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 struct wined3d_gl_info
*gl_info
)
47 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*) shader
;
48 if(!This
->baseShader
.reg_maps
.usesmova
) return FALSE
;
49 return !GL_SUPPORT(NV_VERTEX_PROGRAM2_OPTION
);
52 static BOOL
need_helper_const(const struct wined3d_gl_info
*gl_info
)
54 if (!GL_SUPPORT(NV_VERTEX_PROGRAM
) /* Need to init colors. */
55 || gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
/* Load the immval offset. */
56 || gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) /* Have to init texcoords. */
63 static unsigned int reserved_vs_const(IWineD3DBaseShader
*shader
, const struct wined3d_gl_info
*gl_info
)
66 /* We use one PARAM for the pos fixup, and in some cases one to load
67 * some immediate values into the shader
69 if(need_helper_const(gl_info
)) ret
++;
70 if(need_mova_const(shader
, gl_info
)) ret
++;
74 static inline BOOL
ffp_clip_emul(IWineD3DStateBlockImpl
*stateblock
)
76 return stateblock
->lowest_disabled_stage
< 7;
79 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
80 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
82 return GL_SUPPORT(NV_VERTEX_PROGRAM2_OPTION
);
85 /* Internally used shader constants. Applications can use constants 0 to GL_LIMITS(vshader_constantsF) - 1,
86 * so upload them above that
88 #define ARB_SHADER_PRIVCONST_BASE (GL_LIMITS(vshader_constantsF) - 1)
89 #define ARB_SHADER_PRIVCONST_POS ARB_SHADER_PRIVCONST_BASE + 0
91 /* ARB_program_shader private data */
114 unsigned int loop_no
;
117 struct loop_control loop_control
;
121 struct arb_ps_np2fixup_info
123 struct ps_np2fixup_info super
;
124 /* For ARB we need a offset value:
125 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
126 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
127 * array we need an offset to the index inside the program local parameter array. */
131 struct arb_ps_compile_args
133 struct ps_compile_args super
;
134 DWORD bools
; /* WORD is enough, use DWORD for alignment */
135 unsigned char loop_ctrl
[MAX_CONST_I
][3];
138 struct stb_const_desc
140 unsigned char texunit
;
144 struct arb_ps_compiled_shader
146 struct arb_ps_compile_args args
;
147 struct arb_ps_np2fixup_info np2fixup_info
;
148 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
149 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
150 UINT int_consts
[MAX_CONST_I
];
153 unsigned char numbumpenvmatconsts
;
157 struct arb_vs_compile_args
159 struct vs_compile_args super
;
165 char clip_control
[2];
167 DWORD boolclip_compare
;
172 unsigned char vertex_samplers
[4];
173 DWORD vertex_samplers_compare
;
175 unsigned char loop_ctrl
[MAX_CONST_I
][3];
178 struct arb_vs_compiled_shader
180 struct arb_vs_compile_args args
;
182 UINT int_consts
[MAX_CONST_I
];
184 char need_color_unclamp
;
188 struct recorded_instruction
190 struct wined3d_shader_instruction ins
;
194 struct shader_arb_ctx_priv
199 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
201 /* GL_NV_vertex_progam2_option or GL_NV_fragment_program_option */
203 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
207 const struct arb_vs_compile_args
*cur_vs_args
;
208 const struct arb_ps_compile_args
*cur_ps_args
;
209 const struct arb_ps_compiled_shader
*compiled_fprog
;
210 const struct arb_vs_compiled_shader
*compiled_vprog
;
211 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
212 struct list control_frames
;
216 unsigned int num_loops
, loop_depth
, num_ifcs
;
219 unsigned int vs_clipplanes
;
223 /* For 3.0 vertex shaders */
224 const char *vs_output
[MAX_REG_OUTPUT
];
225 /* For 2.x and earlier vertex shaders */
226 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
228 /* 3.0 pshader input for compatibility with fixed function */
229 const char *ps_input
[MAX_REG_INPUT
];
234 struct wined3d_shader_signature_element
*sig
;
236 struct wine_rb_entry entry
;
239 struct arb_pshader_private
{
240 struct arb_ps_compiled_shader
*gl_shaders
;
241 UINT num_gl_shaders
, shader_array_size
;
242 BOOL has_signature_idx
;
243 DWORD input_signature_idx
;
244 DWORD clipplane_emulation
;
248 struct arb_vshader_private
{
249 struct arb_vs_compiled_shader
*gl_shaders
;
250 UINT num_gl_shaders
, shader_array_size
;
253 struct shader_arb_priv
255 GLuint current_vprogram_id
;
256 GLuint current_fprogram_id
;
257 const struct arb_ps_compiled_shader
*compiled_fprog
;
258 const struct arb_vs_compiled_shader
*compiled_vprog
;
259 GLuint depth_blt_vprogram_id
;
260 GLuint depth_blt_fprogram_id
[tex_type_count
];
261 BOOL use_arbfp_fixed_func
;
262 struct wine_rb_tree fragment_shaders
;
263 BOOL last_ps_const_clamped
;
264 BOOL last_vs_color_unclamp
;
266 struct wine_rb_tree signature_tree
;
270 /********************************************************
271 * ARB_[vertex/fragment]_program helper functions follow
272 ********************************************************/
275 * Loads floating point constants into the currently set ARB_vertex/fragment_program.
276 * When constant_list == NULL, it will load all the constants.
278 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
279 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
281 /* GL locking is done by the caller */
282 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
283 GLuint target_type
, unsigned int max_constants
, const float *constants
, char *dirty_consts
)
285 local_constant
* lconst
;
289 if (TRACE_ON(d3d_shader
)) {
290 for(i
= 0; i
< max_constants
; i
++) {
291 if(!dirty_consts
[i
]) continue;
292 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
293 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
294 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
297 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
298 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& This
->baseShader
.reg_maps
.shader_version
.major
== 1)
301 for(i
= 0; i
< max_constants
; i
++) {
302 if(!dirty_consts
[i
]) continue;
306 if (constants
[j
+ 0] > 1.0f
) lcl_const
[0] = 1.0f
;
307 else if (constants
[j
+ 0] < -1.0f
) lcl_const
[0] = -1.0f
;
308 else lcl_const
[0] = constants
[j
+ 0];
310 if (constants
[j
+ 1] > 1.0f
) lcl_const
[1] = 1.0f
;
311 else if (constants
[j
+ 1] < -1.0f
) lcl_const
[1] = -1.0f
;
312 else lcl_const
[1] = constants
[j
+ 1];
314 if (constants
[j
+ 2] > 1.0f
) lcl_const
[2] = 1.0f
;
315 else if (constants
[j
+ 2] < -1.0f
) lcl_const
[2] = -1.0f
;
316 else lcl_const
[2] = constants
[j
+ 2];
318 if (constants
[j
+ 3] > 1.0f
) lcl_const
[3] = 1.0f
;
319 else if (constants
[j
+ 3] < -1.0f
) lcl_const
[3] = -1.0f
;
320 else lcl_const
[3] = constants
[j
+ 3];
322 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
325 if(GL_SUPPORT(EXT_GPU_PROGRAM_PARAMETERS
)) {
326 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
327 * or just reloading *all* constants at once
329 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, 0, max_constants, constants));
331 for(i
= 0; i
< max_constants
; i
++) {
332 if(!dirty_consts
[i
]) continue;
334 /* Find the next block of dirty constants */
337 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
341 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
344 for(i
= 0; i
< max_constants
; i
++) {
345 if(dirty_consts
[i
]) {
347 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
352 checkGLcall("glProgramEnvParameter4fvARB()");
354 /* Load immediate constants */
355 if(This
->baseShader
.load_local_constsF
) {
356 if (TRACE_ON(d3d_shader
)) {
357 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
358 GLfloat
* values
= (GLfloat
*)lconst
->value
;
359 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
360 values
[0], values
[1], values
[2], values
[3]);
363 /* Immediate constants are clamped for 1.X shaders at loading times */
365 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
366 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
367 ret
= max(ret
, lconst
->idx
+ 1);
368 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
370 checkGLcall("glProgramEnvParameter4fvARB()");
371 return ret
; /* The loaded immediate constants need reloading for the next shader */
373 return 0; /* No constants are dirty now */
378 * Loads the texture dimensions for NP2 fixup into the currently set ARB_[vertex/fragment]_programs.
380 static void shader_arb_load_np2fixup_constants(
381 IWineD3DDevice
* device
,
383 char useVertexShader
) {
385 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
386 const struct shader_arb_priv
* const priv
= (const struct shader_arb_priv
*) deviceImpl
->shader_priv
;
387 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
388 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
390 if (!usePixelShader
) {
391 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
395 if (priv
->compiled_fprog
&& priv
->compiled_fprog
->np2fixup_info
.super
.active
) {
396 const struct arb_ps_np2fixup_info
* const fixup
= &priv
->compiled_fprog
->np2fixup_info
;
398 WORD active
= fixup
->super
.active
;
399 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
401 for (i
= 0; active
; active
>>= 1, ++i
) {
402 const unsigned char idx
= fixup
->super
.idx
[i
];
403 const IWineD3DTextureImpl
* const tex
= (const IWineD3DTextureImpl
*) stateBlock
->textures
[i
];
404 GLfloat
* tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
406 if (!(active
& 1)) continue;
409 FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
414 tex_dim
[2] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[3] = tex
->baseTexture
.pow2Matrix
[5];
416 tex_dim
[0] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[1] = tex
->baseTexture
.pow2Matrix
[5];
420 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
) {
421 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
422 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
427 /* GL locking is done by the caller. */
428 static inline void shader_arb_ps_local_constants(IWineD3DDeviceImpl
* deviceImpl
)
430 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
431 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
433 struct shader_arb_priv
*priv
= deviceImpl
->shader_priv
;
434 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
436 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
438 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
440 /* The state manager takes care that this function is always called if the bump env matrix changes */
441 const float *data
= (const float *)&stateBlock
->textureState
[texunit
][WINED3DTSS_BUMPENVMAT00
];
442 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
444 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
446 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
447 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
448 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
449 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
451 const float *scale
= (const float *)&stateBlock
->textureState
[texunit
][WINED3DTSS_BUMPENVLSCALE
];
452 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->luminanceconst
[i
].const_num
, scale
));
455 checkGLcall("Load bumpmap consts");
457 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
459 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
460 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
465 val
[0] = deviceImpl
->render_offscreen
? 0.0f
: ((IWineD3DSurfaceImpl
*) deviceImpl
->render_targets
[0])->currentDesc
.Height
;
466 val
[1] = deviceImpl
->render_offscreen
? 1.0f
: -1.0f
;
469 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
470 checkGLcall("y correction loading");
473 if(gl_shader
->num_int_consts
== 0) return;
475 for(i
= 0; i
< MAX_CONST_I
; i
++)
477 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
480 val
[0] = stateBlock
->pixelShaderConstantI
[4 * i
];
481 val
[1] = stateBlock
->pixelShaderConstantI
[4 * i
+ 1];
482 val
[2] = stateBlock
->pixelShaderConstantI
[4 * i
+ 2];
485 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
488 checkGLcall("Load ps int consts");
491 /* GL locking is done by the caller. */
492 static inline void shader_arb_vs_local_constants(IWineD3DDeviceImpl
* deviceImpl
)
494 IWineD3DStateBlockImpl
* stateBlock
;
495 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
497 struct shader_arb_priv
*priv
= deviceImpl
->shader_priv
;
498 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
500 /* Upload the position fixup */
501 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, deviceImpl
->posFixup
));
503 if(gl_shader
->num_int_consts
== 0) return;
505 stateBlock
= deviceImpl
->stateBlock
;
507 for(i
= 0; i
< MAX_CONST_I
; i
++)
509 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
512 val
[0] = stateBlock
->vertexShaderConstantI
[4 * i
];
513 val
[1] = stateBlock
->vertexShaderConstantI
[4 * i
+ 1];
514 val
[2] = stateBlock
->vertexShaderConstantI
[4 * i
+ 2];
517 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
520 checkGLcall("Load vs int consts");
524 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
526 * We only support float constants in ARB at the moment, so don't
527 * worry about the Integers or Booleans
529 /* GL locking is done by the caller (state handler) */
530 static void shader_arb_load_constants(
531 IWineD3DDevice
* device
,
533 char useVertexShader
) {
535 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
536 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
537 const struct wined3d_context
*context
= context_get_current();
538 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
540 if (useVertexShader
) {
541 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
543 /* Load DirectX 9 float constants for vertex shader */
544 deviceImpl
->highest_dirty_vs_const
= shader_arb_load_constantsF(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
545 deviceImpl
->highest_dirty_vs_const
, stateBlock
->vertexShaderConstantF
, context
->vshader_const_dirty
);
547 shader_arb_vs_local_constants(deviceImpl
);
550 if (usePixelShader
) {
551 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
553 /* Load DirectX 9 float constants for pixel shader */
554 deviceImpl
->highest_dirty_ps_const
= shader_arb_load_constantsF(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
555 deviceImpl
->highest_dirty_ps_const
, stateBlock
->pixelShaderConstantF
, context
->pshader_const_dirty
);
556 shader_arb_ps_local_constants(deviceImpl
);
560 static void shader_arb_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
562 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
563 struct wined3d_context
*context
= context_get_current();
565 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
566 * context. On a context switch the old context will be fully dirtified */
567 if (!context
|| ((IWineD3DSurfaceImpl
*)context
->surface
)->resource
.wineD3DDevice
!= This
) return;
569 memset(context
->vshader_const_dirty
+ start
, 1, sizeof(*context
->vshader_const_dirty
) * count
);
570 This
->highest_dirty_vs_const
= max(This
->highest_dirty_vs_const
, start
+ count
);
573 static void shader_arb_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
575 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
576 struct wined3d_context
*context
= context_get_current();
578 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
579 * context. On a context switch the old context will be fully dirtified */
580 if (!context
|| ((IWineD3DSurfaceImpl
*)context
->surface
)->resource
.wineD3DDevice
!= This
) return;
582 memset(context
->pshader_const_dirty
+ start
, 1, sizeof(*context
->pshader_const_dirty
) * count
);
583 This
->highest_dirty_ps_const
= max(This
->highest_dirty_ps_const
, start
+ count
);
586 static DWORD
*local_const_mapping(IWineD3DBaseShaderImpl
*This
)
590 const local_constant
*lconst
;
592 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) return NULL
;
594 ret
= HeapAlloc(GetProcessHeap(), 0, sizeof(DWORD
) * This
->baseShader
.limits
.constant_float
);
596 ERR("Out of memory\n");
600 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
601 ret
[lconst
->idx
] = idx
++;
606 /* Generate the variable & register declarations for the ARB_vertex_program output target */
607 static DWORD
shader_generate_arb_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
608 struct wined3d_shader_buffer
*buffer
, const struct wined3d_gl_info
*gl_info
, DWORD
*lconst_map
,
609 DWORD
*num_clipplanes
, struct shader_arb_ctx_priv
*ctx
)
611 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
612 DWORD i
, next_local
= 0;
613 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
614 unsigned max_constantsF
;
615 const local_constant
*lconst
;
617 /* In pixel shaders, all private constants are program local, we don't need anything
618 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
619 * If we need a private constant the GL implementation will squeeze it in somewhere
621 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
622 * immediate values. The posFixup is loaded using program.env for now, so always
623 * subtract one from the number of constants. If the shader uses indirect addressing,
624 * account for the helper const too because we have to declare all availabke d3d constants
625 * and don't know which are actually used.
628 max_constantsF
= GL_LIMITS(pshader_constantsF
);
630 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
631 DWORD highest_constf
= 0, clip_limit
;
632 max_constantsF
= GL_LIMITS(vshader_constantsF
) - reserved_vs_const(iface
, gl_info
);
633 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.integer_constants
);
635 for(i
= 0; i
< This
->baseShader
.limits
.constant_float
; i
++)
638 DWORD shift
= i
& 0x1f;
639 if(reg_maps
->constf
[idx
] & (1 << shift
)) highest_constf
= i
;
642 clip_limit
= GL_LIMITS(clipplanes
);
643 if(ctx
->target_version
== ARB
) clip_limit
= min(clip_limit
, 4);
644 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
645 max_constantsF
-= *num_clipplanes
;
646 if(*num_clipplanes
< clip_limit
)
648 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, GL_LIMITS(clipplanes
));
653 if(ctx
->target_version
>= NV2
) *num_clipplanes
= GL_LIMITS(clipplanes
);
654 else *num_clipplanes
= min(GL_LIMITS(clipplanes
), 4);
655 max_constantsF
= GL_LIMITS(vshader_constantsF
);
659 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
660 if (reg_maps
->temporary
[i
])
661 shader_addline(buffer
, "TEMP R%u;\n", i
);
664 for (i
= 0; i
< This
->baseShader
.limits
.address
; i
++) {
665 if (reg_maps
->address
[i
])
666 shader_addline(buffer
, "ADDRESS A%d;\n", i
);
669 if(pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3) {
670 for(i
= 0; i
< This
->baseShader
.limits
.texcoord
; i
++) {
671 if (reg_maps
->texcoord
[i
] && pshader
)
672 shader_addline(buffer
,"TEMP T%u;\n", i
);
676 /* Load local constants using the program-local space,
677 * this avoids reloading them each time the shader is used
680 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
681 shader_addline(buffer
, "PARAM C%u = program.local[%u];\n", lconst
->idx
,
682 lconst_map
[lconst
->idx
]);
683 next_local
= max(next_local
, lconst_map
[lconst
->idx
] + 1);
687 /* we use the array-based constants array if the local constants are marked for loading,
688 * because then we use indirect addressing, or when the local constant list is empty,
689 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
690 * local constants do not declare the loaded constants as an array because ARB compilers usually
691 * do not optimize unused constants away
693 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
694 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
695 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
696 max_constantsF
, max_constantsF
- 1);
698 for(i
= 0; i
< max_constantsF
; i
++) {
701 mask
= 1 << (i
& 0x1f);
702 if(!shader_constant_is_local(This
, i
) && (This
->baseShader
.reg_maps
.constf
[idx
] & mask
)) {
703 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
711 static const char * const shift_tab
[] = {
712 "dummy", /* 0 (none) */
713 "coefmul.x", /* 1 (x2) */
714 "coefmul.y", /* 2 (x4) */
715 "coefmul.z", /* 3 (x8) */
716 "coefmul.w", /* 4 (x16) */
717 "dummy", /* 5 (x32) */
718 "dummy", /* 6 (x64) */
719 "dummy", /* 7 (x128) */
720 "dummy", /* 8 (d256) */
721 "dummy", /* 9 (d128) */
722 "dummy", /* 10 (d64) */
723 "dummy", /* 11 (d32) */
724 "coefdiv.w", /* 12 (d16) */
725 "coefdiv.z", /* 13 (d8) */
726 "coefdiv.y", /* 14 (d4) */
727 "coefdiv.x" /* 15 (d2) */
730 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
731 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
733 char *ptr
= write_mask
;
735 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
738 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
739 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
740 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
741 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
747 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
749 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
750 * but addressed as "rgba". To fix this we need to swap the register's x
751 * and z components. */
752 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
753 char *ptr
= swizzle_str
;
755 /* swizzle bits fields: wwzzyyxx */
756 DWORD swizzle
= param
->swizzle
;
757 DWORD swizzle_x
= swizzle
& 0x03;
758 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
759 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
760 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
762 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
763 * generate a swizzle string. Unless we need to our own swizzling. */
764 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
767 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
768 *ptr
++ = swizzle_chars
[swizzle_x
];
770 *ptr
++ = swizzle_chars
[swizzle_x
];
771 *ptr
++ = swizzle_chars
[swizzle_y
];
772 *ptr
++ = swizzle_chars
[swizzle_z
];
773 *ptr
++ = swizzle_chars
[swizzle_w
];
780 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
782 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
783 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
785 if(strcmp(priv
->addr_reg
, src
) == 0) return;
787 strcpy(priv
->addr_reg
, src
);
788 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
791 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
792 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
794 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
795 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
797 /* oPos, oFog and oPts in D3D */
798 static const char * const rastout_reg_names
[] = {"TMP_OUT", "result.fogcoord", "result.pointsize"};
799 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
800 BOOL pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
801 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
807 case WINED3DSPR_TEMP
:
808 sprintf(register_name
, "R%u", reg
->idx
);
811 case WINED3DSPR_INPUT
:
814 if(This
->baseShader
.reg_maps
.shader_version
.major
< 3)
816 if (reg
->idx
== 0) strcpy(register_name
, "fragment.color.primary");
817 else strcpy(register_name
, "fragment.color.secondary");
824 shader_arb_get_src_param(ins
, reg
->rel_addr
, 0, rel_reg
);
826 if(strcmp(rel_reg
, "**aL_emul**") == 0)
828 DWORD idx
= ctx
->aL
+ reg
->idx
;
829 if(idx
< MAX_REG_INPUT
)
831 strcpy(register_name
, ctx
->ps_input
[idx
]);
835 ERR("Pixel shader input register out of bounds: %u\n", idx
);
836 sprintf(register_name
, "out_of_bounds_%u", idx
);
839 else if(This
->baseShader
.reg_maps
.input_registers
& 0x0300)
841 /* There are two ways basically:
843 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
844 * That means trouble if the loop also contains a breakc or if the control values
845 * aren't local constants.
