[Clang][AArch64] Add fix vector types to header into SVE (#73258)

This patch is needed for the reduction instructions in sve2.1
 It add a new header to sve with all the fixed vector types.
  The new types are only added if neon is not declared.

clang/lib/Headers/CMakeLists.txt

@@ -387,6 +387,8 @@ if(ARM IN_LIST LLVM_TARGETS_TO_BUILD OR AArch64 IN_LIST LLVM_TARGETS_TO_BUILD)
   clang_generate_header(-gen-arm-mve-header arm_mve.td arm_mve.h)
   # Generate arm_cde.h
   clang_generate_header(-gen-arm-cde-header arm_cde.td arm_cde.h)
+  # Generate arm_vector_types.h
+  clang_generate_header(-gen-arm-vector-type arm_neon.td arm_vector_types.h)
 
   # Add headers to target specific lists
   list(APPEND arm_common_generated_files
@@ -403,6 +405,7 @@ if(ARM IN_LIST LLVM_TARGETS_TO_BUILD OR AArch64 IN_LIST LLVM_TARGETS_TO_BUILD)
     "${CMAKE_CURRENT_BINARY_DIR}/arm_sve.h"
     "${CMAKE_CURRENT_BINARY_DIR}/arm_sme_draft_spec_subject_to_change.h"
     "${CMAKE_CURRENT_BINARY_DIR}/arm_bf16.h"
+    "${CMAKE_CURRENT_BINARY_DIR}/arm_vector_types.h"
     )
 endif()
 if(RISCV IN_LIST LLVM_TARGETS_TO_BUILD)

clang/lib/Sema/SemaType.cpp

@@ -8360,12 +8360,25 @@ static void HandleNeonVectorTypeAttr(QualType &CurType, const ParsedAttr &Attr,
   // not to need a separate attribute)
   if (!(S.Context.getTargetInfo().hasFeature("neon") ||
         S.Context.getTargetInfo().hasFeature("mve") ||
-        IsTargetCUDAAndHostARM)) {
+        S.Context.getTargetInfo().hasFeature("sve") ||
+        S.Context.getTargetInfo().hasFeature("sme") ||
+        IsTargetCUDAAndHostARM) &&
+      VecKind == VectorKind::Neon) {
+    S.Diag(Attr.getLoc(), diag::err_attribute_unsupported)
+        << Attr << "'neon', 'mve', 'sve' or 'sme'";
+    Attr.setInvalid();
+    return;
+  }
+  if (!(S.Context.getTargetInfo().hasFeature("neon") ||
+        S.Context.getTargetInfo().hasFeature("mve") ||
+        IsTargetCUDAAndHostARM) &&
+      VecKind == VectorKind::NeonPoly) {
     S.Diag(Attr.getLoc(), diag::err_attribute_unsupported)
         << Attr << "'neon' or 'mve'";
     Attr.setInvalid();
     return;
   }
+
   // Check the attribute arguments.
   if (Attr.getNumArgs() != 1) {
     S.Diag(Attr.getLoc(), diag::err_attribute_wrong_number_arguments)

