGuard = "sme2,sme-mop4" in {
@@ -345,13 +346,21 @@ multiclass SUMOP4<string mode, string za, string t, string i, list<ImmCheck> che
                               "vidu", t, MergeNone, "aarch64_sme_sumop4" # mode # i # "_wide_1x1",
                               [IsStreaming, IsInOutZA],
                               checks>;
+  def _1x2 : SInst<"svmop4" # mode # "[_1x2]" # za # "[_{d}_{3}]",
+                              "vid2.u", t, MergeNone, "aarch64_sme_sumop4" # mode # i # "_wide_1x2",
+                              [IsStreaming, IsInOutZA],
+                              checks>;
 }
 
 multiclass USMOP4<string mode, string za, string t, string i, list<ImmCheck> checks> {
   def _1x1 : SInst<"svmop4" # mode # "[_1x1]" # za # "[_{d}_{3}]",
                               "vidx", t, MergeNone, "aarch64_sme_usmop4" # mode # i # "_wide_1x1",
                               [IsStreaming, IsInOutZA],
                               checks>;
+  def _1x2 : SInst<"svmop4" # mode # "[_1x2]" # za # "[_{d}_{3}]",
+                              "vid2.x", t, MergeNone, "aarch64_sme_usmop4" # mode # i # "_wide_1x2",
+                              [IsStreaming, IsInOutZA],
+                              checks>;
 }
 
 let SMETargetGuard = "sme2,sme-mop4" in {

clang/include/clang/Basic/arm_sve.td

@@ -2420,8 +2420,8 @@ let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2" in {
 let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2" in {
   def SVSUNPK_X2 : SInst<"svunpk_{d}[_{1}_x2]", "2h",   "sil",    MergeNone, "aarch64_sve_sunpk_x2", [IsStreaming], []>;
   def SVUUNPK_X2 : SInst<"svunpk_{d}[_{1}_x2]", "2h",   "UsUiUl", MergeNone, "aarch64_sve_uunpk_x2", [IsStreaming], []>;
-  def SVSUNPK_X4 : SInst<"svunpk_{d}[_{3}_x4]", "42.h", "sil",    MergeNone, "aarch64_sve_sunpk_x4", [IsStreaming], []>;
-  def SVUUNPK_X4 : SInst<"svunpk_{d}[_{3}_x4]", "42.h", "UsUiUl", MergeNone, "aarch64_sve_uunpk_x4", [IsStreaming], []>;
+  def SVSUNPK_X4 : SInst<"svunpk_{d}[_{1}_x4]", "42.h", "sil",    MergeNone, "aarch64_sve_sunpk_x4", [IsStreaming], []>;
+  def SVUUNPK_X4 : SInst<"svunpk_{d}[_{1}_x4]", "42.h", "UsUiUl", MergeNone, "aarch64_sve_uunpk_x4", [IsStreaming], []>;
 }
 
 let SVETargetGuard = InvalidMode, SMETargetGuard = "sme2,fp8" in {

clang/test/CodeGen/AArch64/sme2-intrinsics/acle_sme2_mop4_1x2.c

@@ -6,27 +6,27 @@
 
 #include <arm_sme.h>
 
-void tests_mop4_imm_s8_s8(svint8_t zn, svint8_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_s8_s8_1x1(svint8_t zn, svint8_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za32_s8_s8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_s8_s8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   return;
 }
 
-void tests_mop4_imm_u8_u8(svuint8_t zn, svuint8_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_u8_u8_1x1(svuint8_t zn, svuint8_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za32_u8_u8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_u8_u8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   return;
 }
 
-void tests_mop4_imm_s8_u8(svint8_t zn, svuint8_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_s8_u8_1x1(svint8_t zn, svuint8_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za32_s8_u8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_s8_u8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4a_1x1_za32_u8_s8(-1, zm, zn); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_u8_s8(-1, zm, zn); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   return;
 }
 
-void tests_mop4_imm_s16_s16(svint16_t zn, svint16_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_s16_s16_1x1(svint16_t zn, svint16_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za32_s16_s16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_s16_s16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
 
@@ -35,7 +35,7 @@ void tests_mop4_imm_s16_s16(svint16_t zn, svint16_t zm) __arm_streaming __arm_in
   return;
 }
 
-void tests_mop4_imm_u16_u16(svuint16_t zn, svuint16_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_u16_u16_1x1(svuint16_t zn, svuint16_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za32_u16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_u16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
 
@@ -44,15 +44,15 @@ void tests_mop4_imm_u16_u16(svuint16_t zn, svuint16_t zm) __arm_streaming __arm_
   return;
 }
 
-void tests_mop4_imm_s16_u16(svint16_t zn, svuint16_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_s16_u16_1x1(svint16_t zn, svuint16_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za64_s16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
   svmop4s_1x1_za64_s16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
   svmop4a_1x1_za64_u16_s16(-1, zm, zn); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
   svmop4s_1x1_za64_u16_s16(-1, zm, zn); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
   return;
 }
 
-void tests_mop4_imm_f16_f16(svfloat16_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_f16_f16_1x1(svfloat16_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za32_f16_f16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_f16_f16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
 
@@ -61,7 +61,7 @@ void tests_mop4_imm_f16_f16(svfloat16_t zn, svfloat16_t zm) __arm_streaming __ar
   return;
 }
 
-void tests_mop4_imm_bf16_bf16(svbfloat16_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_bf16_bf16_1x1(svbfloat16_t zn, svbfloat16_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za32_bf16_bf16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_bf16_bf16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
 
@@ -71,14 +71,98 @@ void tests_mop4_imm_bf16_bf16(svbfloat16_t zn, svbfloat16_t zm) __arm_streaming
 
