Context switching and kernel / user mode boundaries

[mrtroy]

Hi all,

I'm trying to understand the way windows currently operates.. From what knowledge I have gathered the monotholic nature of windows xp means that the device drivers run in kernel space. So I'm just trying to get my head around how the system does a graphics operation like gl_begin().

when I call that in my user space program does a context switch occur when the method in opengl32.dll is invoked? Is there a context switch happening for every GL command?

I'm just trying to understand the architecture so I can properly design the video component of my os.

[mrkaktus]

I think yes, there is context swith. This is why M$ want to write brand new graphic drivers for Windows Vista which will operate in User Level (DPL/CPL=3).

[Da_Maestro]

No there is no context switch. The windows OpenGL libraries use a call-gate to switch to kernel mode in the context of the current task.

Having said that there is performance overhead with a switch from user to kernel mode, but it is negligible compared to the gains you get by running the code in kernel mode.

I think yes, there is context swith. This is why M$ want to write brand new graphic drivers for Windows Vista which will operate in User Level (DPL/CPL=3).

This is a rumor. The original Windows NT has a graphics engine which runs in user mode. They scrapped it and moved graphics into kernel mode because it was too slow.