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 2007-2011, 2013 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
31 #include "wine/port.h"
36 #include "wined3d_private.h"
38 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 /* sRGB correction constants */
43 static const float srgb_cmp
= 0.0031308f
;
44 static const float srgb_mul_low
= 12.92f
;
45 static const float srgb_pow
= 0.41666f
;
46 static const float srgb_mul_high
= 1.055f
;
47 static const float srgb_sub_high
= 0.055f
;
49 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
51 return type
== WINED3D_SHADER_TYPE_PIXEL
;
54 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
56 return type
== WINED3D_SHADER_TYPE_VERTEX
;
59 /* Extract a line. Note that this modifies the source string. */
60 static char *get_line(char **ptr
)
65 if (!(q
= strstr(p
, "\n")))
77 static void shader_arb_dump_program_source(const char *source
)
80 char *ptr
, *line
, *tmp
;
82 source_size
= strlen(source
) + 1;
83 tmp
= HeapAlloc(GetProcessHeap(), 0, source_size
);
86 ERR("Failed to allocate %u bytes for shader source.\n", source_size
);
89 memcpy(tmp
, source
, source_size
);
92 while ((line
= get_line(&ptr
))) FIXME(" %s\n", line
);
95 HeapFree(GetProcessHeap(), 0, tmp
);
109 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
111 if (shader
== WINED3D_SHADER_TYPE_GEOMETRY
)
113 ERR("Geometry shaders are unsupported\n");
117 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
121 case ARB_ZERO
: return "ps_helper_const.x";
122 case ARB_ONE
: return "ps_helper_const.y";
123 case ARB_TWO
: return "coefmul.x";
124 case ARB_0001
: return "ps_helper_const.xxxy";
125 case ARB_EPS
: return "ps_helper_const.z";
133 case ARB_ZERO
: return "helper_const.x";
134 case ARB_ONE
: return "helper_const.y";
135 case ARB_TWO
: return "helper_const.z";
136 case ARB_EPS
: return "helper_const.w";
137 case ARB_0001
: return "helper_const.xxxy";
138 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
141 FIXME("Unmanaged %s shader helper constant requested: %u\n",
142 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
145 case ARB_ZERO
: return "0.0";
146 case ARB_ONE
: return "1.0";
147 case ARB_TWO
: return "2.0";
148 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
149 case ARB_EPS
: return "1e-8";
150 default: return "bad";
154 static inline BOOL
ffp_clip_emul(const struct wined3d_state
*state
)
156 return state
->lowest_disabled_stage
< 7;
159 /* ARB_program_shader private data */
178 struct wined3d_shader_loop_control loop_control
;
182 struct arb_ps_np2fixup_info
184 struct ps_np2fixup_info super
;
185 /* For ARB we need a offset value:
186 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
187 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
188 * array we need an offset to the index inside the program local parameter array. */
192 struct arb_ps_compile_args
194 struct ps_compile_args super
;
196 WORD clip
; /* only a boolean, use a WORD for alignment */
197 unsigned char loop_ctrl
[MAX_CONST_I
][3];
200 struct stb_const_desc
202 unsigned char texunit
;
206 struct arb_ps_compiled_shader
208 struct arb_ps_compile_args args
;
209 struct arb_ps_np2fixup_info np2fixup_info
;
210 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
211 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
212 UINT int_consts
[MAX_CONST_I
];
215 unsigned char numbumpenvmatconsts
;
219 struct arb_vs_compile_args
221 struct vs_compile_args super
;
227 unsigned char clip_texcoord
;
228 unsigned char clipplane_mask
;
230 DWORD boolclip_compare
;
235 unsigned char samplers
[4];
236 DWORD samplers_compare
;
238 unsigned char loop_ctrl
[MAX_CONST_I
][3];
241 struct arb_vs_compiled_shader
243 struct arb_vs_compile_args args
;
245 UINT int_consts
[MAX_CONST_I
];
247 char need_color_unclamp
;
251 struct recorded_instruction
253 struct wined3d_shader_instruction ins
;
257 struct shader_arb_ctx_priv
262 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
264 /* GL_NV_vertex_progam2_option or GL_NV_fragment_program_option */
266 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
270 const struct arb_vs_compile_args
*cur_vs_args
;
271 const struct arb_ps_compile_args
*cur_ps_args
;
272 const struct arb_ps_compiled_shader
*compiled_fprog
;
273 const struct arb_vs_compiled_shader
*compiled_vprog
;
274 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
275 struct list control_frames
;
279 unsigned int num_loops
, loop_depth
, num_ifcs
;
281 BOOL ps_post_process
;
283 unsigned int vs_clipplanes
;
287 /* For 3.0 vertex shaders */
288 const char *vs_output
[MAX_REG_OUTPUT
];
289 /* For 2.x and earlier vertex shaders */
290 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
292 /* 3.0 pshader input for compatibility with fixed function */
