2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
13 * Copyright 2009 Henri Verbeet for CodeWeavers
15 * This library is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU Lesser General Public
17 * License as published by the Free Software Foundation; either
18 * version 2.1 of the License, or (at your option) any later version.
20 * This library is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
23 * Lesser General Public License for more details.
25 * You should have received a copy of the GNU Lesser General Public
26 * License along with this library; if not, write to the Free Software
27 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 #define GLINFO_LOCATION (*gl_info)
44 /* GL locking for state handlers is done by the caller. */
45 static BOOL
need_mova_const(IWineD3DBaseShader
*shader
, const WineD3D_GL_Info
*gl_info
) {
46 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) shader
;
47 if(!This
->baseShader
.reg_maps
.usesmova
) return FALSE
;
48 return !GL_SUPPORT(NV_VERTEX_PROGRAM2_OPTION
);
51 static BOOL
need_helper_const(const WineD3D_GL_Info
*gl_info
) {
52 if(!GL_SUPPORT(NV_VERTEX_PROGRAM
) || /* Need to init colors */
53 gl_info
->arb_vs_offset_limit
|| /* Have to init texcoords */
54 gl_info
->set_texcoord_w
) { /* Load the immval offset */
60 static unsigned int reserved_vs_const(IWineD3DBaseShader
*shader
, const WineD3D_GL_Info
*gl_info
) {
62 /* We use one PARAM for the pos fixup, and in some cases one to load
63 * some immediate values into the shader
65 if(need_helper_const(gl_info
)) ret
++;
66 if(need_mova_const(shader
, gl_info
)) ret
++;
70 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
71 * so upload them above that
73 #define ARB_SHADER_PRIVCONST_BASE (GL_LIMITS(vshader_constantsF) - 1)
74 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
76 /* ARB_program_shader private data */
77 struct shader_arb_priv
{
78 GLuint current_vprogram_id
;
79 GLuint current_fprogram_id
;
80 GLuint depth_blt_vprogram_id
;
81 GLuint depth_blt_fprogram_id
[tex_type_count
];
82 BOOL use_arbfp_fixed_func
;
83 struct hash_table_t
*fragment_shaders
;
92 struct shader_arb_ctx_priv
{
95 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
97 /* GL_NV_vertex_progam2_option or GL_NV_fragment_program_option */
99 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
103 const struct arb_vs_compile_args
*cur_vs_args
;
104 const struct arb_ps_compile_args
*cur_ps_args
;
105 struct list if_frames
;
109 struct arb_ps_compile_args
{
110 struct ps_compile_args super
;
111 DWORD bools
; /* WORD is enough, use DWORD for alignment */
114 struct arb_ps_compiled_shader
{
115 struct arb_ps_compile_args args
;
119 struct arb_pshader_private
{
120 struct arb_ps_compiled_shader
*gl_shaders
;
121 UINT num_gl_shaders
, shader_array_size
;
124 struct arb_vs_compile_args
{
125 struct vs_compile_args super
;
126 DWORD bools
; /* WORD is enough, use DWORD for alignment */
129 struct arb_vs_compiled_shader
{
130 struct arb_vs_compile_args args
;
134 struct arb_vshader_private
{
135 struct arb_vs_compiled_shader
*gl_shaders
;
136 UINT num_gl_shaders
, shader_array_size
;
139 /********************************************************
140 * ARB_[vertex/fragment]_program helper functions follow
141 ********************************************************/
144 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
145 * When constant_list == NULL, it will load all the constants.
147 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
148 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
150 /* GL locking is done by the caller */
151 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl
* This
, const WineD3D_GL_Info
*gl_info
,
152 GLuint target_type
, unsigned int max_constants
, const float *constants
, char *dirty_consts
)
154 local_constant
* lconst
;
158 if (TRACE_ON(d3d_shader
)) {
159 for(i
= 0; i
< max_constants
; i
++) {
160 if(!dirty_consts
[i
]) continue;
161 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
162 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
163 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
166 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
167 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& This
->baseShader
.reg_maps
.shader_version
.major
== 1)
170 for(i
= 0; i
< max_constants
; i
++) {
171 if(!dirty_consts
[i
]) continue;
175 if(constants
[j
+ 0] > 1.0) lcl_const
[0] = 1.0;
176 else if(constants
[j
+ 0] < -1.0) lcl_const
[0] = -1.0;
177 else lcl_const
[0] = constants
[j
+ 0];
179 if(constants
[j
+ 1] > 1.0) lcl_const
[1] = 1.0;
180 else if(constants
[j
+ 1] < -1.0) lcl_const
[1] = -1.0;
181 else lcl_const
[1] = constants
[j
+ 1];
183 if(constants
[j
+ 2] > 1.0) lcl_const
[2] = 1.0;
184 else if(constants
[j
+ 2] < -1.0) lcl_const
[2] = -1.0;
185 else lcl_const
[2] = constants
[j
+ 2];
187 if(constants
[j
+ 3] > 1.0) lcl_const
[3] = 1.0;
188 else if(constants
[j
+ 3] < -1.0) lcl_const
[3] = -1.0;
189 else lcl_const
[3] = constants
[j
+ 3];
191 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
194 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS
)) {
195 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
196 * or just reloading *all* constants at once
198 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
200 for(i
= 0; i
< max_constants
; i
++) {
201 if(!dirty_consts
[i
]) continue;
203 /* Find the next block of dirty constants */
206 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
210 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
213 for(i
= 0; i
< max_constants
; i
++) {
214 if(dirty_consts
[i
]) {
216 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
221 checkGLcall("glProgramEnvParameter4fvARB()");
223 /* Load immediate constants */
224 if(This
->baseShader
.load_local_constsF
) {
225 if (TRACE_ON(d3d_shader
)) {
226 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
227 GLfloat
* values
= (GLfloat
*)lconst
->value
;
228 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
229 values
[0], values
[1], values
[2], values
[3]);
232 /* Immediate constants are clamped for 1.X shaders at loading times */
234 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
235 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
236 ret
= max(ret
, lconst
->idx
+ 1);
237 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
239 checkGLcall("glProgramEnvParameter4fvARB()");
240 return ret
; /* The loaded immediate constants need reloading for the next shader */
242 return 0; /* No constants are dirty now */
247 * Loads the texture dimensions for NP2 fixup into the currently set ARB_[vertex/fragment]_programs.
249 static void shader_arb_load_np2fixup_constants(
250 IWineD3DDevice
* device
,
252 char useVertexShader
) {
253 /* not implemented */
256 static inline void shader_arb_ps_local_constants(IWineD3DDeviceImpl
* deviceImpl
)
258 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
259 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
260 IWineD3DPixelShaderImpl
*psi
= (IWineD3DPixelShaderImpl
*) pshader
;
261 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
264 for(i
= 0; i
< psi
->numbumpenvmatconsts
; i
++)
266 /* The state manager takes care that this function is always called if the bump env matrix changes */
267 const float *data
= (const float *)&stateBlock
->textureState
[(int) psi
->bumpenvmatconst
[i
].texunit
][WINED3DTSS_BUMPENVMAT00
];
268 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->bumpenvmatconst
[i
].const_num
, data
));
270 if (psi
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
272 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
273 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
274 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
275 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
277 const float *scale
= (const float *)&stateBlock
->textureState
[(int) psi
->luminanceconst
[i
].texunit
][WINED3DTSS_BUMPENVLSCALE
];
278 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, psi
->luminanceconst
[i
].const_num
, scale
));
283 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
285 * We only support float constants in ARB at the moment, so don't
286 * worry about the Integers or Booleans
288 /* GL locking is done by the caller (state handler) */
289 static void shader_arb_load_constants(
290 IWineD3DDevice
* device
,
292 char useVertexShader
) {
294 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
295 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
296 const WineD3D_GL_Info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
298 if (useVertexShader
) {
299 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
301 /* Load DirectX 9 float constants for vertex shader */
302 deviceImpl
->highest_dirty_vs_const
= shader_arb_load_constantsF(
303 vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
304 deviceImpl
->highest_dirty_vs_const
,
305 stateBlock
->vertexShaderConstantF
,
306 deviceImpl
->activeContext
->vshader_const_dirty
);
308 /* Upload the position fixup */
309 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, ARB_SHADER_PRIVCONST_POS
, deviceImpl
->posFixup
));
312 if (usePixelShader
) {
313 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
315 /* Load DirectX 9 float constants for pixel shader */
316 deviceImpl
->highest_dirty_ps_const
= shader_arb_load_constantsF(
317 pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
318 deviceImpl
->highest_dirty_ps_const
,
319 stateBlock
->pixelShaderConstantF
,
320 deviceImpl
->activeContext
->pshader_const_dirty
);
321 shader_arb_ps_local_constants(deviceImpl
);
325 static void shader_arb_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
327 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
329 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
330 * context. On a context switch the old context will be fully dirtified */
331 memset(This
->activeContext
->vshader_const_dirty
+ start
, 1,
332 sizeof(*This
->activeContext
->vshader_const_dirty
) * count
);
333 This
->highest_dirty_vs_const
= max(This
->highest_dirty_vs_const
, start
+ count
+ 1);
336 static void shader_arb_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
338 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
340 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
341 * context. On a context switch the old context will be fully dirtified */
342 memset(This
->activeContext
->pshader_const_dirty
+ start
, 1,
343 sizeof(*This
->activeContext
->pshader_const_dirty
) * count
);
344 This
->highest_dirty_ps_const
= max(This
->highest_dirty_ps_const
, start
+ count
+ 1);
347 static DWORD
*local_const_mapping(IWineD3DBaseShaderImpl
*This
)
351 const local_constant
*lconst
;
353 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) return NULL
;
355 ret
= HeapAlloc(GetProcessHeap(), 0, sizeof(DWORD
) * This
->baseShader
.limits
.constant_float
);
357 ERR("Out of memory\n");
361 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
362 ret
[lconst
->idx
] = idx
++;
367 /* Generate the variable & register declarations for the ARB_vertex_program output target */
368 static void shader_generate_arb_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
369 SHADER_BUFFER
*buffer
, const WineD3D_GL_Info
*gl_info
, DWORD
*lconst_map
)
371 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
372 DWORD i
, cur
, next_local
= 0;
373 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
374 unsigned max_constantsF
;
375 const local_constant
*lconst
;
377 /* In pixel shaders, all private constants are program local, we don't need anything
378 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
379 * If we need a private constant the GL implementation will squeeze it in somewhere
381 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
382 * immediate values. The posFixup is loaded using program.env for now, so always
383 * subtract one from the number of constants. If the shader uses indirect addressing,
384 * account for the helper const too because we have to declare all availabke d3d constants
385 * and don't know which are actually used.
388 max_constantsF
= GL_LIMITS(pshader_constantsF
);
390 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
391 max_constantsF
= GL_LIMITS(vshader_constantsF
) - reserved_vs_const(iface
, gl_info
);
393 max_constantsF
= GL_LIMITS(vshader_constantsF
) - 1;
397 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
398 if (reg_maps
->temporary
[i
])
399 shader_addline(buffer
, "TEMP R%u;\n", i
);
402 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
403 if (reg_maps
->address
[i
])
404 shader_addline(buffer
, "ADDRESS A%d;\n", i
);
407 if(pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3) {
408 for(i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
409 if (reg_maps
->texcoord
[i
] && pshader
)
410 shader_addline(buffer
,"TEMP T%u;\n", i
);
414 /* Load local constants using the program-local space,
415 * this avoids reloading them each time the shader is used
418 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
419 shader_addline(buffer
, "PARAM C%u = program.local[%u];\n", lconst
->idx
,
420 lconst_map
[lconst
->idx
]);
421 next_local
= max(next_local
, lconst_map
[lconst
->idx
] + 1);
425 /* we use the array-based constants array if the local constants are marked for loading,
426 * because then we use indirect addressing, or when the local constant list is empty,
427 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
428 * local constants do not declare the loaded constants as an array because ARB compilers usually
429 * do not optimize unused constants away
431 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
432 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
433 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
434 max_constantsF
, max_constantsF
- 1);
436 for(i
= 0; i
< max_constantsF
; i
++) {
439 mask
= 1 << (i
& 0x1f);
440 if(!shader_constant_is_local(This
, i
) && (This
->baseShader
.reg_maps
.constf
[idx
] & mask
)) {
441 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
446 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
447 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) This
;
448 if(!reg_maps
->bumpmat
[i
]) continue;
450 cur
= ps
->numbumpenvmatconsts
;
451 ps
->bumpenvmatconst
[cur
].const_num
= -1;
452 ps
->bumpenvmatconst
[cur
].texunit
= i
;
453 ps
->luminanceconst
[cur
].const_num
= -1;
454 ps
->luminanceconst
[cur
].texunit
= i
;
456 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
457 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
458 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
459 * textures due to conditional NP2 restrictions)
461 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
462 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
463 * their location is shader dependent anyway and they cannot be loaded globally.