846 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
847 * source dynamically. The trouble is that we cannot simply read aL.y because it
848 * is an ADDRESS register. We could however push it, load .zw with a value and use
849 * ADAC to load the condition code register and pop it again afterwards
851 FIXME("Relative input register addressing with more than 8 registers\n");
853 /* This is better than nothing for now */
854 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
856 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
858 /* This is problematic because we'd have to consult the ctx->ps_input strings
859 * for where to find the varying. Some may be "0.0", others can be texcoords or
860 * colors. This needs either a pipeline replacement to make the vertex shader feed
861 * proper varyings, or loop unrolling
863 * For now use the texcoords and hope for the best
865 FIXME("Non-vertex shader varying input with indirect addressing\n");
866 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
870 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
871 * pulls GL_NV_fragment_program2 in
873 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
878 if(reg
->idx
< MAX_REG_INPUT
)
880 strcpy(register_name
, ctx
->ps_input
[reg
->idx
]);
884 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
);
885 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
);
892 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
)) *is_color
= TRUE
;
893 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
);
897 case WINED3DSPR_CONST
:
898 if (!pshader
&& reg
->rel_addr
)
902 UINT rel_offset
= ((IWineD3DVertexShaderImpl
*)This
)->rel_offset
;
903 if(This
->baseShader
.reg_maps
.shader_version
.major
< 2) {
904 sprintf(rel_reg
, "A0.x");
906 shader_arb_get_src_param(ins
, reg
->rel_addr
, 0, rel_reg
);
907 if(ctx
->target_version
== ARB
) {
908 if(strcmp(rel_reg
, "**aL_emul**") == 0) {
911 shader_arb_request_a0(ins
, rel_reg
);
912 sprintf(rel_reg
, "A0.x");
917 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
);
918 else if (reg
->idx
>= rel_offset
)
919 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
- rel_offset
);
921 sprintf(register_name
, "C[%s - %u]", rel_reg
, -reg
->idx
+ rel_offset
);
925 if (This
->baseShader
.reg_maps
.usesrelconstF
)
926 sprintf(register_name
, "C[%u]", reg
->idx
);
928 sprintf(register_name
, "C%u", reg
->idx
);
932 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
934 if(This
->baseShader
.reg_maps
.shader_version
.major
== 1 &&
935 This
->baseShader
.reg_maps
.shader_version
.minor
<= 3) {
936 /* In ps <= 1.3, Tx is a temporary register as destination to all instructions,
937 * and as source to most instructions. For some instructions it is the texcoord
938 * input. Those instructions know about the special use
940 sprintf(register_name
, "T%u", reg
->idx
);
942 /* in ps 1.4 and 2.x Tx is always a (read-only) varying */
943 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
);
948 if(This
->baseShader
.reg_maps
.shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
950 sprintf(register_name
, "A%u", reg
->idx
);
954 sprintf(register_name
, "A%u_SHADOW", reg
->idx
);
959 case WINED3DSPR_COLOROUT
:
960 if(ctx
->cur_ps_args
->super
.srgb_correction
&& reg
->idx
== 0)
962 strcpy(register_name
, "TMP_COLOR");
966 if(ctx
->cur_ps_args
->super
.srgb_correction
) FIXME("sRGB correction on higher render targets\n");
967 if(This
->baseShader
.reg_maps
.highest_render_target
> 0)
969 sprintf(register_name
, "result.color[%u]", reg
->idx
);
973 strcpy(register_name
, "result.color");
978 case WINED3DSPR_RASTOUT
:
979 if(reg
->idx
== 1) sprintf(register_name
, "%s", ctx
->fog_output
);
980 else sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
]);
983 case WINED3DSPR_DEPTHOUT
:
984 strcpy(register_name
, "result.depth");
987 case WINED3DSPR_ATTROUT
:
988 /* case WINED3DSPR_OUTPUT: */
989 if (pshader
) sprintf(register_name
, "oD[%u]", reg
->idx
);
990 else strcpy(register_name
, ctx
->color_output
[reg
->idx
]);
993 case WINED3DSPR_TEXCRDOUT
:
996 sprintf(register_name
, "oT[%u]", reg
->idx
);
1000 if(This
->baseShader
.reg_maps
.shader_version
.major
< 3)
1002 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
]);
1006 strcpy(register_name
, ctx
->vs_output
[reg
->idx
]);
1011 case WINED3DSPR_LOOP
:
1012 if(ctx
->target_version
>= NV2
)
1014 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1015 if(pshader
) sprintf(register_name
, "A0.x");
1016 else sprintf(register_name
, "aL.y");
1020 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1021 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1022 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1023 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1026 sprintf(register_name
, "**aL_emul**");
1031 case WINED3DSPR_CONSTINT
:
1032 sprintf(register_name
, "I%u", reg
->idx
);
1035 case WINED3DSPR_MISCTYPE
:
1038 sprintf(register_name
, "vpos");
1040 else if(reg
->idx
== 1)
1042 sprintf(register_name
, "fragment.facing.x");
1046 FIXME("Unknown MISCTYPE register index %u\n", reg
->idx
);
1051 FIXME("Unhandled register type %#x[%u]\n", reg
->type
, reg
->idx
);
1052 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
);
1057 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1058 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1060 char register_name
[255];
1064 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1065 strcpy(str
, register_name
);
1067 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1068 strcat(str
, write_mask
);
1071 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1073 switch(channel_source
)
1075 case CHANNEL_SOURCE_ZERO
: return "0";
1076 case CHANNEL_SOURCE_ONE
: return "1";
1077 case CHANNEL_SOURCE_X
: return "x";
1078 case CHANNEL_SOURCE_Y
: return "y";
1079 case CHANNEL_SOURCE_Z
: return "z";
1080 case CHANNEL_SOURCE_W
: return "w";
1082 FIXME("Unhandled channel source %#x\n", channel_source
);
1087 static void gen_color_correction(struct wined3d_shader_buffer
*buffer
, const char *reg
,
1088 DWORD dst_mask
, const char *one
, const char *two
, struct color_fixup_desc fixup
)
1092 if (is_yuv_fixup(fixup
))
1094 enum yuv_fixup yuv_fixup
= get_yuv_fixup(fixup
);
1095 FIXME("YUV fixup (%#x) not supported\n", yuv_fixup
);
1100 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1101 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1102 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1103 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1108 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
1109 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1110 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1114 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
1115 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
1116 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
1117 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
1123 char *ptr
= reg_mask
;
1125 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
1128 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1129 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1130 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1131 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1135 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
1139 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1142 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1143 if (!ins
->dst_count
) return "";
1145 mod
= ins
->dst
[0].modifiers
;
1147 /* Silently ignore PARTIALPRECISION if its not supported */
1148 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1150 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1152 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1153 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1158 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1161 case WINED3DSPDM_SATURATE
:
1164 case WINED3DSPDM_PARTIALPRECISION
:
1171 FIXME("Unknown modifiers 0x%08x\n", mod
);
1176 #define TEX_PROJ 0x1
1177 #define TEX_BIAS 0x2
1179 #define TEX_DERIV 0x10
1181 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1182 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1184 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1185 DWORD sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
1186 const char *tex_type
;
1187 BOOL np2_fixup
= FALSE
;
1188 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1189 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1190 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1192 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1194 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1195 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1197 switch(sampler_type
) {
1203 if(device
->stateBlock
->textures
[sampler_idx
] &&
1204 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
1209 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1211 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1213 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1214 else np2_fixup
= TRUE
;
1219 case WINED3DSTT_VOLUME
:
1223 case WINED3DSTT_CUBE
:
1228 ERR("Unexpected texture type %d\n", sampler_type
);
1232 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1233 * so don't use shader_arb_get_modifier
1235 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1238 /* Fragment samplers always have indentity mapping */
1239 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1241 sampler_idx
= priv
->cur_vs_args
->vertex_samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1244 if (flags
& TEX_DERIV
)
1246 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivates\n");
1247 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivates\n");
1248 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1249 dsx
, dsy
,sampler_idx
, tex_type
);
1251 else if(flags
& TEX_LOD
)
1253 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1254 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1255 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1256 sampler_idx
, tex_type
);
1258 else if (flags
& TEX_BIAS
)
1260 /* Shouldn't be possible, but let's check for it */
1261 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1262 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1263 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1265 else if (flags
& TEX_PROJ
)
1267 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1273 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1274 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1275 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1277 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, dst_str
, sampler_idx
, tex_type
);
1280 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1285 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
1286 "one", "coefmul.x", priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
1290 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1291 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1293 /* Generate a line that does the input modifier computation and return the input register to use */
1294 BOOL is_color
= FALSE
;
1298 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1299 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1301 /* Assume a new line will be added */
1304 /* Get register name */
1305 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1306 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1308 switch (src
->modifiers
)
1310 case WINED3DSPSM_NONE
:
1311 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1314 case WINED3DSPSM_NEG
:
1315 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1318 case WINED3DSPSM_BIAS
:
1319 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1321 case WINED3DSPSM_BIASNEG
:
1322 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1324 case WINED3DSPSM_SIGN
:
1325 shader_addline(buffer
, "MAD T%c, %s, coefmul.x, -one.x;\n", 'A' + tmpreg
, regstr
);
1327 case WINED3DSPSM_SIGNNEG
:
1328 shader_addline(buffer
, "MAD T%c, %s, -coefmul.x, one.x;\n", 'A' + tmpreg
, regstr
);
1330 case WINED3DSPSM_COMP
:
1331 shader_addline(buffer
, "SUB T%c, one.x, %s;\n", 'A' + tmpreg
, regstr
);
1333 case WINED3DSPSM_X2
:
1334 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1336 case WINED3DSPSM_X2NEG
:
1337 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1339 case WINED3DSPSM_DZ
:
1340 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1341 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1343 case WINED3DSPSM_DW
:
1344 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1345 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1347 case WINED3DSPSM_ABS
:
1348 if(ctx
->target_version
>= NV2
) {
1349 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1352 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1355 case WINED3DSPSM_ABSNEG
:
1356 if(ctx
->target_version
>= NV2
) {
1357 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1359 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1360 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1365 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1369 /* Return modified or original register, with swizzle */
1371 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1374 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1376 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1377 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1379 char src_name
[2][50];
1380 DWORD sampler_code
= dst
->reg
.idx
;
1382 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1384 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1386 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1387 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1390 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1391 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1392 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1393 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1394 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1396 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1397 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1400 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1405 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1406 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1407 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1408 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1409 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1410 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1411 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1412 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1413 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1414 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1415 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1416 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1417 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1419 FIXME("Unknown modifier %u\n", mod
);
1423 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1425 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1426 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1428 char src_name
[3][50];
1429 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1430 ins
->ctx
->reg_maps
->shader_version
.minor
);
1433 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1434 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1436 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1437 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
)
1439 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1441 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1444 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1445 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1447 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1448 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1449 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1450 /* No modifiers supported on CMP */
1451 shader_addline(buffer
, "CMP %s, TA, %s, %s;\n", dst_name
, src_name
[1], src_name
[2]);
1453 /* _SAT on CMP doesn't make much sense, but it is not a pure NOP */
1454 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
1456 shader_arb_get_register_name(ins
, &dst
->reg
, src_name
[0], &is_color
);
1457 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, dst_name
);
1462 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1464 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1465 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1467 char src_name
[3][50];
1470 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1472 /* Generate input register names (with modifiers) */
1473 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1474 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1475 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1477 /* No modifiers are supported on CMP */
1478 shader_addline(buffer
, "CMP %s, %s, %s, %s;\n", dst_name
,
1479 src_name
[0], src_name
[2], src_name
[1]);
1481 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
1483 shader_arb_get_register_name(ins
, &dst
->reg
, src_name
[0], &is_color
);
1484 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name
[0]);
1488 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1489 * dst = dot2(src0, src1) + src2 */
1490 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1492 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1493 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1495 char src_name
[3][50];
1496 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1498 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1499 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1500 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1502 if(ctx
->target_version
>= NV3
)
1504 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1505 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1506 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1507 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1509 else if(ctx
->target_version
>= NV2
)
1511 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1512 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1513 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1514 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1516 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1518 * .xyxy and other swizzles that we could get with this are not valid in
1519 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1521 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1522 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1523 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1525 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1527 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1528 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1532 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1533 /* 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]
1534 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1536 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1537 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1538 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1539 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1543 /* Map the opcode 1-to-1 to the GL code */
1544 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1546 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1547 const char *instruction
;
1548 char arguments
[256], dst_str
[50];
1550 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1552 switch (ins
->handler_idx
)
1554 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1555 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1556 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1557 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1558 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1559 case WINED3DSIH_DST
: instruction
= "DST"; break;
1560 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1561 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1562 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1563 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1564 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1565 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1566 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1567 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1568 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1569 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1570 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1571 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1572 case WINED3DSIH_SGN
: instruction
= "SSG"; break;
1573 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1574 default: instruction
= "";
1575 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1579 /* Note that shader_arb_add_dst_param() adds spaces. */
1580 arguments
[0] = '\0';
1581 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1582 for (i
= 0; i
< ins
->src_count
; ++i
)
1585 strcat(arguments
, ", ");
1586 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1587 strcat(arguments
, operand
);
1589 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1592 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
)
1594 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1595 shader_addline(buffer
, "NOP;\n");
1598 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1600 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1601 BOOL pshader
= shader_is_pshader_version(shader
->baseShader
.reg_maps
.shader_version
.type
);
1602 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1604 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1605 char src0_param
[256];
1607 if(ins
->handler_idx
== WINED3DSIH_MOVA
) {
1610 if(ctx
->target_version
>= NV2
) {
1611 shader_hw_map2gl(ins
);
1614 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1615 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1617 /* This implements the mova formula used in GLSL. The first two instructions
1618 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1622 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1624 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1625 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1627 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, mova_const.y;\n", write_mask
, src0_param
);
1628 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, mova_const.z, -mova_const.w;\n", write_mask
);
1630 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1631 shader_addline(buffer
, "ADD TA%s, TA, mova_const.x;\n", write_mask
);
1632 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1633 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
)
1635 shader_addline(buffer
, "ADD TA%s, TA, helper_const.z;\n", write_mask
);
1637 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1639 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1640 } else if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
1641 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)
1642 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1644 src0_param
[0] = '\0';
1645 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
)
1647 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1648 shader_addline(buffer
, "ADD TA.x, %s, helper_const.z;\n", src0_param
);
1649 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1653 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1654 * with more than one component. Thus replicate the first source argument over all
1655 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1656 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1657 tmp_src
.swizzle
= (tmp_src
.swizzle
& 0x3) * 0x55;
1658 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1659 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1662 else if(ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& ins
->dst
[0].reg
.idx
== 0 && pshader
)
1664 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) shader
;
1665 if(ctx
->cur_ps_args
->super
.srgb_correction
&& ps
->color0_mov
)
1667 shader_addline(buffer
, "#mov handled in srgb write code\n");
1670 shader_hw_map2gl(ins
);
1674 shader_hw_map2gl(ins
);
1678 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1680 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1681 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1684 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1685 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1687 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1689 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1691 const char *kilsrc
= "TA";
1694 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1695 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1701 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1702 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1703 * masked out components to 0(won't kill)
1705 char x
= '0', y
= '0', z
= '0', w
= '0';
1706 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1707 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1708 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1709 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1710 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1712 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1714 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1715 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1717 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1718 * or pass in any temporary register(in shader phase 2)
1720 if(ins
->ctx
->reg_maps
->shader_version
.minor
<= 3) {
1721 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
);
1723 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1725 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1726 shader_addline(buffer
, "KIL TA;\n");
1730 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1732 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1733 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1734 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1735 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1736 ins
->ctx
->reg_maps
->shader_version
.minor
);
1737 struct wined3d_shader_src_param src
;
1741 DWORD reg_sampler_code
;
1744 /* All versions have a destination register */
1745 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1747 /* 1.0-1.4: Use destination register number as texture code.
1748 2.0+: Use provided sampler number as texure code. */
1749 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1750 reg_sampler_code
= dst
->reg
.idx
;
1752 reg_sampler_code
= ins
->src
[1].reg
.idx
;
1754 /* 1.0-1.3: Use the texcoord varying.
