which only applies to TIA mode (in fullscreen, of course).
Some code/debugging cleanups across the various FrameBuffer classes.
Finally fixed software rendering mode crashes. It now works just as
well as OpenGL mode, with all dialog positioning working the same in
both.
Reactivated OpenGL filtering (GL_LINEAR and GL_NEAREST). Only the TIA
image can be changed; the UI dialogs are always using GL_NEAREST).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1579 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
in OpenGL mode now, and I think I've finally finished the major UI stuff
there. Now for software mode.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1578 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
double-buffered modes don't have flicker. Basically, the core FrameBuffer
class has been modified to draw dialogs multiple times when necessary,
above and beyond when changes are made (ie, if a new surface is to be
overlaid, all surfaces underneath must be drawn to BOTH buffers in
double-buffered mode first).
Software rendering is currently broken in fullscreen mode, but I know
why. It's basically the optimizations I try to make for large software
surfaces (the 'isBase' functionality). This is going to disappear before
the next release. And in case I haven't mentioned it for a while; I
REALLY HATE software mode.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1577 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
and allows you to change it for the next Stella run, but doesn't
actually change it for this run (since Stella now has to be
restarted for toggling between software and OpenGL mode).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1575 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
tested 32-bit mode for now; more testing to come.
Made changes in OSystem so creating a framebuffer is only ever done
once. This means that switching between software and OpenGL mode is
no longer dynamic (ie, it can no longer be done while Stella is running).
With the forthcoming changes to VideoDialog, you'll be able to set the
mode to change the next time Stella runs. It just adds too much
complexity to support dynamic switching, especially when I don't think
many people do it too often.
Updated CommandDialog to be font-resizable (ie, take advantage of the
new UI fonts).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1574 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
something which was the basis for this whole surface restructuring in
the first place; updated fonts in the UI.
Switched out the old normal and larger sized fonts for higher
resolution ones. The surfaces are now mostly larger than 320x210,
which means 1x TIA zoom mode is no longer allowed (which was actually
disabled some time ago).
I'm getting close to the end of a single surface design to what we have
now; a multi-level surface infrastructure. This allows higher-res UI's
(which have just been committed), and optimizations for OpenGL TIA
rendering (which were committed over the last week). In the future,
this will also make it easy to add TIA filters to OpenGL mode (ScalexX,
HQxX, etc).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1573 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Fixed compile warning in converting integer values to hex strings in the
debugger. The resulting code was actually causing crashes in the RIOT
tab of the debugger.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1572 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
now correctly positioned, but there's still an issue with ContextMenus
which are used by PromptWidgets.
Finally, we see the first performance improvements in OpenGL TIA
rendering coming from the new 'surface' infrastructure. Now that TIA and
UI surfaces are rendered separately, we no longer have to double the TIA
width in software; it can be hardware-accelerated (ie, a 160x200 texture
is stretched in hardware).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1571 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
to being finished now (thankfully, since I want to move to other parts
of the codebase).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1570 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
It seems the fix for OpenGL issues in OSX do work, so now the OpenGL
code is working the same on all systems again. So now I can get to
the fullscreen positioning issues ...
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1569 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
behaviour differs between OSX and other systems.
Disabled output of the TIA display while in the debugger; this code
needs to be rewritten.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1567 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
I'm really beginning to tire of the OpenGL stuff right now, since it
seems everything works differently between Linux, OSX and Windows. And
surprise, surprise, there never seems to be any problems in Linux ...
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1564 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
of FBSurfaces should belong solely to FrameBuffer, so that different
parts of the code don't try to delete what they don't own.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1561 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Basically, all textures/surfaces need to be reloaded each time the GL
context changes (which seems to be more often under OSX than on other
systems). This new code takes care of that.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1560 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
textures in OSX. It seems the textures have to be completely reloaded
after the window changes size, even if it's still large enough to hold
the previous textures. I assume the GL context is being reset each
time SDL_SetVideoMode is called (although this doesn't happen in Linux
or Win32).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1559 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Can you believe it was because of ONE #$&#%&** line of code!
Still TODO, a general cleanup of the positioning code (ie, switching
from windowed to fullscreen & vice-versa) used the 'old' coordinates,
so things often aren't centered.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1557 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Fixed crash when displaying 'popup' messages with OpenGL.
Changed OpenGL internal format to be the same as OSX. Based on my
experimentation and research, this is the fastest format independent
of the platform/operating system.
