registers AMD64 provides, this routine still needs to be written as self-
modifying code for maximum performance. The additional registers do allow
for further optimization over the x86 version by allowing all four pixels
to be in flight at the same time. The end result is that AMD64 ASM is about
2.18 times faster than AMD64 C and about 1.06 times faster than x86 ASM.
(For further comparison, AMD64 C and x86 C are practically the same for
this function.) Should I port any more assembly to AMD64, mvlineasm4 is the
most likely candidate, but it's not used enough at this point to bother.
Also, this may or may not work with Linux at the moment, since it doesn't
have the eh_handler metadata. Win64 is easier, since I just need to
structure the function prologue and epilogue properly and use some
assembler directives/macros to automatically generate the metadata. And
that brings up another point: You need YASM to assemble the AMD64 code,
because NASM doesn't support the Win64 metadata directives.
- Added an SSE version of DoBlending. This is strictly C intrinsics.
VC++ still throws around unneccessary register moves. GCC seems to be
pretty close to optimal, requiring only about 2 cycles/color. They're
both faster than my hand-written MMX routine, so I don't need to feel
bad about not hand-optimizing this for x64 builds.
- Removed an extra instruction from DoBlending_MMX, transposed two
instructions, and unrolled it once, shaving off about 80 cycles from the
time required to blend 256 palette entries. Why? Because I tried writing
a C version of the routine using compiler intrinsics and was appalled by
all the extra movq's VC++ added to the code. GCC was better, but still
generated extra instructions. I only wanted a C version because I can't
use inline assembly with VC++'s x64 compiler, and x64 assembly is a bit
of a pain. (It's a pain because Linux and Windows have different calling
conventions, and you need to maintain extra metadata for functions.) So,
the assembly version stays and the C version stays out.
- Removed all the pixel doubling r_detail modes, since the one platform they
were intended to assist (486) actually sees very little benefit from them.
- Rewrote CheckMMX in C and renamed it to CheckCPU.
- Fixed: CPUID function 0x80000005 is specified to return detailed L1 cache
only for AMD processors, so we must not use it on other architectures, or
we end up overwriting the L1 cache line size with 0 or some other number
we don't actually understand.
SVN r1134 (trunk)
- Reorganized the HackHack code so that the image creation was moved into
MakeTexture. This was necessary because Unload deleted the pixel data
and broke the whole thing.
- Fixed: FPatchTexture::HackHack and FDoomStatusbarTexture::DrawToBar used the
obsolete and uninitialized variable Near255.
- Removed the span creation code specific to FPatchTexture. It only has an
advantage when the lump has already been loaded in memory but since that
is no longer the case now the generic version in FTexture is actually better.
- Changed: FTexture::CopyToBlock no longer uses the spans but the pixel buffer
directly. Since most patches in multipatch textures are non transparent
the added overhead from creating the spans far outweighs any savings they
might provide. It is also simpler to handle for mirrored or rotated patches now.
- Changed: Textures only create the spans when really needed. Flats and native
textures, for example, do not and it only created needless overhead that they
were always created along with the pixel buffer.
- Made use of player and actor variables consistent in a_hereticweaps.cpp.
- Fixed: A few calls to P_SpawnPlayerMissile passed 0 as angle
SVN r911 (trunk)
P_FindFloorCeiling never did that.
- Merged Check_Sides and PIT_CrossLine into A_PainShootSkull.
- Replaced P_BlockLinesIterator with FBlockLinesIterator in all places it was
used. This also allowed to remove all the global variable saving in
P_CreateSecNodeList.
- Added a new FBlockLinesIterator class that doesn't need a callback
function because debugging the previous bug proved to be a bit annoying
because it involved a P_BlockLinesIterator loop.
- Fixed: The MBF code to move monsters away from dropoffs did not work as
intended due to some random decisions in P_DoNewChaseDir. When in the
avoiding dropoff mode these are ignored now. This should cure the problem
that monsters hanging over a dropoff tended to drop down.
