[AArch64][SME] Make the overloaded svreinterpret_* functions streaming-compatible.

[sdesmalen]

Summary

Otherwise these functions are not inlined when invoked from streaming
functions.

[david-arm]

clang/include/clang/Basic/Attr.td

418

I'm not sure why this change is included in this patch?

clang/test/CodeGen/attr-arm-sve-vector-bits-types.c

6

I can see this line looks redundant, but is it supposed to be related to the other changes?

[sdesmalen]

Added "aarch64_32" to TargetAArch64.

clang/include/clang/Basic/Attr.td

418

This is needed for clang/test/Sema/aarch64-sve-intrinsics/big_endian.cpp which has aarch64_be as target and includes arm_sve.h.
Without this change the __arm_streaming_compatible attribute is unsupported and leads to loads of error messages, thus failing the test.

clang/include/clang/Basic/Attr.td

class TargetSpec {
   //   const llvm::Triple &T = Target.getTriple();
   code CustomCode = [{}];
 }
 
 class TargetArch<list<string> arches> : TargetSpec {
   let Arches = arches;
 }
 def TargetARM : TargetArch<["arm", "thumb", "armeb", "thumbeb"]>;
-def TargetAArch64 : TargetArch<["aarch64"]>;
+def TargetAArch64 : TargetArch<["aarch64", "aarch64_be"]>;
 def TargetAnyArm : TargetArch<!listconcat(TargetARM.Arches, TargetAArch64.Arches)>;
 def TargetAVR : TargetArch<["avr"]>;
 def TargetBPF : TargetArch<["bpfel", "bpfeb"]>;
 def TargetMips32 : TargetArch<["mips", "mipsel"]>;
 def TargetAnyMips : TargetArch<["mips", "mipsel", "mips64", "mips64el"]>;
 def TargetMSP430 : TargetArch<["msp430"]>;
 def TargetM68k : TargetArch<["m68k"]>;
 def TargetRISCV : TargetArch<["riscv32", "riscv64"]>;

clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_reinterpret_from_streaming_mode.c

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
+// REQUIRES: aarch64-registered-target
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -S -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
+// RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -S -O1 -Werror -Wall -emit-llvm -o - -x c++ %s | FileCheck %s -check-prefix=CPP-CHECK
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -S -O1 -Werror -Wall -o /dev/null %s
+
+// Note: We need to run this test with '-O1' because oddly enough the svreinterpret is always inlined at -O0.
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+// Test that svreinterpret is inlined (because it should be streaming-compatible)
+__attribute__((target("sme")))
+// CHECK-LABEL: @test_svreinterpret_s16_s8_from_streaming_mode(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP:%.*]] to <vscale x 8 x i16>
+// CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z45test_svreinterpret_s16_s8_from_streaming_modeu10__SVInt8_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP:%.*]] to <vscale x 8 x i16>
+// CPP-CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
+//
+svint16_t test_svreinterpret_s16_s8_from_streaming_mode(svint8_t op) __arm_streaming {
+  return SVE_ACLE_FUNC(svreinterpret_s16,_s8,,)(op);
+}
+

clang/test/CodeGen/attr-arm-sve-vector-bits-types.c

 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +bf16 -mvscale-min=1 -mvscale-max=1 -S -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-128
 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +bf16 -mvscale-min=2 -mvscale-max=2 -S -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-256
 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +bf16 -mvscale-min=4 -mvscale-max=4 -S -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-512
 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +bf16 -mvscale-min=8 -mvscale-max=8 -S -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-1024
 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +bf16 -mvscale-min=16 -mvscale-max=16 -S -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-2048
-// RUN: %clang_cc1 -triple aarch64_32-unknown-darwin -target-feature +sve -target-feature +bf16 -mvscale-min=4 -mvscale-max=4 -S -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK-ILP32
 
 // REQUIRES: aarch64-registered-target
 
 #include <arm_sve.h>
 
 #define N __ARM_FEATURE_SVE_BITS
 
 typedef svint8_t fixed_int8_t __attribute__((arm_sve_vector_bits(N)));

clang/utils/TableGen/SveEmitter.cpp

void SVEEmitter::createHeader(raw_ostream &OS) {
 
   // Add reinterpret functions.
   for (auto ShortForm : { false, true } )
     for (const ReinterpretTypeInfo &From : Reinterprets)
       for (const ReinterpretTypeInfo &To : Reinterprets) {
         if (ShortForm) {
           OS << "__aio __attribute__((target(\"sve\"))) " << From.Type
              << " svreinterpret_" << From.Suffix;
-          OS << "(" << To.Type << " op) {\n";
+          OS << "(" << To.Type << " op) __arm_streaming_compatible {\n";
           OS << "  return __builtin_sve_reinterpret_" << From.Suffix << "_"
              << To.Suffix << "(op);\n";
           OS << "}\n\n";
         } else
           OS << "#define svreinterpret_" << From.Suffix << "_" << To.Suffix
              << "(...) __builtin_sve_reinterpret_" << From.Suffix << "_"
              << To.Suffix << "(__VA_ARGS__)\n";
       }