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
13 * Copyright 2009 Henri Verbeet for CodeWeavers
15 * This library is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU Lesser General Public
17 * License as published by the Free Software Foundation; either
18 * version 2.1 of the License, or (at your option) any later version.
20 * This library is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
23 * Lesser General Public License for more details.
25 * You should have received a copy of the GNU Lesser General Public
26 * License along with this library; if not, write to the Free Software
27 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 #define GLINFO_LOCATION (*gl_info)
44 /* We have to subtract any other PARAMs that we might use in our shader programs.
45 * ATI seems to count 2 implicit PARAMs when we use fog and NVIDIA counts 1,
46 * and we reference one row of the PROJECTION matrix which counts as 1 PARAM. */
47 #define ARB_SHADER_RESERVED_VS_CONSTS 3
49 /* The arb shader only loads the bump mapping environment matrix into the shader if it finds
50 * a free constant to do that, so only reduce the number of available constants by 2 for the fog states.
52 #define ARB_SHADER_RESERVED_PS_CONSTS 2
54 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
55 * so upload them above that
57 #define ARB_SHADER_PRIVCONST_BASE (GL_LIMITS(vshader_constantsF) - ARB_SHADER_RESERVED_VS_CONSTS)
58 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
60 /* ARB_program_shader private data */
61 struct shader_arb_priv
{
62 GLuint current_vprogram_id
;
63 GLuint current_fprogram_id
;
64 GLuint depth_blt_vprogram_id
;
65 GLuint depth_blt_fprogram_id
[tex_type_count
];
66 BOOL use_arbfp_fixed_func
;
67 struct hash_table_t
*fragment_shaders
;
70 /********************************************************
71 * ARB_[vertex/fragment]_program helper functions follow
72 ********************************************************/
75 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
76 * When constant_list == NULL, it will load all the constants.
78 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
79 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
81 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl
* This
, const WineD3D_GL_Info
*gl_info
,
82 GLuint target_type
, unsigned int max_constants
, const float *constants
, char *dirty_consts
)
84 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
85 local_constant
* lconst
;
89 if (TRACE_ON(d3d_shader
)) {
90 for(i
= 0; i
< max_constants
; i
++) {
91 if(!dirty_consts
[i
]) continue;
92 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
93 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
94 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
97 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
98 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& WINED3DSHADER_VERSION_MAJOR(shader_version
) == 1)
101 for(i
= 0; i
< max_constants
; i
++) {
102 if(!dirty_consts
[i
]) continue;
106 if(constants
[j
+ 0] > 1.0) lcl_const
[0] = 1.0;
107 else if(constants
[j
+ 0] < -1.0) lcl_const
[0] = -1.0;
108 else lcl_const
[0] = constants
[j
+ 0];
110 if(constants
[j
+ 1] > 1.0) lcl_const
[1] = 1.0;
111 else if(constants
[j
+ 1] < -1.0) lcl_const
[1] = -1.0;
112 else lcl_const
[1] = constants
[j
+ 1];
114 if(constants
[j
+ 2] > 1.0) lcl_const
[2] = 1.0;
115 else if(constants
[j
+ 2] < -1.0) lcl_const
[2] = -1.0;
116 else lcl_const
[2] = constants
[j
+ 2];
118 if(constants
[j
+ 3] > 1.0) lcl_const
[3] = 1.0;
119 else if(constants
[j
+ 3] < -1.0) lcl_const
[3] = -1.0;
120 else lcl_const
[3] = constants
[j
+ 3];
122 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
125 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS
)) {
126 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
127 * or just reloading *all* constants at once
129 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
131 for(i
= 0; i
< max_constants
; i
++) {
132 if(!dirty_consts
[i
]) continue;
134 /* Find the next block of dirty constants */
137 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
141 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
144 for(i
= 0; i
< max_constants
; i
++) {
145 if(dirty_consts
[i
]) {
147 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
152 checkGLcall("glProgramEnvParameter4fvARB()");
154 /* Load immediate constants */
155 if(This
->baseShader
.load_local_constsF
) {
156 if (TRACE_ON(d3d_shader
)) {
157 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
158 GLfloat
* values
= (GLfloat
*)lconst
->value
;
159 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
160 values
[0], values
[1], values
[2], values
[3]);
163 /* Immediate constants are clamped for 1.X shaders at loading times */
165 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
166 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
167 ret
= max(ret
, lconst
->idx
+ 1);
168 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
170 checkGLcall("glProgramEnvParameter4fvARB()");
171 return ret
; /* The loaded immediate constants need reloading for the next shader */
173 return 0; /* No constants are dirty now */
178 * Loads the texture dimensions for NP2 fixup into the currently set ARB_[vertex/fragment]_programs.
180 static void shader_arb_load_np2fixup_constants(
181 IWineD3DDevice
* device
,
183 char useVertexShader
) {
184 /* not implemented */
188 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
190 * We only support float constants in ARB at the moment, so don't
191 * worry about the Integers or Booleans
193 static void shader_arb_load_constants(
194 IWineD3DDevice
* device
,
196 char useVertexShader
) {
198 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
199 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
200 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
203 if (useVertexShader
) {
204 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
206 /* Load DirectX 9 float constants for vertex shader */
207 deviceImpl
->highest_dirty_vs_const
= shader_arb_load_constantsF(
208 vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
209 deviceImpl
->highest_dirty_vs_const
,
210 stateBlock
->vertexShaderConstantF
,
211 deviceImpl
->activeContext
->vshader_const_dirty
);
213 /* Upload the position fixup */
214 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, ARB_SHADER_PRIVCONST_POS
, deviceImpl
->posFixup
));
217 if (usePixelShader
) {
219 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
220 IWineD3DPixelShaderImpl
*psi
= (IWineD3DPixelShaderImpl
*) pshader
;
222 /* Load DirectX 9 float constants for pixel shader */
223 deviceImpl
->highest_dirty_ps_const
= shader_arb_load_constantsF(
224 pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
225 deviceImpl
->highest_dirty_ps_const
,
226 stateBlock
->pixelShaderConstantF
,
227 deviceImpl
->activeContext
->pshader_const_dirty
);
229 for(i
= 0; i
< psi
->numbumpenvmatconsts
; i
++) {
230 /* The state manager takes care that this function is always called if the bump env matrix changes
232 const float *data
= (const float *)&stateBlock
->textureState
[(int) psi
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
233 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->bumpenvmatconst
[i
].const_num
, data
));
234 deviceImpl
->activeContext
->pshader_const_dirty
[psi
->bumpenvmatconst
[i
].const_num
] = 1;
236 if (psi
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
238 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
239 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
240 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
241 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
243 const float *scale
= (const float *)&stateBlock
->textureState
[(int) psi
->luminanceconst
[i
].texunit
][WINED3DTSS_BUMPENVLSCALE
];
244 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->luminanceconst
[i
].const_num
, scale
));
245 deviceImpl
->activeContext
->pshader_const_dirty
[psi
->luminanceconst
[i
].const_num
] = 1;
251 static void shader_arb_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
253 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
255 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
256 * context. On a context switch the old context will be fully dirtified */
257 memset(This
->activeContext
->vshader_const_dirty
+ start
, 1,
258 sizeof(*This
->activeContext
->vshader_const_dirty
) * count
);
259 This
->highest_dirty_vs_const
= max(This
->highest_dirty_vs_const
, start
+ count
+ 1);
262 static void shader_arb_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
264 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
266 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
267 * context. On a context switch the old context will be fully dirtified */
268 memset(This
->activeContext
->pshader_const_dirty
+ start
, 1,
269 sizeof(*This
->activeContext
->pshader_const_dirty
) * count
);
270 This
->highest_dirty_ps_const
= max(This
->highest_dirty_ps_const
, start
+ count
+ 1);
273 /* Generate the variable & register declarations for the ARB_vertex_program output target */
274 static void shader_generate_arb_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
275 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
)
277 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
278 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
280 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
);
281 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
282 (pshader
? GL_LIMITS(pshader_constantsF
) - ARB_SHADER_RESERVED_PS_CONSTS
:
283 GL_LIMITS(vshader_constantsF
) - ARB_SHADER_RESERVED_VS_CONSTS
));
284 UINT extra_constants_needed
= 0;
285 const local_constant
*lconst
;
287 /* Temporary Output register */
288 shader_addline(buffer
, "TEMP TMP_OUT;\n");
290 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
291 if (reg_maps
->temporary
[i
])
292 shader_addline(buffer
, "TEMP R%u;\n", i
);
295 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
296 if (reg_maps
->address
[i
])
297 shader_addline(buffer
, "ADDRESS A%d;\n", i
);
300 for(i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
301 if (reg_maps
->texcoord
[i
])
302 shader_addline(buffer
,"TEMP T%u;\n", i
);
305 /* Texture coordinate registers must be pre-loaded */
306 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
307 if (reg_maps
->texcoord
[i
])
308 shader_addline(buffer
, "MOV T%u, fragment.texcoord[%u];\n", i
, i
);
311 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
312 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) This
;
313 if(!reg_maps
->bumpmat
[i
]) continue;
315 cur
= ps
->numbumpenvmatconsts
;
316 ps
->bumpenvmatconst
[cur
].const_num
= -1;
317 ps
->bumpenvmatconst
[cur
].texunit
= i
;
318 ps
->luminanceconst
[cur
].const_num
= -1;
319 ps
->luminanceconst
[cur
].texunit
= i
;
321 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
322 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
325 if(max_constantsF
+ extra_constants_needed
< GL_LIMITS(pshader_constantsF
) - ARB_SHADER_RESERVED_PS_CONSTS
) {
326 ps
->bumpenvmatconst
[cur
].const_num
= max_constantsF
+ extra_constants_needed
;
327 shader_addline(buffer
, "PARAM bumpenvmat%d = program.env[%d];\n",
328 i
, ps
->bumpenvmatconst
[cur
].const_num
);
329 extra_constants_needed
++;
331 if(reg_maps
->luminanceparams
&& max_constantsF
+ extra_constants_needed
< GL_LIMITS(pshader_constantsF
) - ARB_SHADER_RESERVED_PS_CONSTS
) {
332 ((IWineD3DPixelShaderImpl
*)This
)->luminanceconst
[cur
].const_num
= max_constantsF
+ extra_constants_needed
;
333 shader_addline(buffer
, "PARAM luminance%d = program.env[%d];\n",
334 i
, ps
->luminanceconst
[cur
].const_num
);
335 extra_constants_needed
++;
336 } else if(reg_maps
->luminanceparams
) {
337 FIXME("No free constant to load the luminance parameters\n");
340 FIXME("No free constant found to load environment bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
343 ps
->numbumpenvmatconsts
= cur
+ 1;
346 if(device
->stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
] && pshader
) {
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 shader_addline(buffer
, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
352 srgb_pow
, srgb_pow
, srgb_pow
);
353 shader_addline(buffer
, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
354 srgb_mul_high
, srgb_mul_high
, srgb_mul_high
);
355 shader_addline(buffer
, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
356 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
359 /* Load local constants using the program-local space,
360 * this avoids reloading them each time the shader is used
362 if(!This
->baseShader
.load_local_constsF
) {
363 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
364 shader_addline(buffer
, "PARAM C%u = program.local[%u];\n", lconst
->idx
,
369 /* we use the array-based constants array if the local constants are marked for loading,
370 * because then we use indirect addressing, or when the local constant list is empty,
371 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
372 * local constants do not declare the loaded constants as an array because ARB compilers usually
373 * do not optimize unused constants away
375 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) {
376 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
377 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
378 max_constantsF
, max_constantsF
- 1);
380 for(i
= 0; i
< max_constantsF
; i
++) {
381 if(!shader_constant_is_local(This
, i
)) {
382 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
388 static const char * const shift_tab
[] = {
389 "dummy", /* 0 (none) */
390 "coefmul.x", /* 1 (x2) */
391 "coefmul.y", /* 2 (x4) */
392 "coefmul.z", /* 3 (x8) */
393 "coefmul.w", /* 4 (x16) */
394 "dummy", /* 5 (x32) */
395 "dummy", /* 6 (x64) */
396 "dummy", /* 7 (x128) */
397 "dummy", /* 8 (d256) */
