Lenovo T570, Ubuntu 18.04.4, 5.3.0-53-generic, OneAPI/Vtune 2021.1.0 beta06

I've been looking into building an i915 module so I can try CONFIG_DRM_I915_LOW_LEVEL_TRACEPOINTS=y to get at the low level information using VTune.  I tried the instructions in https://software.intel.com/content/www/us/en/develop/documentation/vtune... and get the following:

root@t570:~/OneAPI# sudo add-apt-repository -s "deb http://ru.archive.ubuntu/ bionic main restricted"
Hit:1 http://us.archive.ubuntu.com/ubuntu bionic InRelease
Get:2 http://us.archive.ubuntu.com/ubuntu bionic-updates InRelease [88.7 kB]                  
Get:3 http://dl.google.com/linux/chrome/deb stable InRelease [1,811 B]                                           
Get:4 http://security.ubuntu.com/ubuntu bionic-security InRelease [88.7 kB]                                      
Get:5 http://us.archive.ubuntu.com/ubuntu bionic-backports InRelease [74.6 kB]                                   
Get:6 http://ru.archive.ubuntu bionic InRelease                                                                  
Err:6 http://ru.archive.ubuntu bionic InRelease                                                                  
  Clearsigned file isn't valid, got 'NOSPLIT' (does the network require authentication?)
Get:7 http://dl.google.com/linux/chrome/deb stable/main amd64 Packages [1,123 B]                                 
Get:8 http://us.archive.ubuntu.com/ubuntu bionic-updates/main i386 Packages [685 kB]                  
Get:9 http://us.archive.ubuntu.com/ubuntu bionic-updates/main amd64 Packages [951 kB]               
Get:10 http://us.archive.ubuntu.com/ubuntu bionic-updates/main Translation-en [324 kB]                           
Get:11 http://us.archive.ubuntu.com/ubuntu bionic-updates/main amd64 DEP-11 Metadata [306 kB]                    
Get:12 http://security.ubuntu.com/ubuntu bionic-security/main amd64 Packages [727 kB]                            
Get:13 http://us.archive.ubuntu.com/ubuntu bionic-updates/main DEP-11 48x48 Icons [81.9 kB]                      
Get:14 http://us.archive.ubuntu.com/ubuntu bionic-updates/main DEP-11 64x64 Icons [157 kB]                       
Get:15 http://us.archive.ubuntu.com/ubuntu bionic-updates/universe i386 Packages [1,018 kB]                      
Get:16 http://us.archive.ubuntu.com/ubuntu bionic-updates/universe amd64 Packages [1,077 kB]                     
Get:17 http://us.archive.ubuntu.com/ubuntu bionic-updates/universe Translation-en [335 kB]                       
Get:18 http://us.archive.ubuntu.com/ubuntu bionic-updates/universe amd64 DEP-11 Metadata [279 kB]                
Get:19 http://us.archive.ubuntu.com/ubuntu bionic-updates/universe DEP-11 48x48 Icons [213 kB]                   
Get:20 http://us.archive.ubuntu.com/ubuntu bionic-updates/universe DEP-11 64x64 Icons [462 kB]                   
Get:21 http://security.ubuntu.com/ubuntu bionic-security/main i386 Packages [474 kB]                             
Get:22 http://us.archive.ubuntu.com/ubuntu bionic-updates/multiverse amd64 DEP-11 Metadata [2,464 B]             
Get:23 http://us.archive.ubuntu.com/ubuntu bionic-backports/universe amd64 DEP-11 Metadata [7,968 B]             
Get:24 http://security.ubuntu.com/ubuntu bionic-security/main Translation-en [230 kB]                            
Ign:25 https://repositories.intel.com/graphics/ubuntu bionic-devel InRelease        
Get:26 https://repositories.intel.com/graphics/ubuntu bionic-devel Release [2,702 B]
Get:27 http://security.ubuntu.com/ubuntu bionic-security/main amd64 DEP-11 Metadata [43.8 kB]
Get:28 http://security.ubuntu.com/ubuntu bionic-security/main DEP-11 48x48 Icons [25.7 kB]
Get:29 http://security.ubuntu.com/ubuntu bionic-security/main DEP-11 64x64 Icons [58.0 kB]
Get:30 http://security.ubuntu.com/ubuntu bionic-security/universe i386 Packages [623 kB]
Get:31 http://security.ubuntu.com/ubuntu bionic-security/universe amd64 Packages [669 kB]
Get:32 http://security.ubuntu.com/ubuntu bionic-security/universe Translation-en [222 kB]
Get:33 http://security.ubuntu.com/ubuntu bionic-security/universe amd64 DEP-11 Metadata [49.2 kB]
Get:34 http://security.ubuntu.com/ubuntu bionic-security/universe DEP-11 48x48 Icons [18.6 kB]
Get:35 http://security.ubuntu.com/ubuntu bionic-security/universe DEP-11 64x64 Icons [107 kB]
Get:36 http://security.ubuntu.com/ubuntu bionic-security/multiverse amd64 DEP-11 Metadata [2,464 B]
Ign:37 https://repositories.intel.com/graphics/ubuntu bionic-devel Release.gpg                   
Get:38 https://repositories.intel.com/graphics/ubuntu bionic-devel/main amd64 Packages [4,164 B]
Reading package lists... Done                           
E: Failed to fetch http://ru.archive.ubuntu/dists/bionic/InRelease  Clearsigned file isn't valid, got 'NOSPLIT' (does the network require authentication?)
E: The repository 'http://ru.archive.ubuntu bionic InRelease' is not signed.
N: Updating from such a repository can't be done securely, and is therefore disabled by default.
N: See apt-secure(8) manpage for repository creation and user configuration details.

