2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
14 * This library is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU Lesser General Public
16 * License as published by the Free Software Foundation; either
17 * version 2.1 of the License, or (at your option) any later version.
19 * This library is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
22 * Lesser General Public License for more details.
24 * You should have received a copy of the GNU Lesser General Public
25 * License along with this library; if not, write to the Free Software
26 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
34 #include "wined3d_private.h"
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 #define GLINFO_LOCATION (*gl_info)
42 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
43 * so upload them above that
45 #define ARB_SHADER_PRIVCONST_BASE GL_LIMITS(vshader_constantsF)
46 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
48 /* ARB_program_shader private data */
49 struct shader_arb_priv
{
50 GLuint current_vprogram_id
;
51 GLuint current_fprogram_id
;
52 GLuint depth_blt_vprogram_id
;
53 GLuint depth_blt_fprogram_id
;
54 BOOL use_arbfp_fixed_func
;
55 struct hash_table_t
*fragment_shaders
;
58 /********************************************************
59 * ARB_[vertex/fragment]_program helper functions follow
60 ********************************************************/
63 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
64 * When constant_list == NULL, it will load all the constants.
66 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
67 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
69 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl
* This
, WineD3D_GL_Info
*gl_info
, GLuint target_type
,
70 unsigned int max_constants
, float* constants
, char *dirty_consts
) {
71 local_constant
* lconst
;
75 if (TRACE_ON(d3d_shader
)) {
76 for(i
= 0; i
< max_constants
; i
++) {
77 if(!dirty_consts
[i
]) continue;
78 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
79 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
80 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
83 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
84 if(target_type
== GL_FRAGMENT_PROGRAM_ARB
&&
85 WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) == 1) {
87 for(i
= 0; i
< max_constants
; i
++) {
88 if(!dirty_consts
[i
]) continue;
92 if(constants
[j
+ 0] > 1.0) lcl_const
[0] = 1.0;
93 else if(constants
[j
+ 0] < -1.0) lcl_const
[0] = -1.0;
94 else lcl_const
[0] = constants
[j
+ 0];
96 if(constants
[j
+ 1] > 1.0) lcl_const
[1] = 1.0;
97 else if(constants
[j
+ 1] < -1.0) lcl_const
[1] = -1.0;
98 else lcl_const
[1] = constants
[j
+ 1];
100 if(constants
[j
+ 2] > 1.0) lcl_const
[2] = 1.0;
101 else if(constants
[j
+ 2] < -1.0) lcl_const
[2] = -1.0;
102 else lcl_const
[2] = constants
[j
+ 2];
104 if(constants
[j
+ 3] > 1.0) lcl_const
[3] = 1.0;
105 else if(constants
[j
+ 3] < -1.0) lcl_const
[3] = -1.0;
106 else lcl_const
[3] = constants
[j
+ 3];
108 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
111 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS
)) {
112 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
113 * or just reloading *all* constants at once
115 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
117 for(i
= 0; i
< max_constants
; i
++) {
118 if(!dirty_consts
[i
]) continue;
120 /* Find the next block of dirty constants */
123 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
127 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
130 for(i
= 0; i
< max_constants
; i
++) {
131 if(dirty_consts
[i
]) {
133 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
138 checkGLcall("glProgramEnvParameter4fvARB()");
140 /* Load immediate constants */
141 if(This
->baseShader
.load_local_constsF
) {
142 if (TRACE_ON(d3d_shader
)) {
143 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
144 GLfloat
* values
= (GLfloat
*)lconst
->value
;
145 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
146 values
[0], values
[1], values
[2], values
[3]);
149 /* Immediate constants are clamped for 1.X shaders at loading times */
151 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
152 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
153 ret
= max(ret
, lconst
->idx
+ 1);
154 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
156 checkGLcall("glProgramEnvParameter4fvARB()");
157 return ret
; /* The loaded immediate constants need reloading for the next shader */
159 return 0; /* No constants are dirty now */
164 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
166 * We only support float constants in ARB at the moment, so don't
167 * worry about the Integers or Booleans
169 static void shader_arb_load_constants(
170 IWineD3DDevice
* device
,
172 char useVertexShader
) {
174 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
175 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
176 WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
179 if (useVertexShader
) {
180 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
182 /* Load DirectX 9 float constants for vertex shader */
183 deviceImpl
->highest_dirty_vs_const
= shader_arb_load_constantsF(
184 vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
185 deviceImpl
->highest_dirty_vs_const
,
186 stateBlock
->vertexShaderConstantF
,
187 deviceImpl
->activeContext
->vshader_const_dirty
);
189 /* Upload the position fixup */
190 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, ARB_SHADER_PRIVCONST_POS
, deviceImpl
->posFixup
));
193 if (usePixelShader
) {
195 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
196 IWineD3DPixelShaderImpl
*psi
= (IWineD3DPixelShaderImpl
*) pshader
;
198 /* Load DirectX 9 float constants for pixel shader */
199 deviceImpl
->highest_dirty_ps_const
= shader_arb_load_constantsF(
200 pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
201 deviceImpl
->highest_dirty_ps_const
,
202 stateBlock
->pixelShaderConstantF
,
203 deviceImpl
->activeContext
->pshader_const_dirty
);
205 for(i
= 0; i
< psi
->numbumpenvmatconsts
; i
++) {
206 /* The state manager takes care that this function is always called if the bump env matrix changes
208 float *data
= (float *) &stateBlock
->textureState
[(int) psi
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
209 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->bumpenvmatconst
[i
].const_num
, data
));
210 deviceImpl
->activeContext
->pshader_const_dirty
[psi
->bumpenvmatconst
[i
].const_num
] = 1;
212 if(psi
->luminanceconst
[i
].const_num
!= -1) {
213 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
214 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
215 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
216 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
218 float *scale
= (float *) &stateBlock
->textureState
[(int) psi
->luminanceconst
[i
].texunit
][WINED3DTSS_BUMPENVLSCALE
];
219 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->luminanceconst
[i
].const_num
, scale
));
220 deviceImpl
->activeContext
->pshader_const_dirty
[psi
->luminanceconst
[i
].const_num
] = 1;
224 if(((IWineD3DPixelShaderImpl
*) pshader
)->srgb_enabled
&&
225 !((IWineD3DPixelShaderImpl
*) pshader
)->srgb_mode_hardcoded
) {
229 if(stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
]) {
230 comparison
[0] = srgb_cmp
; comparison
[1] = srgb_cmp
;
231 comparison
[2] = srgb_cmp
; comparison
[3] = srgb_cmp
;
233 mul_low
[0] = srgb_mul_low
; mul_low
[1] = srgb_mul_low
;
234 mul_low
[2] = srgb_mul_low
; mul_low
[3] = srgb_mul_low
;
236 comparison
[0] = 1.0 / 0.0; comparison
[1] = 1.0 / 0.0;
237 comparison
[2] = 1.0 / 0.0; comparison
[3] = 1.0 / 0.0;
239 mul_low
[0] = 1.0; mul_low
[1] = 1.0;
240 mul_low
[2] = 1.0; mul_low
[3] = 1.0;
242 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->srgb_cmp_const
, comparison
));
243 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->srgb_low_const
, mul_low
));
244 checkGLcall("Load sRGB correction constants\n");
245 deviceImpl
->activeContext
->pshader_const_dirty
[psi
->srgb_low_const
] = 1;
246 deviceImpl
->activeContext
->pshader_const_dirty
[psi
->srgb_cmp_const
] = 1;
252 /* Generate the variable & register declarations for the ARB_vertex_program output target */
253 static void shader_generate_arb_declarations(
254 IWineD3DBaseShader
*iface
,
255 shader_reg_maps
* reg_maps
,
256 SHADER_BUFFER
* buffer
,
257 WineD3D_GL_Info
* gl_info
) {
259 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
260 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
262 char pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
263 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
264 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
265 UINT extra_constants_needed
= 0;
266 local_constant
* lconst
;
268 /* Temporary Output register */
269 shader_addline(buffer
, "TEMP TMP_OUT;\n");
271 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
272 if (reg_maps
->temporary
[i
])
273 shader_addline(buffer
, "TEMP R%u;\n", i
);
276 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
277 if (reg_maps
->address
[i
])
278 shader_addline(buffer
, "ADDRESS A%d;\n", i
);
281 for(i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
282 if (reg_maps
->texcoord
[i
])
283 shader_addline(buffer
,"TEMP T%u;\n", i
);
286 /* Texture coordinate registers must be pre-loaded */
287 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
288 if (reg_maps
->texcoord
[i
])
289 shader_addline(buffer
, "MOV T%u, fragment.texcoord[%u];\n", i
, i
);
292 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
293 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) This
;
294 if(!reg_maps
->bumpmat
[i
]) continue;
296 cur
= ps
->numbumpenvmatconsts
;
297 ps
->bumpenvmatconst
[cur
].const_num
= -1;
298 ps
->bumpenvmatconst
[cur
].texunit
= i
;
299 ps
->luminanceconst
[cur
].const_num
= -1;
300 ps
->luminanceconst
[cur
].texunit
= i
;
302 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
303 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
306 if(max_constantsF
+ extra_constants_needed
< GL_LIMITS(pshader_constantsF
)) {
307 ps
->bumpenvmatconst
[cur
].const_num
= max_constantsF
+ extra_constants_needed
;
308 shader_addline(buffer
, "PARAM bumpenvmat%d = program.env[%d];\n",
309 i
, ps
->bumpenvmatconst
[cur
].const_num
);
310 extra_constants_needed
++;
312 if(reg_maps
->luminanceparams
&& max_constantsF
+ extra_constants_needed
< GL_LIMITS(pshader_constantsF
)) {
313 ((IWineD3DPixelShaderImpl
*)This
)->luminanceconst
[cur
].const_num
= max_constantsF
+ extra_constants_needed
;
314 shader_addline(buffer
, "PARAM luminance%d = program.env[%d];\n",
315 i
, ps
->luminanceconst
[cur
].const_num
);
316 extra_constants_needed
++;
317 } else if(reg_maps
->luminanceparams
) {
318 FIXME("No free constant to load the luminance parameters\n");
321 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
324 ps
->numbumpenvmatconsts
= cur
+ 1;
327 if(device
->stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
] && pshader
) {
328 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
329 /* If there are 2 constants left to use, use them to pass the sRGB correction values in. This way
330 * srgb write correction can be turned on and off dynamically without recompilation. Otherwise
331 * hardcode them. The drawback of hardcoding is that the shader needs recompilation to turn sRGB
334 if(max_constantsF
+ extra_constants_needed
+ 1 < GL_LIMITS(pshader_constantsF
) && FALSE
) {
335 /* The idea is that if srgb is enabled, then disabled, the constant loading code
336 * can effectively disable sRGB correction by passing 1.0 and INF as the multiplication
337 * and comparison constants. If it disables it that way, the shader won't be recompiled
338 * and the code will stay in, so sRGB writing can be turned on again by setting the
339 * constants from the spec
341 ps_impl
->srgb_mode_hardcoded
= 0;
342 ps_impl
->srgb_low_const
= GL_LIMITS(pshader_constantsF
) - extra_constants_needed
;
343 ps_impl
->srgb_cmp_const
= GL_LIMITS(pshader_constantsF
) - extra_constants_needed
- 1;
344 shader_addline(buffer
, "PARAM srgb_mul_low = program.env[%d];\n", ps_impl
->srgb_low_const
);
345 shader_addline(buffer
, "PARAM srgb_comparison = program.env[%d];\n", ps_impl
->srgb_cmp_const
);
347 shader_addline(buffer
, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
348 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
349 shader_addline(buffer
, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
350 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
351 ps_impl
->srgb_mode_hardcoded
= 1;
353 /* These can be hardcoded, they do not cause any harm because no fragment will enter the high
354 * path if the comparison value is set to INF
356 shader_addline(buffer
, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
357 srgb_pow
, srgb_pow
, srgb_pow
);
358 shader_addline(buffer
, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
359 srgb_mul_high
, srgb_mul_high
, srgb_mul_high
);
360 shader_addline(buffer
, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
361 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
362 ps_impl
->srgb_enabled
= 1;
364 IWineD3DPixelShaderImpl
*ps_impl
= (IWineD3DPixelShaderImpl
*) This
;
366 /* Do not write any srgb fixup into the shader to save shader size and processing time.