398 "dummy", /* 9 (d128) */
399 "dummy", /* 10 (d64) */
400 "dummy", /* 11 (d32) */
401 "coefdiv.w", /* 12 (d16) */
402 "coefdiv.z", /* 13 (d8) */
403 "coefdiv.y", /* 14 (d4) */
404 "coefdiv.x" /* 15 (d2) */
407 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
408 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
410 char *ptr
= write_mask
;
411 char vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
);
413 if (vshader
&& dst
->register_type
== WINED3DSPR_ADDR
)
418 else if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
421 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
422 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
423 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
424 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
430 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
432 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
433 * but addressed as "rgba". To fix this we need to swap the register's x
434 * and z components. */
435 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
436 char *ptr
= swizzle_str
;
438 /* swizzle bits fields: wwzzyyxx */
439 DWORD swizzle
= param
->swizzle
;
440 DWORD swizzle_x
= swizzle
& 0x03;
441 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
442 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
443 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
445 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
446 * generate a swizzle string. Unless we need to our own swizzling. */
447 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
450 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
451 *ptr
++ = swizzle_chars
[swizzle_x
];
453 *ptr
++ = swizzle_chars
[swizzle_x
];
454 *ptr
++ = swizzle_chars
[swizzle_y
];
455 *ptr
++ = swizzle_chars
[swizzle_z
];
456 *ptr
++ = swizzle_chars
[swizzle_w
];
463 static void shader_arb_get_register_name(IWineD3DBaseShader
*iface
, WINED3DSHADER_PARAM_REGISTER_TYPE register_type
,
464 UINT register_idx
, BOOL rel_addr
, char *register_name
, BOOL
*is_color
)
466 /* oPos, oFog and oPts in D3D */
467 static const char * const rastout_reg_names
[] = {"TMP_OUT", "result.fogcoord", "result.pointsize"};
468 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)iface
;
469 DWORD shader_version
= This
->baseShader
.reg_maps
.shader_version
;
470 BOOL pshader
= shader_is_pshader_version(shader_version
);
474 switch (register_type
)
476 case WINED3DSPR_TEMP
:
477 sprintf(register_name
, "R%u", register_idx
);
480 case WINED3DSPR_INPUT
:
483 if (register_idx
== 0) strcpy(register_name
, "fragment.color.primary");
484 else strcpy(register_name
, "fragment.color.secondary");
488 if (((IWineD3DVertexShaderImpl
*)This
)->cur_args
->swizzle_map
& (1 << register_idx
)) *is_color
= TRUE
;
489 sprintf(register_name
, "vertex.attrib[%u]", register_idx
);
493 case WINED3DSPR_CONST
:
494 if (!pshader
&& rel_addr
)
496 UINT rel_offset
= ((IWineD3DVertexShaderImpl
*)This
)->rel_offset
;
497 if (register_idx
>= rel_offset
)
498 sprintf(register_name
, "C[A0.x + %u]", register_idx
- rel_offset
);
500 sprintf(register_name
, "C[A0.x - %u]", -register_idx
+ rel_offset
);
504 if (This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
))
505 sprintf(register_name
, "C[%u]", register_idx
);
507 sprintf(register_name
, "C%u", register_idx
);
511 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
512 if (pshader
) sprintf(register_name
, "T%u", register_idx
);
513 else sprintf(register_name
, "A%u", register_idx
);
516 case WINED3DSPR_COLOROUT
:
517 if (register_idx
== 0)
519 strcpy(register_name
, "TMP_COLOR");
523 /* TODO: See GL_ARB_draw_buffers */
524 FIXME("Unsupported write to render target %u\n", register_idx
);
525 sprintf(register_name
, "unsupported_register");
529 case WINED3DSPR_RASTOUT
:
530 sprintf(register_name
, "%s", rastout_reg_names
[register_idx
]);
533 case WINED3DSPR_DEPTHOUT
:
534 strcpy(register_name
, "result.depth");
537 case WINED3DSPR_ATTROUT
:
538 if (pshader
) sprintf(register_name
, "oD[%u]", register_idx
);
539 else if (register_idx
== 0) strcpy(register_name
, "result.color.primary");
540 else strcpy(register_name
, "result.color.secondary");
543 case WINED3DSPR_TEXCRDOUT
:
544 if (pshader
) sprintf(register_name
, "oT[%u]", register_idx
);
545 else sprintf(register_name
, "result.texcoord[%u]", register_idx
);
549 FIXME("Unhandled register type %#x[%u]\n", register_type
, register_idx
);
550 sprintf(register_name
, "unrecognized_register[%u]", register_idx
);
555 static void shader_arb_add_src_param(const struct wined3d_shader_instruction
*ins
,
556 const struct wined3d_shader_src_param
*wined3d_src
, char *str
)
558 char register_name
[255];
562 if (wined3d_src
->modifiers
== WINED3DSPSM_NEG
) strcat(str
, " -");
563 else strcat(str
, " ");
565 shader_arb_get_register_name(ins
->ctx
->shader
, wined3d_src
->register_type
,
566 wined3d_src
->register_idx
, !!wined3d_src
->rel_addr
, register_name
, &is_color
);
567 strcat(str
, register_name
);
569 shader_arb_get_swizzle(wined3d_src
, is_color
, swizzle
);
570 strcat(str
, swizzle
);
573 static void shader_arb_add_dst_param(const struct wined3d_shader_instruction
*ins
,
574 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
576 char register_name
[255];
582 shader_arb_get_register_name(ins
->ctx
->shader
, wined3d_dst
->register_type
,
583 wined3d_dst
->register_idx
, !!wined3d_dst
->rel_addr
, register_name
, &is_color
);
584 strcat(str
, register_name
);
586 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
587 strcat(str
, write_mask
);
590 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
592 switch(channel_source
)
594 case CHANNEL_SOURCE_ZERO
: return "0";
595 case CHANNEL_SOURCE_ONE
: return "1";
596 case CHANNEL_SOURCE_X
: return "x";
597 case CHANNEL_SOURCE_Y
: return "y";
598 case CHANNEL_SOURCE_Z
: return "z";
599 case CHANNEL_SOURCE_W
: return "w";
601 FIXME("Unhandled channel source %#x\n", channel_source
);
606 static void gen_color_correction(SHADER_BUFFER
*buffer
, const char *reg
, DWORD dst_mask
,
607 const char *one
, const char *two
, struct color_fixup_desc fixup
)
611 if (is_yuv_fixup(fixup
))
613 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
614 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
619 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
620 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
621 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
622 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
627 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
628 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
629 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
633 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
634 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
635 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
636 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
642 char *ptr
= reg_mask
;
644 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
647 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
648 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
649 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
650 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
654 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
658 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
659 const char *dst_str
, const char *coord_reg
, BOOL projected
, BOOL bias
)
661 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
662 DWORD sampler_type
= ins
->ctx
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
663 const char *tex_type
;
664 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
665 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
667 switch(sampler_type
) {
673 if(device
->stateBlock
->textures
[sampler_idx
] &&
674 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
679 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
681 const IWineD3DPixelShaderImpl
* const ps
= (const IWineD3DPixelShaderImpl
*)This
;
682 if(ps
->cur_args
->np2_fixup
& (1 << sampler_idx
)) {
683 FIXME("NP2 texcoord fixup is currently not implemented in ARB mode (use GLSL instead).\n");
688 case WINED3DSTT_VOLUME
:
692 case WINED3DSTT_CUBE
:
697 ERR("Unexpected texture type %d\n", sampler_type
);
702 /* Shouldn't be possible, but let's check for it */
703 if(projected
) FIXME("Biased and Projected texture sampling\n");
704 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
705 shader_addline(buffer
, "TXB %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
706 } else if (projected
) {
707 shader_addline(buffer
, "TXP %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
709 shader_addline(buffer
, "TEX %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
712 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
714 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
715 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
716 "one", "coefmul.x", ps
->cur_args
->color_fixup
[sampler_idx
]);
720 static void pshader_gen_input_modifier_line(IWineD3DBaseShader
*iface
, SHADER_BUFFER
*buffer
,
721 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
723 /* Generate a line that does the input modifier computation and return the input register to use */
724 BOOL is_color
= FALSE
;
729 /* Assume a new line will be added */
732 /* Get register name */
733 shader_arb_get_register_name(iface
, src
->register_type
,
734 src
->register_idx
, !!src
->rel_addr
, regstr
, &is_color
);
735 shader_arb_get_swizzle(src
, is_color
, swzstr
);
737 switch (src
->modifiers
)
739 case WINED3DSPSM_NONE
:
740 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
743 case WINED3DSPSM_NEG
:
744 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
747 case WINED3DSPSM_BIAS
:
748 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
750 case WINED3DSPSM_BIASNEG
:
751 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
753 case WINED3DSPSM_SIGN
:
754 shader_addline(buffer
, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg
, regstr
);
756 case WINED3DSPSM_SIGNNEG
:
757 shader_addline(buffer
, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg
, regstr
);
759 case WINED3DSPSM_COMP
:
760 shader_addline(buffer
, "SUB T%c, one.x, %s;\n", 'A' + tmpreg
, regstr
);
763 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
765 case WINED3DSPSM_X2NEG
:
766 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
769 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
770 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
773 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
774 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
777 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
781 /* Return modified or original register, with swizzle */
783 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
786 static inline void pshader_gen_output_modifier_line(SHADER_BUFFER
*buffer
, int saturate
, const char *write_mask
,
787 int shift
, const char *regstr
)
789 /* Generate a line that does the output modifier computation */
790 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", saturate
? "_SAT" : "",
791 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
794 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
796 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
797 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
798 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
800 char src_name
[2][50];
802 DWORD sampler_code
= dst
->register_idx
;
803 BOOL has_bumpmat
= FALSE
;
807 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
808 if (This
->bumpenvmatconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
809 && This
->bumpenvmatconst
[i
].texunit
== sampler_code
)
816 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
817 dst
->register_idx
, !!dst
->rel_addr
, dst_name
, &is_color
);
818 shader_arb_get_write_mask(ins
, dst
, dst_wmask
);
819 strcat(dst_name
, dst_wmask
);
821 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_name
[0]);
822 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[1], 1, src_name
[1]);
825 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
826 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
827 shader_addline(buffer
, "DP3 TMP.r, TMP2, %s;\n", src_name
[1]);
828 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
829 shader_addline(buffer
, "DP3 TMP.g, TMP2, %s;\n", src_name
[1]);
831 shader_addline(buffer
, "ADD %s, %s, TMP;\n", dst_name
, src_name
[0]);
833 shader_addline(buffer
, "MOV %s, %s;\n", dst_name
, src_name
[0]);
837 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
839 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
840 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
843 char src_name
[3][50];
844 BOOL sat
= dst
->modifiers
& WINED3DSPDM_SATURATE
;
845 DWORD shift
= dst
->shift
;
848 /* FIXME: support output modifiers */
850 /* Handle output register */
851 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
852 dst
->register_idx
, !!dst
->rel_addr
, dst_name
, &is_color
);
853 shader_arb_get_write_mask(ins
, dst
, dst_wmask
);
855 /* Generate input register names (with modifiers) */
856 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_name
[0]);
857 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[1], 1, src_name
[1]);
858 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[2], 2, src_name
[2]);
860 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
861 if (ins
->ctx
->reg_maps
->shader_version
<= WINED3DPS_VERSION(1, 3) && ins
->coissue
)
863 shader_addline(buffer
, "MOV%s %s%s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[1]);
865 shader_addline(buffer
, "ADD TMP, -%s, coefdiv.x;\n", src_name
[0]);
866 shader_addline(buffer
, "CMP%s %s%s, TMP, %s, %s;\n",
867 sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[1], src_name
[2]);
870 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
