qzdoom/src/r_draw.cpp

2293 lines
50 KiB
C++
Raw Normal View History

// Emacs style mode select -*- C++ -*-
//-----------------------------------------------------------------------------
//
// $Id:$
//
// Copyright (C) 1993-1996 by id Software, Inc.
//
// This source is available for distribution and/or modification
// only under the terms of the DOOM Source Code License as
// published by id Software. All rights reserved.
//
// The source is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// FITNESS FOR A PARTICULAR PURPOSE. See the DOOM Source Code License
// for more details.
//
// $Log:$
//
// DESCRIPTION:
// The actual span/column drawing functions.
// Here find the main potential for optimization,
// e.g. inline assembly, different algorithms.
//
//-----------------------------------------------------------------------------
#include <stddef.h>
#include "templates.h"
#include "m_alloc.h"
#include "doomdef.h"
#include "i_system.h"
#include "w_wad.h"
#include "r_local.h"
#include "v_video.h"
#include "doomstat.h"
#include "st_stuff.h"
#include "a_hexenglobal.h"
#include "g_game.h"
#include "g_level.h"
#include "r_translate.h"
#include "gi.h"
#include "stats.h"
#include "x86.h"
#undef RANGECHECK
// status bar height at bottom of screen
// [RH] status bar position at bottom of screen
extern int ST_Y;
//
// All drawing to the view buffer is accomplished in this file.
// The other refresh files only know about ccordinates,
// not the architecture of the frame buffer.
// Conveniently, the frame buffer is a linear one,
// and we need only the base address,
// and the total size == width*height*depth/8.,
//
BYTE* viewimage;
extern "C" {
int viewwidth;
int halfviewwidth;
int viewheight;
int ylookup[MAXHEIGHT];
BYTE *dc_destorg;
}
int scaledviewwidth;
int viewwindowx;
int viewwindowy;
// [RH] Pointers to the different column drawers.
// These get changed depending on the current
// screen depth and asm/no asm.
void (*R_DrawColumnHoriz)(void);
void (*R_DrawColumn)(void);
void (*R_DrawFuzzColumn)(void);
void (*R_DrawTranslatedColumn)(void);
void (*R_DrawShadedColumn)(void);
void (*R_DrawSpan)(void);
void (*R_DrawSpanMasked)(void);
void (*R_DrawSpanTranslucent)(void);
void (*R_DrawSpanMaskedTranslucent)(void);
void (STACK_ARGS *rt_map4cols)(int,int,int);
//
// R_DrawColumn
// Source is the top of the column to scale.
//
extern "C" {
int dc_pitch=0xABadCafe; // [RH] Distance between rows
lighttable_t* dc_colormap;
int dc_x;
int dc_yl;
int dc_yh;
fixed_t dc_iscale;
fixed_t dc_texturemid;
fixed_t dc_texturefrac;
int dc_color; // [RH] Color for column filler
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
DWORD dc_srccolor;
DWORD *dc_srcblend; // [RH] Source and destination
DWORD *dc_destblend; // blending lookups
// first pixel in a column (possibly virtual)
const BYTE* dc_source;
BYTE* dc_dest;
int dc_count;
DWORD vplce[4];
DWORD vince[4];
BYTE* palookupoffse[4];
const BYTE* bufplce[4];
// just for profiling
int dccount;
}
int dc_fillcolor;
BYTE *dc_translation;
- Discovered that Shader Model 1.4 clamps my constants, so I can't use 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)
2007-12-26 04:42:15 +00:00
BYTE shadetables[NUMCOLORMAPS*16*256];
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
FDynamicColormap ShadeFakeColormap[16];
BYTE identitymap[256];
// Convert legacy render styles to flexible render styles.
const FRenderStyle LegacyRenderStyles[STYLE_Count] =
{
/* STYLE_None */ {{ STYLEOP_None, STYLEALPHA_Zero, STYLEALPHA_Zero, 0 }},
/* STYLE_Normal */ {{ STYLEOP_Add, STYLEALPHA_Src, STYLEALPHA_InvSrc, STYLEF_Alpha1 }},
/* STYLE_Fuzzy */ {{ STYLEOP_Fuzz, STYLEALPHA_Src, STYLEALPHA_InvSrc, 0 }},
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
/* STYLE_SoulTrans */ {{ STYLEOP_Add, STYLEALPHA_Src, STYLEALPHA_InvSrc, STYLEF_TransSoulsAlpha }},
/* STYLE_OptFuzzy */ {{ STYLEOP_FuzzOrAdd, STYLEALPHA_Src, STYLEALPHA_InvSrc, 0 }},
/* STYLE_Stencil */ {{ STYLEOP_Add, STYLEALPHA_Src, STYLEALPHA_InvSrc, STYLEF_Alpha1 | STYLEF_ColorIsFixed }},
/* STYLE_Translucent */ {{ STYLEOP_Add, STYLEALPHA_Src, STYLEALPHA_InvSrc, 0 }},
/* STYLE_Add */ {{ STYLEOP_Add, STYLEALPHA_Src, STYLEALPHA_One, 0 }},
/* STYLE_Shaded */ {{ STYLEOP_Add, STYLEALPHA_Src, STYLEALPHA_InvSrc, STYLEF_RedIsAlpha | STYLEF_ColorIsFixed }},
/* STYLE_TranslucentStencil */{{ STYLEOP_Add, STYLEALPHA_Src, STYLEALPHA_InvSrc, STYLEF_ColorIsFixed }},
};
EXTERN_CVAR (Int, r_columnmethod)
/************************************/
/* */
/* Palettized drawers (C versions) */
/* */
/************************************/
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#ifndef X86_ASM
//
// A column is a vertical slice/span from a wall texture that,
// given the DOOM style restrictions on the view orientation,
// will always have constant z depth.
// Thus a special case loop for very fast rendering can
// be used. It has also been used with Wolfenstein 3D.
//
void R_DrawColumnP_C (void)
{
int count;
BYTE* dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
// Zero length, column does not exceed a pixel.
if (count <= 0)
return;
// Framebuffer destination address.
dest = dc_dest;
// Determine scaling,
// which is the only mapping to be done.
fracstep = dc_iscale;
frac = dc_texturefrac;
{
// [RH] Get local copies of these variables so that the compiler
// has a better chance of optimizing this well.
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
// Inner loop that does the actual texture mapping,
// e.g. a DDA-lile scaling.
// This is as fast as it gets.
do
{
// Re-map color indices from wall texture column
// using a lighting/special effects LUT.