465 ps
->bumpenvmatconst
[cur
].const_num
= next_local
++;
466 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
467 i
, ps
->bumpenvmatconst
[cur
].const_num
);
468 ps
->numbumpenvmatconsts
= cur
+ 1;
470 if(!reg_maps
->luminanceparams
[i
]) continue;
472 ((IWineD3DPixelShaderImpl
*)This
)->luminanceconst
[cur
].const_num
= next_local
++;
473 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
474 i
, ps
->luminanceconst
[cur
].const_num
);
479 static const char * const shift_tab
[] = {
480 "dummy", /* 0 (none) */
481 "coefmul.x", /* 1 (x2) */
482 "coefmul.y", /* 2 (x4) */
483 "coefmul.z", /* 3 (x8) */
484 "coefmul.w", /* 4 (x16) */
485 "dummy", /* 5 (x32) */
486 "dummy", /* 6 (x64) */
487 "dummy", /* 7 (x128) */
488 "dummy", /* 8 (d256) */
489 "dummy", /* 9 (d128) */
490 "dummy", /* 10 (d64) */
491 "dummy", /* 11 (d32) */
492 "coefdiv.w", /* 12 (d16) */
493 "coefdiv.z", /* 13 (d8) */
494 "coefdiv.y", /* 14 (d4) */
495 "coefdiv.x" /* 15 (d2) */
498 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
499 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
501 char *ptr
= write_mask
;
503 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
506 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
507 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
508 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
509 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
515 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
517 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
518 * but addressed as "rgba". To fix this we need to swap the register's x
519 * and z components. */
520 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
521 char *ptr
= swizzle_str
;
523 /* swizzle bits fields: wwzzyyxx */
524 DWORD swizzle
= param
->swizzle
;
525 DWORD swizzle_x
= swizzle
& 0x03;
526 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
527 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
528 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
530 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
531 * generate a swizzle string. Unless we need to our own swizzling. */
532 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
535 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
536 *ptr
++ = swizzle_chars
[swizzle_x
];
538 *ptr
++ = swizzle_chars
[swizzle_x
];
539 *ptr
++ = swizzle_chars
[swizzle_y
];
540 *ptr
++ = swizzle_chars
[swizzle_z
];
541 *ptr
++ = swizzle_chars
[swizzle_w
];
548 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
550 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
551 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
553 if(strcmp(priv
->addr_reg
, src
) == 0) return;
555 strcpy(priv
->addr_reg
, src
);
556 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
559 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
560 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
562 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
563 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
565 /* oPos, oFog and oPts in D3D */
566 static const char * const rastout_reg_names
[] = {"TMP_OUT", "result.fogcoord", "result.pointsize"};
567 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
568 BOOL pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
569 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
575 case WINED3DSPR_TEMP
:
576 sprintf(register_name
, "R%u", reg
->idx
);
579 case WINED3DSPR_INPUT
:
582 if (reg
->idx
== 0) strcpy(register_name
, "fragment.color.primary");
583 else strcpy(register_name
, "fragment.color.secondary");
587 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
)) *is_color
= TRUE
;
588 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
);
592 case WINED3DSPR_CONST
:
593 if (!pshader
&& reg
->rel_addr
)
596 UINT rel_offset
= ((IWineD3DVertexShaderImpl
*)This
)->rel_offset
;
597 if(This
->baseShader
.reg_maps
.shader_version
.major
< 2) {
598 sprintf(rel_reg
, "A0.x");
600 shader_arb_get_src_param(ins
, reg
->rel_addr
, 0, rel_reg
);
601 if(ctx
->target_version
== ARB
) {
602 shader_arb_request_a0(ins
, rel_reg
);
603 sprintf(rel_reg
, "A0.x");
606 if (reg
->idx
>= rel_offset
)
607 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
- rel_offset
);
609 sprintf(register_name
, "C[%s - %u]", rel_reg
, -reg
->idx
+ rel_offset
);
613 if (This
->baseShader
.reg_maps
.usesrelconstF
)
614 sprintf(register_name
, "C[%u]", reg
->idx
);
616 sprintf(register_name
, "C%u", reg
->idx
);
620 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
622 if(This
->baseShader
.reg_maps
.shader_version
.major
== 1 &&
623 This
->baseShader
.reg_maps
.shader_version
.minor
<= 3) {
624 /* In ps <= 1.3, Tx is a temporary register as destination to all instructions,
625 * and as source to most instructions. For some instructions it is the texcoord
626 * input. Those instructions know about the special use
628 sprintf(register_name
, "T%u", reg
->idx
);
630 /* in ps 1.4 and 2.x Tx is always a (read-only) varying */
631 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
);
636 if(This
->baseShader
.reg_maps
.shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
638 sprintf(register_name
, "A%u", reg
->idx
);
642 sprintf(register_name
, "A%u_SHADOW", reg
->idx
);
647 case WINED3DSPR_COLOROUT
:
650 if(ctx
->cur_ps_args
->super
.srgb_correction
)
652 strcpy(register_name
, "TMP_COLOR");
656 strcpy(register_name
, "result.color");
661 /* TODO: See GL_ARB_draw_buffers */
662 FIXME("Unsupported write to render target %u\n", reg
->idx
);
663 sprintf(register_name
, "unsupported_register");
667 case WINED3DSPR_RASTOUT
:
668 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
]);
671 case WINED3DSPR_DEPTHOUT
:
672 strcpy(register_name
, "result.depth");
675 case WINED3DSPR_ATTROUT
:
676 if (pshader
) sprintf(register_name
, "oD[%u]", reg
->idx
);
677 else if (reg
->idx
== 0) strcpy(register_name
, "result.color.primary");
678 else strcpy(register_name
, "result.color.secondary");
681 case WINED3DSPR_TEXCRDOUT
:
682 if (pshader
) sprintf(register_name
, "oT[%u]", reg
->idx
);
683 else sprintf(register_name
, "result.texcoord[%u]", reg
->idx
);
687 FIXME("Unhandled register type %#x[%u]\n", reg
->type
, reg
->idx
);
688 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
);
693 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
694 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
696 char register_name
[255];
700 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
701 strcpy(str
, register_name
);
703 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
704 strcat(str
, write_mask
);
707 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
709 switch(channel_source
)
711 case CHANNEL_SOURCE_ZERO
: return "0";
712 case CHANNEL_SOURCE_ONE
: return "1";
713 case CHANNEL_SOURCE_X
: return "x";
714 case CHANNEL_SOURCE_Y
: return "y";
715 case CHANNEL_SOURCE_Z
: return "z";
716 case CHANNEL_SOURCE_W
: return "w";
718 FIXME("Unhandled channel source %#x\n", channel_source
);
723 static void gen_color_correction(SHADER_BUFFER
*buffer
, const char *reg
, DWORD dst_mask
,
724 const char *one
, const char *two
, struct color_fixup_desc fixup
)
728 if (is_yuv_fixup(fixup
))
730 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
731 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
736 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
737 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
738 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
739 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
744 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
745 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
746 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
750 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
751 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
752 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
753 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
759 char *ptr
= reg_mask
;
761 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
764 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
765 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
766 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
767 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
771 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
775 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
776 const char *dst_str
, const char *coord_reg
, BOOL projected
, BOOL bias
)
778 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
779 DWORD sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
780 const char *tex_type
;
781 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
782 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
783 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
785 switch(sampler_type
) {
791 if(device
->stateBlock
->textures
[sampler_idx
] &&
792 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
797 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
799 if(priv
->cur_ps_args
->super
.np2_fixup
& (1 << sampler_idx
))
801 FIXME("NP2 texcoord fixup is currently not implemented in ARB mode (use GLSL instead).\n");
806 case WINED3DSTT_VOLUME
:
810 case WINED3DSTT_CUBE
:
815 ERR("Unexpected texture type %d\n", sampler_type
);
820 /* Shouldn't be possible, but let's check for it */
821 if(projected
) FIXME("Biased and Projected texture sampling\n");
822 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
823 shader_addline(buffer
, "TXB %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
824 } else if (projected
) {
825 shader_addline(buffer
, "TXP %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
827 shader_addline(buffer
, "TEX %s, %s, texture[%u], %s;\n", dst_str
, coord_reg
, sampler_idx
, tex_type
);
830 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
832 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
833 "one", "coefmul.x", priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
837 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
838 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
840 /* Generate a line that does the input modifier computation and return the input register to use */
841 BOOL is_color
= FALSE
;
845 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
846 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
848 /* Assume a new line will be added */
851 /* Get register name */
852 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
853 shader_arb_get_swizzle(src
, is_color
, swzstr
);
855 switch (src
->modifiers
)
857 case WINED3DSPSM_NONE
:
858 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
861 case WINED3DSPSM_NEG
:
862 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
865 case WINED3DSPSM_BIAS
:
866 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
868 case WINED3DSPSM_BIASNEG
:
869 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
871 case WINED3DSPSM_SIGN
:
872 shader_addline(buffer
, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg
, regstr
);
874 case WINED3DSPSM_SIGNNEG
:
875 shader_addline(buffer
, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg
, regstr
);
877 case WINED3DSPSM_COMP
:
878 shader_addline(buffer
, "SUB T%c, one.x, %s;\n", 'A' + tmpreg
, regstr
);
881 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
883 case WINED3DSPSM_X2NEG
:
884 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
887 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
888 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
891 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
892 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
894 case WINED3DSPSM_ABS
:
895 if(ctx
->target_version
>= NV2
) {
896 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
899 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
902 case WINED3DSPSM_ABSNEG
:
903 if(ctx
->target_version
>= NV2
) {
904 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
906 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
907 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
912 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
916 /* Return modified or original register, with swizzle */
918 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
921 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
924 const char *ret
= "";
925 if (!ins
->dst_count
) return "";
927 mod
= ins
->dst
[0].modifiers
;
928 if(mod
& WINED3DSPDM_SATURATE
) {
930 mod
&= ~WINED3DSPDM_SATURATE
;
932 if(mod
& WINED3DSPDM_PARTIALPRECISION
) {
933 FIXME("Unhandled modifier WINED3DSPDM_PARTIALPRECISION\n");
934 mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
936 if(mod
& WINED3DSPDM_MSAMPCENTROID
) {
937 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
938 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
941 FIXME("Unknown modifiers 0x%08x\n", mod
);
946 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
948 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
949 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
951 char src_name
[2][50];
952 DWORD sampler_code
= dst
->reg
.idx
;
954 shader_arb_get_dst_param(ins
, dst
, dst_name
);
956 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
958 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
959 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
962 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
963 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
964 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
965 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
966 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
968 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
969 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
972 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
974 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
975 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
977 char src_name
[3][50];
978 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
979 ins
->ctx
->reg_maps
->shader_version
.minor
);
981 shader_arb_get_dst_param(ins
, dst
, dst_name
);
982 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
984 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
985 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
)
987 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
989 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
990 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
991 shader_addline(buffer
, "ADD TA, -%s, coefdiv.x;\n", src_name
[0]);
992 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n",
993 shader_arb_get_modifier(ins
), dst_name
, src_name
[1], src_name
[2]);
997 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
999 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1000 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1002 char src_name
[3][50];
1004 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1006 /* Generate input register names (with modifiers) */
1007 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1008 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1009 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1011 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
,
1012 src_name
[0], src_name
[2], src_name
[1]);
1015 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1016 * dst = dot2(src0, src1) + src2 */
1017 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1019 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1020 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1022 char src_name
[3][50];
1023 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1025 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1026 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1027 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1029 if(ctx
->target_version
>= NV3
)
1031 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1032 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1033 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1034 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1036 else if(ctx
->target_version
>= NV2
)
1038 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1039 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1040 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1041 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1043 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1045 * .xyxy and other swizzles that we could get with this are not valid in
1046 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1048 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1049 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1050 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1052 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1054 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1055 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1059 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1060 /* Emulate a DP2 with a DP3 and 0.0. Don't use the dest as temp register, it could be src[1] or src[2]
1061 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1063 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1064 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1065 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1066 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1070 /* Map the opcode 1-to-1 to the GL code */
1071 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1073 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1074 const char *instruction
;
1075 char arguments
[256], dst_str
[50];
1077 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1079 switch (ins
->handler_idx
)
1081 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1082 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1083 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1084 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1085 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1086 case WINED3DSIH_DST
: instruction
= "DST"; break;
1087 case WINED3DSIH_EXP
: instruction
= "EX2"; break;
1088 case WINED3DSIH_EXPP
: instruction
= "EXP"; break;
1089 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1090 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1091 case WINED3DSIH_LOG
: instruction
= "LG2"; break;
1092 case WINED3DSIH_LOGP
: instruction
= "LOG"; break;
1093 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1094 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1095 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1096 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1097 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1098 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1099 case WINED3DSIH_POW
: instruction
= "POW"; break;
1100 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1101 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1102 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1103 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1104 case WINED3DSIH_SGN
: instruction
= "SSG"; break;
1105 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1106 default: instruction
= "";
1107 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1111 /* Note that shader_arb_add_dst_param() adds spaces. */
1112 arguments
[0] = '\0';
1113 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1114 for (i
= 0; i
< ins
->src_count
; ++i
)
1117 strcat(arguments
, ", ");
1118 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1119 strcat(arguments
, operand
);
1121 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1124 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
)
1126 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1127 shader_addline(buffer
, "NOP;\n");
1130 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1132 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1133 BOOL pshader
= shader_is_pshader_version(shader
->baseShader
.reg_maps
.shader_version
.type
);
1134 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1136 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1137 char src0_param
[256];
1139 if(ins
->handler_idx
== WINED3DSIH_MOVA
) {
1140 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1143 if(ctx
->target_version
>= NV2
) {
1144 shader_hw_map2gl(ins
);
1147 tmp_src
.swizzle
= (tmp_src
.swizzle
& 0x3) * 0x55;
1148 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1149 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1151 /* This implements the mova formula used in GLSL. The first two instructions
1152 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1156 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1158 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1159 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1161 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, mova_const.y;\n", write_mask
, src0_param
);
1162 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, mova_const.z, -mova_const.w;\n", write_mask
);
1164 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1165 shader_addline(buffer
, "ADD TA%s, TA, mova_const.x;\n", write_mask
);
1166 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1167 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1169 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1170 } else if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
1171 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)
1172 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1174 src0_param
[0] = '\0';
1175 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
)
1177 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1178 shader_addline(buffer
, "ADD TA.x, %s, helper_const.z;\n", src0_param
);
1179 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1183 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1184 * with more than one component. Thus replicate the first source argument over all
1185 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1186 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1187 tmp_src
.swizzle
= (tmp_src
.swizzle
& 0x3) * 0x55;
1188 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1189 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1192 else if(ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& ins
->dst
[0].reg
.idx
== 0 && pshader
)
1194 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) shader
;
1195 if(ctx
->cur_ps_args
->super
.srgb_correction
&& ps
->color0_mov
)
1197 shader_addline(buffer
, "#mov handled in srgb write code\n");
1200 shader_hw_map2gl(ins
);
1204 shader_hw_map2gl(ins
);
1208 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1210 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1211 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1214 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1215 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1217 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1219 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1221 /* The arb backend doesn't claim ps 2.0 support, but try to eat what the app feeds to us */
1222 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1223 shader_addline(buffer
, "KIL %s;\n", reg_dest
);
1225 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1226 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1228 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1229 * or pass in any temporary register(in shader phase 2)
1231 if(ins
->ctx
->reg_maps
->shader_version
.minor
<= 3) {
1232 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
);
1234 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1236 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1237 shader_addline(buffer
, "KIL TA;\n");
1241 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1243 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1244 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1245 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1246 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1247 ins
->ctx
->reg_maps
->shader_version
.minor
);
1248 BOOL projected
= FALSE
, bias
= FALSE
;
1249 struct wined3d_shader_src_param src
;
1253 DWORD reg_sampler_code
;
1255 /* All versions have a destination register */
1256 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1258 /* 1.0-1.4: Use destination register number as texture code.