367 * As a consequence, we can't toggle srgb write on without recompilation
369 ps_impl
->srgb_enabled
= 0;
370 ps_impl
->srgb_mode_hardcoded
= 1;
373 /* Load local constants using the program-local space,
374 * this avoids reloading them each time the shader is used
376 if(!This
->baseShader
.load_local_constsF
) {
377 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
378 shader_addline(buffer
, "PARAM C%u = program.local[%u];\n", lconst
->idx
,
383 /* we use the array-based constants array if the local constants are marked for loading,
384 * because then we use indirect addressing, or when the local constant list is empty,
385 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
386 * local constants do not declare the loaded constants as an array because ARB compilers usually
387 * do not optimize unused constants away
389 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) {
390 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
391 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
392 max_constantsF
, max_constantsF
- 1);
394 for(i
= 0; i
< max_constantsF
; i
++) {
395 if(!shader_constant_is_local(This
, i
)) {
396 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
402 static const char * const shift_tab
[] = {
403 "dummy", /* 0 (none) */
404 "coefmul.x", /* 1 (x2) */
405 "coefmul.y", /* 2 (x4) */
406 "coefmul.z", /* 3 (x8) */
407 "coefmul.w", /* 4 (x16) */
408 "dummy", /* 5 (x32) */
409 "dummy", /* 6 (x64) */
410 "dummy", /* 7 (x128) */
411 "dummy", /* 8 (d256) */
412 "dummy", /* 9 (d128) */
413 "dummy", /* 10 (d64) */
414 "dummy", /* 11 (d32) */
415 "coefdiv.w", /* 12 (d16) */
416 "coefdiv.z", /* 13 (d8) */
417 "coefdiv.y", /* 14 (d4) */
418 "coefdiv.x" /* 15 (d2) */
421 static void shader_arb_get_write_mask(SHADER_OPCODE_ARG
* arg
, const DWORD param
, char *write_mask
) {
422 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
423 char *ptr
= write_mask
;
424 char vshader
= shader_is_vshader_version(This
->baseShader
.hex_version
);
426 if(vshader
&& shader_get_regtype(param
) == WINED3DSPR_ADDR
) {
429 } else if ((param
& WINED3DSP_WRITEMASK_ALL
) != WINED3DSP_WRITEMASK_ALL
) {
431 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
432 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
433 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
434 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
440 static void shader_arb_get_swizzle(const DWORD param
, BOOL fixup
, char *swizzle_str
) {
441 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
442 * but addressed as "rgba". To fix this we need to swap the register's x
443 * and z components. */
444 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
445 char *ptr
= swizzle_str
;
447 /* swizzle bits fields: wwzzyyxx */
448 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
449 DWORD swizzle_x
= swizzle
& 0x03;
450 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
451 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
452 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
454 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
455 * generate a swizzle string. Unless we need to our own swizzling. */
456 if ((WINED3DSP_NOSWIZZLE
>> WINED3DSP_SWIZZLE_SHIFT
) != swizzle
|| fixup
) {
458 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
459 *ptr
++ = swizzle_chars
[swizzle_x
];
461 *ptr
++ = swizzle_chars
[swizzle_x
];
462 *ptr
++ = swizzle_chars
[swizzle_y
];
463 *ptr
++ = swizzle_chars
[swizzle_z
];
464 *ptr
++ = swizzle_chars
[swizzle_w
];
471 static void pshader_get_register_name(IWineD3DBaseShader
* iface
,
472 const DWORD param
, char* regstr
) {
474 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
475 DWORD regtype
= shader_get_regtype(param
);
476 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
479 case WINED3DSPR_TEMP
:
480 sprintf(regstr
, "R%u", reg
);
482 case WINED3DSPR_INPUT
:
484 strcpy(regstr
, "fragment.color.primary");
486 strcpy(regstr
, "fragment.color.secondary");
489 case WINED3DSPR_CONST
:
490 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) {
491 sprintf(regstr
, "C[%u]", reg
);
493 sprintf(regstr
, "C%u", reg
);
496 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
497 sprintf(regstr
,"T%u", reg
);
499 case WINED3DSPR_COLOROUT
:
501 sprintf(regstr
, "TMP_COLOR");
503 /* TODO: See GL_ARB_draw_buffers */
504 FIXME("Unsupported write to render target %u\n", reg
);
505 sprintf(regstr
, "unsupported_register");
508 case WINED3DSPR_DEPTHOUT
:
509 sprintf(regstr
, "result.depth");
511 case WINED3DSPR_ATTROUT
:
512 sprintf(regstr
, "oD[%u]", reg
);
514 case WINED3DSPR_TEXCRDOUT
:
515 sprintf(regstr
, "oT[%u]", reg
);
518 FIXME("Unhandled register name Type(%d)\n", regtype
);
519 sprintf(regstr
, "unrecognized_register");
524 /* TODO: merge with pixel shader */
525 static void vshader_program_add_param(SHADER_OPCODE_ARG
*arg
, const DWORD param
, BOOL is_input
, char *hwLine
) {
527 IWineD3DVertexShaderImpl
* This
= (IWineD3DVertexShaderImpl
*) arg
->shader
;
529 /* oPos, oFog and oPts in D3D */
530 static const char * const hwrastout_reg_names
[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
532 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
533 DWORD regtype
= shader_get_regtype(param
);
535 BOOL is_color
= FALSE
;
537 if ((param
& WINED3DSP_SRCMOD_MASK
) == WINED3DSPSM_NEG
) {
538 strcat(hwLine
, " -");
544 case WINED3DSPR_TEMP
:
545 sprintf(tmpReg
, "R%u", reg
);
546 strcat(hwLine
, tmpReg
);
548 case WINED3DSPR_INPUT
:
550 if (vshader_input_is_color((IWineD3DVertexShader
*) This
, reg
))
553 sprintf(tmpReg
, "vertex.attrib[%u]", reg
);
554 strcat(hwLine
, tmpReg
);
556 case WINED3DSPR_CONST
:
557 if(param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
558 if(reg
>= This
->rel_offset
) {
559 sprintf(tmpReg
, "C[A0.x + %u]", reg
- This
->rel_offset
);
561 sprintf(tmpReg
, "C[A0.x - %u]", -reg
+ This
->rel_offset
);
564 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) {
565 sprintf(tmpReg
, "C[%u]", reg
);
567 sprintf(tmpReg
, "C%u", reg
);
570 strcat(hwLine
, tmpReg
);
572 case WINED3DSPR_ADDR
: /*case D3DSPR_TEXTURE:*/
573 sprintf(tmpReg
, "A%u", reg
);
574 strcat(hwLine
, tmpReg
);
576 case WINED3DSPR_RASTOUT
:
577 sprintf(tmpReg
, "%s", hwrastout_reg_names
[reg
]);
578 strcat(hwLine
, tmpReg
);
580 case WINED3DSPR_ATTROUT
:
582 strcat(hwLine
, "result.color.primary");
584 strcat(hwLine
, "result.color.secondary");
587 case WINED3DSPR_TEXCRDOUT
:
588 sprintf(tmpReg
, "result.texcoord[%u]", reg
);
589 strcat(hwLine
, tmpReg
);
592 FIXME("Unknown reg type %d %d\n", regtype
, reg
);
593 strcat(hwLine
, "unrecognized_register");
599 shader_arb_get_write_mask(arg
, param
, write_mask
);
600 strcat(hwLine
, write_mask
);
603 shader_arb_get_swizzle(param
, is_color
, swizzle
);
604 strcat(hwLine
, swizzle
);
608 static void shader_hw_sample(SHADER_OPCODE_ARG
* arg
, DWORD sampler_idx
, const char *dst_str
, const char *coord_reg
, BOOL projected
, BOOL bias
) {
609 SHADER_BUFFER
* buffer
= arg
->buffer
;
610 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
611 const char *tex_type
;
613 switch(sampler_type
) {
620 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
621 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
622 if(device
->stateBlock
->textures
[sampler_idx
] &&
623 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
631 case WINED3DSTT_VOLUME
:
635 case WINED3DSTT_CUBE
:
640 ERR("Unexpected texture type %d\n", sampler_type
);
645 /* Shouldn't be possible, but let's check for it */
646 if(projected
) FIXME("Biased and Projected texture sampling\n");
647 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
648 shader_addline(buffer
, "TXB %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
649 } else if (projected
) {
650 shader_addline(buffer
, "TXP %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
652 shader_addline(buffer
, "TEX %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
656 static void gen_color_correction(SHADER_BUFFER
*buffer
, const char *reg
, const char *writemask
,
657 const char *one
, const char *two
, WINED3DFORMAT fmt
,
658 WineD3D_GL_Info
*gl_info
) {
660 case WINED3DFMT_V8U8
:
661 case WINED3DFMT_V16U16
:
662 if(GL_SUPPORT(NV_TEXTURE_SHADER
) ||
663 (GL_SUPPORT(ATI_ENVMAP_BUMPMAP
) && fmt
== WINED3DFMT_V8U8
)) {
665 /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-)
666 * disabled until an application that needs it is found because it causes unneeded
667 * shader recompilation in some game
669 if(strlen(writemask
) >= 4) {
670 shader_addline(buffer
, "MOV %s.%c, %s;\n", reg
, writemask
[3], one
);
674 /* Correct the sign, but leave the blue as it is - it was loaded correctly already
675 * ARB shaders are a bit picky wrt writemasks and swizzles. If we're free to scale
676 * all registers, do so, this saves an instruction.
678 if(strlen(writemask
) >= 5) {
679 shader_addline(buffer
, "MAD %s, %s, %s, -%s;\n", reg
, reg
, two
, one
);
680 } else if(strlen(writemask
) >= 3) {
681 shader_addline(buffer
, "MAD %s.%c, %s.%c, %s, -%s;\n",
685 shader_addline(buffer
, "MAD %s.%c, %s.%c, %s, -%s;\n",
689 } else if(strlen(writemask
) == 2) {
690 shader_addline(buffer
, "MAD %s.%c, %s.%c, %s, -%s;\n", reg
, writemask
[1],
691 reg
, writemask
[1], two
, one
);
696 case WINED3DFMT_X8L8V8U8
:
697 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
698 /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
699 * and a(X) is always 1.0. Cannot do a full conversion due to L(blue)
701 if(strlen(writemask
) >= 3) {
702 shader_addline(buffer
, "MAD %s.%c, %s.%c, %s, -%s;\n",
706 shader_addline(buffer
, "MAD %s.%c, %s.%c, %s, -%s;\n",
710 } else if(strlen(writemask
) == 2) {
711 shader_addline(buffer
, "MAD %s.%c, %s.%c, %s, -%s;\n",
719 case WINED3DFMT_L6V5U5
:
720 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
721 if(strlen(writemask
) >= 4) {
722 /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
723 shader_addline(buffer
, "MOV TMP.g, %s.%c;\n",
725 shader_addline(buffer
, "MAD %s.%c%c, %s.%c%c%c%c, %s, -%s;\n",
726 reg
, writemask
[1], writemask
[2],
727 reg
, writemask
[3], writemask
[1], writemask
[3], writemask
[1],
729 shader_addline(buffer
, "MOV %s.%c, TMP.g;\n", reg
,
731 } else if(strlen(writemask
) == 3) {
732 /* This is bad: We have VL, but we need VU */
733 FIXME("2 components sampled from a converted L6V5U5 texture\n");
735 shader_addline(buffer
, "MAD %s.%c, %s.%c, %s, -%s;\n",
743 case WINED3DFMT_Q8W8V8U8
:
744 if(!GL_SUPPORT(NV_TEXTURE_SHADER
)) {
745 /* Correct the sign in all channels */
746 switch(strlen(writemask
)) {
748 shader_addline(buffer
, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
753 shader_addline(buffer
, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
758 shader_addline(buffer
, "MAD %s.%c, %s.%c, coefmul.x, -one;\n",
763 /* Should not occur, since it's at minimum '.' and a letter */
765 ERR("Unexpected writemask: \"%s\"\n", writemask
);
770 shader_addline(buffer
, "MAD %s, %s, coefmul.x, -one;\n", reg
, reg
);
775 case WINED3DFMT_ATI2N
:
776 /* GL_ATI_texture_compression_3dc returns the two channels as luminance-alpha,
777 * which means the first one is replicated across .rgb, and the 2nd one is in
778 * .a. We need the 2nd in .g
780 * GL_EXT_texture_compression_rgtc returns the values in .rg, however, they
781 * are swapped compared to d3d. So swap red and green.