SVN r887 (trunk)
- Sbarinfo optimization: Creating and destroying bar textures every frame is
a relatively expensive operation. We can skip the custom texture entirely
and just draw the bars directly to the screen, using the clipping parameters
for DrawTexture(). This also means bars are no longer limited to the game
palette, and the bar itself has the same resolution as the screen.
SVN r731 (trunk)
- Added .txt files to the list of types (wad, zip, and pk3) that can be
loaded without listing them after -file.
- Fonts that are created by the ACS setfont command to wrap a texture now
support animated textures.
- FON2 fonts can now use their full palette for CR_UNTRANSLATED when drawn
with the hardware 2D path instead of being restricted to the game palette.
- Fixed: Toggling vid_vsync would reset the displayed fullscreen gamma to 1
on a Radeon 9000.
- Added back the off-by-one palette handling, but in a much more limited
scope than before. The skipped entry is assumed to always be at 248, and
it is assumed that all Shader Model 1.4 cards suffer from this. That's
because all SM1.4 cards are based on variants of the ATI R200 core, and the
RV250 in a Radeon 9000 craps up like this. I see no reason to assume that
other flavors of the R200 are any different. (Interesting note: With the
Radeon 9000, D3DTADDRESS_CLAMP is an invalid address mode when using the
debug Direct3D 9 runtime, but it works perfectly fine with the retail
Direct3D 9 runtime.) (Insight: The R200 probably uses bytes for all its
math inside pixel shaders. That would explain perfectly why I can't use
constants greater than 1 with PS1.4 and why it can't do an exact mapping to
every entry in the color palette.
- Fixed: The software shaded drawer did not work for 2D, because its selected
"color"map was replaced with the identitymap before being used.
- Fixed: I cannot use Printf to output messages before the framebuffer was
completely setup, meaning that Shader Model 1.4 cards could not change
resolution.
- I have decided to let remap palettes specify variable alpha values for
their colors. D3DFB no longer forces them to 255.
- Updated re2c to version 0.12.3.
- Fixed: A_Wander used threshold as a timer, when it should have used
reactiontime.
- Fixed: A_CustomRailgun would not fire at all for actors without a target
when the aim parameter was disabled.
- Made the warp command work in multiplayer, again courtesy of Karate Chris.
- Fixed: Trying to spawn a bot while not in a game made for a crashing time.
(Patch courtesy of Karate Chris.)
- Removed some floating point math from hu_scores.cpp that somebody's GCC
gave warnings for (not mine, though).
- Fixed: The SBarInfo drawbar command crashed if the sprite image was
unavailable.
- Fixed: FString::operator=(const char *) did not release its old buffer when
being assigned to the null string.
- The scanner no longer has an upper limit on the length of strings it
accepts, though short strings will be faster than long ones.
- Moved all the text scanning functions into a class. Mainly, this means that
multiple script scanner states can be stored without being forced to do so
recursively. I think I might be taking advantage of that in the near
future. Possibly. Maybe.
- Removed some potential buffer overflows from the decal parser.
- Applied Blzut3's SBARINFO update #9:
* Fixed: When using even length values in drawnumber it would cap to a 98
value instead of a 99 as intended.
* The SBarInfo parser can now accept negatives for coordinates. This
doesn't allow much right now, but later I plan to add better fullscreen
hud support in which the negatives will be more useful. This also cleans
up the source a bit since all calls for (x, y) coordinates are with the
function getCoordinates().
- Added support for stencilling actors.
- Added support for non-black colors specified with DTA_ColorOverlay to the
software renderer.
- Fixed: The inverse, gold, red, and green fixed colormaps each allocated
space for 32 different colormaps, even though each only used the first one.
- Added two new blending flags to make reverse subtract blending more useful:
STYLEF_InvertSource and STYLEF_InvertOverlay. These invert the color that
gets blended with the background, since that seems like a good idea for
reverse subtraction. They also work with the other two blending operations.
- Added subtract and reverse subtract blending operations to the renderer.