293 const char *ps_input
[MAX_REG_INPUT
];
298 struct wined3d_shader_signature_element
*sig
;
300 struct wine_rb_entry entry
;
303 struct arb_pshader_private
{
304 struct arb_ps_compiled_shader
*gl_shaders
;
305 UINT num_gl_shaders
, shader_array_size
;
306 DWORD input_signature_idx
;
307 DWORD clipplane_emulation
;
311 struct arb_vshader_private
{
312 struct arb_vs_compiled_shader
*gl_shaders
;
313 UINT num_gl_shaders
, shader_array_size
;
317 struct shader_arb_priv
319 GLuint current_vprogram_id
;
320 GLuint current_fprogram_id
;
321 const struct arb_ps_compiled_shader
*compiled_fprog
;
322 const struct arb_vs_compiled_shader
*compiled_vprog
;
323 GLuint depth_blt_vprogram_id
;
324 GLuint depth_blt_fprogram_id_full
[tex_type_count
];
325 GLuint depth_blt_fprogram_id_masked
[tex_type_count
];
326 BOOL use_arbfp_fixed_func
;
327 struct wine_rb_tree fragment_shaders
;
328 BOOL last_ps_const_clamped
;
329 BOOL last_vs_color_unclamp
;
331 struct wine_rb_tree signature_tree
;
334 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
335 char *vshader_const_dirty
, *pshader_const_dirty
;
336 const struct wined3d_context
*last_context
;
338 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
339 const struct fragment_pipeline
*fragment_pipe
;
340 BOOL ffp_proj_control
;
343 /* Context activation for state handlers is done by the caller. */
345 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
346 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
348 if (shader_data
->rel_offset
) return TRUE
;
349 if (!reg_maps
->usesmova
) return FALSE
;
350 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
353 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
354 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
356 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
357 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
360 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
361 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
363 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
364 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
365 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
366 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
367 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
368 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
369 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
370 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
374 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
375 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
377 unsigned int ret
= 1;
378 /* We use one PARAM for the pos fixup, and in some cases one to load
379 * some immediate values into the shader. */
380 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
381 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
385 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
386 * When constant_list == NULL, it will load all the constants.
388 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
389 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
391 /* Context activation is done by the caller. */
392 static unsigned int shader_arb_load_constantsF(const struct wined3d_shader
*shader
,
393 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
394 const float *constants
, char *dirty_consts
)
396 struct wined3d_shader_lconst
*lconst
;
400 if (TRACE_ON(d3d_constants
))
402 for(i
= 0; i
< max_constants
; i
++) {
403 if(!dirty_consts
[i
]) continue;
404 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
405 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
406 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
412 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
413 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
416 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
417 * shaders, the first 8 constants are marked dirty for reload
419 for(; i
< min(8, max_constants
); i
++) {
420 if(!dirty_consts
[i
]) continue;
424 if (constants
[j
+ 0] > 1.0f
) lcl_const
[0] = 1.0f
;
425 else if (constants
[j
+ 0] < -1.0f
) lcl_const
[0] = -1.0f
;
426 else lcl_const
[0] = constants
[j
+ 0];
428 if (constants
[j
+ 1] > 1.0f
) lcl_const
[1] = 1.0f
;
429 else if (constants
[j
+ 1] < -1.0f
) lcl_const
[1] = -1.0f
;
430 else lcl_const
[1] = constants
[j
+ 1];
432 if (constants
[j
+ 2] > 1.0f
) lcl_const
[2] = 1.0f
;
433 else if (constants
[j
+ 2] < -1.0f
) lcl_const
[2] = -1.0f
;
434 else lcl_const
[2] = constants
[j
+ 2];
436 if (constants
[j
+ 3] > 1.0f
) lcl_const
[3] = 1.0f
;
437 else if (constants
[j
+ 3] < -1.0f
) lcl_const
[3] = -1.0f
;
438 else lcl_const
[3] = constants
[j
+ 3];
440 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
443 /* If further constants are dirty, reload them without clamping.
