[Clang] VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - add MMX/SSE/AVX/AVX512 PMULHRSW intrinsics to be used in constexpr

clang/include/clang/Basic/BuiltinsX86.td

@@ -124,13 +124,13 @@ let Attributes = [Const, NoThrow, RequiredVectorWidth<128>] in {
   }
 
   let Features = "ssse3" in {
-    def pmulhrsw128 : X86Builtin<"_Vector<8, short>(_Vector<8, short>, _Vector<8, short>)">;
     def psignb128 : X86Builtin<"_Vector<16, char>(_Vector<16, char>, _Vector<16, char>)">;
     def psignw128 : X86Builtin<"_Vector<8, short>(_Vector<8, short>, _Vector<8, short>)">;
     def psignd128 : X86Builtin<"_Vector<4, int>(_Vector<4, int>, _Vector<4, int>)">;
   }
 
   let Features = "ssse3", Attributes = [NoThrow, Const, Constexpr, RequiredVectorWidth<128>] in {
+    def pmulhrsw128 : X86Builtin<"_Vector<8, short>(_Vector<8, short>, _Vector<8, short>)">;
     def pmaddubsw128 : X86Builtin<"_Vector<8, short>(_Vector<16, char>, _Vector<16, char>)">;
     def pshufb128 : X86Builtin<"_Vector<16, char>(_Vector<16, char>, _Vector<16, char>)">;
   }
@@ -608,7 +608,6 @@ let Features = "avx2", Attributes = [NoThrow, Const, RequiredVectorWidth<256>] i
   def palignr256 : X86Builtin<"_Vector<32, char>(_Vector<32, char>, _Vector<32, char>, _Constant int)">;
 
   def pmovmskb256 : X86Builtin<"int(_Vector<32, char>)">;
-  def pmulhrsw256 : X86Builtin<"_Vector<16, short>(_Vector<16, short>, _Vector<16, short>)">;
   def psadbw256 : X86Builtin<"_Vector<4, long long int>(_Vector<32, char>, _Vector<32, char>)">;
   def psignb256 : X86Builtin<"_Vector<32, char>(_Vector<32, char>, _Vector<32, char>)">;
   def psignw256 : X86Builtin<"_Vector<16, short>(_Vector<16, short>, _Vector<16, short>)">;
@@ -661,6 +660,7 @@ let Features = "avx2", Attributes = [NoThrow, Const, Constexpr, RequiredVectorWi
   def psrawi256 : X86Builtin<"_Vector<16, short>(_Vector<16, short>, int)">;
   def psradi256 : X86Builtin<"_Vector<8, int>(_Vector<8, int>, int)">;
 
+  def pmulhrsw256 : X86Builtin<"_Vector<16, short>(_Vector<16, short>, _Vector<16, short>)">;
   def pmulhuw256 : X86Builtin<"_Vector<16, unsigned short>(_Vector<16, unsigned short>, _Vector<16, unsigned short>)">;
   def pmulhw256 : X86Builtin<"_Vector<16, short>(_Vector<16, short>, _Vector<16, short>)">;
 
@@ -1386,13 +1386,10 @@ let Features = "avx512bitalg", Attributes = [NoThrow, Const, RequiredVectorWidth
   def vpshufbitqmb512_mask : X86Builtin<"unsigned long long int(_Vector<64, char>, _Vector<64, char>, unsigned long long int)">;
 }
 
-let Features = "avx512bw", Attributes = [NoThrow, Const, RequiredVectorWidth<512>] in {
-  def pmulhrsw512 : X86Builtin<"_Vector<32, short>(_Vector<32, short>, _Vector<32, short>)">;
-}
-
 let Features = "avx512bw", Attributes = [NoThrow, Const, Constexpr, RequiredVectorWidth<512>] in {
   def pavgb512 : X86Builtin<"_Vector<64, unsigned char>(_Vector<64, unsigned char>, _Vector<64, unsigned char>)">;
   def pavgw512 : X86Builtin<"_Vector<32, unsigned short>(_Vector<32, unsigned short>, _Vector<32, unsigned short>)">;
+  def pmulhrsw512 : X86Builtin<"_Vector<32, short>(_Vector<32, short>, _Vector<32, short>)">;
   def pmulhuw512 : X86Builtin<"_Vector<32, unsigned short>(_Vector<32, unsigned short>, _Vector<32, unsigned short>)">;
   def pmulhw512 : X86Builtin<"_Vector<32, short>(_Vector<32, short>, _Vector<32, short>)">;
 }

clang/lib/AST/ByteCode/InterpBuiltin.cpp

@@ -3621,6 +3621,15 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const CallExpr *Call,
           return LHS.isSigned() ? LHS.ssub_sat(RHS) : LHS.usub_sat(RHS);
         });
 
