Russell asked me to try some benchmarks on some of the "memset" optimizations 
in recent kernels.  I thought I'd post them here in case others are 
interested.

I ran lmbench (version 2alpha10 as it happens) twice in quick succession on 
each kernel.  All the kernels were, I hope, identical other than the memset 
thing.  The machine I used was my CATS.

The "a" kernel is 2.3.19 with Russell's unmodified `memset' macro.
The "b" kernel has the `memset' macro removed from <asm/string.h>, which 
should restore things to how they were in older kernels.
The "c" kernel has the slightly modified `memset' macro that is currently 
in CVS.

p.

Processor, Processes - times in microseconds - smaller is better
----------------------------------------------------------------
Host                 OS  Mhz null null      open selct sig  sig  fork exec sh  
                             call  I/O stat clos       inst hndl proc proc proc
--------- ------------- ---- ---- ---- ---- ---- ----- ---- ---- ---- ---- ----
armv4l-li Linux 2.3.19a  233  0.6  1.3   10   12 0.13K  4.7    4 2.9K  11K  49K
armv4l-li Linux 2.3.19a  233  0.6  1.3    9   12 0.13K  4.7    4 2.9K  11K  49K
armv4l-li Linux 2.3.19b  233  0.6  1.3   10   12 0.13K  4.6    4 2.9K  11K  49K
armv4l-li Linux 2.3.19b  233  0.6  1.3   10   12 0.13K  4.6    4 2.9K  11K  49K
armv4l-li Linux 2.3.19c  233  0.6  1.3   10   12 0.13K  4.6    4 2.9K  11K  49K
armv4l-li Linux 2.3.19c  233  0.6  1.3   10   12 0.13K  4.7    4 2.9K  11K  49K

Context switching - times in microseconds - smaller is better
-------------------------------------------------------------
Host                 OS 2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
                        ctxsw  ctxsw  ctxsw ctxsw  ctxsw   ctxsw   ctxsw
--------- ------------- ----- ------ ------ ------ ------ ------- -------
armv4l-li Linux 2.3.19a  208    425    912   443    926     454     952
armv4l-li Linux 2.3.19a  210    425    917   443    921     460     934
armv4l-li Linux 2.3.19b  206    420    911   423    918     455     939
armv4l-li Linux 2.3.19b  206    420    915   447    958     446     936
armv4l-li Linux 2.3.19c  212    416    909   423    972     444     959
armv4l-li Linux 2.3.19c  207    419    911   451    918     445     941

*Local* Communication latencies in microseconds - smaller is better
-------------------------------------------------------------------
Host                 OS 2p/0K  Pipe AF     UDP  RPC/   TCP  RPC/ TCP
                        ctxsw       UNIX         UDP         TCP conn
--------- ------------- ----- ----- ---- ----- ----- ----- ----- ----
armv4l-li Linux 2.3.19a   208   418 1115   764         983       2204
armv4l-li Linux 2.3.19a   210   429 1031   759         986       2188
armv4l-li Linux 2.3.19b   206   417 1147   765         973       2215
armv4l-li Linux 2.3.19b   206   421 1096   761         988       2199
armv4l-li Linux 2.3.19c   212   422 1196   764         983       2207
armv4l-li Linux 2.3.19c   207   428 1027   759         987       2195

File & VM system latencies in microseconds - smaller is better
--------------------------------------------------------------
Host                 OS   0K File      10K File      Mmap    Prot    Page       
                        Create Delete Create Delete  Latency Fault   Fault 
--------- ------------- ------ ------ ------ ------  ------- -----   ----- 
armv4l-li Linux 2.3.19a                                 5053     2    0.0K
armv4l-li Linux 2.3.19a                                 5017     2    0.0K
armv4l-li Linux 2.3.19b                                 5022     2    0.0K
armv4l-li Linux 2.3.19b                                 5055     2    0.0K
armv4l-li Linux 2.3.19c                                 5009     2    0.0K
armv4l-li Linux 2.3.19c                                 5013     2    0.0K


unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]

Reply via email to