gamedevlists-general

RE: [GD-General] Assistance with VTune results required

[Barstow J. <JBa...@uk...>]

> > hall.dll			35%
> > 
> > That 35% in hall.dll breaks down to
> > 	READ_PORT_ULONG
> > and
> > 	HalRequestlpi
> > (among some others)
> 
> A wild guess: "hall" ~= "hardware abstraction layer <something>"
> 
> Another guess: the time is spent waiting for the graphics card to
> complete some operations.  I.e. your CPU has more graphics commands
> for your GPU, but has to wait for the GPU input queue to have some
> room before it can write the new commands.  Meanwhile the GPU is busy
> filling pixels.
> 
> Try rendering to a smaller window to see if the time proportions
> change.

Sorry, I made a typo; it is indeed hal.dll (the kernal hardware abstraction
layer)
I ran the initial test @ 400 x 300 x 16bit windowed.

I've subsequently disabled rendering altogether (the device->Present still
operative)
and sound and dplay and dinput, and the spike is still there.

However, switching to fullscreen and the READ_PORT_ULONG disappears.
(HalRequestlpi is still there.)

RE: [GD-General] Assistance with VTune results required

[Tom F. <to...@mu...>]

More wild guesswork - READ_PORT_ULONG reads a ULONG from a hardware
register? This is a common part of waiting until a blit has completed on the
hardware. This is done so that rendering in a windows doesn't buffer up
hundreds of frames - the driver waits until the blit used by Present()
actually finishes rendering before submitting any more data.

When you go fullscreen, you're using a buffer flip instead, which often uses
a different throttling mechanism.

No idea what HalRequestlpi does. Best ask Richard Huddy - he's the nVidia
driver expert.


Tom Forsyth - purely hypothetical Muckyfoot bloke.

This email is the product of your deranged imagination,
and does not in any way imply existence of the author.

> -----Original Message-----
> From: Barstow Jason [mailto:JBa...@uk...]
> Sent: 06 March 2002 15:50
> To: gam...@li...
> Subject: RE: [GD-General] Assistance with VTune results required
> 
> 
> > > hall.dll			35%
> > > 
> > > That 35% in hall.dll breaks down to
> > > 	READ_PORT_ULONG
> > > and
> > > 	HalRequestlpi
> > > (among some others)
> > 
> > A wild guess: "hall" ~= "hardware abstraction layer <something>"
> > 
> > Another guess: the time is spent waiting for the graphics card to
> > complete some operations.  I.e. your CPU has more graphics commands
> > for your GPU, but has to wait for the GPU input queue to have some
> > room before it can write the new commands.  Meanwhile the 
> GPU is busy
> > filling pixels.
> > 
> > Try rendering to a smaller window to see if the time proportions
> > change.
> 
> Sorry, I made a typo; it is indeed hal.dll (the kernal 
> hardware abstraction
> layer)
> I ran the initial test @ 400 x 300 x 16bit windowed.
> 
> I've subsequently disabled rendering altogether (the 
> device->Present still
> operative)
> and sound and dplay and dinput, and the spike is still there.
> 
> However, switching to fullscreen and the READ_PORT_ULONG disappears.
> (HalRequestlpi is still there.)
> 
> 
> 
> 
> 
> _______________________________________________
> Gamedevlists-general mailing list
> Gam...@li...
> https://lists.sourceforge.net/lists/listinfo/gamedevlists-general
> Archives:
> http://sourceforge.net/mailarchive/forum.php?forum_id=557
>