+  case clang::X86::BI__builtin_ia32_pmulhrsw128:
+  case clang::X86::BI__builtin_ia32_pmulhrsw256:
+  case clang::X86::BI__builtin_ia32_pmulhrsw512:
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call, [](const APSInt &LHS, const APSInt &RHS) {
+          return (llvm::APIntOps::mulsExtended(LHS, RHS).ashr(14) + 1)
+              .extractBits(16, 1);
+        });
+
   case clang::X86::BI__builtin_ia32_pavgb128:
   case clang::X86::BI__builtin_ia32_pavgw128:
   case clang::X86::BI__builtin_ia32_pavgb256:

clang/lib/AST/ExprConstant.cpp

@@ -11819,6 +11819,14 @@ bool VectorExprEvaluator::VisitCallExpr(const CallExpr *E) {
   case clang::X86::BI__builtin_ia32_pavgw512:
     return EvaluateBinOpExpr(llvm::APIntOps::avgCeilU);
 
+  case clang::X86::BI__builtin_ia32_pmulhrsw128:
+  case clang::X86::BI__builtin_ia32_pmulhrsw256:
+  case clang::X86::BI__builtin_ia32_pmulhrsw512:
+    return EvaluateBinOpExpr([](const APSInt &LHS, const APSInt &RHS) {
+      return (llvm::APIntOps::mulsExtended(LHS, RHS).ashr(14) + 1)
+          .extractBits(16, 1);
+    });
+
   case clang::X86::BI__builtin_ia32_pmaddubsw128:
   case clang::X86::BI__builtin_ia32_pmaddubsw256:
   case clang::X86::BI__builtin_ia32_pmaddubsw512:

clang/lib/Headers/avx2intrin.h

@@ -1650,9 +1650,8 @@ _mm256_mul_epi32(__m256i __a, __m256i __b) {
 /// \param __b
 ///    A 256-bit vector of [16 x i16] containing one of the source operands.
 /// \returns A 256-bit vector of [16 x i16] containing the rounded products.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_mulhrs_epi16(__m256i __a, __m256i __b)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_mulhrs_epi16(__m256i __a, __m256i __b) {
   return (__m256i)__builtin_ia32_pmulhrsw256((__v16hi)__a, (__v16hi)__b);
 }
 
@@ -1670,8 +1669,7 @@ _mm256_mulhrs_epi16(__m256i __a, __m256i __b)
 ///    A 256-bit vector of [16 x i16] containing one of the source operands.
 /// \returns A 256-bit vector of [16 x i16] containing the products.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_mulhi_epu16(__m256i __a, __m256i __b)
-{
+_mm256_mulhi_epu16(__m256i __a, __m256i __b) {
   return (__m256i)__builtin_ia32_pmulhuw256((__v16hu)__a, (__v16hu)__b);
 }
 

clang/lib/Headers/avx512bwintrin.h

@@ -1003,23 +1003,20 @@ _mm512_maskz_permutex2var_epi16(__mmask32 __U, __m512i __A, __m512i __I,
                               (__v32hi)_mm512_setzero_si512());
 }
 
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mulhrs_epi16(__m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mulhrs_epi16(__m512i __A, __m512i __B) {
   return (__m512i)__builtin_ia32_pmulhrsw512((__v32hi)__A, (__v32hi)__B);
 }
 
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_mask_mulhrs_epi16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_mulhrs_epi16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
   return (__m512i)__builtin_ia32_selectw_512((__mmask32)__U,
                                          (__v32hi)_mm512_mulhrs_epi16(__A, __B),
                                          (__v32hi)__W);
 }
 
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_maskz_mulhrs_epi16(__mmask32 __U, __m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_maskz_mulhrs_epi16(__mmask32 __U, __m512i __A, __m512i __B) {
   return (__m512i)__builtin_ia32_selectw_512((__mmask32)__U,
                                          (__v32hi)_mm512_mulhrs_epi16(__A, __B),
                                          (__v32hi)_mm512_setzero_si512());

clang/lib/Headers/avx512vlbwintrin.h

@@ -1510,28 +1510,28 @@ _mm256_mask_cvtusepi16_storeu_epi8 (void * __P, __mmask16 __M, __m256i __A)
   __builtin_ia32_pmovuswb256mem_mask ((__v16qi*) __P, (__v16hi) __A, __M);
 }
 