873 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
875 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
876 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
879 char src_name
[3][50];
880 DWORD shift
= dst
->shift
;
881 BOOL sat
= dst
->modifiers
& WINED3DSPDM_SATURATE
;
884 /* FIXME: support output modifiers */
886 /* Handle output register */
887 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
888 dst
->register_idx
, !!dst
->rel_addr
, dst_name
, &is_color
);
889 shader_arb_get_write_mask(ins
, dst
, dst_wmask
);
891 /* Generate input register names (with modifiers) */
892 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_name
[0]);
893 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[1], 1, src_name
[1]);
894 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[2], 2, src_name
[2]);
896 shader_addline(buffer
, "CMP%s %s%s, %s, %s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
897 src_name
[0], src_name
[2], src_name
[1]);
900 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
903 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
904 * dst = dot2(src0, src1) + src2 */
905 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
907 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
908 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
911 char src_name
[3][50];
912 DWORD shift
= dst
->shift
;
913 BOOL sat
= dst
->modifiers
& WINED3DSPDM_SATURATE
;
916 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
917 dst
->register_idx
, !!dst
->rel_addr
, dst_name
, &is_color
);
918 shader_arb_get_write_mask(ins
, dst
, dst_wmask
);
920 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_name
[0]);
921 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[1], 1, src_name
[1]);
922 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[2], 2, src_name
[2]);
924 /* Emulate a DP2 with a DP3 and 0.0 */
925 shader_addline(buffer
, "MOV TMP, %s;\n", src_name
[0]);
926 shader_addline(buffer
, "MOV TMP.z, 0.0;\n");
927 shader_addline(buffer
, "DP3 TMP2, TMP, %s;\n", src_name
[1]);
928 shader_addline(buffer
, "ADD%s %s%s, TMP2, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[2]);
931 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
934 /* Map the opcode 1-to-1 to the GL code */
935 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
937 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
938 const char *instruction
;
942 switch (ins
->handler_idx
)
944 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
945 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
946 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
947 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
948 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
949 case WINED3DSIH_DST
: instruction
= "DST"; break;
950 case WINED3DSIH_EXP
: instruction
= "EX2"; break;
951 case WINED3DSIH_EXPP
: instruction
= "EXP"; break;
952 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
953 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
954 case WINED3DSIH_LOG
: instruction
= "LG2"; break;
955 case WINED3DSIH_LOGP
: instruction
= "LOG"; break;
956 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
957 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
958 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
959 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
960 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
961 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
962 case WINED3DSIH_NOP
: instruction
= "NOP"; break;
963 case WINED3DSIH_POW
: instruction
= "POW"; break;
964 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
965 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
966 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
967 default: instruction
= "";
968 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
972 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
))
974 /* Output token related */
975 const struct wined3d_shader_dst_param
*dst
;
976 char output_rname
[256];
977 char output_wmask
[20];
978 char operands
[4][100];
979 BOOL saturate
= FALSE
;
980 BOOL centroid
= FALSE
;
981 BOOL partialprecision
= FALSE
;
982 const char *modifier
;
986 if (!(ins
->dst_count
+ ins
->src_count
))
988 ERR("Opcode \"%#x\" has no parameters\n", ins
->handler_idx
);
993 /* Process modifiers */
996 DWORD mask
= dst
->modifiers
;
998 saturate
= mask
& WINED3DSPDM_SATURATE
;
999 centroid
= mask
& WINED3DSPDM_MSAMPCENTROID
;
1000 partialprecision
= mask
& WINED3DSPDM_PARTIALPRECISION
;
1001 mask
&= ~(WINED3DSPDM_MSAMPCENTROID
| WINED3DSPDM_PARTIALPRECISION
| WINED3DSPDM_SATURATE
);
1003 FIXME("Unrecognized modifier(%#x)\n", mask
);
1006 FIXME("Unhandled modifier(%#x)\n", mask
);
1009 modifier
= (saturate
&& !shift
) ? "_SAT" : "";
1011 /* Generate input register names (with modifiers) */
1012 for (i
= 0; i
< ins
->src_count
; ++i
)
1014 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[i
], i
, operands
[i
+ 1]);
1017 /* Handle output register */
1018 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1019 dst
->register_idx
, !!dst
->rel_addr
, output_rname
, &is_color
);
1020 strcpy(operands
[0], output_rname
);
1021 shader_arb_get_write_mask(ins
, dst
, output_wmask
);
1022 strcat(operands
[0], output_wmask
);
1024 arguments
[0] = '\0';
1025 strcat(arguments
, operands
[0]);
1026 for (i
= 0; i
< ins
->src_count
; ++i
)
1028 strcat(arguments
, ", ");
1029 strcat(arguments
, operands
[i
+ 1]);
1031 shader_addline(buffer
, "%s%s %s;\n", instruction
, modifier
, arguments
);
1033 /* A shift requires another line. */
1034 if (shift
) pshader_gen_output_modifier_line(buffer
, saturate
, output_wmask
, shift
, output_rname
);
1036 /* Note that shader_arb_add_*_param() adds spaces. */
1038 arguments
[0] = '\0';
1041 shader_arb_add_dst_param(ins
, &ins
->dst
[0], arguments
);
1042 for (i
= 0; i
< ins
->src_count
; ++i
)
1044 strcat(arguments
, ",");
1045 shader_arb_add_src_param(ins
, &ins
->src
[i
], arguments
);
1048 shader_addline(buffer
, "%s%s;\n", instruction
, arguments
);
1052 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1054 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1056 if ((WINED3DSHADER_VERSION_MAJOR(ins
->ctx
->reg_maps
->shader_version
) == 1
1057 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
)
1058 && ins
->dst
[0].register_type
== WINED3DSPR_ADDR
)
1059 || ins
->handler_idx
== WINED3DSIH_MOVA
)
1061 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1062 char src0_param
[256];
1064 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1065 FIXME("mova should round\n");
1067 src0_param
[0] = '\0';
1068 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
)
1070 shader_arb_add_src_param(ins
, &ins
->src
[0], src0_param
);
1071 shader_addline(buffer
, "ADD TMP.x, %s, helper_const.z;\n", src0_param
);
1072 shader_addline(buffer
, "ARL A0.x, TMP.x;\n");
1076 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1077 * with more than one component. Thus replicate the first source argument over all
1078 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1079 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1080 tmp_src
.swizzle
= (tmp_src
.swizzle
& 0x3) * 0x55;
1081 shader_arb_add_src_param(ins
, &tmp_src
, src0_param
);
1082 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1087 shader_hw_map2gl(ins
);
1091 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1093 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1094 DWORD shader_version
= ins
->ctx
->reg_maps
->shader_version
;
1095 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1099 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1100 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1102 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1103 dst
->register_idx
, !!dst
->rel_addr
, reg_dest
, &is_color
);
1105 if (shader_version
>= WINED3DPS_VERSION(2,0))
1107 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1108 shader_addline(buffer
, "KIL %s;\n", reg_dest
);
1110 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1111 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1113 shader_addline(buffer
, "MOV TMP, %s;\n", reg_dest
);
1114 shader_addline(buffer
, "MOV TMP.w, one.w;\n");
1115 shader_addline(buffer
, "KIL TMP;\n");
1119 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1121 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1122 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1123 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1125 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1126 DWORD shader_version
= ins
->ctx
->reg_maps
->shader_version
;
1127 BOOL projected
= FALSE
, bias
= FALSE
;
1131 DWORD reg_sampler_code
;
1133 /* All versions have a destination register */
1134 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1135 dst
->register_idx
, !!dst
->rel_addr
, reg_dest
, &is_color
);
1137 /* 1.0-1.3: Use destination register as coordinate source.
1138 1.4+: Use provided coordinate source register. */
1139 if (shader_version
< WINED3DPS_VERSION(1,4))
1140 strcpy(reg_coord
, reg_dest
);
1142 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, reg_coord
);
1144 /* 1.0-1.4: Use destination register number as texture code.
1145 2.0+: Use provided sampler number as texure code. */
1146 if (shader_version
< WINED3DPS_VERSION(2,0))
1147 reg_sampler_code
= dst
->register_idx
;
1149 reg_sampler_code
= ins
->src
[1].register_idx
;
1152 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1153 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1154 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1156 if (shader_version
< WINED3DPS_VERSION(1,4))
1159 if(reg_sampler_code
< MAX_TEXTURES
) {
1160 flags
= deviceImpl
->stateBlock
->textureState
[reg_sampler_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1162 if (flags
& WINED3DTTFF_PROJECTED
) {
1166 else if (shader_version
< WINED3DPS_VERSION(2,0))
1168 DWORD src_mod
= ins
->src
[0].modifiers
;
1169 if (src_mod
== WINED3DSPSM_DZ
) {
1171 } else if(src_mod
== WINED3DSPSM_DW
) {
1175 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) projected
= TRUE
;
1176 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) bias
= TRUE
;
1178 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, projected
, bias
);
1181 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
1183 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1184 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1187 shader_arb_get_write_mask(ins
, dst
, tmp
);
1188 if (ins
->ctx
->reg_maps
->shader_version
!= WINED3DPS_VERSION(1,4))
1190 DWORD reg
= dst
->register_idx
;
1191 shader_addline(buffer
, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg
, tmp
, reg
);
1195 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, reg_src
);
1196 shader_addline(buffer
, "MOV R%u%s, %s;\n", dst
->register_idx
, tmp
, reg_src
);
1200 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
1202 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1203 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1204 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1207 DWORD reg1
= ins
->dst
[0].register_idx
;
1211 sprintf(dst_str
, "T%u", reg1
);
1212 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_str
);
1213 shader_addline(buffer
, "MOV TMP.x, %s.w;\n", src_str
);
1214 shader_addline(buffer
, "MOV TMP.y, %s.x;\n", src_str
);
1215 flags
= reg1
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg1
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1216 shader_hw_sample(ins
, reg1
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1219 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
1221 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1223 DWORD reg1
= ins
->dst
[0].register_idx
;
1227 sprintf(dst_str
, "T%u", reg1
);
1228 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_str
);
1229 shader_addline(buffer
, "MOV TMP.x, %s.y;\n", src_str
);
1230 shader_addline(buffer
, "MOV TMP.y, %s.z;\n", src_str
);
1231 shader_hw_sample(ins
, reg1
, dst_str
, "TMP", FALSE
, FALSE
);
1234 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
1236 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1237 DWORD reg1
= ins
->dst
[0].register_idx
;
1241 sprintf(dst_str
, "T%u", reg1
);
1242 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_str
);
1243 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, FALSE
, FALSE
);
1246 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
1248 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1249 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1250 BOOL has_bumpmat
= FALSE
;
1251 BOOL has_luminance
= FALSE
;
1255 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1258 DWORD reg_dest_code
;
1260 /* All versions have a destination register */
1261 reg_dest_code
= dst
->register_idx
;
1262 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1263 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1264 dst
->register_idx
, !!dst
->rel_addr
, reg_coord
, &is_color
);
1266 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
1267 if (This
->bumpenvmatconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
1268 && reg_dest_code
== This
->bumpenvmatconst
[i
].texunit
)
1274 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
1275 if (This
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
1276 && reg_dest_code
== This
->luminanceconst
[i
].texunit
)
1278 has_luminance
= TRUE
;
1284 DWORD src
= ins
->src
[0].register_idx
;
1286 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1288 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
1289 shader_addline(buffer
, "DP3 TMP.x, TMP2, T%u;\n", src
);
1290 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
1291 shader_addline(buffer
, "DP3 TMP.y, TMP2, T%u;\n", src
);
1293 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1294 * so we can't let the GL handle this.
