Re: [Dri-devel] Good news/The Man returns.

[Brian P. <br...@va...>]

Gareth Hughes wrote:
> 
> Andreas Ehliar wrote:
> >
> > Is that with or without swapping between every glClear()?
> > I'm asking since clearspd defaults to run glClear() 100 times for every
> > glutSwapBuffers().
> 
> It actually bumps the number up to 25,600 clears per swap due to the
> speed of the clears.  clearspd works out the number of operations needed
> to reach the minimum time period, and then benchmarks with that.
> 
> demos> ./clearspd
> For options:  ./clearspd -help
> Mode:  RGB
> SwapBuffers: no
> Size: 400 x 400
> Buffers: color
> Rate: 25600 clears in 3.138s = 8158.06 clears/s   1.30529e+09 pixels/s
> Rate: 25600 clears in 3.138s = 8158.06 clears/s   1.30529e+09 pixels/s
> Rate: 25600 clears in 3.137s = 8160.66 clears/s   1.30571e+09 pixels/s
> Rate: 25600 clears in 3.137s = 8160.66 clears/s   1.30571e+09 pixels/s
> Rate: 25600 clears in 3.137s = 8160.66 clears/s   1.30571e+09 pixels/s
> Rate: 25600 clears in 3.137s = 8160.66 clears/s   1.30571e+09 pixels/s
> Rate: 25600 clears in 3.138s = 8158.06 clears/s   1.30529e+09 pixels/s


Just a footnote: clearspd was written years ago for testing the
speed of window clears and swapbuffers for Mesa's software drivers.
I'm not sure it's of much value for hardware drivers now.

-Brian