1259 2.0+: Use provided sampler number as texure code. */
1260 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1261 reg_sampler_code
= dst
->reg
.idx
;
1263 reg_sampler_code
= ins
->src
[1].reg
.idx
;
1265 /* 1.0-1.3: Use the texcoord varying.
1266 1.4+: Use provided coordinate source register. */
1267 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1268 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1270 /* TEX is the only instruction that can handle DW and DZ natively */
1272 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1273 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1274 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1278 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1279 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1280 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1282 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1285 if(reg_sampler_code
< MAX_TEXTURES
) {
1286 flags
= deviceImpl
->stateBlock
->textureState
[reg_sampler_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1288 if (flags
& WINED3DTTFF_PROJECTED
) {
1292 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1294 DWORD src_mod
= ins
->src
[0].modifiers
;
1295 if (src_mod
== WINED3DSPSM_DZ
) {
1296 /* TXP cannot handle DZ natively, so move the z coordinate to .w. reg_coord is a read-only
1297 * varying register, so we need a temp reg
1299 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
1300 strcpy(reg_coord
, "TA");
1302 } else if(src_mod
== WINED3DSPSM_DW
) {
1306 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) projected
= TRUE
;
1307 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) bias
= TRUE
;
1309 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, projected
, bias
);
1312 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
1314 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1315 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1316 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1317 ins
->ctx
->reg_maps
->shader_version
.minor
);
1320 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1322 DWORD reg
= dst
->reg
.idx
;
1324 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1325 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
1329 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
1330 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1331 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
1335 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
1337 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1338 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1339 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1342 DWORD reg1
= ins
->dst
[0].reg
.idx
;
1346 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
1347 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1348 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
1349 /* Move .x first in case src_str is "TA" */
1350 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
1351 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
1352 flags
= reg1
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg1
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1353 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1356 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
1358 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1360 DWORD reg1
= ins
->dst
[0].reg
.idx
;
1364 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
1365 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1366 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
1367 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
1368 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
1369 shader_hw_sample(ins
, reg1
, dst_str
, "TA", FALSE
, FALSE
);
1372 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
1374 DWORD reg1
= ins
->dst
[0].reg
.idx
;
1378 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
1379 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1380 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
1381 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, FALSE
, FALSE
);
1384 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
1386 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1387 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1388 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1389 char reg_coord
[40], dst_reg
[50], src_reg
[50];
1390 DWORD reg_dest_code
;
1392 /* All versions have a destination register. The Tx where the texture coordinates come
1393 * from is the varying incarnation of the texture register
1395 reg_dest_code
= dst
->reg
.idx
;
1396 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
1397 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
1398 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
1400 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1401 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
1403 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
1404 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
1405 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
1406 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
1408 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1409 * so we can't let the GL handle this.
1411 if (((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
]
1412 & WINED3DTTFF_PROJECTED
) {
1413 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
1414 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
1415 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
1417 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
1420 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", FALSE
, FALSE
);
1422 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
1424 /* No src swizzles are allowed, so this is ok */
1425 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
1426 src_reg
, reg_dest_code
, reg_dest_code
);
1427 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
1431 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
1433 DWORD reg
= ins
->dst
[0].reg
.idx
;
1434 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1435 char src0_name
[50], dst_name
[50];
1437 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
1439 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1440 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
1441 * T<reg+1> register. Use this register to store the calculated vector
1443 tmp_reg
.idx
= reg
+ 1;
1444 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
1445 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
1448 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
1450 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1451 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1453 DWORD reg
= ins
->dst
[0].reg
.idx
;
1454 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1460 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
1461 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
1463 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1464 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1465 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
1466 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1467 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1470 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
1472 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1473 DWORD reg
= ins
->dst
[0].reg
.idx
;
1474 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1475 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1476 char src0_name
[50], dst_name
[50];
1477 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
1480 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
1481 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
1482 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
1484 tmp_reg
.idx
= reg
+ 2 - current_state
->current_row
;
1485 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
1487 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1488 shader_addline(buffer
, "DP3 %s%u.%c, fragment.texcoord[%u], %s;\n",
1489 dst_name
, tmp_reg
.idx
, 'x' + current_state
->current_row
, reg
, src0_name
);
1490 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
1493 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
1495 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1496 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1498 DWORD reg
= ins
->dst
[0].reg
.idx
;
1499 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1500 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1502 char src0_name
[50], dst_name
[50];
1505 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
1506 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1507 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
1509 /* Sample the texture using the calculated coordinates */
1510 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1511 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1512 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1513 current_state
->current_row
= 0;
1516 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
1518 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1519 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1521 DWORD reg
= ins
->dst
[0].reg
.idx
;
1522 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1523 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1529 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
1530 * components for temporary data storage
1532 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
1533 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1534 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
1536 /* Construct the eye-ray vector from w coordinates */
1537 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[0]);
1538 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[1]);
1539 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
1541 /* Calculate reflection vector
1543 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
1544 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
1545 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
1546 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
1547 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
1548 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
1549 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
1551 /* Sample the texture using the calculated coordinates */
1552 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1553 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1554 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1555 current_state
->current_row
= 0;
1558 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
1560 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1561 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1563 DWORD reg
= ins
->dst
[0].reg
.idx
;
1564 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1565 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1572 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1573 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
1574 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
1575 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
1576 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
1578 /* Calculate reflection vector.
1581 * dst_reg.xyz = 2 * --------- * N - E
1584 * Which normalizes the normal vector
1586 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
1587 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
1588 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
1589 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
1590 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
1591 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
1593 /* Sample the texture using the calculated coordinates */
1594 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1595 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1596 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
, FALSE
);
1597 current_state
->current_row
= 0;
1600 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
1602 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1603 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1606 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
1607 * which is essentially an input, is the destination register because it is the first
1608 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
1609 * here(writemasks/swizzles are not valid on texdepth)
1611 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1613 /* According to the msdn, the source register(must be r5) is unusable after
1614 * the texdepth instruction, so we're free to modify it
1616 shader_addline(buffer
, "MIN %s.y, %s.y, one.y;\n", dst_name
, dst_name
);
1618 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1619 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1620 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1622 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
1623 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
1624 shader_addline(buffer
, "MIN TA.x, TA.x, one.x;\n");
1625 shader_addline(buffer
, "MAX result.depth, TA.x, 0.0;\n");
1628 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
1629 * Take a 3-component dot product of the TexCoord[dstreg] and src,
1630 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
1631 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
1633 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1634 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
1638 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
1639 shader_addline(buffer
, "MOV TB, 0.0;\n");
1640 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
1642 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1643 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", FALSE
/* Only one coord, can't be projected */, FALSE
);
1646 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
1647 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
1648 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
1650 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1653 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1655 /* Handle output register */
1656 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1657 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
1658 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
, src0
);
1661 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
1662 * Perform the 3rd row of a 3x3 matrix multiply */
1663 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
1665 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1666 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1667 char dst_str
[50], dst_name
[50];
1671 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1672 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
1673 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
1674 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
, src0
);
1675 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
1678 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
1679 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
1680 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
1681 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
1683 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
1685 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1686 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1687 char src0
[50], dst_name
[50];
1690 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
1691 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
1692 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
, src0
);
1694 /* How to deal with the special case dst_name.g == 0? if r != 0, then
1695 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
1696 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
1698 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
1699 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
1700 shader_addline(buffer
, "MIN %s.x, %s.x, one.x;\n", dst_name
, dst_name
);
1701 shader_addline(buffer
, "MAX result.depth, %s.x, 0.0;\n", dst_name
);
1704 /** Handles transforming all WINED3DSIO_M?x? opcodes for
1705 Vertex/Pixel shaders to ARB_vertex_program codes */
1706 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
1709 int nComponents
= 0;
1710 struct wined3d_shader_dst_param tmp_dst
= {{0}};
1711 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
1712 struct wined3d_shader_instruction tmp_ins
;
1714 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
1716 /* Set constants for the temporary argument */
1717 tmp_ins
.ctx
= ins
->ctx
;
1718 tmp_ins
.dst_count
= 1;
1719 tmp_ins
.dst
= &tmp_dst
;
1720 tmp_ins
.src_count
= 2;
1721 tmp_ins
.src
= tmp_src
;
1723 switch(ins
->handler_idx
)
1725 case WINED3DSIH_M4x4
:
1727 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
1729 case WINED3DSIH_M4x3
:
1731 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
1733 case WINED3DSIH_M3x4
:
1735 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
1737 case WINED3DSIH_M3x3
:
1739 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
1741 case WINED3DSIH_M3x2
:
1743 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
1746 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1750 tmp_dst
= ins
->dst
[0];
1751 tmp_src
[0] = ins
->src
[0];
1752 tmp_src
[1] = ins
->src
[1];
1753 for (i
= 0; i
< nComponents
; i
++) {
1754 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
1755 shader_hw_map2gl(&tmp_ins
);
1756 ++tmp_src
[1].reg
.idx
;
1760 static void shader_hw_rsq_rcp(const struct wined3d_shader_instruction
*ins
)
1762 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1763 const char *instruction
;
1768 switch(ins
->handler_idx
)
1770 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
1771 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
1772 default: instruction
= "";
1773 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1777 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
1778 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
1779 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
1781 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
1787 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
1790 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
1792 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1795 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1796 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1798 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
1799 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
1801 if(pshader
&& priv
->target_version
>= NV3
)
1803 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
1807 shader_addline(buffer
, "DP3 TA, %s, %s;\n", src_name
, src_name
);
1808 shader_addline(buffer
, "RSQ TA, TA.x;\n");
1809 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
1810 shader_addline(buffer
, "MUL%s %s, %s, TA;\n", shader_arb_get_modifier(ins
), dst_name
,
1815 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
1817 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
1818 * must contain fixed constants. So we need a separate function to filter those constants and
1821 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1822 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1823 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1825 char src_name0
[50], src_name1
[50], src_name2
[50];
1828 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
1829 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
1830 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
1831 shader_addline(buffer
, "SCS%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
,
1833 } else if(priv
->target_version
>= NV2
) {
1834 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
1836 /* Sincos writemask must be .x, .y or .xy */
1837 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
1838 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
1839 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
1840 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
1842 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
1843 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
1845 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
1846 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
1848 * The constants we get are:
1850 * +1 +1, -1 -1 +1 +1 -1 -1
1851 * ---- , ---- , ---- , ----- , ----- , ----- , ------
1852 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
1854 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
1858 * (x/2)^4 = x^4 / 16
1859 * (x/2)^5 = x^5 / 32
1862 * To get the final result:
1863 * sin(x) = 2 * sin(x/2) * cos(x/2)
1864 * cos(x) = cos(x/2)^2 - sin(x/2)^2
1865 * (from sin(x+y) and cos(x+y) rules)
1867 * As per MSDN, dst.z is undefined after the operation, and so is
1868 * dst.x and dst.y if they're masked out by the writemask. Ie
1869 * sincos dst.y, src1, c0, c1
1870 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
1871 * vsa.exe also stops with an error if the dest register is the same register as the source
1872 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
1873 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
1875 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
1876 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
1877 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
1879 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
1880 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
1881 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
1882 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
1883 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
1884 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
1888 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
1889 * properly merge that with MULs in the code above?