445 * The alternative is not to touch them, but then we cannot reset the dirty constant count
446 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
447 * above would always re-check the first 8 constants since max_constant remains at the init
452 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
454 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
455 * or just reloading *all* constants at once
457 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
459 for(; i
< max_constants
; i
++) {
460 if(!dirty_consts
[i
]) continue;
462 /* Find the next block of dirty constants */
465 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
469 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
472 for(; i
< max_constants
; i
++) {
473 if(dirty_consts
[i
]) {
475 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
479 checkGLcall("glProgramEnvParameter4fvARB()");
481 /* Load immediate constants */
482 if (shader
->load_local_constsF
)
484 if (TRACE_ON(d3d_shader
))
486 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
488 GLfloat
* values
= (GLfloat
*)lconst
->value
;
489 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
490 values
[0], values
[1], values
[2], values
[3]);
493 /* Immediate constants are clamped for 1.X shaders at loading times */
495 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
497 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
498 ret
= max(ret
, lconst
->idx
+ 1);
499 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
501 checkGLcall("glProgramEnvParameter4fvARB()");
502 return ret
; /* The loaded immediate constants need reloading for the next shader */
504 return 0; /* No constants are dirty now */
508 /* Loads the texture dimensions for NP2 fixup into the currently set
509 * ARB_[vertex/fragment]_programs. */
510 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
511 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
513 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
514 WORD active
= fixup
->super
.active
;
520 for (i
= 0; active
; active
>>= 1, ++i
)
522 const struct wined3d_texture
*tex
= state
->textures
[i
];
523 unsigned char idx
= fixup
->super
.idx
[i
];
524 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
531 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
537 tex_dim
[2] = tex
->pow2_matrix
[0];
538 tex_dim
[3] = tex
->pow2_matrix
[5];
542 tex_dim
[0] = tex
->pow2_matrix
[0];
543 tex_dim
[1] = tex
->pow2_matrix
[5];
547 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
549 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
550 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
554 /* Context activation is done by the caller. */
555 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
556 const struct wined3d_context
*context
, const struct wined3d_state
*state
, UINT rt_height
)
558 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
561 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
563 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
565 /* The state manager takes care that this function is always called if the bump env matrix changes */
566 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
567 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
568 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
570 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
572 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
573 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
574 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
575 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
577 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
578 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
579 gl_shader
->luminanceconst
[i
].const_num
, scale
));
582 checkGLcall("Load bumpmap consts");
584 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
586 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
587 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
592 val
[0] = context
->render_offscreen
? 0.0f
: (float) rt_height
;
593 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
596 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
597 checkGLcall("y correction loading");
600 if (!gl_shader
->num_int_consts
) return;
602 for(i
= 0; i
< MAX_CONST_I
; i
++)
604 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
607 val
[0] = (float)state
->ps_consts_i
[4 * i
];
608 val
[1] = (float)state
->ps_consts_i
[4 * i
+ 1];
609 val
[2] = (float)state
->ps_consts_i
[4 * i
+ 2];
612 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
615 checkGLcall("Load ps int consts");
618 /* Context activation is done by the caller. */
619 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
620 const struct wined3d_context
*context
, const struct wined3d_state
*state
)
622 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
623 float position_fixup
[4];
626 /* Upload the position fixup */
627 shader_get_position_fixup(context
, state
, position_fixup
);
628 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
630 if (!gl_shader
->num_int_consts
) return;
632 for(i
= 0; i
< MAX_CONST_I
; i
++)
634 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
637 val
[0] = (float)state
->vs_consts_i
[4 * i
];
638 val
[1] = (float)state
->vs_consts_i
[4 * i
+ 1];
639 val
[2] = (float)state
->vs_consts_i
[4 * i
+ 2];
642 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
