2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
14 * This library is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU Lesser General Public
16 * License as published by the Free Software Foundation; either
17 * version 2.1 of the License, or (at your option) any later version.
19 * This library is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
22 * Lesser General Public License for more details.
24 * You should have received a copy of the GNU Lesser General Public
25 * License along with this library; if not, write to the Free Software
26 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
34 #include "wined3d_private.h"
36 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
41 #define GLINFO_LOCATION (*gl_info)
43 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
44 * so upload them above that
46 #define ARB_SHADER_PRIVCONST_BASE GL_LIMITS(vshader_constantsF)
47 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
49 /* ARB_program_shader private data */
50 struct shader_arb_priv
{
51 GLuint current_vprogram_id
;
52 GLuint current_fprogram_id
;
53 GLuint depth_blt_vprogram_id
;
54 GLuint depth_blt_fprogram_id
[tex_type_count
];
55 BOOL use_arbfp_fixed_func
;
56 struct hash_table_t
*fragment_shaders
;
59 /********************************************************
60 * ARB_[vertex/fragment]_program helper functions follow
61 ********************************************************/
64 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
65 * When constant_list == NULL, it will load all the constants.
67 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
68 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
70 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl
* This
, const WineD3D_GL_Info
*gl_info
,
71 GLuint target_type
, unsigned int max_constants
, const float *constants
, char *dirty_consts
)
73 local_constant
* lconst
;
77 if (TRACE_ON(d3d_shader
)) {
78 for(i
= 0; i
< max_constants
; i
++) {
79 if(!dirty_consts
[i
]) continue;
80 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
81 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
82 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
85 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
86 if(target_type
== GL_FRAGMENT_PROGRAM_ARB
&&
87 WINED3DSHADER_VERSION_MAJOR(This
->baseShader
.hex_version
) == 1) {
89 for(i
= 0; i
< max_constants
; i
++) {
90 if(!dirty_consts
[i
]) continue;
94 if(constants
[j
+ 0] > 1.0) lcl_const
[0] = 1.0;
95 else if(constants
[j
+ 0] < -1.0) lcl_const
[0] = -1.0;
96 else lcl_const
[0] = constants
[j
+ 0];
98 if(constants
[j
+ 1] > 1.0) lcl_const
[1] = 1.0;
99 else if(constants
[j
+ 1] < -1.0) lcl_const
[1] = -1.0;
100 else lcl_const
[1] = constants
[j
+ 1];
102 if(constants
[j
+ 2] > 1.0) lcl_const
[2] = 1.0;
103 else if(constants
[j
+ 2] < -1.0) lcl_const
[2] = -1.0;
104 else lcl_const
[2] = constants
[j
+ 2];
106 if(constants
[j
+ 3] > 1.0) lcl_const
[3] = 1.0;
107 else if(constants
[j
+ 3] < -1.0) lcl_const
[3] = -1.0;
108 else lcl_const
[3] = constants
[j
+ 3];
110 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
113 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS
)) {
114 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
115 * or just reloading *all* constants at once
117 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
119 for(i
= 0; i
< max_constants
; i
++) {
120 if(!dirty_consts
[i
]) continue;
122 /* Find the next block of dirty constants */
125 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
129 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
132 for(i
= 0; i
< max_constants
; i
++) {
133 if(dirty_consts
[i
]) {
135 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
140 checkGLcall("glProgramEnvParameter4fvARB()");
142 /* Load immediate constants */
143 if(This
->baseShader
.load_local_constsF
) {
144 if (TRACE_ON(d3d_shader
)) {
145 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
146 GLfloat
* values
= (GLfloat
*)lconst
->value
;
147 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
148 values
[0], values
[1], values
[2], values
[3]);
151 /* Immediate constants are clamped for 1.X shaders at loading times */
153 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
154 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
155 ret
= max(ret
, lconst
->idx
+ 1);
156 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
158 checkGLcall("glProgramEnvParameter4fvARB()");
159 return ret
; /* The loaded immediate constants need reloading for the next shader */
161 return 0; /* No constants are dirty now */
166 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
168 * We only support float constants in ARB at the moment, so don't
169 * worry about the Integers or Booleans
171 static void shader_arb_load_constants(
172 IWineD3DDevice
* device
,
174 char useVertexShader
) {
176 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
177 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
178 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
181 if (useVertexShader
) {
182 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
184 /* Load DirectX 9 float constants for vertex shader */
185 deviceImpl
->highest_dirty_vs_const
= shader_arb_load_constantsF(
186 vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
187 deviceImpl
->highest_dirty_vs_const
,
188 stateBlock
->vertexShaderConstantF
,
189 deviceImpl
->activeContext
->vshader_const_dirty
);
191 /* Upload the position fixup */
192 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, ARB_SHADER_PRIVCONST_POS
, deviceImpl
->posFixup
));
195 if (usePixelShader
) {
197 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
198 IWineD3DPixelShaderImpl
*psi
= (IWineD3DPixelShaderImpl
*) pshader
;
200 /* Load DirectX 9 float constants for pixel shader */
201 deviceImpl
->highest_dirty_ps_const
= shader_arb_load_constantsF(
202 pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
203 deviceImpl
->highest_dirty_ps_const
,
204 stateBlock
->pixelShaderConstantF
,
205 deviceImpl
->activeContext
->pshader_const_dirty
);
207 for(i
= 0; i
< psi
->numbumpenvmatconsts
; i
++) {
208 /* The state manager takes care that this function is always called if the bump env matrix changes
210 const float *data
= (const float *)&stateBlock
->textureState
[(int) psi
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
211 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->bumpenvmatconst
[i
].const_num
, data
));
212 deviceImpl
->activeContext
->pshader_const_dirty
[psi
->bumpenvmatconst
[i
].const_num
] = 1;
214 if(psi
->luminanceconst
[i
].const_num
!= -1) {
215 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
216 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
217 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
218 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
220 const float *scale
= (const float *)&stateBlock
->textureState
[(int) psi
->luminanceconst
[i
].texunit
][WINED3DTSS_BUMPENVLSCALE
];
221 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->luminanceconst
[i
].const_num
, scale
));
222 deviceImpl
->activeContext
->pshader_const_dirty
[psi
->luminanceconst
[i
].const_num
] = 1;
228 /* Generate the variable & register declarations for the ARB_vertex_program output target */
229 static void shader_generate_arb_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
230 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
)
232 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
233 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
235 char pshader
= shader_is_pshader_version(This
->baseShader
.hex_version
);
236 unsigned max_constantsF
= min(This
->baseShader
.limits
.constant_float
,
237 (pshader
? GL_LIMITS(pshader_constantsF
) : GL_LIMITS(vshader_constantsF
)));
238 UINT extra_constants_needed
= 0;
239 const local_constant
*lconst
;
241 /* Temporary Output register */
242 shader_addline(buffer
, "TEMP TMP_OUT;\n");
244 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
245 if (reg_maps
->temporary
[i
])
246 shader_addline(buffer
, "TEMP R%u;\n", i
);
249 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
250 if (reg_maps
->address
[i
])
251 shader_addline(buffer
, "ADDRESS A%d;\n", i
);
254 for(i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
255 if (reg_maps
->texcoord
[i
])
256 shader_addline(buffer
,"TEMP T%u;\n", i
);
259 /* Texture coordinate registers must be pre-loaded */
260 for (i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
261 if (reg_maps
->texcoord
[i
])
262 shader_addline(buffer
, "MOV T%u, fragment.texcoord[%u];\n", i
, i
);
265 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
266 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) This
;
267 if(!reg_maps
->bumpmat
[i
]) continue;
269 cur
= ps
->numbumpenvmatconsts
;
270 ps
->bumpenvmatconst
[cur
].const_num
= -1;
271 ps
->bumpenvmatconst
[cur
].texunit
= i
;
272 ps
->luminanceconst
[cur
].const_num
= -1;
273 ps
->luminanceconst
[cur
].texunit
= i
;
275 /* If the shader does not use all available constants, use the next free constant to load the bump mapping environment matrix from
276 * the stateblock into the shader. If no constant is available don't load, texbem will then just sample the texture without applying
279 if(max_constantsF
+ extra_constants_needed
< GL_LIMITS(pshader_constantsF
)) {
280 ps
->bumpenvmatconst
[cur
].const_num
= max_constantsF
+ extra_constants_needed
;
281 shader_addline(buffer
, "PARAM bumpenvmat%d = program.env[%d];\n",
282 i
, ps
->bumpenvmatconst
[cur
].const_num
);
283 extra_constants_needed
++;
285 if(reg_maps
->luminanceparams
&& max_constantsF
+ extra_constants_needed
< GL_LIMITS(pshader_constantsF
)) {
286 ((IWineD3DPixelShaderImpl
*)This
)->luminanceconst
[cur
].const_num
= max_constantsF
+ extra_constants_needed
;
287 shader_addline(buffer
, "PARAM luminance%d = program.env[%d];\n",
288 i
, ps
->luminanceconst
[cur
].const_num
);
289 extra_constants_needed
++;
290 } else if(reg_maps
->luminanceparams
) {
291 FIXME("No free constant to load the luminance parameters\n");
294 FIXME("No free constant found to load environemnt bump mapping matrix into the shader. texbem instruction will not apply bump mapping\n");
297 ps
->numbumpenvmatconsts
= cur
+ 1;
300 if(device
->stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
] && pshader
) {
301 shader_addline(buffer
, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
302 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
303 shader_addline(buffer
, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
304 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
305 shader_addline(buffer
, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
306 srgb_pow
, srgb_pow
, srgb_pow
);
307 shader_addline(buffer
, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
308 srgb_mul_high
, srgb_mul_high
, srgb_mul_high
);
309 shader_addline(buffer
, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
310 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
313 /* Load local constants using the program-local space,
314 * this avoids reloading them each time the shader is used
316 if(!This
->baseShader
.load_local_constsF
) {
317 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
318 shader_addline(buffer
, "PARAM C%u = program.local[%u];\n", lconst
->idx
,
323 /* we use the array-based constants array if the local constants are marked for loading,
324 * because then we use indirect addressing, or when the local constant list is empty,
325 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
326 * local constants do not declare the loaded constants as an array because ARB compilers usually
327 * do not optimize unused constants away
329 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) {
330 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
331 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
332 max_constantsF
, max_constantsF
- 1);
334 for(i
= 0; i
< max_constantsF
; i
++) {
335 if(!shader_constant_is_local(This
, i
)) {
336 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
342 static const char * const shift_tab
[] = {
343 "dummy", /* 0 (none) */
344 "coefmul.x", /* 1 (x2) */
345 "coefmul.y", /* 2 (x4) */
346 "coefmul.z", /* 3 (x8) */
347 "coefmul.w", /* 4 (x16) */
348 "dummy", /* 5 (x32) */
349 "dummy", /* 6 (x64) */
350 "dummy", /* 7 (x128) */
351 "dummy", /* 8 (d256) */
352 "dummy", /* 9 (d128) */
353 "dummy", /* 10 (d64) */
354 "dummy", /* 11 (d32) */
355 "coefdiv.w", /* 12 (d16) */
356 "coefdiv.z", /* 13 (d8) */
357 "coefdiv.y", /* 14 (d4) */
358 "coefdiv.x" /* 15 (d2) */
361 static void shader_arb_get_write_mask(const SHADER_OPCODE_ARG
*arg
, const DWORD param
, char *write_mask
)
363 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
364 char *ptr
= write_mask
;
365 char vshader
= shader_is_vshader_version(This
