[clang] [clang][x86][bytecode] Refactor BMI intrinsic wrappers to use interp__builtin_elementwise_int_binop (PR #160362)

[Simon Pilgrim via cfe-commits]

================
@@ -3256,11 +3165,37 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const CallExpr *Call,
   case clang::X86::BI__builtin_ia32_bextr_u64:
   case clang::X86::BI__builtin_ia32_bextri_u32:
   case clang::X86::BI__builtin_ia32_bextri_u64:
-    return interp__builtin_ia32_bextr(S, OpPC, Frame, Call);
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call, [](const APSInt &Val, const APSInt &Idx) {
+          unsigned BitWidth = Val.getBitWidth();
+          uint64_t Shift = Idx.extractBitsAsZExtValue(8, 0);
+          uint64_t Length = Idx.extractBitsAsZExtValue(8, 8);
+          if (Length > BitWidth) {
+            Length = BitWidth;
+          }
+
+          // Handle out of bounds cases.
+          if (Length == 0 || Shift >= BitWidth)
+            return APInt(BitWidth, 0);
+
+          uint64_t Result = Val.getZExtValue() >> Shift;
+          Result &= llvm::maskTrailingOnes<uint64_t>(Length);
+          return APInt(BitWidth, Result);
+        });
 
   case clang::X86::BI__builtin_ia32_bzhi_si:
   case clang::X86::BI__builtin_ia32_bzhi_di:
-    return interp__builtin_ia32_bzhi(S, OpPC, Frame, Call);
+    return interp__builtin_elementwise_int_binop(
+        S, OpPC, Call, [](const APSInt &Val, const APSInt &Idx) {
+          unsigned BitWidth = Val.getBitWidth();
+          uint64_t Index = Idx.extractBitsAsZExtValue(8, 0);
+          APSInt Result = Val;
----------------
RKSimon wrote:

`APInt Result = Val;` to avoid any signedness troubles

https://github.com/llvm/llvm-project/pull/160362