645 checkGLcall("Load vs int consts");
648 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
649 const struct wined3d_state
*state
);
652 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
654 * We only support float constants in ARB at the moment, so don't
655 * worry about the Integers or Booleans
657 /* Context activation is done by the caller (state handler). */
658 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
,
659 struct wined3d_context
*context
, const struct wined3d_state
*state
,
660 BOOL usePixelShader
, BOOL useVertexShader
, BOOL from_shader_select
)
662 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
663 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
665 if (!from_shader_select
)
667 const struct wined3d_shader
*vshader
= state
->vertex_shader
, *pshader
= state
->pixel_shader
;
669 && (vshader
->reg_maps
.boolean_constants
670 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
671 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
673 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
674 shader_arb_select(priv
, context
, state
);
677 && (pshader
->reg_maps
.boolean_constants
678 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
679 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
681 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
682 shader_arb_select(priv
, context
, state
);
686 if (context
!= priv
->last_context
)
688 memset(priv
->vshader_const_dirty
, 1,
689 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
690 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
692 memset(priv
->pshader_const_dirty
, 1,
693 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
694 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
696 priv
->last_context
= context
;
701 struct wined3d_shader
*vshader
= state
->vertex_shader
;
702 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
704 /* Load DirectX 9 float constants for vertex shader */
705 priv
->highest_dirty_vs_const
= shader_arb_load_constantsF(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
706 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
707 shader_arb_vs_local_constants(gl_shader
, context
, state
);
712 struct wined3d_shader
*pshader
= state
->pixel_shader
;
713 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
714 UINT rt_height
= state
->fb
->render_targets
[0]->resource
.height
;
716 /* Load DirectX 9 float constants for pixel shader */
717 priv
->highest_dirty_ps_const
= shader_arb_load_constantsF(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
718 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
719 shader_arb_ps_local_constants(gl_shader
, context
, state
, rt_height
);
721 if (context
->constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
722 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
726 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
727 const struct wined3d_state
*state
)
729 BOOL vs
= use_vs(state
);
730 BOOL ps
= use_ps(state
);
732 shader_arb_load_constants_internal(shader_priv
, context
, state
, ps
, vs
, FALSE
);
735 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
737 struct wined3d_context
*context
= context_get_current();
738 struct shader_arb_priv
*priv
= device
->shader_priv
;
741 for (i
= 0; i
< device
->context_count
; ++i
)
743 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
746 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
747 * context. On a context switch the old context will be fully dirtified */
748 if (!context
|| context
->swapchain
->device
!= device
) return;
750 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
751 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
754 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
756 struct wined3d_context
*context
= context_get_current();
757 struct shader_arb_priv
*priv
= device
->shader_priv
;
760 for (i
= 0; i
< device
->context_count
; ++i
)
762 device
->contexts
[i
]->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
765 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
766 * context. On a context switch the old context will be fully dirtified */
767 if (!context
|| context
->swapchain
->device
!= device
) return;
769 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
770 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
773 /* Generate the variable & register declarations for the ARB_vertex_program output target */
774 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
775 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_shader_buffer
*buffer
,
776 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
777 const struct shader_arb_ctx_priv
*ctx
)
780 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
781 const struct wined3d_shader_lconst
*lconst
;
782 unsigned max_constantsF
;
785 /* In pixel shaders, all private constants are program local, we don't need anything
786 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
787 * If we need a private constant the GL implementation will squeeze it in somewhere
789 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
790 * immediate values. The posFixup is loaded using program.env for now, so always
791 * subtract one from the number of constants. If the shader uses indirect addressing,
792 * account for the helper const too because we have to declare all available d3d constants
793 * and don't know which are actually used.