->baseShader
.hex_version
);
367 if(vshader
&& shader_get_regtype(param
) == WINED3DSPR_ADDR
) {
370 } else if ((param
& WINED3DSP_WRITEMASK_ALL
) != WINED3DSP_WRITEMASK_ALL
) {
372 if (param
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
373 if (param
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
374 if (param
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
375 if (param
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
381 static void shader_arb_get_swizzle(const DWORD param
, BOOL fixup
, char *swizzle_str
) {
382 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
383 * but addressed as "rgba". To fix this we need to swap the register's x
384 * and z components. */
385 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
386 char *ptr
= swizzle_str
;
388 /* swizzle bits fields: wwzzyyxx */
389 DWORD swizzle
= (param
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
390 DWORD swizzle_x
= swizzle
& 0x03;
391 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
392 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
393 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
395 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
396 * generate a swizzle string. Unless we need to our own swizzling. */
397 if ((WINED3DSP_NOSWIZZLE
>> WINED3DSP_SWIZZLE_SHIFT
) != swizzle
|| fixup
) {
399 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
400 *ptr
++ = swizzle_chars
[swizzle_x
];
402 *ptr
++ = swizzle_chars
[swizzle_x
];
403 *ptr
++ = swizzle_chars
[swizzle_y
];
404 *ptr
++ = swizzle_chars
[swizzle_z
];
405 *ptr
++ = swizzle_chars
[swizzle_w
];
412 static void pshader_get_register_name(IWineD3DBaseShader
* iface
,
413 const DWORD param
, char* regstr
) {
415 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
416 DWORD regtype
= shader_get_regtype(param
);
417 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) iface
;
420 case WINED3DSPR_TEMP
:
421 sprintf(regstr
, "R%u", reg
);
423 case WINED3DSPR_INPUT
:
425 strcpy(regstr
, "fragment.color.primary");
427 strcpy(regstr
, "fragment.color.secondary");
430 case WINED3DSPR_CONST
:
431 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) {
432 sprintf(regstr
, "C[%u]", reg
);
434 sprintf(regstr
, "C%u", reg
);
437 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
438 sprintf(regstr
,"T%u", reg
);
440 case WINED3DSPR_COLOROUT
:
442 sprintf(regstr
, "TMP_COLOR");
444 /* TODO: See GL_ARB_draw_buffers */
445 FIXME("Unsupported write to render target %u\n", reg
);
446 sprintf(regstr
, "unsupported_register");
449 case WINED3DSPR_DEPTHOUT
:
450 sprintf(regstr
, "result.depth");
452 case WINED3DSPR_ATTROUT
:
453 sprintf(regstr
, "oD[%u]", reg
);
455 case WINED3DSPR_TEXCRDOUT
:
456 sprintf(regstr
, "oT[%u]", reg
);
459 FIXME("Unhandled register name Type(%d)\n", regtype
);
460 sprintf(regstr
, "unrecognized_register");
465 /* TODO: merge with pixel shader */
466 static void vshader_program_add_param(const SHADER_OPCODE_ARG
*arg
, const DWORD param
, BOOL is_input
, char *hwLine
)
468 IWineD3DVertexShaderImpl
* This
= (IWineD3DVertexShaderImpl
*) arg
->shader
;
470 /* oPos, oFog and oPts in D3D */
471 static const char * const hwrastout_reg_names
[] = { "TMP_OUT", "result.fogcoord", "result.pointsize" };
473 DWORD reg
= param
& WINED3DSP_REGNUM_MASK
;
474 DWORD regtype
= shader_get_regtype(param
);
476 BOOL is_color
= FALSE
;
478 if ((param
& WINED3DSP_SRCMOD_MASK
) == WINED3DSPSM_NEG
) {
479 strcat(hwLine
, " -");
485 case WINED3DSPR_TEMP
:
486 sprintf(tmpReg
, "R%u", reg
);
487 strcat(hwLine
, tmpReg
);
489 case WINED3DSPR_INPUT
:
491 if (vshader_input_is_color((IWineD3DVertexShader
*) This
, reg
))
494 sprintf(tmpReg
, "vertex.attrib[%u]", reg
);
495 strcat(hwLine
, tmpReg
);
497 case WINED3DSPR_CONST
:
498 if(param
& WINED3DSHADER_ADDRMODE_RELATIVE
) {
499 if(reg
>= This
->rel_offset
) {
500 sprintf(tmpReg
, "C[A0.x + %u]", reg
- This
->rel_offset
);
502 sprintf(tmpReg
, "C[A0.x - %u]", -reg
+ This
->rel_offset
);
505 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) {
506 sprintf(tmpReg
, "C[%u]", reg
);
508 sprintf(tmpReg
, "C%u", reg
);
511 strcat(hwLine
, tmpReg
);
513 case WINED3DSPR_ADDR
: /*case D3DSPR_TEXTURE:*/
514 sprintf(tmpReg
, "A%u", reg
);
515 strcat(hwLine
, tmpReg
);
517 case WINED3DSPR_RASTOUT
:
518 sprintf(tmpReg
, "%s", hwrastout_reg_names
[reg
]);
519 strcat(hwLine
, tmpReg
);
521 case WINED3DSPR_ATTROUT
:
523 strcat(hwLine
, "result.color.primary");
525 strcat(hwLine
, "result.color.secondary");
528 case WINED3DSPR_TEXCRDOUT
:
529 sprintf(tmpReg
, "result.texcoord[%u]", reg
);
530 strcat(hwLine
, tmpReg
);
533 FIXME("Unknown reg type %d %d\n", regtype
, reg
);
534 strcat(hwLine
, "unrecognized_register");
540 shader_arb_get_write_mask(arg
, param
, write_mask
);
541 strcat(hwLine
, write_mask
);
544 shader_arb_get_swizzle(param
, is_color
, swizzle
);
545 strcat(hwLine
, swizzle
);
549 static void shader_hw_sample(const SHADER_OPCODE_ARG
*arg
, DWORD sampler_idx
, const char *dst_str
,
550 const char *coord_reg
, BOOL projected
, BOOL bias
)
552 SHADER_BUFFER
* buffer
= arg
->buffer
;
553 DWORD sampler_type
= arg
->reg_maps
->samplers
[sampler_idx
] & WINED3DSP_TEXTURETYPE_MASK
;
554 const char *tex_type
;
556 switch(sampler_type
) {
563 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
564 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
565 if(device
->stateBlock
->textures
[sampler_idx
] &&
566 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
574 case WINED3DSTT_VOLUME
:
578 case WINED3DSTT_CUBE
:
583 ERR("Unexpected texture type %d\n", sampler_type
);
588 /* Shouldn't be possible, but let's check for it */
589 if(projected
) FIXME("Biased and Projected texture sampling\n");
590 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
591 shader_addline(buffer
, "TXB %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
592 } else if (projected
) {
593 shader_addline(buffer
, "TXP %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
595 shader_addline(buffer
, "TEX %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
599 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
601 switch(channel_source
)
603 case CHANNEL_SOURCE_ZERO
: return "0";
604 case CHANNEL_SOURCE_ONE
: return "1";
605 case CHANNEL_SOURCE_X
: return "x";
606 case CHANNEL_SOURCE_Y
: return "y";
607 case CHANNEL_SOURCE_Z
: return "z";
608 case CHANNEL_SOURCE_W
: return "w";
610 FIXME("Unhandled channel source %#x\n", channel_source
);
615 static void gen_color_correction(SHADER_BUFFER
*buffer
, const char *reg
, DWORD dst_mask
,
616 const char *one
, const char *two
, struct color_fixup_desc fixup
)
620 if (is_yuv_fixup(fixup
))
622 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
623 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
628 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
629 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
630 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
631 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
636 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
637 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
638 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
642 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
643 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
644 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
645 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
651 char *ptr
= reg_mask
;
653 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
656 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
657 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
658 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
659 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
663 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
667 static void shader_arb_color_correction(const struct SHADER_OPCODE_ARG
* arg
, struct color_fixup_desc fixup
)
670 pshader_get_register_name(arg
->shader
, arg
->dst
, reg
);
671 gen_color_correction(arg
->buffer
, reg
, arg
->dst
& WINED3DSP_WRITEMASK_ALL
, "one", "coefmul.x", fixup
);
674 static void pshader_gen_input_modifier_line (
675 IWineD3DBaseShader
*iface
,
676 SHADER_BUFFER
* buffer
,
681 /* Generate a line that does the input modifier computation and return the input register to use */
686 /* Assume a new line will be added */
689 /* Get register name */
690 pshader_get_register_name(iface
, instr
, regstr
);
691 shader_arb_get_swizzle(instr
, FALSE
, swzstr
);
693 switch (instr
& WINED3DSP_SRCMOD_MASK
) {
694 case WINED3DSPSM_NONE
:
695 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
698 case WINED3DSPSM_NEG
:
699 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
702 case WINED3DSPSM_BIAS
:
703 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
705 case WINED3DSPSM_BIASNEG
:
706 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
708 case WINED3DSPSM_SIGN
:
709 shader_addline(buffer
, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg
, regstr
);
711 case WINED3DSPSM_SIGNNEG
:
712 shader_addline(buffer
, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg
, regstr
);
714 case WINED3DSPSM_COMP
:
715 shader_addline(buffer
, "SUB T%c, one.x, %s;\n", 'A' + tmpreg
, regstr
);
718 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
720 case WINED3DSPSM_X2NEG
:
721 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
724 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
725 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
728 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
729 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
732 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
736 /* Return modified or original register, with swizzle */
738 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
741 static inline void pshader_gen_output_modifier_line(SHADER_BUFFER
*buffer
, int saturate
, const char *write_mask
,
742 int shift
, const char *regstr
)
744 /* Generate a line that does the output modifier computation */
745 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", saturate
? "_SAT" : "",
746 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
749 static void pshader_hw_bem(const SHADER_OPCODE_ARG
*arg
)
751 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
753 SHADER_BUFFER
* buffer
= arg
->buffer
;
755 char src_name
[2][50];
757 DWORD sampler_code
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
758 BOOL has_bumpmat
= FALSE
;
761 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
762 if(This
->bumpenvmatconst
[i
].const_num
!= -1 && This
->bumpenvmatconst
[i
].texunit
== sampler_code
) {
768 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
769 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
770 strcat(dst_name
, dst_wmask
);
772 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
[0]);
773 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[1], 1, src_name
[1]);
776 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
777 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
778 shader_addline(buffer
, "DP3 TMP.r, TMP2, %s;\n", src_name
[1]);
779 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
780 shader_addline(buffer
, "DP3 TMP.g, TMP2, %s;\n", src_name
[1]);
782 shader_addline(buffer
, "ADD %s, %s, TMP;\n", dst_name
, src_name
[0]);
784 shader_addline(buffer
, "MOV %s, %s;\n", dst_name
, src_name
[0]);
788 static void pshader_hw_cnd(const SHADER_OPCODE_ARG
*arg
)
790 IWineD3DBaseShaderImpl
* shader
= (IWineD3DBaseShaderImpl
*) arg
->shader
;
791 SHADER_BUFFER
* buffer
= arg
->buffer
;
794 char src_name
[3][50];
795 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
796 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
798 /* FIXME: support output modifiers */
800 /* Handle output register */
801 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
802 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
804 /* Generate input register names (with modifiers) */
805 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
[0]);
806 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[1], 1, src_name
[1]);
807 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[2], 2, src_name
[2]);
809 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
810 if (shader
->baseShader
.hex_version
<= WINED3DPS_VERSION(1, 3) &&
811 arg
->opcode_token
& WINED3DSI_COISSUE
) {
812 shader_addline(buffer
, "MOV%s %s%s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[1]);
814 shader_addline(buffer
, "ADD TMP, -%s, coefdiv.x;\n", src_name
[0]);
815 shader_addline(buffer
, "CMP%s %s%s, TMP, %s, %s;\n",
816 sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[1], src_name
[2]);
819 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
822 static void pshader_hw_cmp(const SHADER_OPCODE_ARG
*arg
)
824 SHADER_BUFFER
* buffer
= arg
->buffer