clang/test/CodeGen/arm-vector_type-params-returns.c

@@ -0,0 +1,134 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 3
+
+// RUN: %clang_cc1 -DSVE_HEADER -triple aarch64 -target-feature +sve -emit-llvm -O2 -o - %s | opt -S -passes=mem2reg,sroa | FileCheck %s
+// RUN: %clang_cc1 -DSVE_HEADER -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -o - /dev/null %s
+
+// RUN: %clang_cc1 -DNEON_HEADER -triple aarch64 -target-feature +sve -emit-llvm -O2 -o - %s | opt -S -passes=mem2reg,sroa | FileCheck %s
+// RUN: %clang_cc1 -DNEON_HEADER -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -o - /dev/null %s
+
+// RUN: %clang_cc1 -DSVE_HEADER -DNEON_HEADER -triple aarch64 -target-feature +sve -emit-llvm -O2 -o - %s | opt -S -passes=mem2reg,sroa | FileCheck %s
+// RUN: %clang_cc1 -DSVE_HEADER -DNEON_HEADER -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -o - /dev/null %s
+
+// RUN: %clang_cc1 -DNEON_HEADER -DSVE_HEADER2  -triple aarch64 -target-feature +sve -emit-llvm -O2 -o - %s | opt -S -passes=mem2reg,sroa | FileCheck %s
+// RUN: %clang_cc1 -DNEON_HEADER -DSVE_HEADER2 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -o - /dev/null %s
+
+#ifdef SVE_HEADER
+  #include <arm_sve.h>
+#endif
+
+#ifdef NEON_HEADER
+  #include <arm_neon.h>
+#endif
+
+#ifdef SVE_HEADER_2
+  #include <arm_sve.h>
+#endif
+
+// function return types
+// CHECK-LABEL: define dso_local <8 x half> @test_ret_v8f16(
+// CHECK-SAME: <8 x half> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <8 x half> [[V]]
+//
+float16x8_t test_ret_v8f16(float16x8_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <4 x float> @test_ret_v4f32(
+// CHECK-SAME: <4 x float> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <4 x float> [[V]]
+//
+float32x4_t test_ret_v4f32(float32x4_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <2 x double> @test_ret_v2f64(
+// CHECK-SAME: <2 x double> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <2 x double> [[V]]
+//
+float64x2_t test_ret_v2f64(float64x2_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <8 x bfloat> @test_ret_v8bf16(
+// CHECK-SAME: <8 x bfloat> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <8 x bfloat> [[V]]
+//
+bfloat16x8_t test_ret_v8bf16(bfloat16x8_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <16 x i8> @test_ret_v16s8(
+// CHECK-SAME: <16 x i8> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <16 x i8> [[V]]
+//
+int8x16_t test_ret_v16s8(int8x16_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <8 x i16> @test_ret_v8s16(
+// CHECK-SAME: <8 x i16> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <8 x i16> [[V]]
+//
+int16x8_t test_ret_v8s16(int16x8_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <4 x i32> @test_ret_v32s4(
+// CHECK-SAME: <4 x i32> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <4 x i32> [[V]]
+//
+int32x4_t test_ret_v32s4(int32x4_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <2 x i64> @test_ret_v64s2(
+// CHECK-SAME: <2 x i64> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <2 x i64> [[V]]
+//
+int64x2_t test_ret_v64s2(int64x2_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <16 x i8> @test_ret_v16u8(
+// CHECK-SAME: <16 x i8> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <16 x i8> [[V]]
+//
+uint8x16_t test_ret_v16u8(uint8x16_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <8 x i16> @test_ret_v8u16(
+// CHECK-SAME: <8 x i16> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <8 x i16> [[V]]
+//
+uint16x8_t test_ret_v8u16(uint16x8_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <4 x i32> @test_ret_v32u4(
+// CHECK-SAME: <4 x i32> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <4 x i32> [[V]]
+//
+uint32x4_t test_ret_v32u4(uint32x4_t v) {
+  return v;
+}
+
+// CHECK-LABEL: define dso_local <2 x i64> @test_ret_v64u2(
+// CHECK-SAME: <2 x i64> noundef returned [[V:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret <2 x i64> [[V]]
+//
+uint64x2_t test_ret_v64u2(uint64x2_t v) {
+  return v;
+}

clang/test/Sema/aarch64-sve-intrinsics/acle_sve_target.cpp

@@ -1,4 +1,4 @@
-// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -fsyntax-only -verify -emit-llvm -o - -ferror-limit 100 %s
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +neon -fsyntax-only -verify -emit-llvm -o - -ferror-limit 100 %s
 // REQUIRES: aarch64-registered-target
 
 // Test that functions with the correct target attributes can use the correct SVE intrinsics.
@@ -29,4 +29,5 @@ void __attribute__((target("sve2-sha3"))) test_sve2_sha3()
 void __attribute__((target("sve2"))) test_f16(svbool_t pg)
 {
   svlogb_f16_z(pg, svundef_f16());
-}
+}
+

