Quoting [EMAIL PROTECTED] ([EMAIL PROTECTED]):
> Is DirectFB usable on 2.5 yet for use with a Matrox G400 video card?
> Does the 2.5 matroxfb even work at this point?  ISTR the last time I
> asked about this reports were that it didn't even work in standard
> framebuffer mode yet.

I'm currently successfully running Linux 2.5.70-bk7-fusion-32mb
with XDirectFB and several DirectFB apps on the multi app core.

The "feel" of the system has changed. Fast scrolling text in DFBTerm
looks much smoother, it seems like each new line is rendered instead
of a random amount at once because of buffering. But in some rare cases
moving DFBTerm around looks like a cat chasing a mouse, but the buffer
of motion gets flushed within less than a second.

I had to fix two issues regarding spinlocks. I didn't update the patches yet,
so you have to use the partial patch along with the stuff from the linux/
directory (see linux-fusion/README).

However, fusion benchmarks run slower, maybe because of preemptible
support. Without preemptible the spin_lock/unlock are noops on UP systems.

-----------------------------------------------------------------------

Fusion Benchmark (2.4.21-rc2)

mutex lock/unlock (1 threads)         ->  4535.15 k/sec
mutex lock/unlock (2 threads)         ->  4454.34 k/sec
mutex lock/unlock (3 threads)         ->  4451.04 k/sec
mutex lock/unlock (4 threads)         ->  3883.50 k/sec
mutex lock/unlock (5 threads)         ->  3229.97 k/sec

spinlock lock/unlock (1 threads)      -> 17094.02 k/sec
spinlock lock/unlock (2 threads)      -> 11299.44 k/sec
spinlock lock/unlock (3 threads)      ->  8534.85 k/sec
spinlock lock/unlock (4 threads)      ->  8510.64 k/sec
spinlock lock/unlock (5 threads)      ->  7002.80 k/sec

skirmish prevail/dismiss (1 threads)  ->   716.48 k/sec
skirmish prevail/dismiss (2 threads)  ->   731.45 k/sec
skirmish prevail/dismiss (3 threads)  ->   728.66 k/sec
skirmish prevail/dismiss (4 threads)  ->   736.84 k/sec
skirmish prevail/dismiss (5 threads)  ->   737.62 k/sec

property lease/cede                   ->   717.49 k/sec

ref up/down (local)                   ->   690.13 k/sec
ref up/down (global)                  ->   727.80 k/sec

reactor attach/detach                 ->   338.18 k/sec
reactor attach/detach (2nd)           ->   394.63 k/sec
reactor dispatch                      ->   755.29 k/sec  (100% arrived)

-----------------------------------------------------------------------

Fusion Benchmark (2.5.70-bk7 preemptible)

mutex lock/unlock (1 threads)         ->  4454.34 k/sec
mutex lock/unlock (2 threads)         ->  1499.81 k/sec
mutex lock/unlock (3 threads)         ->   564.39 k/sec
mutex lock/unlock (4 threads)         ->   245.35 k/sec
mutex lock/unlock (5 threads)         ->   270.82 k/sec

spinlock lock/unlock (1 threads)      -> 16393.44 k/sec
spinlock lock/unlock (2 threads)      -> 11661.81 k/sec
spinlock lock/unlock (3 threads)      ->  8053.69 k/sec
spinlock lock/unlock (4 threads)      ->  6661.12 k/sec
spinlock lock/unlock (5 threads)      ->  5817.34 k/sec

skirmish prevail/dismiss (1 threads)  ->   666.67 k/sec
skirmish prevail/dismiss (2 threads)  ->   621.12 k/sec
skirmish prevail/dismiss (3 threads)  ->   605.89 k/sec
skirmish prevail/dismiss (4 threads)  ->   566.34 k/sec
skirmish prevail/dismiss (5 threads)  ->   549.97 k/sec

property lease/cede                   ->   641.95 k/sec

ref up/down (local)                   ->   620.93 k/sec
ref up/down (global)                  ->   650.20 k/sec

reactor attach/detach                 ->   349.16 k/sec
reactor attach/detach (2nd)           ->   376.79 k/sec
reactor dispatch                      ->   690.13 k/sec  (100% arrived)

-- 
Best regards,
  Denis Oliver Kropp

.------------------------------------------.
| DirectFB - Hardware accelerated graphics |
| http://www.directfb.org/                 |
"------------------------------------------"

                            Convergence GmbH


-- 
Info:  To unsubscribe send a mail to [EMAIL PROTECTED] with 
"unsubscribe directfb-dev" as subject.

Reply via email to