;
827 char src_name
[3][50];
828 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
829 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
831 /* FIXME: support output modifiers */
833 /* Handle output register */
834 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
835 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
837 /* Generate input register names (with modifiers) */
838 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
[0]);
839 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[1], 1, src_name
[1]);
840 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[2], 2, src_name
[2]);
842 shader_addline(buffer
, "CMP%s %s%s, %s, %s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
843 src_name
[0], src_name
[2], src_name
[1]);
846 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
849 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
850 * dst = dot2(src0, src1) + src2 */
851 static void pshader_hw_dp2add(const SHADER_OPCODE_ARG
*arg
)
853 SHADER_BUFFER
* buffer
= arg
->buffer
;
856 char src_name
[3][50];
857 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
858 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
860 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
861 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
863 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
[0]);
864 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[1], 1, src_name
[1]);
865 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[2], 2, src_name
[2]);
867 /* Emulate a DP2 with a DP3 and 0.0 */
868 shader_addline(buffer
, "MOV TMP, %s;\n", src_name
[0]);
869 shader_addline(buffer
, "MOV TMP.z, 0.0;\n");
870 shader_addline(buffer
, "DP3 TMP2, TMP, %s;\n", src_name
[1]);
871 shader_addline(buffer
, "ADD%s %s%s, TMP2, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
, src_name
[2]);
874 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
877 /* Map the opcode 1-to-1 to the GL code */
878 static void shader_hw_map2gl(const SHADER_OPCODE_ARG
*arg
)
880 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
881 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
882 SHADER_BUFFER
* buffer
= arg
->buffer
;
883 DWORD dst
= arg
->dst
;
884 const DWORD
*src
= arg
->src
;
888 if (shader_is_pshader_version(shader
->baseShader
.hex_version
))
890 /* Output token related */
891 char output_rname
[256];
892 char output_wmask
[20];
893 char operands
[4][100];
894 BOOL saturate
= FALSE
;
895 BOOL centroid
= FALSE
;
896 BOOL partialprecision
= FALSE
;
897 const char *modifier
;
900 if (!curOpcode
->num_params
)
902 ERR("Opcode \"%s\" has no parameters\n", curOpcode
->name
);
906 /* Process modifiers */
907 if (dst
& WINED3DSP_DSTMOD_MASK
)
909 DWORD mask
= dst
& WINED3DSP_DSTMOD_MASK
;
911 saturate
= mask
& WINED3DSPDM_SATURATE
;
912 centroid
= mask
& WINED3DSPDM_MSAMPCENTROID
;
913 partialprecision
= mask
& WINED3DSPDM_PARTIALPRECISION
;
914 mask
&= ~(WINED3DSPDM_MSAMPCENTROID
| WINED3DSPDM_PARTIALPRECISION
| WINED3DSPDM_SATURATE
);
916 FIXME("Unrecognized modifier(%#x)\n", mask
>> WINED3DSP_DSTMOD_SHIFT
);
919 FIXME("Unhandled modifier(%#x)\n", mask
>> WINED3DSP_DSTMOD_SHIFT
);
921 shift
= (dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
922 modifier
= (saturate
&& !shift
) ? "_SAT" : "";
924 /* Generate input register names (with modifiers) */
925 for (i
= 1; i
< curOpcode
->num_params
; ++i
)
926 pshader_gen_input_modifier_line(arg
->shader
, buffer
, src
[i
-1], i
-1, operands
[i
]);
928 /* Handle output register */
929 pshader_get_register_name(arg
->shader
, dst
, output_rname
);
930 strcpy(operands
[0], output_rname
);
931 shader_arb_get_write_mask(arg
, dst
, output_wmask
);
932 strcat(operands
[0], output_wmask
);
935 strcat(arguments
, operands
[0]);
936 for (i
= 1; i
< curOpcode
->num_params
; i
++)
938 strcat(arguments
, ", ");
939 strcat(arguments
, operands
[i
]);
941 shader_addline(buffer
, "%s%s %s;\n", curOpcode
->glname
, modifier
, arguments
);
943 /* A shift requires another line. */
944 if (shift
) pshader_gen_output_modifier_line(buffer
, saturate
, output_wmask
, shift
, output_rname
);
946 /* Note that vshader_program_add_param() adds spaces. */
949 if (curOpcode
->num_params
> 0)
951 vshader_program_add_param(arg
, dst
, FALSE
, arguments
);
952 for (i
= 1; i
< curOpcode
->num_params
; ++i
)
954 strcat(arguments
, ",");
955 vshader_program_add_param(arg
, src
[i
-1], TRUE
, arguments
);
958 shader_addline(buffer
, "%s%s;\n", curOpcode
->glname
, arguments
);
962 static void shader_hw_mov(const SHADER_OPCODE_ARG
*arg
)
964 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
966 if ((WINED3DSHADER_VERSION_MAJOR(shader
->baseShader
.hex_version
) == 1
967 && !shader_is_pshader_version(shader
->baseShader
.hex_version
)
968 && shader_get_regtype(arg
->dst
) == WINED3DSPR_ADDR
)
969 || arg
->opcode
->opcode
== WINED3DSIO_MOVA
)
971 SHADER_BUFFER
*buffer
= arg
->buffer
;
972 char src0_param
[256];
974 if (arg
->opcode
->opcode
== WINED3DSIO_MOVA
)
975 FIXME("mova should round\n");
977 src0_param
[0] = '\0';
978 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
)
980 vshader_program_add_param(arg
, arg
->src
[0], TRUE
, src0_param
);
981 shader_addline(buffer
, "ADD TMP.x, %s, helper_const.z;\n", src0_param
);
982 shader_addline(buffer
, "ARL A0.x, TMP.x;\n");
986 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
987 * with more than one component. Thus replicate the first source argument over all
988 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc)
990 DWORD parm
= arg
->src
[0] & ~(WINED3DVS_SWIZZLE_MASK
);
991 if((arg
->src
[0] & WINED3DVS_X_W
) == WINED3DVS_X_W
)
992 parm
|= WINED3DVS_X_W
| WINED3DVS_Y_W
| WINED3DVS_Z_W
| WINED3DVS_W_W
;
993 else if((arg
->src
[0] & WINED3DVS_X_Z
) == WINED3DVS_X_Z
)
994 parm
|= WINED3DVS_X_Z
| WINED3DVS_Y_Z
| WINED3DVS_Z_Z
| WINED3DVS_W_Z
;
995 else if((arg
->src
[0] & WINED3DVS_X_Y
) == WINED3DVS_X_Y
)
996 parm
|= WINED3DVS_X_Y
| WINED3DVS_Y_Y
| WINED3DVS_Z_Y
| WINED3DVS_W_Y
;
997 else if((arg
->src
[0] & WINED3DVS_X_X
) == WINED3DVS_X_X
)
998 parm
|= WINED3DVS_X_X
| WINED3DVS_Y_X
| WINED3DVS_Z_X
| WINED3DVS_W_X
;
999 vshader_program_add_param(arg
, parm
, TRUE
, src0_param
);
1000 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1005 shader_hw_map2gl(arg
);
1009 static void pshader_hw_texkill(const SHADER_OPCODE_ARG
*arg
)
1011 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1012 DWORD hex_version
= This
->baseShader
.hex_version
;
1013 SHADER_BUFFER
* buffer
= arg
->buffer
;
1016 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1017 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1019 pshader_get_register_name(arg
->shader
, arg
->dst
, reg_dest
);
1021 if(hex_version
>= WINED3DPS_VERSION(2,0)) {
1022 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1023 shader_addline(buffer
, "KIL %s;\n", reg_dest
);
1025 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1026 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1028 shader_addline(buffer
, "MOV TMP, %s;\n", reg_dest
);
1029 shader_addline(buffer
, "MOV TMP.w, one.w;\n");
1030 shader_addline(buffer
, "KIL TMP;\n");
1034 static void pshader_hw_tex(const SHADER_OPCODE_ARG
*arg
)
1036 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1037 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1039 DWORD dst
= arg
->dst
;
1040 const DWORD
*src
= arg
->src
;
1041 SHADER_BUFFER
* buffer
= arg
->buffer
;
1042 DWORD hex_version
= This
->baseShader
.hex_version
;
1043 BOOL projected
= FALSE
, bias
= FALSE
;
1047 DWORD reg_dest_code
;
1048 DWORD reg_sampler_code
;
1050 /* All versions have a destination register */
1051 reg_dest_code
= dst
& WINED3DSP_REGNUM_MASK
;
1052 pshader_get_register_name(arg
->shader
, dst
, reg_dest
);
1054 /* 1.0-1.3: Use destination register as coordinate source.
1055 1.4+: Use provided coordinate source register. */
1056 if (hex_version
< WINED3DPS_VERSION(1,4))
1057 strcpy(reg_coord
, reg_dest
);
1059 pshader_gen_input_modifier_line(arg
->shader
, buffer
, src
[0], 0, reg_coord
);
1061 /* 1.0-1.4: Use destination register number as texture code.
1062 2.0+: Use provided sampler number as texure code. */
1063 if (hex_version
< WINED3DPS_VERSION(2,0))
1064 reg_sampler_code
= reg_dest_code
;
1066 reg_sampler_code
= src
[1] & WINED3DSP_REGNUM_MASK
;
1069 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1070 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1071 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1073 if(hex_version
< WINED3DPS_VERSION(1,4)) {
1075 if(reg_sampler_code
< MAX_TEXTURES
) {
1076 flags
= deviceImpl
->stateBlock
->textureState
[reg_sampler_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1078 if (flags
& WINED3DTTFF_PROJECTED
) {
1081 } else if(hex_version
< WINED3DPS_VERSION(2,0)) {
1082 DWORD src_mod
= arg
->src
[0] & WINED3DSP_SRCMOD_MASK
;
1083 if (src_mod
== WINED3DSPSM_DZ
) {
1085 } else if(src_mod
== WINED3DSPSM_DW
) {
1089 if(arg
->opcode_token
& WINED3DSI_TEXLD_PROJECT
) {
1092 if(arg
->opcode_token
& WINED3DSI_TEXLD_BIAS
) {
1096 shader_hw_sample(arg
, reg_sampler_code
, reg_dest
, reg_coord
, projected
, bias
);
1099 static void pshader_hw_texcoord(const SHADER_OPCODE_ARG
*arg
)
1101 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1102 DWORD dst
= arg
->dst
;
1103 SHADER_BUFFER
* buffer
= arg
->buffer
;
1104 DWORD hex_version
= This
->baseShader
.hex_version
;
1107 shader_arb_get_write_mask(arg
, dst
, tmp
);
1108 if (hex_version
!= WINED3DPS_VERSION(1,4)) {
1109 DWORD reg
= dst
& WINED3DSP_REGNUM_MASK
;
1110 shader_addline(buffer
, "MOV_SAT T%u%s, fragment.texcoord[%u];\n", reg
, tmp
, reg
);
1112 DWORD reg1
= dst
& WINED3DSP_REGNUM_MASK
;
1115 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, reg_src
);
1116 shader_addline(buffer
, "MOV R%u%s, %s;\n", reg1
, tmp
, reg_src
);
1120 static void pshader_hw_texreg2ar(const SHADER_OPCODE_ARG
*arg
)
1122 SHADER_BUFFER
* buffer
= arg
->buffer
;
1123 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1124 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1127 DWORD reg1
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1131 sprintf(dst_str
, "T%u", reg1
);
1132 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_str
);
1133 shader_addline(buffer
, "MOV TMP.r, %s.a;\n", src_str
);
1134 shader_addline(buffer
, "MOV TMP.g, %s.r;\n", src_str
);
1135 flags
= reg1
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg1
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1136 shader_hw_sample(arg
, reg1
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1139 static void pshader_hw_texreg2gb(const SHADER_OPCODE_ARG
*arg
)
1141 SHADER_BUFFER
* buffer
= arg
->buffer
;
1143 DWORD reg1
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1147 sprintf(dst_str
, "T%u", reg1
);
1148 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_str
);
1149 shader_addline(buffer
, "MOV TMP.r, %s.g;\n", src_str
);
1150 shader_addline(buffer
, "MOV TMP.g, %s.b;\n", src_str
);
1151 shader_hw_sample(arg
, reg1
, dst_str
, "TMP", FALSE
, FALSE
);
1154 static void pshader_hw_texreg2rgb(const SHADER_OPCODE_ARG
*arg
)
1156 SHADER_BUFFER
* buffer
= arg
->buffer
;
1157 DWORD reg1
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1161 sprintf(dst_str
, "T%u", reg1
);
1162 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_str
);
1163 shader_hw_sample(arg
, reg1
, dst_str
, src_str
, FALSE
, FALSE
);
1166 static void pshader_hw_texbem(const SHADER_OPCODE_ARG
*arg
)
1168 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1169 BOOL has_bumpmat
= FALSE
;
1170 BOOL has_luminance
= FALSE
;
1173 DWORD dst
= arg
->dst
;
1174 DWORD src
= arg
->src
[0] & WINED3DSP_REGNUM_MASK
;
1175 SHADER_BUFFER
* buffer
= arg
->buffer
;
1178 DWORD reg_dest_code
;
1180 /* All versions have a destination register */
1181 reg_dest_code
= dst
& WINED3DSP_REGNUM_MASK
;
1182 /* Can directly use the name because texbem is only valid for <= 1.3 shaders */
1183 pshader_get_register_name(arg
->shader
, dst
, reg_coord
);
1185 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
1186 if(This
->bumpenvmatconst
[i
].const_num
!= -1 && reg_dest_code
== This
->bumpenvmatconst
[i
].texunit
) {
1191 for(i
= 0; i
< This
->numbumpenvmatconsts
; i
++) {
1192 if(This
->luminanceconst
[i
].const_num
!= -1 && reg_dest_code
== This
->luminanceconst
[i
].texunit
) {
1193 has_luminance
= TRUE
;
1199 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed */
1201 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
1202 shader_addline(buffer
, "DP3 TMP.r, TMP2, T%u;\n", src
);
1203 shader_addline(buffer
, "SWZ TMP2, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
1204 shader_addline(buffer
, "DP3 TMP.g, TMP2, T%u;\n", src
);
1206 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1207 * so we can't let the GL handle this.