*dest = colormap[source[frac>>FRACBITS]];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#endif
// [RH] Just fills a column with a color
void R_FillColumnP (void)
{
int count;
BYTE* dest;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
{
int pitch = dc_pitch;
BYTE color = dc_color;
do
{
*dest = color;
dest += pitch;
} while (--count);
}
}
void R_FillAddColumn (void)
{
int count;
BYTE *dest;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
DWORD *bg2rgb;
DWORD fg;
bg2rgb = dc_destblend;
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
fg = dc_srccolor;
int pitch = dc_pitch;
do
{
DWORD bg;
bg = (fg + bg2rgb[*dest]) | 0x1f07c1f;
*dest = RGB32k[0][0][bg & (bg>>15)];
dest += pitch;
} while (--count);
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
void R_FillAddClampColumn (void)
{
int count;
BYTE *dest;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
DWORD *bg2rgb;
DWORD fg;
bg2rgb = dc_destblend;
fg = dc_srccolor;
int pitch = dc_pitch;
do
{
DWORD a = fg + bg2rgb[*dest];
DWORD b = a;
a |= 0x01f07c1f;
b &= 0x40100400;
a &= 0x3fffffff;
b = b - (b >> 5);
a |= b;
*dest = RGB32k[0][0][a & (a>>15)];
dest += pitch;
} while (--count);
}
void R_FillSubClampColumn (void)
{
int count;
BYTE *dest;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
DWORD *bg2rgb;
DWORD fg;
bg2rgb = dc_destblend;
fg = dc_srccolor | 0x40100400;
int pitch = dc_pitch;
do
{
DWORD a = fg - bg2rgb[*dest];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
*dest = RGB32k[0][0][a & (a>>15)];
dest += pitch;
} while (--count);
}
void R_FillRevSubClampColumn (void)
{
int count;
BYTE *dest;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
DWORD *bg2rgb;
DWORD fg;
bg2rgb = dc_destblend;
fg = dc_srccolor;
int pitch = dc_pitch;
do
{
DWORD a = (bg2rgb[*dest] | 0x40100400) - fg;
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
*dest = RGB32k[0][0][a & (a>>15)];
dest += pitch;
} while (--count);
}
//
// Spectre/Invisibility.
//
#define FUZZTABLE 50
extern "C"
{
int fuzzoffset[FUZZTABLE+1]; // [RH] +1 for the assembly routine
int fuzzpos = 0;
int fuzzviewheight;
}
/*
FUZZOFF,-FUZZOFF,FUZZOFF,-FUZZOFF,FUZZOFF,FUZZOFF,-FUZZOFF,
FUZZOFF,FUZZOFF,-FUZZOFF,FUZZOFF,FUZZOFF,FUZZOFF,-FUZZOFF,
FUZZOFF,FUZZOFF,FUZZOFF,-FUZZOFF,-FUZZOFF,-FUZZOFF,-FUZZOFF,
FUZZOFF,-FUZZOFF,-FUZZOFF,FUZZOFF,FUZZOFF,FUZZOFF,FUZZOFF,-FUZZOFF,
FUZZOFF,-FUZZOFF,FUZZOFF,FUZZOFF,-FUZZOFF,-FUZZOFF,FUZZOFF,
FUZZOFF,-FUZZOFF,-FUZZOFF,-FUZZOFF,-FUZZOFF,FUZZOFF,FUZZOFF,
FUZZOFF,FUZZOFF,-FUZZOFF,FUZZOFF,FUZZOFF,-FUZZOFF,FUZZOFF
*/
static const signed char fuzzinit[FUZZTABLE] = {
1,-1, 1,-1, 1, 1,-1,
1, 1,-1, 1, 1, 1,-1,
1, 1, 1,-1,-1,-1,-1,
1,-1,-1, 1, 1, 1, 1,-1,
1,-1, 1, 1,-1,-1, 1,
1,-1,-1,-1,-1, 1, 1,
1, 1,-1, 1, 1,-1, 1
};
void R_InitFuzzTable (int fuzzoff)
{
int i;
for (i = 0; i < FUZZTABLE; i++)
{
fuzzoffset[i] = fuzzinit[i] * fuzzoff;
}
}
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#ifndef X86_ASM
//
// Creates a fuzzy image by copying pixels from adjacent ones above and below.
// Used with an all black colormap, this could create the SHADOW effect,
// i.e. spectres and invisible players.
//
void R_DrawFuzzColumnP_C (void)
{
int count;
BYTE *dest;
// Adjust borders. Low...
if (dc_yl == 0)
dc_yl = 1;
// .. and high.
if (dc_yh > fuzzviewheight)
dc_yh = fuzzviewheight;
count = dc_yh - dc_yl;
// Zero length.
if (count < 0)
return;
count++;
dest = ylookup[dc_yl] + dc_x + dc_destorg;
// colormap #6 is used for shading (of 0-31, a bit brighter than average)
{
// [RH] Make local copies of global vars to try and improve
// the optimizations made by the compiler.
int pitch = dc_pitch;
int fuzz = fuzzpos;
int cnt;
BYTE *map = &NormalLight.Maps[6*256];
// [RH] Split this into three separate loops to minimize
// the number of times fuzzpos needs to be clamped.
if (fuzz)
{
cnt = MIN(FUZZTABLE-fuzz,count);
count -= cnt;
do
{
*dest = map[dest[fuzzoffset[fuzz++]]];
dest += pitch;
} while (--cnt);
}
if (fuzz == FUZZTABLE || count > 0)
{
while (count >= FUZZTABLE)
{
fuzz = 0;
cnt = FUZZTABLE;
count -= FUZZTABLE;
do
{
*dest = map[dest[fuzzoffset[fuzz++]]];
dest += pitch;
} while (--cnt);
}
fuzz = 0;
if (count > 0)
{
do
{
*dest = map[dest[fuzzoffset[fuzz++]]];
dest += pitch;
} while (--count);
}
}
fuzzpos = fuzz;
}
}
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#endif
//
// R_DrawTranlucentColumn
//
/*
[RH] This translucency algorithm is based on DOSDoom 0.65's, but uses
a 32k RGB table instead of an 8k one. At least on my machine, it's
slightly faster (probably because it uses only one shift instead of
two), and it looks considerably less green at the ends of the
translucency range. The extra size doesn't appear to be an issue.
The following note is from DOSDoom 0.65:
New translucency algorithm, by Erik Sandberg:
Basically, we compute the red, green and blue values for each pixel, and
then use a RGB table to check which one of the palette colours that best
represents those RGB values. The RGB table is 8k big, with 4 R-bits,
5 G-bits and 4 B-bits. A 4k table gives a bit too bad precision, and a 32k
table takes up more memory and results in more cache misses, so an 8k
table seemed to be quite ultimate.
The computation of the RGB for each pixel is accelerated by using two
1k tables for each translucency level.
The xth element of one of these tables contains the r, g and b values for
the colour x, weighted for the current translucency level (for example,
the weighted rgb values for background colour at 75% translucency are 1/4
of the original rgb values). The rgb values are stored as three
low-precision fixed point values, packed into one long per colour:
Bit 0-4: Frac part of blue (5 bits)
Bit 5-8: Int part of blue (4 bits)
Bit 9-13: Frac part of red (5 bits)
Bit 14-17: Int part of red (4 bits)
Bit 18-22: Frac part of green (5 bits)
Bit 23-27: Int part of green (5 bits)
Bit 28-31: All zeros (4 bits)
The point of this format is that the two colours now can be added, and
then be converted to a RGB table index very easily: First, we just set
all the frac bits and the four upper zero bits to 1. It's now possible
to get the RGB table index by anding the current value >> 5 with the
current value >> 19. When asm-optimised, this should be the fastest
algorithm that uses RGB tables.
*/
void R_DrawAddColumnP_C (void)
{
int count;
BYTE *dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
do
{
DWORD fg = colormap[source[frac>>FRACBITS]];
DWORD bg = *dest;
fg = fg2rgb[fg];
bg = bg2rgb[bg];
fg = (fg+bg) | 0x1f07c1f;
*dest = RGB32k[0][0][fg & (fg>>15)];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
//
// R_DrawTranslatedColumn
// Used to draw player sprites with the green colorramp mapped to others.
