[clang][x86] Add constexpr support for BZHI intrinsics (#110508)

Author: RKSimon

clang/docs/ReleaseNotes.rst

@@ -515,6 +515,7 @@ X86 Support
   all lzcnt intrinsics in lzcntintrin.h 
   all bextr intrinsics in bmiintrin.h
   all tzcnt intrinsics in bmiintrin.h
+  all bzhi intrinsics in bmi2intrin.h
   all bextr intrinsics in tbmintrin.h
 
 Arm and AArch64 Support

clang/include/clang/Basic/BuiltinsX86.def

@@ -560,7 +560,7 @@ TARGET_BUILTIN(__builtin_ia32_tzcnt_u16, "UsUs", "ncE", "")
 TARGET_BUILTIN(__builtin_ia32_tzcnt_u32, "UiUi", "ncE", "")
 
 // BMI2
-TARGET_BUILTIN(__builtin_ia32_bzhi_si, "UiUiUi", "nc", "bmi2")
+TARGET_BUILTIN(__builtin_ia32_bzhi_si, "UiUiUi", "ncE", "bmi2")
 TARGET_BUILTIN(__builtin_ia32_pdep_si, "UiUiUi", "nc", "bmi2")
 TARGET_BUILTIN(__builtin_ia32_pext_si, "UiUiUi", "nc", "bmi2")
 

clang/include/clang/Basic/BuiltinsX86_64.def

@@ -73,7 +73,7 @@ TARGET_BUILTIN(__builtin_ia32_rdseed64_step, "UiUOi*", "n", "rdseed")
 TARGET_BUILTIN(__builtin_ia32_lzcnt_u64, "UOiUOi", "ncE", "lzcnt")
 TARGET_BUILTIN(__builtin_ia32_bextr_u64, "UOiUOiUOi", "ncE", "bmi")
 TARGET_BUILTIN(__builtin_ia32_tzcnt_u64, "UOiUOi", "ncE", "")
-TARGET_BUILTIN(__builtin_ia32_bzhi_di, "UOiUOiUOi", "nc", "bmi2")
+TARGET_BUILTIN(__builtin_ia32_bzhi_di, "UOiUOiUOi", "ncE", "bmi2")
 TARGET_BUILTIN(__builtin_ia32_pdep_di, "UOiUOiUOi", "nc", "bmi2")
 TARGET_BUILTIN(__builtin_ia32_pext_di, "UOiUOiUOi", "nc", "bmi2")
 TARGET_BUILTIN(__builtin_ia32_bextri_u64, "UOiUOiIUOi", "ncE", "tbm")

clang/lib/AST/ExprConstant.cpp

@@ -13487,6 +13487,20 @@ bool IntExprEvaluator::VisitBuiltinCallExpr(const CallExpr *E,
     return Success(Result, E);
   }
 
+  case clang::X86::BI__builtin_ia32_bzhi_si:
+  case clang::X86::BI__builtin_ia32_bzhi_di: {
+    APSInt Val, Idx;
+    if (!EvaluateInteger(E->getArg(0), Val, Info) ||
+        !EvaluateInteger(E->getArg(1), Idx, Info))
+      return false;
+
+    unsigned BitWidth = Val.getBitWidth();
+    unsigned Index = Idx.extractBitsAsZExtValue(8, 0);
+    if (Index < BitWidth)
+      Val.clearHighBits(BitWidth - Index);
+    return Success(Val, E);
+  }
+
   case clang::X86::BI__builtin_ia32_lzcnt_u16:
   case clang::X86::BI__builtin_ia32_lzcnt_u32:
   case clang::X86::BI__builtin_ia32_lzcnt_u64: {

clang/lib/Headers/bmi2intrin.h

@@ -17,6 +17,12 @@
 /* Define the default attributes for the functions in this file. */
 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("bmi2")))
 
