[X86] Replace avx512f integer mul reduction builtins with generic builtin

[RKSimon]

Summary

D117829 added the generic "__builtin_reduce_mul" which we can use to replace the x86 specific integer mul reduction builtins - internally these were mapping to the same intrinsic already so there are no test changes required.

[pengfei]

LGTM.

clang/include/clang/Basic/BuiltinsX86.def

 TARGET_BUILTIN(__builtin_ia32_reduce_fmin_ph512, "xV32x", "ncV:512:", "avx512fp16")
 TARGET_BUILTIN(__builtin_ia32_reduce_fmin_ph256, "xV16x", "ncV:256:", "avx512fp16,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_reduce_fmin_ph128, "xV8x", "ncV:128:", "avx512fp16,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_reduce_fmul_pd512, "ddV8d", "ncV:512:", "avx512f")
 TARGET_BUILTIN(__builtin_ia32_reduce_fmul_ps512, "ffV16f", "ncV:512:", "avx512f")
 TARGET_BUILTIN(__builtin_ia32_reduce_fmul_ph512, "xxV32x", "ncV:512:", "avx512fp16")
 TARGET_BUILTIN(__builtin_ia32_reduce_fmul_ph256, "xxV16x", "ncV:256:", "avx512fp16,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_reduce_fmul_ph128, "xxV8x", "ncV:128:", "avx512fp16,avx512vl")
-TARGET_BUILTIN(__builtin_ia32_reduce_mul_d512, "iV16i", "ncV:512:", "avx512f")
-TARGET_BUILTIN(__builtin_ia32_reduce_mul_q512, "OiV8Oi", "ncV:512:", "avx512f")
 
 // MONITORX/MWAITX
 TARGET_BUILTIN(__builtin_ia32_monitorx, "vvC*UiUi", "n", "mwaitx")
 TARGET_BUILTIN(__builtin_ia32_mwaitx, "vUiUiUi", "n", "mwaitx")
 
 // WAITPKG
 TARGET_BUILTIN(__builtin_ia32_umonitor, "vvC*", "n", "waitpkg")
 TARGET_BUILTIN(__builtin_ia32_umwait, "UcUiUiUi", "n", "waitpkg")

clang/lib/CodeGen/CGBuiltin.cpp

   case X86::BI__builtin_ia32_reduce_fmin_ph512:
   case X86::BI__builtin_ia32_reduce_fmin_ph256:
   case X86::BI__builtin_ia32_reduce_fmin_ph128: {
     Function *F =
         CGM.getIntrinsic(Intrinsic::vector_reduce_fmin, Ops[0]->getType());
     Builder.getFastMathFlags().setNoNaNs();
     return Builder.CreateCall(F, {Ops[0]});
   }
-  case X86::BI__builtin_ia32_reduce_mul_d512:
-  case X86::BI__builtin_ia32_reduce_mul_q512: {
-    Function *F =
-        CGM.getIntrinsic(Intrinsic::vector_reduce_mul, Ops[0]->getType());
-    return Builder.CreateCall(F, {Ops[0]});
-  }
 
   // 3DNow!
   case X86::BI__builtin_ia32_pswapdsf:
   case X86::BI__builtin_ia32_pswapdsi: {
     llvm::Type *MMXTy = llvm::Type::getX86_MMXTy(getLLVMContext());
     Ops[0] = Builder.CreateBitCast(Ops[0], MMXTy, "cast");
     llvm::Function *F = CGM.getIntrinsic(Intrinsic::x86_3dnowa_pswapd);
     return Builder.CreateCall(F, Ops, "pswapd");

clang/lib/Headers/avx512fintrin.h

  * This takes log2(n) steps where n is the number of elements in the vector.
  */
 
 static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_add_epi64(__m512i __W) {
   return __builtin_reduce_add((__v8di)__W);
 }
 
 static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_mul_epi64(__m512i __W) {
-  return __builtin_ia32_reduce_mul_q512(__W);
+  return __builtin_reduce_mul((__v8di)__W);
 }
 
 static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_and_epi64(__m512i __W) {
   return __builtin_reduce_and((__v8di)__W);
 }
 
 static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_or_epi64(__m512i __W) {
   return __builtin_reduce_or((__v8di)__W);
 }
 
 static __inline__ long long __DEFAULT_FN_ATTRS512
 _mm512_mask_reduce_add_epi64(__mmask8 __M, __m512i __W) {
   __W = _mm512_maskz_mov_epi64(__M, __W);
   return __builtin_reduce_add((__v8di)__W);
 }
 
 static __inline__ long long __DEFAULT_FN_ATTRS512
 _mm512_mask_reduce_mul_epi64(__mmask8 __M, __m512i __W) {
   __W = _mm512_mask_mov_epi64(_mm512_set1_epi64(1), __M, __W);
-  return __builtin_ia32_reduce_mul_q512(__W);
+  return __builtin_reduce_mul((__v8di)__W);
 }
 
 static __inline__ long long __DEFAULT_FN_ATTRS512
 _mm512_mask_reduce_and_epi64(__mmask8 __M, __m512i __W) {
   __W = _mm512_mask_mov_epi64(_mm512_set1_epi64(-1LL), __M, __W);
   return __builtin_reduce_and((__v8di)__W);
 }
 

 
 static __inline__ int __DEFAULT_FN_ATTRS512
 _mm512_reduce_add_epi32(__m512i __W) {
   return __builtin_reduce_add((__v16si)__W);
 }
 
 static __inline__ int __DEFAULT_FN_ATTRS512
 _mm512_reduce_mul_epi32(__m512i __W) {
-  return __builtin_ia32_reduce_mul_d512((__v16si)__W);
+  return __builtin_reduce_mul((__v16si)__W);
 }
 
 static __inline__ int __DEFAULT_FN_ATTRS512
 _mm512_reduce_and_epi32(__m512i __W) {
   return __builtin_reduce_and((__v16si)__W);
 }
 
 static __inline__ int __DEFAULT_FN_ATTRS512
 _mm512_reduce_or_epi32(__m512i __W) {
   return __builtin_reduce_or((__v16si)__W);
 }
 
 static __inline__ int __DEFAULT_FN_ATTRS512
 _mm512_mask_reduce_add_epi32( __mmask16 __M, __m512i __W) {
   __W = _mm512_maskz_mov_epi32(__M, __W);
   return __builtin_reduce_add((__v16si)__W);
 }
 
 static __inline__ int __DEFAULT_FN_ATTRS512
 _mm512_mask_reduce_mul_epi32( __mmask16 __M, __m512i __W) {
   __W = _mm512_mask_mov_epi32(_mm512_set1_epi32(1), __M, __W);
-  return __builtin_ia32_reduce_mul_d512((__v16si)__W);
+  return __builtin_reduce_mul((__v16si)__W);
 }
 
 static __inline__ int __DEFAULT_FN_ATTRS512
 _mm512_mask_reduce_and_epi32( __mmask16 __M, __m512i __W) {
   __W = _mm512_mask_mov_epi32(_mm512_set1_epi32(-1), __M, __W);
   return __builtin_reduce_and((__v16si)__W);
 }