1755 1.4+: Use provided coordinate source register. */
1756 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1757 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1759 /* TEX is the only instruction that can handle DW and DZ natively */
1761 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1762 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1763 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1767 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1768 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1769 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1771 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1774 if(reg_sampler_code
< MAX_TEXTURES
) {
1775 flags
= deviceImpl
->stateBlock
->textureState
[reg_sampler_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1777 if (flags
& WINED3DTTFF_PROJECTED
) {
1778 myflags
|= TEX_PROJ
;
1781 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1783 DWORD src_mod
= ins
->src
[0].modifiers
;
1784 if (src_mod
== WINED3DSPSM_DZ
) {
1785 /* TXP cannot handle DZ natively, so move the z coordinate to .w. reg_coord is a read-only
1786 * varying register, so we need a temp reg
1788 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
1789 strcpy(reg_coord
, "TA");
1790 myflags
|= TEX_PROJ
;
1791 } else if(src_mod
== WINED3DSPSM_DW
) {
1792 myflags
|= TEX_PROJ
;
1795 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
1796 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
1798 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
1801 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
1803 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1804 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1805 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1806 ins
->ctx
->reg_maps
->shader_version
.minor
);
1809 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1811 DWORD reg
= dst
->reg
.idx
;
1813 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1814 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
1818 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
1819 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1820 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
1824 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
1826 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1827 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1828 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1831 DWORD reg1
= ins
->dst
[0].reg
.idx
;
1835 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
1836 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1837 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
1838 /* Move .x first in case src_str is "TA" */
1839 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
1840 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
1841 flags
= reg1
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg1
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1842 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
1845 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
1847 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1849 DWORD reg1
= ins
->dst
[0].reg
.idx
;
1853 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
1854 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1855 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
1856 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
1857 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
1858 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
1861 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
1863 DWORD reg1
= ins
->dst
[0].reg
.idx
;
1867 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
1868 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1869 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
1870 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
1873 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
1875 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1876 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1877 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1878 char reg_coord
[40], dst_reg
[50], src_reg
[50];
1879 DWORD reg_dest_code
;
1881 /* All versions have a destination register. The Tx where the texture coordinates come
1882 * from is the varying incarnation of the texture register
1884 reg_dest_code
= dst
->reg
.idx
;
1885 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
1886 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
1887 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
1889 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1890 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
1892 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
1893 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
1895 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
1896 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
1899 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
1900 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
1901 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
1902 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
1904 /* with projective textures, texbem only divides the static texture coord, not the displacement,
1905 * so we can't let the GL handle this.
1907 if (((IWineD3DDeviceImpl
*) This
->baseShader
.device
)->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
]
1908 & WINED3DTTFF_PROJECTED
) {
1909 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
1910 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
1911 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
1913 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
1916 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
1918 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
1920 /* No src swizzles are allowed, so this is ok */
1921 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
1922 src_reg
, reg_dest_code
, reg_dest_code
);
1923 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
1927 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
1929 DWORD reg
= ins
->dst
[0].reg
.idx
;
1930 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1931 char src0_name
[50], dst_name
[50];
1933 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
1935 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1936 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
1937 * T<reg+1> register. Use this register to store the calculated vector
1939 tmp_reg
.idx
= reg
+ 1;
1940 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
1941 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
1944 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
1946 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1947 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1949 DWORD reg
= ins
->dst
[0].reg
.idx
;
1950 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1956 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
1957 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
1959 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1960 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1961 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
1962 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1963 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
1966 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
1968 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1969 DWORD reg
= ins
->dst
[0].reg
.idx
;
1970 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1971 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1972 char src0_name
[50], dst_name
[50];
1973 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
1976 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
1977 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
1978 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
1980 tmp_reg
.idx
= reg
+ 2 - current_state
->current_row
;
1981 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
1983 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
1984 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
1985 dst_name
, 'x' + current_state
->current_row
, reg
, src0_name
);
1986 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
1989 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
1991 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
1992 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1994 DWORD reg
= ins
->dst
[0].reg
.idx
;
1995 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1996 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
1998 char src0_name
[50], dst_name
[50];
2001 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2002 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2003 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2005 /* Sample the texture using the calculated coordinates */
2006 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2007 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
2008 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2009 current_state
->current_row
= 0;
2012 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2014 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2015 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2017 DWORD reg
= ins
->dst
[0].reg
.idx
;
2018 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2019 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2025 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2026 * components for temporary data storage
2028 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2029 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2030 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2032 /* Construct the eye-ray vector from w coordinates */
2033 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[0]);
2034 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[1]);
2035 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2037 /* Calculate reflection vector
2039 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2040 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2041 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2042 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2043 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2044 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2045 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2047 /* Sample the texture using the calculated coordinates */
2048 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2049 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
2050 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2051 current_state
->current_row
= 0;
2054 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2056 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*)ins
->ctx
->shader
;
2057 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
2059 DWORD reg
= ins
->dst
[0].reg
.idx
;
2060 SHADER_PARSE_STATE
* current_state
= &This
->baseShader
.parse_state
;
2061 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2068 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2069 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2070 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2071 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2072 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2074 /* Calculate reflection vector.
2077 * dst_reg.xyz = 2 * --------- * N - E
2080 * Which normalizes the normal vector
2082 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2083 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2084 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2085 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2086 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2087 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2089 /* Sample the texture using the calculated coordinates */
2090 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2091 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
2092 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2093 current_state
->current_row
= 0;
2096 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2098 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2099 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2102 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2103 * which is essentially an input, is the destination register because it is the first
2104 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2105 * here(writemasks/swizzles are not valid on texdepth)
2107 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2109 /* According to the msdn, the source register(must be r5) is unusable after
2110 * the texdepth instruction, so we're free to modify it
2112 shader_addline(buffer
, "MIN %s.y, %s.y, one.y;\n", dst_name
, dst_name
);
2114 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2115 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2116 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2118 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2119 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2120 shader_addline(buffer
, "MIN TA.x, TA.x, one.x;\n");
2121 shader_addline(buffer
, "MAX result.depth, TA.x, 0.0;\n");
2124 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2125 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2126 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2127 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2129 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2130 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
2134 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2135 shader_addline(buffer
, "MOV TB, 0.0;\n");
2136 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2138 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2139 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2142 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2143 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2144 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2146 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2149 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2151 /* Handle output register */
2152 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2153 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2154 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
, src0
);
2157 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2158 * Perform the 3rd row of a 3x3 matrix multiply */
2159 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2161 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2162 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2163 char dst_str
[50], dst_name
[50];
2167 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2168 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2169 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2170 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
, src0
);
2171 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2174 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2175 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2176 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2177 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2179 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2181 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2182 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2183 char src0
[50], dst_name
[50];
2186 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2187 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2188 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
, src0
);
2190 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2191 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2192 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2194 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2195 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2196 shader_addline(buffer
, "MIN %s.x, %s.x, one.x;\n", dst_name
, dst_name
);
2197 shader_addline(buffer
, "MAX result.depth, %s.x, 0.0;\n", dst_name
);
2200 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2201 Vertex/Pixel shaders to ARB_vertex_program codes */
2202 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2205 int nComponents
= 0;
2206 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2207 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2208 struct wined3d_shader_instruction tmp_ins
;
2210 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2212 /* Set constants for the temporary argument */
2213 tmp_ins
.ctx
= ins
->ctx
;
2214 tmp_ins
.dst_count
= 1;
2215 tmp_ins
.dst
= &tmp_dst
;
2216 tmp_ins
.src_count
= 2;
2217 tmp_ins
.src
= tmp_src
;
2219 switch(ins
->handler_idx
)
2221 case WINED3DSIH_M4x4
:
2223 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2225 case WINED3DSIH_M4x3
:
2227 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2229 case WINED3DSIH_M3x4
:
2231 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2233 case WINED3DSIH_M3x3
:
2235 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2237 case WINED3DSIH_M3x2
:
2239 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2242 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2246 tmp_dst
= ins
->dst
[0];
2247 tmp_src
[0] = ins
->src
[0];
2248 tmp_src
[1] = ins
->src
[1];
2249 for (i
= 0; i
< nComponents
; i
++) {
2250 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2251 shader_hw_map2gl(&tmp_ins
);
2252 ++tmp_src
[1].reg
.idx
;
2256 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2258 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2259 const char *instruction
;
2264 switch(ins
->handler_idx
)
2266 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2267 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2268 case WINED3DSIH_EXP
: instruction
= "EX2"; break;
2269 case WINED3DSIH_EXPP
: instruction
= "EXP"; break;
2270 default: instruction
= "";
2271 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2275 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2276 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2277 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2279 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2285 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2288 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2290 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2293 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2294 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2296 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2297 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2299 if(pshader
&& priv
->target_version
>= NV3
)
2301 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2305 shader_addline(buffer
, "DP3 TA, %s, %s;\n", src_name
, src_name
);
2306 shader_addline(buffer
, "RSQ TA, TA.x;\n");
2307 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2308 shader_addline(buffer
, "MUL%s %s, %s, TA;\n", shader_arb_get_modifier(ins
), dst_name
,
2313 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2315 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2317 char src_name
[3][50];
2319 /* ARB_fragment_program has a convenient LRP instruction */
2320 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2321 shader_hw_map2gl(ins
);
2325 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2326 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2327 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2328 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2330 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2331 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2332 dst_name
, src_name
[0], src_name
[2]);
2335 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2337 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2338 * must contain fixed constants. So we need a separate function to filter those constants and
2341 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2342 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2343 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2345 char src_name0
[50], src_name1
[50], src_name2
[50];
2348 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2349 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2350 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2351 /* No modifiers are supported on SCS */
2352 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2354 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2356 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2357 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2359 } else if(priv
->target_version
>= NV2
) {
2360 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2362 /* Sincos writemask must be .x, .y or .xy */
2363 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2364 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2365 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2366 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2368 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2369 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2371 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2372 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2374 * The constants we get are:
2376 * +1 +1, -1 -1 +1 +1 -1 -1
2377 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2378 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2380 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2384 * (x/2)^4 = x^4 / 16
2385 * (x/2)^5 = x^5 / 32
2388 * To get the final result:
2389 * sin(x) = 2 * sin(x/2) * cos(x/2)
2390 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2391 * (from sin(x+y) and cos(x+y) rules)
2393 * As per MSDN, dst.z is undefined after the operation, and so is
2394 * dst.x and dst.y if they're masked out by the writemask. Ie
2395 * sincos dst.y, src1, c0, c1
2396 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2397 * vsa.exe also stops with an error if the dest register is the same register as the source
2398 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2399 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2401 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2402 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2403 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2405 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2406 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2407 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2408 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2409 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2410 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2414 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2415 * properly merge that with MULs in the code above?
2416 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2417 * we can merge the sine and cosine MAD rows to calculate them together.
2419 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2420 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2421 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2422 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2425 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2426 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2427 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2429 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2431 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2432 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2434 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2436 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2437 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2442 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2444 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2447 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2449 /* SGN is only valid in vertex shaders */
2450 if(ctx
->target_version
== NV2
) {
2451 shader_hw_map2gl(ins
);
2454 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2455 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2457 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2458 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2460 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2461 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2463 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2464 * Then use TA, and calculate the final result
2466 * Not reading from TA? Store the first result in TA to avoid overwriting the
2467 * destination if src reg = dst reg
2469 if(strstr(src_name
, "TA"))
2471 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2472 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2473 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2477 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2478 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2479 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2484 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2486 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2492 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2493 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2494 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2496 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2497 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2500 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2506 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2507 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2508 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2509 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2510 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2511 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2512 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2513 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2514 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2515 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2516 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2517 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2518 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2520 FIXME("Unknown modifier %u\n", mod
);
2524 static void shader_hw_log_pow(const struct wined3d_shader_instruction
*ins
)
2526 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2527 char src0
[50], src1
[50], dst
[50];
2528 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2529 BOOL need_abs
= FALSE
;
2533 switch(ins
->handler_idx
)
2535 case WINED3DSIH_LOG
: instr
= "LG2"; break;
2536 case WINED3DSIH_LOGP
: instr
= "LOG"; break;
2537 case WINED3DSIH_POW
: instr
= "POW"; arg2
= TRUE
; break;
2539 ERR("Unexpected instruction %d\n", ins
->handler_idx
);
2543 /* LOG, LOGP and POW operate on the absolute value of the input */
2544 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2546 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2547 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2548 if(arg2
) shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2552 shader_addline(buffer
, "ABS TA, %s;\n", src0
);
2555 shader_addline(buffer
, "%s%s %s, TA, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src1
);
2559 shader_addline(buffer
, "%s%s %s, TA;\n", instr
, shader_arb_get_modifier(ins
), dst
);
2564 shader_addline(buffer
, "%s%s %s, %s, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src0
, src1
);
2568 shader_addline(buffer
, "%s%s %s, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src0
);
2572 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2574 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2576 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2579 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2583 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2584 struct list
*e
= list_head(&priv
->control_frames
);
2585 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2587 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2588 /* The constant loader makes sure to load -1 into iX.w */
2589 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2590 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->loop_no
);
2591 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->loop_no
);
2595 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2599 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2601 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2603 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2605 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2607 /* The constant loader makes sure to load -1 into iX.w */
2610 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2611 struct list
*e
= list_head(&priv
->control_frames
);
2612 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2614 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2616 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2617 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->loop_no
);
2618 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->loop_no
);
2622 shader_addline(buffer
, "REP %s;\n", src_name
);
2626 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2628 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2629 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2633 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2634 struct list
*e
= list_head(&priv
->control_frames
);
2635 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2637 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2638 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->loop_no
);
2639 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->loop_no
);
2641 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2645 shader_addline(buffer
, "ENDLOOP;\n");
2649 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2651 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2652 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2656 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2657 struct list
*e
= list_head(&priv
->control_frames
);
2658 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2660 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2661 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->loop_no
);
2662 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->loop_no
);
2664 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2668 shader_addline(buffer
, "ENDREP;\n");
2672 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2674 struct control_frame
*control_frame
;
2676 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2678 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2680 ERR("Could not find loop for break\n");
2684 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2686 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2687 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2688 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2692 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->loop_no
);
2696 shader_addline(buffer
, "BRK;\n");
2700 static const char *get_compare(COMPARISON_TYPE flags
)
2704 case COMPARISON_GT
: return "GT";
2705 case COMPARISON_EQ
: return "EQ";
2706 case COMPARISON_GE
: return "GE";
2707 case COMPARISON_LT
: return "LT";
2708 case COMPARISON_NE
: return "NE";
2709 case COMPARISON_LE
: return "LE";
2711 FIXME("Unrecognized comparison value: %u\n", flags
);
2716 static COMPARISON_TYPE
invert_compare(COMPARISON_TYPE flags
)
2720 case COMPARISON_GT
: return COMPARISON_LE
;
2721 case COMPARISON_EQ
: return COMPARISON_NE
;
2722 case COMPARISON_GE
: return COMPARISON_LT
;
2723 case COMPARISON_LT
: return COMPARISON_GE
;
2724 case COMPARISON_NE
: return COMPARISON_EQ
;
2725 case COMPARISON_LE
: return COMPARISON_GT
;
2727 FIXME("Unrecognized comparison value: %u\n", flags
);
2732 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
2734 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2735 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2736 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2739 const char *comp
= get_compare(ins
->flags
);
2741 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2742 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2746 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
2747 * away the subtraction result
2749 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2750 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->loop_no
, comp
);
2754 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2755 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
2759 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
2761 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2762 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2763 struct list
*e
= list_head(&priv
->control_frames
);
2764 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2768 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2770 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2771 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2775 /* Invert the flag. We jump to the else label if the condition is NOT true */
2776 comp
= get_compare(invert_compare(ins
->flags
));
2777 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2778 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->ifc_no
, comp
);
2782 comp
= get_compare(ins
->flags
);
2783 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2784 shader_addline(buffer
, "IF %s.x;\n", comp
);
2788 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
2790 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2791 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2792 struct list
*e
= list_head(&priv
->control_frames
);
2793 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2794 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2798 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->ifc_no
);
2799 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->ifc_no
);
2800 control_frame
->had_else
= TRUE
;
2804 shader_addline(buffer
, "ELSE;\n");
2808 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
2810 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2811 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2812 struct list
*e
= list_head(&priv
->control_frames
);
2813 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2814 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2818 if(control_frame
->had_else
)
2820 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->ifc_no
);
2824 shader_addline(buffer
, "#No else branch. else is endif\n");
2825 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->ifc_no
);
2830 shader_addline(buffer
, "ENDIF;\n");
2834 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
2836 DWORD sampler_idx
= ins
->src
[1].reg
.idx
;
2838 char reg_src
[3][40];
2839 DWORD flags
= TEX_DERIV
;
2841 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
2842 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
2843 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
2844 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
2846 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
2847 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
2849 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
2852 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
2854 DWORD sampler_idx
= ins
->src
[1].reg
.idx
;
2857 DWORD flags
= TEX_LOD
;
2859 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
2860 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
2862 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
2863 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
2865 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
2868 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
2870 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2871 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2873 priv
->in_main_func
= FALSE
;
2874 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
2875 * subroutine, don't generate a label that will make GL complain
2877 if(priv
->target_version
== ARB
) return;
2879 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
);
2882 static void vshader_add_footer(IWineD3DVertexShaderImpl
*This
, struct wined3d_shader_buffer
*buffer
,
2883 const struct arb_vs_compile_args
*args
, struct shader_arb_ctx_priv
*priv_ctx
)
2885 const shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
2886 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
2887 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
2890 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
2891 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
2892 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
2893 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
2895 if(args
->super
.fog_src
== VS_FOG_Z
) {
2896 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
2897 } else if (!reg_maps
->fog
) {
2898 /* posFixup.x is always 1.0, so we can savely use it */
2899 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
2902 /* Write the final position.