783 if(GL_SUPPORT(EXT_TEXTURE_COMPRESSION_RGTC
)) {
784 shader_addline(buffer
, "SWZ %s, %s, %c, %c, 1, 0;\n",
785 reg
, reg
, writemask
[2], writemask
[1]);
787 if(strlen(writemask
) == 5) {
788 shader_addline(buffer
, "MOV %s.%c, %s.%c;\n",
789 reg
, writemask
[2], reg
, writemask
[4]);
790 } else if(strlen(writemask
) == 2) {
793 /* This is bad: We have VL, but we need VU */
794 FIXME("2 or 3 components sampled from a converted ATI2N texture\n");
799 /* stupid compiler */
805 static void shader_arb_color_correction(SHADER_OPCODE_ARG
* arg
) {
806 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
807 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) shader
->baseShader
.device
;
808 WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
810 WINED3DFORMAT conversion_group
;
811 IWineD3DBaseTextureImpl
*texture
;
813 BOOL recorded
= FALSE
;
815 DWORD hex_version
= shader
->baseShader
.hex_version
;
819 switch(arg
->opcode
->opcode
) {
821 if (hex_version
< WINED3DPS_VERSION(2,0)) {
822 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
824 sampler_idx
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
828 case WINED3DSIO_TEXLDL
:
829 FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
832 case WINED3DSIO_TEXDP3TEX
:
833 case WINED3DSIO_TEXM3x3TEX
:
834 case WINED3DSIO_TEXM3x3SPEC
:
835 case WINED3DSIO_TEXM3x3VSPEC
:
836 case WINED3DSIO_TEXBEM
:
837 case WINED3DSIO_TEXREG2AR
:
838 case WINED3DSIO_TEXREG2GB
:
839 case WINED3DSIO_TEXREG2RGB
:
840 sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
844 /* Not a texture sampling instruction, nothing to do */
848 texture
= (IWineD3DBaseTextureImpl
*) deviceImpl
->stateBlock
->textures
[sampler_idx
];
850 fmt
= texture
->resource
.format
;
851 conversion_group
= texture
->baseTexture
.shader_conversion_group
;
853 fmt
= WINED3DFMT_UNKNOWN
;
854 conversion_group
= WINED3DFMT_UNKNOWN
;
857 /* before doing anything, record the sampler with the format in the format conversion list,
858 * but check if it's not there already
860 for(i
= 0; i
< shader
->baseShader
.num_sampled_samplers
; i
++) {
861 if(shader
->baseShader
.sampled_samplers
[i
] == sampler_idx
) {
866 shader
->baseShader
.sampled_samplers
[shader
->baseShader
.num_sampled_samplers
] = sampler_idx
;
867 shader
->baseShader
.num_sampled_samplers
++;
868 shader
->baseShader
.sampled_format
[sampler_idx
] = conversion_group
;
871 pshader_get_register_name(arg
->shader
, arg
->dst
, reg
);
872 shader_arb_get_write_mask(arg
, arg
->dst
, writemask
);
873 if(strlen(writemask
) == 0) strcpy(writemask
, ".xyzw");
875 gen_color_correction(arg
->buffer
, reg
, writemask
, "one", "coefmul.x", fmt
, gl_info
);
880 static void pshader_gen_input_modifier_line (
881 IWineD3DBaseShader
*iface
,
882 SHADER_BUFFER
* buffer
,
887 /* Generate a line that does the input modifier computation and return the input register to use */
892 /* Assume a new line will be added */
895 /* Get register name */
896 pshader_get_register_name(iface
, instr
, regstr
);
897 shader_arb_get_swizzle(instr
, FALSE
, swzstr
);
899 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
900 case WINED3DSPSM_NONE
:
901 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
904 case WINED3DSPSM_NEG
:
905 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
908 case WINED3DSPSM_BIAS
:
909 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
911 case WINED3DSPSM_BIASNEG
:
912 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
914 case WINED3DSPSM_SIGN
:
915 shader_addline(buffer
, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg
, regstr
);
917 case WINED3DSPSM_SIGNNEG
:
918 shader_addline(buffer
, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg
, regstr
);
920 case WINED3DSPSM_COMP
:
921 shader_addline(buffer
, "SUB T%c, one.x, %s;\n", 'A' + tmpreg
, regstr
);
924 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
926 case WINED3DSPSM_X2NEG
:
927 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
930 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
931 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
934 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
935 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
938 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
942 /* Return modified or original register, with swizzle */
944 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
947 static inline void pshader_gen_output_modifier_line(
948 SHADER_BUFFER
* buffer
,
954 /* Generate a line that does the output modifier computation */
955 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", saturate
? "_SAT" : "",
956 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
959 static void pshader_hw_bem(SHADER_OPCODE_ARG
* arg
) {
960 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
962 SHADER_BUFFER
* buffer
= arg
->buffer
;
964 char src_name
[2][50];
966 DWORD sampler_code
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
967 BOOL has_bumpmat
= FALSE
;
970 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
971 if(This
->bumpenvmatconst
[i
].const_num
!= -1 && This
->bumpenvmatconst
[i
].texunit
== sampler_code
) {
977 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
978 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
979 strcat(dst_name
, dst_wmask
);
981 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
[0]);
982 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[1], 1, src_name
[1]);
985 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
986 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
987 shader_addline(buffer
, "DP3 TMP.r, TMP2, %s;\n", src_name
[1]);
988 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
989 shader_addline(buffer
, "DP3 TMP.g, TMP2, %s;\n", src_name
[1]);
991 shader_addline(buffer
, "ADD %s, %s, TMP;\n", dst_name
, src_name
[0]);
993 shader_addline(buffer
, "MOV %s, %s;\n", dst_name
, src_name
[0]);
997 static void pshader_hw_cnd(SHADER_OPCODE_ARG
* arg
) {
999 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
1000 SHADER_BUFFER
* buffer
= arg
->buffer
;
1003 char src_name
[3][50];
1004 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
1005 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1007 /* FIXME: support output modifiers */
1009 /* Handle output register */
1010 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1011 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
1013 /* Generate input register names (with modifiers) */
1014 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
[0]);
1015 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[1], 1, src_name
[1]);
1016 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[2], 2, src_name
[2]);
1018 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1019 if (shader
->baseShader
.hex_version
<= WINED3DPS_VERSION(1, 3) &&
1020 arg
->opcode_token
& WINED3DSI_COISSUE
) {
1021 shader_addline(buffer
, "MOV%s %s%s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[1]);
1023 shader_addline(buffer
, "ADD TMP, -%s, coefdiv.x;\n", src_name
[0]);
1024 shader_addline(buffer
, "CMP%s %s%s, TMP, %s, %s;\n",
1025 sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[1], src_name
[2]);
1028 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1031 static void pshader_hw_cmp(SHADER_OPCODE_ARG
* arg
) {
1033 SHADER_BUFFER
* buffer
= arg
->buffer
;
1036 char src_name
[3][50];
1037 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1038 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
1040 /* FIXME: support output modifiers */
1042 /* Handle output register */
1043 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1044 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
1046 /* Generate input register names (with modifiers) */
1047 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
[0]);
1048 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[1], 1, src_name
[1]);
1049 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[2], 2, src_name
[2]);
1051 shader_addline(buffer
, "CMP%s %s%s, %s, %s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
1052 src_name
[0], src_name
[2], src_name
[1]);
1055 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1058 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1059 * dst = dot2(src0, src1) + src2 */
1060 static void pshader_hw_dp2add(SHADER_OPCODE_ARG
* arg
) {
1061 SHADER_BUFFER
* buffer
= arg
->buffer
;
1064 char src_name
[3][50];
1065 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1066 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
1068 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1069 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
1071 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
[0]);
1072 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[1], 1, src_name
[1]);
1073 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[2], 2, src_name
[2]);
1075 /* Emulate a DP2 with a DP3 and 0.0 */
1076 shader_addline(buffer
, "MOV TMP, %s;\n", src_name
[0]);
1077 shader_addline(buffer
, "MOV TMP.z, 0.0;\n");
1078 shader_addline(buffer
, "DP3 TMP2, TMP, %s;\n", src_name
[1]);
1079 shader_addline(buffer
, "ADD%s %s%s, TMP2, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[2]);
1082 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1085 /* Map the opcode 1-to-1 to the GL code */
1086 static void shader_hw_map2gl(SHADER_OPCODE_ARG
* arg
)
1088 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1089 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1090 SHADER_BUFFER
* buffer
= arg
->buffer
;
1091 DWORD dst
= arg
->dst
;
1092 DWORD
* src
= arg
->src
;
1097 if (shader_is_pshader_version(shader
->baseShader
.hex_version
))
1099 BOOL saturate
= FALSE
;
1100 BOOL centroid
= FALSE
;
1101 BOOL partialprecision
= FALSE
;
1104 strcpy(tmpLine
, curOpcode
->glname
);
1106 /* Process modifiers */
1107 if (dst
& WINED3DSP_DSTMOD_MASK
)
1109 DWORD mask
= dst
& WINED3DSP_DSTMOD_MASK
;
1111 saturate
= mask
& WINED3DSPDM_SATURATE
;
1112 centroid
= mask
& WINED3DSPDM_MSAMPCENTROID
;
1113 partialprecision
= mask
& WINED3DSPDM_PARTIALPRECISION
;
1114 mask
&= ~(WINED3DSPDM_MSAMPCENTROID
| WINED3DSPDM_PARTIALPRECISION
| WINED3DSPDM_SATURATE
);
1116 FIXME("Unrecognized modifier(%#x)\n", mask
>> WINED3DSP_DSTMOD_SHIFT
);
1119 FIXME("Unhandled modifier(%#x)\n", mask
>> WINED3DSP_DSTMOD_SHIFT
);
1121 shift
= (dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1123 /* Generate input and output registers */
1124 if (curOpcode
->num_params
> 0)
1126 /* Output token related */
1127 char output_rname
[256];
1128 char output_wmask
[20];
1129 char operands
[4][100];
1131 /* Generate input register names (with modifiers) */
1132 for (i
= 1; i
< curOpcode
->num_params
; ++i
)
1133 pshader_gen_input_modifier_line(arg
->shader
, buffer
, src
[i
-1], i
-1, operands
[i
]);
1135 /* Handle output register */
1136 pshader_get_register_name(arg
->shader
, dst
, output_rname
);
1137 strcpy(operands
[0], output_rname
);
1138 shader_arb_get_write_mask(arg
, dst
, output_wmask
);
1139 strcat(operands
[0], output_wmask
);
1141 if (saturate
&& (shift
== 0))
1142 strcat(tmpLine
, "_SAT");
1143 strcat(tmpLine
, " ");
1144 strcat(tmpLine
, operands
[0]);
1145 for (i
= 1; i
< curOpcode
->num_params
; i
++)
1147 strcat(tmpLine
, ", ");
1148 strcat(tmpLine
, operands
[i
]);
1150 strcat(tmpLine
,";\n");
1151 shader_addline(buffer
, tmpLine
);
1153 /* A shift requires another line. */
1155 pshader_gen_output_modifier_line(buffer
, saturate
, output_wmask
, shift
, output_rname
);
1158 if ((curOpcode
->opcode
== WINED3DSIO_MOV
&& shader_get_regtype(dst
) == WINED3DSPR_ADDR
)
1159 || curOpcode
->opcode
== WINED3DSIO_MOVA
) {
1160 memset(tmpLine
, 0, sizeof(tmpLine
));
1161 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
)
1163 vshader_program_add_param(arg
, src
[0], TRUE
, tmpLine
);
1164 shader_addline(buffer
, "ADD TMP.x, %s, helper_const.z;\n", tmpLine
);
1165 shader_addline(buffer
, "ARL A0.x, TMP.x;\n");
1167 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1168 * with more than one component. Thus replicate the first source argument over all
1169 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
1171 DWORD parm
= src
[0] & ~(WINED3DVS_SWIZZLE_MASK
);
1172 if((src
[0] & WINED3DVS_X_W
) == WINED3DVS_X_W
) {
1173 parm
|= WINED3DVS_X_W
| WINED3DVS_Y_W
| WINED3DVS_Z_W
| WINED3DVS_W_W
;
1174 } else if((src
[0] & WINED3DVS_X_Z
) == WINED3DVS_X_Z
) {
1175 parm
|= WINED3DVS_X_Z
| WINED3DVS_Y_Z
| WINED3DVS_Z_Z
| WINED3DVS_W_Z
;
1176 } else if((src
[0] & WINED3DVS_X_Y
) == WINED3DVS_X_Y
) {
1177 parm
|= WINED3DVS_X_Y
| WINED3DVS_Y_Y
| WINED3DVS_Z_Y
| WINED3DVS_W_Y
;
1178 } else if((src
[0] & WINED3DVS_X_X
) == WINED3DVS_X_X
) {
1179 parm
|= WINED3DVS_X_X
| WINED3DVS_Y_X
| WINED3DVS_Z_X
| WINED3DVS_W_X
;
1181 vshader_program_add_param(arg
, parm
, TRUE
, tmpLine
);
1182 shader_addline(buffer
, "ARL A0.x, %s;\n", tmpLine
);
1186 strcpy(tmpLine
, curOpcode
->glname
);
1188 if (curOpcode
->num_params
> 0)
1190 vshader_program_add_param(arg
, dst
, FALSE
, tmpLine
);
1191 for (i
= 1; i
< curOpcode
->num_params
; ++i
)
1193 strcat(tmpLine
, ",");
1194 vshader_program_add_param(arg
, src
[i
-1], TRUE
, tmpLine
);
1197 shader_addline(buffer
, "%s;\n", tmpLine
);
1201 static void pshader_hw_texkill(SHADER_OPCODE_ARG
* arg
) {
1202 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1203 DWORD hex_version
= This
->baseShader
.hex_version
;
1204 SHADER_BUFFER
* buffer
= arg
->buffer
;
1207 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1208 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1210 pshader_get_register_name(arg
->shader
, arg
->dst
, reg_dest
);
1212 if(hex_version
>= WINED3DPS_VERSION(2,0)) {
1213 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1214 shader_addline(buffer
, "KIL %s;\n", reg_dest
);
1216 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1217 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1219 shader_addline(buffer
, "MOV TMP, %s;\n", reg_dest
);
1220 shader_addline(buffer
, "MOV TMP.w, one.w;\n");
1221 shader_addline(buffer
, "KIL TMP;\n");
1225 static void pshader_hw_tex(SHADER_OPCODE_ARG
* arg
) {
1226 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1227 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1229 DWORD dst
= arg
->dst
;
1230 DWORD
* src
= arg
->src
;
1231 SHADER_BUFFER
* buffer
= arg
->buffer
;
1232 DWORD hex_version
= This
->baseShader
.hex_version
;
1233 BOOL projected
= FALSE
, bias
= FALSE
;
1237 DWORD reg_dest_code
;
1238 DWORD reg_sampler_code
;
1240 /* All versions have a destination register */
1241 reg_dest_code
= dst
& WINED3DSP_REGNUM_MASK
;
1242 pshader_get_register_name(arg
->shader
, dst
, reg_dest
);
1244 /* 1.0-1.3: Use destination register as coordinate source.
1245 1.4+: Use provided coordinate source register. */
1246 if (hex_version
< WINED3DPS_VERSION(1,4))
1247 strcpy(reg_coord
, reg_dest
);
1249 pshader_gen_input_modifier_line(arg
->shader
, buffer
, src
[0], 0, reg_coord
);
1251 /* 1.0-1.4: Use destination register number as texture code.