1209 if (((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
]
1210 & WINED3DTTFF_PROJECTED
) {
1211 shader_addline(buffer
, "RCP TMP2.a, %s.a;\n", reg_coord
);
1212 shader_addline(buffer
, "MUL TMP2.rg, %s, TMP2.a;\n", reg_coord
);
1213 shader_addline(buffer
, "ADD TMP.rg, TMP, TMP2;\n");
1215 shader_addline(buffer
, "ADD TMP.rg, TMP, %s;\n", reg_coord
);
1218 shader_hw_sample(arg
, reg_dest_code
, reg_coord
, "TMP", FALSE
, FALSE
);
1220 if(arg
->opcode
->opcode
== WINED3DSIO_TEXBEML
&& has_luminance
) {
1221 shader_addline(buffer
, "MAD TMP, T%u.z, luminance%d.x, luminance%d.y;\n",
1222 src
, reg_dest_code
, reg_dest_code
);
1223 shader_addline(buffer
, "MUL %s, %s, TMP;\n", reg_coord
, reg_coord
);
1228 if(reg_dest_code
< MAX_TEXTURES
) {
1229 tf
= ((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1233 /* Without a bump matrix loaded, just sample with the unmodified coordinates */
1234 shader_hw_sample(arg
, reg_dest_code
, reg_coord
, reg_coord
, tf
& WINED3DTTFF_PROJECTED
, FALSE
);
1238 static void pshader_hw_texm3x2pad(const SHADER_OPCODE_ARG
*arg
)
1240 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1241 SHADER_BUFFER
* buffer
= arg
->buffer
;
1244 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1245 shader_addline(buffer
, "DP3 TMP.x, T%u, %s;\n", reg
, src0_name
);
1248 static void pshader_hw_texm3x2tex(const SHADER_OPCODE_ARG
*arg
)
1250 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1251 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1253 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1254 SHADER_BUFFER
* buffer
= arg
->buffer
;
1258 sprintf(dst_str
, "T%u", reg
);
1259 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1260 shader_addline(buffer
, "DP3 TMP.y, T%u, %s;\n", reg
, src0_name
);
1261 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1262 shader_hw_sample(arg
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1265 static void pshader_hw_texm3x3pad(const SHADER_OPCODE_ARG
*arg
)
1267 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1268 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1269 SHADER_BUFFER
* buffer
= arg
->buffer
;
1270 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1273 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1274 shader_addline(buffer
, "DP3 TMP.%c, T%u, %s;\n", 'x' + current_state
->current_row
, reg
, src0_name
);
1275 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
1278 static void pshader_hw_texm3x3tex(const SHADER_OPCODE_ARG
*arg
)
1280 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1281 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1283 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1284 SHADER_BUFFER
* buffer
= arg
->buffer
;
1285 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1289 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1290 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1292 /* Sample the texture using the calculated coordinates */
1293 sprintf(dst_str
, "T%u", reg
);
1294 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1295 shader_hw_sample(arg
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1296 current_state
->current_row
= 0;
1299 static void pshader_hw_texm3x3vspec(const SHADER_OPCODE_ARG
*arg
)
1301 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1302 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1304 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1305 SHADER_BUFFER
* buffer
= arg
->buffer
;
1306 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1310 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1311 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1313 /* Construct the eye-ray vector from w coordinates */
1314 shader_addline(buffer
, "MOV TMP2.x, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[0]);
1315 shader_addline(buffer
, "MOV TMP2.y, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[1]);
1316 shader_addline(buffer
, "MOV TMP2.z, fragment.texcoord[%u].w;\n", reg
);
1318 /* Calculate reflection vector
1320 shader_addline(buffer
, "DP3 TMP.w, TMP, TMP2;\n");
1321 /* The .w is ignored when sampling, so I can use TMP2.w to calculate dot(N, N) */
1322 shader_addline(buffer
, "DP3 TMP2.w, TMP, TMP;\n");
1323 shader_addline(buffer
, "RCP TMP2.w, TMP2.w;\n");
1324 shader_addline(buffer
, "MUL TMP.w, TMP.w, TMP2.w;\n");
1325 shader_addline(buffer
, "MUL TMP, TMP.w, TMP;\n");
1326 shader_addline(buffer
, "MAD TMP, coefmul.x, TMP, -TMP2;\n");
1328 /* Sample the texture using the calculated coordinates */
1329 sprintf(dst_str
, "T%u", reg
);
1330 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1331 shader_hw_sample(arg
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1332 current_state
->current_row
= 0;
1335 static void pshader_hw_texm3x3spec(const SHADER_OPCODE_ARG
*arg
)
1337 IWineD3DPixelShaderImpl
* This
= (IWineD3DPixelShaderImpl
*) arg
->shader
;
1338 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1340 DWORD reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1341 DWORD reg3
= arg
->src
[1] & WINED3DSP_REGNUM_MASK
;
1342 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1343 SHADER_BUFFER
* buffer
= arg
->buffer
;
1347 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0_name
);
1348 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", reg
, src0_name
);
1350 /* Calculate reflection vector.
1353 * TMP.xyz = 2 * --------- * N - E
1356 * Which normalizes the normal vector
1358 shader_addline(buffer
, "DP3 TMP.w, TMP, C[%u];\n", reg3
);
1359 shader_addline(buffer
, "DP3 TMP2.w, TMP, TMP;\n");
1360 shader_addline(buffer
, "RCP TMP2.w, TMP2.w;\n");
1361 shader_addline(buffer
, "MUL TMP.w, TMP.w, TMP2.w;\n");
1362 shader_addline(buffer
, "MUL TMP, TMP.w, TMP;\n");
1363 shader_addline(buffer
, "MAD TMP, coefmul.x, TMP, -C[%u];\n", reg3
);
1365 /* Sample the texture using the calculated coordinates */
1366 sprintf(dst_str
, "T%u", reg
);
1367 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1368 shader_hw_sample(arg
, reg
, dst_str
, "TMP", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1369 current_state
->current_row
= 0;
1372 static void pshader_hw_texdepth(const SHADER_OPCODE_ARG
*arg
)
1374 SHADER_BUFFER
* buffer
= arg
->buffer
;
1377 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1378 * which is essentially an input, is the destination register because it is the first
1379 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1382 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1384 /* According to the msdn, the source register(must be r5) is unusable after
1385 * the texdepth instruction, so we're free to modify it
1387 shader_addline(buffer
, "MIN %s.g, %s.g, one.g;\n", dst_name
, dst_name
);
1389 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1390 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1391 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1393 shader_addline(buffer
, "RCP %s.g, %s.g;\n", dst_name
, dst_name
);
1394 shader_addline(buffer
, "MUL TMP.x, %s.r, %s.g;\n", dst_name
, dst_name
);
1395 shader_addline(buffer
, "MIN TMP.x, TMP.x, one.r;\n");
1396 shader_addline(buffer
, "MAX result.depth, TMP.x, 0.0;\n");
1399 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1400 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1401 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1402 static void pshader_hw_texdp3tex(const SHADER_OPCODE_ARG
*arg
)
1404 SHADER_BUFFER
* buffer
= arg
->buffer
;
1405 DWORD sampler_idx
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1409 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0
);
1410 shader_addline(buffer
, "MOV TMP, 0.0;\n");
1411 shader_addline(buffer
, "DP3 TMP.x, T%u, %s;\n", sampler_idx
, src0
);
1413 sprintf(dst_str
, "T%u", sampler_idx
);
1414 shader_hw_sample(arg
, sampler_idx
, dst_str
, "TMP", FALSE
/* Only one coord, can't be projected */, FALSE
);
1417 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1418 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1419 static void pshader_hw_texdp3(const SHADER_OPCODE_ARG
*arg
)
1424 DWORD dstreg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1425 SHADER_BUFFER
* buffer
= arg
->buffer
;
1427 /* Handle output register */
1428 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_str
);
1429 shader_arb_get_write_mask(arg
, arg
->dst
, dst_mask
);
1431 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0
);
1432 shader_addline(buffer
, "DP3 %s%s, T%u, %s;\n", dst_str
, dst_mask
, dstreg
, src0
);
1434 /* TODO: Handle output modifiers */
1437 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1438 * Perform the 3rd row of a 3x3 matrix multiply */
1439 static void pshader_hw_texm3x3(const SHADER_OPCODE_ARG
*arg
)
1441 SHADER_BUFFER
* buffer
= arg
->buffer
;
1445 DWORD dst_reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1447 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_str
);
1448 shader_arb_get_write_mask(arg
, arg
->dst
, dst_mask
);
1450 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0
);
1451 shader_addline(buffer
, "DP3 TMP.z, T%u, %s;\n", dst_reg
, src0
);
1452 shader_addline(buffer
, "MOV %s%s, TMP;\n", dst_str
, dst_mask
);
1454 /* TODO: Handle output modifiers */
1457 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1458 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1459 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1460 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1462 static void pshader_hw_texm3x2depth(const SHADER_OPCODE_ARG
*arg
)
1464 SHADER_BUFFER
* buffer
= arg
->buffer
;
1465 DWORD dst_reg
= arg
->dst
& WINED3DSP_REGNUM_MASK
;
1468 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src0
);
1469 shader_addline(buffer
, "DP3 TMP.y, T%u, %s;\n", dst_reg
, src0
);
1471 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1472 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1473 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1475 shader_addline(buffer
, "RCP TMP.y, TMP.y;\n");
1476 shader_addline(buffer
, "MUL TMP.x, TMP.x, TMP.y;\n");
1477 shader_addline(buffer
, "MIN TMP.x, TMP.x, one.r;\n");
1478 shader_addline(buffer
, "MAX result.depth, TMP.x, 0.0;\n");
1481 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1482 Vertex/Pixel shaders to ARB_vertex_program codes */
1483 static void shader_hw_mnxn(const SHADER_OPCODE_ARG
*arg
)
1486 int nComponents
= 0;
1487 SHADER_OPCODE_ARG tmpArg
;
1488 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)arg
->shader
;
1489 const SHADER_OPCODE
*opcode_table
= shader
->baseShader
.shader_ins
;
1490 DWORD shader_version
= shader
->baseShader
.hex_version
;
1492 memset(&tmpArg
, 0, sizeof(SHADER_OPCODE_ARG
));
1494 /* Set constants for the temporary argument */
1495 tmpArg
.shader
= arg
->shader
;
1496 tmpArg
.buffer
= arg
->buffer
;
1497 tmpArg
.src
[0] = arg
->src
[0];
1498 tmpArg
.src_addr
[0] = arg
->src_addr
[0];
1499 tmpArg
.src_addr
[1] = arg
->src_addr
[1];
1500 tmpArg
.reg_maps
= arg
->reg_maps
;
1502 switch(arg
->opcode
->opcode
) {
1503 case WINED3DSIO_M4x4
:
1505 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
1507 case WINED3DSIO_M4x3
:
1509 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP4
);
1511 case WINED3DSIO_M3x4
:
1513 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1515 case WINED3DSIO_M3x3
:
1517 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1519 case WINED3DSIO_M3x2
:
1521 tmpArg
.opcode
= shader_get_opcode(opcode_table
, shader_version
, WINED3DSIO_DP3
);
1527 for (i
= 0; i
< nComponents
; i
++) {
1528 tmpArg
.dst
= ((arg
->dst
) & ~WINED3DSP_WRITEMASK_ALL
)|(WINED3DSP_WRITEMASK_0
<<i
);
1529 tmpArg
.src
[1] = arg
->src
[1]+i
;
1530 shader_hw_map2gl(&tmpArg
);
1534 static void vshader_hw_rsq_rcp(const SHADER_OPCODE_ARG
*arg
)
1536 CONST SHADER_OPCODE
* curOpcode
= arg
->opcode
;
1537 SHADER_BUFFER
* buffer
= arg
->buffer
;
1538 DWORD dst
= arg
->dst
;
1539 DWORD src
= arg
->src
[0];
1540 DWORD swizzle
= (src
& WINED3DSP_SWIZZLE_MASK
) >> WINED3DSP_SWIZZLE_SHIFT
;
1544 strcpy(tmpLine
, curOpcode
->glname
); /* Opcode */
1545 vshader_program_add_param(arg
, dst
, FALSE
, tmpLine
); /* Destination */
1546 strcat(tmpLine
, ",");
1547 vshader_program_add_param(arg
, src
, TRUE
, tmpLine
);
1548 if ((WINED3DSP_NOSWIZZLE
>> WINED3DSP_SWIZZLE_SHIFT
) == swizzle
) {
1549 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1552 strcat(tmpLine
, ".w");
1555 shader_addline(buffer
, "%s;\n", tmpLine
);
1558 static void shader_hw_nrm(const SHADER_OPCODE_ARG
*arg
)
1560 SHADER_BUFFER
* buffer
= arg
->buffer
;
1564 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1565 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
1567 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1568 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
1570 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
);
1571 shader_addline(buffer
, "DP3 TMP, %s, %s;\n", src_name
, src_name
);
1572 shader_addline(buffer
, "RSQ TMP, TMP.x;\n");
1573 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1574 shader_addline(buffer
, "MUL%s %s%s, %s, TMP;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
1578 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1581 static void shader_hw_sincos(const SHADER_OPCODE_ARG
*arg
)
1583 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1584 * must contain fixed constants. So we need a separate function to filter those constants and
1587 SHADER_BUFFER
* buffer
= arg
->buffer
;
1591 DWORD shift
= (arg
->dst
& WINED3DSP_DSTSHIFT_MASK
) >> WINED3DSP_DSTSHIFT_SHIFT
;
1592 BOOL sat
= (arg
->dst
& WINED3DSP_DSTMOD_MASK
) & WINED3DSPDM_SATURATE
;
1594 pshader_get_register_name(arg
->shader
, arg
->dst
, dst_name
);
1595 shader_arb_get_write_mask(arg
, arg
->dst
, dst_wmask
);
1597 pshader_gen_input_modifier_line(arg
->shader
, buffer
, arg
->src
[0], 0, src_name
);
1598 shader_addline(buffer
, "SCS%s %s%s, %s;\n", sat
? "_SAT" : "", dst_name
, dst_wmask
,
1602 pshader_gen_output_modifier_line(buffer
, FALSE
, dst_wmask
, shift
, dst_name
);
1606 static GLuint
create_arb_blt_vertex_program(const WineD3D_GL_Info
*gl_info
)
1608 GLuint program_id
= 0;
1609 const char *blt_vprogram
=
1611 "PARAM c[1] = { { 1, 0.5 } };\n"
1612 "MOV result.position, vertex.position;\n"
1613 "MOV result.color, c[0].x;\n"
1614 "MOV result.texcoord[0], vertex.texcoord[0];\n"
1617 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
1618 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
1619 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_vprogram
), blt_vprogram
));
1621 if (glGetError() == GL_INVALID_OPERATION
) {