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
 _mm_mask_mulhrs_epi16(__m128i __W, __mmask8 __U, __m128i __X, __m128i __Y) {
   return (__m128i)__builtin_ia32_selectw_128((__mmask8)__U,
                                              (__v8hi)_mm_mulhrs_epi16(__X, __Y),
                                              (__v8hi)__W);
 }
 
-static __inline__ __m128i __DEFAULT_FN_ATTRS128
+static __inline__ __m128i __DEFAULT_FN_ATTRS128_CONSTEXPR
 _mm_maskz_mulhrs_epi16(__mmask8 __U, __m128i __X, __m128i __Y) {
   return (__m128i)__builtin_ia32_selectw_128((__mmask8)__U,
                                              (__v8hi)_mm_mulhrs_epi16(__X, __Y),
                                              (__v8hi)_mm_setzero_si128());
 }
 
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_mask_mulhrs_epi16(__m256i __W, __mmask16 __U, __m256i __X, __m256i __Y) {
   return (__m256i)__builtin_ia32_selectw_256((__mmask16)__U,
                                          (__v16hi)_mm256_mulhrs_epi16(__X, __Y),
                                          (__v16hi)__W);
 }
 
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_maskz_mulhrs_epi16(__mmask16 __U, __m256i __X, __m256i __Y) {
   return (__m256i)__builtin_ia32_selectw_256((__mmask16)__U,
                                          (__v16hi)_mm256_mulhrs_epi16(__X, __Y),

clang/lib/Headers/tmmintrin.h

@@ -544,7 +544,7 @@ _mm_maddubs_pi16(__m64 __a, __m64 __b) {
 ///    A 128-bit vector of [8 x i16] containing one of the source operands.
 /// \returns A 128-bit vector of [8 x i16] containing the rounded and scaled
 ///    products of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mulhrs_epi16(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_mulhrs_epi16(__m128i __a,
                                                               __m128i __b) {
   return (__m128i)__builtin_ia32_pmulhrsw128((__v8hi)__a, (__v8hi)__b);
 }
@@ -563,11 +563,10 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mulhrs_epi16(__m128i __a,
 ///    A 64-bit vector of [4 x i16] containing one of the source operands.
 /// \returns A 64-bit vector of [4 x i16] containing the rounded and scaled
 ///    products of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_mulhrs_pi16(__m64 __a, __m64 __b)
-{
-    return __trunc64(__builtin_ia32_pmulhrsw128((__v8hi)__anyext128(__a),
-                                                (__v8hi)__anyext128(__b)));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_mulhrs_pi16(__m64 __a, __m64 __b) {
+  return __trunc64(__builtin_ia32_pmulhrsw128((__v8hi)__zext128(__a),
+                                              (__v8hi)__zext128(__b)));
 }
 
 /// Copies the 8-bit integers from a 128-bit integer vector to the

clang/test/CodeGen/X86/avx2-builtins.c

@@ -1038,6 +1038,7 @@ __m256i test_mm256_mulhrs_epi16(__m256i a, __m256i b) {
   // CHECK: call <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16> %{{.*}}, <16 x i16> %{{.*}})
   return _mm256_mulhrs_epi16(a, b);
 }
+TEST_CONSTEXPR(match_v16hi(_mm256_mulhrs_epi16((__m256i)(__v16hi){+100, +200, -300, -400, +500, +600, -700, +800, -900, -1000, +1100, +1200, -1300, -1400, +1500, +1600}, (__m256i)(__v16hi){+1600, -1500, +1400, -1300, +1200, -1100, +1000, -900, +800, -700, +600, -500, +400, -300, +200, -100}), +5, -9, -13, +16, +18, -20, -21, -22, -22, +21, +20, -18, -16, +13, +9, -5));
 