Since the ERenderStyle enumeration was getting rather unwieldy, I converted
it into a new FRenderStyle structure that lets each parameter of the
blending equation be set separately. This simplified the set up for the
blend quite a bit, and it means a number of new combinations are available
by setting the parameters properly.
SVN r710 (trunk)
of the frame buffer so I can get rid of the last remaining renderer check
outside the display code in GZDoom.
- made V_SetResolution a virtual function in IVideo. It's probably not a perfect
solution but at least it allows overriding it (which I need in GZDoom.)
Note: There's a lot of redundancy between hardware.cpp/h in the SDL and Win32
folders so some cleaning up might be a good idea.
SVN r692 (trunk)
can share the same hardware texture. This greatly reduces the number of
DrawPrimitive calls that need to be made when drawing text (or any 2D
graphics in general), so now hardware text is much faster than software text
all around. (As an example, one scenario went from 315 fps to over 1635 fps
for hardware, compared to 540 fps for software.)
SVN r687 (trunk)
- Modified M_DrawFrame() and R_DrawTopBorder() so that they call FlatFill() to
draw the edges of the frames. This at least seems a bit faster for hardware
2D.
- Implemented FlatFill() for D3DFB. It seems to be exactly as fast as the
default implementation that just calls DrawTexture() to tile the pieces onto
the screen, so I'm not sure it was worth the bother.
SVN r686 (trunk)
vertex buffer, made line batching automatic, and added an index buffer for
use when batching quads. The index buffer actually offered more of a
performance boost than simply batching the quads alone did.
SVN r685 (trunk)
- Bumped the maximum resolution up to 2560x1600.
- Fixed: DCanvas::DrawTexture() only expanded virtual screen sizes for widescreen
resolutions but left 5:4 modes alone. This fix neccessitated the addition of
DTA_Bottom320x200 for the status bar to use so that it could be flush with the
bottom of the screen rather than sitting slightly above it.
- Fixed: FConfigFile::ReadConfig()'s definition of whitespace was too broad.
- Fixed: Defining custom translation ranges in descending order and/or with gaps
made for crashes.
SVN r676 (trunk)
and into the vertex data.
- Added functions for doing line drawing with Direct3D, including a new pair
of functions to do batched line drawing so that the map can actually be
drawn faster in hardware than in software (instead of an order of magnitude
slower).
SVN r663 (trunk)
shader seems to be producing crappy output, so it's disabled for now.
Specifically, it produces distorted output at regular intervals for
textures that aren't power-of-2-sized, and it's still doing visible
filtering when the texture is rendered at its original size, so
obviously it's not doing something right.
- Fixed the use of power-of-2-sized native textures for smaller game
textures again.
- Fixed: D3DFB did not restore all the state it needed to after resetting
the device.
- Fixed: R_DrawTopBorder() must clip itself around the 3D view, since it's
now drawn later.
- With full software rendering, palette flashes once again effect the whole
screen.
Changes I neglected to put in the previous commit log:
- Moved the view border drawing into the 2D mode part. When using Begin2D()
now, the only part of the software buffer that gets updated to the screen
is the part with the actual 3D scene and only if you tell it to.
- Fixed a D3D memory leak on every frame in windowed mode and the same thing
for the screen wipes. Note to self: If it's an interface, be sure to
Release it, because it will be AddRef'ed before being returned to you.
- Moved the BlendView() call out of FBaseStatusBar::Draw() so that it can be
applied before copying the 3D scene to the screen underneath the 2D parts.
- Restored the console's darkening level to its old table-based amount.
- Fixed D3DFB::SetColorOverlay()'s incorrect calculations.
- Fixed the D3D screen wipes for letterboxed modes.
SVN r662 (trunk)
since I couldn't think of any reason why it should be grabbed at any other
time. (This only applies to windowed mode, where it makes sense to let the
OS have control of the pointer.)
SVN r661 (trunk)
drawn to a texture, then that texture is copied to the real back buffer
using a gamma-correcting pixel shader. In fullscreen mode, SetGammaRamp
is used.