797 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
798 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
799 if (max_constantsF
< 24)
800 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
804 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
805 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
806 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
807 * Also prevents max_constantsF from becoming less than 0 and
809 if (max_constantsF
< 96)
810 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
812 if (reg_maps
->usesrelconstF
)
814 DWORD highest_constf
= 0, clip_limit
;
816 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
817 max_constantsF
-= count_bits(reg_maps
->integer_constants
);
818 max_constantsF
-= gl_info
->reserved_arb_constants
;
820 for (i
= 0; i
< shader
->limits
.constant_float
; ++i
)
823 DWORD shift
= i
& 0x1f;
824 if(reg_maps
->constf
[idx
] & (1 << shift
)) highest_constf
= i
;
827 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
829 if(ctx
->cur_vs_args
->super
.clip_enabled
)
830 clip_limit
= gl_info
->limits
.clipplanes
;
836 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
837 clip_limit
= min(count_bits(mask
), 4);
839 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
840 max_constantsF
-= *num_clipplanes
;
841 if(*num_clipplanes
< clip_limit
)
843 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
848 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
849 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
853 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
855 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
858 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
860 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
863 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
865 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
867 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
871 if (!shader
->load_local_constsF
)
873 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
876 value
= (const float *)lconst
->value
;
877 shader_addline(buffer
, "PARAM C%u = {%.8e, %.8e, %.8e, %.8e};\n", lconst
->idx
,
878 value
[0], value
[1], value
[2], value
[3]);
882 /* After subtracting privately used constants from the hardware limit(they are loaded as
883 * local constants), make sure the shader doesn't violate the env constant limit
887 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
891 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
894 /* Avoid declaring more constants than needed */
895 max_constantsF
= min(max_constantsF
, shader
->limits
.constant_float
);
897 /* we use the array-based constants array if the local constants are marked for loading,
898 * because then we use indirect addressing, or when the local constant list is empty,
899 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
900 * local constants do not declare the loaded constants as an array because ARB compilers usually
901 * do not optimize unused constants away
903 if (reg_maps
->usesrelconstF
)
905 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
906 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
907 max_constantsF
, max_constantsF
- 1);
909 for(i
= 0; i
< max_constantsF
; i
++) {
912 mask
= 1 << (i
& 0x1f);
913 if (!shader_constant_is_local(shader
, i
) && (reg_maps
->constf
[idx
] & mask
))
915 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
921 static const char * const shift_tab
[] = {
922 "dummy", /* 0 (none) */
923 "coefmul.x", /* 1 (x2) */
924 "coefmul.y", /* 2 (x4) */
925 "coefmul.z", /* 3 (x8) */
926 "coefmul.w", /* 4 (x16) */
927 "dummy", /* 5 (x32) */
928 "dummy", /* 6 (x64) */
929 "dummy", /* 7 (x128) */
930 "dummy", /* 8 (d256) */
931 "dummy", /* 9 (d128) */
932 "dummy", /* 10 (d64) */
933 "dummy", /* 11 (d32) */
934 "coefdiv.w", /* 12 (d16) */
935 "coefdiv.z", /* 13 (d8) */
936 "coefdiv.y", /* 14 (d4) */
937 "coefdiv.x" /* 15 (d2) */
940 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
941 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
943 char *ptr
= write_mask
;
945 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
948 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
949 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
950 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
951 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
957 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
959 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
960 * but addressed as "rgba". To fix this we need to swap the register's x
961 * and z components. */
962 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
963 char *ptr
= swizzle_str
;
965 /* swizzle bits fields: wwzzyyxx */
966 DWORD swizzle
= param
->swizzle
;
967 DWORD swizzle_x
= swizzle
& 0x03;
968 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
969 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
970 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
