topic I also have the below in Analyzers

error while reporting

Surya_Narayanan_N_ — Mon, 20 Jan 2014 06:48:34 GMT

I get the following error, when i generate report for some of my configurations. Can you please help me sorting this out?

amplxe: Executing actions 14 % Loading data files

amplxe: Warning: Cannot load data file `/home/snataraj/test/threads/bfs/bw128/data.0/tbs1818746128.tb6' (tb6 trace file is corrupted!).
amplxe: Executing actions 50 % done
amplxe: Error: Error 0x4000001e (Cannot load raw collector data)
amplxe: Using result path `/home/snataraj/test/threads/bfs/bw128'
amplxe: Executing actions 0 %
amplxe: Warning: The result contains a lot of raw data. Finalization may take a long time to complete.
amplxe: Executing actions 14 % Loading data files
amplxe: Warning: Cannot load data file `/home/snataraj/test/threads/bfs/bw128/data.0/tbs1818746128.tb6' (tb6 trace file is corrupted!).
amplxe: Executing actions 50 % done
amplxe: Error: Error 0x4000001e (Cannot load raw collector data)

This is a very interesting

Peter_W_Intel — Mon, 20 Jan 2014 12:06:08 GMT

This is a very interesting issue.

1. Are you sure that your user account has permission to create result directory, /home/snataraj/test/threads/bfs/bw128 ?

2. It seemed that collection duration is long, do you have any problem to run collection shortly? Specify "-duration 60" for example.

3. Do you have any problem to collect with other simple application? For example, "amplxe-cl -collect advanced-hotspots -r /home/snataraj/test/threads/bfs/bw128 -- /bin/ls"

Regards, Peter

If collector was not exited

Peter_W_Intel — Mon, 20 Jan 2014 12:30:42 GMT

If collector was not exited abnormally (killed by user, ctrl-C is allowed), Verify if tb6 file is corrupted, use Finalizing:

amplxe-cl -finalize -r /home/snataraj/test/threads/bfs/bw128

Ok this problem was mainly

Surya_Narayanan_N_ — Wed, 22 Jan 2014 14:44:15 GMT

Ok this problem was mainly due to the temporary file storage limitation. I moved it to another disk and its working fine. I followed the procedure in this

https://www.google.co.in/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CCMQFjAA&url=http%3A%2F%2Fsoftware.intel.com%2Fen-us%2Farticles%2Fhow-to-configure-an-alternate-temporary-scratch-directory-for-vtune-amplifier-xe-on-linux&ei=F9nfUvK0L4in0wWdsICIAw&usg=AFQjCNG7-J8buVIcbbaCsZkYVeB6P4USyQ&bvm=bv.59568121,d.d2k

I also have the below

Surya_Narayanan_N_ — Wed, 22 Jan 2014 15:55:01 GMT

I also have the below mentioned problem while collecting. Can you help me out with this?

PARSEC Benchmark Suite Version 3.0-beta-20130728
read 65536 points
ERROR Reading data from socket: Connection reset by peer
ERROR Reading data from socket: Connection reset by peer
ERROR Reading data from socket: Connection reset by peer
ERROR reading read_record header size 1 from socket: Connection reset by peer
ERROR reading read_record header size 2 from socket: Connection reset by peer
scif_send failed with err -1
scif_recv failed with err -1
scif_send failed with err -1
scif_recv failed with err -1
scif_send failed with err -1
scif_recv failed with err -1
scif_send failed with err -1
scif_recv failed with err -1
ERROR writing data to socket: Connection reset by peer
ERROR writing data to socket: Connection reset by peer
ERROR writing data to socket: Connection reset by peer
ERROR writing data to socket: Connection reset by peer

Did you mean above error

Peter_W_Intel — Thu, 23 Jan 2014 07:46:48 GMT

Did you mean above error message was generated by program itself? If program worked without VTune, it ran smoothly?

It seemed that was due to data send/receive time-out, my opinion is to use PMU event-based sampling instead of user-mode sampling to reduce overhead, and note not to enable call stack collection which might add overheads. Another way is to enlarge SAV value, to reduce sample counts.

Hope it helps.

Regards, Peter

@Surya

Bernard — Thu, 23 Jan 2014 21:04:36 GMT

@Surya

Are you collecting data over network?

I log into the network...i,e

Surya_Narayanan_N_ — Thu, 23 Jan 2014 21:11:55 GMT

I log into the network...i,e the xeon host...then I use the vtune which is in host and launch my collect from the host...where the script runs the natie application in phi.

amplxe-cl -collect-with runsa-knc -knob event-config=CPU_CLK_UNHALTED:sa=100000,HWP_L2MISS:sa=100000 -- ssh mic0 "script.sh"

It works but is uncertain..

Surya_Narayanan_N_ — Thu, 23 Jan 2014 21:13:14 GMT

It works but is uncertain...at times i get this error for certain configurations and am not sure for which ones.

I would like to advise you to

Bernard — Thu, 23 Jan 2014 21:30:51 GMT

I would like to advise you to use wireshark to monitor connection between host and Phi coprocessor(if I understood your description correctly).