- Fixed flashing of vid_fps display when fps > 1000.
- Fixed loading of RGB textures for native 2D mode.
- Changed the first rotozoomer's data because it just became too obvious when
the backdrop is drawn with a full 256 distinct colors available.
- Set the player backdrop to update no more frequently than 35 FPS, so opening
the player setup menu before starting a game won't produce a very fast
moving backdrop.
- Changed the player backdrop into a texture so that it can be drawn like
anything else.
SVN r648 (trunk)
palettes smaller than 256 entries with the shader I wrote for it. Is there
a list of gotchas like this listed some where? I'd really like to see it.
Well, when compiled with SM2.0, the PalTex shader seems to be every-so-
slightly faster on my GF7950GT than the SM1.4 version, so I guess it's a
minor win for cards that support it.
- Fixed: ST_Endoom() failed to free the bitmap it used.
- Added the DTA_ColorOverlay attribute to blend a color with the texture
being drawn. For software, this (currently) only works with black. For
hardware, it works with any color. The motiviation for this was so I could
rewrite the status bar calls that passed DIM_MAP to DTA_Translation to
draw darker icons into something that didn't require making a whole new
remap table.
- After having an "OMG! How could I have been so stupid?" moment, I have
removed the off-by-one check from D3DFB. I had thought the off-by-one error
was caused by rounding errors by the shader hardware. Not so. Rather, I
wasn't sampling what I thought I was sampling. A texture that uses palette
index 255 passes the value 1.0 to the shader. The shader needs to adjust the
range of its palette indexes, or it will end up trying to read color 256
from the palette texture when it should be reading color 255. Doh!
- The TranslationToTable() function has been added to map from translation
numbers used by actors to the tables those numbers represent. This function
performs validation for the input and returns NULL if the input value
is invalid.
- Major changes to the way translation tables work: No longer are they each a
256-byte array. Instead, the FRemapTable structure is used to represent each
one. It includes a remap array for the software renderer, a palette array
for a hardware renderer, and a native texture pointer for D3DFB. The
translationtables array itself is now an array of TArrays that point to the
real tables. The DTA_Translation attribute must also be passed a pointer
to a FRemapTable, not a byte array as previously.
- Modified DFrameBuffer::DrawRateStuff() so that it can do its thing properly
for D3DFB's 2D mode. Before, any fullscreen graphics (like help images)
covered it up.
SVN r640 (trunk)
- moved the AM line drawer into DCanvas as a virtual function. While testing
this code I discovered that the antialias precalculation was never used
except for the very first frame of AM drawing. However, since I couldn't
detect even a marginal performance improvement using this code on 2 computers
I just disabled it completely because it severely complicates a more generic
implementation. I also disabled am_ovtrans in the process because I couldn't
see any positive effects of using this cvar. All it does is adding some
ugly distortion to the lines it affects without any apparent benefits.
* Added fix by Karate Chris containing:
- Added a 'No team changing' DMFlag2 which prevents players from changing teams
unless they are not on a team.
- Added a 'No respawn' DMFlag2 which prevents a player from respawning after
they have died.
- Added a 'Keep frags gained' DMFlag2 which allows you to choose whether you
want to reset the frags of each player next level or not.
- Added a small visual enhancement to the cooperative scoreboard to show
if a player has died.
- Fixed: If the 'teamplay' console variable was set to 'true' in a cooperative
game, the scoreboard would show team play related items as opposed to
cooperative items.
- Fixed: The 'bot_observer' console variable should not work in network games.
- Fixed: Bots made intermission skip really fast.
SVN r634 (trunk)
- Fixed: ASkyViewpoint::Destroy was missing the super call.
- Added SnowKate709's A_LookEx update.
- Moved M_DrawPlayerBackdrop into DCanvas as a virtual function. Since this
directly accesses the frame buffer it must be overridable for renderers
that work differently.
SVN r632 (trunk)
passing renderer specific data to the function. Also added DTA_Font so
that the renderer can fetch font translations from the proper font.