1623 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
1624 FIXME("Vertex program error at position %d: %s\n", pos
,
1625 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
1631 static GLuint
create_arb_blt_fragment_program(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
1633 GLuint program_id
= 0;
1634 const char *blt_fprograms
[tex_type_count
] =
1641 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1642 "MOV result.depth.z, R0.x;\n"
1649 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1650 "MOV result.depth.z, R0.x;\n"
1655 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1656 "MOV result.depth.z, R0.x;\n"
1660 if (!blt_fprograms
[tex_type
])
1662 FIXME("tex_type %#x not supported\n", tex_type
);
1666 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
1667 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
1668 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_fprograms
[tex_type
]), blt_fprograms
[tex_type
]));
1670 if (glGetError() == GL_INVALID_OPERATION
) {
1672 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
1673 FIXME("Fragment program error at position %d: %s\n", pos
,
1674 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
1680 static void shader_arb_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
1681 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1682 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->shader_priv
;
1683 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1686 TRACE("Using vertex shader\n");
1687 IWineD3DVertexShaderImpl_CompileShader(This
->stateBlock
->vertexShader
);
1689 priv
->current_vprogram_id
= ((IWineD3DVertexShaderImpl
*)This
->stateBlock
->vertexShader
)->prgId
;
1691 /* Bind the vertex program */
1692 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
1693 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1695 /* Enable OpenGL vertex programs */
1696 glEnable(GL_VERTEX_PROGRAM_ARB
);
1697 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1698 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
1699 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
1700 priv
->current_vprogram_id
= 0;
1701 glDisable(GL_VERTEX_PROGRAM_ARB
);
1702 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1706 struct ps_compile_args compile_args
;
1707 TRACE("Using pixel shader\n");
1708 find_ps_compile_args((IWineD3DPixelShaderImpl
*) This
->stateBlock
->pixelShader
, This
->stateBlock
, &compile_args
);
1709 priv
->current_fprogram_id
= find_gl_pshader((IWineD3DPixelShaderImpl
*) This
->stateBlock
->pixelShader
,
1712 /* Bind the fragment program */
1713 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
1714 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1716 if(!priv
->use_arbfp_fixed_func
) {
1717 /* Enable OpenGL fragment programs */
1718 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1719 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1721 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
1722 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
) && !priv
->use_arbfp_fixed_func
) {
1723 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
1724 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
1725 * replacement shader
1727 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
1728 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1729 priv
->current_fprogram_id
= 0;
1733 static void shader_arb_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
1734 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1735 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->shader_priv
;
1736 GLuint
*blt_fprogram
= &priv
->depth_blt_fprogram_id
[tex_type
];
1737 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1739 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
1740 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
1741 glEnable(GL_VERTEX_PROGRAM_ARB
);
1743 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
);
1744 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
1745 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1748 static void shader_arb_deselect_depth_blt(IWineD3DDevice
*iface
) {
1749 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1750 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->shader_priv
;
1751 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1753 if (priv
->current_vprogram_id
) {
1754 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
1755 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
1757 glEnable(GL_VERTEX_PROGRAM_ARB
);
1758 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
1760 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
1762 glDisable(GL_VERTEX_PROGRAM_ARB
);
1763 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
1766 if (priv
->current_fprogram_id
) {
1767 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
1768 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
1770 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
1771 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
1773 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
1775 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
1776 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
1780 static void shader_arb_cleanup(IWineD3DDevice
*iface
) {
1781 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1782 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1783 if (GL_SUPPORT(ARB_VERTEX_PROGRAM
)) glDisable(GL_VERTEX_PROGRAM_ARB
);
1784 if (GL_SUPPORT(ARB_FRAGMENT_PROGRAM
)) glDisable(GL_FRAGMENT_PROGRAM_ARB
);
1787 static void shader_arb_destroy(IWineD3DBaseShader
*iface
) {
1788 IWineD3DBaseShaderImpl
*baseShader
= (IWineD3DBaseShaderImpl
*) iface
;
1789 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)baseShader
->baseShader
.device
)->adapter
->gl_info
;
1790 char pshader
= shader_is_pshader_version(baseShader
->baseShader
.hex_version
);
1793 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
1797 for(i
= 0; i
< This
->num_gl_shaders
; i
++) {
1798 GL_EXTCALL(glDeleteProgramsARB(1, &This
->gl_shaders
[i
].prgId
));
1799 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->gl_shaders[i].prgId))");
1802 HeapFree(GetProcessHeap(), 0, This
->gl_shaders
);
1803 This
->gl_shaders
= NULL
;
1804 This
->num_gl_shaders
= 0;
1806 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*) iface
;
1809 GL_EXTCALL(glDeleteProgramsARB(1, &This
->prgId
));
1810 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &This->prgId))");
1811 ((IWineD3DVertexShaderImpl
*) This
)->prgId
= 0;
1814 baseShader
->baseShader
.is_compiled
= FALSE
;
1817 static HRESULT
shader_arb_alloc(IWineD3DDevice
*iface
) {
1818 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1819 This
->shader_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
1823 static void shader_arb_free(IWineD3DDevice
*iface
) {
1824 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
1825 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
1826 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->shader_priv
;
1829 if(priv
->depth_blt_vprogram_id
) {
1830 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
1832 for (i
= 0; i
< tex_type_count
; ++i
) {
1833 if (priv
->depth_blt_fprogram_id
[i
]) {
1834 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id
[i
]));
1838 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
1841 static BOOL
shader_arb_dirty_const(IWineD3DDevice
*iface
) {
1845 static void arbfp_add_sRGB_correction(SHADER_BUFFER
*buffer
, const char *fragcolor
, const char *tmp1
,
1846 const char *tmp2
, const char *tmp3
, const char *tmp4
) {
1847 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
1849 /* Calculate the > 0.0031308 case */
1850 shader_addline(buffer
, "POW %s.x, %s.x, srgb_pow.x;\n", tmp1
, fragcolor
);
1851 shader_addline(buffer
, "POW %s.y, %s.y, srgb_pow.y;\n", tmp1
, fragcolor
);
1852 shader_addline(buffer
, "POW %s.z, %s.z, srgb_pow.z;\n", tmp1
, fragcolor
);
1853 shader_addline(buffer
, "MUL %s, %s, srgb_mul_hi;\n", tmp1
, tmp1
);
1854 shader_addline(buffer
, "SUB %s, %s, srgb_sub_hi;\n", tmp1
, tmp1
);
1855 /* Calculate the < case */
1856 shader_addline(buffer
, "MUL %s, srgb_mul_low, %s;\n", tmp2
, fragcolor
);
1857 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
1858 shader_addline(buffer
, "SLT %s, srgb_comparison, %s;\n", tmp3
, fragcolor
);
1859 shader_addline(buffer
, "SGE %s, srgb_comparison, %s;\n", tmp4
, fragcolor
);
1860 /* Store the components > 0.0031308 in the destination */
1861 shader_addline(buffer
, "MUL %s, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
1862 /* Add the components that are < 0.0031308 */
1863 shader_addline(buffer
, "MAD result.color.xyz, %s, %s, %s;\n", tmp2
, tmp4
, fragcolor
);
1864 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
1867 static GLuint
shader_arb_generate_pshader(IWineD3DPixelShader
*iface
, SHADER_BUFFER
*buffer
) {
1868 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)iface
;
1869 const shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
1870 CONST DWORD
*function
= This
->baseShader
.function
;
1871 const char *fragcolor
;
1872 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
1873 const local_constant
*lconst
;
1876 /* Create the hw ARB shader */
1877 shader_addline(buffer
, "!!ARBfp1.0\n");
1879 shader_addline(buffer
, "TEMP TMP;\n"); /* Used in matrix ops */
1880 shader_addline(buffer
, "TEMP TMP2;\n"); /* Used in matrix ops */
1881 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
1882 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
1883 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
1884 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
1885 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
1886 shader_addline(buffer
, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
1888 /* Base Declarations */
1889 shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
1891 /* We need two variables for fog blending */
1892 shader_addline(buffer
, "TEMP TMP_FOG;\n");
1893 if (This
->baseShader
.hex_version
>= WINED3DPS_VERSION(2,0)) {
1894 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
1897 /* Base Shader Body */
1898 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
1900 /* calculate fog and blend it
1901 * NOTE: state.fog.params.y and state.fog.params.z don't hold fog start s and end e but
1902 * -1/(e-s) and e/(e-s) respectively.
1904 shader_addline(buffer
, "MAD_SAT TMP_FOG, fragment.fogcoord, state.fog.params.y, state.fog.params.z;\n");
1906 if (This
->baseShader
.hex_version
< WINED3DPS_VERSION(2,0)) {
1909 fragcolor
= "TMP_COLOR";
1911 if(((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->stateBlock
->renderState
[WINED3DRS_SRGBWRITEENABLE
]) {
1912 arbfp_add_sRGB_correction(buffer
, fragcolor
, "TMP", "TMP2", "TA", "TB");
1914 if (This
->baseShader
.hex_version
< WINED3DPS_VERSION(3,0)) {
1915 shader_addline(buffer
, "LRP result.color.rgb, TMP_FOG.x, %s, state.fog.color;\n", fragcolor
);
1916 shader_addline(buffer
, "MOV result.color.a, %s.a;\n", fragcolor
);
1919 shader_addline(buffer
, "END\n");
1921 /* TODO: change to resource.glObjectHandle or something like that */
1922 GL_EXTCALL(glGenProgramsARB(1, &retval
));
1924 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
1925 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
1927 TRACE("Created hw pixel shader, prg=%d\n", retval
);
1928 /* Create the program and check for errors */
1929 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
1930 buffer
->bsize
, buffer
->buffer
));
1932 if (glGetError() == GL_INVALID_OPERATION
) {
1934 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
1935 FIXME("HW PixelShader Error at position %d: %s\n",
1936 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
1940 /* Load immediate constants */
1941 if(!This
->baseShader
.load_local_constsF
) {
1942 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
1943 const float *value
= (const float *)lconst
->value
;
1944 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, lconst
->idx
, value
));
1945 checkGLcall("glProgramLocalParameter4fvARB");
1952 static void shader_arb_generate_vshader(IWineD3DVertexShader
*iface
, SHADER_BUFFER
*buffer
) {
1953 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*)iface
;
1954 const shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
1955 CONST DWORD
*function
= This
->baseShader
.function
;
1956 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
1957 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
1958 const local_constant
*lconst
;
1960 /* Create the hw ARB shader */
1961 shader_addline(buffer
, "!!ARBvp1.0\n");
1962 shader_addline(buffer
, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This
->rel_offset
);
1964 /* Mesa supports only 95 constants */
1965 if (GL_VEND(MESA
) || GL_VEND(WINE
))
1966 This
->baseShader
.limits
.constant_float
=
1967 min(95, This
->baseShader
.limits
.constant_float
);
1969 shader_addline(buffer
, "TEMP TMP;\n");
1971 /* Base Declarations */
1972 shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
);
1974 /* We need a constant to fixup the final position */
1975 shader_addline(buffer
, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS
);
1977 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
1978 * for output parameters. D3D in theory does not do that either, but some applications depend on a
1979 * proper initialization of the secondary color, and programs using the fixed function pipeline without
1980 * a replacement shader depend on the texcoord.w being set properly.
1982 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
1983 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
1984 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
1985 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
1986 * this can eat a number of instructions, so skip it unless this cap is set as well
1988 if(!GL_SUPPORT(NV_VERTEX_PROGRAM
)) {
1989 shader_addline(buffer
, "MOV result.color.secondary, -helper_const.wwwy;\n");
1991 if((GLINFO_LOCATION
).set_texcoord_w
&& !device
->frag_pipe
->ffp_proj_control
) {
1993 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
1994 if(This
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
1995 This
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
1996 shader_addline(buffer
, "MOV result.texcoord[%u].w, -helper_const.y;\n", i
);
2002 /* Base Shader Body */
2003 shader_generate_main( (IWineD3DBaseShader
*) This
, buffer
, reg_maps
, function
);
2005 /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
2007 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
2009 /* Write the final position.
