fix benchmark docs errors and update more instructions

benchmark/README.md

@@ -30,6 +30,8 @@ cmake -Bbuild -DLLVM_DIR=/usr/lib/llvm-15/cmake -DCMAKE_BUILD_TYPE:STRING=RelWit
 cmake --build build --config RelWithDebInfo --target install -j
 ```
 
+If you fail to build , notice LLVM version.
+
 ## build and run at a click
 
 Build the agent first. In project root:
@@ -195,7 +197,7 @@ Tested on `kernel version 6.2` and `Intel(R) Xeon(R) Gold 5418Y` CPU.
 
 ### Uprobe and read/write with `bpf_probe_write_user` and `bpf_probe_read_user`
 
-Userspace:
+Kernelspace:
 
 ```txt
 Benchmarking __bench_uprobe_uretprobe in thread 1
@@ -233,6 +235,8 @@ Benchmarking __bench_write in thread 1
 Average time usage 389.037170 ns, iter 100000 times
 ```
 
+noted that the performance of `bpf_probe_read_user` and `bpf_probe_write_user` will be influenced by the option of `ENABLE_PROBE_WRITE_USER` and `ENABLE_PROBE_READ_USER`.
+
 ### maps operations
 
 Run the map op 1000 times in one function. Userspace map op is also faster than the kernel in the current version. Current version is 10x faster than stupid old version.