 }
 
-void tests_mop4_imm_f32_f32(svfloat32_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_f32_f32_1x1(svfloat32_t zn, svfloat32_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za32_f32_f32(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   svmop4s_1x1_za32_f32_f32(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
   return;
 }
 
-void tests_mop4_imm_f64_f64(svfloat64_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {
+void tests_mop4_imm_f64_f64_1x1(svfloat64_t zn, svfloat64_t zm) __arm_streaming __arm_inout("za") {
   svmop4a_1x1_za64_f64_f64(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
   svmop4s_1x1_za64_f64_f64(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
   return;
 }
+
+void tests_mop4_imm_s8_s8_1x2(svint8_t zn, svint8x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_s8_s8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_s8_s8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  return;
+}
+
+void tests_mop4_imm_u8_u8_1x2(svuint8_t zn, svuint8x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_u8_u8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_u8_u8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  return;
+}
+
+void tests_mop4_imm_s8_u8_1x2(svint8_t zn, svuint8x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_s8_u8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_s8_u8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  return;
+}
+
+void tests_mop4_imm_u8_s8_1x2(svuint8_t zn, svint8x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_u8_s8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_u8_s8(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  return;
+}
+
+void tests_mop4_imm_s16_s16_1x2(svint16_t zn, svint16x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_s16_s16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_s16_s16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+
+  svmop4a_1x2_za64_s16_s16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  svmop4s_1x2_za64_s16_s16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  return;
+}
+
+void tests_mop4_imm_u16_u16_1x2(svuint16_t zn, svuint16x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_u16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_u16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+
+  svmop4a_1x2_za64_u16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  svmop4s_1x2_za64_u16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  return;
+}
+
+void tests_mop4_imm_s16_u16_1x2(svint16_t zn, svuint16x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za64_s16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  svmop4s_1x2_za64_s16_u16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  return;
+}
+
+void tests_mop4_imm_u16_s16_1x2(svuint16_t zn, svint16x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za64_u16_s16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  svmop4s_1x2_za64_u16_s16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  return;
+}
+
+void tests_mop4_imm_f16_f16_1x2(svfloat16_t zn, svfloat16x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_f16_f16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_f16_f16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+
+  svmop4a_1x2_za16_f16_f16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 1]}}
+  svmop4s_1x2_za16_f16_f16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 1]}}
+  return;
+}
+
+void tests_mop4_imm_bf16_bf16_1x2(svbfloat16_t zn, svbfloat16x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_bf16_bf16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_bf16_bf16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+
+  svmop4a_1x2_za16_bf16_bf16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 1]}}
+  svmop4s_1x2_za16_bf16_bf16(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 1]}}
+  return;
+}
+
+void tests_mop4_imm_f32_f32_1x2(svfloat32_t zn, svfloat32x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za32_f32_f32(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  svmop4s_1x2_za32_f32_f32(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 3]}}
+  return;
+}
+
+void tests_mop4_imm_f64_f64_1x2(svfloat64_t zn, svfloat64x2_t zm) __arm_streaming __arm_inout("za") {
+  svmop4a_1x2_za64_f64_f64(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  svmop4s_1x2_za64_f64_f64(-1, zn, zm); // expected-error {{argument value 18446744073709551615 is outside the valid range [0, 7]}}
+  return;
+}

clang/test/Sema/aarch64-sme2p2-instrinsics/acle_sme2p2_imm.cpp

@@ -1043,7 +1043,10 @@ std::string Intrinsic::replaceTemplatedArgs(std::string Name, TypeSpec TS,
     case '1':
     case '2':
     case '3':
-      T = SVEType(TS, Proto[C - '0']);
+      // Extract the modifier before passing to SVEType to handle numeric
+      // modifiers
+      auto [Mod, NumVectors] = getProtoModifier(Proto, (C - '0'));
+      T = SVEType(TS, Mod);
       break;
     }
 

clang/utils/TableGen/SveEmitter.cpp

@@ -3070,11 +3070,19 @@ let TargetPrefix = "aarch64" in {
           llvm_anyvector_ty,
           LLVMMatchType<0>], [ImmArg<ArgIndex<0>>, IntrNoMem, IntrHasSideEffects]>;
 
+  class SME_OuterProduct_QuarterTile_Single_Multi
+      : DefaultAttrsIntrinsic<[],
+          [llvm_i32_ty,
+          llvm_anyvector_ty,
+          LLVMMatchType<0>,
+          LLVMMatchType<0>], [ImmArg<ArgIndex<0>>, IntrNoMem, IntrHasSideEffects]>;
+
   // 2-way and 4-way multi-vector signed/unsigned Quarter Tile Quarter Product A/S
   foreach mode = ["s", "a"] in {
     foreach za = ["", "_za64"] in {
       foreach ty = ["s", "u", "su", "us"] in {
         def int_aarch64_sme_ # ty # "mop4" # mode # za # "_wide_1x1" : SME_OuterProduct_QuarterTile_Single_Single;
+        def int_aarch64_sme_ # ty # "mop4" # mode # za # "_wide_1x2" : SME_OuterProduct_QuarterTile_Single_Multi;
       }
     }
   }
@@ -3083,9 +3091,10 @@ let TargetPrefix = "aarch64" in {
   foreach mode = ["s", "a"] in {
     foreach wide = ["", "_wide"] in {
       def int_aarch64_sme_mop4 # mode # wide # "_1x1" : SME_OuterProduct_QuarterTile_Single_Single;
+      def int_aarch64_sme_mop4 # mode # wide # "_1x2" : SME_OuterProduct_QuarterTile_Single_Multi;
     }
   }
-
+  
   class SME_AddVectorToTile_Intrinsic
       : DefaultAttrsIntrinsic<[],
           [llvm_i32_ty,