- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi, I'm trying to use VTune to view some google json trace data. I've custom converted it into csv which looks like this (truncated for easy reproduction):
name,start_tsc.UTC,end_tsc,pid,tid
E:/edward/dev17.5/hfs/houdini/help/vexhelp,2019-08-26 16:50:14.879000,2019-08-26 16:51:32.232000,,
E:/edward/dev17.5/hfs/houdini/help/command.help,2019-08-26 16:50:14.878000,2019-08-26 16:50:40.690000,,
E:/edward/dev17.5/hfs/houdini/help/exprhelp,2019-08-26 16:50:14.879000,2019-08-26 16:50:39.200000,,
E:/edward/dev17.5/hfs/houdini/help/config.help,2019-08-26 16:50:14.880000,2019-08-26 16:50:16.405000,,
From the command line I've tried:
amplxe-cl -import path_to_file.csv
But all I can get is this:
amplxe: Importing a new result 100 % done
amplxe: Using result path `C:\Users\edward\Documents\Amplifier XE\Projects\dev17.5 build\r012'
amplxe: Executing actions 50 % done
amplxe: Error: 0x4000002a (Database interface error) -- Precompute error
What am I missing?
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Currently you can't import custom data into new result. You can only add it to an existing result. The idea behind this is that such external data complements the one that VTune collects - i.e. you can see it alongside with all kinds of profiling information VTune provides out of the box like PMU events, memory/upi/pci bandwidth, etc.
So what you can do is collect the google json traces in parallel with some VTune collection. After that you can import you custom data into VTune result. You can even automate this to some extent using custom collector VTune feature - e.g. see this article as an example: https://software.intel.com/en-us/articles/profiling-tensorflow-workloads-with-intel-vtune-amplifier
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I tried that but it's not practical in my case because I'm profiling a build process that takes over 2 hours. So I don't want to fake a "collection" that runs two hours. If I just tell it to run some process that finishes immediately, then I run into "Timestamp out of range problems", so it skips all the data. I then tried to import the data again using the "Import using CSV" using the generated result which gave no errors or warnings but I still could see any of the data.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I'm afraid you have to have VTune result and it must cover the whole running time. This is the only way.
Can you tell us more about the data you are trying to analyze and why you want to use VTune to visualize it?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I want to use VTune to visualize it because Chrome's "about:tracing" UI is not nearly as good, plus I already have VTune at my disposal.
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page