Re: [tracecompass-dev] Use Trace compass internal storage for large trac

[Matthew Khouzam <matthew.khouzam@xxxxxxxxxxxx>]

Hi, I am out of it, I just want to give you the heads up:

TmfTimestamp.fromNanos(System.currentTimeMillis() * 1000) should be TmfTimetamp.fromMillis(System.currentTimeMillis())

---

From: Vinod Appu <vinod.appu@xxxxxxxxxxx>
Sent: Monday, August 5, 2024 1:16 AM
To: Matthew Khouzam <matthew.khouzam@xxxxxxxxxxxx>; tracecompass-dev@xxxxxxxxxxx <tracecompass-dev@xxxxxxxxxxx>
Subject: Use Trace compass internal storage for large trace streaming mode

 

Hi,

 

I’m trying to implement a streaming function flow trace (function entry and exit), where the below specific case is getting handled.

 

1. Now we keep function entry and exit it in a list in heap which is the data to be displayed, but obviously for large data, this is not enough.
2. If we store the function entry and exit again into a file, which will affect the performance.

 

Below is what I’m trying to achieve in theory.

 

1. Use state system itself as the cache, stream events and feed to the state system.
2. Let the state system persist data to hard disk in its own format.
3. One-time sequential analysis, I’ll discard the original data once analysis is done (Need to identify a point where this can be done too.)

 

Please share your thoughts, Could you please share the major areas I must investigate?

 

~Vinod