2011 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2012 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2013 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2014 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2016 shader_addline(buffer
, "MUL TMP, posFixup, TMP_OUT.w;\n");
2017 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TMP.z;\n");
2018 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TMP.w;\n");
2020 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2021 * and the glsl equivalent
2023 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2025 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
2027 shader_addline(buffer
, "END\n");
2029 /* TODO: change to resource.glObjectHandle or something like that */
2030 GL_EXTCALL(glGenProgramsARB(1, &This
->prgId
));
2032 TRACE("Creating a hw vertex shader, prg=%d\n", This
->prgId
);
2033 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, This
->prgId
));
2035 TRACE("Created hw vertex shader, prg=%d\n", This
->prgId
);
2036 /* Create the program and check for errors */
2037 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
2038 buffer
->bsize
, buffer
->buffer
));
2040 if (glGetError() == GL_INVALID_OPERATION
) {
2042 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
2043 FIXME("HW VertexShader Error at position %d: %s\n",
2044 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2048 /* Load immediate constants */
2049 if(!This
->baseShader
.load_local_constsF
) {
2050 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
2051 const float *value
= (const float *)lconst
->value
;
2052 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, lconst
->idx
, value
));
2057 static void shader_arb_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
2059 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2060 * then overwrite the shader specific ones
2062 none_shader_backend
.shader_get_caps(devtype
, gl_info
, pCaps
);
2064 if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
2065 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(1,1);
2066 TRACE_(d3d_caps
)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2067 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
2070 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
)) {
2071 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(1,4);
2072 pCaps
->PixelShader1xMaxValue
= 8.0;
2073 TRACE_(d3d_caps
)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2077 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
2079 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
2081 TRACE("Checking support for color_fixup:\n");
2082 dump_color_fixup_desc(fixup
);
2085 /* We support everything except YUV conversions. */
2086 if (!is_yuv_fixup(fixup
))
2092 TRACE("[FAILED]\n");
2096 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
2098 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
2099 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
2100 /* WINED3DSIH_BEM */ pshader_hw_bem
,
2101 /* WINED3DSIH_BREAK */ NULL
,
2102 /* WINED3DSIH_BREAKC */ NULL
,
2103 /* WINED3DSIH_BREAKP */ NULL
,
2104 /* WINED3DSIH_CALL */ NULL
,
2105 /* WINED3DSIH_CALLNZ */ NULL
,
2106 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
2107 /* WINED3DSIH_CND */ pshader_hw_cnd
,
2108 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
2109 /* WINED3DSIH_DCL */ NULL
,
2110 /* WINED3DSIH_DEF */ NULL
,
2111 /* WINED3DSIH_DEFB */ NULL
,
2112 /* WINED3DSIH_DEFI */ NULL
,
2113 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
2114 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
2115 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
2116 /* WINED3DSIH_DST */ shader_hw_map2gl
,
2117 /* WINED3DSIH_DSX */ NULL
,
2118 /* WINED3DSIH_DSY */ NULL
,
2119 /* WINED3DSIH_ELSE */ NULL
,
2120 /* WINED3DSIH_ENDIF */ NULL
,
2121 /* WINED3DSIH_ENDLOOP */ NULL
,
2122 /* WINED3DSIH_ENDREP */ NULL
,
2123 /* WINED3DSIH_EXP */ shader_hw_map2gl
,
2124 /* WINED3DSIH_EXPP */ shader_hw_map2gl
,
2125 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
2126 /* WINED3DSIH_IF */ NULL
,
2127 /* WINED3DSIH_IFC */ NULL
,
2128 /* WINED3DSIH_LABEL */ NULL
,
2129 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
2130 /* WINED3DSIH_LOG */ shader_hw_map2gl
,
2131 /* WINED3DSIH_LOGP */ shader_hw_map2gl
,
2132 /* WINED3DSIH_LOOP */ NULL
,
2133 /* WINED3DSIH_LRP */ shader_hw_map2gl
,
2134 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
2135 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
2136 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
2137 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
2138 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
2139 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
2140 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
2141 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
2142 /* WINED3DSIH_MOV */ shader_hw_mov
,
2143 /* WINED3DSIH_MOVA */ shader_hw_mov
,
2144 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
2145 /* WINED3DSIH_NOP */ shader_hw_map2gl
,
2146 /* WINED3DSIH_NRM */ shader_hw_nrm
,
2147 /* WINED3DSIH_PHASE */ NULL
,
2148 /* WINED3DSIH_POW */ shader_hw_map2gl
,
2149 /* WINED3DSIH_RCP */ vshader_hw_rsq_rcp
,
2150 /* WINED3DSIH_REP */ NULL
,
2151 /* WINED3DSIH_RET */ NULL
,
2152 /* WINED3DSIH_RSQ */ vshader_hw_rsq_rcp
,
2153 /* WINED3DSIH_SETP */ NULL
,
2154 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
2155 /* WINED3DSIH_SGN */ NULL
,
2156 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
2157 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
2158 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
2159 /* WINED3DSIH_TEX */ pshader_hw_tex
,
2160 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
2161 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
2162 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
2163 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
2164 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
2165 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
2166 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
2167 /* WINED3DSIH_TEXLDD */ NULL
,
2168 /* WINED3DSIH_TEXLDL */ NULL
,
2169 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
2170 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
2171 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
2172 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
2173 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
2174 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
2175 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
2176 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
2177 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
2178 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
2179 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
2180 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
2183 const shader_backend_t arb_program_shader_backend
= {
2184 shader_arb_instruction_handler_table
,
2186 shader_arb_select_depth_blt
,
2187 shader_arb_deselect_depth_blt
,
2188 shader_arb_load_constants
,
2190 shader_arb_color_correction
,
2194 shader_arb_dirty_const
,
2195 shader_arb_generate_pshader
,
2196 shader_arb_generate_vshader
,
2197 shader_arb_get_caps
,
2198 shader_arb_color_fixup_supported
,
2201 /* ARB_fragment_program fixed function pipeline replacement definitions */
2202 #define ARB_FFP_CONST_TFACTOR 0
2203 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2204 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2205 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2206 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2208 struct arbfp_ffp_desc
2210 struct ffp_frag_desc parent
;
2212 unsigned int num_textures_used
;
2215 static void arbfp_enable(IWineD3DDevice
*iface
, BOOL enable
) {
2217 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
2218 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2220 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
2221 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2225 static HRESULT
arbfp_alloc(IWineD3DDevice
*iface
) {
2226 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
2227 struct shader_arb_priv
*priv
;
2228 /* Share private data between the shader backend and the pipeline replacement, if both
2229 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2230 * if no pixel shader is bound or not
2232 if(This
->shader_backend
== &arb_program_shader_backend
) {
2233 This
->fragment_priv
= This
->shader_priv
;
2235 This
->fragment_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
2236 if(!This
->fragment_priv
) return E_OUTOFMEMORY
;
2238 priv
= (struct shader_arb_priv
*) This
->fragment_priv
;
2239 priv
->fragment_shaders
= hash_table_create(ffp_frag_program_key_hash
, ffp_frag_program_key_compare
);
2240 priv
->use_arbfp_fixed_func
= TRUE
;
2244 static void arbfp_free_ffpshader(void *value
, void *gli
) {
2245 const WineD3D_GL_Info
*gl_info
= gli
;
2246 struct arbfp_ffp_desc
*entry_arb
= value
;
2249 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
2250 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2251 HeapFree(GetProcessHeap(), 0, entry_arb
);
2255 static void arbfp_free(IWineD3DDevice
*iface
) {
2256 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
2257 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) This
->fragment_priv
;
2259 hash_table_destroy(priv
->fragment_shaders
, arbfp_free_ffpshader
, &This
->adapter
->gl_info
);
2260 priv
->use_arbfp_fixed_func
= FALSE
;
2262 if(This
->shader_backend
!= &arb_program_shader_backend
) {
2263 HeapFree(GetProcessHeap(), 0, This
->fragment_priv
);
2267 static void arbfp_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct fragment_caps
*caps
)
2269 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
2270 WINED3DTEXOPCAPS_SELECTARG1
|
2271 WINED3DTEXOPCAPS_SELECTARG2
|
2272 WINED3DTEXOPCAPS_MODULATE4X
|
2273 WINED3DTEXOPCAPS_MODULATE2X
|
2274 WINED3DTEXOPCAPS_MODULATE
|
2275 WINED3DTEXOPCAPS_ADDSIGNED2X
|
2276 WINED3DTEXOPCAPS_ADDSIGNED
|
2277 WINED3DTEXOPCAPS_ADD
|
2278 WINED3DTEXOPCAPS_SUBTRACT
|
2279 WINED3DTEXOPCAPS_ADDSMOOTH
|
2280 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
2281 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
2282 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
2283 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
2284 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
2285 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
2286 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
2287 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
2288 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
2289 WINED3DTEXOPCAPS_DOTPRODUCT3
|
2290 WINED3DTEXOPCAPS_MULTIPLYADD
|
2291 WINED3DTEXOPCAPS_LERP
|
2292 WINED3DTEXOPCAPS_BUMPENVMAP
|
2293 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
2295 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
2297 caps
->MaxTextureBlendStages
= 8;
2298 caps
->MaxSimultaneousTextures
= min(GL_LIMITS(fragment_samplers
), 8);
2300 caps
->PrimitiveMiscCaps
|= WINED3DPMISCCAPS_TSSARGTEMP
;
2302 #undef GLINFO_LOCATION
2304 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
2305 static void state_texfactor_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2307 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2309 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2310 * application provided constants
2312 if(device
->shader_backend
== &arb_program_shader_backend
) {
2313 if(use_ps(device
)) return;
2315 device
= stateblock
->wineD3DDevice
;
2316 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
2317 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
2320 D3DCOLORTOGLFLOAT4(stateblock
->renderState
[WINED3DRS_TEXTUREFACTOR
], col
);
2321 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
2322 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
2326 static void state_arb_specularenable(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2328 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2330 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
2331 * application provided constants
2333 if(device
->shader_backend
== &arb_program_shader_backend
) {
2334 if(use_ps(device
)) return;
2336 device
= stateblock
->wineD3DDevice
;
2337 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
2338 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
2341 if(stateblock
->renderState
[WINED3DRS_SPECULARENABLE
]) {
2342 /* The specular color has no alpha */
2343 col
[0] = 1.0; col
[1] = 1.0;
2344 col
[2] = 1.0; col
[3] = 0.0;
2346 col
[0] = 0.0; col
[1] = 0.0;
2347 col
[2] = 0.0; col
[3] = 0.0;
2349 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
2350 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
2353 static void set_bumpmat_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2354 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE
;
2355 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2358 if(use_ps(device
)) {
2360 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.bumpmat
[stage
]) {
2361 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
2364 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
2365 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
2369 if(device
->shader_backend
== &arb_program_shader_backend
) {
2370 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2373 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
2374 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
2375 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
2378 mat
[0][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT00
]);
2379 mat
[0][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT01
]);
2380 mat
[1][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT10
]);
2381 mat
[1][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT11
]);
2383 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
2384 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
2387 static void tex_bumpenvlum_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2388 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / WINED3D_HIGHEST_TEXTURE_STATE
;
2389 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2392 if(use_ps(device
)) {
2394 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.luminanceparams
[stage
]) {
2395 /* The pixel shader has to know the luminance offset. Do a constants update if it
2396 * isn't scheduled anyway
2398 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
2399 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
2403 if(device
->shader_backend
== &arb_program_shader_backend
) {
2404 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
2407 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
2408 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
2409 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
2412 param
[0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
]);
2413 param
[1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
]);
2417 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
2418 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
2421 static const char *get_argreg(SHADER_BUFFER
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
) {
2424 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
2426 switch(arg
& WINED3DTA_SELECTMASK
) {
2427 case WINED3DTA_DIFFUSE
:
2428 ret
= "fragment.color.primary"; break;
2430 case WINED3DTA_CURRENT
:
2431 if(stage
== 0) ret
= "fragment.color.primary";
2435 case WINED3DTA_TEXTURE
:
2437 case 0: ret
= "tex0"; break;
2438 case 1: ret
= "tex1"; break;
2439 case 2: ret
= "tex2"; break;
2440 case 3: ret
= "tex3"; break;
2441 case 4: ret
= "tex4"; break;
2442 case 5: ret
= "tex5"; break;
2443 case 6: ret
= "tex6"; break;
2444 case 7: ret
= "tex7"; break;
2445 default: ret
= "unknown texture";
2449 case WINED3DTA_TFACTOR
:
2450 ret
= "tfactor"; break;
2452 case WINED3DTA_SPECULAR
:
2453 ret
= "fragment.color.secondary"; break;
2455 case WINED3DTA_TEMP
:
2456 ret
= "tempreg"; break;
2458 case WINED3DTA_CONSTANT
:
2459 FIXME("Implement perstage constants\n");
2461 case 0: ret
= "const0"; break;
2462 case 1: ret
= "const1"; break;
2463 case 2: ret
= "const2"; break;
2464 case 3: ret
= "const3"; break;
2465 case 4: ret
= "const4"; break;
2466 case 5: ret
= "const5"; break;
2467 case 6: ret
= "const6"; break;
2468 case 7: ret
= "const7"; break;
2469 default: ret
= "unknown constant";
2477 if(arg
& WINED3DTA_COMPLEMENT
) {
2478 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
2479 if(argnum
== 0) ret
= "arg0";
2480 if(argnum
== 1) ret
= "arg1";
2481 if(argnum
== 2) ret
= "arg2";
2483 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
2484 shader_addline(buffer
, "MOV arg%u, %s.a;\n", argnum
, ret
);
2485 if(argnum
== 0) ret
= "arg0";
2486 if(argnum
== 1) ret
= "arg1";
2487 if(argnum
== 2) ret
= "arg2";
2492 static void gen_ffp_instr(SHADER_BUFFER
*buffer
, unsigned int stage
, BOOL color
, BOOL alpha
,
2493 DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
) {
2494 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
2495 unsigned int mul
= 1;
2496 BOOL mul_final_dest
= FALSE
;
2498 if(color
&& alpha
) dstmask
= "";
2499 else if(color
) dstmask
= ".rgb";
2500 else dstmask
= ".a";
2502 if(dst
== tempreg
) dstreg
= "tempreg";
2503 else dstreg
= "ret";
2505 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
2506 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
2507 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
2510 case WINED3DTOP_DISABLE
:
2511 if(stage
== 1) shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
2514 case WINED3DTOP_SELECTARG2
:
2516 case WINED3DTOP_SELECTARG1
:
2517 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
2520 case WINED3DTOP_MODULATE4X
:
2522 case WINED3DTOP_MODULATE2X
:
2524 if(strcmp(dstreg
, "result.color") == 0) {
2526 mul_final_dest
= TRUE
;
2528 case WINED3DTOP_MODULATE
:
2529 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2532 case WINED3DTOP_ADDSIGNED2X
:
2534 if(strcmp(dstreg
, "result.color") == 0) {
2536 mul_final_dest
= TRUE
;
2538 case WINED3DTOP_ADDSIGNED
:
2539 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
2541 case WINED3DTOP_ADD
:
2542 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2545 case WINED3DTOP_SUBTRACT
:
2546 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
2549 case WINED3DTOP_ADDSMOOTH
:
2550 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
2551 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2554 case WINED3DTOP_BLENDCURRENTALPHA
:
2555 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
2556 shader_addline(buffer
, "LRP %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2558 case WINED3DTOP_BLENDFACTORALPHA
:
2559 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
2560 shader_addline(buffer
, "LRP %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2562 case WINED3DTOP_BLENDTEXTUREALPHA
:
2563 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
2564 shader_addline(buffer
, "LRP %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2566 case WINED3DTOP_BLENDDIFFUSEALPHA
:
2567 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
2568 shader_addline(buffer
, "LRP %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2571 case WINED3DTOP_BLENDTEXTUREALPHAPM
:
2572 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
2573 shader_addline(buffer
, "SUB arg0.a, const.x, %s.a;\n", arg0
);
2574 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.a, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2577 /* D3DTOP_PREMODULATE ???? */
2579 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR
:
2580 shader_addline(buffer
, "SUB arg0.a, const.x, %s;\n", arg1
);
2581 shader_addline(buffer
, "MAD_SAT %s%s, arg0.a, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
2583 case WINED3DTOP_MODULATEALPHA_ADDCOLOR
:
2584 shader_addline(buffer
, "MAD_SAT %s%s, %s.a, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
2586 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA
:
2587 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
2588 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.a;\n", dstreg
, dstmask
, arg2
, arg1
);
2590 case WINED3DTOP_MODULATECOLOR_ADDALPHA
:
2591 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.a;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
2594 case WINED3DTOP_DOTPRODUCT3
:
2596 if(strcmp(dstreg
, "result.color") == 0) {
2598 mul_final_dest
= TRUE
;
2600 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
2601 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
2602 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
2605 case WINED3DTOP_MULTIPLYADD
:
2606 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
2609 case WINED3DTOP_LERP
:
2610 /* The msdn is not quite right here */
2611 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
2614 case WINED3DTOP_BUMPENVMAP
:
2615 case WINED3DTOP_BUMPENVMAPLUMINANCE
:
2616 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
2620 FIXME("Unhandled texture op %08x\n", op
);
2624 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
2625 } else if(mul
== 4) {
2626 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
2630 /* The stateblock is passed for GLINFO_LOCATION */
2631 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, IWineD3DStateBlockImpl
*stateblock
)
2634 SHADER_BUFFER buffer
;
2635 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2636 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2637 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
2638 const char *textype
;
2639 const char *instr
, *sat
;
2640 char colorcor_dst
[8];
2642 DWORD arg0
, arg1
, arg2
;
2643 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
2645 const char *final_combiner_src
= "ret";
2647 /* Find out which textures are read */
2648 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2649 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) break;
2650 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
2651 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
2652 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
2653 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2654 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2655 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2657 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHA
) tex_read
[stage
] = TRUE
;
2658 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHAPM
) tex_read
[stage
] = TRUE
;
2659 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
) {
2660 bump_used
[stage
] = TRUE
;
2661 tex_read
[stage
] = TRUE
;
2663 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2664 bump_used
[stage
] = TRUE
;
2665 tex_read
[stage
] = TRUE
;
2666 luminance_used
[stage
] = TRUE
;
2667 } else if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDFACTORALPHA
) {
2668 tfactor_used
= TRUE
;
2671 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
2672 tfactor_used
= TRUE
;
2675 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
2676 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
2677 tempreg_used
= TRUE
;
2680 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) continue;