1296 if (((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
]
1297 & WINED3DTTFF_PROJECTED
) {
1298 shader_addline(buffer
, "RCP TMP2.w, %s.w;\n", reg_coord
);
1299 shader_addline(buffer
, "MUL TMP2.xy, %s, TMP2.w;\n", reg_coord
);
1300 shader_addline(buffer
, "ADD TMP.xy, TMP, TMP2;\n");
1302 shader_addline(buffer
, "ADD TMP.xy, TMP, %s;\n", reg_coord
);
1305 shader_hw_sample(ins
, reg_dest_code
, reg_coord
, "TMP", FALSE
, FALSE
);
1307 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
&& has_luminance
)
1309 shader_addline(buffer
, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1310 src
, reg_dest_code
, reg_dest_code
);
1311 shader_addline(buffer
, "MUL %s, %s, TMP;\n", reg_coord
, reg_coord
);
1316 if(reg_dest_code
< MAX_TEXTURES
) {
1317 tf
= ((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1321 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1322 shader_hw_sample(ins
, reg_dest_code
, reg_coord
, reg_coord
, tf
& WINED3DTTFF_PROJECTED
, FALSE
);
1326 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
1328 DWORD reg
= ins
->dst
[0].register_idx
;
1329 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1332 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0_name
);
1333 shader_addline(buffer
, "DP3 TMP.x, T%u, %s;\n", reg
, src0_name
);
1336 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
1338 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1339 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1341 DWORD reg
= ins
->dst
[0].register_idx
;
1342 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1346 sprintf(dst_str
, "T%u", reg
);
1347 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0_name
);
1348 shader_addline(buffer
, "DP3 TMP.y, T%u, %s;\n", reg
, src0_name
);
1349 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1350 shader_hw_sample(ins
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1353 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
1355 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1356 DWORD reg
= ins
->dst
[0].register_idx
;
1357 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1358 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1361 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0_name
);
1362 shader_addline(buffer
, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state
->current_row
, reg
, src0_name
);
1363 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
1366 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
1368 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1369 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1371 DWORD reg
= ins
->dst
[0].register_idx
;
1372 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1373 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1377 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0_name
);
1378 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1380 /* Sample the texture using the calculated coordinates */
1381 sprintf(dst_str
, "T%u", reg
);
1382 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1383 shader_hw_sample(ins
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1384 current_state
->current_row
= 0;
1387 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
1389 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1390 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1392 DWORD reg
= ins
->dst
[0].register_idx
;
1393 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1394 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1398 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0_name
);
1399 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1401 /* Construct the eye-ray vector from w coordinates */
1402 shader_addline(buffer
, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[0]);
1403 shader_addline(buffer
, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[1]);
1404 shader_addline(buffer
, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg
);
1406 /* Calculate reflection vector
1408 shader_addline(buffer
, "DP3 TMP.w, TMP, TMP2;\n");
1409 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1410 shader_addline(buffer
, "DP3 TMP2.w, TMP, TMP;\n");
1411 shader_addline(buffer
, "RCP TMP2.w, TMP2.w;\n");
1412 shader_addline(buffer
, "MUL TMP.w, TMP.w, TMP2.w;\n");
1413 shader_addline(buffer
, "MUL TMP, TMP.w, TMP;\n");
1414 shader_addline(buffer
, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1416 /* Sample the texture using the calculated coordinates */
1417 sprintf(dst_str
, "T%u", reg
);
1418 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1419 shader_hw_sample(ins
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1420 current_state
->current_row
= 0;
1423 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
1425 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1426 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1428 DWORD reg
= ins
->dst
[0].register_idx
;
1429 DWORD reg3
= ins
->src
[1].register_idx
;
1430 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1431 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1435 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0_name
);
1436 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1438 /* Calculate reflection vector.
1441 * TMP.xyz = 2 * --------- * N - E
1444 * Which normalizes the normal vector
1446 shader_addline(buffer
, "DP3 TMP.w, TMP, C[%u];\n", reg3
);
1447 shader_addline(buffer
, "DP3 TMP2.w, TMP, TMP;\n");
1448 shader_addline(buffer
, "RCP TMP2.w, TMP2.w;\n");
1449 shader_addline(buffer
, "MUL TMP.w, TMP.w, TMP2.w;\n");
1450 shader_addline(buffer
, "MUL TMP, TMP.w, TMP;\n");
1451 shader_addline(buffer
, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3
);
1453 /* Sample the texture using the calculated coordinates */
1454 sprintf(dst_str
, "T%u", reg
);
1455 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1456 shader_hw_sample(ins
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1457 current_state
->current_row
= 0;
1460 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
1462 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1463 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1467 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1468 * which is essentially an input, is the destination register because it is the first
1469 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1472 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1473 dst
->register_idx
, !!dst
->rel_addr
, dst_name
, &is_color
);
1475 /* According to the msdn, the source register(must be r5) is unusable after
1476 * the texdepth instruction, so we're free to modify it
1478 shader_addline(buffer
, "MIN %s.y, %s.y, one.y;\n", dst_name
, dst_name
);
1480 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1481 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1482 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1484 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
1485 shader_addline(buffer
, "MUL TMP.x, %s.x, %s.y;\n", dst_name
, dst_name
);
1486 shader_addline(buffer
, "MIN TMP.x, TMP.x, one.x;\n");
1487 shader_addline(buffer
, "MAX result.depth, TMP.x, 0.0;\n");
1490 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1491 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1492 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1493 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
1495 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1496 DWORD sampler_idx
= ins
->dst
[0].register_idx
;
1500 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0
);
1501 shader_addline(buffer
, "MOV TMP, 0.0;\n");
1502 shader_addline(buffer
, "DP3 TMP.x, T%u, %s;\n", sampler_idx
, src0
);
1504 sprintf(dst_str
, "T%u", sampler_idx
);
1505 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TMP", FALSE
/* Only one coord, can't be projected */, FALSE
);
1508 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1509 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1510 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
1512 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1516 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1519 /* Handle output register */
1520 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1521 dst
->register_idx
, !!dst
->rel_addr
, dst_str
, &is_color
);
1522 shader_arb_get_write_mask(ins
, dst
, dst_mask
);
1524 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0
);
1525 shader_addline(buffer
, "DP3 %s%s, T%u, %s;\n", dst_str
, dst_mask
, dst
->register_idx
, src0
);
1527 /* TODO: Handle output modifiers */
1530 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1531 * Perform the 3rd row of a 3x3 matrix multiply */
1532 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
1534 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1535 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1541 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1542 dst
->register_idx
, !!dst
->rel_addr
, dst_str
, &is_color
);
1543 shader_arb_get_write_mask(ins
, dst
, dst_mask
);
1545 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0
);
1546 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", dst
->register_idx
, src0
);
1547 shader_addline(buffer
, "MOV %s%s, TMP;\n", dst_str
, dst_mask
);
1549 /* TODO: Handle output modifiers */
1552 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1553 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1554 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1555 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1557 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
1559 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1560 DWORD dst_reg
= ins
->dst
[0].register_idx
;
1563 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src0
);
1564 shader_addline(buffer
, "DP3 TMP.y, T%u, %s;\n", dst_reg
, src0
);
1566 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1567 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1568 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1570 shader_addline(buffer
, "RCP TMP.y, TMP.y;\n");
1571 shader_addline(buffer
, "MUL TMP.x, TMP.x, TMP.y;\n");
1572 shader_addline(buffer
, "MIN TMP.x, TMP.x, one.x;\n");
1573 shader_addline(buffer
, "MAX result.depth, TMP.x, 0.0;\n");
1576 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1577 Vertex/Pixel shaders to ARB_vertex_program codes */
1578 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
1581 int nComponents
= 0;
1582 struct wined3d_shader_dst_param tmp_dst
= {0};
1583 struct wined3d_shader_src_param tmp_src
[2] = {{0}};
1584 struct wined3d_shader_instruction tmp_ins
;
1586 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
1588 /* Set constants for the temporary argument */
1589 tmp_ins
.ctx
= ins
->ctx
;
1590 tmp_ins
.dst_count
= 1;
1591 tmp_ins
.dst
= &tmp_dst
;
1592 tmp_ins
.src_count
= 2;
1593 tmp_ins
.src
= tmp_src
;
1595 switch(ins
->handler_idx
)
1597 case WINED3DSIH_M4x4
:
1599 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
1601 case WINED3DSIH_M4x3
:
1603 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
1605 case WINED3DSIH_M3x4
:
1607 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
1609 case WINED3DSIH_M3x3
:
1611 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
1613 case WINED3DSIH_M3x2
:
1615 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
1618 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1622 tmp_dst
= ins
->dst
[0];
1623 tmp_src
[0] = ins
->src
[0];
1624 tmp_src
[1] = ins
->src
[1];
1625 for (i
= 0; i
< nComponents
; i
++) {
1626 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
1627 shader_hw_map2gl(&tmp_ins
);
1628 ++tmp_src
[1].register_idx
;
1632 static void vshader_hw_rsq_rcp(const struct wined3d_shader_instruction
*ins
)
1634 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1635 const char *instruction
;
1639 switch(ins
->handler_idx
)
1641 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
1642 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
1643 default: instruction
= "";
1644 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1648 strcpy(tmpLine
, instruction
);
1649 shader_arb_add_dst_param(ins
, &ins
->dst
[0], tmpLine
); /* Destination */
1650 strcat(tmpLine
, ",");
1651 shader_arb_add_src_param(ins
, &ins
->src
[0], tmpLine
);
1652 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
1654 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1657 strcat(tmpLine
, ".w");
1660 shader_addline(buffer
, "%s;\n", tmpLine
);
1663 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
1665 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1666 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1670 DWORD shift
= dst
->shift
;
1671 BOOL sat
= dst
->modifiers
& WINED3DSPDM_SATURATE
;
1674 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1675 dst
->register_idx
, !!dst
->rel_addr
, dst_name
, &is_color
);
1676 shader_arb_get_write_mask(ins
, dst
, dst_wmask
);
1678 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_name
);
1679 shader_addline(buffer
, "DP3 TMP, %s, %s;\n", src_name
, src_name
);
1680 shader_addline(buffer
, "RSQ TMP, TMP.x;\n");
1681 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1682 shader_addline(buffer
, "MUL%s %s%s, %s, TMP;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
1686 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1689 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
1691 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1692 * must contain fixed constants. So we need a separate function to filter those constants and
1695 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1696 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1700 DWORD shift
= dst
->shift
;
1701 BOOL sat
= dst
->modifiers
& WINED3DSPDM_SATURATE
;
1704 shader_arb_get_register_name(ins
->ctx
->shader
, dst
->register_type
,
1705 dst
->register_idx
, !!dst
->rel_addr
, dst_name
, &is_color
);
1706 shader_arb_get_write_mask(ins
, dst
, dst_wmask
);
1708 pshader_gen_input_modifier_line(ins
->ctx
->shader
, buffer
, &ins
->src
[0], 0, src_name
);
1709 shader_addline(buffer
, "SCS%s %s%s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
1713 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1717 static GLuint
create_arb_blt_vertex_program(const WineD3D_GL_Info
*gl_info
)
1719 GLuint program_id
= 0;
1720 const char *blt_vprogram
=
1722 "PARAM c[1] = { { 1, 0.5 } };\n"
1723 "MOV result.position, vertex.position;\n"
1724 "MOV result.color, c[0].x;\n"
1725 "MOV result.texcoord[0], vertex.texcoord[0];\n"
