[UPDATE] Performance analysis of Linux Kernel Markers 0.20 for 2.6.17

[Mathieu Desnoyers <compudj@krystal.dyndns.org>]

Here are two minor corrections of my results :

* Mathieu Desnoyers (compudj@krystal.dyndns.org) wrote:
[...]
> * Micro-benchmarks
[...]
> The following tests are done with the "optimized" markers only
>
[...]
> - Execute a loop with a marker enabled, with an empty probe. Var args argument
>   setup, probe empty. No preemption disabling.
> NR_LOOPS : 100000
> time delta (cycles): 3363450
> cycles per loop : 33.63
- cycles per loop to disable preemption : 44.08-33.63=10.45
+ cycles per loop to disable preemption : 52.11-33.63=18.48
[...]
> * Size (x86)
> 
> This is the size added by each marker to the memory image :
> 
> - Optimized
> 
> .text section : instructions
> Adds 6 bytes in the "likely" path.
> Adds 32 bytes in the "unlikely" path.
> .data section : r/w data
> 0 byte
+ 4 bytes for the call address
> .rodata.str1 : strings
> Length of the marker name
> .debug_str : strings (if loaded..)
> Length of the marker name + 7 bytes (__mark_)
> .markers
> 8 bytes (2 pointers)
> .markers.c
> 12 bytes (3 pointers)
> 
> - Generic
> 
> .text section : instructions
> Adds 11 bytes in the "likely" path.
> Adds 32 bytes in the "unlikely" path.
> .data section : r/w data
> 1 byte (the activation flag)
+ 4 bytes for the call address
> .rodata.str1 : strings
> Length of the marker name
> .debug_str : strings (if loaded..)
> Length of the marker name + 7 bytes (__mark_)
> .markers
> 8 bytes (2 pointers)
> .markers.c
> 12 bytes (3 pointers)
> 


Mathieu


OpenPGP public key:              http://krystal.dyndns.org:8080/key/compudj.gpg
Key fingerprint:     8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68