// Could be used with different translation tables, e.g. the lighter colored
// version of the BaronOfHell, the HellKnight, uses identical sprites, kinda
// brightened up.
//
void R_DrawTranslatedColumnP_C (void)
{
int count;
BYTE* dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
// [RH] Local copies of global vars to improve compiler optimizations
BYTE *colormap = dc_colormap;
BYTE *translation = dc_translation;
const BYTE *source = dc_source;
int pitch = dc_pitch;
do
{
*dest = colormap[translation[source[frac>>FRACBITS]]];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
// Draw a column that is both translated and translucent
void R_DrawTlatedAddColumnP_C (void)
{
int count;
BYTE *dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
BYTE *translation = dc_translation;
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
do
{
DWORD fg = colormap[translation[source[frac>>FRACBITS]]];
DWORD bg = *dest;
fg = fg2rgb[fg];
bg = bg2rgb[bg];
fg = (fg+bg) | 0x1f07c1f;
*dest = RGB32k[0][0][fg & (fg>>15)];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
// Draw a column whose "color" values are actually translucency
// levels for a base color stored in dc_color.
void R_DrawShadedColumnP_C (void)
{
int count;
BYTE *dest;
fixed_t frac, fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
const BYTE *source = dc_source;
BYTE *colormap = dc_colormap;
int pitch = dc_pitch;
DWORD *fgstart = &Col2RGB8[0][dc_color];
do
{
DWORD val = colormap[source[frac>>FRACBITS]];
DWORD fg = fgstart[val<<8];
val = (Col2RGB8[64-val][*dest] + fg) | 0x1f07c1f;
*dest = RGB32k[0][0][val & (val>>15)];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
// Add source to destination, clamping it to white
void R_DrawAddClampColumnP_C ()
{
int count;
BYTE *dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
DWORD a = fg2rgb[colormap[source[frac>>FRACBITS]]] + bg2rgb[*dest];
DWORD b = a;
a |= 0x01f07c1f;
b &= 0x40100400;
a &= 0x3fffffff;
b = b - (b >> 5);
a |= b;
*dest = RGB32k[0][0][a & (a>>15)];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
// Add translated source to destination, clamping it to white
void R_DrawAddClampTranslatedColumnP_C ()
{
int count;
BYTE *dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
BYTE *translation = dc_translation;
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
DWORD a = fg2rgb[colormap[translation[source[frac>>FRACBITS]]]] + bg2rgb[*dest];
DWORD b = a;
a |= 0x01f07c1f;
b &= 0x40100400;
a &= 0x3fffffff;
b = b - (b >> 5);
a |= b;
*dest = RGB32k[0][0][(a>>15) & a];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
// Subtract destination from source, clamping it to black
void R_DrawSubClampColumnP_C ()
{
int count;
BYTE *dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
DWORD a = (fg2rgb[colormap[source[frac>>FRACBITS]]] | 0x40100400) - bg2rgb[*dest];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
*dest = RGB32k[0][0][a & (a>>15)];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
// Subtract destination from source, clamping it to black
void R_DrawSubClampTranslatedColumnP_C ()
{
int count;
BYTE *dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
BYTE *translation = dc_translation;
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
DWORD a = (fg2rgb[colormap[translation[source[frac>>FRACBITS]]]] | 0x40100400) - bg2rgb[*dest];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
*dest = RGB32k[0][0][(a>>15) & a];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
// Subtract source from destination, clamping it to black
void R_DrawRevSubClampColumnP_C ()
{
int count;
BYTE *dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
DWORD a = (bg2rgb[*dest] | 0x40100400) - fg2rgb[colormap[source[frac>>FRACBITS]]];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
*dest = RGB32k[0][0][a & (a>>15)];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
// Subtract source from destination, clamping it to black
void R_DrawRevSubClampTranslatedColumnP_C ()
{
int count;
BYTE *dest;
fixed_t frac;
fixed_t fracstep;
count = dc_count;
if (count <= 0)
return;
dest = dc_dest;
fracstep = dc_iscale;
frac = dc_texturefrac;
{
BYTE *translation = dc_translation;
BYTE *colormap = dc_colormap;
const BYTE *source = dc_source;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
DWORD a = (bg2rgb[*dest] | 0x40100400) - fg2rgb[colormap[translation[source[frac>>FRACBITS]]]];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
*dest = RGB32k[0][0][(a>>15) & a];
dest += pitch;
frac += fracstep;
} while (--count);
}
}
//
// R_DrawSpan
// With DOOM style restrictions on view orientation,
// the floors and ceilings consist of horizontal slices
// or spans with constant z depth.
// However, rotation around the world z axis is possible,
// thus this mapping, while simpler and faster than
// perspective correct texture mapping, has to traverse
// the texture at an angle in all but a few cases.
// In consequence, flats are not stored by column (like walls),
// and the inner loop has to step in texture space u and v.
//
// [RH] I'm not sure who wrote this, but floor/ceiling mapping
// *is* perspective correct for spans of constant z depth, which
// Doom guarantees because it does not let you change your pitch.
// Also, because of the new texture system, flats *are* stored by
// column to make it easy to use them on walls too. To accomodate
// this, the use of x/u and y/v in R_DrawSpan just needs to be
// swapped.
//
extern "C" {
int ds_color; // [RH] color for non-textured spans
int ds_y;
int ds_x1;
int ds_x2;
lighttable_t* ds_colormap;
dsfixed_t ds_xfrac;
dsfixed_t ds_yfrac;
dsfixed_t ds_xstep;
dsfixed_t ds_ystep;
int ds_xbits;
int ds_ybits;
// start of a floor/ceiling tile image
const BYTE* ds_source;
// just for profiling
int dscount;
}
//
// Draws the actual span.
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#ifndef X86_ASM
void R_DrawSpanP_C (void)
{
dsfixed_t xfrac;
dsfixed_t yfrac;
dsfixed_t xstep;
dsfixed_t ystep;
BYTE* dest;
const BYTE* source = ds_source;
const BYTE* colormap = ds_colormap;
int count;
int spot;
#ifdef RANGECHECK
if (ds_x2 < ds_x1 || ds_x1 < 0
|| ds_x2 >= screen->width || ds_y > screen->height)
{
I_Error ("R_DrawSpan: %i to %i at %i", ds_x1, ds_x2, ds_y);
}
// dscount++;
#endif
xfrac = ds_xfrac;
yfrac = ds_yfrac;
dest = ylookup[ds_y] + ds_x1 + dc_destorg;
count = ds_x2 - ds_x1 + 1;
xstep = ds_xstep;
ystep = ds_ystep;
if (ds_xbits == 6 && ds_ybits == 6)
{
// 64x64 is the most common case by far, so special case it.
do
{
// Current texture index in u,v.
spot = ((xfrac>>(32-6-6))&(63*64)) + (yfrac>>(32-6));
// Lookup pixel from flat texture tile,
// re-index using light/colormap.
