Re: [SAMPLE][PATCH 1/3]BTI: Binary Transport Interface for SystemTap

[Masami Hiramatsu <hiramatu@sdl.hitachi.co.jp>]

Hi, Frank

Frank Ch. Eigler wrote:
> hiramatu wrote:
>>[...]  The systemtap script with BTI (*1) has about 1.4 micro secs
>>of processing time. The systemtap script with ATI (*2) has about 4
>>micro secs of processing time. [...]
> 
>>(*2)
>>probe kernel.function("sys_gettimeofday") {
>>      log(string(get_tsc()) . string(get_cpu()) . string(pid()) . string(1)=
>> .
>>string(2). string(3). string(4). string(5));
>>}
> 
> With the newer "print" statements, this expression is no longer state of the
> art.  How would this compare?

OK. I measured the script using printf(). And the processing time is
about 2.6 micro secs. So, I expects BTI still has low overhead.

>   printf("%d %d %d %d %d %d %d\n", get_tsc() /* we have that? */,
>          get_cpu(), pid(), 1, 2, 3, 4, 5);

Oh, I forgot to write these functions. They are here:

function get_tsc:long () %{
        rdtscll(THIS->__retvalue);
%}
function get_cpu:long () %{
        THIS->__retvalue = smp_processor_id();
%}


-- 
Masami HIRAMATSU
2nd Research Dept.
Hitachi, Ltd., Systems Development Laboratory
E-mail: hiramatu@sdl.hitachi.co.jp