 __m256i test_mm256_mullo_epi16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_mullo_epi16

clang/test/CodeGen/X86/avx512bw-builtins.c

@@ -1596,18 +1596,24 @@ __m512i test_mm512_mulhrs_epi16(__m512i __A, __m512i __B) {
   // CHECK: @llvm.x86.avx512.pmul.hr.sw.512
   return _mm512_mulhrs_epi16(__A,__B); 
 }
+TEST_CONSTEXPR(match_v32hi(_mm512_mulhrs_epi16((__m512i)(__v32hi){+100, +200, -300, -400, +500, +600, -700, +800, -900, -1000, +1100, +1200, -1300, -1400, +1500, +1600, -1700, -1800, +1900, +2000, -2100, -2200, +2300, +2400, -2500, -2600, +2700, +2800, -2900, -3000, +3100, +3200}, (__m512i)(__v32hi){+3200, -3100, +3000, -2900, +2800, -2700, +2600, -2500, +2400, -2300, +2200, -2100, +2000, -1900, +1800, -1700, +1600, -1500, +1400, -1300, +1200, -1100, +1000, -900, +800, -700, +600, -500, +400, -300, +200, -100}), +10, -19, -27, +35, +43, -49, -56, -61, -66, +70, +74, -77, -79, +81, +82, -83, -83, +82, +81, -79, -77, +74, +70, -66, -61, +56, +49, -43, -35, +27, +19, -10));
+
 __m512i test_mm512_mask_mulhrs_epi16(__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: test_mm512_mask_mulhrs_epi16
   // CHECK: @llvm.x86.avx512.pmul.hr.sw.512
   // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
   return _mm512_mask_mulhrs_epi16(__W,__U,__A,__B); 
 }
+TEST_CONSTEXPR(match_v32hi(_mm512_mask_mulhrs_epi16(_mm512_set1_epi16(1), 0x0000FFFF, (__m512i)(__v32hi){+100, +200, -300, -400, +500, +600, -700, +800, -900, -1000, +1100, +1200, -1300, -1400, +1500, +1600, -1700, -1800, +1900, +2000, -2100, -2200, +2300, +2400, -2500, -2600, +2700, +2800, -2900, -3000, +3100, +3200}, (__m512i)(__v32hi){+3200, -3100, +3000, -2900, +2800, -2700, +2600, -2500, +2400, -2300, +2200, -2100, +2000, -1900, +1800, -1700, +1600, -1500, +1400, -1300, +1200, -1100, +1000, -900, +800, -700, +600, -500, +400, -300, +200, -100}), +10, -19, -27, +35, +43, -49, -56, -61, -66, +70, +74, -77, -79, +81, +82, -83, +1, +1, +1, +1, +1, +1, +1, +1, +1, +1, +1, +1, +1, +1, +1, +1));
+
 __m512i test_mm512_maskz_mulhrs_epi16(__mmask32 __U, __m512i __A, __m512i __B) {
   // CHECK-LABEL: test_mm512_maskz_mulhrs_epi16
   // CHECK: @llvm.x86.avx512.pmul.hr.sw.512
   // CHECK: select <32 x i1> %{{.*}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}
   return _mm512_maskz_mulhrs_epi16(__U,__A,__B); 
 }
+TEST_CONSTEXPR(match_v32hi(_mm512_maskz_mulhrs_epi16(0x0000FFFF, (__m512i)(__v32hi){+100, +200, -300, -400, +500, +600, -700, +800, -900, -1000, +1100, +1200, -1300, -1400, +1500, +1600, -1700, -1800, +1900, +2000, -2100, -2200, +2300, +2400, -2500, -2600, +2700, +2800, -2900, -3000, +3100, +3200}, (__m512i)(__v32hi){+3200, -3100, +3000, -2900, +2800, -2700, +2600, -2500, +2400, -2300, +2200, -2100, +2000, -1900, +1800, -1700, +1600, -1500, +1400, -1300, +1200, -1100, +1000, -900, +800, -700, +600, -500, +400, -300, +200, -100}), +10, -19, -27, +35, +43, -49, -56, -61, -66, +70, +74, -77, -79, +81, +82, -83, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0));
+
 __m512i test_mm512_mulhi_epi16(__m512i __A, __m512i __B) {
   // CHECK-LABEL: test_mm512_mulhi_epi16
   // CHECK: @llvm.x86.avx512.pmulh.w.512

clang/test/CodeGen/X86/avx512vlbw-builtins.c

@@ -2061,27 +2061,31 @@ __m128i test_mm_mask_mulhrs_epi16(__m128i __W, __mmask8 __U, __m128i __X, __m128
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_mask_mulhrs_epi16(__W, __U, __X, __Y); 
 }
+TEST_CONSTEXPR(match_v8hi(_mm_mask_mulhrs_epi16(_mm_set1_epi16(1), 0x0F, (__m128i)(__v8hi){+100, +200, -300, -400, +500, +600, -700, +800}, (__m128i)(__v8hi){+8000, -7000, +6000, -5000, +4000, -3000, +2000, -1000}), +24, -43, -55, +61, +1, +1, +1, +1));
 