2904 * OpenGL coordinates specify the center of the pixel while d3d coords specify
2905 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
2906 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
2907 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
2909 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
2910 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
2911 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
2913 if(use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
2915 for(i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
2917 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
2920 else if(args
->boolclip
.clip_control
[0])
2922 unsigned int cur_clip
= 0;
2923 char component
[4] = {'x', 'y', 'z', 'w'};
2925 for(i
= 0; i
< GL_LIMITS(clipplanes
); i
++)
2927 if(args
->boolclip
.clip_control
[1] & (1 << i
))
2929 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
2930 component
[cur_clip
++], i
);
2936 shader_addline(buffer
, "MOV TA, -helper_const.w;\n");
2939 shader_addline(buffer
, "MOV TA.yzw, -helper_const.w;\n");
2942 shader_addline(buffer
, "MOV TA.zw, -helper_const.w;\n");
2945 shader_addline(buffer
, "MOV TA.w, -helper_const.w;\n");
2948 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
2949 args
->boolclip
.clip_control
[0] - 1);
2952 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
2953 * and the glsl equivalent
2955 if(need_helper_const(gl_info
)) {
2956 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, helper_const.x, -TMP_OUT.w;\n");
2958 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
2959 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
2962 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
2964 priv_ctx
->footer_written
= TRUE
;
2967 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
2969 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2970 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2971 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*) ins
->ctx
->shader
;
2972 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2974 if(priv
->target_version
== ARB
) return;
2978 if(priv
->in_main_func
) vshader_add_footer((IWineD3DVertexShaderImpl
*) shader
, buffer
, priv
->cur_vs_args
, priv
);
2981 shader_addline(buffer
, "RET;\n");
2984 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
2986 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2987 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
);
2990 /* GL locking is done by the caller */
2991 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
2993 GLuint program_id
= 0;
2994 const char *blt_vprogram
=
2996 "PARAM c[1] = { { 1, 0.5 } };\n"
2997 "MOV result.position, vertex.position;\n"
2998 "MOV result.color, c[0].x;\n"
2999 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3002 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3003 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3004 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_vprogram
), blt_vprogram
));
3006 if (glGetError() == GL_INVALID_OPERATION
) {
3008 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3009 FIXME("Vertex program error at position %d: %s\n", pos
,
3010 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3016 /* GL locking is done by the caller */
3017 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
, enum tex_types tex_type
)
3019 GLuint program_id
= 0;
3020 static const char * const blt_fprograms
[tex_type_count
] =
3027 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3028 "MOV result.depth.z, R0.x;\n"
3035 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3036 "MOV result.depth.z, R0.x;\n"
3041 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3042 "MOV result.depth.z, R0.x;\n"
3046 if (!blt_fprograms
[tex_type
])
3048 FIXME("tex_type %#x not supported\n", tex_type
);
3052 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3053 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3054 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(blt_fprograms
[tex_type
]), blt_fprograms
[tex_type
]));
3056 if (glGetError() == GL_INVALID_OPERATION
) {
3058 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3059 FIXME("Fragment program error at position %d: %s\n", pos
,
3060 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3066 static void arbfp_add_sRGB_correction(struct wined3d_shader_buffer
*buffer
, const char *fragcolor
,
3067 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3069 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3073 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3074 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.y;\n", tmp1
, fragcolor
);
3075 /* Calculate the > 0.0031308 case */
3076 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3077 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3078 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3079 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts1.w;\n", fragcolor
, fragcolor
);
3080 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts2.x;\n", fragcolor
, fragcolor
);
3081 /* Calculate the < case */
3082 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts1.x, %s;\n", fragcolor
, fragcolor
);
3086 /* Calculate the > 0.0031308 case */
3087 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts1.z;\n", tmp1
, fragcolor
);
3088 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts1.z;\n", tmp1
, fragcolor
);
3089 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts1.z;\n", tmp1
, fragcolor
);
3090 shader_addline(buffer
, "MUL %s, %s, srgb_consts1.w;\n", tmp1
, tmp1
);
3091 shader_addline(buffer
, "SUB %s, %s, srgb_consts2.x;\n", tmp1
, tmp1
);
3092 /* Calculate the < case */
3093 shader_addline(buffer
, "MUL %s, srgb_consts1.x, %s;\n", tmp2
, fragcolor
);
3094 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3095 shader_addline(buffer
, "SLT %s, srgb_consts1.y, %s;\n", tmp3
, fragcolor
);
3096 shader_addline(buffer
, "SGE %s, srgb_consts1.y, %s;\n", tmp4
, fragcolor
);
3097 /* Store the components > 0.0031308 in the destination */
3098 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3099 /* Add the components that are < 0.0031308 */
3100 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3101 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3102 * result.color writes(.rgb first, then .a), or handle overwriting already written
3103 * components. The assembler uses a temporary register in this case, which is usually
3104 * not allocated from one of our registers that were used earlier.
3107 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3108 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3111 static const DWORD
*find_loop_control_values(IWineD3DBaseShaderImpl
*This
, DWORD idx
)
3113 const local_constant
*constant
;
3115 LIST_FOR_EACH_ENTRY(constant
, &This
->baseShader
.constantsI
, local_constant
, entry
)
3117 if (constant
->idx
== idx
)
3119 return constant
->value
;
3125 static void init_ps_input(const IWineD3DPixelShaderImpl
*This
, const struct arb_ps_compile_args
*args
,
3126 struct shader_arb_ctx_priv
*priv
)
3128 const char *texcoords
[8] =
3130 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3131 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3134 const struct wined3d_shader_signature_element
*sig
= This
->input_signature
;
3135 const char *semantic_name
;
3138 switch(args
->super
.vp_mode
)
3140 case pretransformed
:
3142 /* The pixelshader has to collect the varyings on its own. In any case properly load
3143 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3144 * other attribs to 0.0.
3146 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3147 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3148 * load the texcoord attrib pointers to match the pixel shader signature
3150 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3152 semantic_name
= sig
[i
].semantic_name
;
3153 semantic_idx
= sig
[i
].semantic_idx
;
3154 if(semantic_name
== NULL
) continue;
3156 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3158 if(semantic_idx
== 0) priv
->ps_input
[i
] = "fragment.color.primary";
3159 else if(semantic_idx
== 1) priv
->ps_input
[i
] = "fragment.color.secondary";
3160 else priv
->ps_input
[i
] = "0.0";
3162 else if(args
->super
.vp_mode
== fixedfunction
)
3164 priv
->ps_input
[i
] = "0.0";
3166 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3168 if(semantic_idx
< 8) priv
->ps_input
[i
] = texcoords
[semantic_idx
];
3169 else priv
->ps_input
[i
] = "0.0";
3171 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3173 if(semantic_idx
== 0) priv
->ps_input
[i
] = "fragment.fogcoord";
3174 else priv
->ps_input
[i
] = "0.0";
3178 priv
->ps_input
[i
] = "0.0";
3181 TRACE("v%u, semantic %s%u is %s\n", i
, semantic_name
, semantic_idx
, priv
->ps_input
[i
]);
3186 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3189 for(i
= 0; i
< 8; i
++)
3191 priv
->ps_input
[i
] = texcoords
[i
];
3193 priv
->ps_input
[8] = "fragment.color.primary";
3194 priv
->ps_input
[9] = "fragment.color.secondary";
3199 /* GL locking is done by the caller */
3200 static GLuint
shader_arb_generate_pshader(IWineD3DPixelShaderImpl
*This
, struct wined3d_shader_buffer
*buffer
,
3201 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3203 const shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
3204 CONST DWORD
*function
= This
->baseShader
.function
;
3205 const struct wined3d_gl_info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3206 const local_constant
*lconst
;
3209 DWORD
*lconst_map
= local_const_mapping((IWineD3DBaseShaderImpl
*) This
), next_local
, cur
;
3210 struct shader_arb_ctx_priv priv_ctx
;
3211 BOOL dcl_tmp
= args
->super
.srgb_correction
, dcl_td
= FALSE
;
3212 BOOL want_nv_prog
= FALSE
;
3213 struct arb_pshader_private
*shader_priv
= This
->backend_priv
;
3216 unsigned int i
, found
= 0;
3218 for(i
= 0; i
< This
->baseShader
.limits
.temporary
; i
++) {
3220 /* Don't overwrite the color source */
3221 if(This
->color0_mov
&& i
== This
->color0_reg
) continue;
3222 else if(reg_maps
->shader_version
.major
< 2 && i
== 0) continue;
3224 if(reg_maps
->temporary
[i
]) {
3225 sprintf(srgbtmp
[found
], "R%u", i
);
3227 if(found
== 4) break;
3232 case 4: dcl_tmp
= FALSE
; break;
3234 sprintf(srgbtmp
[0], "TA");
3235 sprintf(srgbtmp
[1], "TB");
3236 sprintf(srgbtmp
[2], "TC");
3237 sprintf(srgbtmp
[3], "TD");
3241 sprintf(srgbtmp
[1], "TA");
3242 sprintf(srgbtmp
[2], "TB");
3243 sprintf(srgbtmp
[3], "TC");
3246 sprintf(srgbtmp
[2], "TA");
3247 sprintf(srgbtmp
[3], "TB");
3250 sprintf(srgbtmp
[3], "TA");
3254 /* Create the hw ARB shader */
3255 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3256 priv_ctx
.cur_ps_args
= args
;
3257 priv_ctx
.compiled_fprog
= compiled
;
3258 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3259 init_ps_input(This
, args
, &priv_ctx
);
3260 list_init(&priv_ctx
.control_frames
);
3262 /* Avoid enabling NV_fragment_program* if we do not need it.
3264 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3265 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3266 * is faster than what we gain from using higher native instructions. There are some things though
3267 * that cannot be emulated. In that case enable the extensions.
3268 * If the extension is enabled, instruction handlers that support both ways will use it.
3270 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3271 * So enable the best we can get.
3273 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3274 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3276 want_nv_prog
= TRUE
;
3279 shader_addline(buffer
, "!!ARBfp1.0\n");
3280 if(want_nv_prog
&& GL_SUPPORT(NV_FRAGMENT_PROGRAM2
)) {
3281 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3282 priv_ctx
.target_version
= NV3
;
3283 } else if(want_nv_prog
&& GL_SUPPORT(NV_FRAGMENT_PROGRAM_OPTION
)) {
3284 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3285 priv_ctx
.target_version
= NV2
;
3289 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3292 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3295 priv_ctx
.target_version
= ARB
;
3298 if(This
->baseShader
.reg_maps
.highest_render_target
> 0)
3300 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3303 if (reg_maps
->shader_version
.major
< 3)
3305 switch(args
->super
.fog
) {
3309 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3312 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3315 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3320 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3321 * unused temps away(but occupies them for the whole shader if they're used once). Always
3322 * declaring them avoids tricky bookkeeping work
3324 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3325 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3326 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3327 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3328 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3329 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3330 shader_addline(buffer
, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
3332 if (reg_maps
->shader_version
.major
< 2)
3334 strcpy(fragcolor
, "R0");
3336 if(args
->super
.srgb_correction
) {
3337 if(This
->color0_mov
) {
3338 sprintf(fragcolor
, "R%u", This
->color0_reg
);
3340 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3341 strcpy(fragcolor
, "TMP_COLOR");
3344 strcpy(fragcolor
, "result.color");
3348 if(args
->super
.srgb_correction
) {
3349 shader_addline(buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
3350 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
3351 shader_addline(buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
3352 srgb_sub_high
, 0.0, 0.0, 0.0);
3355 /* Base Declarations */
3356 next_local
= shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
,
3357 lconst_map
, NULL
, &priv_ctx
);
3359 for(i
= 0; i
< (sizeof(reg_maps
->bumpmat
) / sizeof(reg_maps
->bumpmat
[0])); i
++) {
3360 if(!reg_maps
->bumpmat
[i
]) continue;
3362 cur
= compiled
->numbumpenvmatconsts
;
3363 compiled
->bumpenvmatconst
[cur
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3364 compiled
->bumpenvmatconst
[cur
].texunit
= i
;
3365 compiled
->luminanceconst
[cur
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3366 compiled
->luminanceconst
[cur
].texunit
= i
;
3368 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3369 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3370 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3371 * textures due to conditional NP2 restrictions)
3373 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3374 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3375 * their location is shader dependent anyway and they cannot be loaded globally.
3377 compiled
->bumpenvmatconst
[cur
].const_num
= next_local
++;
3378 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3379 i
, compiled
->bumpenvmatconst
[cur
].const_num
);
3380 compiled
->numbumpenvmatconsts
= cur
+ 1;
3382 if(!reg_maps
->luminanceparams
[i
]) continue;
3384 compiled
->luminanceconst
[cur
].const_num
= next_local
++;
3385 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3386 i
, compiled
->luminanceconst
[cur
].const_num
);
3389 for(i
= 0; i
< MAX_CONST_I
; i
++)
3391 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3392 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3394 const DWORD
*control_values
= find_loop_control_values((IWineD3DBaseShaderImpl
*) This
, i
);
3398 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3399 control_values
[0], control_values
[1], control_values
[2]);
3403 compiled
->int_consts
[i
] = next_local
;
3404 compiled
->num_int_consts
++;
3405 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3410 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3412 compiled
->ycorrection
= next_local
;
3413 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3417 shader_addline(buffer
, "TEMP vpos;\n");
3418 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3419 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3420 * ycorrection.z: 1.0
3421 * ycorrection.w: 0.0
3423 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3424 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3429 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3432 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3433 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3434 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3435 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3436 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3437 * shader compilation errors and the subsequent errors when drawing with this shader. */
3438 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3440 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3441 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3442 const UINT max_lconsts
= gl_info
->ps_arb_max_local_constants
;
3444 fixup
->offset
= next_local
;
3445 fixup
->super
.active
= 0;
3448 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3449 if (!(map
& (1 << i
))) continue;
3451 if (fixup
->offset
+ (cur
>> 1) < max_lconsts
) {
3452 fixup
->super
.active
|= (1 << i
);
3453 fixup
->super
.idx
[i
] = cur
++;
3455 FIXME("No free constant found to load NP2 fixup data into shader. "
3456 "Sampling from this texture will probably look wrong.\n");
3461 fixup
->super
.num_consts
= (cur
+ 1) >> 1;
3462 if (fixup
->super
.num_consts
) {
3463 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3464 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3467 next_local
+= fixup
->super
.num_consts
;
3470 if(shader_priv
->clipplane_emulation
)
3472 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
- 1);
3475 /* Base Shader Body */
3476 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
3478 if(args
->super
.srgb_correction
) {
3479 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3480 priv_ctx
.target_version
>= NV2
);
3481 } else if(reg_maps
->shader_version
.major
< 2) {
3482 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3484 shader_addline(buffer
, "END\n");
3486 /* TODO: change to resource.glObjectHandle or something like that */
3487 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3489 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3490 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3492 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3493 /* Create the program and check for errors */
3494 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3495 buffer
->bsize
, buffer
->buffer
));
3497 if (glGetError() == GL_INVALID_OPERATION
) {
3499 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
3500 FIXME("HW PixelShader Error at position %d: %s\n",
3501 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3505 /* Load immediate constants */
3507 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
3508 const float *value
= (const float *)lconst
->value
;
3509 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, lconst_map
[lconst
->idx
], value
));
3510 checkGLcall("glProgramLocalParameter4fvARB");
3512 HeapFree(GetProcessHeap(), 0, lconst_map
);
3518 static int compare_sig(const struct wined3d_shader_signature_element
*sig1
, const struct wined3d_shader_signature_element
*sig2
)
3523 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3525 if(sig1
[i
].semantic_name
== NULL
|| sig2
[i
].semantic_name
== NULL
)
3527 /* Compare pointers, not contents. One string is NULL(element does not exist), the other one is not NULL */
3528 if(sig1
[i
].semantic_name
!= sig2
[i
].semantic_name
) return sig1
[i
].semantic_name
< sig2
[i
].semantic_name
? -1 : 1;
3532 ret
= strcmp(sig1
[i
].semantic_name
, sig2
[i
].semantic_name
);
3533 if(ret
!= 0) return ret
;
3534 if(sig1
[i
].semantic_idx
!= sig2
[i
].semantic_idx
) return sig1
[i
].semantic_idx
< sig2
[i
].semantic_idx
? -1 : 1;
3535 if(sig1
[i
].sysval_semantic
!= sig2
[i
].sysval_semantic
) return sig1
[i
].sysval_semantic
< sig2
[i
].sysval_semantic
? -1 : 1;
3536 if(sig1
[i
].component_type
!= sig2
[i
].component_type
) return sig1
[i
].sysval_semantic
< sig2
[i
].component_type
? -1 : 1;
3537 if(sig1
[i
].register_idx
!= sig2
[i
].register_idx
) return sig1
[i
].register_idx
< sig2
[i
].register_idx
? -1 : 1;
3538 if(sig1
[i
].mask
!= sig2
->mask
) return sig1
[i
].mask
< sig2
[i
].mask
? -1 : 1;
3543 static struct wined3d_shader_signature_element
*clone_sig(const struct wined3d_shader_signature_element
*sig
)
3545 struct wined3d_shader_signature_element
*new;
3549 new = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*new) * MAX_REG_INPUT
);
3550 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3552 if(sig
[i
].semantic_name
== NULL
)
3558 /* Clone the semantic string */
3559 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
[i
].semantic_name
) + 1);
3560 strcpy(name
, sig
[i
].semantic_name
);
3561 new[i
].semantic_name
= name
;
3566 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature_element
*sig
)
3568 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3569 struct ps_signature
*found_sig
;
3573 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3574 TRACE("Found existing signature %u\n", found_sig
->idx
);
3575 return found_sig
->idx
;
3577 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*sig
));
3578 found_sig
->sig
= clone_sig(sig
);
3579 found_sig
->idx
= priv
->ps_sig_number
++;
3580 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3581 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3583 ERR("Failed to insert program entry.\n");
3585 return found_sig
->idx
;
3588 static void init_output_registers(IWineD3DVertexShaderImpl
*shader
, DWORD sig_num
, struct shader_arb_ctx_priv
*priv_ctx
,
3589 struct arb_vs_compiled_shader
*compiled
)
3592 static const char *texcoords
[8] =
3594 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3595 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3597 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) shader
->baseShader
.device
;
3598 const struct wined3d_shader_signature_element
*sig
;
3599 const char *semantic_name
;
3600 DWORD semantic_idx
, reg_idx
;
3602 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3603 * and varying 9 to result.color.secondary
3605 const char *decl_idx_to_string
[MAX_REG_INPUT
] =
3607 texcoords
[0], texcoords
[1], texcoords
[2], texcoords
[3],
3608 texcoords
[4], texcoords
[5], texcoords
[6], texcoords
[7],
3609 "result.color.primary", "result.color.secondary"
3614 TRACE("Pixel shader uses builtin varyings\n");
3615 /* Map builtins to builtins */
3616 for(i
= 0; i
< 8; i
++)
3618 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3620 priv_ctx
->color_output
[0] = "result.color.primary";
3621 priv_ctx
->color_output
[1] = "result.color.secondary";
3622 priv_ctx
->fog_output
= "result.fogcoord";
3624 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3625 for(i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); i
++)
3627 semantic_name
= shader
->output_signature
[i
].semantic_name
;
3628 if(semantic_name
== NULL
) continue;
3630 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3632 TRACE("o%u is TMP_OUT\n", i
);
3633 if(shader
->output_signature
[i
].semantic_idx
== 0) priv_ctx
->vs_output
[i
] = "TMP_OUT";
3634 else priv_ctx
->vs_output
[i
] = "TA";
3636 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3638 TRACE("o%u is result.pointsize\n", i
);
3639 if(shader
->output_signature
[i
].semantic_idx
== 0) priv_ctx
->vs_output
[i
] = "result.pointsize";
3640 else priv_ctx
->vs_output
[i
] = "TA";
3642 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3644 TRACE("o%u is result.color.?, idx %u\n", i
, shader
->output_signature
[i
].semantic_idx
);
3645 if(shader
->output_signature
[i
].semantic_idx
== 0) priv_ctx
->vs_output
[i
] = "result.color.primary";
3646 else if(shader
->output_signature
[i
].semantic_idx
== 1) priv_ctx
->vs_output
[i
] = "result.color.secondary";
3647 else priv_ctx
->vs_output
[i
] = "TA";
3649 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3651 TRACE("o%u is %s\n", i
, texcoords
[shader
->output_signature
[i
].semantic_idx
]);
3652 if(shader
->output_signature
[i
].semantic_idx
>= 8) priv_ctx
->vs_output
[i
] = "TA";
3653 else priv_ctx
->vs_output
[i
] = texcoords
[shader
->output_signature
[i
].semantic_idx
];
3655 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3657 TRACE("o%u is result.fogcoord\n", i
);
3658 if(shader
->output_signature
[i
].semantic_idx
> 0) priv_ctx
->vs_output
[i
] = "TA";
3659 else priv_ctx
->vs_output
[i
] = "result.fogcoord";
3663 priv_ctx
->vs_output
[i
] = "TA";
3669 /* Instead of searching for the signature in the signature list, read the one from the current pixel shader.