1890 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
1891 * we can merge the sine and cosine MAD rows to calculate them together.
1893 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
1894 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
1895 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
1896 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
1899 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
1900 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
1901 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
1903 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
1905 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
1906 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
1908 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
1910 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
1911 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
1916 /* GL locking is done by the caller */
1917 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
1919 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
1922 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1924 /* SGN is only valid in vertex shaders */
1925 if(ctx
->target_version
== NV2
) {
1926 shader_hw_map2gl(ins
);
1929 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
1930 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
1932 FIXME("Emulated SGN untested\n");
1933 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
1934 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
1936 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
1937 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
1939 shader_addline(buffer
, "SLT TB, -%s, %s;\n", src_name
, src_name
);
1940 shader_addline(buffer
, "SLT TC, %s, -%s;\n", src_name
, src_name
);
1941 shader_addline(buffer
, "ADD %s, TB, -TC;\n", dst_name
);
1945 static GLuint
create_arb_blt_vertex_program(const WineD3D_GL_Info
*gl_info
)
1947 GLuint program_id
= 0;
1948 const char *blt_vprogram
=
1950 "PARAM c[1] = { { 1, 0.5 } };\n"
1951 "MOV result.position, vertex.position;\n"
1952 "MOV result.color, c[0].x;\n"
1953 "MOV result.texcoord[0], vertex.texcoord[0];\n"
1956 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
1957 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
1958 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_vprogram
), blt_vprogram
));
1960 if (glGetError() == GL_INVALID_OPERATION
) {
1962 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
1963 FIXME("Vertex program error at position %d: %s\n", pos
,
1964 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
1970 /* GL locking is done by the caller */
1971 static GLuint
create_arb_blt_fragment_program(const WineD3D_GL_Info
*gl_info
, enum tex_types tex_type
)
1973 GLuint program_id
= 0;
1974 static const char * const blt_fprograms
[tex_type_count
] =
1981 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
1982 "MOV result.depth.z, R0.x;\n"
1989 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
1990 "MOV result.depth.z, R0.x;\n"
1995 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
1996 "MOV result.depth.z, R0.x;\n"
2000 if (!blt_fprograms
[tex_type
])
2002 FIXME("tex_type %#x not supported\n", tex_type
);
2006 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
2007 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
2008 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_fprograms
[tex_type
]), blt_fprograms
[tex_type
]));
2010 if (glGetError() == GL_INVALID_OPERATION
) {
2012 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
2013 FIXME("Fragment program error at position %d: %s\n", pos
,
2014 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2020 static void arbfp_add_sRGB_correction(SHADER_BUFFER
*buffer
, const char *fragcolor
, const char *tmp1
,
2021 const char *tmp2
, const char *tmp3
) {
2022 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
2024 /* Calculate the > 0.0031308 case */
2025 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts1.z;\n", tmp1
, fragcolor
);
2026 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts1.z;\n", tmp1
, fragcolor
);
2027 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts1.z;\n", tmp1
, fragcolor
);
2028 shader_addline(buffer
, "MUL %s, %s, srgb_consts1.w;\n", tmp1
, tmp1
);
2029 shader_addline(buffer
, "SUB %s, %s, srgb_consts2.x;\n", tmp1
, tmp1
);
2030 /* Calculate the < case */
2031 shader_addline(buffer
, "MUL %s, srgb_consts1.x, %s;\n", tmp2
, fragcolor
);
2032 /* Subtract the comparison value from the fragcolor and use CMP to pick either the > case * or the < case */
2033 shader_addline(buffer
, "SUB %s, %s, srgb_consts1.y;\n", tmp3
, fragcolor
);
2034 shader_addline(buffer
, "CMP result.color.xyz, %s, %s, %s;\n", tmp3
, tmp2
, tmp1
);
2037 /* GL locking is done by the caller */
2038 static GLuint
shader_arb_generate_pshader(IWineD3DPixelShaderImpl
*This
,
2039 SHADER_BUFFER
*buffer
, const struct arb_ps_compile_args
*args
)
2041 const shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
2042 CONST DWORD
*function
= This
->baseShader
.function
;
2043 const WineD3D_GL_Info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
2044 const local_constant
*lconst
;
2047 DWORD
*lconst_map
= local_const_mapping((IWineD3DBaseShaderImpl
*) This
);
2048 struct shader_arb_ctx_priv priv_ctx
;
2050 /* Create the hw ARB shader */
2051 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
2052 priv_ctx
.cur_ps_args
= args
;
2053 list_init(&priv_ctx
.if_frames
);
2055 shader_addline(buffer
, "!!ARBfp1.0\n");
2056 if(GL_SUPPORT(NV_FRAGMENT_PROGRAM2
)) {
2057 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
2058 priv_ctx
.target_version
= NV3
;
2059 } else if(GL_SUPPORT(NV_FRAGMENT_PROGRAM_OPTION
)) {
2060 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
2061 priv_ctx
.target_version
= NV2
;
2063 priv_ctx
.target_version
= ARB
;
2066 if (reg_maps
->shader_version
.major
< 3)
2068 switch(args
->super
.fog
) {
2072 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
2075 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
2078 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
2083 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
2084 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
2085 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
2086 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
2087 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
2088 shader_addline(buffer
, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
2090 if (reg_maps
->shader_version
.major
< 2)
2092 strcpy(fragcolor
, "R0");
2094 if(args
->super
.srgb_correction
) {
2095 if(This
->color0_mov
) {
2096 sprintf(fragcolor
, "R%u", This
->color0_reg
);
2098 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
2099 strcpy(fragcolor
, "TMP_COLOR");
2102 strcpy(fragcolor
, "result.color");
2106 if(args
->super
.srgb_correction
) {
2107 shader_addline(buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
2108 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
2109 shader_addline(buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
2110 srgb_sub_high
, 0.0, 0.0, 0.0);
2113 /* Base Declarations */
2114 shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, lconst_map
);
2116 /* Base Shader Body */
2117 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
2119 if(args
->super
.srgb_correction
) {
2120 arbfp_add_sRGB_correction(buffer
, fragcolor
, "TA", "TB", "TC");
2121 shader_addline(buffer
, "MOV result.color.a, %s;\n", fragcolor
);
2122 } else if(reg_maps
->shader_version
.major
< 2) {
2123 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
2125 shader_addline(buffer
, "END\n");
2127 /* TODO: change to resource.glObjectHandle or something like that */
2128 GL_EXTCALL(glGenProgramsARB(1, &retval
));
2130 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
2131 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
2133 TRACE("Created hw pixel shader, prg=%d\n", retval
);
2134 /* Create the program and check for errors */
2135 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
2136 buffer
->bsize
, buffer
->buffer
));
2138 if (glGetError() == GL_INVALID_OPERATION
) {
2140 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
2141 FIXME("HW PixelShader Error at position %d: %s\n",
2142 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2146 /* Load immediate constants */
2148 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
2149 const float *value
= (const float *)lconst
->value
;
2150 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, lconst_map
[lconst
->idx
], value
));
2151 checkGLcall("glProgramLocalParameter4fvARB");
2153 HeapFree(GetProcessHeap(), 0, lconst_map
);
2159 /* GL locking is done by the caller */
2160 static GLuint
shader_arb_generate_vshader(IWineD3DVertexShaderImpl
*This
,
2161 SHADER_BUFFER
*buffer
, const struct arb_vs_compile_args
*args
)
2163 const shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
2164 CONST DWORD
*function
= This
->baseShader
.function
;
2165 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
2166 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
2167 const local_constant
*lconst
;
2169 DWORD
*lconst_map
= local_const_mapping((IWineD3DBaseShaderImpl
*) This
);
2170 struct shader_arb_ctx_priv priv_ctx
;
2172 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
2173 priv_ctx
.cur_vs_args
= args
;
2174 list_init(&priv_ctx
.if_frames
);
2176 /* Create the hw ARB shader */
2177 shader_addline(buffer
, "!!ARBvp1.0\n");
2179 if(GL_SUPPORT(NV_VERTEX_PROGRAM2_OPTION
)) {
2180 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
2181 priv_ctx
.target_version
= NV2
;
2183 priv_ctx
.target_version
= ARB
;
2186 shader_addline(buffer
, "TEMP TMP_OUT;\n");
2187 if(need_helper_const(gl_info
)) {
2188 shader_addline(buffer
, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This
->rel_offset
);
2190 if(need_mova_const((IWineD3DBaseShader
*) This
, gl_info
)) {
2191 shader_addline(buffer
, "PARAM mova_const = { 0.5, 0.0, 2.0, 1.0 };\n");
2192 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
2195 shader_addline(buffer
, "TEMP TA;\n");
2197 /* Base Declarations */
2198 shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
, lconst_map
);
2200 /* We need a constant to fixup the final position */
2201 shader_addline(buffer
, "PARAM posFixup = program.env[%d];\n", ARB_SHADER_PRIVCONST_POS
);
2203 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
2204 * for output parameters. D3D in theory does not do that either, but some applications depend on a
2205 * proper initialization of the secondary color, and programs using the fixed function pipeline without
2206 * a replacement shader depend on the texcoord.w being set properly.
2208 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
2209 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
2210 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
2211 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
2212 * this can eat a number of instructions, so skip it unless this cap is set as well
2214 if(!GL_SUPPORT(NV_VERTEX_PROGRAM
)) {
2215 shader_addline(buffer
, "MOV result.color.secondary, -helper_const.wwwy;\n");
2217 if((GLINFO_LOCATION
).set_texcoord_w
&& !device
->frag_pipe
->ffp_proj_control
) {
2219 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
2220 if(This
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
2221 This
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
2222 shader_addline(buffer
, "MOV result.texcoord[%u].w, -helper_const.y;\n", i
);
2228 /* Base Shader Body */
2229 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
2231 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
2232 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
2233 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
2234 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
2236 if(args
->super
.fog_src
== VS_FOG_Z
) {
2237 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
2238 } else if (!reg_maps
->fog
) {
2239 /* posFixup.x is always 1.0, so we can savely use it */
2240 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
2243 /* Write the final position.