1728 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
1729 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
1730 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_vprogram
), blt_vprogram
));
1732 if (glGetError() == GL_INVALID_OPERATION
) {
1734 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
1735 FIXME("Vertex program error at position %d: %s\n", pos
,
1736 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
1742 static GLuint
create_arb_blt_fragment_program(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
1744 GLuint program_id
= 0;
1745 static const char * const blt_fprograms
[tex_type_count
] =
1752 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1753 "MOV result.depth.z, R0.x;\n"
1760 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1761 "MOV result.depth.z, R0.x;\n"
1766 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1767 "MOV result.depth.z, R0.x;\n"
1771 if (!blt_fprograms
[tex_type
])
1773 FIXME("tex_type %#x not supported\n", tex_type
);
1777 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
1778 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
1779 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_fprograms
[tex_type
]), blt_fprograms
[tex_type
]));
1781 if (glGetError() == GL_INVALID_OPERATION
) {
1783 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
1784 FIXME("Fragment program error at position %d: %s\n", pos
,
1785 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
1791 static void shader_arb_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
1792 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1793 struct shader_arb_priv
*priv
= This
->shader_priv
;
1794 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1797 struct vs_compile_args compile_args
;
1799 TRACE("Using vertex shader\n");
1800 find_vs_compile_args((IWineD3DVertexShaderImpl
*) This
->stateBlock
->vertexShader
, This
->stateBlock
, &compile_args
);
1801 priv
->current_vprogram_id
= find_gl_vshader((IWineD3DVertexShaderImpl
*) This
->stateBlock
->vertexShader
, &compile_args
);
1803 /* Bind the vertex program */
1804 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
1805 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
1807 /* Enable OpenGL vertex programs */
1808 glEnable(GL_VERTEX_PROGRAM_ARB
);
1809 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1810 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
1811 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
1812 priv
->current_vprogram_id
= 0;
1813 glDisable(GL_VERTEX_PROGRAM_ARB
);
1814 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1818 struct ps_compile_args compile_args
;
1819 TRACE("Using pixel shader\n");
1820 find_ps_compile_args((IWineD3DPixelShaderImpl
*) This
->stateBlock
->pixelShader
, This
->stateBlock
, &compile_args
);
1821 priv
->current_fprogram_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) This
->stateBlock
->pixelShader
,
1824 /* Bind the fragment program */
1825 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
1826 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
1828 if(!priv
->use_arbfp_fixed_func
) {
1829 /* Enable OpenGL fragment programs */
1830 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1831 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1833 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
1834 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
) && !priv
->use_arbfp_fixed_func
) {
1835 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1836 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1837 * replacement shader
1839 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
1840 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1841 priv
->current_fprogram_id
= 0;
1845 static void shader_arb_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
1846 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1847 struct shader_arb_priv
*priv
= This
->shader_priv
;
1848 GLuint
*blt_fprogram
= &priv
->depth_blt_fprogram_id
[tex_type
];
1849 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1851 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
1852 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
1853 glEnable(GL_VERTEX_PROGRAM_ARB
);
1855 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
);
1856 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
1857 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1860 static void shader_arb_deselect_depth_blt(IWineD3DDevice
*iface
) {
1861 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1862 struct shader_arb_priv
*priv
= This
->shader_priv
;
1863 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1865 if (priv
->current_vprogram_id
) {
1866 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
1867 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1869 glEnable(GL_VERTEX_PROGRAM_ARB
);
1870 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1872 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
1874 glDisable(GL_VERTEX_PROGRAM_ARB
);
1875 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1878 if (priv
->current_fprogram_id
) {
1879 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
1880 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1882 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1883 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1885 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
1887 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
1888 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1892 static void shader_arb_destroy(IWineD3DBaseShader
*iface
) {
1893 IWineD3DBaseShaderImpl
*baseShader
= (IWineD3DBaseShaderImpl
*) iface
;
1894 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)baseShader
->baseShader
.device
)->adapter
->gl_info
;
1896 if (shader_is_pshader_version(baseShader
->baseShader
.reg_maps
.shader_version
))
1898 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
1902 for(i
= 0; i
< This
->num_gl_shaders
; i
++) {
1903 GL_EXTCALL(glDeleteProgramsARB(1, &This
->gl_shaders
[i
].prgId
));
1904 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1907 HeapFree(GetProcessHeap(), 0, This
->gl_shaders
);
1908 This
->gl_shaders
= NULL
;
1909 This
->num_gl_shaders
= 0;
1910 This
->shader_array_size
= 0;
1912 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*) iface
;
1916 for(i
= 0; i
< This
->num_gl_shaders
; i
++) {
1917 GL_EXTCALL(glDeleteProgramsARB(1, &This
->gl_shaders
[i
].prgId
));
1918 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1921 HeapFree(GetProcessHeap(), 0, This
->gl_shaders
);
1922 This
->gl_shaders
= NULL
;
1923 This
->num_gl_shaders
= 0;
1924 This
->shader_array_size
= 0;
1928 static HRESULT
shader_arb_alloc(IWineD3DDevice
*iface
) {
1929 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1930 This
->shader_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
1934 static void shader_arb_free(IWineD3DDevice
*iface
) {
1935 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1936 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1937 struct shader_arb_priv
*priv
= This
->shader_priv
;
1940 if(priv
->depth_blt_vprogram_id
) {
1941 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
1943 for (i
= 0; i
< tex_type_count
; ++i
) {
1944 if (priv
->depth_blt_fprogram_id
[i
]) {
1945 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id
[i
]));
1949 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
1952 static BOOL
shader_arb_dirty_const(IWineD3DDevice
*iface
) {
1956 static void arbfp_add_sRGB_correction(SHADER_BUFFER
*buffer
, const char *fragcolor
, const char *tmp1
,
1957 const char *tmp2
, const char *tmp3
, const char *tmp4
) {
1958 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1960 /* Calculate the > 0.0031308 case */
1961 shader_addline(buffer
, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1
, fragcolor
);
1962 shader_addline(buffer
, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1
, fragcolor
);
1963 shader_addline(buffer
, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1
, fragcolor
);
1964 shader_addline(buffer
, "MUL %s, %s, srgb_mul_hi;\n", tmp1
, tmp1
);
1965 shader_addline(buffer
, "SUB %s, %s, srgb_sub_hi;\n", tmp1
, tmp1
);
1966 /* Calculate the < case */
1967 shader_addline(buffer
, "MUL %s, srgb_mul_low, %s;\n", tmp2
, fragcolor
);
1968 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1969 shader_addline(buffer
, "SLT %s, srgb_comparison, %s;\n", tmp3
, fragcolor
);
1970 shader_addline(buffer
, "SGE %s, srgb_comparison, %s;\n", tmp4
, fragcolor
);
1971 /* Store the components > 0.0031308 in the destination */
1972 shader_addline(buffer
, "MUL %s, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
1973 /* Add the components that are < 0.0031308 */
1974 shader_addline(buffer
, "MAD result.color.xyz, %s, %s, %s;\n", tmp2
, tmp4
, fragcolor
);
1975 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1978 static GLuint
shader_arb_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
, const struct ps_compile_args
*args
) {
1979 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
1980 const shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
1981 CONST DWORD
*function
= This
->baseShader
.function
;
1982 DWORD shader_version
= reg_maps
->shader_version
;
1983 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
1984 const local_constant
*lconst
;
1986 const char *fragcolor
;
1988 /* Create the hw ARB shader */
1989 shader_addline(buffer
, "!!ARBfp1.0\n");
1991 if (shader_version
< WINED3DPS_VERSION(3,0)) {
1996 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
1999 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
2002 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
2007 shader_addline(buffer
, "TEMP TMP;\n"); /* Used in matrix ops */
2008 shader_addline(buffer
, "TEMP TMP2;\n"); /* Used in matrix ops */
2009 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
2010 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
2011 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
2012 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
2013 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
2014 shader_addline(buffer
, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
2016 if (shader_version
< WINED3DPS_VERSION(2,0)) {
2019 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
2020 fragcolor
= "TMP_COLOR";
2023 /* Base Declarations */
2024 shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
2026 /* Base Shader Body */
2027 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
2029 if(args
->srgb_correction
) {
2030 arbfp_add_sRGB_correction(buffer
, fragcolor
, "TMP", "TMP2", "TA", "TB");
2032 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
2033 shader_addline(buffer
, "END\n");
2035 /* TODO: change to resource.glObjectHandle or something like that */
2036 GL_EXTCALL(glGenProgramsARB(1, &retval
));
2038 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
2039 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
2041 TRACE("Created hw pixel shader, prg=%d\n", retval
);
2042 /* Create the program and check for errors */
2043 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
2044 buffer
->bsize
, buffer
->buffer
));
2046 if (glGetError() == GL_INVALID_OPERATION
) {
2048 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
2049 FIXME("HW PixelShader Error at position %d: %s\n",
2050 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2054 /* Load immediate constants */
2055 if(!This
->baseShader
.load_local_constsF
) {
2056 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
2057 const float *value
= (const float *)lconst
->value
;
2058 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, lconst
->idx
, value
));
2059 checkGLcall("glProgramLocalParameter4fvARB");
2066 static GLuint
shader_arb_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
, const struct vs_compile_args
*args
) {
2067 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
2068 const shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
2069 CONST DWORD
*function
= This
->baseShader
.function
;
2070 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
2071 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
2072 const local_constant
*lconst
;
2075 /* Create the hw ARB shader */
2076 shader_addline(buffer
, "!!ARBvp1.0\n");
2077 shader_addline(buffer
, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This
->rel_offset
);
2079 /* Mesa supports only 95 constants */
2080 if (GL_VEND(MESA
) || GL_VEND(WINE
))
2081 This
->baseShader
.limits
.constant_float
=
2082 min(95, This
->baseShader
.limits
.constant_float
);
2084 shader_addline(buffer
, "TEMP TMP;\n");
2086 /* Base Declarations */
2087 shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
2089 /* We need a constant to fixup the final position */
2090 shader_addline(buffer
, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS
);
2092 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2093 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2094 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2095 * a replacement shader depend on the texcoord.w being set properly.
2097 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2098 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2099 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2100 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2101 * this can eat a number of instructions, so skip it unless this cap is set as well
2103 if(!GL_SUPPORT(NV_VERTEX_PROGRAM
)) {
2104 shader_addline(buffer
, "MOV result.color.secondary, -helper_const.wwwy;\n");
2106 if((GLINFO_LOCATION
).set_texcoord_w
&& !device
->frag_pipe
->ffp_proj_control
) {
2108 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
2109 if(This
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
2110 This
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
2111 shader_addline(buffer
, "MOV result.texcoord[%u].w, -helper_const.y;\n", i
);
2117 /* Base Shader Body */
2118 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
2120 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
2121 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
2122 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
2123 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
2125 if(args
->fog_src
== VS_FOG_Z
) {
2126 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
2127 } else if (!reg_maps
->fog
) {
2128 shader_addline(buffer
, "MOV result.fogcoord, helper_const.w;\n");
2131 /* Write the final position.