3670 * Its maybe not the shader where the signature came from, but it is the same signature and faster to find
3672 sig
= ((IWineD3DPixelShaderImpl
*)device
->stateBlock
->pixelShader
)->input_signature
;
3673 TRACE("Pixel shader uses declared varyings\n");
3675 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
3676 for(i
= 0; i
< 8; i
++)
3678 priv_ctx
->texcrd_output
[i
] = "TA";
3680 priv_ctx
->color_output
[0] = "TA";
3681 priv_ctx
->color_output
[1] = "TA";
3682 priv_ctx
->fog_output
= "TA";
3684 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3686 semantic_name
= sig
[i
].semantic_name
;
3687 semantic_idx
= sig
[i
].semantic_idx
;
3688 reg_idx
= sig
[i
].register_idx
;
3689 if(semantic_name
== NULL
) continue;
3691 /* If a declared input register is not written by builtin arguments, don't write to it.
3692 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
3694 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
3695 * to TMP_OUT in any case
3697 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3699 if(semantic_idx
< 8) priv_ctx
->texcrd_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
3701 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3703 if(semantic_idx
< 2) priv_ctx
->color_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
3705 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3707 if(semantic_idx
== 0) priv_ctx
->fog_output
= decl_idx_to_string
[reg_idx
];
3714 if(strcmp(decl_idx_to_string
[reg_idx
], "result.color.primary") == 0 ||
3715 strcmp(decl_idx_to_string
[reg_idx
], "result.color.secondary") == 0)
3717 compiled
->need_color_unclamp
= TRUE
;
3721 /* Map declared to declared */
3722 for(i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); i
++)
3724 /* Write unread output to TA to throw them away */
3725 priv_ctx
->vs_output
[i
] = "TA";
3726 semantic_name
= shader
->output_signature
[i
].semantic_name
;
3727 if(semantic_name
== NULL
)
3732 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
) &&
3733 shader
->output_signature
[i
].semantic_idx
== 0)
3735 priv_ctx
->vs_output
[i
] = "TMP_OUT";
3738 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
) &&
3739 shader
->output_signature
[i
].semantic_idx
== 0)
3741 priv_ctx
->vs_output
[i
] = "result.pointsize";
3745 for(j
= 0; j
< MAX_REG_INPUT
; j
++)
3747 if(sig
[j
].semantic_name
== NULL
)
3752 if(strcmp(sig
[j
].semantic_name
, semantic_name
) == 0 &&
3753 sig
[j
].semantic_idx
== shader
->output_signature
[i
].semantic_idx
)
3755 priv_ctx
->vs_output
[i
] = decl_idx_to_string
[sig
[j
].register_idx
];
3757 if(strcmp(priv_ctx
->vs_output
[i
], "result.color.primary") == 0 ||
3758 strcmp(priv_ctx
->vs_output
[i
], "result.color.secondary") == 0)
3760 compiled
->need_color_unclamp
= TRUE
;
3767 /* GL locking is done by the caller */
3768 static GLuint
shader_arb_generate_vshader(IWineD3DVertexShaderImpl
*This
, struct wined3d_shader_buffer
*buffer
,
3769 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
)
3771 const shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3772 CONST DWORD
*function
= This
->baseShader
.function
;
3773 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3774 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
3775 const local_constant
*lconst
;
3777 DWORD next_local
, *lconst_map
= local_const_mapping((IWineD3DBaseShaderImpl
*) This
);
3778 struct shader_arb_ctx_priv priv_ctx
;
3781 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3782 priv_ctx
.cur_vs_args
= args
;
3783 list_init(&priv_ctx
.control_frames
);
3784 init_output_registers(This
, args
->ps_signature
, &priv_ctx
, compiled
);
3786 /* Create the hw ARB shader */
3787 shader_addline(buffer
, "!!ARBvp1.0\n");
3789 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
3790 * mesurable performance penalty, and we can always make use of it for clipplanes.
3792 if(GL_SUPPORT(NV_VERTEX_PROGRAM3
)) {
3793 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
3794 priv_ctx
.target_version
= NV3
;
3795 shader_addline(buffer
, "ADDRESS aL;\n");
3796 } else if(GL_SUPPORT(NV_VERTEX_PROGRAM2_OPTION
)) {
3797 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
3798 priv_ctx
.target_version
= NV2
;
3799 shader_addline(buffer
, "ADDRESS aL;\n");
3801 priv_ctx
.target_version
= ARB
;
3804 shader_addline(buffer
, "TEMP TMP_OUT;\n");
3805 if(need_helper_const(gl_info
)) {
3806 shader_addline(buffer
, "PARAM helper_const = { 2.0, -1.0, %d.0, 0.0 };\n", This
->rel_offset
);
3808 if(need_mova_const((IWineD3DBaseShader
*) This
, gl_info
)) {
3809 shader_addline(buffer
, "PARAM mova_const = { 0.5, 0.0, 2.0, 1.0 };\n");
3810 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
3813 shader_addline(buffer
, "TEMP TA;\n");
3815 /* Base Declarations */
3816 next_local
= shader_generate_arb_declarations( (IWineD3DBaseShader
*) This
, reg_maps
, buffer
, &GLINFO_LOCATION
,
3817 lconst_map
, &priv_ctx
.vs_clipplanes
, &priv_ctx
);
3819 for(i
= 0; i
< MAX_CONST_I
; i
++)
3821 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3822 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3824 const DWORD
*control_values
= find_loop_control_values((IWineD3DBaseShaderImpl
*) This
, i
);
3828 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3829 control_values
[0], control_values
[1], control_values
[2]);
3833 compiled
->int_consts
[i
] = next_local
;
3834 compiled
->num_int_consts
++;
3835 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3840 /* We need a constant to fixup the final position */
3841 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
3842 compiled
->pos_fixup
= next_local
++;
3844 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
3845 * for output parameters. D3D in theory does not do that either, but some applications depend on a
3846 * proper initialization of the secondary color, and programs using the fixed function pipeline without
3847 * a replacement shader depend on the texcoord.w being set properly.
3849 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
3850 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
3851 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
3852 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
3853 * this can eat a number of instructions, so skip it unless this cap is set as well
3855 if(!GL_SUPPORT(NV_VERTEX_PROGRAM
)) {
3856 shader_addline(buffer
, "MOV result.color.secondary, -helper_const.wwwy;\n");
3858 if ((GLINFO_LOCATION
).quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !device
->frag_pipe
->ffp_proj_control
)
3861 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++) {
3862 if(This
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
3863 This
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
3864 shader_addline(buffer
, "MOV result.texcoord[%u].w, -helper_const.y;\n", i
);
3870 /* The shader starts with the main function */
3871 priv_ctx
.in_main_func
= TRUE
;
3872 /* Base Shader Body */
3873 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
3875 if(!priv_ctx
.footer_written
) vshader_add_footer(This
, buffer
, args
, &priv_ctx
);
3877 shader_addline(buffer
, "END\n");
3879 /* TODO: change to resource.glObjectHandle or something like that */
3880 GL_EXTCALL(glGenProgramsARB(1, &ret
));
3882 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
3883 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
3885 TRACE("Created hw vertex shader, prg=%d\n", ret
);
3886 /* Create the program and check for errors */
3887 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3888 buffer
->bsize
, buffer
->buffer
));
3890 if (glGetError() == GL_INVALID_OPERATION
) {
3892 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
3893 FIXME("HW VertexShader Error at position %d: %s\n",
3894 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3897 /* Load immediate constants */
3899 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
3900 const float *value
= (const float *)lconst
->value
;
3901 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, lconst_map
[lconst
->idx
], value
));
3905 HeapFree(GetProcessHeap(), 0, lconst_map
);
3910 static void find_clip_texcoord(IWineD3DPixelShaderImpl
*ps
)
3912 struct arb_pshader_private
*shader_priv
= ps
->backend_priv
;
3914 const struct wined3d_gl_info
*gl_info
= &((IWineD3DDeviceImpl
*)ps
->baseShader
.device
)->adapter
->gl_info
;
3916 /* See if we can use fragment.texcoord[7] for clipplane emulation
3918 * Don't do this if it is not supported, or fragment.texcoord[7] is used
3920 if(ps
->baseShader
.reg_maps
.shader_version
.major
< 3)
3922 for(i
= GL_LIMITS(texture_stages
); i
> 0; i
--)
3924 if(!ps
->baseShader
.reg_maps
.texcoord
[i
- 1])
3926 shader_priv
->clipplane_emulation
= i
;
3930 WARN("Did not find a free clip reg(2.0)\n");
3934 for(i
= GL_LIMITS(texture_stages
); i
> 0; i
--)
3936 if(!(ps
->baseShader
.reg_maps
.input_registers
& (1 << (i
- 1))))
3938 shader_priv
->clipplane_emulation
= i
;
3942 WARN("Did not find a free clip reg(3.0)\n");
3946 /* GL locking is done by the caller */
3947 static struct arb_ps_compiled_shader
*find_arb_pshader(IWineD3DPixelShaderImpl
*shader
, const struct arb_ps_compile_args
*args
)
3951 struct arb_ps_compiled_shader
*new_array
;
3952 struct wined3d_shader_buffer buffer
;
3953 struct arb_pshader_private
*shader_data
;
3956 if(!shader
->backend_priv
) {
3957 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) shader
->baseShader
.device
;
3958 struct shader_arb_priv
*priv
= device
->shader_priv
;
3960 shader
->backend_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
3961 shader_data
= shader
->backend_priv
;
3962 shader_data
->clamp_consts
= shader
->baseShader
.reg_maps
.shader_version
.major
== 1;
3964 if(shader
->baseShader
.reg_maps
.shader_version
.major
< 3) shader_data
->input_signature_idx
= ~0;
3965 else shader_data
->input_signature_idx
= find_input_signature(priv
, shader
->input_signature
);
3967 shader_data
->has_signature_idx
= TRUE
;
3968 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
3970 if(!device
->vs_clipping
) find_clip_texcoord(shader
);
3972 shader_data
= shader
->backend_priv
;
3974 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
3975 * so a linear search is more performant than a hashmap or a binary search
3976 * (cache coherency etc)
3978 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
3979 if(memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)) == 0) {
3980 return &shader_data
->gl_shaders
[i
];
3984 TRACE("No matching GL shader found, compiling a new shader\n");
3985 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
3986 if (shader_data
->num_gl_shaders
)
3988 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
3989 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
3990 new_size
* sizeof(*shader_data
->gl_shaders
));
3992 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
3997 ERR("Out of memory\n");
4000 shader_data
->gl_shaders
= new_array
;
4001 shader_data
->shader_array_size
= new_size
;
4004 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4006 pixelshader_update_samplers(&shader
->baseShader
.reg_maps
,
4007 ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->stateBlock
->textures
);
4009 if (!shader_buffer_init(&buffer
))
4011 ERR("Failed to initialize shader buffer.\n");
4015 ret
= shader_arb_generate_pshader(shader
, &buffer
, args
,
4016 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4017 shader_buffer_free(&buffer
);
4018 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4020 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4023 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4024 const DWORD use_map
, BOOL skip_int
) {
4025 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4026 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4027 if(stored
->boolclip_compare
!= new->boolclip_compare
) return FALSE
;
4028 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4029 if(stored
->vertex_samplers_compare
!= new->vertex_samplers_compare
) return FALSE
;
4030 if(skip_int
) return TRUE
;
4032 return memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
)) == 0;
4035 static struct arb_vs_compiled_shader
*find_arb_vshader(IWineD3DVertexShaderImpl
*shader
, const struct arb_vs_compile_args
*args
)
4039 struct arb_vs_compiled_shader
*new_array
;
4040 DWORD use_map
= ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->strided_streams
.use_map
;
4041 struct wined3d_shader_buffer buffer
;
4042 struct arb_vshader_private
*shader_data
;
4044 const struct wined3d_gl_info
*gl_info
= &((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->adapter
->gl_info
;
4046 if(!shader
->backend_priv
) {
4047 shader
->backend_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4049 shader_data
= shader
->backend_priv
;
4051 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4052 * so a linear search is more performant than a hashmap or a binary search
4053 * (cache coherency etc)
4055 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4056 if(vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
, use_map
, GL_SUPPORT(NV_VERTEX_PROGRAM2_OPTION
))) {
4057 return &shader_data
->gl_shaders
[i
];
4061 TRACE("No matching GL shader found, compiling a new shader\n");
4063 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4064 if (shader_data
->num_gl_shaders
)
4066 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4067 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4068 new_size
* sizeof(*shader_data
->gl_shaders
));
4070 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4075 ERR("Out of memory\n");
4078 shader_data
->gl_shaders
= new_array
;
4079 shader_data
->shader_array_size
= new_size
;
4082 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4084 if (!shader_buffer_init(&buffer
))
4086 ERR("Failed to initialize shader buffer.\n");
4090 ret
= shader_arb_generate_vshader(shader
, &buffer
, args
,
4091 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4092 shader_buffer_free(&buffer
);
4093 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4095 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4098 static inline void find_arb_ps_compile_args(IWineD3DPixelShaderImpl
*shader
, IWineD3DStateBlockImpl
*stateblock
,
4099 struct arb_ps_compile_args
*args
)
4103 const struct wined3d_gl_info
*gl_info
= &((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->adapter
->gl_info
;
4104 find_ps_compile_args(shader
, stateblock
, &args
->super
);
4106 /* This forces all local boolean constants to 1 to make them stateblock independent */
4107 args
->bools
= shader
->baseShader
.reg_maps
.local_bool_consts
;
4109 for(i
= 0; i
< MAX_CONST_B
; i
++)
4111 if(stateblock
->pixelShaderConstantB
[i
]) args
->bools
|= ( 1 << i
);
4114 /* Skip if unused or local, or supported natively */
4115 int_skip
= ~shader
->baseShader
.reg_maps
.integer_constants
| shader
->baseShader
.reg_maps
.local_int_consts
;
4116 if(int_skip
== 0xffff || GL_SUPPORT(NV_FRAGMENT_PROGRAM_OPTION
))
4118 memset(&args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4122 for(i
= 0; i
< MAX_CONST_I
; i
++)
4124 if(int_skip
& (1 << i
))
4126 args
->loop_ctrl
[i
][0] = 0;
4127 args
->loop_ctrl
[i
][1] = 0;
4128 args
->loop_ctrl
[i
][2] = 0;
4132 args
->loop_ctrl
[i
][0] = stateblock
->pixelShaderConstantI
[i
* 4];
4133 args
->loop_ctrl
[i
][1] = stateblock
->pixelShaderConstantI
[i
* 4 + 1];
4134 args
->loop_ctrl
[i
][2] = stateblock
->pixelShaderConstantI
[i
* 4 + 2];
4139 static inline void find_arb_vs_compile_args(IWineD3DVertexShaderImpl
*shader
, IWineD3DStateBlockImpl
*stateblock
,
4140 struct arb_vs_compile_args
*args
)
4144 IWineD3DDeviceImpl
*dev
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
4145 const struct wined3d_gl_info
*gl_info
= &dev
->adapter
->gl_info
;
4146 find_vs_compile_args(shader
, stateblock
, &args
->super
);
4148 args
->boolclip_compare
= 0;
4149 if(use_ps(stateblock
))
4151 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
;
4152 struct arb_pshader_private
*shader_priv
= ps
->backend_priv
;
4153 args
->ps_signature
= shader_priv
->input_signature_idx
;
4155 args
->boolclip
.clip_control
[0] = shader_priv
->clipplane_emulation
;
4159 args
->ps_signature
= ~0;
4160 if(!dev
->vs_clipping
)
4162 args
->boolclip
.clip_control
[0] = ffp_clip_emul(stateblock
) ? GL_LIMITS(texture_stages
) : 0;
4164 /* Otherwise: Setting boolclip_compare set clip_control[0] to 0 */
4167 if(args
->boolclip
.clip_control
[0])
4169 if(stateblock
->renderState
[WINED3DRS_CLIPPING
])
4171 args
->boolclip
.clip_control
[1] = stateblock
->renderState
[WINED3DRS_CLIPPLANEENABLE
];
4173 /* clip_control[1] was set to 0 by setting boolclip_compare to 0 */
4176 /* This forces all local boolean constants to 1 to make them stateblock independent */
4177 args
->boolclip
.bools
= shader
->baseShader
.reg_maps
.local_bool_consts
;
4178 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4179 for(i
= 0; i
< MAX_CONST_B
; i
++)
4181 if(stateblock
->vertexShaderConstantB
[i
]) args
->boolclip
.bools
|= ( 1 << i
);
4184 args
->vertex_samplers
[0] = dev
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 0];
4185 args
->vertex_samplers
[1] = dev
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 1];
4186 args
->vertex_samplers
[2] = dev
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 2];
4187 args
->vertex_samplers
[3] = 0;
4189 /* Skip if unused or local */
4190 int_skip
= ~shader
->baseShader
.reg_maps
.integer_constants
| shader
->baseShader
.reg_maps
.local_int_consts
;
4191 if(int_skip
== 0xffff || GL_SUPPORT(NV_VERTEX_PROGRAM2_OPTION
)) /* This is about flow control, not clipping */
4193 memset(&args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4197 for(i
= 0; i
< MAX_CONST_I
; i
++)
4199 if(int_skip
& (1 << i
))
4201 args
->loop_ctrl
[i
][0] = 0;
4202 args
->loop_ctrl
[i
][1] = 0;
4203 args
->loop_ctrl
[i
][2] = 0;
4207 args
->loop_ctrl
[i
][0] = stateblock
->vertexShaderConstantI
[i
* 4];
4208 args
->loop_ctrl
[i
][1] = stateblock
->vertexShaderConstantI
[i
* 4 + 1];
4209 args
->loop_ctrl
[i
][2] = stateblock
->vertexShaderConstantI
[i
* 4 + 2];
4214 /* GL locking is done by the caller */
4215 static void shader_arb_select(IWineD3DDevice
*iface
, BOOL usePS
, BOOL useVS
) {
4216 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4217 struct shader_arb_priv
*priv
= This
->shader_priv
;
4218 struct wined3d_context
*context
= context_get_current();
4219 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4222 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4224 struct arb_ps_compile_args compile_args
;
4225 struct arb_ps_compiled_shader
*compiled
;
4226 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) This
->stateBlock
->pixelShader
;
4228 TRACE("Using pixel shader %p\n", This
->stateBlock
->pixelShader
);
4229 find_arb_ps_compile_args(ps
, This
->stateBlock
, &compile_args
);
4230 compiled
= find_arb_pshader(ps
, &compile_args
);
4231 priv
->current_fprogram_id
= compiled
->prgId
;
4232 priv
->compiled_fprog
= compiled
;
4234 /* Bind the fragment program */
4235 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4236 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4238 if(!priv
->use_arbfp_fixed_func
) {
4239 /* Enable OpenGL fragment programs */
4240 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4241 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4243 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
4245 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4246 * a 1.x and newer shader, reload the first 8 constants
4248 if(priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*) ps
->backend_priv
)->clamp_consts
)
4250 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*) ps
->backend_priv
)->clamp_consts
;
4251 This
->highest_dirty_ps_const
= max(This
->highest_dirty_ps_const
, 8);
4252 for(i
= 0; i
< 8; i
++)
4254 context
->pshader_const_dirty
[i
] = 1;
4256 /* Also takes care of loading local constants */
4257 shader_arb_load_constants(iface
, TRUE
, FALSE
);
4261 shader_arb_ps_local_constants(This
);
4264 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4265 if (compiled
->np2fixup_info
.super
.active
) This
->shader_backend
->shader_load_np2fixup_constants(iface
, usePS
, useVS
);
4266 } else if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
) && !priv
->use_arbfp_fixed_func
) {
4267 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
4268 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
4269 * replacement shader
4271 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4272 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4273 priv
->current_fprogram_id
= 0;
4277 struct arb_vs_compile_args compile_args
;
4278 struct arb_vs_compiled_shader
*compiled
;
4279 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) This
->stateBlock
->vertexShader
;
4281 TRACE("Using vertex shader %p\n", This
->stateBlock
->vertexShader
);
4282 find_arb_vs_compile_args(vs
, This
->stateBlock
, &compile_args
);
4283 compiled
= find_arb_vshader(vs
, &compile_args
);
4284 priv
->current_vprogram_id
= compiled
->prgId
;
4285 priv
->compiled_vprog
= compiled
;
4287 /* Bind the vertex program */
4288 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4289 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4291 /* Enable OpenGL vertex programs */
4292 glEnable(GL_VERTEX_PROGRAM_ARB
);
4293 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4294 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
4295 shader_arb_vs_local_constants(This
);
4297 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4298 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4300 if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT
)) {
4301 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4302 checkGLcall("glClampColorARB");
4304 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4307 } else if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
4308 priv
->current_vprogram_id
= 0;
4309 glDisable(GL_VERTEX_PROGRAM_ARB
);
4310 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4314 /* GL locking is done by the caller */
4315 static void shader_arb_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
) {
4316 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4317 struct shader_arb_priv
*priv
= This
->shader_priv
;
4318 GLuint
*blt_fprogram
= &priv
->depth_blt_fprogram_id
[tex_type
];
4319 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4321 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4322 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4323 glEnable(GL_VERTEX_PROGRAM_ARB
);
4325 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
);
4326 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4327 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4330 /* GL locking is done by the caller */
4331 static void shader_arb_deselect_depth_blt(IWineD3DDevice
*iface
) {
4332 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4333 struct shader_arb_priv
*priv
= This
->shader_priv
;
4334 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4336 if (priv
->current_vprogram_id
) {
4337 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4338 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4340 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
4342 glDisable(GL_VERTEX_PROGRAM_ARB
);
4343 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4346 if (priv
->current_fprogram_id
) {
4347 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4348 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4350 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
4351 } else if(!priv
->use_arbfp_fixed_func
) {
4352 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4353 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4357 static void shader_arb_destroy(IWineD3DBaseShader
*iface
) {
4358 IWineD3DBaseShaderImpl
*baseShader
= (IWineD3DBaseShaderImpl
*) iface
;
4359 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)baseShader
->baseShader
.device
;
4360 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4362 if (shader_is_pshader_version(baseShader
->baseShader
.reg_maps
.shader_version
.type
))
4364 IWineD3DPixelShaderImpl
*This
= (IWineD3DPixelShaderImpl
*) iface
;
4365 struct arb_pshader_private
*shader_data
= This
->backend_priv
;
4368 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4371 if(shader_data
->num_gl_shaders
) ActivateContext(device
, NULL
, CTXUSAGE_RESOURCELOAD
);
4373 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4374 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4375 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4378 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4379 HeapFree(GetProcessHeap(), 0, shader_data
);
4380 This
->backend_priv
= NULL
;
4382 IWineD3DVertexShaderImpl
*This
= (IWineD3DVertexShaderImpl
*) iface
;
4383 struct arb_vshader_private
*shader_data
= This
->backend_priv
;
4386 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4389 if(shader_data
->num_gl_shaders