2245 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2246 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2247 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2248 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2250 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
2251 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
2252 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
2254 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2255 * and the glsl equivalent
2257 if(need_helper_const(gl_info
)) {
2258 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2260 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
2261 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
2264 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
2266 shader_addline(buffer
, "END\n");
2268 /* TODO: change to resource.glObjectHandle or something like that */
2269 GL_EXTCALL(glGenProgramsARB(1, &ret
));
2271 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
2272 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
2274 TRACE("Created hw vertex shader, prg=%d\n", ret
);
2275 /* Create the program and check for errors */
2276 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
2277 buffer
->bsize
, buffer
->buffer
));
2279 if (glGetError() == GL_INVALID_OPERATION
) {
2281 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
2282 FIXME("HW VertexShader Error at position %d: %s\n",
2283 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
2286 /* Load immediate constants */
2288 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
2289 const float *value
= (const float *)lconst
->value
;
2290 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, lconst_map
[lconst
->idx
], value
));
2294 HeapFree(GetProcessHeap(), 0, lconst_map
);
2299 /* GL locking is done by the caller */
2300 static GLuint
find_arb_pshader(IWineD3DPixelShaderImpl
*shader
, const struct arb_ps_compile_args
*args
)
2304 struct arb_ps_compiled_shader
*new_array
;
2305 SHADER_BUFFER buffer
;
2306 struct arb_pshader_private
*shader_data
;
2309 if(!shader
->backend_priv
) {
2310 shader
->backend_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
2312 shader_data
= shader
->backend_priv
;
2314 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
2315 * so a linear search is more performant than a hashmap or a binary search
2316 * (cache coherency etc)
2318 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
2319 if(memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)) == 0) {
2320 return shader_data
->gl_shaders
[i
].prgId
;
2324 TRACE("No matching GL shader found, compiling a new shader\n");
2325 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
2326 if (shader_data
->num_gl_shaders
)
2328 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
2329 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
2330 new_size
* sizeof(*shader_data
->gl_shaders
));
2332 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
2337 ERR("Out of memory\n");
2340 shader_data
->gl_shaders
= new_array
;
2341 shader_data
->shader_array_size
= new_size
;
2344 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
2346 pixelshader_update_samplers(&shader
->baseShader
.reg_maps
,
2347 ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->stateBlock
->textures
);
2349 shader_buffer_init(&buffer
);
2350 ret
= shader_arb_generate_pshader(shader
, &buffer
, args
);
2351 shader_buffer_free(&buffer
);
2352 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
2357 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
2358 const DWORD use_map
) {
2359 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
2360 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
2361 return stored
->bools
== new->bools
;
2364 static GLuint
find_arb_vshader(IWineD3DVertexShaderImpl
*shader
, const struct arb_vs_compile_args
*args
)
2368 struct arb_vs_compiled_shader
*new_array
;
2369 DWORD use_map
= ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->strided_streams
.use_map
;
2370 SHADER_BUFFER buffer
;
2371 struct arb_vshader_private
*shader_data
;
2374 if(!shader
->backend_priv
) {
2375 shader
->backend_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
2377 shader_data
= shader
->backend_priv
;
2379 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
2380 * so a linear search is more performant than a hashmap or a binary search
2381 * (cache coherency etc)
2383 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
2384 if(vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
, use_map
)) {
2385 return shader_data
->gl_shaders
[i
].prgId
;
2389 TRACE("No matching GL shader found, compiling a new shader\n");
2391 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
2392 if (shader_data
->num_gl_shaders
)
2394 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
2395 new_array
= HeapReAlloc(GetProcessHeap(), 0, shader_data
->gl_shaders
,
2396 new_size
* sizeof(*shader_data
->gl_shaders
));
2398 new_array
= HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data
->gl_shaders
));
2403 ERR("Out of memory\n");
2406 shader_data
->gl_shaders
= new_array
;
2407 shader_data
->shader_array_size
= new_size
;
2410 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
2412 shader_buffer_init(&buffer
);
2413 ret
= shader_arb_generate_vshader(shader
, &buffer
, args
);
2414 shader_buffer_free(&buffer
);
2415 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++].prgId
= ret
;
2420 static inline void find_arb_ps_compile_args(IWineD3DPixelShaderImpl
*shader
, IWineD3DStateBlockImpl
*stateblock
,
2421 struct arb_ps_compile_args
*args
)
2424 find_ps_compile_args(shader
, stateblock
, &args
->super
);
2426 /* This forces all local boolean constants to 1 to make them stateblock independent */
2427 args
->bools
= shader
->baseShader
.reg_maps
.local_bool_consts
;
2429 for(i
= 0; i
< MAX_CONST_B
; i
++)
2431 if(stateblock
->pixelShaderConstantB
[i
]) args
->bools
|= ( 1 << i
);
2436 static inline void find_arb_vs_compile_args(IWineD3DVertexShaderImpl
*shader
, IWineD3DStateBlockImpl
*stateblock
,
2437 struct arb_vs_compile_args
*args
)
2440 find_vs_compile_args(shader
, stateblock
, &args
->super
);
2442 /* This forces all local boolean constants to 1 to make them stateblock independent */
2443 args
->bools
= shader
->baseShader
.reg_maps
.local_bool_consts
;
2445 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
2446 for(i
= 0; i
< MAX_CONST_B
; i
++)
2448 if(stateblock
->vertexShaderConstantB
[i
]) args
->bools
|= ( 1 << i
);
2453 /* GL locking is done by the caller */
2454 static void shader_arb_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
2455 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
2456 struct shader_arb_priv
*priv
= This
->shader_priv
;
2457 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
2460 struct arb_vs_compile_args compile_args
;
2462 TRACE("Using vertex shader\n");
2463 find_arb_vs_compile_args((IWineD3DVertexShaderImpl
*) This
->stateBlock
->vertexShader
, This
->stateBlock
, &compile_args
);
2464 priv
->current_vprogram_id
= find_arb_vshader((IWineD3DVertexShaderImpl
*) This
->stateBlock
->vertexShader
, &compile_args
);
2466 /* Bind the vertex program */
2467 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
2468 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
2470 /* Enable OpenGL vertex programs */
2471 glEnable(GL_VERTEX_PROGRAM_ARB
);
2472 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
2473 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
2474 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
2475 priv
->current_vprogram_id
= 0;
2476 glDisable(GL_VERTEX_PROGRAM_ARB
);
2477 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
2481 struct arb_ps_compile_args compile_args
;
2482 TRACE("Using pixel shader\n");
2483 find_arb_ps_compile_args((IWineD3DPixelShaderImpl
*) This
->stateBlock
->pixelShader
, This
->stateBlock
, &compile_args
);
2484 priv
->current_fprogram_id
= find_arb_pshader((IWineD3DPixelShaderImpl
*) This
->stateBlock
->pixelShader
,
2487 /* Bind the fragment program */
2488 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
2489 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
2491 if(!priv
->use_arbfp_fixed_func
) {
2492 /* Enable OpenGL fragment programs */
2493 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
2494 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
2496 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
2498 shader_arb_ps_local_constants(This
);
2499 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
) && !priv
->use_arbfp_fixed_func
) {
2500 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
2501 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
2502 * replacement shader
2504 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
2505 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2506 priv
->current_fprogram_id
= 0;
2510 /* GL locking is done by the caller */
2511 static void shader_arb_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
2512 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
2513 struct shader_arb_priv
*priv
= This
->shader_priv
;
2514 GLuint
*blt_fprogram
= &priv
->depth_blt_fprogram_id
[tex_type
];
2515 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
2517 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
2518 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
2519 glEnable(GL_VERTEX_PROGRAM_ARB
);
2521 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
);
2522 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
2523 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
2526 /* GL locking is done by the caller */
2527 static void shader_arb_deselect_depth_blt(IWineD3DDevice
*iface
) {
2528 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
2529 struct shader_arb_priv
*priv
= This
->shader_priv
;
2530 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
2532 if (priv
->current_vprogram_id
) {
2533 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
2534 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
2536 glEnable(GL_VERTEX_PROGRAM_ARB
);
2537 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
2539 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
2541 glDisable(GL_VERTEX_PROGRAM_ARB
);
2542 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
2545 if (priv
->current_fprogram_id
) {
2546 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
2547 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
2549 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
2550 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
2552 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
2554 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
2555 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2559 static void shader_arb_destroy(IWineD3DBaseShader
*iface
) {
2560 IWineD3DBaseShaderImpl
*baseShader
= (IWineD3DBaseShaderImpl
*) iface
;
2561 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)baseShader
->baseShader
.device
;
2562 const WineD3D_GL_Info
*gl_info
= &device
->adapter
->gl_info
;
2564 ActivateContext(device
, device
->lastActiveRenderTarget
, CTXUSAGE_RESOURCELOAD
);
2566 if (shader_is_pshader_version(baseShader
->baseShader
.reg_maps
.shader_version
.type
))
2568 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
2569 struct arb_pshader_private
*shader_data
= This
->backend_priv
;
2572 if(!shader_data
) return; /* This can happen if a shader was never compiled */
2574 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
2575 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
2576 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
2579 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
2580 HeapFree(GetProcessHeap(), 0, shader_data
);
2581 This
->backend_priv
= NULL
;
2583 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*) iface
;
2584 struct arb_vshader_private
*shader_data
= This
->backend_priv
;
2587 if(!shader_data
) return; /* This can happen if a shader was never compiled */
2589 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
2590 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
2591 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
2594 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
2595 HeapFree(GetProcessHeap(), 0, shader_data
);
2596 This
->backend_priv
= NULL
;
2600 static HRESULT
shader_arb_alloc(IWineD3DDevice
*iface
) {
2601 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
2602 This
->shader_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
2606 static void shader_arb_free(IWineD3DDevice
*iface
) {
2607 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
2608 const WineD3D_GL_Info
*gl_info
= &This
->adapter
->gl_info
;
2609 struct shader_arb_priv
*priv
= This
->shader_priv
;
2613 if(priv
->depth_blt_vprogram_id
) {
2614 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
2616 for (i
= 0; i
< tex_type_count
; ++i
) {
2617 if (priv
->depth_blt_fprogram_id
[i
]) {
2618 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id
[i
]));
2623 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
2626 static BOOL
shader_arb_dirty_const(IWineD3DDevice
*iface
) {
2630 static void shader_arb_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct shader_caps
*pCaps
)
2632 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
2633 * then overwrite the shader specific ones
2635 none_shader_backend
.shader_get_caps(devtype
, gl_info
, pCaps
);
2637 if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
2638 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(1,1);
2639 TRACE_(d3d_caps
)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
2640 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
) - 1;
2643 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
)) {
2644 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(1,4);
2645 pCaps
->PixelShader1xMaxValue
= 8.0;
2646 TRACE_(d3d_caps
)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
2647 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
);
2650 pCaps
->VSClipping
= FALSE
; /* TODO: GL_NV_vertex_program2_option provides this */
2653 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
2655 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
2657 TRACE("Checking support for color_fixup:\n");
2658 dump_color_fixup_desc(fixup
);
2661 /* We support everything except YUV conversions. */
2662 if (!is_yuv_fixup(fixup
))
2668 TRACE("[FAILED]\n");
2672 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
2674 char write_mask
[20], regstr
[50];
2675 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2676 BOOL is_color
= FALSE
;
2677 const struct wined3d_shader_dst_param
*dst
;
2679 if (!ins
->dst_count
) return;
2683 if(shift
== 0) return; /* Saturate alone is handled by the instructions */
2685 shader_arb_get_write_mask(ins
, dst
, write_mask
);
2686 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
2688 /* Generate a line that does the output modifier computation
2689 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
2690 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
2692 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
2693 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
2696 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
2698 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
2699 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
2700 /* WINED3DSIH_BEM */ pshader_hw_bem
,
2701 /* WINED3DSIH_BREAK */ NULL
,
2702 /* WINED3DSIH_BREAKC */ NULL
,
2703 /* WINED3DSIH_BREAKP */ NULL
,
2704 /* WINED3DSIH_CALL */ NULL
,
2705 /* WINED3DSIH_CALLNZ */ NULL
,
2706 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
2707 /* WINED3DSIH_CND */ pshader_hw_cnd
,
2708 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
2709 /* WINED3DSIH_DCL */ NULL
,
2710 /* WINED3DSIH_DEF */ NULL
,
2711 /* WINED3DSIH_DEFB */ NULL
,
2712 /* WINED3DSIH_DEFI */ NULL
,
2713 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
2714 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
2715 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
2716 /* WINED3DSIH_DST */ shader_hw_map2gl
,
2717 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
2718 /* WINED3DSIH_DSY */ NULL
,
2719 /* WINED3DSIH_ELSE */ NULL
,
2720 /* WINED3DSIH_ENDIF */ NULL
,
2721 /* WINED3DSIH_ENDLOOP */ NULL
,
2722 /* WINED3DSIH_ENDREP */ NULL
,
2723 /* WINED3DSIH_EXP */ shader_hw_map2gl
,
2724 /* WINED3DSIH_EXPP */ shader_hw_map2gl
,
2725 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
2726 /* WINED3DSIH_IF */ NULL
,
2727 /* WINED3DSIH_IFC */ NULL
,
2728 /* WINED3DSIH_LABEL */ NULL
,
2729 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
2730 /* WINED3DSIH_LOG */ shader_hw_map2gl
,
2731 /* WINED3DSIH_LOGP */ shader_hw_map2gl
,
2732 /* WINED3DSIH_LOOP */ NULL
,
2733 /* WINED3DSIH_LRP */ shader_hw_map2gl
,
2734 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
2735 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
2736 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
2737 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
2738 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
2739 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
2740 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
2741 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
2742 /* WINED3DSIH_MOV */ shader_hw_mov
,
2743 /* WINED3DSIH_MOVA */ shader_hw_mov
,
2744 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
2745 /* WINED3DSIH_NOP */ shader_hw_nop
,
2746 /* WINED3DSIH_NRM */ shader_hw_nrm
,
2747 /* WINED3DSIH_PHASE */ NULL
,
2748 /* WINED3DSIH_POW */ shader_hw_map2gl
,
2749 /* WINED3DSIH_RCP */ shader_hw_rsq_rcp
,
2750 /* WINED3DSIH_REP */ NULL
,
2751 /* WINED3DSIH_RET */ NULL
,
2752 /* WINED3DSIH_RSQ */ shader_hw_rsq_rcp
,
2753 /* WINED3DSIH_SETP */ NULL
,
2754 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
2755 /* WINED3DSIH_SGN */ shader_hw_sgn
,
2756 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
2757 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
2758 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
2759 /* WINED3DSIH_TEX */ pshader_hw_tex
,
2760 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
2761 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
2762 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
2763 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
2764 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
2765 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
2766 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
2767 /* WINED3DSIH_TEXLDD */ NULL
,
2768 /* WINED3DSIH_TEXLDL */ NULL
,
2769 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
2770 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
2771 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
2772 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
2773 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
2774 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
2775 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
2776 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
2777 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
2778 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
2779 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
2780 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
2783 static inline BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
, IWineD3DBaseShaderImpl
*This
, DWORD idx
)
2785 BOOL vshader
= shader_is_vshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
2787 WORD flag
= (1 << idx
);
2788 const local_constant
*constant
;
2789 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2791 if(This
->baseShader
.reg_maps
.local_bool_consts
& flag
)
2793 /* What good is a if(bool) with a hardcoded local constant? I don't know, but handle it */
2794 LIST_FOR_EACH_ENTRY(constant
, &This
->baseShader
.constantsB
, local_constant
, entry
)
2796 if (constant
->idx
== idx
)
2798 return constant
->value
[0];
2801 ERR("Local constant not found\n");
2806 if(vshader
) bools
= priv
->cur_vs_args
->bools
;
2807 else bools
= priv
->cur_ps_args
->bools
;
2808 return bools
& flag
;
2812 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
2813 SHADER_HANDLER hw_fct
;
2814 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2815 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2816 struct if_frame
*if_frame
;
2817 SHADER_BUFFER
*buffer
= ins
->ctx
->buffer
;
2820 if(ins
->handler_idx
== WINED3DSIH_IF
)
2822 if_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*if_frame
));
2823 list_add_head(&priv
->if_frames
, &if_frame
->entry
);
2825 if(!priv
->muted
&& get_bool_const(ins
, This
, ins
->src
[0].reg
.idx
) == FALSE
)
2827 shader_addline(buffer
, "#if(FALSE){\n");