I'm already worried I'm gonna brick my laptop doing this, so I stopped.   I then thought it might make sense to point to another source repository so I enabled "deb-src http://us.archive.ubuntu.com/ubuntu/ bionic main restricted" in /etc/apt/sources.list.  (maybe this is doomed, but it seemed like a reasonable try.)  It didn't help.  The result was:

root@t570:~/OneAPI# apt build-dep linux linux-image-$(uname -r)
Reading package lists... Done
Picking 'linux-signed-hwe' as source package instead of 'linux-image-5.3.0-53-generic'
E: Unable to find a source package for linux-image-5.3.0-53-generic

So, I'm kinda stuck.  Anyone got some advice for me?  Is building a new i915 module really worth it?

Hi,

For the storage performance snapshot tool, has the queue depth output in the results been removed?

https://software.intel.com/sites/products/snapshots/storage-snapshot/

The getting started guide for the tool mentions queue depth

https://www.intel.com/content/www/us/en/develop/documentation/get-starte...

I have run the tool with a test workload and can't see the queue depth graph that used to exist in the tool previously.

Any help would be much appreciated!

Kind Regards,

Conor

Hi there,

I have been adding some tasks to my code and I was trying to create sub-tasks that are within a parent task. I did a bit of searching on the web and I noticed that within the Intel Graphics Performance Analyzers the ITT APIs appear to offer such functionality. (https://software.intel.com/content/www/us/en/develop/documentation/gpa-u...).

I started looking through the header file for ITT I have in VTune (ittnotify.h), and I found references to 'overlapped' tasks, as well as '__itt_id_make' and '__itt_id_create' functions. I had a quick attempt at trying to make use of these, but as there is little documentation I struggled to make any head way.

Is it possible to create sub-tasks and assign them to parent task_ids, and is there any documentation or examples anywhere I can use to crib usage from?

Thanks,

Ewan

Hello,

Could I get some help with how to analyze the performance of my code? I know I can make reports for FPGAs, but what do I do for GPUs to analyze my code?

Is there a better way to get performance output other than reports?

Thank you!

Lenovo T570, Ubuntu 18.04.4, Intel(R) oneAPI VTune(TM) Profiler 2021.1.0 beta06 (build 608503) Command Line Tool

Given the following toy code:

#include <unistd.h>
#include <ittnotify.h>
int 
main(int argc, char *argv[])
{
  auto itt_d = __itt_domain_create("Domain");
  auto itt_str1 = __itt_string_handle_create("Task 1");
  unsigned __int64 counter = 5;
  auto itt_counter = __itt_counter_create("acounter", "cDomain");
  __itt_task_begin(itt_d, __itt_null, __itt_null, itt_str1);
  sleep(1);
  __itt_counter_set_value(itt_counter, &counter);
  sleep(1);
  __itt_counter_inc(itt_counter);
  sleep(1);
  __itt_counter_inc(itt_counter);
  sleep(1);
  __itt_counter_dec(itt_counter);
  sleep(1);
  __itt_counter_destroy(itt_counter);
  __itt_task_end(itt_d);
}

Compiled as:

dpcpp -g -o itt itt.cpp -I/local/opt/intel/inteloneapi/vtune/2021.1-beta06/include/pkgconfig/lib64/../../ -L/local/opt/intel/inteloneapi/vtune/2021.1-beta06/include/pkgconfig/lib64/../../../lib64 -littnotify -ldl   -lOpenCL -lsycl

I then run itt in vtune-gui and get the attached (I don't know how to put it inline).

The issue is that, based on the code, I would expect the counter to be undefined/default from 0..1 seconds, 5 from 1..2, 6 from 2..3, 7, from 3..4 and 6 from 4..5 seconds.  That's not what shows up.  Instead the five regions show the counter to be 0, 6, 7, 6, 0.

There's not a lot of examples or discussion in the VTUNE manual concerning use of the ITT API, so what am I missing?

Extra credit: Eventually I'm going to want to do this :^). Is there a way to instrument SYCL kernels with the ITT API?

Is there a simple way to make vtune (2019) count the number of times a function has been called?

I am new to VTune and I am using it to analyse a simple algorithm like finding min and max of a large array. I used the microarchitecture exploration mode hoping to investigate if it is memory bound or front end bound. So in this situation, what does it mean to be 100% core bound? Or more likely, what am I doing wrong while measuring? Sorry if I am asking a stupid question and thank you for taking the time to read my question.

Other info:

Linux kernel version: 4.14.81.bm.21-amd64

LSB: Debian GNU/Linux 9.12 (stretch)

Model name: Intel(R) Xeon(R) Gold 6130 CPU @ 2.10GHz (Skylake)

Code I am running, the source array can go up to 1GB (== source size of 2^28 as each data is 4 bytes): 

__attribute__((noinline))
void findMinMax(int32_t *src, uint32_t src_size, int32_t &min, int32_t &max) {
	min = src[0];
	max = src[0];
	for (int i = 1; i < src_size; i++) {
		auto current = src[i];
		if (current < min) {
			min = current;
		}
		if (current > max) {
			max = current;
		}
	}
}

The executable is also compiled using GCC 9.3 with -O3 flag. I also have tried different levels of optimisation flags, including no optimisation flag, and all of them yield 100% core bound.

I also have tried on various other simple algorithms and they seem to give the same 100% core bound issue.