) ActivateContext(device
, NULL
, CTXUSAGE_RESOURCELOAD
);
4391 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4392 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4393 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4396 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4397 HeapFree(GetProcessHeap(), 0, shader_data
);
4398 This
->backend_priv
= NULL
;
4402 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4404 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4405 return compare_sig(key
, e
->sig
);
4408 struct wine_rb_functions sig_tree_functions
=
4416 static HRESULT
shader_arb_alloc(IWineD3DDevice
*iface
) {
4417 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4418 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4419 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4421 ERR("RB tree init failed\n");
4422 HeapFree(GetProcessHeap(), 0, priv
);
4423 return E_OUTOFMEMORY
;
4425 This
->shader_priv
= priv
;
4429 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4431 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4433 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4435 HeapFree(GetProcessHeap(), 0, (char *) sig
->sig
[i
].semantic_name
);
4437 HeapFree(GetProcessHeap(), 0, sig
->sig
);
4438 HeapFree(GetProcessHeap(), 0, sig
);
4441 /* Context activation is done by the caller. */
4442 static void shader_arb_free(IWineD3DDevice
*iface
) {
4443 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4444 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4445 struct shader_arb_priv
*priv
= This
->shader_priv
;
4449 if(priv
->depth_blt_vprogram_id
) {
4450 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
4452 for (i
= 0; i
< tex_type_count
; ++i
) {
4453 if (priv
->depth_blt_fprogram_id
[i
]) {
4454 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id
[i
]));
4459 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4460 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
4463 static BOOL
shader_arb_dirty_const(IWineD3DDevice
*iface
) {
4467 static void shader_arb_get_caps(WINED3DDEVTYPE devtype
, const struct wined3d_gl_info
*gl_info
,
4468 struct shader_caps
*pCaps
)
4470 /* We don't have an ARB fixed function pipeline yet, so let the none backend set its caps,
4471 * then overwrite the shader specific ones
4473 none_shader_backend
.shader_get_caps(devtype
, gl_info
, pCaps
);
4475 if(GL_SUPPORT(ARB_VERTEX_PROGRAM
)) {
4476 if(GL_SUPPORT(NV_VERTEX_PROGRAM3
))
4478 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4479 TRACE_(d3d_caps
)("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
4481 else if(GL_LIMITS(vshader_constantsF
) >= 256)
4483 /* Shader Model 2.0 requires at least 256 vertex shader constants */
4484 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4485 TRACE_(d3d_caps
)("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
4489 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(1,1);
4490 TRACE_(d3d_caps
)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
4492 pCaps
->MaxVertexShaderConst
= GL_LIMITS(vshader_constantsF
);
4495 if(GL_SUPPORT(ARB_FRAGMENT_PROGRAM
)) {
4496 if(GL_SUPPORT(NV_FRAGMENT_PROGRAM2
))
4498 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4499 TRACE_(d3d_caps
)("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
4501 else if(GL_LIMITS(pshader_constantsF
) >= 32)
4503 /* Shader Model 2.0 requires at least 32 pixel shader constants */
4504 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4505 TRACE_(d3d_caps
)("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
4509 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(1,4);
4510 TRACE_(d3d_caps
)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
4512 pCaps
->PixelShader1xMaxValue
= 8.0f
;
4513 pCaps
->MaxPixelShaderConst
= GL_LIMITS(pshader_constantsF
);
4516 pCaps
->VSClipping
= use_nv_clip(gl_info
);
4519 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
4521 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4523 TRACE("Checking support for color_fixup:\n");
4524 dump_color_fixup_desc(fixup
);
4527 /* We support everything except YUV conversions. */
4528 if (!is_yuv_fixup(fixup
))
4534 TRACE("[FAILED]\n");
4538 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
4540 char write_mask
[20], regstr
[50];
4541 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
4542 BOOL is_color
= FALSE
;
4543 const struct wined3d_shader_dst_param
*dst
;
4545 if (!ins
->dst_count
) return;
4549 if(shift
== 0) return; /* Saturate alone is handled by the instructions */
4551 shader_arb_get_write_mask(ins
, dst
, write_mask
);
4552 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
4554 /* Generate a line that does the output modifier computation
4555 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
4556 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
4558 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
4559 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
4562 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4564 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
4565 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
4566 /* WINED3DSIH_BEM */ pshader_hw_bem
,
4567 /* WINED3DSIH_BREAK */ shader_hw_break
,
4568 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
4569 /* WINED3DSIH_BREAKP */ NULL
,
4570 /* WINED3DSIH_CALL */ shader_hw_call
,
4571 /* WINED3DSIH_CALLNZ */ NULL
,
4572 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
4573 /* WINED3DSIH_CND */ pshader_hw_cnd
,
4574 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
4575 /* WINED3DSIH_DCL */ NULL
,
4576 /* WINED3DSIH_DEF */ NULL
,
4577 /* WINED3DSIH_DEFB */ NULL
,
4578 /* WINED3DSIH_DEFI */ NULL
,
4579 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
4580 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
4581 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
4582 /* WINED3DSIH_DST */ shader_hw_map2gl
,
4583 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
4584 /* WINED3DSIH_DSY */ shader_hw_dsy
,
4585 /* WINED3DSIH_ELSE */ shader_hw_else
,
4586 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
4587 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
4588 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
4589 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
4590 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
4591 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
4592 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
4593 /* WINED3DSIH_IFC */ shader_hw_ifc
,
4594 /* WINED3DSIH_LABEL */ shader_hw_label
,
4595 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
4596 /* WINED3DSIH_LOG */ shader_hw_log_pow
,
4597 /* WINED3DSIH_LOGP */ shader_hw_log_pow
,
4598 /* WINED3DSIH_LOOP */ shader_hw_loop
,
4599 /* WINED3DSIH_LRP */ shader_hw_lrp
,
4600 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
4601 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
4602 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
4603 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
4604 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
4605 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
4606 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
4607 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
4608 /* WINED3DSIH_MOV */ shader_hw_mov
,
4609 /* WINED3DSIH_MOVA */ shader_hw_mov
,
4610 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
4611 /* WINED3DSIH_NOP */ shader_hw_nop
,
4612 /* WINED3DSIH_NRM */ shader_hw_nrm
,
4613 /* WINED3DSIH_PHASE */ NULL
,
4614 /* WINED3DSIH_POW */ shader_hw_log_pow
,
4615 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
4616 /* WINED3DSIH_REP */ shader_hw_rep
,
4617 /* WINED3DSIH_RET */ shader_hw_ret
,
4618 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
4619 /* WINED3DSIH_SETP */ NULL
,
4620 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
4621 /* WINED3DSIH_SGN */ shader_hw_sgn
,
4622 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
4623 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
4624 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
4625 /* WINED3DSIH_TEX */ pshader_hw_tex
,
4626 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
4627 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
4628 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
4629 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
4630 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
4631 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
4632 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
4633 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
4634 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
4635 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
4636 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
4637 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
4638 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
4639 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
4640 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
4641 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
4642 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
4643 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
4644 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
4645 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
4646 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
4649 static inline BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
, IWineD3DBaseShaderImpl
*This
, DWORD idx
)
4651 BOOL vshader
= shader_is_vshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
4653 WORD flag
= (1 << idx
);
4654 const local_constant
*constant
;
4655 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
4657 if(This
->baseShader
.reg_maps
.local_bool_consts
& flag
)
4659 /* What good is a if(bool) with a hardcoded local constant? I don't know, but handle it */
4660 LIST_FOR_EACH_ENTRY(constant
, &This
->baseShader
.constantsB
, local_constant
, entry
)
4662 if (constant
->idx
== idx
)
4664 return constant
->value
[0];
4667 ERR("Local constant not found\n");
4672 if(vshader
) bools
= priv
->cur_vs_args
->boolclip
.bools
;
4673 else bools
= priv
->cur_ps_args
->bools
;
4674 return bools
& flag
;
4678 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
4679 IWineD3DBaseShaderImpl
*This
, UINT idx
, struct loop_control
*loop_control
)
4681 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
4683 /* Integer constants can either be a local constant, or they can be stored in the shader
4684 * type specific compile args. */
4685 if (This
->baseShader
.reg_maps
.local_int_consts
& (1 << idx
))
4687 const local_constant
*constant
;
4689 LIST_FOR_EACH_ENTRY(constant
, &This
->baseShader
.constantsI
, local_constant
, entry
)
4691 if (constant
->idx
== idx
)
4693 loop_control
->count
= constant
->value
[0];
4694 loop_control
->start
= constant
->value
[1];
4695 /* Step is signed. */
4696 loop_control
->step
= (int)constant
->value
[2];
4700 /* If this happens the flag was set incorrectly */
4701 ERR("Local constant not found\n");
4702 loop_control
->count
= 0;
4703 loop_control
->start
= 0;
4704 loop_control
->step
= 0;
4708 switch (This
->baseShader
.reg_maps
.shader_version
.type
)
4710 case WINED3D_SHADER_TYPE_VERTEX
:
4711 /* Count and aL start value are unsigned */
4712 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
4713 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
4714 /* Step is signed. */
4715 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
4718 case WINED3D_SHADER_TYPE_PIXEL
:
4719 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
4720 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
4721 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
4725 FIXME("Unhandled shader type %#x.\n", This
->baseShader
.reg_maps
.shader_version
.type
);
4730 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
4733 struct wined3d_shader_dst_param
*dst_param
= NULL
;
4734 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
4735 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
4738 ERR("Out of memory\n");
4743 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
4744 if(!dst_param
) goto free
;
4745 *dst_param
= *ins
->dst
;
4746 if(ins
->dst
->reg
.rel_addr
)
4748 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
->reg
.rel_addr
));
4749 if(!rel_addr
) goto free
;
4750 *rel_addr
= *ins
->dst
->reg
.rel_addr
;
4751 dst_param
->reg
.rel_addr
= rel_addr
;
4753 rec
->ins
.dst
= dst_param
;
4755 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
4756 if(!src_param
) goto free
;
4757 for(i
= 0; i
< ins
->src_count
; i
++)
4759 src_param
[i
] = ins
->src
[i
];
4760 if(ins
->src
[i
].reg
.rel_addr
)
4762 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
4763 if(!rel_addr
) goto free
;
4764 *rel_addr
= *ins
->src
[i
].reg
.rel_addr
;
4765 src_param
[i
].reg
.rel_addr
= rel_addr
;
4768 rec
->ins
.src
= src_param
;
4769 list_add_tail(list
, &rec
->entry
);
4773 ERR("Out of memory\n");
4776 HeapFree(GetProcessHeap(), 0, (void *) dst_param
->reg
.rel_addr
);
4777 HeapFree(GetProcessHeap(), 0, dst_param
);
4781 for(i
= 0; i
< ins
->src_count
; i
++)
4783 HeapFree(GetProcessHeap(), 0, (void *) src_param
[i
].reg
.rel_addr
);
4785 HeapFree(GetProcessHeap(), 0, src_param
);
4787 HeapFree(GetProcessHeap(), 0, rec
);
4790 static void free_recorded_instruction(struct list
*list
)
4792 struct recorded_instruction
*rec_ins
, *entry2
;
4795 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
4797 list_remove(&rec_ins
->entry
);
4798 if(rec_ins
->ins
.dst
)
4800 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.dst
->reg
.rel_addr
);
4801 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.dst
);
4803 if(rec_ins
->ins
.src
)
4805 for(i
= 0; i
< rec_ins
->ins
.src_count
; i
++)
4807 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.src
[i
].reg
.rel_addr
);
4809 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.src
);
4811 HeapFree(GetProcessHeap(), 0, rec_ins
);
4815 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
4816 SHADER_HANDLER hw_fct
;
4817 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
4818 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
4819 struct control_frame
*control_frame
;
4820 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
4823 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
4825 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
4826 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
4828 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
4829 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
4831 if(priv
->target_version
>= NV2
)
4833 control_frame
->loop_no
= priv
->num_loops
++;
4838 /* Don't bother recording when we're in a not used if branch */
4844 if(!priv
->recording
)
4846 list_init(&priv
->record
);
4847 priv
->recording
= TRUE
;
4848 control_frame
->outer_loop
= TRUE
;
4849 get_loop_control_const(ins
, This
, ins
->src
[0].reg
.idx
, &control_frame
->loop_control
);
4850 return; /* Instruction is handled */
4852 /* Record this loop in the outer loop's recording */
4855 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
4857 if(priv
->target_version
>= NV2
)
4859 /* Nothing to do. The control frame is popped after the HW instr handler */
4863 struct list
*e
= list_head(&priv
->control_frames
);
4864 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
4865 list_remove(&control_frame
->entry
);
4867 if(control_frame
->outer_loop
)
4869 int iteration
, aL
= 0;
4872 /* Turn off recording before playback */
4873 priv
->recording
= FALSE
;
4875 /* Move the recorded instructions to a separate list and get them out of the private data
4876 * structure. If there are nested loops, the shader_arb_handle_instruction below will
4877 * be recorded again, thus priv->record might be overwritten
4880 list_move_tail(©
, &priv
->record
);
4881 list_init(&priv
->record
);
4883 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
4885 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
4886 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
4887 control_frame
->loop_control
.step
);
4888 aL
= control_frame
->loop_control
.start
;
4892 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
4895 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
4897 struct recorded_instruction
*rec_ins
;
4898 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
4901 shader_addline(buffer
, "#Iteration %d, aL=%d\n", iteration
, aL
);
4905 shader_addline(buffer
, "#Iteration %d\n", iteration
);
4908 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
4910 shader_arb_handle_instruction(&rec_ins
->ins
);
4913 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
4915 aL
+= control_frame
->loop_control
.step
;
4918 shader_addline(buffer
, "#end loop/rep\n");
4920 free_recorded_instruction(©
);
4921 HeapFree(GetProcessHeap(), 0, control_frame
);
4922 return; /* Instruction is handled */
4926 /* This is a nested loop. Proceed to the normal recording function */
4927 HeapFree(GetProcessHeap(), 0, control_frame
);
4934 record_instruction(&priv
->record
, ins
);
4939 if(ins
->handler_idx
== WINED3DSIH_IF
)
4941 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
4942 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
4943 control_frame
->type
= IF
;
4945 bool_const
= get_bool_const(ins
, This
, ins
->src
[0].reg
.idx
);
4946 if(ins
->src
[0].modifiers
== WINED3DSPSM_NOT
) bool_const
= !bool_const
;
4947 if(!priv
->muted
&& bool_const
== FALSE
)
4949 shader_addline(buffer
, "#if(FALSE){\n");
4951 control_frame
->muting
= TRUE
;
4953 else shader_addline(buffer
, "#if(TRUE) {\n");
4955 return; /* Instruction is handled */
4957 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
4959 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
4960 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
4961 control_frame
->type
= IFC
;
4962 control_frame
->ifc_no
= priv
->num_ifcs
++;
4963 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
4965 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
4967 struct list
*e
= list_head(&priv
->control_frames
);
4968 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
4970 if(control_frame
->type
== IF
)
4972 shader_addline(buffer
, "#} else {\n");
4973 if(!priv
->muted
&& !control_frame
->muting
)
4976 control_frame
->muting
= TRUE
;
4978 else if(control_frame
->muting
) priv
->muted
= FALSE
;
4979 return; /* Instruction is handled. */
4981 /* In case of an ifc, generate a HW shader instruction */
4983 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
4985 struct list
*e
= list_head(&priv
->control_frames
);
4986 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
4988 if(control_frame
->type
== IF
)
4990 shader_addline(buffer
, "#} endif\n");
4991 if(control_frame
->muting
) priv
->muted
= FALSE
;
4992 list_remove(&control_frame
->entry
);
4993 HeapFree(GetProcessHeap(), 0, control_frame
);
4994 return; /* Instruction is handled */
4998 if(priv
->muted
) return;
5000 /* Select handler */
5001 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5003 /* Unhandled opcode */
5006 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5011 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5013 struct list
*e
= list_head(&priv
->control_frames
);
5014 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5015 list_remove(&control_frame
->entry
);
5016 HeapFree(GetProcessHeap(), 0, control_frame
);
5019 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5021 /* Non-ifc ENDIFs don't reach that place because of the return in the if block above */
5022 struct list
*e
= list_head(&priv
->control_frames
);
5023 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5024 list_remove(&control_frame
->entry
);
5025 HeapFree(GetProcessHeap(), 0, control_frame
);
5029 shader_arb_add_instruction_modifiers(ins
);
5032 const shader_backend_t arb_program_shader_backend
= {
5033 shader_arb_handle_instruction
,
5035 shader_arb_select_depth_blt
,
5036 shader_arb_deselect_depth_blt
,
5037 shader_arb_update_float_vertex_constants
,
5038 shader_arb_update_float_pixel_constants
,
5039 shader_arb_load_constants
,
5040 shader_arb_load_np2fixup_constants
,
5044 shader_arb_dirty_const
,
5045 shader_arb_get_caps
,
5046 shader_arb_color_fixup_supported
,
5049 /* ARB_fragment_program fixed function pipeline replacement definitions */
5050 #define ARB_FFP_CONST_TFACTOR 0
5051 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
5052 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5053 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5054 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5056 struct arbfp_ffp_desc
5058 struct ffp_frag_desc parent
;
5060 unsigned int num_textures_used
;
5063 /* Context activation is done by the caller. */
5064 static void arbfp_enable(IWineD3DDevice
*iface
, BOOL enable
) {
5067 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5068 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5070 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5071 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5076 static HRESULT
arbfp_alloc(IWineD3DDevice
*iface
) {
5077 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
5078 struct shader_arb_priv
*priv
;
5079 /* Share private data between the shader backend and the pipeline replacement, if both
5080 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
5081 * if no pixel shader is bound or not
5083 if(This
->shader_backend
== &arb_program_shader_backend
) {
5084 This
->fragment_priv
= This
->shader_priv
;
5086 This
->fragment_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
5087 if(!This
->fragment_priv
) return E_OUTOFMEMORY
;
5089 priv
= This
->fragment_priv
;
5090 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5092 ERR("Failed to initialize rbtree.\n");
5093 HeapFree(GetProcessHeap(), 0, This
->fragment_priv
);
5094 return E_OUTOFMEMORY
;
5096 priv
->use_arbfp_fixed_func
= TRUE
;
5100 /* Context activation is done by the caller. */
5101 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5103 const struct wined3d_gl_info
*gl_info
= context
;
5104 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5107 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5108 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5109 HeapFree(GetProcessHeap(), 0, entry_arb
);
5113 /* Context activation is done by the caller. */
5114 static void arbfp_free(IWineD3DDevice
*iface
) {
5115 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
5116 struct shader_arb_priv
*priv
= This
->fragment_priv
;
5118 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &This
->adapter
->gl_info
);
5119 priv
->use_arbfp_fixed_func
= FALSE
;
5121 if(This
->shader_backend
!= &arb_program_shader_backend
) {
5122 HeapFree(GetProcessHeap(), 0, This
->fragment_priv
);
5126 static void arbfp_get_caps(WINED3DDEVTYPE devtype
, const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5128 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5129 WINED3DTEXOPCAPS_SELECTARG1
|
5130 WINED3DTEXOPCAPS_SELECTARG2
|
5131 WINED3DTEXOPCAPS_MODULATE4X
|
5132 WINED3DTEXOPCAPS_MODULATE2X
|
5133 WINED3DTEXOPCAPS_MODULATE
|
5134 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5135 WINED3DTEXOPCAPS_ADDSIGNED
|
5136 WINED3DTEXOPCAPS_ADD
|
5137 WINED3DTEXOPCAPS_SUBTRACT
|
5138 WINED3DTEXOPCAPS_ADDSMOOTH
|
5139 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5140 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5141 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5142 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5143 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5144 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5145 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5146 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5147 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5148 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5149 WINED3DTEXOPCAPS_MULTIPLYADD
|
5150 WINED3DTEXOPCAPS_LERP
|
5151 WINED3DTEXOPCAPS_BUMPENVMAP
|
5152 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5154 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5156 caps
->MaxTextureBlendStages
= 8;
5157 caps
->MaxSimultaneousTextures
= min(GL_LIMITS(fragment_samplers
), 8);
5159 caps
->PrimitiveMiscCaps
|= WINED3DPMISCCAPS_TSSARGTEMP
;
5161 #undef GLINFO_LOCATION
5163 #define GLINFO_LOCATION stateblock->wineD3DDevice->adapter->gl_info
5164 static void state_texfactor_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5167 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
5169 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5170 * application provided constants
5172 if(device
->shader_backend
== &arb_program_shader_backend
) {
5173 if (use_ps(stateblock
)) return;
5175 device
= stateblock
->wineD3DDevice
;
5176 context
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5177 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5180 D3DCOLORTOGLFLOAT4(stateblock