2829 if_frame
->muting
= TRUE
;
2831 else shader_addline(buffer
, "#if(TRUE) {\n");
2833 return; /* Instruction is handled */
2835 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
2837 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
2838 if_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*if_frame
));
2839 if_frame
->ifc
= TRUE
;
2840 list_add_head(&priv
->if_frames
, &if_frame
->entry
);
2842 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
2844 struct list
*e
= list_head(&priv
->if_frames
);
2845 if_frame
= LIST_ENTRY(e
, struct if_frame
, entry
);
2847 if(if_frame
->ifc
== FALSE
)
2849 shader_addline(buffer
, "#} else {\n");
2850 if(!priv
->muted
&& !if_frame
->muting
)
2853 if_frame
->muting
= TRUE
;
2855 else if(if_frame
->muting
) priv
->muted
= FALSE
;
2856 return; /* Instruction is handled. */
2858 /* In case of an ifc, generate a HW shader instruction */
2860 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
2862 struct list
*e
= list_head(&priv
->if_frames
);
2863 if_frame
= LIST_ENTRY(e
, struct if_frame
, entry
);
2867 shader_addline(buffer
, "#} endif\n");
2868 if(if_frame
->muting
) priv
->muted
= FALSE
;
2869 list_remove(&if_frame
->entry
);
2870 HeapFree(GetProcessHeap(), 0, if_frame
);
2871 return; /* Instruction is handled */
2875 list_remove(&if_frame
->entry
);
2876 HeapFree(GetProcessHeap(), 0, if_frame
);
2877 /* ifc - generate a hw endif */
2881 if(priv
->muted
) return;
2883 /* Select handler */
2884 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
2886 /* Unhandled opcode */
2889 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
2894 shader_arb_add_instruction_modifiers(ins
);
2897 const shader_backend_t arb_program_shader_backend
= {
2898 shader_arb_handle_instruction
,
2900 shader_arb_select_depth_blt
,
2901 shader_arb_deselect_depth_blt
,
2902 shader_arb_update_float_vertex_constants
,
2903 shader_arb_update_float_pixel_constants
,
2904 shader_arb_load_constants
,
2905 shader_arb_load_np2fixup_constants
,
2909 shader_arb_dirty_const
,
2910 shader_arb_get_caps
,
2911 shader_arb_color_fixup_supported
,
2914 /* ARB_fragment_program fixed function pipeline replacement definitions */
2915 #define ARB_FFP_CONST_TFACTOR 0
2916 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
2917 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
2918 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
2919 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
2921 struct arbfp_ffp_desc
2923 struct ffp_frag_desc parent
;
2925 unsigned int num_textures_used
;
2928 static void arbfp_enable(IWineD3DDevice
*iface
, BOOL enable
) {
2931 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
2932 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
2934 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
2935 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
2940 static HRESULT
arbfp_alloc(IWineD3DDevice
*iface
) {
2941 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
2942 struct shader_arb_priv
*priv
;
2943 /* Share private data between the shader backend and the pipeline replacement, if both
2944 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
2945 * if no pixel shader is bound or not
2947 if(This
->shader_backend
== &arb_program_shader_backend
) {
2948 This
->fragment_priv
= This
->shader_priv
;
2950 This
->fragment_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
2951 if(!This
->fragment_priv
) return E_OUTOFMEMORY
;
2953 priv
= This
->fragment_priv
;
2954 priv
->fragment_shaders
= hash_table_create(ffp_frag_program_key_hash
, ffp_frag_program_key_compare
);
2955 priv
->use_arbfp_fixed_func
= TRUE
;
2959 static void arbfp_free_ffpshader(void *value
, void *gli
) {
2960 const WineD3D_GL_Info
*gl_info
= gli
;
2961 struct arbfp_ffp_desc
*entry_arb
= value
;
2964 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
2965 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
2966 HeapFree(GetProcessHeap(), 0, entry_arb
);
2970 static void arbfp_free(IWineD3DDevice
*iface
) {
2971 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
2972 struct shader_arb_priv
*priv
= This
->fragment_priv
;
2974 hash_table_destroy(priv
->fragment_shaders
, arbfp_free_ffpshader
, &This
->adapter
->gl_info
);
2975 priv
->use_arbfp_fixed_func
= FALSE
;
2977 if(This
->shader_backend
!= &arb_program_shader_backend
) {
2978 HeapFree(GetProcessHeap(), 0, This
->fragment_priv
);
2982 static void arbfp_get_caps(WINED3DDEVTYPE devtype
, const WineD3D_GL_Info
*gl_info
, struct fragment_caps
*caps
)
2984 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
2985 WINED3DTEXOPCAPS_SELECTARG1
|
2986 WINED3DTEXOPCAPS_SELECTARG2
|
2987 WINED3DTEXOPCAPS_MODULATE4X
|
2988 WINED3DTEXOPCAPS_MODULATE2X
|
2989 WINED3DTEXOPCAPS_MODULATE
|
2990 WINED3DTEXOPCAPS_ADDSIGNED2X
|
2991 WINED3DTEXOPCAPS_ADDSIGNED
|
2992 WINED3DTEXOPCAPS_ADD
|
2993 WINED3DTEXOPCAPS_SUBTRACT
|
2994 WINED3DTEXOPCAPS_ADDSMOOTH
|
2995 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
2996 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
2997 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
2998 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
2999 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
3000 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
3001 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
3002 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
3003 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
3004 WINED3DTEXOPCAPS_DOTPRODUCT3
|
3005 WINED3DTEXOPCAPS_MULTIPLYADD
|
3006 WINED3DTEXOPCAPS_LERP
|
3007 WINED3DTEXOPCAPS_BUMPENVMAP
|
3008 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
3010 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
3012 caps
->MaxTextureBlendStages
= 8;
3013 caps
->MaxSimultaneousTextures
= min(GL_LIMITS(fragment_samplers
), 8);
3015 caps
->PrimitiveMiscCaps
|= WINED3DPMISCCAPS_TSSARGTEMP
;
3017 #undef GLINFO_LOCATION
3019 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
3020 static void state_texfactor_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3022 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
3024 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
3025 * application provided constants
3027 if(device
->shader_backend
== &arb_program_shader_backend
) {
3028 if (use_ps(stateblock
)) return;
3030 device
= stateblock
->wineD3DDevice
;
3031 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
3032 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
3035 D3DCOLORTOGLFLOAT4(stateblock
->renderState
[WINED3DRS_TEXTUREFACTOR
], col
);
3036 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
3037 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
3041 static void state_arb_specularenable(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3043 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
3045 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
3046 * application provided constants
3048 if(device
->shader_backend
== &arb_program_shader_backend
) {
3049 if (use_ps(stateblock
)) return;
3051 device
= stateblock
->wineD3DDevice
;
3052 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
3053 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
3056 if(stateblock
->renderState
[WINED3DRS_SPECULARENABLE
]) {
3057 /* The specular color has no alpha */
3058 col
[0] = 1.0; col
[1] = 1.0;
3059 col
[2] = 1.0; col
[3] = 0.0;
3061 col
[0] = 0.0; col
[1] = 0.0;
3062 col
[2] = 0.0; col
[3] = 0.0;
3064 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
3065 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
3068 static void set_bumpmat_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3069 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
3070 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
3073 if (use_ps(stateblock
))
3076 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.bumpmat
[stage
]) {
3077 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
3080 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
3081 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
3085 if(device
->shader_backend
== &arb_program_shader_backend
) {
3086 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
3089 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
3090 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
3091 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
3094 mat
[0][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT00
]);
3095 mat
[0][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT01
]);
3096 mat
[1][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT10
]);
3097 mat
[1][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT11
]);
3099 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
3100 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
3103 static void tex_bumpenvlum_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3104 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
3105 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
3108 if (use_ps(stateblock
))
3111 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.luminanceparams
[stage
]) {
3112 /* The pixel shader has to know the luminance offset. Do a constants update if it
3113 * isn't scheduled anyway
3115 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
3116 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
3120 if(device
->shader_backend
== &arb_program_shader_backend
) {
3121 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
3124 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
3125 device
->activeContext
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
3126 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
3129 param
[0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
]);
3130 param
[1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
]);
3134 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
3135 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
3138 static const char *get_argreg(SHADER_BUFFER
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
) {
3141 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
3143 switch(arg
& WINED3DTA_SELECTMASK
) {
3144 case WINED3DTA_DIFFUSE
:
3145 ret
= "fragment.color.primary"; break;
3147 case WINED3DTA_CURRENT
:
3148 if(stage
== 0) ret
= "fragment.color.primary";
3152 case WINED3DTA_TEXTURE
:
3154 case 0: ret
= "tex0"; break;
3155 case 1: ret
= "tex1"; break;
3156 case 2: ret
= "tex2"; break;
3157 case 3: ret
= "tex3"; break;
3158 case 4: ret
= "tex4"; break;
3159 case 5: ret
= "tex5"; break;
3160 case 6: ret
= "tex6"; break;
3161 case 7: ret
= "tex7"; break;
3162 default: ret
= "unknown texture";
3166 case WINED3DTA_TFACTOR
:
3167 ret
= "tfactor"; break;
3169 case WINED3DTA_SPECULAR
:
3170 ret
= "fragment.color.secondary"; break;
3172 case WINED3DTA_TEMP
:
3173 ret
= "tempreg"; break;
3175 case WINED3DTA_CONSTANT
:
3176 FIXME("Implement perstage constants\n");
3178 case 0: ret
= "const0"; break;
3179 case 1: ret
= "const1"; break;
3180 case 2: ret
= "const2"; break;
3181 case 3: ret
= "const3"; break;
3182 case 4: ret
= "const4"; break;
3183 case 5: ret
= "const5"; break;
3184 case 6: ret
= "const6"; break;
3185 case 7: ret
= "const7"; break;
3186 default: ret
= "unknown constant";
3194 if(arg
& WINED3DTA_COMPLEMENT
) {
3195 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
3196 if(argnum
== 0) ret
= "arg0";
3197 if(argnum
== 1) ret
= "arg1";
3198 if(argnum
== 2) ret
= "arg2";
3200 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
3201 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
3202 if(argnum
== 0) ret
= "arg0";
3203 if(argnum
== 1) ret
= "arg1";
3204 if(argnum
== 2) ret
= "arg2";
3209 static void gen_ffp_instr(SHADER_BUFFER
*buffer
, unsigned int stage
, BOOL color
, BOOL alpha
,
3210 DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
) {
3211 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
3212 unsigned int mul
= 1;
3213 BOOL mul_final_dest
= FALSE
;
3215 if(color
&& alpha
) dstmask
= "";
3216 else if(color
) dstmask
= ".xyz";
3217 else dstmask
= ".w";
3219 if(dst
== tempreg
) dstreg
= "tempreg";
3220 else dstreg
= "ret";
3222 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
3223 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
3224 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
3227 case WINED3DTOP_DISABLE
:
3228 if(stage
== 0) shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
3231 case WINED3DTOP_SELECTARG2
:
3233 case WINED3DTOP_SELECTARG1
:
3234 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
3237 case WINED3DTOP_MODULATE4X
:
3239 case WINED3DTOP_MODULATE2X
:
3241 if(strcmp(dstreg
, "result.color") == 0) {
3243 mul_final_dest
= TRUE
;
3245 case WINED3DTOP_MODULATE
:
3246 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
3249 case WINED3DTOP_ADDSIGNED2X
:
3251 if(strcmp(dstreg
, "result.color") == 0) {
3253 mul_final_dest
= TRUE
;
3255 case WINED3DTOP_ADDSIGNED
:
3256 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
3258 case WINED3DTOP_ADD
:
3259 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
3262 case WINED3DTOP_SUBTRACT
:
3263 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
3266 case WINED3DTOP_ADDSMOOTH
:
3267 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
3268 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
3271 case WINED3DTOP_BLENDCURRENTALPHA
:
3272 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
3273 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
3275 case WINED3DTOP_BLENDFACTORALPHA
:
3276 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
3277 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
3279 case WINED3DTOP_BLENDTEXTUREALPHA
:
3280 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
3281 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
3283 case WINED3DTOP_BLENDDIFFUSEALPHA
:
3284 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
3285 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
3288 case WINED3DTOP_BLENDTEXTUREALPHAPM
:
3289 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
3290 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
3291 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
3294 /* D3DTOP_PREMODULATE ???? */
3296 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR
:
3297 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
3298 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
3300 case WINED3DTOP_MODULATEALPHA_ADDCOLOR
:
3301 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
3303 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA
:
3304 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
3305 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
3307 case WINED3DTOP_MODULATECOLOR_ADDALPHA
:
3308 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
3311 case WINED3DTOP_DOTPRODUCT3
:
3313 if(strcmp(dstreg
, "result.color") == 0) {
3315 mul_final_dest
= TRUE
;
3317 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
3318 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
3319 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
3322 case WINED3DTOP_MULTIPLYADD
:
3323 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
3326 case WINED3DTOP_LERP
:
3327 /* The msdn is not quite right here */
3328 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
3331 case WINED3DTOP_BUMPENVMAP
:
3332 case WINED3DTOP_BUMPENVMAPLUMINANCE
:
3333 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
3337 FIXME("Unhandled texture op %08x\n", op
);
3341 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
3342 } else if(mul
== 4) {
3343 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
3347 /* The stateblock is passed for GLINFO_LOCATION */
3348 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, IWineD3DStateBlockImpl
*stateblock
)
3351 SHADER_BUFFER buffer
;
3352 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
3353 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
3354 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
3355 const char *textype
;
3356 const char *instr
, *sat
;
3357 char colorcor_dst
[8];
3359 DWORD arg0
, arg1
, arg2
;
3360 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
3362 const char *final_combiner_src
= "ret";
3364 /* Find out which textures are read */
3365 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
3366 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) break;
3367 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
3368 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
3369 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
3370 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
3371 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
3372 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
3374 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHA
) tex_read
[stage
] = TRUE
;
3375 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHAPM
) tex_read
[stage
] = TRUE
;
3376 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
) {
3377 bump_used
[stage
] = TRUE
;
3378 tex_read
[stage
] = TRUE
;
3380 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
3381 bump_used
[stage
] = TRUE
;
3382 tex_read
[stage
] = TRUE
;
3383 luminance_used
[stage
] = TRUE
;
3384 } else if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDFACTORALPHA
) {
3385 tfactor_used
= TRUE
;
3388 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
3389 tfactor_used
= TRUE
;
3392 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
3393 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
3394 tempreg_used
= TRUE
;
3397 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) continue;
3398 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
3399 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
3400 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
3401 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
3402 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
3403 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
3405 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
3406 tempreg_used
= TRUE
;
3408 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
3409 tfactor_used
= TRUE
;
3414 shader_buffer_init(&buffer
);
3416 shader_addline(&buffer
, "!!ARBfp1.0\n");
3418 switch(settings
->fog
) {
3419 case FOG_OFF
: break;
3420 case FOG_LINEAR
: shader_addline(&buffer
, "OPTION ARB_fog_linear;\n"); break;
3421 case FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
3422 case FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
3423 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
3426 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
3427 shader_addline(&buffer
, "TEMP TMP;\n");
3428 shader_addline(&buffer
, "TEMP ret;\n");
3429 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
3430 shader_addline(&buffer
, "TEMP arg0;\n");
3431 shader_addline(&buffer
, "TEMP arg1;\n");
3432 shader_addline(&buffer
, "TEMP arg2;\n");
3433 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
3434 if(!tex_read
[stage
]) continue;
3435 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
3436 if(!bump_used
[stage
]) continue;
3437 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
3438 if(!luminance_used
[stage
]) continue;
3439 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
3442 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
3444 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
3446 if(settings
->sRGB_write
) {
3447 shader_addline(&buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
3448 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
3449 shader_addline(&buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
3450 srgb_sub_high
, 0.0, 0.0, 0.0);
3453 /* Generate texture sampling instructions) */
3454 for(stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3DTOP_DISABLE
; stage
++) {
3455 if(!tex_read
[stage
]) continue;
3457 switch(settings
->op
[stage
].tex_type
) {
3458 case tex_1d
: textype
= "1D"; break;
3459 case tex_2d
: textype
= "2D"; break;
3460 case tex_3d
: textype
= "3D"; break;
3461 case tex_cube
: textype
= "CUBE"; break;
3462 case tex_rect
: textype
= "RECT"; break;
3463 default: textype
= "unexpected_textype"; break;
3466 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
||
3467 settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
3473 if(settings
->op
[stage
].projected
== proj_none
) {
3475 } else if(settings
->op
[stage
].projected
== proj_count4
||
3476 settings
->op
[stage
].projected
== proj_count3
) {
3479 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
3484 (settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAP
||
3485 settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
)) {
3486 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