2133 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2134 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2135 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2136 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2138 shader_addline(buffer
, "MUL TMP, posFixup, TMP_OUT.w;\n");
2139 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2140 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2142 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2143 * and the glsl equivalent
2145 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2147 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
2149 shader_addline(buffer
, "END\n");
2151 /* TODO: change to resource.glObjectHandle or something like that */
2152 GL_EXTCALL(glGenProgramsARB(1, &ret
));
2154 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
2155 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
2157 TRACE("Created hw vertex shader, prg=%d\n", ret
);
2158 /* Create the program and check for errors */
2159 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
2160 buffer
->bsize
, buffer
->buffer
));
2162 if (glGetError() == GL_INVALID_OPERATION
) {
2164 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
2165 FIXME("HW VertexShader Error at position %d: %s\n",
2166 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2169 /* Load immediate constants */
2170 if(!This
->baseShader
.load_local_constsF
) {
2171 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
2172 const float *value
= (const float *)lconst
->value
;
2173 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, lconst
->idx
, value
));
2180 static void shader_arb_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
2182 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2183 * then overwrite the shader specific ones
2185 none_shader_backend
.shader_get_caps(devtype
, gl_info
, pCaps
);
2187 if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
2188 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(1,1);
2189 TRACE_(d3d_caps
)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2190 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
) - ARB_SHADER_RESERVED_VS_CONSTS
;
2193 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
)) {
2194 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(1,4);
2195 pCaps
->PixelShader1xMaxValue
= 8.0;
2196 TRACE_(d3d_caps
)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2197 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
) - ARB_SHADER_RESERVED_PS_CONSTS
;
2201 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
2203 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
2205 TRACE("Checking support for color_fixup:\n");
2206 dump_color_fixup_desc(fixup
);
2209 /* We support everything except YUV conversions. */
2210 if (!is_yuv_fixup(fixup
))
2216 TRACE("[FAILED]\n");
2220 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
2222 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
2223 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
2224 /* WINED3DSIH_BEM */ pshader_hw_bem
,
2225 /* WINED3DSIH_BREAK */ NULL
,
2226 /* WINED3DSIH_BREAKC */ NULL
,
2227 /* WINED3DSIH_BREAKP */ NULL
,
2228 /* WINED3DSIH_CALL */ NULL
,
2229 /* WINED3DSIH_CALLNZ */ NULL
,
2230 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
2231 /* WINED3DSIH_CND */ pshader_hw_cnd
,
2232 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
2233 /* WINED3DSIH_DCL */ NULL
,
2234 /* WINED3DSIH_DEF */ NULL
,
2235 /* WINED3DSIH_DEFB */ NULL
,
2236 /* WINED3DSIH_DEFI */ NULL
,
2237 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
2238 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
2239 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
2240 /* WINED3DSIH_DST */ shader_hw_map2gl
,
2241 /* WINED3DSIH_DSX */ NULL
,
2242 /* WINED3DSIH_DSY */ NULL
,
2243 /* WINED3DSIH_ELSE */ NULL
,
2244 /* WINED3DSIH_ENDIF */ NULL
,
2245 /* WINED3DSIH_ENDLOOP */ NULL
,
2246 /* WINED3DSIH_ENDREP */ NULL
,
2247 /* WINED3DSIH_EXP */ shader_hw_map2gl
,
2248 /* WINED3DSIH_EXPP */ shader_hw_map2gl
,
2249 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
2250 /* WINED3DSIH_IF */ NULL
,
2251 /* WINED3DSIH_IFC */ NULL
,
2252 /* WINED3DSIH_LABEL */ NULL
,
2253 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
2254 /* WINED3DSIH_LOG */ shader_hw_map2gl
,
2255 /* WINED3DSIH_LOGP */ shader_hw_map2gl
,
2256 /* WINED3DSIH_LOOP */ NULL
,
2257 /* WINED3DSIH_LRP */ shader_hw_map2gl
,
2258 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
2259 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
2260 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
2261 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
2262 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
2263 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
2264 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
2265 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
2266 /* WINED3DSIH_MOV */ shader_hw_mov
,
2267 /* WINED3DSIH_MOVA */ shader_hw_mov
,
2268 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
2269 /* WINED3DSIH_NOP */ shader_hw_map2gl
,
2270 /* WINED3DSIH_NRM */ shader_hw_nrm
,
2271 /* WINED3DSIH_PHASE */ NULL
,
2272 /* WINED3DSIH_POW */ shader_hw_map2gl
,
2273 /* WINED3DSIH_RCP */ vshader_hw_rsq_rcp
,
2274 /* WINED3DSIH_REP */ NULL
,
2275 /* WINED3DSIH_RET */ NULL
,
2276 /* WINED3DSIH_RSQ */ vshader_hw_rsq_rcp
,
2277 /* WINED3DSIH_SETP */ NULL
,
2278 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
2279 /* WINED3DSIH_SGN */ NULL
,
2280 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
2281 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
2282 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
2283 /* WINED3DSIH_TEX */ pshader_hw_tex
,
2284 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
2285 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
2286 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
2287 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
2288 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
2289 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
2290 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
2291 /* WINED3DSIH_TEXLDD */ NULL
,
2292 /* WINED3DSIH_TEXLDL */ NULL
,
2293 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
2294 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
2295 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
2296 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
2297 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
2298 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
2299 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
2300 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
2301 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
2302 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
2303 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
2304 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
2307 const shader_backend_t arb_program_shader_backend
= {
2308 shader_arb_instruction_handler_table
,
2310 shader_arb_select_depth_blt
,
2311 shader_arb_deselect_depth_blt
,
2312 shader_arb_update_float_vertex_constants
,
2313 shader_arb_update_float_pixel_constants
,
2314 shader_arb_load_constants
,
2315 shader_arb_load_np2fixup_constants
,
2319 shader_arb_dirty_const
,
2320 shader_arb_generate_pshader
,
2321 shader_arb_generate_vshader
,
2322 shader_arb_get_caps
,
2323 shader_arb_color_fixup_supported
,
2326 /* ARB_fragment_program fixed function pipeline replacement definitions */
2327 #define ARB_FFP_CONST_TFACTOR 0
2328 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2329 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2330 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2331 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2333 struct arbfp_ffp_desc
2335 struct ffp_frag_desc parent
;
2337 unsigned int num_textures_used
;
2340 static void arbfp_enable(IWineD3DDevice
*iface
, BOOL enable
) {
2342 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
2343 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2345 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
2346 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2350 static HRESULT
arbfp_alloc(IWineD3DDevice
*iface
) {
2351 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
2352 struct shader_arb_priv
*priv
;
2353 /* Share private data between the shader backend and the pipeline replacement, if both
2354 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2355 * if no pixel shader is bound or not
2357 if(This
->shader_backend
== &arb_program_shader_backend
) {
2358 This
->fragment_priv
= This
->shader_priv
;
2360 This
->fragment_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
2361 if(!This
->fragment_priv
) return E_OUTOFMEMORY
;
2363 priv
= This
->fragment_priv
;
2364 priv
->fragment_shaders
= hash_table_create(ffp_frag_program_key_hash
, ffp_frag_program_key_compare
);
2365 priv
->use_arbfp_fixed_func
= TRUE
;
2369 static void arbfp_free_ffpshader(void *value
, void *gli
) {
2370 const WineD3D_GL_Info
*gl_info
= gli
;
2371 struct arbfp_ffp_desc
*entry_arb
= value
;
2374 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
2375 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2376 HeapFree(GetProcessHeap(), 0, entry_arb
);
2380 static void arbfp_free(IWineD3DDevice
*iface
) {
2381 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
2382 struct shader_arb_priv
*priv
= This
->fragment_priv
;
2384 hash_table_destroy(priv
->fragment_shaders
, arbfp_free_ffpshader
, &This
->adapter
->gl_info
);
2385 priv
->use_arbfp_fixed_func
= FALSE
;
2387 if(This
->shader_backend
!= &arb_program_shader_backend
) {
2388 HeapFree(GetProcessHeap(), 0, This
->fragment_priv
);
2392 static void arbfp_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct fragment_caps
*caps
)
2394 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
2395 WINED3DTEXOPCAPS_SELECTARG1
|
2396 WINED3DTEXOPCAPS_SELECTARG2
|
2397 WINED3DTEXOPCAPS_MODULATE4X
|
2398 WINED3DTEXOPCAPS_MODULATE2X
|
2399 WINED3DTEXOPCAPS_MODULATE
|
2400 WINED3DTEXOPCAPS_ADDSIGNED2X
|
2401 WINED3DTEXOPCAPS_ADDSIGNED
|
2402 WINED3DTEXOPCAPS_ADD
|
2403 WINED3DTEXOPCAPS_SUBTRACT
|
2404 WINED3DTEXOPCAPS_ADDSMOOTH
|
2405 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
2406 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
2407 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
2408 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
2409 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
2410 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
2411 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
2412 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
2413 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
2414 WINED3DTEXOPCAPS_DOTPRODUCT3
|
2415 WINED3DTEXOPCAPS_MULTIPLYADD
|
2416 WINED3DTEXOPCAPS_LERP
|
2417 WINED3DTEXOPCAPS_BUMPENVMAP
|
2418 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
2420 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2422 caps
->MaxTextureBlendStages
= 8;
2423 caps
->MaxSimultaneousTextures
= min(GL_LIMITS(fragment_samplers
), 8);
2425 caps
->PrimitiveMiscCaps
|= WINED3DPMISCCAPS_TSSARGTEMP
;
2427 #undef GLINFO_LOCATION
2429 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2430 static void state_texfactor_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2432 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2434 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2435 * application provided constants
2437 if(device
->shader_backend
== &arb_program_shader_backend
) {
2438 if (use_ps(stateblock
)) return;
2440 device
= stateblock
->wineD3DDevice
;
2441 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
2442 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
2445 D3DCOLORTOGLFLOAT4(stateblock
->renderState
[WINED3DRS_TEXTUREFACTOR
], col
);
2446 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
2447 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2451 static void state_arb_specularenable(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2453 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2455 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2456 * application provided constants
2458 if(device
->shader_backend
== &arb_program_shader_backend
) {
2459 if (use_ps(stateblock
)) return;
2461 device
= stateblock
->wineD3DDevice
;
2462 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
2463 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
2466 if(stateblock
->renderState
[WINED3DRS_SPECULARENABLE
]) {
2467 /* The specular color has no alpha */
2468 col
[0] = 1.0; col
[1] = 1.0;
2469 col
[2] = 1.0; col
[3] = 0.0;
2471 col
[0] = 0.0; col
[1] = 0.0;
2472 col
[2] = 0.0; col
[3] = 0.0;
2474 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
2475 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2478 static void set_bumpmat_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2479 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
2480 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2483 if (use_ps(stateblock
))
2486 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.bumpmat
[stage
]) {
2487 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2490 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
2491 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
2495 if(device
->shader_backend
== &arb_program_shader_backend
) {
2496 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2499 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
2500 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
2501 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
2504 mat
[0][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT00
]);
2505 mat
[0][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT01
]);
2506 mat
[1][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT10
]);
2507 mat
[1][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT11
]);
2509 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
2510 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2513 static void tex_bumpenvlum_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2514 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
2515 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2518 if (use_ps(stateblock
))
2521 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.luminanceparams
[stage
]) {
2522 /* The pixel shader has to know the luminance offset. Do a constants update if it
2523 * isn't scheduled anyway
2525 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
2526 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
2530 if(device
->shader_backend
== &arb_program_shader_backend
) {
2531 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2534 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
2535 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
2536 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
2539 param
[0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
]);
2540 param
[1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
]);
2544 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
2545 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2548 static const char *get_argreg(SHADER_BUFFER
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
) {
2551 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
2553 switch(arg
& WINED3DTA_SELECTMASK
) {
2554 case WINED3DTA_DIFFUSE
:
2555 ret
= "fragment.color.primary"; break;
2557 case WINED3DTA_CURRENT
:
2558 if(stage
== 0) ret
= "fragment.color.primary";
2562 case WINED3DTA_TEXTURE
:
2564 case 0: ret
= "tex0"; break;
2565 case 1: ret
= "tex1"; break;
2566 case 2: ret
= "tex2"; break;
2567 case 3: ret
= "tex3"; break;
2568 case 4: ret
= "tex4"; break;
2569 case 5: ret
= "tex5"; break;
2570 case 6: ret
= "tex6"; break;
2571 case 7: ret
= "tex7"; break;
2572 default: ret
= "unknown texture";
2576 case WINED3DTA_TFACTOR
:
2577 ret
= "tfactor"; break;
2579 case WINED3DTA_SPECULAR
:
2580 ret
= "fragment.color.secondary"; break;
2582 case WINED3DTA_TEMP
:
2583 ret
= "tempreg"; break;
2585 case WINED3DTA_CONSTANT
:
2586 FIXME("Implement perstage constants\n");
2588 case 0: ret
= "const0"; break;
2589 case 1: ret
= "const1"; break;
2590 case 2: ret
= "const2"; break;
2591 case 3: ret
= "const3"; break;
2592 case 4: ret
= "const4"; break;
2593 case 5: ret
= "const5"; break;
2594 case 6: ret
= "const6"; break;
2595 case 7: ret
= "const7"; break;
2596 default: ret
= "unknown constant";
2604 if(arg
& WINED3DTA_COMPLEMENT
) {
2605 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
2606 if(argnum
== 0) ret
= "arg0";
2607 if(argnum
== 1) ret
= "arg1";
2608 if(argnum
== 2) ret
= "arg2";
2610 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
2611 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
2612 if(argnum
== 0) ret
= "arg0";
2613 if(argnum
== 1) ret
= "arg1";
2614 if(argnum
== 2) ret
= "arg2";
2619 static void gen_ffp_instr(SHADER_BUFFER
*buffer
, unsigned int stage
, BOOL color
, BOOL alpha
,
2620 DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
) {
2621 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
2622 unsigned int mul
= 1;
2623 BOOL mul_final_dest
= FALSE
;
2625 if(color
&& alpha
) dstmask
= "";
2626 else if(color
) dstmask
= ".xyz";
2627 else dstmask
= ".w";
2629 if(dst
== tempreg
) dstreg
= "tempreg";
2630 else dstreg
= "ret";
2632 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
2633 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
2634 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
2637 case WINED3DTOP_DISABLE
:
2638 if(stage
== 0) shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
2641 case WINED3DTOP_SELECTARG2
:
2643 case WINED3DTOP_SELECTARG1
:
2644 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
2647 case WINED3DTOP_MODULATE4X
:
2649 case WINED3DTOP_MODULATE2X
:
2651 if(strcmp(dstreg
, "result.color") == 0) {
2653 mul_final_dest
= TRUE
;
2655 case WINED3DTOP_MODULATE
:
2656 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2659 case WINED3DTOP_ADDSIGNED2X
:
2661 if(strcmp(dstreg
, "result.color") == 0) {
2663 mul_final_dest
= TRUE
;
2665 case WINED3DTOP_ADDSIGNED
:
2666 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
2668 case WINED3DTOP_ADD
:
2669 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2672 case WINED3DTOP_SUBTRACT
:
2673 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2676 case WINED3DTOP_ADDSMOOTH
:
2677 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
2678 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2681 case WINED3DTOP_BLENDCURRENTALPHA
:
2682 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
2683 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2685 case WINED3DTOP_BLENDFACTORALPHA
:
2686 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
2687 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2689 case WINED3DTOP_BLENDTEXTUREALPHA
:
2690 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
2691 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2693 case WINED3DTOP_BLENDDIFFUSEALPHA
:
2694 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
2695 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2698 case WINED3DTOP_BLENDTEXTUREALPHAPM
:
2699 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
2700 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
2701 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2704 /* D3DTOP_PREMODULATE ???? */
2706 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR
:
2707 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
2708 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2710 case WINED3DTOP_MODULATEALPHA_ADDCOLOR
:
2711 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
2713 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA
:
2714 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
2715 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
2717 case WINED3DTOP_MODULATECOLOR_ADDALPHA
:
2718 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
2721 case WINED3DTOP_DOTPRODUCT3
:
2723 if(strcmp(dstreg
, "result.color") == 0) {
2725 mul_final_dest
= TRUE
;
2727 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
2728 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
2729 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
2732 case WINED3DTOP_MULTIPLYADD
:
2733 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
2736 case WINED3DTOP_LERP
:
2737 /* The msdn is not quite right here */
2738 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2741 case WINED3DTOP_BUMPENVMAP
:
2742 case WINED3DTOP_BUMPENVMAPLUMINANCE
:
2743 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2747 FIXME("Unhandled texture op %08x\n", op
);
2751 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
2752 } else if(mul
== 4) {
2753 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
2757 /* The stateblock is passed for GLINFO_LOCATION */
2758 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, IWineD3DStateBlockImpl
*stateblock
)
2761 SHADER_BUFFER buffer
;
2762 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2763 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2764 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2765 const char *textype
;
2766 const char *instr
, *sat
;
2767 char colorcor_dst
[8];
2769 DWORD arg0
, arg1
, arg2
;
2770 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
2772 const char *final_combiner_src
= "ret";
2774 /* Find out which textures are read */
2775 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2776 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) break;
2777 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
2778 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
2779 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
2780 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2781 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2782 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2784 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHA
) tex_read
[stage
] = TRUE
;
2785 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHAPM
) tex_read
[stage
] = TRUE
;
2786 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
) {
2787 bump_used
[stage
] = TRUE
;
2788 tex_read
[stage
] = TRUE
;
2790 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2791 bump_used
[stage
] = TRUE
;
2792 tex_read
[stage
] = TRUE
;
2793 luminance_used
[stage
] = TRUE
;
2794 } else if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDFACTORALPHA
) {
2795 tfactor_used
= TRUE
;
2798 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
2799 tfactor_used
= TRUE
;
2802 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
2803 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
2804 tempreg_used
= TRUE
;
2807 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) continue;
2808 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
2809 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
2810 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
2811 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2812 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2813 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2815 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
2816 tempreg_used
= TRUE
;
2818 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
2819 tfactor_used
= TRUE
;
2824 shader_buffer_init(&buffer
);
2826 shader_addline(&buffer
, "!!ARBfp1.0\n");
2828 switch(settings
->fog
) {
2829 case FOG_OFF