->renderState
[WINED3DRS_TEXTUREFACTOR
], col
);
5181 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5182 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5186 static void state_arb_specularenable(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5189 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
5191 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5192 * application provided constants
5194 if(device
->shader_backend
== &arb_program_shader_backend
) {
5195 if (use_ps(stateblock
)) return;
5197 device
= stateblock
->wineD3DDevice
;
5198 context
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5199 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5202 if(stateblock
->renderState
[WINED3DRS_SPECULARENABLE
]) {
5203 /* The specular color has no alpha */
5204 col
[0] = 1.0f
; col
[1] = 1.0f
;
5205 col
[2] = 1.0f
; col
[3] = 0.0f
;
5207 col
[0] = 0.0f
; col
[1] = 0.0f
;
5208 col
[2] = 0.0f
; col
[3] = 0.0f
;
5210 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5211 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5214 static void set_bumpmat_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5216 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5217 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
5220 if (use_ps(stateblock
))
5223 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.bumpmat
[stage
]) {
5224 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
5227 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
5228 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
5232 if(device
->shader_backend
== &arb_program_shader_backend
) {
5233 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
5236 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
5237 context
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5238 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5241 mat
[0][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT00
]);
5242 mat
[0][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT01
]);
5243 mat
[1][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT10
]);
5244 mat
[1][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT11
]);
5246 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5247 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5250 static void tex_bumpenvlum_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5252 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5253 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
5256 if (use_ps(stateblock
))
5259 ((IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
)->baseShader
.reg_maps
.luminanceparams
[stage
]) {
5260 /* The pixel shader has to know the luminance offset. Do a constants update if it
5261 * isn't scheduled anyway
5263 if(!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
)) {
5264 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
5268 if(device
->shader_backend
== &arb_program_shader_backend
) {
5269 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
5272 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
5273 context
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5274 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5277 param
[0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
]);
5278 param
[1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
]);
5282 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5283 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5286 static const char *get_argreg(struct wined3d_shader_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
5290 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
5292 switch(arg
& WINED3DTA_SELECTMASK
) {
5293 case WINED3DTA_DIFFUSE
:
5294 ret
= "fragment.color.primary"; break;
5296 case WINED3DTA_CURRENT
:
5297 if(stage
== 0) ret
= "fragment.color.primary";
5301 case WINED3DTA_TEXTURE
:
5303 case 0: ret
= "tex0"; break;
5304 case 1: ret
= "tex1"; break;
5305 case 2: ret
= "tex2"; break;
5306 case 3: ret
= "tex3"; break;
5307 case 4: ret
= "tex4"; break;
5308 case 5: ret
= "tex5"; break;
5309 case 6: ret
= "tex6"; break;
5310 case 7: ret
= "tex7"; break;
5311 default: ret
= "unknown texture";
5315 case WINED3DTA_TFACTOR
:
5316 ret
= "tfactor"; break;
5318 case WINED3DTA_SPECULAR
:
5319 ret
= "fragment.color.secondary"; break;
5321 case WINED3DTA_TEMP
:
5322 ret
= "tempreg"; break;
5324 case WINED3DTA_CONSTANT
:
5325 FIXME("Implement perstage constants\n");
5327 case 0: ret
= "const0"; break;
5328 case 1: ret
= "const1"; break;
5329 case 2: ret
= "const2"; break;
5330 case 3: ret
= "const3"; break;
5331 case 4: ret
= "const4"; break;
5332 case 5: ret
= "const5"; break;
5333 case 6: ret
= "const6"; break;
5334 case 7: ret
= "const7"; break;
5335 default: ret
= "unknown constant";
5343 if(arg
& WINED3DTA_COMPLEMENT
) {
5344 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
5345 if(argnum
== 0) ret
= "arg0";
5346 if(argnum
== 1) ret
= "arg1";
5347 if(argnum
== 2) ret
= "arg2";
5349 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
5350 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
5351 if(argnum
== 0) ret
= "arg0";
5352 if(argnum
== 1) ret
= "arg1";
5353 if(argnum
== 2) ret
= "arg2";
5358 static void gen_ffp_instr(struct wined3d_shader_buffer
*buffer
, unsigned int stage
, BOOL color
,
5359 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
5361 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
5362 unsigned int mul
= 1;
5363 BOOL mul_final_dest
= FALSE
;
5365 if(color
&& alpha
) dstmask
= "";
5366 else if(color
) dstmask
= ".xyz";
5367 else dstmask
= ".w";
5369 if(dst
== tempreg
) dstreg
= "tempreg";
5370 else dstreg
= "ret";
5372 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
5373 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
5374 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
5377 case WINED3DTOP_DISABLE
:
5378 if(stage
== 0) shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
5381 case WINED3DTOP_SELECTARG2
:
5383 case WINED3DTOP_SELECTARG1
:
5384 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
5387 case WINED3DTOP_MODULATE4X
:
5389 case WINED3DTOP_MODULATE2X
:
5391 if(strcmp(dstreg
, "result.color") == 0) {
5393 mul_final_dest
= TRUE
;
5395 case WINED3DTOP_MODULATE
:
5396 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5399 case WINED3DTOP_ADDSIGNED2X
:
5401 if(strcmp(dstreg
, "result.color") == 0) {
5403 mul_final_dest
= TRUE
;
5405 case WINED3DTOP_ADDSIGNED
:
5406 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
5408 case WINED3DTOP_ADD
:
5409 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5412 case WINED3DTOP_SUBTRACT
:
5413 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5416 case WINED3DTOP_ADDSMOOTH
:
5417 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
5418 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5421 case WINED3DTOP_BLENDCURRENTALPHA
:
5422 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
5423 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5425 case WINED3DTOP_BLENDFACTORALPHA
:
5426 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
5427 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5429 case WINED3DTOP_BLENDTEXTUREALPHA
:
5430 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
5431 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5433 case WINED3DTOP_BLENDDIFFUSEALPHA
:
5434 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
5435 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5438 case WINED3DTOP_BLENDTEXTUREALPHAPM
:
5439 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
5440 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
5441 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5444 /* D3DTOP_PREMODULATE ???? */
5446 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR
:
5447 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
5448 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5450 case WINED3DTOP_MODULATEALPHA_ADDCOLOR
:
5451 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
5453 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA
:
5454 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
5455 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
5457 case WINED3DTOP_MODULATECOLOR_ADDALPHA
:
5458 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
5461 case WINED3DTOP_DOTPRODUCT3
:
5463 if(strcmp(dstreg
, "result.color") == 0) {
5465 mul_final_dest
= TRUE
;
5467 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
5468 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
5469 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
5472 case WINED3DTOP_MULTIPLYADD
:
5473 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
5476 case WINED3DTOP_LERP
:
5477 /* The msdn is not quite right here */
5478 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5481 case WINED3DTOP_BUMPENVMAP
:
5482 case WINED3DTOP_BUMPENVMAPLUMINANCE
:
5483 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
5487 FIXME("Unhandled texture op %08x\n", op
);
5491 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
5492 } else if(mul
== 4) {
5493 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
5497 /* The stateblock is passed for GLINFO_LOCATION */
5498 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, IWineD3DStateBlockImpl
*stateblock
)
5501 struct wined3d_shader_buffer buffer
;
5502 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5503 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5504 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5505 const char *textype
;
5506 const char *instr
, *sat
;
5507 char colorcor_dst
[8];
5509 DWORD arg0
, arg1
, arg2
;
5510 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
5512 const char *final_combiner_src
= "ret";
5514 /* Find out which textures are read */
5515 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
5516 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) break;
5517 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
5518 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
5519 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
5520 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5521 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5522 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5524 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHA
) tex_read
[stage
] = TRUE
;
5525 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHAPM
) tex_read
[stage
] = TRUE
;
5526 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
) {
5527 bump_used
[stage
] = TRUE
;
5528 tex_read
[stage
] = TRUE
;
5530 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
5531 bump_used
[stage
] = TRUE
;
5532 tex_read
[stage
] = TRUE
;
5533 luminance_used
[stage
] = TRUE
;
5534 } else if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDFACTORALPHA
) {
5535 tfactor_used
= TRUE
;
5538 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
5539 tfactor_used
= TRUE
;
5542 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
5543 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
5544 tempreg_used
= TRUE
;
5547 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) continue;
5548 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
5549 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
5550 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
5551 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5552 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5553 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5555 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
5556 tempreg_used
= TRUE
;
5558 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
5559 tfactor_used
= TRUE
;
5564 if (!shader_buffer_init(&buffer
))
5566 ERR("Failed to initialize shader buffer.\n");
5570 shader_addline(&buffer
, "!!ARBfp1.0\n");
5572 switch(settings
->fog
) {
5573 case FOG_OFF
: break;
5574 case FOG_LINEAR
: shader_addline(&buffer
, "OPTION ARB_fog_linear;\n"); break;
5575 case FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
5576 case FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
5577 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
5580 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
5581 shader_addline(&buffer
, "TEMP TMP;\n");
5582 shader_addline(&buffer
, "TEMP ret;\n");
5583 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
5584 shader_addline(&buffer
, "TEMP arg0;\n");
5585 shader_addline(&buffer
, "TEMP arg1;\n");
5586 shader_addline(&buffer
, "TEMP arg2;\n");
5587 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
5588 if(!tex_read
[stage
]) continue;
5589 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
5590 if(!bump_used
[stage
]) continue;
5591 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
5592 if(!luminance_used
[stage
]) continue;
5593 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
5596 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
5598 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
5600 if(settings
->sRGB_write
) {
5601 shader_addline(&buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
5602 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
5603 shader_addline(&buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
5604 srgb_sub_high
, 0.0, 0.0, 0.0);
5607 if(ffp_clip_emul(stateblock
) && settings
->emul_clipplanes
) shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
5609 /* Generate texture sampling instructions) */
5610 for(stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3DTOP_DISABLE
; stage
++) {
5611 if(!tex_read
[stage
]) continue;
5613 switch(settings
->op
[stage
].tex_type
) {
5614 case tex_1d
: textype
= "1D"; break;
5615 case tex_2d
: textype
= "2D"; break;
5616 case tex_3d
: textype
= "3D"; break;
5617 case tex_cube
: textype
= "CUBE"; break;
5618 case tex_rect
: textype
= "RECT"; break;
5619 default: textype
= "unexpected_textype"; break;
5622 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
||
5623 settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
5629 if(settings
->op
[stage
].projected
== proj_none
) {
5631 } else if(settings
->op
[stage
].projected
== proj_count4
||
5632 settings
->op
[stage
].projected
== proj_count3
) {
5635 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
5640 (settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAP
||
5641 settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
)) {
5642 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
5643 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
5644 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
5645 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
5647 /* with projective textures, texbem only divides the static texture coord, not the displacement,
5648 * so multiply the displacement with the dividing parameter before passing it to TXP
5650 if (settings
->op
[stage
].projected
!= proj_none
) {
5651 if(settings
->op
[stage
].projected
== proj_count4
) {
5652 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
5653 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
5655 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
5656 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
5659 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
5662 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
5663 instr
, sat
, stage
, stage
, textype
);
5664 if(settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
5665 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
5666 stage
- 1, stage
- 1, stage
- 1);
5667 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
5669 } else if(settings
->op
[stage
].projected
== proj_count3
) {
5670 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
5671 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
5672 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
5673 instr
, sat
, stage
, stage
, textype
);
5675 shader_addline(&buffer
, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
5676 instr
, sat
, stage
, stage
, stage
, textype
);
5679 sprintf(colorcor_dst
, "tex%u", stage
);
5680 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
5681 settings
->op
[stage
].color_fixup
);
5684 /* Generate the main shader */
5685 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
5686 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) {
5688 final_combiner_src
= "fragment.color.primary";
5693 if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
5694 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
5695 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
5696 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
5697 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
5698 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
5699 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
5700 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
5701 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
5702 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
5703 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
5704 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
5706 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
&&
5707 settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
&&
5708 settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
&&
5709 settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
5712 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) {
5713 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
5714 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
5715 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
5717 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
5719 } else if(op_equal
) {
5720 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
5721 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
5722 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
5724 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
5725 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
5726 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
5727 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
5728 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
5729 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
5733 if(settings
->sRGB_write
) {
5734 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
5735 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
5737 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
5741 shader_addline(&buffer
, "END\n");
5743 /* Generate the shader */
5744 GL_EXTCALL(glGenProgramsARB(1, &ret
));
5745 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
5746 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
5748 if (glGetError() == GL_INVALID_OPERATION
) {
5750 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
5751 FIXME("Fragment program error at position %d: %s\n", pos
,
5752 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
5754 shader_buffer_free(&buffer
);
5758 static void fragment_prog_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5760 IWineD3DDeviceImpl
*device
= stateblock
->wineD3DDevice
;
5761 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5762 BOOL use_pshader
= use_ps(stateblock
);
5763 BOOL use_vshader
= use_vs(stateblock
);
5764 struct ffp_frag_settings settings
;
5765 const struct arbfp_ffp_desc
*desc
;
5768 TRACE("state %#x, stateblock %p, context %p\n", state
, stateblock
, context
);
5770 if(isStateDirty(context
, STATE_RENDER(WINED3DRS_FOGENABLE
))) {
5771 if(!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
5772 /* Reload fixed function constants since they collide with the pixel shader constants */
5773 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
5774 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
5776 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
5777 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
5778 } else if(use_pshader
&& !isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
5779 device
->shader_backend
->shader_select((IWineD3DDevice
*)stateblock
->wineD3DDevice
, use_pshader
, use_vshader
);
5785 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
5786 gen_ffp_frag_op(stateblock
, &settings
, FALSE
);
5787 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
5789 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
5792 ERR("Out of memory\n");
5795 new_desc
->num_textures_used
= 0;
5796 for(i
= 0; i
< GL_LIMITS(texture_stages
); i
++) {
5797 if(settings
.op
[i
].cop
== WINED3DTOP_DISABLE
) break;
5798 new_desc
->num_textures_used
= i
;
5801 memcpy(&new_desc
->parent
.settings
, &settings
, sizeof(settings
));
5802 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, stateblock
);
5803 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
5804 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
5808 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
5809 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
5812 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
5813 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
5814 priv
->current_fprogram_id
= desc
->shader
;
5816 if(device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
5817 /* Reload fixed function constants since they collide with the pixel shader constants */
5818 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
5819 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
5821 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
5822 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
5824 context
->last_was_pshader
= FALSE
;
5826 context
->last_was_pshader
= TRUE
;
5829 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
5830 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
5831 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
5832 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
5833 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
5835 * Don't call shader_select if the vertex shader is dirty, because it will be called later on by the vertex
5838 if(!isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
5839 device
->shader_backend
->shader_select((IWineD3DDevice
*)stateblock
->wineD3DDevice
, use_pshader
, use_vshader
);
5841 if (!isStateDirty(context
, STATE_VERTEXSHADERCONSTANT
) && (use_vshader
|| use_pshader
)) {
5842 device
->StateTable
[STATE_VERTEXSHADERCONSTANT
].apply(STATE_VERTEXSHADERCONSTANT
, stateblock
, context
);
5846 device
->StateTable
[STATE_PIXELSHADERCONSTANT
].apply(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
5850 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
5851 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
5852 * state table, so we need to handle that with a forwarding function. The other invisible side effect
5853 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
5854 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
5856 static void state_arbfp_fog(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5858 enum fogsource new_source
;
5860 TRACE("state %#x, stateblock %p, context %p\n", state
, stateblock
, context
);
5862 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
5863 fragment_prog_arbfp(state
, stateblock
, context
);
5866 if(!stateblock
->renderState
[WINED3DRS_FOGENABLE
]) return;
5868 if(stateblock
->renderState
[WINED3DRS_FOGTABLEMODE
] == WINED3DFOG_NONE
) {
5869 if(use_vs(stateblock
)) {
5870 new_source
= FOGSOURCE_VS
;
5872 if(stateblock
->renderState
[WINED3DRS_FOGVERTEXMODE
] == WINED3DFOG_NONE
|| context
->last_was_rhw
) {
5873 new_source
= FOGSOURCE_COORD
;
5875 new_source
= FOGSOURCE_FFP
;
5879 new_source
= FOGSOURCE_FFP
;
5881 if(new_source
!= context
->fog_source
) {
5882 context
->fog_source
= new_source
;
5883 state_fogstartend(STATE_RENDER(WINED3DRS_FOGSTART
), stateblock
, context
);
5887 static void textransform(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5889 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
5890 fragment_prog_arbfp(state
, stateblock
, context
);
5894 #undef GLINFO_LOCATION
5896 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] = {
5897 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