3487 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
3488 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
3489 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
3491 /* with projective textures, texbem only divides the static texture coord, not the displacement,
3492 * so multiply the displacement with the dividing parameter before passing it to TXP
3494 if (settings
->op
[stage
].projected
!= proj_none
) {
3495 if(settings
->op
[stage
].projected
== proj_count4
) {
3496 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
3497 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
3499 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
3500 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
3503 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
3506 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
3507 instr
, sat
, stage
, stage
, textype
);
3508 if(settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
3509 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
3510 stage
- 1, stage
- 1, stage
- 1);
3511 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
3513 } else if(settings
->op
[stage
].projected
== proj_count3
) {
3514 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
3515 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
3516 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
3517 instr
, sat
, stage
, stage
, textype
);
3519 shader_addline(&buffer
, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
3520 instr
, sat
, stage
, stage
, stage
, textype
);
3523 sprintf(colorcor_dst
, "tex%u", stage
);
3524 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
3525 settings
->op
[stage
].color_fixup
);
3528 /* Generate the main shader */
3529 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
3530 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) {
3532 final_combiner_src
= "fragment.color.primary";
3537 if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
3538 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
3539 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
3540 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
3541 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
3542 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
3543 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
3544 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
3545 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
3546 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
3547 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
3548 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
3550 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
&&
3551 settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
&&
3552 settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
&&
3553 settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
3556 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) {
3557 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
3558 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
3559 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
3561 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
3563 } else if(op_equal
) {
3564 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
3565 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
3566 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
3568 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
3569 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
3570 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
3571 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
3572 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
3573 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
3577 if(settings
->sRGB_write
) {
3578 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
3579 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2");
3580 shader_addline(&buffer
, "MOV result.color.w, ret.w;\n");
3582 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
3586 shader_addline(&buffer
, "END\n");
3588 /* Generate the shader */
3589 GL_EXTCALL(glGenProgramsARB(1, &ret
));
3590 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
3591 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
3593 if (glGetError() == GL_INVALID_OPERATION
) {
3595 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3596 FIXME("Fragment program error at position %d: %s\n", pos
,
3597 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3599 shader_buffer_free(&buffer
);
3603 static void fragment_prog_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3604 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
3605 struct shader_arb_priv
*priv
= device
->fragment_priv
;
3606 BOOL use_pshader
= use_ps(stateblock
);
3607 BOOL use_vshader
= use_vs(stateblock
);
3608 struct ffp_frag_settings settings
;
3609 const struct arbfp_ffp_desc
*desc
;
3612 TRACE("state %#x, stateblock %p, context %p\n", state
, stateblock
, context
);
3614 if(isStateDirty(context
, STATE_RENDER(WINED3DRS_FOGENABLE
))) {
3615 if(!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
3616 /* Reload fixed function constants since they collide with the pixel shader constants */
3617 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
3618 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
3620 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
3621 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
3622 } else if(use_pshader
&& !isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
3623 device
->shader_backend
->shader_select((IWineD3DDevice
*)stateblock
->wineD3DDevice
, use_pshader
, use_vshader
);
3629 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
3630 gen_ffp_frag_op(stateblock
, &settings
, FALSE
);
3631 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(priv
->fragment_shaders
, &settings
);
3633 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
3636 ERR("Out of memory\n");
3639 new_desc
->num_textures_used
= 0;
3640 for(i
= 0; i
< GL_LIMITS(texture_stages
); i
++) {
3641 if(settings
.op
[i
].cop
== WINED3DTOP_DISABLE
) break;
3642 new_desc
->num_textures_used
= i
;
3645 memcpy(&new_desc
->parent
.settings
, &settings
, sizeof(settings
));
3646 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, stateblock
);
3647 add_ffp_frag_shader(priv
->fragment_shaders
, &new_desc
->parent
);
3648 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
3652 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
3653 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
3656 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
3657 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
3658 priv
->current_fprogram_id
= desc
->shader
;
3660 if(device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
3661 /* Reload fixed function constants since they collide with the pixel shader constants */
3662 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
3663 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
3665 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
3666 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
3668 context
->last_was_pshader
= FALSE
;
3670 context
->last_was_pshader
= TRUE
;
3673 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
3674 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
3675 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
3676 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
3677 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
3679 * Don't call shader_select if the vertex shader is dirty, because it will be called later on by the vertex
3682 if(!isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
3683 device
->shader_backend
->shader_select((IWineD3DDevice
*)stateblock
->wineD3DDevice
, use_pshader
, use_vshader
);
3685 if (!isStateDirty(context
, STATE_VERTEXSHADERCONSTANT
) && (use_vshader
|| use_pshader
)) {
3686 device
->StateTable
[STATE_VERTEXSHADERCONSTANT
].apply(STATE_VERTEXSHADERCONSTANT
, stateblock
, context
);
3690 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
3694 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
3695 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
3696 * state table, so we need to handle that with a forwarding function. The other invisible side effect
3697 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
3698 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
3700 static void state_arbfp_fog(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3701 enum fogsource new_source
;
3703 TRACE("state %#x, stateblock %p, context %p\n", state
, stateblock
, context
);
3705 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
3706 fragment_prog_arbfp(state
, stateblock
, context
);
3709 if(!stateblock
->renderState
[WINED3DRS_FOGENABLE
]) return;
3711 if(stateblock
->renderState
[WINED3DRS_FOGTABLEMODE
] == WINED3DFOG_NONE
) {
3712 if(use_vs(stateblock
)) {
3713 new_source
= FOGSOURCE_VS
;
3715 if(stateblock
->renderState
[WINED3DRS_FOGVERTEXMODE
] == WINED3DFOG_NONE
|| context
->last_was_rhw
) {
3716 new_source
= FOGSOURCE_COORD
;
3718 new_source
= FOGSOURCE_FFP
;
3722 new_source
= FOGSOURCE_FFP
;
3724 if(new_source
!= context
->fog_source
) {
3725 context
->fog_source
= new_source
;
3726 state_fogstartend(STATE_RENDER(WINED3DRS_FOGSTART
), stateblock
, context
);
3730 static void textransform(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, WineD3DContext
*context
) {
3731 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
3732 fragment_prog_arbfp(state
, stateblock
, context
);
3736 #undef GLINFO_LOCATION
3738 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] = {
3739 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), state_texfactor_arbfp
}, 0 },
3740 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3741 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3742 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3743 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3744 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3745 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3746 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3747 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3748 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3749 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3750 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3751 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3752 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3753 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3754 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3755 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3756 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3757 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3758 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3759 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3760 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3761 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3762 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3763 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3764 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3765 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3766 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3767 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3768 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3769 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3770 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3771 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3772 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3773 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3774 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3775 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3776 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3777 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3778 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3779 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3780 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3781 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3782 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3783 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3784 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3785 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3786 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3787 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3788 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3789 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3790 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3791 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3792 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3793 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3794 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3795 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3796 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3797 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3798 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3799 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3800 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3801 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3802 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3803 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3804 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3805 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3806 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3807 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3808 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3809 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3810 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3811 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3812 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3813 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3814 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3815 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3816 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3817 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3818 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3819 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3820 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3821 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3822 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3823 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3824 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3825 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3826 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3827 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3828 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3829 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3830 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3831 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3832 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3833 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3834 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3835 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3836 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3837 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3838 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3839 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3840 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3841 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3842 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3843 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3844 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3845 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3846 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3847 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3848 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3849 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3850 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3851 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3852 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3853 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3854 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3855 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3856 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3857 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, 0 },
3858 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3859 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, 0 },
3860 {STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim
}, 0 },
3861 {STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim
}, 0 },
3862 {STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim
}, 0 },
3863 {STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim
}, 0 },
3864 {STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim
}, 0 },
3865 {STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim
}, 0 },
3866 {STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim
}, 0 },
3867 {STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim
}, 0 },
3868 {STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3869 {STATE_RENDER(WINED3DRS_FOGENABLE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3870 {STATE_RENDER(WINED3DRS_FOGTABLEMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3871 {STATE_RENDER(WINED3DRS_FOGVERTEXMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, 0 },
3872 {STATE_RENDER(WINED3DRS_FOGSTART
), { STATE_RENDER(WINED3DRS_FOGSTART
), state_fogstartend
}, 0 },
3873 {STATE_RENDER(WINED3DRS_FOGEND
), { STATE_RENDER(WINED3DRS_FOGSTART
), state_fogstartend
}, 0 },
3874 {STATE_RENDER(WINED3DRS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, 0 },
3875 {STATE_RENDER(WINED3DRS_FOGCOLOR
), { STATE_RENDER(WINED3DRS_FOGCOLOR
), state_fogcolor
}, 0 },
3876 {STATE_RENDER(WINED3DRS_FOGDENSITY
), { STATE_RENDER(WINED3DRS_FOGDENSITY
), state_fogdensity
}, 0 },
3877 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3878 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3879 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3880 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3881 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3882 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3883 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3884 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, 0 },
3885 {STATE_RENDER(WINED3DRS_SPECULARENABLE
), { STATE_RENDER(WINED3DRS_SPECULARENABLE
), state_arb_specularenable
}, 0 },
3886 {0 /* Terminate */, { 0, 0 }, 0 },
3889 const struct fragment_pipeline arbfp_fragment_pipeline
= {
3894 shader_arb_color_fixup_supported
,
3895 arbfp_fragmentstate_template
,
3896 TRUE
/* We can disable projected textures */
3899 #define GLINFO_LOCATION device->adapter->gl_info
3901 struct arbfp_blit_priv
{
3902 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
3903 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
3904 GLenum yv12_rect_shader
, yv12_2d_shader
;
3907 static HRESULT
arbfp_blit_alloc(IWineD3DDevice
*iface
) {
3908 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3909 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
3910 if(!device
->blit_priv
) {
3911 ERR("Out of memory\n");
3912 return E_OUTOFMEMORY
;
3916 static void arbfp_blit_free(IWineD3DDevice
*iface
) {
3917 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
3918 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
3921 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
3922 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
3923 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
3924 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
3925 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
3926 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
3927 checkGLcall("Delete yuv programs\n");
3931 static BOOL
gen_planar_yuv_read(SHADER_BUFFER
*buffer
, enum yuv_fixup yuv_fixup
, GLenum textype
, char *luminance
)
3934 const char *tex
, *texinstr
;
3936 if (yuv_fixup
== YUV_FIXUP_UYVY
) {
3944 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
3945 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
3947 /* This is more tricky than just replacing the texture type - we have to navigate
3948 * properly in the texture to find the correct chroma values
3950 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
3954 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
3955 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
3956 * filtering when we sample the texture.