clang/test/Sema/arm-vector-types-support.c

@@ -1,5 +1,5 @@
 // RUN: %clang_cc1 %s -triple armv7 -fsyntax-only -verify
 
-typedef __attribute__((neon_vector_type(2))) int int32x2_t; // expected-error{{'neon_vector_type' attribute is not supported on targets missing 'neon' or 'mve'; specify an appropriate -march= or -mcpu=}}
+typedef __attribute__((neon_vector_type(2))) int int32x2_t; // expected-error{{'neon_vector_type' attribute is not supported on targets missing 'neon', 'mve', 'sve' or 'sme'; specify an appropriate -march= or -mcpu=}}
 typedef __attribute__((neon_polyvector_type(16))) short poly8x16_t; // expected-error{{'neon_polyvector_type' attribute is not supported on targets missing 'neon' or 'mve'; specify an appropriate -march= or -mcpu=}}
 typedef __attribute__((arm_sve_vector_bits(256))) void nosveflag; // expected-error{{'arm_sve_vector_bits' attribute is not supported on targets missing 'sve'; specify an appropriate -march= or -mcpu=}}

clang/test/SemaCUDA/neon-attrs.cu

@@ -15,7 +15,8 @@
 
 // quiet-no-diagnostics
 typedef __attribute__((neon_vector_type(4))) float float32x4_t;
-// expected-error@-1 {{'neon_vector_type' attribute is not supported on targets missing 'neon' or 'mve'}}
+// expected-error@-1 {{'neon_vector_type' attribute is not supported on targets missing 'neon', 'mve', 'sve' or 'sme'}}
+// expect
 typedef unsigned char poly8_t;
 typedef __attribute__((neon_polyvector_type(8))) poly8_t poly8x8_t;
 // expected-error@-1 {{'neon_polyvector_type' attribute is not supported on targets missing 'neon' or 'mve'}}

clang/utils/TableGen/NeonEmitter.cpp

@@ -593,6 +593,8 @@ public:
   // Emit arm_bf16.h.inc
   void runBF16(raw_ostream &o);
 
+  void runVectorTypes(raw_ostream &o);
+
   // Emit all the __builtin prototypes used in arm_neon.h, arm_fp16.h and
   // arm_bf16.h
   void runHeader(raw_ostream &o);
@@ -2355,13 +2357,7 @@ void NeonEmitter::run(raw_ostream &OS) {
 
   OS << "#include <arm_bf16.h>\n";
 
-  // Emit NEON-specific scalar typedefs.
-  OS << "typedef float float32_t;\n";
-  OS << "typedef __fp16 float16_t;\n";
-
-  OS << "#ifdef __aarch64__\n";
-  OS << "typedef double float64_t;\n";
-  OS << "#endif\n\n";
+  OS << "#include <arm_vector_types.h>\n";
 
   // For now, signedness of polynomial types depends on target
   OS << "#ifdef __aarch64__\n";
@@ -2374,10 +2370,7 @@ void NeonEmitter::run(raw_ostream &OS) {
   OS << "typedef int16_t poly16_t;\n";
   OS << "typedef int64_t poly64_t;\n";
   OS << "#endif\n";
-
-  emitNeonTypeDefs("cQcsQsiQilQlUcQUcUsQUsUiQUiUlQUlhQhfQfdQdPcQPcPsQPsPlQPl", OS);
-
-  emitNeonTypeDefs("bQb", OS);
+  emitNeonTypeDefs("PcQPcPsQPsPlQPl", OS);
 
   OS << "#define __ai static __inline__ __attribute__((__always_inline__, "
         "__nodebug__))\n\n";
@@ -2546,6 +2539,38 @@ void NeonEmitter::runFP16(raw_ostream &OS) {
   OS << "#endif /* __ARM_FP16_H */\n";
 }
 
+void NeonEmitter::runVectorTypes(raw_ostream &OS) {
+  OS << "/*===---- arm_vector_types - ARM vector type "
+        "------===\n"
+        " *\n"
+        " *\n"
+        " * Part of the LLVM Project, under the Apache License v2.0 with LLVM "
+        "Exceptions.\n"
+        " * See https://llvm.org/LICENSE.txt for license information.\n"
+        " * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception\n"
+        " *\n"
+        " *===-----------------------------------------------------------------"
+        "------===\n"
+        " */\n\n";
+  OS << "#if !defined(__ARM_NEON_H) && !defined(__ARM_SVE_H)\n";
+  OS << "#error \"This file should not be used standalone. Please include"
+        " arm_neon.h or arm_sve.h instead\"\n\n";
+  OS << "#endif\n";
+  OS << "#ifndef __ARM_NEON_TYPES_H\n";
+  OS << "#define __ARM_NEON_TYPES_H\n";
+  OS << "typedef float float32_t;\n";
+  OS << "typedef __fp16 float16_t;\n";
+
+  OS << "#ifdef __aarch64__\n";
+  OS << "typedef double float64_t;\n";
+  OS << "#endif\n\n";
+
+  emitNeonTypeDefs("cQcsQsiQilQlUcQUcUsQUsUiQUiUlQUlhQhfQfdQd", OS);
+
+  emitNeonTypeDefs("bQb", OS);
+  OS << "#endif // __ARM_NEON_TYPES_H\n";
+}
+
 void NeonEmitter::runBF16(raw_ostream &OS) {
   OS << "/*===---- arm_bf16.h - ARM BF16 intrinsics "
         "-----------------------------------===\n"
@@ -2640,6 +2665,10 @@ void clang::EmitNeonSema(RecordKeeper &Records, raw_ostream &OS) {
   NeonEmitter(Records).runHeader(OS);
 }
 