5898 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5899 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5900 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5901 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5902 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5903 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5904 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5905 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5906 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5907 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5908 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5909 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5910 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5911 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5912 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5913 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5914 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5915 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5916 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5917 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5918 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5919 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5920 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5921 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5922 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5923 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5924 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5925 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5926 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5927 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5928 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5929 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5930 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5931 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5932 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5933 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5934 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5935 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5936 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5937 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5938 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5939 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5940 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5941 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5942 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5943 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5944 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5945 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5946 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5947 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5948 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5949 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5950 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5951 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5952 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5953 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5954 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5955 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5956 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5957 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5958 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5959 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5960 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5961 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5962 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5963 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5964 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5965 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5966 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5967 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5968 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5969 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5970 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5971 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5972 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5973 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5974 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5975 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5976 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5977 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5978 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5979 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5980 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5981 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5982 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5983 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5984 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5985 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5986 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5987 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
5988 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5989 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5990 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5991 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5992 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5993 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5994 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5995 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5996 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
5997 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5998 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
5999 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6000 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6001 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6002 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6003 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6004 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6005 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6006 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6007 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6008 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6009 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6010 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6011 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6012 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6013 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6014 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6015 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6016 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6017 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6018 {STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6019 {STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6020 {STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6021 {STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6022 {STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6023 {STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6024 {STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6025 {STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6026 {STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6027 {STATE_RENDER(WINED3DRS_FOGENABLE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6028 {STATE_RENDER(WINED3DRS_FOGTABLEMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6029 {STATE_RENDER(WINED3DRS_FOGVERTEXMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6030 {STATE_RENDER(WINED3DRS_FOGSTART
), { STATE_RENDER(WINED3DRS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6031 {STATE_RENDER(WINED3DRS_FOGEND
), { STATE_RENDER(WINED3DRS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6032 {STATE_RENDER(WINED3DRS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6033 {STATE_RENDER(WINED3DRS_FOGCOLOR
), { STATE_RENDER(WINED3DRS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6034 {STATE_RENDER(WINED3DRS_FOGDENSITY
), { STATE_RENDER(WINED3DRS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6035 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6036 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6037 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6038 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6039 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6040 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6041 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6042 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6043 {STATE_RENDER(WINED3DRS_SPECULARENABLE
), { STATE_RENDER(WINED3DRS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6044 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6047 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6052 shader_arb_color_fixup_supported
,
6053 arbfp_fragmentstate_template
,
6054 TRUE
/* We can disable projected textures */
6057 #define GLINFO_LOCATION device->adapter->gl_info
6059 struct arbfp_blit_priv
{
6060 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
6061 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
6062 GLenum yv12_rect_shader
, yv12_2d_shader
;
6065 static HRESULT
arbfp_blit_alloc(IWineD3DDevice
*iface
) {
6066 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
6067 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
6068 if(!device
->blit_priv
) {
6069 ERR("Out of memory\n");
6070 return E_OUTOFMEMORY
;
6075 /* Context activation is done by the caller. */
6076 static void arbfp_blit_free(IWineD3DDevice
*iface
) {
6077 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
6078 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6081 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
6082 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
6083 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
6084 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
6085 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
6086 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
6087 checkGLcall("Delete yuv programs");
6091 static BOOL
gen_planar_yuv_read(struct wined3d_shader_buffer
*buffer
, enum yuv_fixup yuv_fixup
,
6092 GLenum textype
, char *luminance
)
6095 const char *tex
, *texinstr
;
6097 if (yuv_fixup
== YUV_FIXUP_UYVY
) {
6105 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
6106 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
6108 /* This is more tricky than just replacing the texture type - we have to navigate
6109 * properly in the texture to find the correct chroma values
6111 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
6115 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6116 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6117 * filtering when we sample the texture.
6119 * These are the rules for reading the chroma:
6125 * So we have to get the sampling x position in non-normalized coordinates in integers
6127 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
6128 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6129 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6131 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6133 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6134 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6137 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6138 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6140 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6141 * even and odd pixels respectively
6143 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6144 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6146 /* Sample Pixel 1 */
6147 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6149 /* Put the value into either of the chroma values */
6150 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6151 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6152 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6153 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6155 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6156 * the pixel right to the current one. Otherwise, sample the left pixel.
6157 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6159 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6160 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6161 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6163 /* Put the value into the other chroma */
6164 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6165 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
6166 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6167 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
6169 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
6170 * the current one and lerp the two U and V values
6173 /* This gives the correctly filtered luminance value */
6174 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
6179 static BOOL
gen_yv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
, char *luminance
)
6184 case GL_TEXTURE_2D
: tex
= "2D"; break;
6185 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
6187 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
6191 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
6192 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
6193 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
6194 * pitch of the luminance plane, the packing into the gl texture is a bit
6195 * unfortunate. If the whole texture is interpreted as luminance data it looks
6196 * approximately like this:
6198 * +----------------------------------+----
6210 * +----------------+-----------------+----
6212 * | U even rows | U odd rows |
6214 * +----------------+------------------ -
6216 * | V even rows | V odd rows |
6218 * +----------------+-----------------+----
6222 * So it appears as if there are 4 chroma images, but in fact the odd rows
6223 * in the chroma images are in the same row as the even ones. So its is
6224 * kinda tricky to read
6226 * When reading from rectangle textures, keep in mind that the input y coordinates
6227 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
6229 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
6230 2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
);
6232 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6233 /* the chroma planes have only half the width */
6234 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
6236 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
6237 * the coordinate. Also read the right side of the image when reading odd lines
6239 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
6242 if(textype
== GL_TEXTURE_2D
) {
6244 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
6246 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
6248 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
6249 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
6251 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6252 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6253 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6254 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6255 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6257 /* clamp, keep the half pixel origin in mind */
6258 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
6259 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6260 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
6261 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6263 /* Read from [size - size+size/4] */
6264 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6265 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
6267 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6268 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6269 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6270 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6271 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
6272 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6274 /* Make sure to read exactly from the pixel center */
6275 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6276 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
6279 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
6280 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
6281 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6282 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
6283 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6285 /* Read the texture, put the result into the output register */
6286 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6287 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
6289 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
6290 * No need to clamp because we're just reusing the already clamped value from above
6292 if(textype
== GL_TEXTURE_2D
) {
6293 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
6295 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
6297 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6298 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
6300 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
6301 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
6302 * values due to filtering
6304 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6305 if(textype
== GL_TEXTURE_2D
) {
6306 /* Multiply the y coordinate by 2/3 and clamp it */
6307 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
6308 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
6309 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6310 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
6312 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
6313 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
6316 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
6317 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
6318 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
6325 /* Context activation is done by the caller. */
6326 static GLuint
gen_yuv_shader(IWineD3DDeviceImpl
*device
, enum yuv_fixup yuv_fixup
, GLenum textype
)
6329 struct wined3d_shader_buffer buffer
;
6330 char luminance_component
;
6331 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6334 if (!shader_buffer_init(&buffer
))
6336 ERR("Failed to initialize shader buffer.\n");
6341 GL_EXTCALL(glGenProgramsARB(1, &shader
));
6342 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
6343 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
6344 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
6347 shader_buffer_free(&buffer
);
6351 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
6352 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
6353 * two chroma(U and V) values. Each macropixel has two luminance values, one for
6354 * each single pixel it contains, and one U and one V value shared between both
6357 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
6358 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
6359 * take the format into account when generating the read swizzles
6361 * Reading the Y value is straightforward - just sample the texture. The hardware
6362 * takes care of filtering in the horizontal and vertical direction.
6364 * Reading the U and V values is harder. We have to avoid filtering horizontally,
6365 * because that would mix the U and V values of one pixel or two adjacent pixels.
6366 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
6367 * regardless of the filtering setting. Vertical filtering works automatically
6368 * though - the U and V values of two rows are mixed nicely.
6370 * Appart of avoiding filtering issues, the code has to know which value it just
6371 * read, and where it can find the other one. To determine this, it checks if
6372 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
6374 * Handling horizontal filtering of U and V values requires reading a 2nd pair
6375 * of pixels, extracting U and V and mixing them. This is not implemented yet.
6377 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
6378 * with width / 2. This way one read gives all 3 values, finding U and V is easy
6379 * in an unfiltered situation. Finding the luminance on the other hand requires
6380 * finding out if it is an odd or even pixel. The real drawback of this approach
6381 * is filtering. This would have to be emulated completely in the shader, reading
6382 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
6383 * vertically. Beyond that it would require adjustments to the texture handling
6384 * code to deal with the width scaling
6386 shader_addline(&buffer
, "!!ARBfp1.0\n");
6387 shader_addline(&buffer
, "TEMP luminance;\n");
6388 shader_addline(&buffer
, "TEMP temp;\n");
6389 shader_addline(&buffer
, "TEMP chroma;\n");
6390 shader_addline(&buffer
, "TEMP texcrd;\n");
6391 shader_addline(&buffer
, "TEMP texcrd2;\n");
6392 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
6393 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
6394 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
6398 case YUV_FIXUP_UYVY
:
6399 case YUV_FIXUP_YUY2
:
6400 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
6402 shader_buffer_free(&buffer
);
6407 case YUV_FIXUP_YV12
:
6408 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
6410 shader_buffer_free(&buffer
);
6416 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
6417 shader_buffer_free(&buffer
);
6421 /* Calculate the final result. Formula is taken from
6422 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
6423 * ranges from -0.5 to 0.5
6425 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
6427 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
6428 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
6429 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
6430 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
6431 shader_addline(&buffer
, "END\n");
6434 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(buffer
.buffer
), buffer
.buffer
));
6436 if (glGetError() == GL_INVALID_OPERATION
) {
6438 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6439 FIXME("Fragment program error at position %d: %s\n", pos
,
6440 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6442 shader_buffer_free(&buffer
);
6447 case YUV_FIXUP_YUY2
:
6448 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yuy2_rect_shader
= shader
;
6449 else priv
->yuy2_2d_shader
= shader
;
6452 case YUV_FIXUP_UYVY
:
6453 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->uyvy_rect_shader
= shader
;
6454 else priv
->uyvy_2d_shader
= shader
;
6457 case YUV_FIXUP_YV12
:
6458 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yv12_rect_shader
= shader
;
6459 else priv
->yv12_2d_shader
= shader
;
6466 /* Context activation is done by the caller. */
6467 static HRESULT
arbfp_blit_set(IWineD3DDevice
*iface
, const struct GlPixelFormatDesc
*format_desc
,
6468 GLenum textype
, UINT width
, UINT height
)
6471 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
6472 float size
[4] = {width
, height
, 1, 1};
6473 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6474 enum yuv_fixup yuv_fixup
;
6476 if (!is_yuv_fixup(format_desc
->color_fixup
))
6479 dump_color_fixup_desc(format_desc
->color_fixup
);
6480 /* Don't bother setting up a shader for unconverted formats */
6483 checkGLcall("glEnable(textype)");
6488 yuv_fixup
= get_yuv_fixup(format_desc
->color_fixup
);
6492 case YUV_FIXUP_YUY2
:
6493 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yuy2_rect_shader
: priv
->yuy2_2d_shader
;
6496 case YUV_FIXUP_UYVY
:
6497 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->uyvy_rect_shader
: priv
->uyvy_2d_shader
;
6500 case YUV_FIXUP_YV12
:
6501 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yv12_rect_shader
: priv
->yv12_2d_shader
;
6505 FIXME("Unsupported YUV fixup %#x, not setting a shader\n", yuv_fixup
);
6508 checkGLcall("glEnable(textype)");
6513 if (!shader
) shader
= gen_yuv_shader(device
, yuv_fixup
, textype
);
6516 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
6517 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
6518 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
6519 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
6520 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
6521 checkGLcall("glProgramLocalParameter4fvARB");
6527 /* Context activation is done by the caller. */
6528 static void arbfp_blit_unset(IWineD3DDevice
*iface
) {
6529 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
6532 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
6533 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
6534 glDisable(GL_TEXTURE_2D
);
6535 checkGLcall("glDisable(GL_TEXTURE_2D)");
6536 if(GL_SUPPORT(ARB_TEXTURE_CUBE_MAP
)) {
6537 glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
6538 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
6540 if(GL_SUPPORT(ARB_TEXTURE_RECTANGLE
)) {
6541 glDisable(GL_TEXTURE_RECTANGLE_ARB
);
6542 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
6547 static BOOL
arbfp_blit_color_fixup_supported(struct color_fixup_desc fixup
)
6549 enum yuv_fixup yuv_fixup
;
6551 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
6553 TRACE("Checking support for fixup:\n");
6554 dump_color_fixup_desc(fixup
);
6557 if (is_identity_fixup(fixup
))
6563 /* We only support YUV conversions. */
6564 if (!is_yuv_fixup(fixup
))
6566 TRACE("[FAILED]\n");
6570 yuv_fixup
= get_yuv_fixup(fixup
);
6573 case YUV_FIXUP_YUY2
:
6574 case YUV_FIXUP_UYVY
:
6575 case YUV_FIXUP_YV12
:
6580 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
6581 TRACE("[FAILED]\n");
6586 const struct blit_shader arbfp_blit
= {
6591 arbfp_blit_color_fixup_supported
,
6594 #undef GLINFO_LOCATION