3958 * These are the rules for reading the chroma:
3964 * So we have to get the sampling x position in non-normalized coordinates in integers
3966 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
3967 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
3968 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
3970 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
3972 /* We must not allow filtering between pixel x and x+1, this would mix U and V
3973 * Vertical filtering is ok. However, bear in mind that the pixel center is at
3976 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
3977 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
3979 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
3980 * even and odd pixels respectively
3982 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
3983 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
3985 /* Sample Pixel 1 */
3986 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
3988 /* Put the value into either of the chroma values */
3989 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
3990 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
3991 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
3992 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
3994 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
3995 * the pixel right to the current one. Otherwise, sample the left pixel.
3996 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
3998 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
3999 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
4000 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
4002 /* Put the value into the other chroma */
4003 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
4004 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
4005 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
4006 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
4008 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
4009 * the current one and lerp the two U and V values
4012 /* This gives the correctly filtered luminance value */
4013 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
4018 static BOOL
gen_yv12_read(SHADER_BUFFER
*buffer
, GLenum textype
, char *luminance
)
4023 case GL_TEXTURE_2D
: tex
= "2D"; break;
4024 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
4026 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
4030 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
4031 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
4032 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
4033 * pitch of the luminance plane, the packing into the gl texture is a bit
4034 * unfortunate. If the whole texture is interpreted as luminance data it looks
4035 * approximately like this:
4037 * +----------------------------------+----
4049 * +----------------+-----------------+----
4051 * | U even rows | U odd rows |
4053 * +----------------+------------------ -
4055 * | V even rows | V odd rows |
4057 * +----------------+-----------------+----
4061 * So it appears as if there are 4 chroma images, but in fact the odd rows
4062 * in the chroma images are in the same row as the even ones. So its is
4063 * kinda tricky to read
4065 * When reading from rectangle textures, keep in mind that the input y coordinates
4066 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
4068 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
4069 2.0 / 3.0, 1.0 / 6.0, (2.0 / 3.0) + (1.0 / 6.0), 1.0 / 3.0);
4071 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
4072 /* the chroma planes have only half the width */
4073 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
4075 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
4076 * the coordinate. Also read the right side of the image when reading odd lines
4078 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
4081 if(textype
== GL_TEXTURE_2D
) {
4083 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
4085 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
4087 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
4088 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
4090 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
4091 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
4092 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
4093 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
4094 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
4096 /* clamp, keep the half pixel origin in mind */
4097 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
4098 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
4099 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
4100 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
4102 /* Read from [size - size+size/4] */
4103 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
4104 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
4106 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
4107 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
4108 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
4109 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
4110 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
4111 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
4113 /* Make sure to read exactly from the pixel center */
4114 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
4115 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
4118 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
4119 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
4120 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
4121 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
4122 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
4124 /* Read the texture, put the result into the output register */
4125 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
4126 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
4128 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
4129 * No need to clamp because we're just reusing the already clamped value from above
4131 if(textype
== GL_TEXTURE_2D
) {
4132 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
4134 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
4136 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
4137 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
4139 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
4140 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
4141 * values due to filtering
4143 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
4144 if(textype
== GL_TEXTURE_2D
) {
4145 /* Multiply the y coordinate by 2/3 and clamp it */
4146 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
4147 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
4148 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
4149 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
4151 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
4152 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
4155 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
4156 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
4157 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
4164 static GLuint
gen_yuv_shader(IWineD3DDeviceImpl
*device
, enum yuv_fixup yuv_fixup
, GLenum textype
)
4167 SHADER_BUFFER buffer
;
4168 char luminance_component
;
4169 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
4172 shader_buffer_init(&buffer
);
4175 GL_EXTCALL(glGenProgramsARB(1, &shader
));
4176 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
4177 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
4178 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
4181 shader_buffer_free(&buffer
);
4185 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
4186 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
4187 * two chroma(U and V) values. Each macropixel has two luminance values, one for
4188 * each single pixel it contains, and one U and one V value shared between both
4191 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
4192 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
4193 * take the format into account when generating the read swizzles
4195 * Reading the Y value is straightforward - just sample the texture. The hardware
4196 * takes care of filtering in the horizontal and vertical direction.
4198 * Reading the U and V values is harder. We have to avoid filtering horizontally,
4199 * because that would mix the U and V values of one pixel or two adjacent pixels.
4200 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
4201 * regardless of the filtering setting. Vertical filtering works automatically
4202 * though - the U and V values of two rows are mixed nicely.
4204 * Appart of avoiding filtering issues, the code has to know which value it just
4205 * read, and where it can find the other one. To determine this, it checks if
4206 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
4208 * Handling horizontal filtering of U and V values requires reading a 2nd pair
4209 * of pixels, extracting U and V and mixing them. This is not implemented yet.
4211 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
4212 * with width / 2. This way one read gives all 3 values, finding U and V is easy
4213 * in an unfiltered situation. Finding the luminance on the other hand requires
4214 * finding out if it is an odd or even pixel. The real drawback of this approach
4215 * is filtering. This would have to be emulated completely in the shader, reading
4216 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
4217 * vertically. Beyond that it would require adjustments to the texture handling
4218 * code to deal with the width scaling
4220 shader_addline(&buffer
, "!!ARBfp1.0\n");
4221 shader_addline(&buffer
, "TEMP luminance;\n");
4222 shader_addline(&buffer
, "TEMP temp;\n");
4223 shader_addline(&buffer
, "TEMP chroma;\n");
4224 shader_addline(&buffer
, "TEMP texcrd;\n");
4225 shader_addline(&buffer
, "TEMP texcrd2;\n");
4226 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
4227 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
4228 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
4232 case YUV_FIXUP_UYVY
:
4233 case YUV_FIXUP_YUY2
:
4234 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
4236 shader_buffer_free(&buffer
);
4241 case YUV_FIXUP_YV12
:
4242 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
4244 shader_buffer_free(&buffer
);
4250 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
4251 shader_buffer_free(&buffer
);
4255 /* Calculate the final result. Formula is taken from
4256 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
4257 * ranges from -0.5 to 0.5
4259 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
4261 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
4262 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
4263 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
4264 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
4265 shader_addline(&buffer
, "END\n");
4268 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
4270 if (glGetError() == GL_INVALID_OPERATION
) {
4272 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
4273 FIXME("Fragment program error at position %d: %s\n", pos
,
4274 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
4276 shader_buffer_free(&buffer
);
4281 case YUV_FIXUP_YUY2
:
4282 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yuy2_rect_shader
= shader
;
4283 else priv
->yuy2_2d_shader
= shader
;
4286 case YUV_FIXUP_UYVY
:
4287 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->uyvy_rect_shader
= shader
;
4288 else priv
->uyvy_2d_shader
= shader
;
4291 case YUV_FIXUP_YV12
:
4292 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yv12_rect_shader
= shader
;
4293 else priv
->yv12_2d_shader
= shader
;
4300 static HRESULT
arbfp_blit_set(IWineD3DDevice
*iface
, const struct GlPixelFormatDesc
*format_desc
,
4301 GLenum textype
, UINT width
, UINT height
)
4304 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
4305 float size
[4] = {width
, height
, 1, 1};
4306 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
4307 enum yuv_fixup yuv_fixup
;
4309 if (!is_yuv_fixup(format_desc
->color_fixup
))
4312 dump_color_fixup_desc(format_desc
->color_fixup
);
4313 /* Don't bother setting up a shader for unconverted formats */
4316 checkGLcall("glEnable(textype)");
4321 yuv_fixup
= get_yuv_fixup(format_desc
->color_fixup
);
4325 case YUV_FIXUP_YUY2
:
4326 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yuy2_rect_shader
: priv
->yuy2_2d_shader
;
4329 case YUV_FIXUP_UYVY
:
4330 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->uyvy_rect_shader
: priv
->uyvy_2d_shader
;
4333 case YUV_FIXUP_YV12
:
4334 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yv12_rect_shader
: priv
->yv12_2d_shader
;
4338 FIXME("Unsupported YUV fixup %#x, not setting a shader\n", yuv_fixup
);
4341 checkGLcall("glEnable(textype)");
4346 if (!shader
) shader
= gen_yuv_shader(device
, yuv_fixup
, textype
);
4349 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4350 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
4351 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
4352 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
4353 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
4354 checkGLcall("glProgramLocalParameter4fvARB");
4360 static void arbfp_blit_unset(IWineD3DDevice
*iface
) {
4361 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
4364 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4365 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4366 glDisable(GL_TEXTURE_2D
);
4367 checkGLcall("glDisable(GL_TEXTURE_2D)");
4368 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP
)) {
4369 glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
4370 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
4372 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
4373 glDisable(GL_TEXTURE_RECTANGLE_ARB
);
4374 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
4379 static BOOL
arbfp_blit_color_fixup_supported(struct color_fixup_desc fixup
)
4381 enum yuv_fixup yuv_fixup
;
4383 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4385 TRACE("Checking support for fixup:\n");
4386 dump_color_fixup_desc(fixup
);
4389 if (is_identity_fixup(fixup
))
4395 /* We only support YUV conversions. */
4396 if (!is_yuv_fixup(fixup
))
4398 TRACE("[FAILED]\n");
4402 yuv_fixup
= get_yuv_fixup(fixup
);
4405 case YUV_FIXUP_YUY2
:
4406 case YUV_FIXUP_UYVY
:
4407 case YUV_FIXUP_YV12
:
4412 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
4413 TRACE("[FAILED]\n");
4418 const struct blit_shader arbfp_blit
= {
4423 arbfp_blit_color_fixup_supported
,
4426 #undef GLINFO_LOCATION