*dest++ = colormap[source[spot]];
// Next step in u,v.
xfrac += xstep;
yfrac += ystep;
} while (--count);
}
else
{
BYTE yshift = 32 - ds_ybits;
BYTE xshift = yshift - ds_xbits;
int xmask = ((1 << ds_xbits) - 1) << ds_ybits;
do
{
// Current texture index in u,v.
spot = ((xfrac >> xshift) & xmask) + (yfrac >> yshift);
// Lookup pixel from flat texture tile,
// re-index using light/colormap.
*dest++ = colormap[source[spot]];
// Next step in u,v.
xfrac += xstep;
yfrac += ystep;
} while (--count);
}
}
// [RH] Draw a span with holes
void R_DrawSpanMaskedP_C (void)
{
dsfixed_t xfrac;
dsfixed_t yfrac;
dsfixed_t xstep;
dsfixed_t ystep;
BYTE* dest;
const BYTE* source = ds_source;
const BYTE* colormap = ds_colormap;
int count;
int spot;
xfrac = ds_xfrac;
yfrac = ds_yfrac;
dest = ylookup[ds_y] + ds_x1 + dc_destorg;
count = ds_x2 - ds_x1 + 1;
xstep = ds_xstep;
ystep = ds_ystep;
if (ds_xbits == 6 && ds_ybits == 6)
{
// 64x64 is the most common case by far, so special case it.
do
{
BYTE texdata;
spot = ((xfrac>>(32-6-6))&(63*64)) + (yfrac>>(32-6));
texdata = source[spot];
if (texdata != 0)
{
*dest = colormap[texdata];
}
dest++;
xfrac += xstep;
yfrac += ystep;
} while (--count);
}
else
{
BYTE yshift = 32 - ds_ybits;
BYTE xshift = yshift - ds_xbits;
int xmask = ((1 << ds_xbits) - 1) << ds_ybits;
do
{
BYTE texdata;
spot = ((xfrac >> xshift) & xmask) + (yfrac >> yshift);
texdata = source[spot];
if (texdata != 0)
{
*dest = colormap[texdata];
}
dest++;
xfrac += xstep;
yfrac += ystep;
} while (--count);
}
}
#endif
void R_DrawSpanTranslucentP_C (void)
{
dsfixed_t xfrac;
dsfixed_t yfrac;
dsfixed_t xstep;
dsfixed_t ystep;
BYTE* dest;
const BYTE* source = ds_source;
const BYTE* colormap = ds_colormap;
int count;
int spot;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
xfrac = ds_xfrac;
yfrac = ds_yfrac;
dest = ylookup[ds_y] + ds_x1 + dc_destorg;
count = ds_x2 - ds_x1 + 1;
xstep = ds_xstep;
ystep = ds_ystep;
if (ds_xbits == 6 && ds_ybits == 6)
{
// 64x64 is the most common case by far, so special case it.
do
{
spot = ((xfrac>>(32-6-6))&(63*64)) + (yfrac>>(32-6));
DWORD fg = colormap[source[spot]];
DWORD bg = *dest;
fg = fg2rgb[fg];
bg = bg2rgb[bg];
fg = (fg+bg) | 0x1f07c1f;
*dest++ = RGB32k[0][0][fg & (fg>>15)];
xfrac += xstep;
yfrac += ystep;
} while (--count);
}
else
{
BYTE yshift = 32 - ds_ybits;
BYTE xshift = yshift - ds_xbits;
int xmask = ((1 << ds_xbits) - 1) << ds_ybits;
do
{
spot = ((xfrac >> xshift) & xmask) + (yfrac >> yshift);
DWORD fg = colormap[source[spot]];
DWORD bg = *dest;
fg = fg2rgb[fg];
bg = bg2rgb[bg];
fg = (fg+bg) | 0x1f07c1f;
*dest++ = RGB32k[0][0][fg & (fg>>15)];
xfrac += xstep;
yfrac += ystep;
} while (--count);
}
}
void R_DrawSpanMaskedTranslucentP_C (void)
{
dsfixed_t xfrac;
dsfixed_t yfrac;
dsfixed_t xstep;
dsfixed_t ystep;
BYTE* dest;
const BYTE* source = ds_source;
const BYTE* colormap = ds_colormap;
int count;
int spot;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
xfrac = ds_xfrac;
yfrac = ds_yfrac;
dest = ylookup[ds_y] + ds_x1 + dc_destorg;
count = ds_x2 - ds_x1 + 1;
xstep = ds_xstep;
ystep = ds_ystep;
if (ds_xbits == 6 && ds_ybits == 6)
{
// 64x64 is the most common case by far, so special case it.
do
{
BYTE texdata;
spot = ((xfrac>>(32-6-6))&(63*64)) + (yfrac>>(32-6));
texdata = source[spot];
if (texdata != 0)
{
DWORD fg = colormap[texdata];
DWORD bg = *dest;
fg = fg2rgb[fg];
bg = bg2rgb[bg];
fg = (fg+bg) | 0x1f07c1f;
*dest = RGB32k[0][0][fg & (fg>>15)];
}
dest++;
xfrac += xstep;
yfrac += ystep;
} while (--count);
}
else
{
BYTE yshift = 32 - ds_ybits;
BYTE xshift = yshift - ds_xbits;
int xmask = ((1 << ds_xbits) - 1) << ds_ybits;
do
{
BYTE texdata;
spot = ((xfrac >> xshift) & xmask) + (yfrac >> yshift);
texdata = source[spot];
if (texdata != 0)
{
DWORD fg = colormap[texdata];
DWORD bg = *dest;
fg = fg2rgb[fg];
bg = bg2rgb[bg];
fg = (fg+bg) | 0x1f07c1f;
*dest = RGB32k[0][0][fg & (fg>>15)];
}
dest++;
xfrac += xstep;
yfrac += ystep;
} while (--count);
}
}
// [RH] Just fill a span with a color
void R_FillSpan (void)
{
memset (ylookup[ds_y] + ds_x1 + dc_destorg, ds_color, ds_x2 - ds_x1 + 1);
}
/****************************************************/
/****************************************************/
// wallscan stuff, in C
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#ifndef X86_ASM
static DWORD STACK_ARGS vlinec1 ();
static int vlinebits;
DWORD (STACK_ARGS *dovline1)() = vlinec1;
DWORD (STACK_ARGS *doprevline1)() = vlinec1;
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#ifdef X64_ASM
extern "C" void vlinetallasm4();
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#define dovline4 vlinetallasm4
extern "C" void setupvlinetallasm (int);
#else
static void STACK_ARGS vlinec4 ();
void (STACK_ARGS *dovline4)() = vlinec4;
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#endif
static DWORD STACK_ARGS mvlinec1();
static void STACK_ARGS mvlinec4();
static int mvlinebits;
DWORD (STACK_ARGS *domvline1)() = mvlinec1;
void (STACK_ARGS *domvline4)() = mvlinec4;
#else
extern "C"
{
DWORD STACK_ARGS vlineasm1 ();
DWORD STACK_ARGS prevlineasm1 ();
DWORD STACK_ARGS vlinetallasm1 ();