DIM_MAP/ConShade had to be made a regular translation table to make it
work.
- Added Karate Chris's fix for scoreboard displaying team play related data
in non teamplay games.
- Fixed: The team selection menu didn't work.
- Fixed: UpdateTeam passed an FString to Printf.
SVN r623 (trunk)
- Added a new color parameter to DCanvas::Clear() that specifies the
ARGB value of the color. This is used if the old color parameter,
which specifies a palette entry, is -1.
SVN r617 (trunk)
if present.
- Modified GZDoom's true color texture copy functions and added them
to generate 32 bit D3D textures. Paletted TGAs and PCXs are also handled
this way but I don't think these 2 formats are worth some more special
handling.
(Question: Is it worth it to implement special handling for paletted PNGs
so that they are used as 8 bit textures internally?)
SVN r608 (trunk)
both dest and src. Also changed this function so that it is owned by the
destination canvas of the operation which is necessary if it needs to
be overridden by subclasses.
SVN r607 (trunk)
They are not actually drawn with it yet, nor is it complete, but it's
something to start with.
- Split up DCanvas::DrawTexture() into more pieces to make it easier to
virtualize.
- Removed support for non-32-bit palette textures from D3DFB. What kind of
card supports pixel shaders but not 32-bit textures?
SVN r605 (trunk)
crouching motion.
- Fixed: The keypress for nightmare mode verification appeared in the console
if the fullscreen console was open while using the menu.
- Added support for scaling down large console fonts in the customize
controls menu.
- Fixed: Joining a negative team was possible.
SVN r590 (trunk)
warnings. At first, I was going to try and clean them all up. Then I decided
that was a worthless cause and went about just acting on the ones that
might actually be helpful:
C4189 (local variable is initialized but not referenced)
C4702 (unreachable code)
C4512 (assignment operator could not be generated)
SVN r420 (trunk)
be surprised if it doesn't work.
- Reorganized the network startup loops so now they are event driven. There is
a single function that gets called to drive it, and it uses callbacks to
perform the different stages of the synchronization. This lets me have a nice,
responsive abort button instead of the previous unannounced hit-escape-to-
abort behavior, and I think the rearranged code is slightly easier to
understand too.
- Increased the number of bytes for version info during D_ArbitrateNetStart(),
in preparation for the day when NETGAMEVERSION requires more than one byte.
- I noticed an issue with Vista RC1 and the new fatal error setup. Even after
releasing a DirectDraw or Direct3D interface, the DWM can still use the
last image drawn using them when it composites the window. It doesn't always
do it but it does often enough that it is a real problem. At this point, I
don't know if it's a problem with the release version of Vista or not.
After messing around, I discovered the problem was caused by ~Win32Video()
hiding the window and then having it immediately shown soon after. The DWM
kept an image of the window to do the transition effect with, and then when
it didn't get a chance to do the transition, it didn't properly forget about
its saved image and kept plastering it on top of everything else
underneath.
- Added a network synchronization panel to the window during netgame startup.
- Fixed: PClass::CreateDerivedClass() must initialize StateList to NULL.
Otherwise, classic DECORATE definitions generate a big, fat crash.
- Resurrected the R_Init progress bar, now as a standard Windows control.
- Removed the sound failure dialog. The FMOD setup already defaulted to no
sound if initialization failed, so this only applies when snd_output is set
to "alternate" which now also falls back to no sound. In addition, it wasn't
working right, and I didn't feel like fixing it for the probably 0% of users
it affected.
- Fixed: The edit control used for logging output added text in reverse order
on Win9x.
- Went back to the roots and made graphics initialization one of the last
things to happen during setup. Now the startup text is visible again. More
importantly, the main window is no longer created invisible, which seems
to cause trouble with it not always appearing in the taskbar. The fatal
error dialog is now also embedded in the main window instead of being a
separate modal dialog, so you can play with the log window to see any
problems that might be reported there.