2681 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
2682 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
2683 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
2684 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2685 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2686 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
2688 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
2689 tempreg_used
= TRUE
;
2691 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
2692 tfactor_used
= TRUE
;
2699 buffer
.newline
= TRUE
;
2700 buffer
.buffer
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, SHADER_PGMSIZE
);
2702 shader_addline(&buffer
, "!!ARBfp1.0\n");
2704 switch(settings
->fog
) {
2705 case FOG_OFF
: break;
2706 case FOG_LINEAR
: shader_addline(&buffer
, "OPTION ARB_fog_linear;\n"); break;
2707 case FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
2708 case FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
2709 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
2712 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
2713 shader_addline(&buffer
, "TEMP TMP;\n");
2714 shader_addline(&buffer
, "TEMP ret;\n");
2715 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
2716 shader_addline(&buffer
, "TEMP arg0;\n");
2717 shader_addline(&buffer
, "TEMP arg1;\n");
2718 shader_addline(&buffer
, "TEMP arg2;\n");
2719 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2720 if(!tex_read
[stage
]) continue;
2721 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
2722 if(!bump_used
[stage
]) continue;
2723 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
2724 if(!luminance_used
[stage
]) continue;
2725 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
2728 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
2730 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
2732 if(settings
->sRGB_write
) {
2733 shader_addline(&buffer
, "PARAM srgb_mul_low = {%f, %f, %f, 1.0};\n",
2734 srgb_mul_low
, srgb_mul_low
, srgb_mul_low
);
2735 shader_addline(&buffer
, "PARAM srgb_comparison = {%f, %f, %f, %f};\n",
2736 srgb_cmp
, srgb_cmp
, srgb_cmp
, srgb_cmp
);
2737 shader_addline(&buffer
, "PARAM srgb_pow = {%f, %f, %f, 1.0};\n",
2738 srgb_pow
, srgb_pow
, srgb_pow
);
2739 shader_addline(&buffer
, "PARAM srgb_mul_hi = {%f, %f, %f, 1.0};\n",
2740 srgb_mul_high
, srgb_mul_high
, srgb_mul_high
);
2741 shader_addline(&buffer
, "PARAM srgb_sub_hi = {%f, %f, %f, 0.0};\n",
2742 srgb_sub_high
, srgb_sub_high
, srgb_sub_high
);
2745 /* Generate texture sampling instructions) */
2746 for(stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3DTOP_DISABLE
; stage
++) {
2747 if(!tex_read
[stage
]) continue;
2749 switch(settings
->op
[stage
].tex_type
) {
2750 case tex_1d
: textype
= "1D"; break;
2751 case tex_2d
: textype
= "2D"; break;
2752 case tex_3d
: textype
= "3D"; break;
2753 case tex_cube
: textype
= "CUBE"; break;
2754 case tex_rect
: textype
= "RECT"; break;
2755 default: textype
= "unexpected_textype"; break;
2758 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
||
2759 settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2765 if(settings
->op
[stage
].projected
== proj_none
) {
2767 } else if(settings
->op
[stage
].projected
== proj_count4
||
2768 settings
->op
[stage
].projected
== proj_count3
) {
2771 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
2776 (settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAP
||
2777 settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
)) {
2778 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
2779 shader_addline(&buffer
, "DP3 ret.r, arg1, tex%u;\n", stage
- 1);
2780 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
2781 shader_addline(&buffer
, "DP3 ret.g, arg1, tex%u;\n", stage
- 1);
2783 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2784 * so multiply the displacement with the dividing parameter before passing it to TXP
2786 if (settings
->op
[stage
].projected
!= proj_none
) {
2787 if(settings
->op
[stage
].projected
== proj_count4
) {
2788 shader_addline(&buffer
, "MOV ret.a, fragment.texcoord[%u].a;\n", stage
);
2789 shader_addline(&buffer
, "MUL ret.rgb, ret, fragment.texcoord[%u].a, fragment.texcoord[%u];\n", stage
, stage
);
2791 shader_addline(&buffer
, "MOV ret.a, fragment.texcoord[%u].b;\n", stage
);
2792 shader_addline(&buffer
, "MAD ret.rgb, ret, fragment.texcoord[%u].b, fragment.texcoord[%u];\n", stage
, stage
);
2795 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
2798 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
2799 instr
, sat
, stage
, stage
, textype
);
2800 if(settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
2801 shader_addline(&buffer
, "MAD_SAT ret.r, tex%u.b, luminance%u.r, luminance%u.g;\n",
2802 stage
- 1, stage
- 1, stage
- 1);
2803 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.r;\n", stage
, stage
);
2805 } else if(settings
->op
[stage
].projected
== proj_count3
) {
2806 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
2807 shader_addline(&buffer
, "MOV ret.a, ret.b;\n");
2808 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
2809 instr
, sat
, stage
, stage
, textype
);
2811 shader_addline(&buffer
, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
2812 instr
, sat
, stage
, stage
, stage
, textype
);
2815 sprintf(colorcor_dst
, "tex%u", stage
);
2816 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
2817 settings
->op
[stage
].color_correction
);
2820 /* Generate the main shader */
2821 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
2822 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) {
2824 final_combiner_src
= "fragment.color.primary";
2829 if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
2830 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
2831 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
2832 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
2833 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
2834 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
2835 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
2836 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
2837 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
2838 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
2839 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
2840 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
2842 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
&&
2843 settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
&&
2844 settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
&&
2845 settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
2848 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) {
2849 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
2850 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2851 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2853 shader_addline(&buffer
, "MOV ret.a, fragment.color.primary.a;\n");
2855 } else if(op_equal
) {
2856 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
2857 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2858 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2860 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
2861 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
2862 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
2863 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
2864 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
2865 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
2869 if(settings
->sRGB_write
) {
2870 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
2871 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg");
2872 shader_addline(&buffer
, "MOV result.color.a, ret.a;\n");
2874 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
2878 shader_addline(&buffer
, "END\n");
2880 /* Generate the shader */
2881 GL_EXTCALL(glGenProgramsARB(1, &ret
));
2882 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
2883 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
2885 if (glGetError() == GL_INVALID_OPERATION
) {
2887 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
2888 FIXME("Fragment program error at position %d: %s\n", pos
,
2889 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2891 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
2895 static void fragment_prog_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2896 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
2897 struct shader_arb_priv
*priv
= (struct shader_arb_priv
*) device
->fragment_priv
;
2898 BOOL use_pshader
= use_ps(device
);
2899 BOOL use_vshader
= use_vs(device
);
2900 struct ffp_frag_settings settings
;
2901 const struct arbfp_ffp_desc
*desc
;
2904 if(isStateDirty(context
, STATE_RENDER(WINED3DRS_FOGENABLE
))) {
2905 if(!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
2906 /* Reload fixed function constants since they collide with the pixel shader constants */
2907 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
2908 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
2910 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
2911 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
2917 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
2918 gen_ffp_frag_op(stateblock
, &settings
, FALSE
);
2919 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(priv
->fragment_shaders
, &settings
);
2921 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
2924 ERR("Out of memory\n");
2927 new_desc
->num_textures_used
= 0;
2928 for(i
= 0; i
< GL_LIMITS(texture_stages
); i
++) {
2929 if(settings
.op
[i
].cop
== WINED3DTOP_DISABLE
) break;
2930 new_desc
->num_textures_used
= i
;
2933 memcpy(&new_desc
->parent
.settings
, &settings
, sizeof(settings
));
2934 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, stateblock
);
2935 add_ffp_frag_shader(priv
->fragment_shaders
, &new_desc
->parent
);
2936 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
2940 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
2941 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
2944 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
2945 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
2946 priv
->current_fprogram_id
= desc
->shader
;
2948 if(device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
2949 /* Reload fixed function constants since they collide with the pixel shader constants */
2950 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
2951 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
2953 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
2954 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
2958 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
2959 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
2960 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
2961 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
2962 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
2964 * Don't call shader_select if the vertex shader is dirty, because some shader backends(GLSL) need both shaders
2965 * to be compiled before activating them(needs some cleanups in the shader backend interface)
2967 if(!isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
2968 device
->shader_backend
->shader_select((IWineD3DDevice
*)stateblock
->wineD3DDevice
, use_pshader
, use_vshader
);
2970 if (!isStateDirty(context
, STATE_VERTEXSHADERCONSTANT
) && (use_vshader
|| use_pshader
)) {
2971 device
->StateTable
[STATE_VERTEXSHADERCONSTANT
].apply(STATE_VERTEXSHADERCONSTANT
, stateblock
, context
);
2975 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
2979 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
2980 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
2981 * state table, so we need to handle that with a forwarding function. The other invisible side effect
2982 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
2983 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
2985 static void state_arbfp_fog(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2986 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
2987 fragment_prog_arbfp(state
, stateblock
, context
);
2991 static void textransform(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
2992 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
2993 fragment_prog_arbfp(state
, stateblock
, context
);
2997 #undef GLINFO_LOCATION
2999 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] = {
3000 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), state_texfactor_arbfp
}, 0 },
3001 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3002 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3003 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3004 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3005 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3006 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3007 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3008 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3009 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3010 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3011 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3012 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3013 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3014 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3015 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3016 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3017 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3018 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3019 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3020 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3021 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3022 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3023 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3024 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3025 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3026 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3027 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3028 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3029 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3030 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3031 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3032 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3033 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3034 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3035 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3036 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3037 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3038 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3039 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3040 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3041 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3042 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3043 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3044 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3045 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3046 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3047 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3048 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3049 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3050 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3051 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3052 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3053 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3054 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3055 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3056 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3057 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3058 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3059 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3060 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3061 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3062 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3063 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3064 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3065 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3066 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3067 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3068 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3069 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3070 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3071 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3072 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3073 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3074 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3075 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3076 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3077 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3078 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3079 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3080 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3081 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3082 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3083 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3084 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3085 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3086 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3087 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3088 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3089 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3090 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3091 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3092 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3093 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3094 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3095 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3096 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3097 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3098 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3099 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3100 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3101 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3102 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3103 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3104 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3105 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3106 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3107 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3108 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3109 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3110 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3111 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3112 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3113 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3114 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3115 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3116 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3117 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3118 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3119 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3120 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3121 { STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim
}, 0 },
3122 { STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim
}, 0 },
3123 { STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim
}, 0 },
3124 { STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim
}, 0 },
3125 { STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim
}, 0 },
3126 { STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim
}, 0 },
3127 { STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim
}, 0 },
3128 { STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim
}, 0 },
3129 { STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3130 { STATE_RENDER(WINED3DRS_FOGENABLE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3131 { STATE_RENDER(WINED3DRS_FOGTABLEMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3132 { STATE_RENDER(WINED3DRS_FOGVERTEXMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3133 { STATE_RENDER(WINED3DRS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3134 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3135 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3136 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3137 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3138 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3139 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3140 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3141 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3142 { STATE_RENDER(WINED3DRS_SPECULARENABLE
), { STATE_RENDER(WINED3DRS_SPECULARENABLE
), state_arb_specularenable
}, 0 },
3143 {0 /* Terminate */, { 0, 0 }, 0 },
3146 const struct fragment_pipeline arbfp_fragment_pipeline
= {
3151 shader_arb_color_fixup_supported
,
3152 arbfp_fragmentstate_template
,
3153 TRUE
/* We can disable projected textures */
3156 #define GLINFO_LOCATION device->adapter->gl_info
3158 struct arbfp_blit_priv
{
3159 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
3160 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
3161 GLenum yv12_rect_shader
, yv12_2d_shader
;
3164 static HRESULT
arbfp_blit_alloc(IWineD3DDevice
*iface
) {
3165 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3166 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
3167 if(!device
->blit_priv
) {
3168 ERR("Out of memory\n");
3169 return E_OUTOFMEMORY
;
3173 static void arbfp_blit_free(IWineD3DDevice
*iface
) {
3174 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3175 struct arbfp_blit_priv
*priv
= (struct arbfp_blit_priv
*) device
->blit_priv
;
3178 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
3179 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
3180 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
3181 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
3182 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
3183 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
3184 checkGLcall("Delete yuv programs\n");
3188 static BOOL
gen_planar_yuv_read(SHADER_BUFFER
*buffer
, enum yuv_fixup yuv_fixup
, GLenum textype
, char *luminance
)
3191 const char *tex
, *texinstr
;
3193 if (yuv_fixup
== YUV_FIXUP_UYVY
) {
3201 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
3202 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
3204 /* This is more tricky than just replacing the texture type - we have to navigate
3205 * properly in the texture to find the correct chroma values
3207 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3211 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3212 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3213 * filtering when we sample the texture.
