[CIR] Backport fixsing ShuffleDynamicOp maskbits logic (#1649)

Backport the calculation of maskbits in the lowering from `N - 1` to
`NextPowerOf2(numElements - 1) - 1`, similar to Clang CG.

Backport from
[#141411](llvm/llvm-project#141411)

Author: AmrDeveloper

clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp

@@ -2021,9 +2021,12 @@ mlir::LogicalResult CIRToLLVMVecShuffleDynamicOpLowering::matchAndRewrite(
   //     __builtin_shufflevector(V, I)
   // is implemented as this pseudocode, where the for loop is unrolled
   // and N is the number of elements:
-  //     masked = I & (N-1)
-  //     for (i in 0 <= i < N)
-  //       result[i] = V[masked[i]]
+  //
+  // result = undef
+  // maskbits = NextPowerOf2(N - 1)
+  // masked = I & maskbits
+  // for (i in 0 <= i < N)
+  //    result[i] = V[masked[i]]
   auto loc = op.getLoc();
   mlir::Value input = adaptor.getVec();
   mlir::Type llvmIndexVecType =
@@ -2032,8 +2035,9 @@ mlir::LogicalResult CIRToLLVMVecShuffleDynamicOpLowering::matchAndRewrite(
       elementTypeIfVector(op.getIndices().getType()));
   uint64_t numElements =
       mlir::cast<cir::VectorType>(op.getVec().getType()).getSize();
-  mlir::Value maskValue = rewriter.create<mlir::LLVM::ConstantOp>(
-      loc, llvmIndexType, numElements - 1);
+  uint64_t maskBits = llvm::NextPowerOf2(numElements - 1) - 1;
+  mlir::Value maskValue =
+      rewriter.create<mlir::LLVM::ConstantOp>(loc, llvmIndexType, maskBits);
   mlir::Value maskVector =
       rewriter.create<mlir::LLVM::UndefOp>(loc, llvmIndexVecType);
   for (uint64_t i = 0; i < numElements; ++i) {

clang/test/CIR/CodeGen/vectype-ext.cpp

@@ -4,6 +4,7 @@
 // RUN: FileCheck --input-file=%t.ll %s -check-prefix=LLVM
 
 typedef int vi4 __attribute__((ext_vector_type(4)));
+typedef int vi6 __attribute__((ext_vector_type(6)));
 typedef unsigned int uvi4 __attribute__((ext_vector_type(4)));
 typedef int vi3 __attribute__((ext_vector_type(3)));
 typedef int vi2 __attribute__((ext_vector_type(2)));
@@ -542,7 +543,17 @@ void vector_integers_shifts_test() {
   uvi4 b = {5u, 6u, 7u, 8u};
 
   vi4 shl = a << b;
-  // CHECK:  %{{[0-9]+}} = cir.shift(left, %{{[0-9]+}} : !cir.vector<!s32i x 4>, %{{[0-9]+}} : !cir.vector<!u32i x 4>) -> !cir.vector<!s32i x 4>
+  // CIR:  %{{[0-9]+}} = cir.shift(left, %{{[0-9]+}} : !cir.vector<!s32i x 4>, %{{[0-9]+}} : !cir.vector<!u32i x 4>) -> !cir.vector<!s32i x 4>
   uvi4 shr = b >> a;
-  // CHECK: %{{[0-9]+}} = cir.shift(right, %{{[0-9]+}} : !cir.vector<!u32i x 4>, %{{[0-9]+}} : !cir.vector<!s32i x 4>) -> !cir.vector<!u32i x 4>
+  // CIR: %{{[0-9]+}} = cir.shift(right, %{{[0-9]+}} : !cir.vector<!u32i x 4>, %{{[0-9]+}} : !cir.vector<!s32i x 4>) -> !cir.vector<!u32i x 4>
+}
+
+void vector_shuffle_dynamic_mask_test() {
+  vi6 a;
+  vi6 b;
+  vi6 r = __builtin_shufflevector(a, b);
+
+  // CIR: %{{[0-9]+}} = cir.vec.shuffle.dynamic %{{[0-9]+}} : !cir.vector<!s32i x 6>, %{{[0-9]+}} : !cir.vector<!s32i x 6>
+
+  // LLVM: {{.*}} = and <6 x i32> {{.*}}, splat (i32 7)
 }