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS constexpr
+#else
+#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS
+#endif
+
 /// Copies the unsigned 32-bit integer \a __X and zeroes the upper bits
 ///    starting at bit number \a __Y.
 ///
@@ -37,7 +43,7 @@
 /// \param __Y
 ///    The lower 8 bits specify the bit number of the lowest bit to zero.
 /// \returns The partially zeroed 32-bit value.
-static __inline__ unsigned int __DEFAULT_FN_ATTRS
+static __inline__ unsigned int __DEFAULT_FN_ATTRS_CONSTEXPR
 _bzhi_u32(unsigned int __X, unsigned int __Y)
 {
   return __builtin_ia32_bzhi_si(__X, __Y);
@@ -153,7 +159,7 @@ _mulx_u32(unsigned int __X, unsigned int __Y, unsigned int *__P)
 /// \param __Y
 ///    The lower 8 bits specify the bit number of the lowest bit to zero.
 /// \returns The partially zeroed 64-bit value.
-static __inline__ unsigned long long __DEFAULT_FN_ATTRS
+static __inline__ unsigned long long __DEFAULT_FN_ATTRS_CONSTEXPR
 _bzhi_u64(unsigned long long __X, unsigned long long __Y)
 {
   return __builtin_ia32_bzhi_di(__X, __Y);

clang/test/CodeGen/X86/bmi2-builtins.c

@@ -1,5 +1,7 @@
-// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s
-// RUN: %clang_cc1 -ffreestanding %s -triple=i386-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s --check-prefix=B32
+// RUN: %clang_cc1 -x c -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 -x c -ffreestanding %s -triple=i386-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s --check-prefix=B32
+// RUN: %clang_cc1 -x c++ -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s
+// RUN: %clang_cc1 -x c++ -ffreestanding %s -triple=i386-apple-darwin -target-feature +bmi2 -emit-llvm -o - | FileCheck %s --check-prefix=B32
 
 
 #include <immintrin.h>
@@ -22,7 +24,6 @@ unsigned int test_pext_u32(unsigned int __X, unsigned int __Y) {
 #ifdef __i386__
 unsigned int test_mulx_u32(unsigned int __X, unsigned int __Y,
                                  unsigned int *__P) {
-  // B32: @test_mulx_u32
   // B32: mul i64
   return _mulx_u32(__X, __Y, __P);
 }
@@ -46,8 +47,24 @@ unsigned long long test_pext_u64(unsigned long long __X, unsigned long long __Y)
 
 unsigned long long test_mulx_u64(unsigned long long __X, unsigned long long __Y,
                                  unsigned long long *__P) {
-  // CHECK: @test_mulx_u64
   // CHECK: mul i128
   return _mulx_u64(__X, __Y, __P);
 }
 #endif
+
+// Test constexpr handling.
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+char bzhi32_0[_bzhi_u32(0x89ABCDEF,   0) == 0x00000000 ? 1 : -1];
+char bzhi32_1[_bzhi_u32(0x89ABCDEF,  16) == 0x0000CDEF ? 1 : -1];
+char bzhi32_2[_bzhi_u32(0x89ABCDEF,  31) == 0x09ABCDEF ? 1 : -1];
+char bzhi32_3[_bzhi_u32(0x89ABCDEF,  32) == 0x89ABCDEF ? 1 : -1];
+char bzhi32_4[_bzhi_u32(0x89ABCDEF,  99) == 0x89ABCDEF ? 1 : -1];
+char bzhi32_5[_bzhi_u32(0x89ABCDEF, 260) == 0x0000000F ? 1 : -1];
+
+#ifdef __x86_64__
+char bzhi64_0[_bzhi_u64(0x0123456789ABCDEFULL,   0) == 0x0000000000000000ULL ? 1 : -1];
+char bzhi64_1[_bzhi_u64(0x0123456789ABCDEFULL,  32) == 0x0000000089ABCDEFULL ? 1 : -1];
+char bzhi64_2[_bzhi_u64(0x0123456789ABCDEFULL,  99) == 0x0123456789ABCDEFULL ? 1 : -1];
+char bzhi64_3[_bzhi_u64(0x0123456789ABCDEFULL, 520) == 0x00000000000000EFULL ? 1 : -1];
+#endif
+#endif