: break;
2830 case FOG_LINEAR
: shader_addline(&buffer
, "OPTION ARB_fog_linear;\n"); break;
2831 case FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
2832 case FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
2833 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
2836 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
2837 shader_addline(&buffer
, "TEMP TMP;\n");
2838 shader_addline(&buffer
, "TEMP ret;\n");
2839 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
2840 shader_addline(&buffer
, "TEMP arg0;\n");
2841 shader_addline(&buffer
, "TEMP arg1;\n");
2842 shader_addline(&buffer
, "TEMP arg2;\n");
2843 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2844 if(!tex_read
[stage
]) continue;
2845 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
2846 if(!bump_used
[stage
]) continue;
2847 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
2848 if(!luminance_used
[stage
]) continue;
2849 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
2852 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
2854 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
2856 if(settings
->sRGB_write
) {
2857 shader_addline(&buffer
, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2858 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
2859 shader_addline(&buffer
, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2860 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
2861 shader_addline(&buffer
, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2862 srgb_pow
, srgb_pow
, srgb_pow
);
2863 shader_addline(&buffer
, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2864 srgb_mul_high
, srgb_mul_high
, srgb_mul_high
);
2865 shader_addline(&buffer
, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2866 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
2869 /* Generate texture sampling instructions) */
2870 for(stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3DTOP_DISABLE
; stage
++) {
2871 if(!tex_read
[stage
]) continue;
2873 switch(settings
->op
[stage
].tex_type
) {
2874 case tex_1d
: textype
= "1D"; break;
2875 case tex_2d
: textype
= "2D"; break;
2876 case tex_3d
: textype
= "3D"; break;
2877 case tex_cube
: textype
= "CUBE"; break;
2878 case tex_rect
: textype
= "RECT"; break;
2879 default: textype
= "unexpected_textype"; break;
2882 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
||
2883 settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2889 if(settings
->op
[stage
].projected
== proj_none
) {
2891 } else if(settings
->op
[stage
].projected
== proj_count4
||
2892 settings
->op
[stage
].projected
== proj_count3
) {
2895 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
2900 (settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAP
||
2901 settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
)) {
2902 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
2903 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
2904 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
2905 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
2907 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2908 * so multiply the displacement with the dividing parameter before passing it to TXP
2910 if (settings
->op
[stage
].projected
!= proj_none
) {
2911 if(settings
->op
[stage
].projected
== proj_count4
) {
2912 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
2913 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
2915 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
2916 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
2919 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
2922 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
2923 instr
, sat
, stage
, stage
, textype
);
2924 if(settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2925 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
2926 stage
- 1, stage
- 1, stage
- 1);
2927 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
2929 } else if(settings
->op
[stage
].projected
== proj_count3
) {
2930 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
2931 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
2932 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
2933 instr
, sat
, stage
, stage
, textype
);
2935 shader_addline(&buffer
, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2936 instr
, sat
, stage
, stage
, stage
, textype
);
2939 sprintf(colorcor_dst
, "tex%u", stage
);
2940 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
2941 settings
->op
[stage
].color_fixup
);
2944 /* Generate the main shader */
2945 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2946 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) {
2948 final_combiner_src
= "fragment.color.primary";
2953 if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
2954 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
2955 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
2956 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
2957 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
2958 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
2959 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
2960 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
2961 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
2962 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
2963 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
2964 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
2966 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
&&
2967 settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
&&
2968 settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
&&
2969 settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
2972 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) {
2973 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
2974 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2975 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2977 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
2979 } else if(op_equal
) {
2980 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
2981 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2982 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2984 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
2985 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2986 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2987 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
2988 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
2989 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
2993 if(settings
->sRGB_write
) {
2994 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
2995 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg");
2996 shader_addline(&buffer
, "MOV result.color.w, ret.w;\n");
2998 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
3002 shader_addline(&buffer
, "END\n");
3004 /* Generate the shader */
3005 GL_EXTCALL(glGenProgramsARB(1, &ret
));
3006 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
3007 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
3009 if (glGetError() == GL_INVALID_OPERATION
) {
3011 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3012 FIXME("Fragment program error at position %d: %s\n", pos
,
3013 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3015 shader_buffer_free(&buffer
);
3019 static void fragment_prog_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3020 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
3021 struct shader_arb_priv
*priv
= device
->fragment_priv
;
3022 BOOL use_pshader
= use_ps(stateblock
);
3023 BOOL use_vshader
= use_vs(stateblock
);
3024 struct ffp_frag_settings settings
;
3025 const struct arbfp_ffp_desc
*desc
;
3028 TRACE("state %#x, stateblock %p, context %p\n", state
, stateblock
, context
);
3030 if(isStateDirty(context
, STATE_RENDER(WINED3DRS_FOGENABLE
))) {
3031 if(!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
3032 /* Reload fixed function constants since they collide with the pixel shader constants */
3033 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
3034 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
3036 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
3037 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
3038 } else if(use_pshader
&& !isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
3039 device
->shader_backend
->shader_select((IWineD3DDevice
*)stateblock
->wineD3DDevice
, use_pshader
, use_vshader
);
3045 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
3046 gen_ffp_frag_op(stateblock
, &settings
, FALSE
);
3047 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(priv
->fragment_shaders
, &settings
);
3049 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
3052 ERR("Out of memory\n");
3055 new_desc
->num_textures_used
= 0;
3056 for(i
= 0; i
< GL_LIMITS(texture_stages
); i
++) {
3057 if(settings
.op
[i
].cop
== WINED3DTOP_DISABLE
) break;
3058 new_desc
->num_textures_used
= i
;
3061 memcpy(&new_desc
->parent
.settings
, &settings
, sizeof(settings
));
3062 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, stateblock
);
3063 add_ffp_frag_shader(priv
->fragment_shaders
, &new_desc
->parent
);
3064 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
3068 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
3069 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
3072 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
3073 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
3074 priv
->current_fprogram_id
= desc
->shader
;
3076 if(device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
3077 /* Reload fixed function constants since they collide with the pixel shader constants */
3078 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
3079 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
3081 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
3082 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
3084 context
->last_was_pshader
= FALSE
;
3086 context
->last_was_pshader
= TRUE
;
3089 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
3090 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
3091 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
3092 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
3093 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
3095 * Don't call shader_select if the vertex shader is dirty, because it will be called later on by the vertex
3098 if(!isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
3099 device
->shader_backend
->shader_select((IWineD3DDevice
*)stateblock
->wineD3DDevice
, use_pshader
, use_vshader
);
3101 if (!isStateDirty(context
, STATE_VERTEXSHADERCONSTANT
) && (use_vshader
|| use_pshader
)) {
3102 device
->StateTable
[STATE_VERTEXSHADERCONSTANT
].apply(STATE_VERTEXSHADERCONSTANT
, stateblock
, context
);
3106 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
3110 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
3111 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
3112 * state table, so we need to handle that with a forwarding function. The other invisible side effect
3113 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
3114 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
3116 static void state_arbfp_fog(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3117 enum fogsource new_source
;
3119 TRACE("state %#x, stateblock %p, context %p\n", state
, stateblock
, context
);
3121 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
3122 fragment_prog_arbfp(state
, stateblock
, context
);
3125 if(!stateblock
->renderState
[WINED3DRS_FOGENABLE
]) return;
3127 if(stateblock
->renderState
[WINED3DRS_FOGTABLEMODE
] == WINED3DFOG_NONE
) {
3128 if(use_vs(stateblock
)) {
3129 new_source
= FOGSOURCE_VS
;
3131 if(stateblock
->renderState
[WINED3DRS_FOGVERTEXMODE
] == WINED3DFOG_NONE
|| context
->last_was_rhw
) {
3132 new_source
= FOGSOURCE_COORD
;
3134 new_source
= FOGSOURCE_FFP
;
3138 new_source
= FOGSOURCE_FFP
;
3140 if(new_source
!= context
->fog_source
) {
3141 context
->fog_source
= new_source
;
3142 state_fogstartend(STATE_RENDER(WINED3DRS_FOGSTART
), stateblock
, context
);
3146 static void textransform(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3147 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
3148 fragment_prog_arbfp(state
, stateblock
, context
);
3152 #undef GLINFO_LOCATION
3154 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] = {
3155 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), state_texfactor_arbfp
}, 0 },
3156 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3157 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3158 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3159 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3160 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3161 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3162 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3163 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3164 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3165 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3166 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3167 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3168 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3169 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3170 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3171 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3172 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3173 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3174 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3175 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3176 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3177 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3178 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3179 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3180 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3181 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3182 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3183 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3184 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3185 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3186 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3187 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3188 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3189 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3190 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3191 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3192 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3193 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3194 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3195 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3196 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3197 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3198 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3199 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3200 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3201 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3202 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3203 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3204 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3205 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3206 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3207 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3208 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3209 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3210 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3211 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3212 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3213 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3214 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3215 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3216 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3217 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3218 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3219 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3220 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3221 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3222 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3223 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3224 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3225 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3226 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3227 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3228 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3229 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3230 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3231 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3232 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3233 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3234 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3235 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3236 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3237 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3238 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3239 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3240 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3241 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3242 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3243 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3244 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3245 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3246 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3247 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3248 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3249 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3250 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3251 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3252 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3253 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3254 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3255 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3256 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3257 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3258 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3259 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3260 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3261 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3262 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3263 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3264 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3265 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3266 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3267 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3268 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3269 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3270 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3271 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3272 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3273 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3274 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3275 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3276 {STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim
}, 0 },
3277 {STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim
}, 0 },
3278 {STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim
}, 0 },
3279 {STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim
}, 0 },
3280 {STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim
}, 0 },
3281 {STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim
}, 0 },
3282 {STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim
}, 0 },
3283 {STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim
}, 0 },
3284 {STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3285 {STATE_RENDER(WINED3DRS_FOGENABLE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3286 {STATE_RENDER(WINED3DRS_FOGTABLEMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3287 {STATE_RENDER(WINED3DRS_FOGVERTEXMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3288 {STATE_RENDER(WINED3DRS_FOGSTART
), { STATE_RENDER(WINED3DRS_FOGSTART
), state_fogstartend
}, 0 },
3289 {STATE_RENDER(WINED3DRS_FOGEND
), { STATE_RENDER(WINED3DRS_FOGSTART
), state_fogstartend
}, 0 },
3290 {STATE_RENDER(WINED3DRS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3291 {STATE_RENDER(WINED3DRS_FOGCOLOR
), { STATE_RENDER(WINED3DRS_FOGCOLOR
), state_fogcolor
}, 0 },
3292 {STATE_RENDER(WINED3DRS_FOGDENSITY
), { STATE_RENDER(WINED3DRS_FOGDENSITY
), state_fogdensity
}, 0 },
3293 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3294 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3295 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3296 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3297 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3298 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3299 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3300 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3301 {STATE_RENDER(WINED3DRS_SPECULARENABLE
), { STATE_RENDER(WINED3DRS_SPECULARENABLE
), state_arb_specularenable
}, 0 },
3302 {0 /* Terminate */, { 0, 0 }, 0 },
3305 const struct fragment_pipeline arbfp_fragment_pipeline
= {
3310 shader_arb_color_fixup_supported
,
3311 arbfp_fragmentstate_template
,
3312 TRUE
/* We can disable projected textures */
3315 #define GLINFO_LOCATION device->adapter->gl_info
3317 struct arbfp_blit_priv
{
3318 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
3319 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
3320 GLenum yv12_rect_shader
, yv12_2d_shader
;
3323 static HRESULT
arbfp_blit_alloc(IWineD3DDevice
*iface
) {
3324 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3325 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
3326 if(!device
->blit_priv
) {
3327 ERR("Out of memory\n");
3328 return E_OUTOFMEMORY
;
3332 static void arbfp_blit_free(IWineD3DDevice
*iface
) {
3333 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3334 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
3337 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
3338 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
3339 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
3340 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
3341 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
3342 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
3343 checkGLcall("Delete yuv programs\n");
3347 static BOOL
gen_planar_yuv_read(SHADER_BUFFER
*buffer
, enum yuv_fixup yuv_fixup
, GLenum textype
, char *luminance
)
3350 const char *tex
, *texinstr
;
3352 if (yuv_fixup
== YUV_FIXUP_UYVY
) {
3360 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
3361 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
3363 /* This is more tricky than just replacing the texture type - we have to navigate
3364 * properly in the texture to find the correct chroma values
3366 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3370 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3371 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3372 * filtering when we sample the texture.