+void clang::EmitVectorTypes(RecordKeeper &Records, raw_ostream &OS) {
+  NeonEmitter(Records).runVectorTypes(OS);
+}
+
 void clang::EmitNeonTest(RecordKeeper &Records, raw_ostream &OS) {
   llvm_unreachable("Neon test generation no longer implemented!");
 }

clang/utils/TableGen/SveEmitter.cpp

@@ -1286,6 +1286,7 @@ void SVEEmitter::createHeader(raw_ostream &OS) {
   OS << "typedef __SVBfloat16_t svbfloat16_t;\n";
 
   OS << "#include <arm_bf16.h>\n";
+  OS << "#include <arm_vector_types.h>\n";
 
   OS << "typedef __SVFloat32_t svfloat32_t;\n";
   OS << "typedef __SVFloat64_t svfloat64_t;\n";
@@ -1730,4 +1731,5 @@ void EmitSmeBuiltinCG(RecordKeeper &Records, raw_ostream &OS) {
 void EmitSmeRangeChecks(RecordKeeper &Records, raw_ostream &OS) {
   SVEEmitter(Records).createSMERangeChecks(OS);
 }
+
 } // End namespace clang

clang/utils/TableGen/TableGen.cpp

@@ -73,6 +73,7 @@ enum ActionType {
   GenArmNeon,
   GenArmFP16,
   GenArmBF16,
+  GenArmVectorType,
   GenArmNeonSema,
   GenArmNeonTest,
   GenArmMveHeader,
@@ -229,6 +230,8 @@ cl::opt<ActionType> Action(
         clEnumValN(GenArmNeon, "gen-arm-neon", "Generate arm_neon.h for clang"),
         clEnumValN(GenArmFP16, "gen-arm-fp16", "Generate arm_fp16.h for clang"),
         clEnumValN(GenArmBF16, "gen-arm-bf16", "Generate arm_bf16.h for clang"),
+        clEnumValN(GenArmVectorType, "gen-arm-vector-type",
+                   "Generate arm_vector_types.h for clang"),
         clEnumValN(GenArmNeonSema, "gen-arm-neon-sema",
                    "Generate ARM NEON sema support for clang"),
         clEnumValN(GenArmNeonTest, "gen-arm-neon-test",
@@ -449,6 +452,9 @@ bool ClangTableGenMain(raw_ostream &OS, RecordKeeper &Records) {
   case GenArmFP16:
     EmitFP16(Records, OS);
     break;
+  case GenArmVectorType:
+    EmitVectorTypes(Records, OS);
+    break;
   case GenArmBF16:
     EmitBF16(Records, OS);
     break;

clang/utils/TableGen/TableGenBackends.h

@@ -97,6 +97,7 @@ void EmitNeon(llvm::RecordKeeper &Records, llvm::raw_ostream &OS);
 void EmitFP16(llvm::RecordKeeper &Records, llvm::raw_ostream &OS);
 void EmitBF16(llvm::RecordKeeper &Records, llvm::raw_ostream &OS);
 void EmitNeonSema(llvm::RecordKeeper &Records, llvm::raw_ostream &OS);
+void EmitVectorTypes(llvm::RecordKeeper &Records, llvm::raw_ostream &OS);
 void EmitNeonTest(llvm::RecordKeeper &Records, llvm::raw_ostream &OS);
 
 void EmitSveHeader(llvm::RecordKeeper &Records, llvm::raw_ostream &OS);