[clang] [llvm] [RISCV] Add packed narrowing convert intrinsics (PR #210389)

[Hongyu Chen via cfe-commits]

================
@@ -134,6 +134,42 @@ typedef uint32_t uint32x2_t __attribute__((__vector_size__(8)));
                                         7);                                    \
   }
 
+#if __riscv_xlen == 64
----------------
XChy wrote:

Why do we split them into RV64 and RV32. Aren't they semantically equivalent on both RV64 and RV32?

https://github.com/llvm/llvm-project/pull/210389