DWORD STACK_ARGS prevlinetallasm1 ();
void STACK_ARGS vlineasm4 ();
void STACK_ARGS vlinetallasmathlon4 ();
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
void STACK_ARGS vlinetallasm4 ();
void STACK_ARGS setupvlineasm (int);
void STACK_ARGS setupvlinetallasm (int);
DWORD STACK_ARGS mvlineasm1();
void STACK_ARGS mvlineasm4();
void STACK_ARGS setupmvlineasm (int);
}
DWORD (STACK_ARGS *dovline1)() = vlinetallasm1;
DWORD (STACK_ARGS *doprevline1)() = prevlinetallasm1;
void (STACK_ARGS *dovline4)() = vlinetallasm4;
DWORD (STACK_ARGS *domvline1)() = mvlineasm1;
void (STACK_ARGS *domvline4)() = mvlineasm4;
#endif
void setupvline (int fracbits)
{
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#ifdef X86_ASM
if (CPU.Family <= 5)
{
if (fracbits >= 24)
{
setupvlineasm (fracbits);
dovline4 = vlineasm4;
dovline1 = vlineasm1;
doprevline1 = prevlineasm1;
}
else
{
setupvlinetallasm (fracbits);
dovline1 = vlinetallasm1;
doprevline1 = prevlinetallasm1;
dovline4 = vlinetallasm4;
}
}
else
{
setupvlinetallasm (fracbits);
if (CPU.bIsAMD && CPU.AMDFamily >= 7)
{
dovline4 = vlinetallasmathlon4;
}
}
#else
vlinebits = fracbits;
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#ifdef X64_ASM
setupvlinetallasm(fracbits);
#endif
#endif
}
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#if !defined(X86_ASM)
DWORD STACK_ARGS vlinec1 ()
{
DWORD fracstep = dc_iscale;
DWORD frac = dc_texturefrac;
BYTE *colormap = dc_colormap;
int count = dc_count;
const BYTE *source = dc_source;
BYTE *dest = dc_dest;
int bits = vlinebits;
int pitch = dc_pitch;
do
{
*dest = colormap[source[frac>>bits]];
frac += fracstep;
dest += pitch;
} while (--count);
return frac;
}
void STACK_ARGS vlinec4 ()
{
BYTE *dest = dc_dest;
int count = dc_count;
int bits = vlinebits;
DWORD place;
do
{
dest[0] = palookupoffse[0][bufplce[0][(place=vplce[0])>>bits]]; vplce[0] = place+vince[0];
dest[1] = palookupoffse[1][bufplce[1][(place=vplce[1])>>bits]]; vplce[1] = place+vince[1];
dest[2] = palookupoffse[2][bufplce[2][(place=vplce[2])>>bits]]; vplce[2] = place+vince[2];
dest[3] = palookupoffse[3][bufplce[3][(place=vplce[3])>>bits]]; vplce[3] = place+vince[3];
dest += dc_pitch;
} while (--count);
}
#endif
void setupmvline (int fracbits)
{
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#if defined(X86_ASM)
setupmvlineasm (fracbits);
domvline1 = mvlineasm1;
domvline4 = mvlineasm4;
#else
mvlinebits = fracbits;
#endif
}
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#if !defined(X86_ASM)
DWORD STACK_ARGS mvlinec1 ()
{
DWORD fracstep = dc_iscale;
DWORD frac = dc_texturefrac;
BYTE *colormap = dc_colormap;
int count = dc_count;
const BYTE *source = dc_source;
BYTE *dest = dc_dest;
int bits = mvlinebits;
int pitch = dc_pitch;
do
{
BYTE pix = source[frac>>bits];
if (pix != 0)
{
*dest = colormap[pix];
}
frac += fracstep;
dest += pitch;
} while (--count);
return frac;
}
void STACK_ARGS mvlinec4 ()
{
BYTE *dest = dc_dest;
int count = dc_count;
int bits = mvlinebits;
DWORD place;
do
{
BYTE pix;
pix = bufplce[0][(place=vplce[0])>>bits]; if(pix) dest[0] = palookupoffse[0][pix]; vplce[0] = place+vince[0];
pix = bufplce[1][(place=vplce[1])>>bits]; if(pix) dest[1] = palookupoffse[1][pix]; vplce[1] = place+vince[1];
pix = bufplce[2][(place=vplce[2])>>bits]; if(pix) dest[2] = palookupoffse[2][pix]; vplce[2] = place+vince[2];
pix = bufplce[3][(place=vplce[3])>>bits]; if(pix) dest[3] = palookupoffse[3][pix]; vplce[3] = place+vince[3];
dest += dc_pitch;
} while (--count);
}
#endif
extern "C" short spanend[MAXHEIGHT];
extern fixed_t rw_light;
extern fixed_t rw_lightstep;
extern int wallshade;
static void R_DrawFogBoundarySection (int y, int y2, int x1)
{
BYTE *colormap = dc_colormap;
BYTE *dest = ylookup[y] + dc_destorg;
for (; y < y2; ++y)
{
int x2 = spanend[y];
int x = x1;
do
{
dest[x] = colormap[dest[x]];
} while (++x <= x2);
dest += dc_pitch;
}
}
static void R_DrawFogBoundaryLine (int y, int x)
{
int x2 = spanend[y];
BYTE *colormap = dc_colormap;
BYTE *dest = ylookup[y] + dc_destorg;
do
{
dest[x] = colormap[dest[x]];
} while (++x <= x2);
}
void R_DrawFogBoundary (int x1, int x2, short *uclip, short *dclip)
{
// This is essentially the same as R_MapVisPlane but with an extra step
// to create new horizontal spans whenever the light changes enough that
// we need to use a new colormap.
fixed_t lightstep = rw_lightstep;
fixed_t light = rw_light+lightstep*(x2-x1);
int x = x2;
int t2 = uclip[x];
int b2 = dclip[x];
int rcolormap = GETPALOOKUP (light, wallshade);
int lcolormap;
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
BYTE *basecolormapdata = basecolormap->Maps;
if (b2 > t2)
{
clearbufshort (spanend+t2, b2-t2, x);
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
dc_colormap = basecolormapdata + (rcolormap << COLORMAPSHIFT);
for (--x; x >= x1; --x)
{
int t1 = uclip[x];
int b1 = dclip[x];
const int xr = x+1;
int stop;
light -= rw_lightstep;
lcolormap = GETPALOOKUP (light, wallshade);
if (lcolormap != rcolormap)
{
if (t2 < b2 && rcolormap != 0)
{ // Colormap 0 is always the identity map, so rendering it is
// just a waste of time.
