Fix incorrect name-mangling for llvm.experimental.vp.strided.load (#7654

)

These ops are only used for RISCV codegen at present, and this one tended to only happen for complex patterns that we don't test in our very limited crosscompilation tests.

src/CodeGen_LLVM.cpp

@@ -2439,7 +2439,7 @@ llvm::Value *CodeGen_LLVM::codegen_vector_load(const Type &type, const std::stri
             }
             if (try_vector_predication_intrinsic("llvm.experimental.vp.strided.load", VPResultType(slice_type, 0),
                                                  slice_lanes, vp_slice_mask,
-                                                 {VPArg(vec_ptr, 1, align_bytes), VPArg(stride, 1)})) {
+                                                 {VPArg(vec_ptr, 1, align_bytes), VPArg(stride, 2)})) {
                 load_inst = dyn_cast<Instruction>(value);
             } else {
                 internal_error << "Vector predicated strided load should not be requested if not supported.\n";