Performances of Kernel mode vs User mode

Performances of Kernel mode vs User mode

Post by sloh » Sun, 11 Apr 2004 01:39:43

Do you mean all-kernel-mode vs. IMGNOTALLKMODE? Or do you mean just
temporarily putting your thread into kernel mode by calling SetKMode()?

SetKMode lets your thread make system calls faster, by sort of using a fast
path. But it also gives your thread access to memory it shouldn't be
touching, etc. So only threads from trusted processes can call SetKMode.
All-kernel-mode sacrifices security for performance by letting all threads
be in kernel mode all the time.

I don't think we provide any perf tests specifically aimed at timing kmode
vs. normal. Perhaps the "osbench" test that comes with PB. If you're
talking about all-kernel-mode then you can just test it by running some
user level benchmark, eg. I-Bench from VeriTest is a good one, if you have
a browser.

XXXX@XXXXX.COM (remove "online" from reply-to address)
This posting is provided "AS IS" with no warranties, and confers no rights.

Windows Embedded Developer Center ( )
Windows Embedded FAQ
( )

Have an opinion on the effectiveness of Microsoft Embedded newsgroups? Let
us know!