R_DrawFogBoundarySection (t2, b2, xr);
}
if (t1 < t2) t2 = t1;
if (b1 > b2) b2 = b1;
if (t2 < b2)
{
clearbufshort (spanend+t2, b2-t2, x);
}
rcolormap = lcolormap;
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
dc_colormap = basecolormapdata + (lcolormap << COLORMAPSHIFT);
}
else
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
if (dc_colormap != basecolormapdata)
{
stop = MIN (t1, b2);
while (t2 < stop)
{
R_DrawFogBoundaryLine (t2++, xr);
}
stop = MAX (b1, t2);
while (b2 > stop)
{
R_DrawFogBoundaryLine (--b2, xr);
}
}
else
{
t2 = MAX (t2, MIN (t1, b2));
b2 = MIN (b2, MAX (b1, t2));
}
stop = MIN (t2, b1);
while (t1 < stop)
{
spanend[t1++] = x;
}
stop = MAX (b2, t2);
while (b1 > stop)
{
spanend[--b1] = x;
}
}
t2 = uclip[x];
b2 = dclip[x];
}
if (t2 < b2 && rcolormap != 0)
{
R_DrawFogBoundarySection (t2, b2, x1);
}
}
int tmvlinebits;
void setuptmvline (int bits)
{
tmvlinebits = bits;
}
fixed_t tmvline1_add ()
{
DWORD fracstep = dc_iscale;
DWORD frac = dc_texturefrac;
BYTE *colormap = dc_colormap;
int count = dc_count;
const BYTE *source = dc_source;
BYTE *dest = dc_dest;
int bits = tmvlinebits;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
BYTE pix = source[frac>>bits];
if (pix != 0)
{
DWORD fg = fg2rgb[colormap[pix]];
DWORD bg = bg2rgb[*dest];
fg = (fg+bg) | 0x1f07c1f;
*dest = RGB32k[0][0][fg & (fg>>15)];
}
frac += fracstep;
dest += pitch;
} while (--count);
return frac;
}
void tmvline4_add ()
{
BYTE *dest = dc_dest;
int count = dc_count;
int bits = tmvlinebits;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
for (int i = 0; i < 4; ++i)
{
BYTE pix = bufplce[i][vplce[i] >> bits];
if (pix != 0)
{
DWORD fg = fg2rgb[palookupoffse[i][pix]];
DWORD bg = bg2rgb[dest[i]];
fg = (fg+bg) | 0x1f07c1f;
dest[i] = RGB32k[0][0][fg & (fg>>15)];
}
vplce[i] += vince[i];
}
dest += dc_pitch;
} while (--count);
}
fixed_t tmvline1_addclamp ()
{
DWORD fracstep = dc_iscale;
DWORD frac = dc_texturefrac;
BYTE *colormap = dc_colormap;
int count = dc_count;
const BYTE *source = dc_source;
BYTE *dest = dc_dest;
int bits = tmvlinebits;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
BYTE pix = source[frac>>bits];
if (pix != 0)
{
DWORD a = fg2rgb[colormap[pix]] + bg2rgb[*dest];
DWORD b = a;
a |= 0x01f07c1f;
b &= 0x40100400;
a &= 0x3fffffff;
b = b - (b >> 5);
a |= b;
*dest = RGB32k[0][0][a & (a>>15)];
}
frac += fracstep;
dest += pitch;
} while (--count);
return frac;
}
void tmvline4_addclamp ()
{
BYTE *dest = dc_dest;
int count = dc_count;
int bits = tmvlinebits;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
for (int i = 0; i < 4; ++i)
{
BYTE pix = bufplce[i][vplce[i] >> bits];
if (pix != 0)
{
DWORD a = fg2rgb[palookupoffse[i][pix]] + bg2rgb[dest[i]];
DWORD b = a;
a |= 0x01f07c1f;
b &= 0x40100400;
a &= 0x3fffffff;
b = b - (b >> 5);
a |= b;
dest[i] = RGB32k[0][0][a & (a>>15)];
}
vplce[i] += vince[i];
}
dest += dc_pitch;
} while (--count);
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
fixed_t tmvline1_subclamp ()
{
DWORD fracstep = dc_iscale;
DWORD frac = dc_texturefrac;
BYTE *colormap = dc_colormap;
int count = dc_count;
const BYTE *source = dc_source;
BYTE *dest = dc_dest;
int bits = tmvlinebits;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
BYTE pix = source[frac>>bits];
if (pix != 0)
{
DWORD a = (fg2rgb[colormap[pix]] | 0x40100400) - bg2rgb[*dest];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
*dest = RGB32k[0][0][a & (a>>15)];
}
frac += fracstep;
dest += pitch;
} while (--count);
return frac;
}
void tmvline4_subclamp ()
{
BYTE *dest = dc_dest;
int count = dc_count;
int bits = tmvlinebits;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
for (int i = 0; i < 4; ++i)
{
BYTE pix = bufplce[i][vplce[i] >> bits];
if (pix != 0)
{
DWORD a = (fg2rgb[palookupoffse[i][pix]] | 0x40100400) - bg2rgb[dest[i]];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
dest[i] = RGB32k[0][0][a & (a>>15)];
}
vplce[i] += vince[i];
}
dest += dc_pitch;
} while (--count);
}
fixed_t tmvline1_revsubclamp ()
{
DWORD fracstep = dc_iscale;
DWORD frac = dc_texturefrac;
BYTE *colormap = dc_colormap;
int count = dc_count;
const BYTE *source = dc_source;
BYTE *dest = dc_dest;
int bits = tmvlinebits;
int pitch = dc_pitch;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
BYTE pix = source[frac>>bits];
if (pix != 0)
{
DWORD a = (bg2rgb[*dest] | 0x40100400) - fg2rgb[colormap[pix]];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
*dest = RGB32k[0][0][a & (a>>15)];
}
frac += fracstep;
dest += pitch;
} while (--count);
return frac;
}
void tmvline4_revsubclamp ()
{
BYTE *dest = dc_dest;
int count = dc_count;
int bits = tmvlinebits;
DWORD *fg2rgb = dc_srcblend;
DWORD *bg2rgb = dc_destblend;
do
{
for (int i = 0; i < 4; ++i)
{
BYTE pix = bufplce[i][vplce[i] >> bits];
if (pix != 0)
{
DWORD a = (bg2rgb[dest[i]] | 0x40100400) - fg2rgb[palookupoffse[i][pix]];
DWORD b = a;
b &= 0x40100400;
b = b - (b >> 5);
a &= b;
a |= 0x01f07c1f;
dest[i] = RGB32k[0][0][a & (a>>15)];
}
vplce[i] += vince[i];
}
dest += dc_pitch;
} while (--count);
}
void R_DrawBorder (int x1, int y1, int x2, int y2)
{
FTextureID picnum;
if (level.info != NULL)
{
picnum = TexMan.CheckForTexture (level.info->bordertexture, FTexture::TEX_Flat);
}
else
{
picnum = TexMan.CheckForTexture (gameinfo.borderFlat, FTexture::TEX_Flat);
}
if (picnum.isValid())
{
screen->FlatFill (x1, y1, x2, y2, TexMan(picnum));
}
else
{
screen->Clear (x1, y1, x2, y2, 0, 0);
}
}
/*
==================
=
= R_DrawViewBorder
=
= Draws the border around the view for different size windows
==================
*/
int BorderNeedRefresh;
void V_MarkRect (int x, int y, int width, int height);
void M_DrawFrame (int x, int y, int width, int height);
void R_DrawViewBorder (void)
{
// [RH] Redraw the status bar if SCREENWIDTH > status bar width.