1252 2.0+: Use provided sampler number as texure code. */
1253 if (hex_version
< WINED3DPS_VERSION(2,0))
1254 reg_sampler_code
= reg_dest_code
;
1256 reg_sampler_code
= src
[1] & WINED3DSP_REGNUM_MASK
;
1259 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1260 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1261 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1263 if(hex_version
< WINED3DPS_VERSION(1,4)) {
1265 if(reg_sampler_code
< MAX_TEXTURES
) {
1266 flags
= deviceImpl
->stateBlock
->textureState
[reg_sampler_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1268 if (flags
& WINED3DTTFF_PROJECTED
) {
1271 } else if(hex_version
< WINED3DPS_VERSION(2,0)) {
1272 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
1273 if (src_mod
== WINED3DSPSM_DZ
) {
1275 } else if(src_mod
== WINED3DSPSM_DW
) {
1279 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
1282 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
1286 shader_hw_sample(arg
, reg_sampler_code
, reg_dest
, reg_coord
, projected
, bias
);
1289 static void pshader_hw_texcoord(SHADER_OPCODE_ARG
* arg
) {
1291 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1292 DWORD dst
= arg
->dst
;
1293 SHADER_BUFFER
* buffer
= arg
->buffer
;
1294 DWORD hex_version
= This
->baseShader
.hex_version
;
1297 shader_arb_get_write_mask(arg
, dst
, tmp
);
1298 if (hex_version
!= WINED3DPS_VERSION(1,4)) {
1299 DWORD reg
= dst
& WINED3DSP_REGNUM_MASK
;
1300 shader_addline(buffer
, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg
, tmp
, reg
);
1302 DWORD reg1
= dst
& WINED3DSP_REGNUM_MASK
;
1305 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, reg_src
);
1306 shader_addline(buffer
, "MOV R%u%s, %s;\n", reg1
, tmp
, reg_src
);
1310 static void pshader_hw_texreg2ar(SHADER_OPCODE_ARG
* arg
) {
1312 SHADER_BUFFER
* buffer
= arg
->buffer
;
1313 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1314 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1317 DWORD reg1
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1321 sprintf(dst_str
, "T%u", reg1
);
1322 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_str
);
1323 shader_addline(buffer
, "MOV TMP.r, %s.a;\n", src_str
);
1324 shader_addline(buffer
, "MOV TMP.g, %s.r;\n", src_str
);
1325 flags
= reg1
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg1
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1326 shader_hw_sample(arg
, reg1
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1329 static void pshader_hw_texreg2gb(SHADER_OPCODE_ARG
* arg
) {
1331 SHADER_BUFFER
* buffer
= arg
->buffer
;
1333 DWORD reg1
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1337 sprintf(dst_str
, "T%u", reg1
);
1338 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_str
);
1339 shader_addline(buffer
, "MOV TMP.r, %s.g;\n", src_str
);
1340 shader_addline(buffer
, "MOV TMP.g, %s.b;\n", src_str
);
1341 shader_hw_sample(arg
, reg1
, dst_str
, "TMP", FALSE
, FALSE
);
1344 static void pshader_hw_texreg2rgb(SHADER_OPCODE_ARG
* arg
) {
1346 SHADER_BUFFER
* buffer
= arg
->buffer
;
1347 DWORD reg1
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1351 sprintf(dst_str
, "T%u", reg1
);
1352 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_str
);
1353 shader_hw_sample(arg
, reg1
, dst_str
, src_str
, FALSE
, FALSE
);
1356 static void pshader_hw_texbem(SHADER_OPCODE_ARG
* arg
) {
1357 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1358 BOOL has_bumpmat
= FALSE
;
1359 BOOL has_luminance
= FALSE
;
1362 DWORD dst
= arg
->dst
;
1363 DWORD src
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1364 SHADER_BUFFER
* buffer
= arg
->buffer
;
1367 DWORD reg_dest_code
;
1369 /* All versions have a destination register */
1370 reg_dest_code
= dst
& WINED3DSP_REGNUM_MASK
;
1371 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1372 pshader_get_register_name(arg
->shader
, dst
, reg_coord
);
1374 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
1375 if(This
->bumpenvmatconst
[i
].const_num
!= -1 && reg_dest_code
== This
->bumpenvmatconst
[i
].texunit
) {
1380 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
1381 if(This
->luminanceconst
[i
].const_num
!= -1 && reg_dest_code
== This
->luminanceconst
[i
].texunit
) {
1382 has_luminance
= TRUE
;
1388 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1390 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
1391 shader_addline(buffer
, "DP3 TMP.r, TMP2, T%u;\n", src
);
1392 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
1393 shader_addline(buffer
, "DP3 TMP.g, TMP2, T%u;\n", src
);
1395 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1396 * so we can't let the GL handle this.
1398 if (((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
]
1399 & WINED3DTTFF_PROJECTED
) {
1400 shader_addline(buffer
, "RCP TMP2.a, %s.a;\n", reg_coord
);
1401 shader_addline(buffer
, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord
);
1402 shader_addline(buffer
, "ADD TMP.rg, TMP, TMP2;\n");
1404 shader_addline(buffer
, "ADD TMP.rg, TMP, %s;\n", reg_coord
);
1407 shader_hw_sample(arg
, reg_dest_code
, reg_coord
, "TMP", FALSE
, FALSE
);
1409 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
&& has_luminance
) {
1410 shader_addline(buffer
, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1411 src
, reg_dest_code
, reg_dest_code
);
1412 shader_addline(buffer
, "MUL %s, %s, TMP;\n", reg_coord
, reg_coord
);
1417 if(reg_dest_code
< MAX_TEXTURES
) {
1418 tf
= ((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1422 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1423 shader_hw_sample(arg
, reg_dest_code
, reg_coord
, reg_coord
, tf
& WINED3DTTFF_PROJECTED
, FALSE
);
1427 static void pshader_hw_texm3x2pad(SHADER_OPCODE_ARG
* arg
) {
1429 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1430 SHADER_BUFFER
* buffer
= arg
->buffer
;
1433 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1434 shader_addline(buffer
, "DP3 TMP.x, T%u, %s;\n", reg
, src0_name
);
1437 static void pshader_hw_texm3x2tex(SHADER_OPCODE_ARG
* arg
) {
1439 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1440 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1442 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1443 SHADER_BUFFER
* buffer
= arg
->buffer
;
1447 sprintf(dst_str
, "T%u", reg
);
1448 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1449 shader_addline(buffer
, "DP3 TMP.y, T%u, %s;\n", reg
, src0_name
);
1450 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1451 shader_hw_sample(arg
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1454 static void pshader_hw_texm3x3pad(SHADER_OPCODE_ARG
* arg
) {
1456 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1457 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1458 SHADER_BUFFER
* buffer
= arg
->buffer
;
1459 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1462 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1463 shader_addline(buffer
, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state
->current_row
, reg
, src0_name
);
1464 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
1467 static void pshader_hw_texm3x3tex(SHADER_OPCODE_ARG
* arg
) {
1469 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1470 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1472 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1473 SHADER_BUFFER
* buffer
= arg
->buffer
;
1474 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1478 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1479 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1481 /* Sample the texture using the calculated coordinates */
1482 sprintf(dst_str
, "T%u", reg
);
1483 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1484 shader_hw_sample(arg
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1485 current_state
->current_row
= 0;
1488 static void pshader_hw_texm3x3vspec(SHADER_OPCODE_ARG
* arg
) {
1490 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1491 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1493 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1494 SHADER_BUFFER
* buffer
= arg
->buffer
;
1495 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1499 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1500 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1502 /* Construct the eye-ray vector from w coordinates */
1503 shader_addline(buffer
, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[0]);
1504 shader_addline(buffer
, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[1]);
1505 shader_addline(buffer
, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg
);
1507 /* Calculate reflection vector
1509 shader_addline(buffer
, "DP3 TMP.w, TMP, TMP2;\n");
1510 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1511 shader_addline(buffer
, "DP3 TMP2.w, TMP, TMP;\n");
1512 shader_addline(buffer
, "RCP TMP2.w, TMP2.w;\n");
1513 shader_addline(buffer
, "MUL TMP.w, TMP.w, TMP2.w;\n");
1514 shader_addline(buffer
, "MUL TMP, TMP.w, TMP;\n");
1515 shader_addline(buffer
, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1517 /* Sample the texture using the calculated coordinates */
1518 sprintf(dst_str
, "T%u", reg
);
1519 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1520 shader_hw_sample(arg
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1521 current_state
->current_row
= 0;
1524 static void pshader_hw_texm3x3spec(SHADER_OPCODE_ARG
* arg
) {
1526 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1527 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1529 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1530 DWORD reg3
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1531 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1532 SHADER_BUFFER
* buffer
= arg
->buffer
;
1536 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1537 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1539 /* Calculate reflection vector.
1542 * TMP.xyz = 2 * --------- * N - E
1545 * Which normalizes the normal vector
1547 shader_addline(buffer
, "DP3 TMP.w, TMP, C[%u];\n", reg3
);
1548 shader_addline(buffer
, "DP3 TMP2.w, TMP, TMP;\n");
1549 shader_addline(buffer
, "RCP TMP2.w, TMP2.w;\n");
1550 shader_addline(buffer
, "MUL TMP.w, TMP.w, TMP2.w;\n");
1551 shader_addline(buffer
, "MUL TMP, TMP.w, TMP;\n");
1552 shader_addline(buffer
, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3
);
1554 /* Sample the texture using the calculated coordinates */
1555 sprintf(dst_str
, "T%u", reg
);
1556 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1557 shader_hw_sample(arg
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1558 current_state
->current_row
= 0;
1561 static void pshader_hw_texdepth(SHADER_OPCODE_ARG
* arg
) {
1562 SHADER_BUFFER
* buffer
= arg
->buffer
;
1565 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1566 * which is essentially an input, is the destination register because it is the first
1567 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1570 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1572 /* According to the msdn, the source register(must be r5) is unusable after
1573 * the texdepth instruction, so we're free to modify it
1575 shader_addline(buffer
, "MIN %s.g, %s.g, one.g;\n", dst_name
, dst_name
);
1577 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1578 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1579 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1581 shader_addline(buffer
, "RCP %s.g, %s.g;\n", dst_name
, dst_name
);
1582 shader_addline(buffer
, "MUL TMP.x, %s.r, %s.g;\n", dst_name
, dst_name
);
1583 shader_addline(buffer
, "MIN TMP.x, TMP.x, one.r;\n");
1584 shader_addline(buffer
, "MAX result.depth, TMP.x, 0.0;\n");
1587 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1588 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1589 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1590 static void pshader_hw_texdp3tex(SHADER_OPCODE_ARG
* arg
) {
1591 SHADER_BUFFER
* buffer
= arg
->buffer
;
1592 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1596 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0
);
1597 shader_addline(buffer
, "MOV TMP, 0.0;\n");
1598 shader_addline(buffer
, "DP3 TMP.x, T%u, %s;\n", sampler_idx
, src0
);
1600 sprintf(dst_str
, "T%u", sampler_idx
);
1601 shader_hw_sample(arg
, sampler_idx
, dst_str
, "TMP", FALSE
/* Only one coord, can't be projected */, FALSE
);
1604 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1605 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1606 static void pshader_hw_texdp3(SHADER_OPCODE_ARG
* arg
) {
1610 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1611 SHADER_BUFFER
* buffer
= arg
->buffer
;
1613 /* Handle output register */
1614 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_str
);
1615 shader_arb_get_write_mask(arg
, arg
->dst
, dst_mask
);
1617 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0
);
1618 shader_addline(buffer
, "DP3 %s%s, T%u, %s;\n", dst_str
, dst_mask
, dstreg
, src0
);
1620 /* TODO: Handle output modifiers */
1623 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1624 * Perform the 3rd row of a 3x3 matrix multiply */
1625 static void pshader_hw_texm3x3(SHADER_OPCODE_ARG
* arg
) {
1626 SHADER_BUFFER
* buffer
= arg
->buffer
;
1630 DWORD dst_reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1632 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_str
);
1633 shader_arb_get_write_mask(arg
, arg
->dst
, dst_mask
);
1635 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0
);
1636 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", dst_reg
, src0
);
1637 shader_addline(buffer
, "MOV %s%s, TMP;\n", dst_str
, dst_mask
);
1639 /* TODO: Handle output modifiers */
1642 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1643 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1644 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1645 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1647 static void pshader_hw_texm3x2depth(SHADER_OPCODE_ARG
* arg
) {
1648 SHADER_BUFFER
* buffer
= arg
->buffer
;
1649 DWORD dst_reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1652 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0
);
1653 shader_addline(buffer
, "DP3 TMP.y, T%u, %s;\n", dst_reg
, src0
);
1655 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1656 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1657 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1659 shader_addline(buffer
, "RCP TMP.y, TMP.y;\n");
1660 shader_addline(buffer
, "MUL TMP.x, TMP.x, TMP.y;\n");
1661 shader_addline(buffer
, "MIN TMP.x, TMP.x, one.r;\n");
1662 shader_addline(buffer
, "MAX result.depth, TMP.x, 0.0;\n");
1665 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1666 Vertex/Pixel shaders to ARB_vertex_program codes */
1667 static void shader_hw_mnxn(SHADER_OPCODE_ARG
* arg
) {
1670 int nComponents
= 0;
1671 SHADER_OPCODE_ARG tmpArg
;
1673 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1675 /* Set constants for the temporary argument */
1676 tmpArg
.shader
= arg
->shader
;
1677 tmpArg
.buffer
= arg
->buffer
;
1678 tmpArg
.src
[0] = arg
->src
[0];
1679 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1680 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1681 tmpArg
.reg_maps
= arg
->reg_maps
;
1683 switch(arg
->opcode
->opcode
) {
1684 case WINED3DSIO_M4x4
:
1686 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP4
);
1688 case WINED3DSIO_M4x3
:
1690 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP4
);
1692 case WINED3DSIO_M3x4
:
1694 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1696 case WINED3DSIO_M3x3
:
1698 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1700 case WINED3DSIO_M3x2
:
1702 tmpArg
.opcode
= shader_get_opcode(arg
->shader
, WINED3DSIO_DP3
);
1708 for (i
= 0; i
< nComponents
; i
++) {
1709 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
1710 tmpArg
.src
[1] = arg
->src
[1]+i
;
1711 shader_hw_map2gl(&tmpArg
);