Rather than completely restoring the original startup order, I tried to
keep things as close to the way they were with early graphics startup. In
particular, V_Init() now creates a dummy screen so that things that need
screen dimensions can get them. It gets replaced by the real screen later
in I_InitGraphics(). Will need to check this under Linux to make sure it
didn't cause any problems there.
- Removed the following stubs that just called functions in Video:
- I_StartModeIterator()
- I_NextMode()
- I_DisplayType()
I_FullscreenChanged() was also removed, and a new fullscreen parameter
was added to IVideo::StartModeIterator(), since that's all it controlled.
- Renamed I_InitHardware() back to I_InitGraphics(), since that's all it's
initialized post-1.22.
SVN r416 (trunk)
on the automap instead. If its first argument is zero, the map marker itself
appears on the automap, otherwise it is drawn on top of any actor with a TID
matching that argument. If the second argument is one, then the map marker
will only be appear if the player has previously seen the sector it is one.
You can use Thing_Activate and Thing_Deactivate on markers to turn them on
and off. And if you subclass MapMarker with DECORATE, you can easily make
your own custom markers.
- Fixed: Map markers could not be drawn partially off the map. They were
drawn either fully or not at all.
- Fixed: Map markers appeared in the wrong place on a rotated overlay map if
screenblocks < 10.
SVN r356 (trunk)
- The stat meters now return an FString instead of sprintfing into a fixed
output buffer.
- NOASM is now automatically defined when compiling for a non-x86 target.
- Some changes have been made to the integral types in doomtype.h:
- For consistancy with the other integral types, byte is no longer a
synonym for BYTE.
- Most uses of BOOL have been change to the standard C++ bool type. Those
that weren't were changed to INTBOOL to indicate they may contain values
other than 0 or 1 but are still used as a boolean.
- Compiler-provided types with explicit bit sizes are now used. In
particular, DWORD is no longer a long so it will work with both 64-bit
Windows and Linux.
- Since some files need to include Windows headers, uint32 is a synonym
for the non-Windows version of DWORD.
- Removed d_textur.h. The pic_t struct it defined was used nowhere, and that
was all it contained.
SVN r326 (trunk)
data file.
- Removed the setcolor CCMD. It's been obsolete for years, ever since color-
aware cvars were added.
- Changed V_GetColorStringByName() to return an FString, because it did a
copystring() call.
- Extended V_GetColorFromString() so that it accepts HTML-style #RRGGBB and
#RGB color strings.
SVN r313 (trunk)
data enough (2 bits instead of 8), so it was super loud and aliased.
- Fixes for GCC 4.1: Several type-punned pointer warnings, but more
importantly, declaring a friend function inside a class body is no longer
enough to declare that function globally; you must declare it again outside
the class.
- Upgraded FArchive::SerializePointer so that it can store 32-bit indices.
- ACS printing pcodes now build their string in an FSttring instead of a fixed
sized buffer on the stack.
SVN r145 (trunk)
functions added to the exit chain with atterm so that they can be called
in a deterministic order and not whatever order the linker decides to put
them in.
- Fixed: DCajunMaster did not free its getspawned.
- Fixed: P_FreeLevelData() did not free ACS scripts.
- Fixed: Level snapshots were not freed at exit.
- Fixed: The save/load menu list was not freed at exit.
- Fixed: FCompressedMemFile needs a destructor to free the m_ImplodedBuffer.
- Fixed: G_DoLoadGame() did not free the engine string.
- Fixed: M_ReadSaveStrings() did not free the engine string.
- Fixed: Processing DEM_SAVEGAME did not free the pathname string.
- Added a check for truncated flats to FFlatTexture::MakeTexture() because
Heretic's F_SKY1 is only four bytes long.
- Added a dump of the offending state to the "Cannot find state..." diagnostic.
- Fixed: FCompressedFile did not initialize m_Mode in its default constructor.
- Fixed: Heretic and Hexen status bars did not initialize ArtiRefresh.
- Fixed: PNGHandle destructor should use delete[] to free TextChunks.
SVN r111 (trunk)