 __m128i test_mm_maskz_mulhrs_epi16(__mmask8 __U, __m128i __X, __m128i __Y) {
   // CHECK-LABEL: test_mm_maskz_mulhrs_epi16
   // CHECK: @llvm.x86.ssse3.pmul.hr.sw
   // CHECK: select <8 x i1> %{{.*}}, <8 x i16> %{{.*}}, <8 x i16> %{{.*}}
   return _mm_maskz_mulhrs_epi16(__U, __X, __Y); 
 }
+TEST_CONSTEXPR(match_v8hi(_mm_maskz_mulhrs_epi16(0x0F, (__m128i)(__v8hi){+100, +200, -300, -400, +500, +600, -700, +800}, (__m128i)(__v8hi){+8000, -7000, +6000, -5000, +4000, -3000, +2000, -1000}), +24, -43, -55, +61, 0, 0, 0, 0));
 
 __m256i test_mm256_mask_mulhrs_epi16(__m256i __W, __mmask16 __U, __m256i __X, __m256i __Y) {
   // CHECK-LABEL: test_mm256_mask_mulhrs_epi16
   // CHECK: @llvm.x86.avx2.pmul.hr.sw
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_mask_mulhrs_epi16(__W, __U, __X, __Y); 
 }
+TEST_CONSTEXPR(match_v16hi(_mm256_mask_mulhrs_epi16(_mm256_set1_epi16(1), 0xF00F, (__m256i)(__v16hi){+100, +200, -300, -400, +500, +600, -700, +800, -900, -1000, +1100, +1200, -1300, -1400, +1500, +1600}, (__m256i)(__v16hi){+1600, -1500, +1400, -1300, +1200, -1100, +1000, -900, +800, -700, +600, -500, +400, -300, +200, -100}), +5, -9, -13, +16, +1, +1, +1, +1, +1, +1, +1, +1, -16, +13, +9, -5));
 
 __m256i test_mm256_maskz_mulhrs_epi16(__mmask16 __U, __m256i __X, __m256i __Y) {
   // CHECK-LABEL: test_mm256_maskz_mulhrs_epi16
   // CHECK: @llvm.x86.avx2.pmul.hr.sw
   // CHECK: select <16 x i1> %{{.*}}, <16 x i16> %{{.*}}, <16 x i16> %{{.*}}
   return _mm256_maskz_mulhrs_epi16(__U, __X, __Y); 
 }
+TEST_CONSTEXPR(match_v16hi(_mm256_maskz_mulhrs_epi16(0xF00F, (__m256i)(__v16hi){+100, +200, -300, -400, +500, +600, -700, +800, -900, -1000, +1100, +1200, -1300, -1400, +1500, +1600}, (__m256i)(__v16hi){+1600, -1500, +1400, -1300, +1200, -1100, +1000, -900, +800, -700, +600, -500, +400, -300, +200, -100}), +5, -9, -13, +16, 0, 0, 0, 0, 0, 0, 0, 0, -16, +13, +9, -5));
 
 __m128i test_mm_mask_mulhi_epu16(__m128i __W, __mmask8 __U, __m128i __A, __m128i __B) {
   // CHECK-LABEL: test_mm_mask_mulhi_epu16

clang/test/CodeGen/X86/mmx-builtins.c

@@ -438,6 +438,7 @@ __m64 test_mm_mulhrs_pi16(__m64 a, __m64 b) {
   // CHECK: call <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(
   return _mm_mulhrs_pi16(a, b);
 }
+TEST_CONSTEXPR(match_v4hi(_mm_mulhrs_pi16((__m64)(__v4hi){+100, +200, -300, -400}, (__m64)(__v4hi){+30000, -20000, +10000, -5000}), +92, -122, -92, +61));
 
 __m64 test_mm_mullo_pi16(__m64 a, __m64 b) {
   // CHECK-LABEL: test_mm_mullo_pi16

clang/test/CodeGen/X86/ssse3-builtins.c

@@ -110,6 +110,7 @@ __m128i test_mm_mulhrs_epi16(__m128i a, __m128i b) {
   // CHECK: call <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16> %{{.*}}, <8 x i16> %{{.*}})
   return _mm_mulhrs_epi16(a, b);
 }
+TEST_CONSTEXPR(match_v8hi(_mm_mulhrs_epi16((__m128i)(__v8hi){+100, +200, -300, -400, +500, +600, -700, +800}, (__m128i)(__v8hi){+8000, -7000, +6000, -5000, +4000, -3000, +2000, -1000}), +24, -43, -55, +61, +61, -55, -43, -24));
 
 __m128i test_mm_shuffle_epi8(__m128i a, __m128i b) {
   // CHECK-LABEL: test_mm_shuffle_epi8