Menu
▾
▴
basilisk-devel
|
From: Brian J. <bjj...@us...> - 2001-07-11 17:49:35
|
--- Gwenole Beauchesne <gb...@di...> wrote:
>
> While performing a few tests with Speedometer 4.x, in "all tests"
> (cmd-'A') mode, the first graphics test will switch the depth and
> then restores it back for the disk and fpu performance test.
>
> Then, "Benchmark mix" and "FPU benchmark" tests are run. Finally, in
> "QuickDraw tests" the depth is switched again, test is carried out,
> but when it's time to switch depth back to its original value, B2
> crashes with the following.
>
> do_handle_screen_fault: unhandled address 0x8
> ...
I haven't seen that particular problem; running just the QuickDraw
section of the Speedometer tests seems to work fine on my box. But I
do get a BadMatch error whenever I try to switch into monochrome mode:
Gdk-ERROR **: BadMatch (invalid parameter attributes)
serial 3461 error_code 8 request_code 1 minor_code 0
The other video modes seem to work just great, and B2 finds an
appropriate visual for all of them (my machine supports 8, 12, 15, and
24 plane visuals.) They are even pretty fast, as reported by
Speedometer. Although the basic CPU performance seems to have taken a
hit in recent weeks. I should run some more performance profiles...
(BTW, we should be able to greatly increase blitter performance by
doing 64-bit loads and stores on machines which can do them directly.
Something to think about.)
One other oddity: Maelstrom seems to run slowly, perhaps about half
the speed it should. This is under either a 30hz or 60hz refresh
rate. I think the latest round of changes to the video and timing
code introduced the problem; I believe it was fine (and even a bit too
fast) with the first iterations of the depth switching code.
Brian
--------------------------------------------------------------------
"Great works are performed not by power, but by perseverance."
-- Samuel Johnson
|
|
From: Brian J. <bjj...@us...> - 2001-07-18 18:16:46
|
--- Christian Bauer <cb...@st...> wrote:
>
> On Wed, Jul 11, 2001 at 12:49:08PM -0500, Brian Johnson wrote:
> > But I do get a BadMatch error whenever I try to switch into monochrome
> > mode:
>
> I think I just fixed that.
Yep, it works now. Thanks!
> > One other oddity: Maelstrom seems to run slowly, perhaps about half
> > the speed it should.
>
> Does the caret cursor ("General Controls") blink faster or slower?
It's quite slow. But I noticed that if I move the mouse constantly,
it seems to speed up to its usual rate. Something to do with the new
X event loop?
Brian
--------------------------------------------------------------------
"We are Microsoft. UNIX is irrelevant. OS/2 is irrelevant. Openness
is futile. Prepare to be assimilated."
-- pr...@tu...
-- (quoted by Eric Berggren)
|
|
From: Christian B. <cb...@st...> - 2001-07-31 14:15:45
|
Hi! On Wed, Jul 18, 2001 at 01:16:37PM -0500, Brian Johnson wrote: > It's quite slow. But I noticed that if I move the mouse constantly, > it seems to speed up to its usual rate. Something to do with the new > X event loop? Try switching the threaded vs. POSIX.4 timer 60Hz interrupt around in unix_main.cpp. Bye, Christian -- / Coding on PowerPC and proud of it \/ http://www.uni-mainz.de/~bauec002/ |
|
From: Brian J. <bjj...@us...> - 2001-08-06 19:30:55
|
--- Christian Bauer <cb...@st...> wrote:
>
> On Wed, Jul 18, 2001 at 01:16:37PM -0500, Brian Johnson wrote:
> > It's quite slow. But I noticed that if I move the mouse
> > constantly, it seems to speed up to its usual rate. Something to
> > do with the new X event loop?
>
> Try switching the threaded vs. POSIX.4 timer 60Hz interrupt around
> in unix_main.cpp.
That didn't work so well: one_tick() takes the interrupt flag lock
via SetInterruptFlag(), so whenever the thread servicing the timer
interrupt happened to be holding the lock already (eg. via EmulOp()),
BII would deadlock. I tried changing timer_event.sigev_notify from
SIGEV_SIGNAL to SIGEV_THREAD, which should cause pthreads to start a
new thread to run one_tick() when the timer expires. But I somehow
ran into a segv with SIGSEGV blocked or ignored, making IRIX SIGKILL
the process before dbx could get a look at it.
I also tried using the threaded code, but rearranging the loop in
tick_func() to look like the one in redraw_func(). That didn't
change anything.
As a hack, I tried calling one_tick() from redraw_func(), right after
the video refresh. That changed the behavior: the cursor flashed at
its "correct" rate when the mouse was still, but went faster than it
should have when I moved the mouse around.
The code in redraw_func() sure looks correct to me, no matter how long
select() delays.... perhaps I'm getting bad values from
GetTicks_usec()? On a whim, I tried hacking GetTicks_usec() to use
gettimeofday() instead of clock_gettime(). No change.
Brian J. Johnson
--------------------------------------------------------------------
"mv emacs /vmunix"
-- The future according to Michael Rawdon
|
|
From: <gb...@di...> - 2001-07-14 06:02:53
|
Hi,
> I haven't seen that particular problem; running just the QuickDraw
> section of the Speedometer tests seems to work fine on my box. But I
> do get a BadMatch error whenever I try to switch into monochrome mode:
>=20
> Gdk-ERROR **: BadMatch (invalid parameter attributes)
> serial 3461 error_code 8 request_code 1 minor_code 0
Indeed, I so that in the train on a Linux/ppc box too. Actually, it
seems we can take the default colormap in 1-bit mode only if
vis =3D=3D DefaultVisual(x_depth, screen) as well
[see the window driver just befor XCreateWindow() is called]
> (BTW, we should be able to greatly increase blitter performance by
> doing 64-bit loads and stores on machines which can do them directly.
> Something to think about.)
Yes, I will have a look at it now that I have access to a full 64-bit
system ;-)
Bye,
Gwenol=E9.
|
|
From: Christian B. <cb...@st...> - 2001-07-14 15:13:17
|
Hi!
On Wed, Jul 11, 2001 at 12:49:08PM -0500, Brian Johnson wrote:
> But I do get a BadMatch error whenever I try to switch into monochrome
> mode:
I think I just fixed that.
> One other oddity: Maelstrom seems to run slowly, perhaps about half
> the speed it should.
Does the caret cursor ("General Controls") blink faster or slower?
Bye,
Christian
--
/ Coding on PowerPC and proud of it
\/ http://www.uni-mainz.de/~bauec002/
|
|
From: <gb...@di...> - 2001-07-14 18:41:41
|
Hi, > > But I do get a BadMatch error whenever I try to switch into monochrome > > mode: >=20 > I think I just fixed that. Indeed. So does the crash due to double deallocation. Thanks, Gwenol=E9. |
|
From: Gwenole B. <gb...@di...> - 2001-08-19 17:44:21
|
On Wed, 11 Jul 2001, Brian Johnson wrote: > (BTW, we should be able to greatly increase blitter performance by > doing 64-bit loads and stores on machines which can do them directly. > Something to think about.) Done in CVS but I don't see any improvement on the ia64. :-/ Probably not significant since gcc is known to generate sub-optimal code for now... Bye, Gwenole. |
×
Want the latest updates on software, tech news, and AI?
Get latest updates about software, tech news, and AI from SourceForge directly in your inbox once a month.
Thanks for helping keep SourceForge clean.
X