Updated TODO marking the reload ROM listing functionality as being
complete.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1556 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
some time; I guess it's time to make it public. I expect to add many
of the top items for the next release (at this point hopefully in
January 2009).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1554 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Partial fixes to F8 bankswitching code, also adding hardcoded support
for HES Challenge/Surfers Paradise double cart (the lower bank was
never accessible). More work is required in this area.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1552 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
the UI stuff, but at least emulation mode is now working. But for
some strange reason, starting a ROM from the commandline gives a blank
screen, while it works fine from the ROM launcher.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1549 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
surfaces are still separate and blitted separately, but the actual
screen updates are now done all at once, at the end of the frame (as in
the previous code). This makes things a little faster, and eliminates
an annoying flicker because of SDL_UpdateRect being called at different
times.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1547 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
that randomization was only done once per game invocation, and not each
time Device::reset() was called (which according to its definition is
when it should be done). This has the side-effect of fixing A. Davies
notBoulderdash ROM, which didn't work if display-type was set to
auto-detect.
Some fixes for dynamic menus generated by PopUpWidget and its
associated dialog/surfaces.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1546 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Those widgets now take two items per entry; the string to display, and
the one to use (in essence a key/value pair). This really shortens the
code in quite a few places.
TIA graphical filters are now selectable from the UI, and correctly change
video mode (in TIA mode).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1545 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
fullscreen resolutions will be automatically chosen based on the required
size for the window. The image will be centered and keep the same aspect
ratio, however, so operation will still work correctly on widescreen
monitors. 'Auto' will be the new default. Otherwise, if a specific
resolution is requested, Stella will try to accomodate it *only* if it fits
into the resolution; otherwise the smallest resolution that fits will be
used.
Removed 'zoom_ui' and 'zoom_tia'. The UI can now only be at 1x mode.
Aded 'tia_filter' commandline argument, which specifies to the filter
to use when rendering the tia image. For now, these accept 'zoom1x',
'zoom2x'..., up to 'zoom10x', and duplicate previous behaviour. Eventually,
Scalexx and HQxx filters may be added. Still TODO is add this to the UI.
First pass at making the standard build use a minimum of zoom2x for the TIA,
so the UI can be larger and use a better looking font. There's still work
to do in this area, especially for those ports with limited hardware that
support zoom1x only.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1544 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
run at all, since I haven't even started on the OpenGL stuff yet.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1543 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
and OpenGL mode (windowed and fullscreen) isn't working at all.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1542 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
onscreen, and don't cause a segfault, but there's still a positioning
problem when toggling fullscreen/windowed modes and when the base surface
has 'unusable' area (ie, an 800x600 window centered in a 1024x768 screen).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1540 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
stats and general messages are now shown correctly. ContextMenu and
PopUpWidget are now drawn correctly, and take image 'centering' into
account. Fixed many 'offset' bugs in rendering surfaces. Fixed
long-standing issue where pressing Alt-Enter to toggle fullscreen mode
actually passed the Enter to the underlying dialog (it's now properly
swallowed). Probably many more items I'm forgetting ...
Still TODO is work on other dialogs that aren't centered by default
(InputTextDialog, etc). My plan is to stabilize these changes and then
fix OpenGL mode, and only after I've achieved parity with 2.6.1 to
actually make use of the new infrastructure (larger fonts, etc).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1539 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
certain number. This is similar to how the old PopUpDialog worked.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1538 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
in-game dialogs are now rendered into separate surfaces, and then layered
onto the SDL screen when necessary. The biggest advantage of this is
now each 'window' is separate, and won't be affected by the scaling or
filtering of other windows. For example, zooming the TIA image no longer
changes the size of UI text overlaid on the TIA. Similarly, eventually
graphical filters will be added, and filters applied to the TIA won't
affect the UI.
There's too many changes to list, so for now I'll list what doesn't work:
* OpenGL mode (don't even try it)
* Several popup dialogs in the debugger
* RomInfo stuff is completely broken
* Scanline count and FrameBuffer 'messages' aren't shown
* Larger fonts - fonts are no longer scaled; we have to use larger ones,
and update the UI accordingly
* Probably a lot of other stuff I forgot to mention
This is a huge undertaking, similar in scale to when the UI was first
added in 2.0 release. So it might take some time for this to stabilize ...
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1537 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
doesn't happen very often (which is why I didn't notice it before), but
at least now if it *does* happen, an error message will be printed
instead of just segfaulting.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1536 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Further tweaking to the RIOT timer values. I reverted to the z26
'0x40000' number, even though I'm not entirely clear why it's being
used. Without it, several PAL ROMs fail in an infinite loop of
reading from INTIM and checking for zero. I really hope that's the
last of the RIOT stuff, since I've been looking at it for the past
two months (the AVox and SaveKey issues started there as well).
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1531 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
Removed scanline autodetection from the Console startup code. It makes
more sense to just assume the Console is either NTSC or PAL, and let the
TIA auto-adjust the framerate if necessary.
git-svn-id: svn://svn.code.sf.net/p/stella/code/trunk@1530 8b62c5a3-ac7e-4cc8-8f21-d9a121418aba
stephena