972 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
973 * generate a swizzle string. Unless we need to our own swizzling. */
974 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
977 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
978 *ptr
++ = swizzle_chars
[swizzle_x
];
980 *ptr
++ = swizzle_chars
[swizzle_x
];
981 *ptr
++ = swizzle_chars
[swizzle_y
];
982 *ptr
++ = swizzle_chars
[swizzle_z
];
983 *ptr
++ = swizzle_chars
[swizzle_w
];
990 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
992 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
993 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
995 if (!strcmp(priv
->addr_reg
, src
)) return;
997 strcpy(priv
->addr_reg
, src
);
998 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
1001 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1002 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
1004 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
1005 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
1007 /* oPos, oFog and oPts in D3D */
1008 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
1009 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1010 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1011 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1012 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1018 case WINED3DSPR_TEMP
:
1019 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1022 case WINED3DSPR_INPUT
:
1025 if (reg_maps
->shader_version
.major
< 3)
1027 if (!reg
->idx
[0].offset
)
1028 strcpy(register_name
, "fragment.color.primary");
1030 strcpy(register_name
, "fragment.color.secondary");
1034 if (reg
->idx
[0].rel_addr
)
1037 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1039 if (!strcmp(rel_reg
, "**aL_emul**"))
1041 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1042 if(idx
< MAX_REG_INPUT
)
1044 strcpy(register_name
, ctx
->ps_input
[idx
]);
1048 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1049 sprintf(register_name
, "out_of_bounds_%u", idx
);
1052 else if (reg_maps
->input_registers
& 0x0300)
1054 /* There are two ways basically:
1056 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1057 * That means trouble if the loop also contains a breakc or if the control values
1058 * aren't local constants.
1059 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1060 * source dynamically. The trouble is that we cannot simply read aL.y because it
1061 * is an ADDRESS register. We could however push it, load .zw with a value and use
1062 * ADAC to load the condition code register and pop it again afterwards
1064 FIXME("Relative input register addressing with more than 8 registers\n");
1066 /* This is better than nothing for now */
1067 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1069 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
1071 /* This is problematic because we'd have to consult the ctx->ps_input strings
1072 * for where to find the varying. Some may be "0.0", others can be texcoords or
1073 * colors. This needs either a pipeline replacement to make the vertex shader feed
1074 * proper varyings, or loop unrolling
1076 * For now use the texcoords and hope for the best
1078 FIXME("Non-vertex shader varying input with indirect addressing\n");
1079 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1083 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1084 * pulls GL_NV_fragment_program2 in
1086 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1091 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1093 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1097 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1098 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1105 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
[0].offset
))
1107 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1111 case WINED3DSPR_CONST
:
1112 if (!pshader
&& reg
->idx
[0].rel_addr
)
1114 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1115 UINT rel_offset
= shader_data
->rel_offset
;
1118 if (reg_maps
->shader_version
.major
< 2)
1120 sprintf(rel_reg
, "A0.x");
1124 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1125 if (ctx
->target_version
== ARB
)
1127 if (!strcmp(rel_reg
, "**aL_emul**"))
1131 shader_arb_request_a0(ins
, rel_reg
);
1132 sprintf(rel_reg
, "A0.x");
1137 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1138 else if (reg
->idx
[0].offset
>= rel_offset
)
1139 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1141 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1145 if (reg_maps
->usesrelconstF
)
1146 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1148 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1152 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1155 if (reg_maps
->shader_version
.major
== 1
1156 && reg_maps
->shader_version
.minor
<= 3)
1157 /* In ps <= 1.3, Tx is a temporary register as destination
1158 * to all instructions, and as source to most instructions.