3374 * These are the rules for reading the chroma:
3380 * So we have to get the sampling x position in non-normalized coordinates in integers
3382 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
3383 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
3384 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
3386 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3388 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3389 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3392 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
3393 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
3395 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3396 * even and odd pixels respectively
3398 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
3399 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
3401 /* Sample Pixel 1 */
3402 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
3404 /* Put the value into either of the chroma values */
3405 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
3406 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
3407 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
3408 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
3410 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3411 * the pixel right to the current one. Otherwise, sample the left pixel.
3412 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3414 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3415 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
3416 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
3418 /* Put the value into the other chroma */
3419 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
3420 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
3421 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
3422 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
3424 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3425 * the current one and lerp the two U and V values
3428 /* This gives the correctly filtered luminance value */
3429 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
3434 static BOOL
gen_yv12_read(SHADER_BUFFER
*buffer
, GLenum textype
, char *luminance
)
3439 case GL_TEXTURE_2D
: tex
= "2D"; break;
3440 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
3442 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3446 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3447 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3448 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3449 * pitch of the luminance plane, the packing into the gl texture is a bit
3450 * unfortunate. If the whole texture is interpreted as luminance data it looks
3451 * approximately like this:
3453 * +----------------------------------+----
3465 * +----------------+-----------------+----
3467 * | U even rows | U odd rows |
3469 * +----------------+------------------ -
3471 * | V even rows | V odd rows |
3473 * +----------------+-----------------+----
3477 * So it appears as if there are 4 chroma images, but in fact the odd rows
3478 * in the chroma images are in the same row as the even ones. So its is
3479 * kinda tricky to read
3481 * When reading from rectangle textures, keep in mind that the input y coordinates
3482 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3484 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3485 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3487 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3488 /* the chroma planes have only half the width */
3489 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
3491 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3492 * the coordinate. Also read the right side of the image when reading odd lines
3494 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3497 if(textype
== GL_TEXTURE_2D
) {
3499 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
3501 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
3503 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
3504 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3506 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3507 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3508 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
3509 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3510 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3512 /* clamp, keep the half pixel origin in mind */
3513 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3514 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
3515 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3516 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3518 /* Read from [size - size+size/4] */
3519 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
3520 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3522 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3523 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3524 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
3525 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3526 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3527 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3529 /* Make sure to read exactly from the pixel center */
3530 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
3531 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
3534 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
3535 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
3536 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3537 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
3538 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
3540 /* Read the texture, put the result into the output register */
3541 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
3542 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
3544 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3545 * No need to clamp because we're just reusing the already clamped value from above
3547 if(textype
== GL_TEXTURE_2D
) {
3548 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3550 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3552 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
3553 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
3555 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3556 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3557 * values due to filtering
3559 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3560 if(textype
== GL_TEXTURE_2D
) {
3561 /* Multiply the y coordinate by 2/3 and clamp it */
3562 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3563 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3564 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3565 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
3567 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
3568 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3571 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
3572 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
3573 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
3580 static GLuint
gen_yuv_shader(IWineD3DDeviceImpl
*device
, enum yuv_fixup yuv_fixup
, GLenum textype
)
3583 SHADER_BUFFER buffer
;
3584 char luminance_component
;
3585 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
3588 shader_buffer_init(&buffer
);
3591 GL_EXTCALL(glGenProgramsARB(1, &shader
));
3592 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3593 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
3594 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3597 shader_buffer_free(&buffer
);
3601 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3602 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3603 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3604 * each single pixel it contains, and one U and one V value shared between both
3607 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3608 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3609 * take the format into account when generating the read swizzles
3611 * Reading the Y value is straightforward - just sample the texture. The hardware
3612 * takes care of filtering in the horizontal and vertical direction.
3614 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3615 * because that would mix the U and V values of one pixel or two adjacent pixels.
3616 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3617 * regardless of the filtering setting. Vertical filtering works automatically
3618 * though - the U and V values of two rows are mixed nicely.
3620 * Appart of avoiding filtering issues, the code has to know which value it just
3621 * read, and where it can find the other one. To determine this, it checks if
3622 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3624 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3625 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3627 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3628 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3629 * in an unfiltered situation. Finding the luminance on the other hand requires
3630 * finding out if it is an odd or even pixel. The real drawback of this approach
3631 * is filtering. This would have to be emulated completely in the shader, reading
3632 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3633 * vertically. Beyond that it would require adjustments to the texture handling
3634 * code to deal with the width scaling
3636 shader_addline(&buffer
, "!!ARBfp1.0\n");
3637 shader_addline(&buffer
, "TEMP luminance;\n");
3638 shader_addline(&buffer
, "TEMP temp;\n");
3639 shader_addline(&buffer
, "TEMP chroma;\n");
3640 shader_addline(&buffer
, "TEMP texcrd;\n");
3641 shader_addline(&buffer
, "TEMP texcrd2;\n");
3642 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3643 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3644 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
3648 case YUV_FIXUP_UYVY
:
3649 case YUV_FIXUP_YUY2
:
3650 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
3652 shader_buffer_free(&buffer
);
3657 case YUV_FIXUP_YV12
:
3658 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
3660 shader_buffer_free(&buffer
);
3666 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
3667 shader_buffer_free(&buffer
);
3671 /* Calculate the final result. Formula is taken from
3672 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3673 * ranges from -0.5 to 0.5
3675 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
3677 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
3678 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
3679 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
3680 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
3681 shader_addline(&buffer
, "END\n");
3684 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
3686 if (glGetError() == GL_INVALID_OPERATION
) {
3688 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3689 FIXME("Fragment program error at position %d: %s\n", pos
,
3690 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3692 shader_buffer_free(&buffer
);
3697 case YUV_FIXUP_YUY2
:
3698 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yuy2_rect_shader
= shader
;
3699 else priv
->yuy2_2d_shader
= shader
;
3702 case YUV_FIXUP_UYVY
:
3703 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->uyvy_rect_shader
= shader
;
3704 else priv
->uyvy_2d_shader
= shader
;
3707 case YUV_FIXUP_YV12
:
3708 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yv12_rect_shader
= shader
;
3709 else priv
->yv12_2d_shader
= shader
;
3716 static HRESULT
arbfp_blit_set(IWineD3DDevice
*iface
, const struct GlPixelFormatDesc
*format_desc
,
3717 GLenum textype
, UINT width
, UINT height
)
3720 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3721 float size
[4] = {width
, height
, 1, 1};
3722 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
3723 enum yuv_fixup yuv_fixup
;
3725 if (!is_yuv_fixup(format_desc
->color_fixup
))
3728 dump_color_fixup_desc(format_desc
->color_fixup
);
3729 /* Don't bother setting up a shader for unconverted formats */
3732 checkGLcall("glEnable(textype)");
3737 yuv_fixup
= get_yuv_fixup(format_desc
->color_fixup
);
3741 case YUV_FIXUP_YUY2
:
3742 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yuy2_rect_shader
: priv
->yuy2_2d_shader
;
3745 case YUV_FIXUP_UYVY
:
3746 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->uyvy_rect_shader
: priv
->uyvy_2d_shader
;
3749 case YUV_FIXUP_YV12
:
3750 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yv12_rect_shader
: priv
->yv12_2d_shader
;
3754 FIXME("Unsupported YUV fixup %#x, not setting a shader\n", yuv_fixup
);
3757 checkGLcall("glEnable(textype)");
3762 if (!shader
) shader
= gen_yuv_shader(device
, yuv_fixup
, textype
);
3765 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
3766 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3767 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
3768 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3769 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
3770 checkGLcall("glProgramLocalParameter4fvARB");
3776 static void arbfp_blit_unset(IWineD3DDevice
*iface
) {
3777 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3780 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
3781 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3782 glDisable(GL_TEXTURE_2D
);
3783 checkGLcall("glDisable(GL_TEXTURE_2D)");
3784 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP
)) {
3785 glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
3786 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3788 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3789 glDisable(GL_TEXTURE_RECTANGLE_ARB
);
3790 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3795 static BOOL
arbfp_blit_color_fixup_supported(struct color_fixup_desc fixup
)
3797 enum yuv_fixup yuv_fixup
;
3799 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
3801 TRACE("Checking support for fixup:\n");
3802 dump_color_fixup_desc(fixup
);
3805 if (is_identity_fixup(fixup
))
3811 /* We only support YUV conversions. */
3812 if (!is_yuv_fixup(fixup
))
3814 TRACE("[FAILED]\n");
3818 yuv_fixup
= get_yuv_fixup(fixup
);
3821 case YUV_FIXUP_YUY2
:
3822 case YUV_FIXUP_UYVY
:
3823 case YUV_FIXUP_YV12
:
3828 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
3829 TRACE("[FAILED]\n");
3834 const struct blit_shader arbfp_blit
= {
3839 arbfp_blit_color_fixup_supported
,
3842 #undef GLINFO_LOCATION