[Bug 664778] [souphttpsrc] chunk allocator: memory waste

[GStreamer (bugzilla.gnome.org)]

https://bugzilla.gnome.org/show_bug.cgi?id=664778
  GStreamer | gst-plugins-good | unspecified

--- Comment #6 from Davy <realbrightster at gmail.com> 2011-11-30 00:40:28 UTC ---
Thanks for all kindly feedbacks. 

In my opinion, we should consider performance trade-off between “socket
performance” and “malloc performance”.

Regarding, socket performance related to block size,
I performed some internal test to compare socket’s throughput (bps), increasing
the block size start form default size (4KB)

In case of LAN, the throughput was significantly improved in proportion to
buffer size increment.
e.g. 24 KB has pretty good performance compare with 4KB. The gap was almost 2.5
times.
Another profit of larger block size, it reduce the loop count of
gst_base_src_loop(). 
The only problem is memory waste. 

So my final opinion is..
I don’t want to give up both socket performance and memory utilization.
On the other hand, obviously I should to endure the additional memcpy()
overhead and latency of making full buffer. – it will occurred when network
bandwidth is too low cause “socket read” is just a copy buffer from Kernel to
user-space.

-- 
Configure bugmail: https://bugzilla.gnome.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
You are the assignee for the bug.