1715 static void vshader_hw_rsq_rcp(SHADER_OPCODE_ARG
* arg
) {
1716 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1717 SHADER_BUFFER
* buffer
= arg
->buffer
;
1718 DWORD dst
= arg
->dst
;
1719 DWORD src
= arg
->src
[0];
1720 DWORD swizzle
= (src
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1724 strcpy(tmpLine
, curOpcode
->glname
); /* Opcode */
1725 vshader_program_add_param(arg
, dst
, FALSE
, tmpLine
); /* Destination */
1726 strcat(tmpLine
, ",");
1727 vshader_program_add_param(arg
, src
, TRUE
, tmpLine
);
1728 if ((WINED3DSP_NOSWIZZLE
>> WINED3DSP_SWIZZLE_SHIFT
) == swizzle
) {
1729 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1732 strcat(tmpLine
, ".w");
1735 shader_addline(buffer
, "%s;\n", tmpLine
);
1738 static void shader_hw_nrm(SHADER_OPCODE_ARG
* arg
) {
1739 SHADER_BUFFER
* buffer
= arg
->buffer
;
1743 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1744 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
1746 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1747 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
1749 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
);
1750 shader_addline(buffer
, "DP3 TMP, %s, %s;\n", src_name
, src_name
);
1751 shader_addline(buffer
, "RSQ TMP, TMP.x;\n");
1752 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1753 shader_addline(buffer
, "MUL%s %s%s, %s, TMP;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
1757 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1760 static void shader_hw_sincos(SHADER_OPCODE_ARG
* arg
) {
1761 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1762 * must contain fixed constants. So we need a separate function to filter those constants and
1765 SHADER_BUFFER
* buffer
= arg
->buffer
;
1769 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1770 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
1772 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1773 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
1775 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
);
1776 shader_addline(buffer
, "SCS%s %s%s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
1780 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1784 static GLuint
create_arb_blt_vertex_program(WineD3D_GL_Info
*gl_info
) {
1785 GLuint program_id
= 0;
1786 const char *blt_vprogram
=
1788 "PARAM c[1] = { { 1, 0.5 } };\n"
1789 "MOV result.position, vertex.position;\n"
1790 "MOV result.color, c[0].x;\n"
1791 "MAD result.texcoord[0].y, -vertex.position, c[0], c[0];\n"
1792 "MAD result.texcoord[0].x, vertex.position, c[0].y, c[0].y;\n"
1795 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
1796 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
1797 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_vprogram
), blt_vprogram
));
1799 if (glGetError() == GL_INVALID_OPERATION
) {
1801 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
1802 FIXME("Vertex program error at position %d: %s\n", pos
,
1803 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
1809 static GLuint
create_arb_blt_fragment_program(WineD3D_GL_Info
*gl_info
) {
1810 GLuint program_id
= 0;
1811 const char *blt_fprogram
=
1814 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1815 "MOV result.depth.z, R0.x;\n"
1818 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
1819 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
1820 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_fprogram
), blt_fprogram
));
1822 if (glGetError() == GL_INVALID_OPERATION
) {
1824 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
1825 FIXME("Fragment program error at position %d: %s\n", pos
,
1826 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
1832 static void shader_arb_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
1833 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1834 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->shader_priv
;
1835 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1838 TRACE("Using vertex shader\n");
1840 priv
->current_vprogram_id
= ((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
)->baseShader
.prgId
;
1842 /* Bind the vertex program */
1843 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
1844 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1846 /* Enable OpenGL vertex programs */
1847 glEnable(GL_VERTEX_PROGRAM_ARB
);
1848 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1849 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
1850 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
1851 priv
->current_vprogram_id
= 0;
1852 glDisable(GL_VERTEX_PROGRAM_ARB
);
1853 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1857 TRACE("Using pixel shader\n");
1859 priv
->current_fprogram_id
= ((IWineD3DPixelShaderImpl
*)This
->stateBlock
->pixelShader
)->baseShader
.prgId
;
1861 /* Bind the fragment program */
1862 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
1863 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1865 /* Enable OpenGL fragment programs */
1866 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1867 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1868 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
1870 priv
->current_fprogram_id
= 0;
1872 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
) && !priv
->use_arbfp_fixed_func
) {
1873 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1874 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1875 * replacement shader
1877 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
1878 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1883 static void shader_arb_select_depth_blt(IWineD3DDevice
*iface
) {
1884 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1885 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->shader_priv
;
1886 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1888 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
1889 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
1890 glEnable(GL_VERTEX_PROGRAM_ARB
);
1892 if (!priv
->depth_blt_fprogram_id
) priv
->depth_blt_fprogram_id
= create_arb_blt_fragment_program(gl_info
);
1893 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->depth_blt_fprogram_id
));
1894 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1897 static void shader_arb_deselect_depth_blt(IWineD3DDevice
*iface
) {
1898 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1899 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->shader_priv
;
1900 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1902 if (priv
->current_vprogram_id
) {
1903 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
1904 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1906 glEnable(GL_VERTEX_PROGRAM_ARB
);
1907 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1909 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
1911 glDisable(GL_VERTEX_PROGRAM_ARB
);
1912 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1915 if (priv
->current_fprogram_id
) {
1916 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
1917 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1919 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1920 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1922 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
1923 } else if(!priv
->use_arbfp_fixed_func
) {
1924 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
1925 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1929 static void shader_arb_cleanup(IWineD3DDevice
*iface
) {
1930 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1931 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1932 if (GL_SUPPORT(ARB_VERTEX_PROGRAM
)) glDisable(GL_VERTEX_PROGRAM_ARB
);
1933 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM
)) glDisable(GL_FRAGMENT_PROGRAM_ARB
);
1936 static void shader_arb_destroy(IWineD3DBaseShader
*iface
) {
1937 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
1938 WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->adapter
->gl_info
;
1941 GL_EXTCALL(glDeleteProgramsARB(1, &This
->baseShader
.prgId
));
1942 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->baseShader.prgId))");
1944 This
->baseShader
.prgId
= 0;
1945 This
->baseShader
.is_compiled
= FALSE
;
1948 static HRESULT
shader_arb_alloc(IWineD3DDevice
*iface
) {
1949 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1950 This
->shader_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
1954 static void shader_arb_free(IWineD3DDevice
*iface
) {
1955 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1956 WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1957 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->shader_priv
;
1959 if(priv
->depth_blt_vprogram_id
) {
1960 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
1962 if(priv
->depth_blt_fprogram_id
) {
1963 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id
));
1966 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
1969 static BOOL
shader_arb_dirty_const(IWineD3DDevice
*iface
) {
1973 static void arbfp_add_sRGB_correction(SHADER_BUFFER
*buffer
, const char *fragcolor
, const char *tmp1
,
1974 const char *tmp2
, const char *tmp3
, const char *tmp4
) {
1975 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1977 /* Calculate the > 0.0031308 case */
1978 shader_addline(buffer
, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1
, fragcolor
);
1979 shader_addline(buffer
, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1
, fragcolor
);
1980 shader_addline(buffer
, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1
, fragcolor
);
1981 shader_addline(buffer
, "MUL %s, %s, srgb_mul_hi;\n", tmp1
, tmp1
);
1982 shader_addline(buffer
, "SUB %s, %s, srgb_sub_hi;\n", tmp1
, tmp1
);
1983 /* Calculate the < case */
1984 shader_addline(buffer
, "MUL %s, srgb_mul_low, %s;\n", tmp2
, fragcolor
);
1985 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1986 shader_addline(buffer
, "SLT %s, srgb_comparison, %s;\n", tmp3
, fragcolor
);
1987 shader_addline(buffer
, "SGE %s, srgb_comparison, %s;\n", tmp4
, fragcolor
);
1988 /* Store the components > 0.0031308 in the destination */
1989 shader_addline(buffer
, "MUL %s, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
1990 /* Add the components that are < 0.0031308 */
1991 shader_addline(buffer
, "MAD result.color.xyz, %s, %s, %s;\n", tmp2
, tmp4
, fragcolor
);
1992 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1995 static void shader_arb_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
) {
1996 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
1997 shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
1998 CONST DWORD
*function
= This
->baseShader
.function
;
1999 const char *fragcolor
;
2000 WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
2001 local_constant
* lconst
;
2003 /* Create the hw ARB shader */
2004 shader_addline(buffer
, "!!ARBfp1.0\n");
2006 shader_addline(buffer
, "TEMP TMP;\n"); /* Used in matrix ops */
2007 shader_addline(buffer
, "TEMP TMP2;\n"); /* Used in matrix ops */
2008 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
2009 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
2010 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
2011 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
2012 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
2013 shader_addline(buffer
, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
2015 /* Base Declarations */
2016 shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
2018 /* We need two variables for fog blending */
2019 shader_addline(buffer
, "TEMP TMP_FOG;\n");
2020 if (This
->baseShader
.hex_version
>= WINED3DPS_VERSION(2,0)) {
2021 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
2024 /* Base Shader Body */
2025 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
2027 /* calculate fog and blend it
2028 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
2029 * -1/(e-s) and e/(e-s) respectively.
2031 shader_addline(buffer
, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
2033 if (This
->baseShader
.hex_version
< WINED3DPS_VERSION(2,0)) {
2036 fragcolor
= "TMP_COLOR";
2038 if(This
->srgb_enabled
) {
2039 arbfp_add_sRGB_correction(buffer
, fragcolor
, "TMP", "TMP2", "TA", "TB");
2041 if (This
->baseShader
.hex_version
< WINED3DPS_VERSION(3,0)) {
2042 shader_addline(buffer
, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor
);
2043 shader_addline(buffer
, "MOV result.color.a, %s.a;\n", fragcolor
);
2046 shader_addline(buffer
, "END\n");
2048 /* TODO: change to resource.glObjectHandle or something like that */
2049 GL_EXTCALL(glGenProgramsARB(1, &This
->baseShader
.prgId
));
2051 TRACE("Creating a hw pixel shader, prg=%d\n", This
->baseShader
.prgId
);
2052 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, This
->baseShader
.prgId
));
2054 TRACE("Created hw pixel shader, prg=%d\n", This
->baseShader
.prgId
);
2055 /* Create the program and check for errors */
2056 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
2057 buffer
->bsize
, buffer
->buffer
));
2059 if (glGetError() == GL_INVALID_OPERATION
) {
2061 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
2062 FIXME("HW PixelShader Error at position %d: %s\n",
2063 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2064 This
->baseShader
.prgId
= -1;
2067 /* Load immediate constants */
2068 if(!This
->baseShader
.load_local_constsF
) {
2069 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
2070 float *value
= (float *) lconst
->value
;
2071 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, lconst
->idx
, value
));
2072 checkGLcall("glProgramLocalParameter4fvARB");
2077 static void shader_arb_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
) {
2078 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
2079 shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
2080 CONST DWORD
*function
= This
->baseShader
.function
;
2081 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
2082 WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
2083 local_constant
* lconst
;
2085 /* Create the hw ARB shader */
2086 shader_addline(buffer
, "!!ARBvp1.0\n");
2087 shader_addline(buffer
, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This
->rel_offset
);
2089 /* Mesa supports only 95 constants */
2090 if (GL_VEND(MESA
) || GL_VEND(WINE
))
2091 This
->baseShader
.limits
.constant_float
=
2092 min(95, This
->baseShader
.limits
.constant_float
);
2094 shader_addline(buffer
, "TEMP TMP;\n");
2096 /* Base Declarations */
2097 shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
2099 /* We need a constant to fixup the final position */
2100 shader_addline(buffer
, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS
);
2102 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2103 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2104 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2105 * a replacement shader depend on the texcoord.w being set properly.
2107 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2108 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2109 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2110 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2111 * this can eat a number of instructions, so skip it unless this cap is set as well
2113 if(!GL_SUPPORT(NV_VERTEX_PROGRAM
)) {
2114 shader_addline(buffer
, "MOV result.color.secondary, -helper_const.wwwy;\n");
2116 if((GLINFO_LOCATION
).set_texcoord_w
&& !device
->frag_pipe
->ffp_proj_control
) {
2118 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
2119 if(This
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
2120 This
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
2121 shader_addline(buffer
, "MOV result.texcoord[%u].w, -helper_const.y;\n", i
);
2127 /* Base Shader Body */
2128 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
2130 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2132 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
2134 /* Write the final position.
