As I was attempting to get better performance out of my PIN program, I decided to measure the base performance of PIN for a simple application (inscount1,...
Hi Marco, Do you have a test case you can send us? What version of Pin are you using (kit number, target, compiler version)? Is there any output message...
Hi Ravi, This happens because we only call the ThreadBegin call-back when the application creates a new thread (e.g. via the _clone system call). The ...
I think I get your point of the context/checkpoint difference, but still have no clue for the Pin_Resume hang. Actually, I write my pintool based on the...
You can use conditional inlining for something like inscount, so that the small counting routine can be inlined and the bigger analysis routine can be called...
What do you mean by program hang? Is it executing instructions but not making apparent forward progress? Perhaps you could change your pintool so print the ip...
... It's an unavoidable fact of life with binary instrumentation: if you want to do anything complicated you have to use function calls, and function calls...
Nicholas Nethercote
njn@...
Jun 2, 2006 10:24 pm
861
I did. No more execution is executed. Not only the instruction from original application, but also that from the instrumented code. No more progress. It is...
Hi, I want to measure my pintool's overhead. It seems that pintool always have large 'start-up' time due to its 'dynamic' instrumentation. Is ther e some way...
Try using the gdb attach command to see what it is doing. You can get the pc this way. Then look at /proc/<pid>/maps to see which file the current pc is in. If...
Can't seem to download Pin from the downloads page at http://rogue.colorado.edu/Pin/downloads.php. On clicking any of the download links I get : Top Downloads...
Actually, image-granularity instrumentation won't do what you expect. All instrumentation (even image-granularity) is delayed until the code is about to be...
Shan, here is another approach that may be more suitable for your timing, it relies on using two new callbacks: (1) CODECACHE_AddCodeCacheEnteredFunction and...
... --snip--- ... that the ... routine can ... 4-5X for ... But if you are JUST running inscount, PIN should be inlining it automatically. I do see how...
Hi group, I was trying to instrument MOV instructions for IA-32 using INS_IsMov () and i noticed that MOVZX is not classified as a MOV instruction. I used...
The INS_IsMov() only gets the vanilla MOV instructions. It does not return true for MOVZX, MOVSZ, any of the numerous SSE or SSE2 aligned / unaliagned mov's,...
Hi all, In my code, I am creating just 1 thread. However, a the pin tool detects 2. Please explain me of what is happening. [Code which I am Analysis] #include...
Hi, I can't supply a testcase, but I've found a workaround. Originally the pin-tool was started from a C-program, using a shell-script (the C-program doing...
... Just to add a bit more. For pthreads the extra thread will be created depending on thread model used. If the thread model is NPTL I don't think an "extra"...
Cristiano Pereira
ligieri2006@...
Jun 7, 2006 7:26 pm
877
Hello, I'm trying to figure out a way to simulate certain system calls in software and not allow them to be handled by the operating system. For example I'd...
here is one way to do it. You can use the ExecuteAt api to "skip" the system calls that you want. You have to instrument the system call with IPOINT_BEFORE....
Cristiano Pereira
ligieri2006@...
Jun 7, 2006 8:47 pm
879
... That's an interesting question -- with Pin's instrumentation inlining, how do you tell that something you think should be getting inlined actually is? Is...
Nicholas Nethercote
njn@...
Jun 8, 2006 2:04 am
880
I'm having some trouble using the included pthread simulator in pin/Pthread from pin-2.0-3585 on IA32/Linux (Fedora Core 3): I've successfully built a tool...
... At the moment, Pin inlines code w/o control flow. However, work is currently being done to inline code w/ control flow. ... I think the answer to this...
... But there's no way to know for sure that it has occurred, right? Apart from looking at the instrumented code, which I assume you can do with a debugging ...
Nicholas Nethercote
njn@...
Jun 8, 2006 10:46 pm
883
Yea there is no way to be informed whether an anlysis routine has been successfully inlined. ... -- &veej....