3215 * These are the rules for reading the chroma:
3221 * So we have to get the sampling x position in non-normalized coordinates in integers
3223 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
3224 shader_addline(buffer
, "MUL texcrd.rg, fragment.texcoord[0], size.x;\n");
3225 shader_addline(buffer
, "MOV texcrd.a, size.x;\n");
3227 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3229 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3230 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3233 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
3234 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
3236 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3237 * even and odd pixels respectively
3239 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
3240 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
3242 /* Sample Pixel 1 */
3243 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
3245 /* Put the value into either of the chroma values */
3246 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
3247 shader_addline(buffer
, "MUL chroma.r, luminance.%c, temp.x;\n", chroma
);
3248 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
3249 shader_addline(buffer
, "MUL chroma.g, luminance.%c, temp.x;\n", chroma
);
3251 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3252 * the pixel right to the current one. Otherwise, sample the left pixel.
3253 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3255 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3256 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
3257 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
3259 /* Put the value into the other chroma */
3260 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
3261 shader_addline(buffer
, "MAD chroma.g, luminance.%c, temp.x, chroma.g;\n", chroma
);
3262 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
3263 shader_addline(buffer
, "MAD chroma.r, luminance.%c, temp.x, chroma.r;\n", chroma
);
3265 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
3266 * the current one and lerp the two U and V values
3269 /* This gives the correctly filtered luminance value */
3270 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
3275 static BOOL
gen_yv12_read(SHADER_BUFFER
*buffer
, GLenum textype
, char *luminance
)
3280 case GL_TEXTURE_2D
: tex
= "2D"; break;
3281 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
3283 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
3287 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
3288 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
3289 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
3290 * pitch of the luminance plane, the packing into the gl texture is a bit
3291 * unfortunate. If the whole texture is interpreted as luminance data it looks
3292 * approximately like this:
3294 * +----------------------------------+----
3306 * +----------------+-----------------+----
3308 * | U even rows | U odd rows |
3310 * +----------------+------------------ -
3312 * | V even rows | V odd rows |
3314 * +----------------+-----------------+----
3318 * So it appears as if there are 4 chroma images, but in fact the odd rows
3319 * in the chroma images are in the same row as the even ones. So its is
3320 * kinda tricky to read
3322 * When reading from rectangle textures, keep in mind that the input y coordinates
3323 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
3325 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
3326 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
3328 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3329 /* the chroma planes have only half the width */
3330 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
3332 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
3333 * the coordinate. Also read the right side of the image when reading odd lines
3335 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
3338 if(textype
== GL_TEXTURE_2D
) {
3340 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
3342 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
3344 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
3345 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
3347 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3348 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3349 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
3350 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3351 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3353 /* clamp, keep the half pixel origin in mind */
3354 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
3355 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
3356 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
3357 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3359 /* Read from [size - size+size/4] */
3360 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
3361 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
3363 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
3364 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
3365 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
3366 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
3367 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
3368 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
3370 /* Make sure to read exactly from the pixel center */
3371 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
3372 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
3375 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
3376 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
3377 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3378 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
3379 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
3381 /* Read the texture, put the result into the output register */
3382 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
3383 shader_addline(buffer
, "MOV chroma.r, temp.a;\n");
3385 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
3386 * No need to clamp because we're just reusing the already clamped value from above
3388 if(textype
== GL_TEXTURE_2D
) {
3389 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
3391 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
3393 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
3394 shader_addline(buffer
, "MOV chroma.g, temp.a;\n");
3396 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
3397 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
3398 * values due to filtering
3400 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3401 if(textype
== GL_TEXTURE_2D
) {
3402 /* Multiply the y coordinate by 2/3 and clamp it */
3403 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
3404 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
3405 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
3406 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
3408 /* Reading from texture_rectangles is pretty streightforward, just use the unmodified
3409 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
3412 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
3413 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
3414 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
3421 static GLuint
gen_yuv_shader(IWineD3DDeviceImpl
*device
, enum yuv_fixup yuv_fixup
, GLenum textype
)
3424 SHADER_BUFFER buffer
;
3425 char luminance_component
;
3426 struct arbfp_blit_priv
*priv
= (struct arbfp_blit_priv
*) device
->blit_priv
;
3431 buffer
.newline
= TRUE
;
3432 buffer
.buffer
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, SHADER_PGMSIZE
);
3435 GL_EXTCALL(glGenProgramsARB(1, &shader
));
3436 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
3437 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
3438 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3441 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3445 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
3446 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
3447 * two chroma(U and V) values. Each macropixel has two luminance values, one for
3448 * each single pixel it contains, and one U and one V value shared between both
3451 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
3452 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
3453 * take the format into account when generating the read swizzles
3455 * Reading the Y value is streightforward - just sample the texture. The hardware
3456 * takes care of filtering in the horizontal and vertical direction.
3458 * Reading the U and V values is harder. We have to avoid filtering horizontally,
3459 * because that would mix the U and V values of one pixel or two adjacent pixels.
3460 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
3461 * regardless of the filtering setting. Vertical filtering works automatically
3462 * though - the U and V values of two rows are mixed nicely.
3464 * Appart of avoiding filtering issues, the code has to know which value it just
3465 * read, and where it can find the other one. To determine this, it checks if
3466 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
3468 * Handling horizontal filtering of U and V values requires reading a 2nd pair
3469 * of pixels, extracting U and V and mixing them. This is not implemented yet.
3471 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
3472 * with width / 2. This way one read gives all 3 values, finding U and V is easy
3473 * in an unfiltered situation. Finding the luminance on the other hand requires
3474 * finding out if it is an odd or even pixel. The real drawback of this approach
3475 * is filtering. This would have to be emulated completely in the shader, reading
3476 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
3477 * vertically. Beyond that it would require adjustments to the texture handling
3478 * code to deal with the width scaling
3480 shader_addline(&buffer
, "!!ARBfp1.0\n");
3481 shader_addline(&buffer
, "TEMP luminance;\n");
3482 shader_addline(&buffer
, "TEMP temp;\n");
3483 shader_addline(&buffer
, "TEMP chroma;\n");
3484 shader_addline(&buffer
, "TEMP texcrd;\n");
3485 shader_addline(&buffer
, "TEMP texcrd2;\n");
3486 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
3487 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
3488 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
3492 case YUV_FIXUP_UYVY
:
3493 case YUV_FIXUP_YUY2
:
3494 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
3496 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3501 case YUV_FIXUP_YV12
:
3502 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
3504 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3510 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
3511 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3515 /* Calculate the final result. Formula is taken from
3516 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
3517 * ranges from -0.5 to 0.5
3519 shader_addline(&buffer
, "SUB chroma.rg, chroma, coef.y;\n");
3521 shader_addline(&buffer
, "MAD result.color.r, chroma.r, yuv_coef.x, luminance.%c;\n", luminance_component
);
3522 shader_addline(&buffer
, "MAD temp.r, -chroma.g, yuv_coef.y, luminance.%c;\n", luminance_component
);
3523 shader_addline(&buffer
, "MAD result.color.g, -chroma.r, yuv_coef.z, temp.r;\n");
3524 shader_addline(&buffer
, "MAD result.color.b, chroma.g, yuv_coef.w, luminance.%c;\n", luminance_component
);
3525 shader_addline(&buffer
, "END\n");
3528 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
3530 if (glGetError() == GL_INVALID_OPERATION
) {
3532 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3533 FIXME("Fragment program error at position %d: %s\n", pos
,
3534 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3536 HeapFree(GetProcessHeap(), 0, buffer
.buffer
);
3541 case YUV_FIXUP_YUY2
:
3542 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yuy2_rect_shader
= shader
;
3543 else priv
->yuy2_2d_shader
= shader
;
3546 case YUV_FIXUP_UYVY
:
3547 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->uyvy_rect_shader
= shader
;
3548 else priv
->uyvy_2d_shader
= shader
;
3551 case YUV_FIXUP_YV12
:
3552 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yv12_rect_shader
= shader
;
3553 else priv
->yv12_2d_shader
= shader
;
3560 static HRESULT
arbfp_blit_set(IWineD3DDevice
*iface
, WINED3DFORMAT fmt
, GLenum textype
, UINT width
, UINT height
) {
3562 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3563 float size
[4] = {width
, height
, 1, 1};
3564 struct arbfp_blit_priv
*priv
= (struct arbfp_blit_priv
*) device
->blit_priv
;
3565 const struct GlPixelFormatDesc
*glDesc
;
3566 enum yuv_fixup yuv_fixup
;
3568 getFormatDescEntry(fmt
, &GLINFO_LOCATION
, &glDesc
);
3570 if (!is_yuv_fixup(glDesc
->color_fixup
))
3573 dump_color_fixup_desc(glDesc
->color_fixup
);
3574 /* Don't bother setting up a shader for unconverted formats */
3577 checkGLcall("glEnable(textype)");
3582 yuv_fixup
= get_yuv_fixup(glDesc
->color_fixup
);
3586 case YUV_FIXUP_YUY2
:
3587 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yuy2_rect_shader
: priv
->yuy2_2d_shader
;
3590 case YUV_FIXUP_UYVY
:
3591 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->uyvy_rect_shader
: priv
->uyvy_2d_shader
;
3594 case YUV_FIXUP_YV12
:
3595 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yv12_rect_shader
: priv
->yv12_2d_shader
;
3599 FIXME("Unsupported YUV fixup %#x, not setting a shader\n", yuv_fixup
);
3602 checkGLcall("glEnable(textype)");
3607 if (!shader
) shader
= gen_yuv_shader(device
, yuv_fixup
, textype
);
3610 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
3611 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
3612 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
3613 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
3614 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
3615 checkGLcall("glProgramLocalParameter4fvARB");
3621 static void arbfp_blit_unset(IWineD3DDevice
*iface
) {
3622 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3625 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
3626 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
3627 glDisable(GL_TEXTURE_2D
);
3628 checkGLcall("glDisable(GL_TEXTURE_2D)");
3629 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP
)) {
3630 glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
3631 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
3633 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
3634 glDisable(GL_TEXTURE_RECTANGLE_ARB
);
3635 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
3640 static BOOL
arbfp_blit_color_fixup_supported(struct color_fixup_desc fixup
)
3642 enum yuv_fixup yuv_fixup
;
3644 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
3646 TRACE("Checking support for fixup:\n");
3647 dump_color_fixup_desc(fixup
);
3650 if (is_identity_fixup(fixup
))
3656 /* We only support YUV conversions. */
3657 if (!is_yuv_fixup(fixup
))
3659 TRACE("[FAILED]\n");
3663 yuv_fixup
= get_yuv_fixup(fixup
);
3666 case YUV_FIXUP_YUY2
:
3667 case YUV_FIXUP_UYVY
:
3668 case YUV_FIXUP_YV12
:
3673 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
3674 TRACE("[FAILED]\n");
3679 const struct blit_shader arbfp_blit
= {
3684 arbfp_blit_color_fixup_supported
,
3687 #undef GLINFO_LOCATION