2136 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2137 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2138 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2139 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2141 shader_addline(buffer
, "MUL TMP, posFixup, TMP_OUT.w;\n");
2142 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2143 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2145 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2146 * and the glsl equivalent
2148 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2150 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
2152 shader_addline(buffer
, "END\n");
2154 /* TODO: change to resource.glObjectHandle or something like that */
2155 GL_EXTCALL(glGenProgramsARB(1, &This
->baseShader
.prgId
));
2157 TRACE("Creating a hw vertex shader, prg=%d\n", This
->baseShader
.prgId
);
2158 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, This
->baseShader
.prgId
));
2160 TRACE("Created hw vertex shader, prg=%d\n", This
->baseShader
.prgId
);
2161 /* Create the program and check for errors */
2162 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
2163 buffer
->bsize
, buffer
->buffer
));
2165 if (glGetError() == GL_INVALID_OPERATION
) {
2167 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
2168 FIXME("HW VertexShader Error at position %d: %s\n",
2169 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2170 This
->baseShader
.prgId
= -1;
2173 /* Load immediate constants */
2174 if(!This
->baseShader
.load_local_constsF
) {
2175 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
2176 float *value
= (float *) lconst
->value
;
2177 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, lconst
->idx
, value
));
2182 static void shader_arb_get_caps(WINED3DDEVTYPE devtype
, WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
) {
2183 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2184 * then overwrite the shader specific ones
2186 none_shader_backend
.shader_get_caps(devtype
, gl_info
, pCaps
);
2188 if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
2189 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(1,1);
2190 TRACE_(d3d_caps
)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2191 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
2194 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
)) {
2195 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(1,4);
2196 pCaps
->PixelShader1xMaxValue
= 8.0;
2197 TRACE_(d3d_caps
)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2201 static BOOL
shader_arb_conv_supported(WINED3DFORMAT fmt
) {
2202 TRACE("Checking shader format support for format %s:", debug_d3dformat(fmt
));
2204 case WINED3DFMT_V8U8
:
2205 case WINED3DFMT_V16U16
:
2206 case WINED3DFMT_X8L8V8U8
:
2207 case WINED3DFMT_L6V5U5
:
2208 case WINED3DFMT_Q8W8V8U8
:
2209 case WINED3DFMT_ATI2N
:
2218 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
2220 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
2221 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
2222 /* WINED3DSIH_BEM */ pshader_hw_bem
,
2223 /* WINED3DSIH_BREAK */ NULL
,
2224 /* WINED3DSIH_BREAKC */ NULL
,
2225 /* WINED3DSIH_BREAKP */ NULL
,
2226 /* WINED3DSIH_CALL */ NULL
,
2227 /* WINED3DSIH_CALLNZ */ NULL
,
2228 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
2229 /* WINED3DSIH_CND */ pshader_hw_cnd
,
2230 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
2231 /* WINED3DSIH_DCL */ NULL
,
2232 /* WINED3DSIH_DEF */ NULL
,
2233 /* WINED3DSIH_DEFB */ NULL
,
2234 /* WINED3DSIH_DEFI */ NULL
,
2235 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
2236 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
2237 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
2238 /* WINED3DSIH_DST */ shader_hw_map2gl
,
2239 /* WINED3DSIH_DSX */ NULL
,
2240 /* WINED3DSIH_DSY */ NULL
,
2241 /* WINED3DSIH_ELSE */ NULL
,
2242 /* WINED3DSIH_ENDIF */ NULL
,
2243 /* WINED3DSIH_ENDLOOP */ NULL
,
2244 /* WINED3DSIH_ENDREP */ NULL
,
2245 /* WINED3DSIH_EXP */ shader_hw_map2gl
,
2246 /* WINED3DSIH_EXPP */ shader_hw_map2gl
,
2247 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
2248 /* WINED3DSIH_IF */ NULL
,
2249 /* WINED3DSIH_IFC */ NULL
,
2250 /* WINED3DSIH_LABEL */ NULL
,
2251 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
2252 /* WINED3DSIH_LOG */ shader_hw_map2gl
,
2253 /* WINED3DSIH_LOGP */ shader_hw_map2gl
,
2254 /* WINED3DSIH_LOOP */ NULL
,
2255 /* WINED3DSIH_LRP */ shader_hw_map2gl
,
2256 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
2257 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
2258 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
2259 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
2260 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
2261 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
2262 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
2263 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
2264 /* WINED3DSIH_MOV */ shader_hw_map2gl
,
2265 /* WINED3DSIH_MOVA */ shader_hw_map2gl
,
2266 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
2267 /* WINED3DSIH_NOP */ shader_hw_map2gl
,
2268 /* WINED3DSIH_NRM */ shader_hw_nrm
,
2269 /* WINED3DSIH_PHASE */ NULL
,
2270 /* WINED3DSIH_POW */ shader_hw_map2gl
,
2271 /* WINED3DSIH_RCP */ vshader_hw_rsq_rcp
,
2272 /* WINED3DSIH_REP */ NULL
,
2273 /* WINED3DSIH_RET */ NULL
,
2274 /* WINED3DSIH_RSQ */ vshader_hw_rsq_rcp
,
2275 /* WINED3DSIH_SETP */ NULL
,
2276 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
2277 /* WINED3DSIH_SGN */ NULL
,
2278 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
2279 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
2280 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
2281 /* WINED3DSIH_TEX */ pshader_hw_tex
,
2282 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
2283 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
2284 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
2285 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
2286 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
2287 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
2288 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
2289 /* WINED3DSIH_TEXLDD */ NULL
,
2290 /* WINED3DSIH_TEXLDL */ NULL
,
2291 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
2292 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
2293 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
2294 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
2295 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
2296 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
2297 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
2298 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
2299 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
2300 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
2301 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
2302 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
2305 const shader_backend_t arb_program_shader_backend
= {
2306 shader_arb_instruction_handler_table
,
2308 shader_arb_select_depth_blt
,
2309 shader_arb_deselect_depth_blt
,
2310 shader_arb_load_constants
,
2312 shader_arb_color_correction
,
2316 shader_arb_dirty_const
,
2317 shader_arb_generate_pshader
,
2318 shader_arb_generate_vshader
,
2319 shader_arb_get_caps
,
2320 shader_arb_conv_supported
,
2323 /* ARB_fragment_program fixed function pipeline replacement definitions */
2324 #define ARB_FFP_CONST_TFACTOR 0
2325 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2326 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2327 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2328 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2330 struct arbfp_ffp_desc
2332 struct ffp_desc parent
;
2334 unsigned int num_textures_used
;
2337 static void arbfp_enable(IWineD3DDevice
*iface
, BOOL enable
) {
2339 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
2340 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2342 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
2343 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2347 static HRESULT
arbfp_alloc(IWineD3DDevice
*iface
) {
2348 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
2349 struct shader_arb_priv
*priv
;
2350 /* Share private data between the shader backend and the pipeline replacement, if both
2351 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2352 * if no pixel shader is bound or not
2354 if(This
->shader_backend
== &arb_program_shader_backend
) {
2355 This
->fragment_priv
= This
->shader_priv
;
2357 This
->fragment_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
2358 if(!This
->fragment_priv
) return E_OUTOFMEMORY
;
2360 priv
= (struct shader_arb_priv
*) This
->fragment_priv
;
2361 priv
->fragment_shaders
= hash_table_create(ffp_program_key_hash
, ffp_program_key_compare
);
2362 priv
->use_arbfp_fixed_func
= TRUE
;
2366 static void arbfp_free_ffpshader(void *value
, void *gli
) {
2367 WineD3D_GL_Info
*gl_info
= gli
;
2368 struct arbfp_ffp_desc
*entry_arb
= value
;
2371 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
2372 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2373 HeapFree(GetProcessHeap(), 0, entry_arb
);
2377 static void arbfp_free(IWineD3DDevice
*iface
) {
2378 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
2379 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->fragment_priv
;
2381 hash_table_destroy(priv
->fragment_shaders
, arbfp_free_ffpshader
, &This
->adapter
->gl_info
);
2382 priv
->use_arbfp_fixed_func
= FALSE
;
2384 if(This
->shader_backend
!= &arb_program_shader_backend
) {
2385 HeapFree(GetProcessHeap(), 0, This
->fragment_priv
);
2389 static void arbfp_get_caps(WINED3DDEVTYPE devtype
, WineD3D_GL_Info
*gl_info
, struct fragment_caps
*caps
) {
2390 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
2391 WINED3DTEXOPCAPS_SELECTARG1
|
2392 WINED3DTEXOPCAPS_SELECTARG2
|
2393 WINED3DTEXOPCAPS_MODULATE4X
|
2394 WINED3DTEXOPCAPS_MODULATE2X
|
2395 WINED3DTEXOPCAPS_MODULATE
|
2396 WINED3DTEXOPCAPS_ADDSIGNED2X
|
2397 WINED3DTEXOPCAPS_ADDSIGNED
|
2398 WINED3DTEXOPCAPS_ADD
|
2399 WINED3DTEXOPCAPS_SUBTRACT
|
2400 WINED3DTEXOPCAPS_ADDSMOOTH
|
2401 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
2402 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
2403 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
2404 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
2405 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
2406 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
2407 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
2408 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
2409 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
2410 WINED3DTEXOPCAPS_DOTPRODUCT3
|
2411 WINED3DTEXOPCAPS_MULTIPLYADD
|
2412 WINED3DTEXOPCAPS_LERP
|
2413 WINED3DTEXOPCAPS_BUMPENVMAP
|
2414 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
2416 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2418 caps
->MaxTextureBlendStages
= 8;
2419 caps
->MaxSimultaneousTextures
= min(GL_LIMITS(fragment_samplers
), 8);
2421 caps
->PrimitiveMiscCaps
|= WINED3DPMISCCAPS_TSSARGTEMP
;
2423 #undef GLINFO_LOCATION
2425 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2426 static void state_texfactor_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2428 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2430 /* Do not overwrite pixel shader constants if a pshader is in use */
2431 if(use_ps(device
)) return;
2433 D3DCOLORTOGLFLOAT4(stateblock
->renderState
[WINED3DRS_TEXTUREFACTOR
], col
);
2434 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
2435 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2437 if(device
->shader_backend
== &arb_program_shader_backend
) {
2438 device
= stateblock
->wineD3DDevice
;
2439 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
2440 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
2444 static void state_arb_specularenable(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2446 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2448 /* Do not overwrite pixel shader constants if a pshader is in use */
2449 if(use_ps(device
)) return;
2451 if(stateblock
->renderState
[WINED3DRS_SPECULARENABLE
]) {
2452 /* The specular color has no alpha */
2453 col
[0] = 1.0; col
[1] = 1.0;
2454 col
[2] = 1.0; col
[3] = 0.0;
2456 col
[0] = 0.0; col
[1] = 0.0;
2457 col
[2] = 0.0; col
[3] = 0.0;
2459 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
2460 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2462 if(device
->shader_backend
== &arb_program_shader_backend
) {
2463 device
= stateblock
->wineD3DDevice
;
2464 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
2465 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
2469 static void set_bumpmat_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2470 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE
;
2471 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2474 if(use_ps(device
)) {
2476 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.bumpmat
[stage
]) {
2477 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2480 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
2481 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
2484 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2488 mat
[0][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT00
]);
2489 mat
[0][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT01
]);
2490 mat
[1][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT10
]);
2491 mat
[1][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT11
]);
2493 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
2494 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2496 if(device
->shader_backend
== &arb_program_shader_backend
) {
2497 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
2498 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
2502 static void tex_bumpenvlum_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2503 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE
;
2504 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2507 if(use_ps(device
)) {
2509 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.luminanceparams
[stage
]) {
2510 /* The pixel shader has to know the luminance offset. Do a constants update if it
2511 * isn't scheduled anyway
2513 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
2514 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
2517 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2521 param
[0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
]);
2522 param
[1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
]);
2526 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
2527 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2529 if(device
->shader_backend
== &arb_program_shader_backend
) {
2530 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
2531 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
2535 static const char *get_argreg(SHADER_BUFFER
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
) {
2538 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
2540 switch(arg
& WINED3DTA_SELECTMASK
) {
2541 case WINED3DTA_DIFFUSE
:
2542 ret
= "fragment.color.primary"; break;
2544 case WINED3DTA_CURRENT
:
2545 if(stage
== 0) ret
= "fragment.color.primary";
2549 case WINED3DTA_TEXTURE
:
2551 case 0: ret
= "tex0"; break;
2552 case 1: ret
= "tex1"; break;
2553 case 2: ret
= "tex2"; break;
2554 case 3: ret
= "tex3"; break;
2555 case 4: ret
= "tex4"; break;
2556 case 5: ret
= "tex5"; break;
2557 case 6: ret
= "tex6"; break;
2558 case 7: ret
= "tex7"; break;
2559 default: ret
= "unknown texture";
2563 case WINED3DTA_TFACTOR
:
2564 ret
= "tfactor"; break;
2566 case WINED3DTA_SPECULAR
:
2567 ret
= "fragment.color.secondary"; break;
2569 case WINED3DTA_TEMP
:
2570 ret
= "tempreg"; break;
2572 case WINED3DTA_CONSTANT
:
2573 FIXME("Implement perstage constants\n");
2575 case 0: ret
= "const0"; break;
2576 case 1: ret
= "const1"; break;
2577 case 2: ret
= "const2"; break;
2578 case 3: ret
= "const3"; break;
2579 case 4: ret
= "const4"; break;
2580 case 5: ret
= "const5"; break;
2581 case 6: ret
= "const6"; break;
2582 case 7: ret
= "const7"; break;
2588 if(arg
& WINED3DTA_COMPLEMENT
) {
2589 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
2590 if(argnum
== 0) ret
= "arg0";
2591 if(argnum
== 1) ret
= "arg1";
2592 if(argnum
== 2) ret
= "arg2";
2594 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
2595 shader_addline(buffer
, "MOV arg%u, %s.a;\n", argnum
, ret
);
2596 if(argnum
== 0) ret
= "arg0";
2597 if(argnum
== 1) ret
= "arg1";
2598 if(argnum
== 2) ret
= "arg2";
2603 static void gen_ffp_instr(SHADER_BUFFER
*buffer
, unsigned int stage
, BOOL color
, BOOL alpha
,
2604 DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
) {
2605 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
2606 unsigned int mul
= 1;
2607 BOOL mul_final_dest
= FALSE
;
2609 if(color
&& alpha
) dstmask
= "";
2610 else if(color
) dstmask
= ".rgb";
2611 else dstmask
= ".a";
2613 if(dst
== tempreg
) dstreg
= "tempreg";
2614 else dstreg
= "ret";
2616 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
2617 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
2618 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
2621 case WINED3DTOP_DISABLE
:
2622 if(stage
== 1) shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
2625 case WINED3DTOP_SELECTARG2
:
2627 case WINED3DTOP_SELECTARG1
:
2628 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
2631 case WINED3DTOP_MODULATE4X
:
2633 case WINED3DTOP_MODULATE2X
:
2635 if(strcmp(dstreg
, "result.color") == 0) {
2637 mul_final_dest
= TRUE
;
2639 case WINED3DTOP_MODULATE
:
2640 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2643 case WINED3DTOP_ADDSIGNED2X
:
2645 if(strcmp(dstreg
, "result.color") == 0) {
2647 mul_final_dest
= TRUE
;
2649 case WINED3DTOP_ADDSIGNED
:
2650 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
2652 case WINED3DTOP_ADD
:
2653 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2656 case WINED3DTOP_SUBTRACT
:
2657 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2660 case WINED3DTOP_ADDSMOOTH
:
2661 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
2662 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2665 case WINED3DTOP_BLENDCURRENTALPHA
:
2666 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
2667 shader_addline(buffer
, "LRP %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2669 case WINED3DTOP_BLENDFACTORALPHA
:
2670 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
2671 shader_addline(buffer
, "LRP %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2673 case WINED3DTOP_BLENDTEXTUREALPHA
:
2674 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
2675 shader_addline(buffer
, "LRP %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2677 case WINED3DTOP_BLENDDIFFUSEALPHA
:
2678 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
2679 shader_addline(buffer
, "LRP %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2682 case WINED3DTOP_BLENDTEXTUREALPHAPM
:
2683 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
2684 shader_addline(buffer
, "SUB arg0.a, const.x, %s.a;\n", arg0
);
2685 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.a, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2688 /* D3DTOP_PREMODULATE ???? */
2690 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR
:
2691 shader_addline(buffer
, "SUB arg0.a, const.x, %s;\n", arg1
);
2692 shader_addline(buffer
, "MAD_SAT %s%s, arg0.a, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2694 case WINED3DTOP_MODULATEALPHA_ADDCOLOR
:
2695 shader_addline(buffer
, "MAD_SAT %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
2697 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA
:
2698 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
2699 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.a;\n", dstreg
, dstmask
, arg2
, arg1
);
2701 case WINED3DTOP_MODULATECOLOR_ADDALPHA
:
2702 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.a;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
2705 case WINED3DTOP_DOTPRODUCT3
:
2707 if(strcmp(dstreg
, "result.color") == 0) {
2709 mul_final_dest
= TRUE
;
2711 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
2712 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
2713 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
2716 case WINED3DTOP_MULTIPLYADD
:
2717 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
2720 case WINED3DTOP_LERP
:
2721 /* The msdn is not quite right here */
2722 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2725 case WINED3DTOP_BUMPENVMAP
:
2726 case WINED3DTOP_BUMPENVMAPLUMINANCE
:
2727 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2731 FIXME("Unhandled texture op %08x\n", op
);
2735 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
2736 } else if(mul
== 4) {
2737 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
2741 /* The stateblock is passed for GLINFO_LOCATION */
2742 static GLuint
gen_arbfp_ffp_shader(struct ffp_settings
*settings
, IWineD3DStateBlockImpl
*stateblock
) {
2744 SHADER_BUFFER buffer
;
2745 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2746 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2747 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2748 const char *textype
;
2749 const char *instr
, *sat
;
2750 char colorcor_dst
[8];
2752 DWORD arg0
, arg1
, arg2
;
2753 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
2755 const char *final_combiner_src
= "ret";
2757 /* Find out which textures are read */
2758 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2759 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) break;
2760 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
2761 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
2762 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
2763 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2764 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2765 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2767 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHA
) tex_read
[stage
] = TRUE
;
2768 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHAPM
) tex_read
[stage
] = TRUE
;
2769 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
) {
2770 bump_used
[stage
] = TRUE
;
2771 tex_read
[stage
] = TRUE
;
2773 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2774 bump_used
[stage
] = TRUE
;
2775 tex_read
[stage
] = TRUE
;
2776 luminance_used
[stage
] = TRUE
;
2777 } else if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDFACTORALPHA
) {
2778 tfactor_used
= TRUE
;
2781 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
2782 tfactor_used
= TRUE
;
2785 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
2786 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
2787 tempreg_used
= TRUE
;
2790 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) continue;
2791 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