// Will draw borders around itself, too.
if (SCREENWIDTH > 320)
{
SB_state = screen->GetPageCount ();
}
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
if (viewwidth == SCREENWIDTH)
{
return;
}
R_DrawBorder (0, 0, SCREENWIDTH, viewwindowy);
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
R_DrawBorder (0, viewwindowy, viewwindowx, viewheight + viewwindowy);
R_DrawBorder (viewwindowx + viewwidth, viewwindowy, SCREENWIDTH, viewheight + viewwindowy);
R_DrawBorder (0, viewwindowy + viewheight, SCREENWIDTH, ST_Y);
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
M_DrawFrame (viewwindowx, viewwindowy, viewwidth, viewheight);
V_MarkRect (0, 0, SCREENWIDTH, ST_Y);
}
/*
==================
=
= R_DrawTopBorder
=
= Draws the top border around the view for different size windows
==================
*/
int BorderTopRefresh;
void R_DrawTopBorder ()
{
FTexture *p;
int offset;
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
if (viewwidth == SCREENWIDTH)
return;
offset = gameinfo.border->offset;
- Tried adding bilinear filtering support for paletted textures, but the 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)
2008-01-03 05:39:36 +00:00
if (viewwindowy < 34)
{
- Tried adding bilinear filtering support for paletted textures, but the 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)
2008-01-03 05:39:36 +00:00
R_DrawBorder (0, 0, viewwindowx, 34);
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
R_DrawBorder (viewwindowx, 0, viewwindowx + viewwidth, viewwindowy);
R_DrawBorder (viewwindowx + viewwidth, 0, SCREENWIDTH, 34);
p = TexMan(gameinfo.border->t);
screen->FlatFill(viewwindowx, viewwindowy - p->GetHeight(),
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
viewwindowx + viewwidth, viewwindowy, p, true);
p = TexMan(gameinfo.border->l);
screen->FlatFill(viewwindowx - p->GetWidth(), viewwindowy,
viewwindowx, 35, p, true);
p = TexMan(gameinfo.border->r);
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
screen->FlatFill(viewwindowx + viewwidth, viewwindowy,
viewwindowx + viewwidth + p->GetWidth(), 35, p, true);
p = TexMan(gameinfo.border->tl);
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
screen->DrawTexture (p, viewwindowx - offset, viewwindowy - offset, TAG_DONE);
p = TexMan(gameinfo.border->tr);
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
screen->DrawTexture (p, viewwindowx + viewwidth, viewwindowy - offset, TAG_DONE);
}
- Tried adding bilinear filtering support for paletted textures, but the 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)
2008-01-03 05:39:36 +00:00
else
{
R_DrawBorder (0, 0, SCREENWIDTH, 34);
}
}
// [RH] Initialize the column drawer pointers
void R_InitColumnDrawers ()
{
- Ported vlinetallasm4 to AMD64 assembly. Even with the increased number of 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)
2008-08-09 03:13:43 +00:00
#ifdef X86_ASM
R_DrawColumn = R_DrawColumnP_ASM;
R_DrawColumnHoriz = R_DrawColumnHorizP_ASM;
R_DrawFuzzColumn = R_DrawFuzzColumnP_ASM;
R_DrawTranslatedColumn = R_DrawTranslatedColumnP_C;
R_DrawShadedColumn = R_DrawShadedColumnP_C;
R_DrawSpan = R_DrawSpanP_ASM;
R_DrawSpanMasked = R_DrawSpanMaskedP_ASM;
if (CPU.Family <= 5)
{
rt_map4cols = rt_map4cols_asm2;
}
else
{
rt_map4cols = rt_map4cols_asm1;
}
#else
R_DrawColumnHoriz = R_DrawColumnHorizP_C;
R_DrawColumn = R_DrawColumnP_C;
R_DrawFuzzColumn = R_DrawFuzzColumnP_C;
R_DrawTranslatedColumn = R_DrawTranslatedColumnP_C;
R_DrawShadedColumn = R_DrawShadedColumnP_C;
R_DrawSpan = R_DrawSpanP_C;
R_DrawSpanMasked = R_DrawSpanMaskedP_C;
rt_map4cols = rt_map4cols_c;
#endif
R_DrawSpanTranslucent = R_DrawSpanTranslucentP_C;
R_DrawSpanMaskedTranslucent = R_DrawSpanMaskedTranslucentP_C;
}
// [RH] Choose column drawers in a single place
EXTERN_CVAR (Bool, r_drawfuzz)
EXTERN_CVAR (Float, transsouls)
CVAR (Bool, r_drawtrans, true, 0)
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
static FDynamicColormap *basecolormapsave;
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
static bool R_SetBlendFunc (int op, fixed_t fglevel, fixed_t bglevel, int flags)
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
// r_drawtrans is a seriously bad thing to turn off. I wonder if I should
// just remove it completely.
if (!r_drawtrans || (op == STYLEOP_Add && fglevel == FRACUNIT && bglevel == 0 && !(flags & STYLEF_InvertSource)))
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
if (flags & STYLEF_ColorIsFixed)
{
colfunc = R_FillColumnP;
hcolfunc_post1 = rt_copy1col;
hcolfunc_post4 = rt_copy4cols;
}
else if (dc_translation == NULL)
{
colfunc = basecolfunc;
hcolfunc_post1 = rt_map1col;
hcolfunc_post4 = rt_map4cols;
}
else
{
colfunc = transcolfunc;
hcolfunc_post1 = rt_tlate1col;
hcolfunc_post4 = rt_tlate4cols;
}
return true;
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
if (flags & STYLEF_InvertSource)
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
dc_srcblend = Col2RGB8_Inverse[fglevel>>10];
dc_destblend = Col2RGB8_LessPrecision[bglevel>>10];
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
else if (op == STYLEOP_Add && fglevel + bglevel <= FRACUNIT)
{
dc_srcblend = Col2RGB8[fglevel>>10];
dc_destblend = Col2RGB8[bglevel>>10];
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
}
else
{
dc_srcblend = Col2RGB8_LessPrecision[fglevel>>10];
dc_destblend = Col2RGB8_LessPrecision[bglevel>>10];
}
switch (op)
{
case STYLEOP_Add:
if (fglevel == 0 && bglevel == FRACUNIT)
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
return false;
}
if (fglevel + bglevel <= FRACUNIT)
{ // Colors won't overflow when added
if (flags & STYLEF_ColorIsFixed)
{
colfunc = R_FillAddColumn;
hcolfunc_post1 = rt_add1col;
hcolfunc_post4 = rt_add4cols;
}
else if (dc_translation == NULL)
{
colfunc = R_DrawAddColumnP_C;
hcolfunc_post1 = rt_add1col;
hcolfunc_post4 = rt_add4cols;
}
else
{
colfunc = R_DrawTlatedAddColumnP_C;
hcolfunc_post1 = rt_tlateadd1col;
hcolfunc_post4 = rt_tlateadd4cols;
}
}
else
{ // Colors might overflow when added
if (flags & STYLEF_ColorIsFixed)
{
colfunc = R_FillAddClampColumn;
hcolfunc_post1 = rt_addclamp1col;
hcolfunc_post4 = rt_addclamp4cols;
}
else if (dc_translation == NULL)
{
colfunc = R_DrawAddClampColumnP_C;
hcolfunc_post1 = rt_addclamp1col;
hcolfunc_post4 = rt_addclamp4cols;
}
else
{
colfunc = R_DrawAddClampTranslatedColumnP_C;
hcolfunc_post1 = rt_tlateaddclamp1col;
hcolfunc_post4 = rt_tlateaddclamp4cols;
}
}
return true;
case STYLEOP_Sub:
if (flags & STYLEF_ColorIsFixed)
{
colfunc = R_FillSubClampColumn;
hcolfunc_post1 = rt_subclamp1col;
hcolfunc_post4 = rt_subclamp4cols;
}
else if (dc_translation == NULL)
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
colfunc = R_DrawSubClampColumnP_C;
hcolfunc_post1 = rt_subclamp1col;
hcolfunc_post4 = rt_subclamp4cols;
}
else
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
colfunc = R_DrawSubClampTranslatedColumnP_C;
hcolfunc_post1 = rt_tlatesubclamp1col;
hcolfunc_post4 = rt_tlatesubclamp4cols;
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
return true;
case STYLEOP_RevSub:
if (fglevel == 0 && bglevel == FRACUNIT)
{
return false;
}
if (flags & STYLEF_ColorIsFixed)
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
colfunc = R_FillRevSubClampColumn;
hcolfunc_post1 = rt_subclamp1col;
hcolfunc_post4 = rt_subclamp4cols;
}
else if (dc_translation == NULL)
{
colfunc = R_DrawRevSubClampColumnP_C;
hcolfunc_post1 = rt_revsubclamp1col;
hcolfunc_post4 = rt_revsubclamp4cols;
}
else
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
colfunc = R_DrawRevSubClampTranslatedColumnP_C;
hcolfunc_post1 = rt_tlaterevsubclamp1col;
hcolfunc_post4 = rt_tlaterevsubclamp4cols;
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
return true;
default:
return false;
}
}
static fixed_t GetAlpha(int type, fixed_t alpha)
{
switch (type)
{
case STYLEALPHA_Zero: return 0;
case STYLEALPHA_One: return FRACUNIT;
case STYLEALPHA_Src: return alpha;
case STYLEALPHA_InvSrc: return FRACUNIT - alpha;
default: return 0;
}
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
ESPSResult R_SetPatchStyle (FRenderStyle style, fixed_t alpha, int translation, DWORD color)
{
fixed_t fglevel, bglevel;
style.CheckFuzz();
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
if (style.Flags & STYLEF_TransSoulsAlpha)
{
alpha = fixed_t(transsouls * FRACUNIT);
}
else if (style.Flags & STYLEF_Alpha1)
{
alpha = FRACUNIT;
}
else
{
alpha = clamp<fixed_t> (alpha, 0, FRACUNIT);
}
- Discovered that Shader Model 1.4 clamps my constants, so I can't use 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)
2007-12-26 04:42:15 +00:00
dc_translation = NULL;
if (translation != 0)
{
- Discovered that Shader Model 1.4 clamps my constants, so I can't use 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)
2007-12-26 04:42:15 +00:00
FRemapTable *table = TranslationToTable(translation);
if (table != NULL)
{
dc_translation = table->Remap;
}
}
basecolormapsave = basecolormap;
hcolfunc_pre = R_DrawColumnHoriz;
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
// Check for special modes
if (style.BlendOp == STYLEOP_Fuzz)
{
colfunc = fuzzcolfunc;
return DoDraw0;
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
}
else if (style == LegacyRenderStyles[STYLE_Shaded])
{
// Shaded drawer only gets 16 levels because it saves memory.
if ((alpha >>= 12) == 0)
return DontDraw;
colfunc = R_DrawShadedColumn;
hcolfunc_post1 = rt_shaded1col;
hcolfunc_post4 = rt_shaded4cols;
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
dc_color = fixedcolormap ? fixedcolormap[APART(color)] : basecolormap->Maps[APART(color)];
dc_colormap = (basecolormap = &ShadeFakeColormap[16-alpha])->Maps;
if (fixedlightlev && !fixedcolormap)
{
dc_colormap += fixedlightlev;
}
return r_columnmethod ? DoDraw1 : DoDraw0;
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
fglevel = GetAlpha(style.SrcAlpha, alpha);
bglevel = GetAlpha(style.DestAlpha, alpha);
if (style.Flags & STYLEF_ColorIsFixed)
{
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
int x = fglevel >> 10;
int r = RPART(color);
int g = GPART(color);
int b = BPART(color);
// dc_color is used by the rt_* routines. It is indexed into dc_srcblend.
dc_color = RGB32k[r>>3][g>>3][b>>3];
if (style.Flags & STYLEF_InvertSource)
{
r = 255 - r;
g = 255 - g;
b = 255 - b;
}
// dc_srccolor is used by the R_Fill* routines. It is premultiplied
// with the alpha.
dc_srccolor = ((((r*x)>>4)<<20) | ((g*x)>>4) | ((((b)*x)>>4)<<10)) & 0x3feffbff;
hcolfunc_pre = R_FillColumnHorizP;
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
dc_colormap = identitymap;
}
if (!R_SetBlendFunc (style.BlendOp, fglevel, bglevel, style.Flags))
{
return DontDraw;
}
return r_columnmethod ? DoDraw1 : DoDraw0;
}
void R_FinishSetPatchStyle ()
{
basecolormap = basecolormapsave;
}
bool R_GetTransMaskDrawers (fixed_t (**tmvline1)(), void (**tmvline4)())
{
if (colfunc == R_DrawAddColumnP_C)
{
*tmvline1 = tmvline1_add;
*tmvline4 = tmvline4_add;
return true;
}
if (colfunc == R_DrawAddClampColumnP_C)
{
*tmvline1 = tmvline1_addclamp;
*tmvline4 = tmvline4_addclamp;
return true;
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
if (colfunc == R_DrawSubClampColumnP_C)
{
*tmvline1 = tmvline1_subclamp;
*tmvline4 = tmvline4_subclamp;
return true;
}
if (colfunc == R_DrawRevSubClampColumnP_C)
{
*tmvline1 = tmvline1_revsubclamp;
*tmvline4 = tmvline4_revsubclamp;
return true;
}
return false;
}
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
//==========================================================================
//
// FRenderStyle :: IsVisible
//
// Coupled with the given alpha, will this render style produce something
// visible on-screen?
//
//==========================================================================
2008-01-26 04:47:58 +00:00
bool FRenderStyle::IsVisible(fixed_t alpha) const throw()
- Updated lempar.c to v1.31. - 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)
2008-01-25 23:57:44 +00:00
{
if (BlendOp == STYLEOP_None)
{
return false;
}
if (BlendOp == STYLEOP_Add || BlendOp == STYLEOP_RevSub)
{
if (Flags & STYLEF_Alpha1)
{
alpha = FRACUNIT;
}
else
{
alpha = clamp(alpha, 0, FRACUNIT);
}
return GetAlpha(SrcAlpha, alpha) != 0 || GetAlpha(DestAlpha, alpha) != FRACUNIT;
}
// Treat anything else as visible.
return true;
}
//==========================================================================
//
// FRenderStyle :: CheckFuzz
//
// Adjusts settings based on r_drawfuzz CVAR
//
//==========================================================================
void FRenderStyle::CheckFuzz()
{
if (BlendOp == STYLEOP_FuzzOrAdd)
{
BlendOp = (r_drawfuzz || !r_drawtrans) ? STYLEOP_Fuzz : STYLEOP_Add;
}
else if (BlendOp == STYLEOP_FuzzOrSub)
{
BlendOp = (r_drawfuzz || !r_drawtrans) ? STYLEOP_Fuzz : STYLEOP_Sub;
}
else if (BlendOp == STYLEOP_FuzzOrRevSub)
{
BlendOp = (r_drawfuzz || !r_drawtrans) ? STYLEOP_Fuzz : STYLEOP_RevSub;
}
}