2792 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
2793 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
2794 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2795 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2796 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2798 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
2799 tempreg_used
= TRUE
;
2801 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
2802 tfactor_used
= TRUE
;
2809 buffer
.newline
= TRUE
;
2810 buffer
.buffer
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, SHADER_PGMSIZE
);
2812 shader_addline(&buffer
, "!!ARBfp1.0\n");
2814 switch(settings
->fog
) {
2815 case FOG_OFF
: break;
2816 case FOG_LINEAR
: shader_addline(&buffer
, "OPTION ARB_fog_linear;\n"); break;
2817 case FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
2818 case FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
2819 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
2822 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
2823 shader_addline(&buffer
, "TEMP TMP;\n");
2824 shader_addline(&buffer
, "TEMP ret;\n");
2825 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
2826 shader_addline(&buffer
, "TEMP arg0;\n");
2827 shader_addline(&buffer
, "TEMP arg1;\n");
2828 shader_addline(&buffer
, "TEMP arg2;\n");
2829 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2830 if(!tex_read
[stage
]) continue;
2831 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
2832 if(!bump_used
[stage
]) continue;
2833 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
2834 if(!luminance_used
[stage
]) continue;
2835 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
2838 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
2840 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
2842 if(settings
->sRGB_write
) {
2843 shader_addline(&buffer
, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2844 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
2845 shader_addline(&buffer
, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2846 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
2847 shader_addline(&buffer
, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2848 srgb_pow
, srgb_pow
, srgb_pow
);
2849 shader_addline(&buffer
, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2850 srgb_mul_high
, srgb_mul_high
, srgb_mul_high
);
2851 shader_addline(&buffer
, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2852 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
2855 /* Generate texture sampling instructions) */
2856 for(stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3DTOP_DISABLE
; stage
++) {
2857 if(!tex_read
[stage
]) continue;
2859 switch(settings
->op
[stage
].tex_type
) {
2860 case tex_1d
: textype
= "1D"; break;
2861 case tex_2d
: textype
= "2D"; break;
2862 case tex_3d
: textype
= "3D"; break;
2863 case tex_cube
: textype
= "CUBE"; break;
2864 case tex_rect
: textype
= "RECT"; break;
2865 default: textype
= "unexpected_textype"; break;
2868 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
||
2869 settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2875 if(settings
->op
[stage
].projected
== proj_none
) {
2877 } else if(settings
->op
[stage
].projected
== proj_count4
||
2878 settings
->op
[stage
].projected
== proj_count3
) {
2881 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
2886 (settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAP
||
2887 settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
)) {
2888 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
2889 shader_addline(&buffer
, "DP3 ret.r, arg1, tex%u;\n", stage
- 1);
2890 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
2891 shader_addline(&buffer
, "DP3 ret.g, arg1, tex%u;\n", stage
- 1);
2893 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2894 * so multiply the displacement with the dividing parameter before passing it to TXP
2896 if (settings
->op
[stage
].projected
!= proj_none
) {
2897 if(settings
->op
[stage
].projected
== proj_count4
) {
2898 shader_addline(&buffer
, "MOV ret.a, fragment.texcoord[%u].a;\n", stage
);
2899 shader_addline(&buffer
, "MUL ret.rgb, ret, fragment.texcoord[%u].a, fragment.texcoord[%u];\n", stage
, stage
);
2901 shader_addline(&buffer
, "MOV ret.a, fragment.texcoord[%u].b;\n", stage
);
2902 shader_addline(&buffer
, "MAD ret.rgb, ret, fragment.texcoord[%u].b, fragment.texcoord[%u];\n", stage
, stage
);
2905 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
2908 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
2909 instr
, sat
, stage
, stage
, textype
);
2910 if(settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2911 shader_addline(&buffer
, "MAD_SAT ret.r, tex%u.b, luminance%u.r, luminance%u.g;\n",
2912 stage
- 1, stage
- 1, stage
- 1);
2913 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.r;\n", stage
, stage
);
2915 } else if(settings
->op
[stage
].projected
== proj_count3
) {
2916 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
2917 shader_addline(&buffer
, "MOV ret.a, ret.b;\n");
2918 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
2919 instr
, sat
, stage
, stage
, textype
);
2921 shader_addline(&buffer
, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2922 instr
, sat
, stage
, stage
, stage
, textype
);
2925 sprintf(colorcor_dst
, "tex%u", stage
);
2926 gen_color_correction(&buffer
, colorcor_dst
, ".rgba", "const.x", "const.y",
2927 settings
->op
[stage
].color_correction
, &GLINFO_LOCATION
);
2930 /* Generate the main shader */
2931 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2932 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) {
2934 final_combiner_src
= "fragment.color.primary";
2939 if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
2940 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
2941 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
2942 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
2943 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
2944 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
2945 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
2946 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
2947 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
2948 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
2949 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
2950 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
2952 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
&&
2953 settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
&&
2954 settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
&&
2955 settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
2958 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) {
2959 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
2960 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2961 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2963 shader_addline(&buffer
, "MOV ret.a, fragment.color.primary.a;\n");
2965 } else if(op_equal
) {
2966 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
2967 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2968 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2970 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
2971 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2972 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2973 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
2974 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
2975 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
2979 if(settings
->sRGB_write
) {
2980 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
2981 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg");
2982 shader_addline(&buffer
, "MOV result.color.a, ret.a;\n");
2984 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
2988 shader_addline(&buffer
, "END\n");
2990 /* Generate the shader */
2991 GL_EXTCALL(glGenProgramsARB(1, &ret
));
2992 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
2993 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
2995 if (glGetError() == GL_INVALID_OPERATION
) {
2997 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
2998 FIXME("Fragment program error at position %d: %s\n", pos
,
2999 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3001 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3005 static void fragment_prog_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3006 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
3007 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) device
->fragment_priv
;
3008 BOOL use_pshader
= use_ps(device
);
3009 BOOL use_vshader
= use_vs(device
);
3010 struct ffp_settings settings
;
3011 struct arbfp_ffp_desc
*desc
;
3014 if(isStateDirty(context
, STATE_RENDER(WINED3DRS_FOGENABLE
))) {
3016 IWineD3DPixelShader_CompileShader(stateblock
->pixelShader
);
3017 } else if(device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
3018 /* Reload fixed function constants since they collide with the pixel shader constants */
3019 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
3020 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
3022 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
3023 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
3029 IWineD3DPixelShader_CompileShader(stateblock
->pixelShader
);
3031 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
3032 gen_ffp_op(stateblock
, &settings
, FALSE
);
3033 desc
= (struct arbfp_ffp_desc
*) find_ffp_shader(priv
->fragment_shaders
, &settings
);
3035 desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*desc
));
3037 ERR("Out of memory\n");
3040 desc
->num_textures_used
= 0;
3041 for(i
= 0; i
< GL_LIMITS(texture_stages
); i
++) {
3042 if(settings
.op
[i
].cop
== WINED3DTOP_DISABLE
) break;
3043 desc
->num_textures_used
= i
;
3046 memcpy(&desc
->parent
.settings
, &settings
, sizeof(settings
));
3047 desc
->shader
= gen_arbfp_ffp_shader(&settings
, stateblock
);
3048 add_ffp_shader(priv
->fragment_shaders
, &desc
->parent
);
3049 TRACE("Allocated fixed function replacement shader descriptor %p\n", desc
);
3052 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
3053 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
3056 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
3057 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
3059 if(device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
3060 /* Reload fixed function constants since they collide with the pixel shader constants */
3061 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
3062 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
3064 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
3065 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
3069 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
3070 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
3071 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
3072 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
3073 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
3075 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
3076 * to be compiled before activating them(needs some cleanups in the shader backend interface)
3078 if(!isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
3079 device
->shader_backend
->shader_select((IWineD3DDevice
*)stateblock
->wineD3DDevice
, use_pshader
, use_vshader
);
3081 if (!isStateDirty(context
, STATE_VERTEXSHADERCONSTANT
) && (use_vshader
|| use_pshader
)) {
3082 device
->StateTable
[STATE_VERTEXSHADERCONSTANT
].apply(STATE_VERTEXSHADERCONSTANT
, stateblock
, context
);
3086 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
3090 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
3091 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
3092 * state table, so we need to handle that with a forwarding function. The other invisible side effect
3093 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
3094 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
3096 static void state_arbfp_fog(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3097 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
3098 fragment_prog_arbfp(state
, stateblock
, context
);
3102 static void textransform(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3103 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
3104 fragment_prog_arbfp(state
, stateblock
, context
);
3108 #undef GLINFO_LOCATION
3110 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] = {
3111 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), state_texfactor_arbfp
}, 0 },
3112 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3113 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3114 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3115 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3116 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3117 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3118 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3119 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3120 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3121 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3122 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3123 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3124 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3125 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3126 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3127 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3128 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3129 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3130 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3131 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3132 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3133 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3134 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3135 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3136 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3137 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3138 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3139 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3140 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3141 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3142 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3143 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3144 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3145 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3146 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3147 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3148 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3149 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3150 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3151 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3152 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3153 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3154 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3155 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3156 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3157 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3158 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3159 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3160 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3161 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3162 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3163 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3164 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3165 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3166 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3167 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3168 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3169 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3170 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3171 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3172 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3173 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3174 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3175 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3176 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3177 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3178 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3179 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3180 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3181 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3182 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3183 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3184 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3185 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3186 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3187 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3188 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3189 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3190 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3191 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3192 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3193 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3194 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3195 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3196 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3197 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3198 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3199 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3200 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3201 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3202 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3203 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3204 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3205 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3206 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3207 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3208 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3209 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3210 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3211 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3212 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3213 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3214 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3215 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3216 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3217 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3218 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3219 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3220 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3221 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3222 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3223 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3224 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3225 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3226 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3227 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3228 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3229 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3230 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3231 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3232 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim
}, 0 },
3233 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim
}, 0 },
3234 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim
}, 0 },
3235 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim
}, 0 },
3236 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim
}, 0 },
3237 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim
}, 0 },
3238 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim
}, 0 },
3239 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim
}, 0 },
3240 { STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3241 { STATE_RENDER(WINED3DRS_FOGENABLE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3242 { STATE_RENDER(WINED3DRS_FOGTABLEMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3243 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3244 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3245 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3246 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3247 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3248 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3249 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3250 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3251 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3252 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3253 { STATE_RENDER(WINED3DRS_SPECULARENABLE
), { STATE_RENDER(WINED3DRS_SPECULARENABLE
), state_arb_specularenable
}, 0 },
3254 {0 /* Terminate */, { 0, 0 }, 0 },
3257 const struct fragment_pipeline arbfp_fragment_pipeline
= {
3262 shader_arb_conv_supported
,
3263 arbfp_fragmentstate_template
,
3264 TRUE
/* We can disable projected textures */
3267 #define GLINFO_LOCATION device->adapter->gl_info
3269 struct arbfp_blit_priv
{
3270 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
3271 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
3272 GLenum yv12_rect_shader
, yv12_2d_shader
;
3275 static HRESULT
arbfp_blit_alloc(IWineD3DDevice
*iface
) {
3276 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3277 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
3278 if(!device
->blit_priv
) {
3279 ERR("Out of memory\n");
3280 return E_OUTOFMEMORY
;
3284 static void arbfp_blit_free(IWineD3DDevice
*iface
) {
3285 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3286 struct arbfp_blit_priv
*priv
= (struct arbfp_blit_priv
*) device
->blit_priv
;
3289 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
3290 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
3291 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
3292 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
3293 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
3294 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
3295 checkGLcall("Delete yuv programs\n");
3299 static BOOL
gen_planar_yuv_read(SHADER_BUFFER
*buffer
, WINED3DFORMAT fmt
, GLenum textype
, char *luminance
) {
3301 const char *tex
, *texinstr
;
3303 if(fmt
== WINED3DFMT_UYVY
) {
3311 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
3312 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
3314 /* This is more tricky than just replacing the texture type - we have to navigate
3315 * properly in the texture to find the correct chroma values
3317 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3321 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3322 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3323 * filtering when we sample the texture.
3325 * These are the rules for reading the chroma:
3331 * So we have to get the sampling x position in non-normalized coordinates in integers
3333 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
3334 shader_addline(buffer
, "MUL texcrd.rg, fragment.texcoord[0], size.x;\n");
3335 shader_addline(buffer
, "MOV texcrd.a, size.x;\n");
3337 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3339 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3340 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3343 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
3344 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
3346 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3347 * even and odd pixels respectively
3349 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
3350 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
3352 /* Sample Pixel 1 */
3353 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
3355 /* Put the value into either of the chroma values */
3356 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
3357 shader_addline(buffer
, "MUL chroma.r, luminance.%c, temp.x;\n", chroma
);
3358 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
3359 shader_addline(buffer
, "MUL chroma.g, luminance.%c, temp.x;\n", chroma
);
3361 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3362 * the pixel right to the current one. Otherwise, sample the left pixel.
3363 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3365 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3366 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
3367 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
3369 /* Put the value into the other chroma */
3370 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
3371 shader_addline(buffer
, "MAD chroma.g, luminance.%c, temp.x, chroma.g;\n", chroma
);
3372 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
3373 shader_addline(buffer
, "MAD chroma.r, luminance.%c, temp.x, chroma.r;\n", chroma
);
3375 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3376 * the current one and lerp the two U and V values
3379 /* This gives the correctly filtered luminance value */
3380 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
3385 static BOOL
gen_yv12_read(SHADER_BUFFER
*buffer
, WINED3DFORMAT fmt
, GLenum textype
, char *luminance
) {
3389 case GL_TEXTURE_2D
: tex
= "2D"; break;
3390 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
3392 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3396 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3397 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3398 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3399 * pitch of the luminance plane, the packing into the gl texture is a bit
3400 * unfortunate. If the whole texture is interpreted as luminance data it looks
3401 * approximately like this:
3403 * +----------------------------------+----
3415 * +----------------+-----------------+----
3417 * | U even rows | U odd rows |
3419 * +----------------+------------------ -
3421 * | V even rows | V odd rows |
3423 * +----------------+-----------------+----
3427 * So it appears as if there are 4 chroma images, but in fact the odd rows
3428 * in the chroma images are in the same row as the even ones. So its is
3429 * kinda tricky to read
3431 * When reading from rectangle textures, keep in mind that the input y coordinates
3432 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3434 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3435 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3437 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3438 /* the chroma planes have only half the width */
3439 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
3441 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3442 * the coordinate. Also read the right side of the image when reading odd lines
3444 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3447 if(textype
== GL_TEXTURE_2D
) {
3449 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
3451 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
3453 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
3454 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3456 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3457 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3458 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
3459 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3460 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3462 /* clamp, keep the half pixel origin in mind */
3463 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3464 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
3465 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3466 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3468 /* Read from [size - size+size/4] */
3469 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
3470 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3472 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3473 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3474 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
3475 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3476 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3477 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3479 /* Make sure to read exactly from the pixel center */
3480 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
3481 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
3484 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
3485 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
3486 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3487 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
3488 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
3490 /* Read the texture, put the result into the output register */
3491 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
3492 shader_addline(buffer
, "MOV chroma.r, temp.a;\n");
3494 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3495 * No need to clamp because we're just reusing the already clamped value from above
3497 if(textype
== GL_TEXTURE_2D
) {
3498 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3500 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3502 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
3503 shader_addline(buffer
, "MOV chroma.g, temp.a;\n");
3505 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3506 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3507 * values due to filtering
3509 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3510 if(textype
== GL_TEXTURE_2D
) {
3511 /* Multiply the y coordinate by 2/3 and clamp it */
3512 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3513 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3514 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3515 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
3517 /* Reading from texture_rectangles is pretty streightforward, just use the unmodified
3518 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3521 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
3522 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
3523 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
3530 static GLuint
gen_yuv_shader(IWineD3DDeviceImpl
*device
, WINED3DFORMAT fmt
, GLenum textype
) {
3532 SHADER_BUFFER buffer
;
3533 char luminance_component
;
3534 struct arbfp_blit_priv
*priv
= (struct arbfp_blit_priv
*) device
->blit_priv
;
3539 buffer
.newline
= TRUE
;
3540 buffer
.buffer
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, SHADER_PGMSIZE
);
3542 GL_EXTCALL(glGenProgramsARB(1, &shader
));
3543 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3544 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
3545 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3547 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3551 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3552 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3553 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3554 * each single pixel it contains, and one U and one V value shared between both
3557 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3558 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3559 * take the format into account when generating the read swizzles
3561 * Reading the Y value is streightforward - just sample the texture. The hardware
3562 * takes care of filtering in the horizontal and vertical direction.
3564 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3565 * because that would mix the U and V values of one pixel or two adjacent pixels.
3566 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3567 * regardless of the filtering setting. Vertical filtering works automatically
3568 * though - the U and V values of two rows are mixed nicely.
3570 * Appart of avoiding filtering issues, the code has to know which value it just
3571 * read, and where it can find the other one. To determine this, it checks if
3572 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3574 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3575 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3577 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3578 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3579 * in an unfiltered situation. Finding the luminance on the other hand requires
3580 * finding out if it is an odd or even pixel. The real drawback of this approach
3581 * is filtering. This would have to be emulated completely in the shader, reading
3582 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3583 * vertically. Beyond that it would require adjustments to the texture handling
3584 * code to deal with the width scaling
3586 shader_addline(&buffer
, "!!ARBfp1.0\n");
3587 shader_addline(&buffer
, "TEMP luminance;\n");
3588 shader_addline(&buffer
, "TEMP temp;\n");
3589 shader_addline(&buffer
, "TEMP chroma;\n");
3590 shader_addline(&buffer
, "TEMP texcrd;\n");
3591 shader_addline(&buffer
, "TEMP texcrd2;\n");
3592 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3593 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3594 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
3596 if(fmt
== WINED3DFMT_UYVY
|| fmt
==WINED3DFMT_YUY2
) {
3597 if(gen_planar_yuv_read(&buffer
, fmt
, textype
, &luminance_component
) == FALSE
) {
3598 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3602 if(gen_yv12_read(&buffer
, fmt
, textype
, &luminance_component
) == FALSE
) {
3603 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3608 /* Calculate the final result. Formula is taken from
3609 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3610 * ranges from -0.5 to 0.5
3612 shader_addline(&buffer
, "SUB chroma.rg, chroma, coef.y;\n");
3614 shader_addline(&buffer
, "MAD result.color.r, chroma.r, yuv_coef.x, luminance.%c;\n", luminance_component
);
3615 shader_addline(&buffer
, "MAD temp.r, -chroma.g, yuv_coef.y, luminance.%c;\n", luminance_component
);
3616 shader_addline(&buffer
, "MAD result.color.g, -chroma.r, yuv_coef.z, temp.r;\n");
3617 shader_addline(&buffer
, "MAD result.color.b, chroma.g, yuv_coef.w, luminance.%c;\n", luminance_component
);
3618 shader_addline(&buffer
, "END\n");
3620 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
3622 if (glGetError() == GL_INVALID_OPERATION
) {
3624 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3625 FIXME("Fragment program error at position %d: %s\n", pos
,
3626 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3628 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3630 if(fmt
== WINED3DFMT_YUY2
) {
3631 if(textype
== GL_TEXTURE_RECTANGLE_ARB
) {
3632 priv
->yuy2_rect_shader
= shader
;
3634 priv
->yuy2_2d_shader
= shader
;
3636 } else if(fmt
== WINED3DFMT_UYVY
) {
3637 if(textype
== GL_TEXTURE_RECTANGLE_ARB
) {
3638 priv
->uyvy_rect_shader
= shader
;
3640 priv
->uyvy_2d_shader
= shader
;
3643 if(textype
== GL_TEXTURE_RECTANGLE_ARB
) {
3644 priv
->yv12_rect_shader
= shader
;
3646 priv
->yv12_2d_shader
= shader
;
3652 static HRESULT
arbfp_blit_set(IWineD3DDevice
*iface
, WINED3DFORMAT fmt
, GLenum textype
, UINT width
, UINT height
) {
3654 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3655 float size
[4] = {width
, height
, 1, 1};
3656 struct arbfp_blit_priv
*priv
= (struct arbfp_blit_priv
*) device
->blit_priv
;
3657 const GlPixelFormatDesc
*glDesc
;
3659 getFormatDescEntry(fmt
, &GLINFO_LOCATION
, &glDesc
);
3661 if(glDesc
->conversion_group
!= WINED3DFMT_YUY2
&& glDesc
->conversion_group
!= WINED3DFMT_UYVY
&&
3662 glDesc
->conversion_group
!= WINED3DFMT_YV12
) {
3663 TRACE("Format: %s\n", debug_d3dformat(glDesc
->conversion_group
));
3664 /* Don't bother setting up a shader for unconverted formats */
3666 checkGLcall("glEnable(textype)");
3670 if(glDesc
->conversion_group
== WINED3DFMT_YUY2
) {
3671 if(textype
== GL_TEXTURE_RECTANGLE_ARB
) {
3672 shader
= priv
->yuy2_rect_shader
;
3674 shader
= priv
->yuy2_2d_shader
;
3676 } else if(glDesc
->conversion_group
== WINED3DFMT_UYVY
) {
3677 if(textype
== GL_TEXTURE_RECTANGLE_ARB
) {
3678 shader
= priv
->uyvy_rect_shader
;
3680 shader
= priv
->uyvy_2d_shader
;
3683 if(textype
== GL_TEXTURE_RECTANGLE_ARB
) {
3684 shader
= priv
->yv12_rect_shader
;
3686 shader
= priv
->yv12_2d_shader
;
3691 shader
= gen_yuv_shader(device
, glDesc
->conversion_group
, textype
);
3694 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
3695 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3696 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
3697 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3698 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
3699 checkGLcall("glProgramLocalParameter4fvARB");
3704 static void arbfp_blit_unset(IWineD3DDevice
*iface
) {
3705 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3706 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
3707 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3708 glDisable(GL_TEXTURE_2D
);
3709 checkGLcall("glDisable(GL_TEXTURE_2D)");
3710 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP
)) {
3711 glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
3712 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3714 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3715 glDisable(GL_TEXTURE_RECTANGLE_ARB
);
3716 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3720 static BOOL
arbfp_blit_conv_supported(WINED3DFORMAT fmt
) {
3721 TRACE("Checking blit format support for format %s:", debug_d3dformat(fmt
));
3723 case WINED3DFMT_YUY2
:
3724 case WINED3DFMT_UYVY
:
3725 case WINED3DFMT_YV12
:
3729 TRACE("[FAILED]\n");
3734 const struct blit_shader arbfp_blit
= {
3739 arbfp_blit_conv_supported
3742 #undef GLINFO_LOCATION