1159 * For some instructions it is the texcoord input. Those
1160 * instructions know about the special use. */
1161 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1163 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1164 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1168 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1169 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1171 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1175 case WINED3DSPR_COLOROUT
:
1176 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1178 strcpy(register_name
, "TMP_COLOR");
1182 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1183 FIXME("sRGB correction on higher render targets.\n");
1184 if (reg_maps
->rt_mask
> 1)
1185 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1187 strcpy(register_name
, "result.color");
1191 case WINED3DSPR_RASTOUT
:
1192 if (reg
->idx
[0].offset
== 1)
1193 sprintf(register_name
, "%s", ctx
->fog_output
);
1195 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1198 case WINED3DSPR_DEPTHOUT
:
1199 strcpy(register_name
, "result.depth");
1202 case WINED3DSPR_ATTROUT
:
1203 /* case WINED3DSPR_OUTPUT: */
1205 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1207 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1210 case WINED3DSPR_TEXCRDOUT
:
1212 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1213 else if (reg_maps
->shader_version
.major
< 3)
1214 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1216 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1219 case WINED3DSPR_LOOP
:
1220 if(ctx
->target_version
>= NV2
)
1222 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1223 if(pshader
) sprintf(register_name
, "A0.x");
1224 else sprintf(register_name
, "aL.y");
1228 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1229 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1230 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1231 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1234 sprintf(register_name
, "**aL_emul**");
1239 case WINED3DSPR_CONSTINT
:
1240 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1243 case WINED3DSPR_MISCTYPE
:
1244 if (!reg
->idx
[0].offset
)
1245 sprintf(register_name
, "vpos");
1246 else if (reg
->idx
[0].offset
== 1)
1247 sprintf(register_name
, "fragment.facing.x");
1249 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1253 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1254 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1259 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1260 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1262 char register_name
[255];
1266 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1267 strcpy(str
, register_name
);
1269 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1270 strcat(str
, write_mask
);
1273 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1275 switch(channel_source
)
1277 case CHANNEL_SOURCE_ZERO
: return "0";
1278 case CHANNEL_SOURCE_ONE
: return "1";
1279 case CHANNEL_SOURCE_X
: return "x";
1280 case CHANNEL_SOURCE_Y
: return "y";
1281 case CHANNEL_SOURCE_Z
: return "z";
1282 case CHANNEL_SOURCE_W
: return "w";
1284 FIXME("Unhandled channel source %#x\n", channel_source
);
1289 static void gen_color_correction(struct wined3d_shader_buffer
*buffer
, const char *reg
,
1290 DWORD dst_mask
, const char *one
, const char *two
, struct color_fixup_desc fixup
)
1294 if (is_complex_fixup(fixup
))
1296 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1297 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1302 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1303 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1304 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1305 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1310 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
1311 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1312 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1316 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
1317 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
1318 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
1319 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
1325 char *ptr
= reg_mask
;
1327 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
1330 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1331 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1332 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1333 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1337 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
1341 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1344 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1345 if (!ins
->dst_count
) return "";
1347 mod
= ins
->dst
[0].modifiers
;
1349 /* Silently ignore PARTIALPRECISION if its not supported */
1350 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1352 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1354 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1355 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1360 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1363 case WINED3DSPDM_SATURATE
:
1366 case WINED3DSPDM_PARTIALPRECISION
:
1373 FIXME("Unknown modifiers 0x%08x\n", mod
);
1378 #define TEX_PROJ 0x1
1379 #define TEX_BIAS 0x2
1381 #define TEX_DERIV 0x10
1383 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1384 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1386 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1387 DWORD sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
1388 const char *tex_type
;
1389 BOOL np2_fixup
= FALSE
;
1390 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1392 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1393 const struct wined3d_shader
*shader
;
1394 const struct wined3d_device
*device
;
1395 const struct wined3d_gl_info
*gl_info
;
1397 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1398 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1400 switch(sampler_type
) {
1406 shader
= ins
->ctx
->shader
;
1407 device
= shader
->device
;
1408 gl_info
= &device
->adapter
->gl_info
;
1410 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1 << sampler_idx
)
1411 && gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1415 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1417 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1419 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1420 else np2_fixup
= TRUE
;
1425 case WINED3DSTT_VOLUME
:
1429 case WINED3DSTT_CUBE
:
1434 ERR("Unexpected texture type %d\n", sampler_type
);
1438 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1439 * so don't use shader_arb_get_modifier
1441 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1444 /* Fragment samplers always have indentity mapping */
1445 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1447 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1450 if (flags
& TEX_DERIV
)
1452 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1453 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1454 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1455 dsx
, dsy
,sampler_idx
, tex_type
);
1457 else if(flags
& TEX_LOD
)
1459 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1460 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1461 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1462 sampler_idx
, tex_type
);
1464 else if (flags
& TEX_BIAS
)
1466 /* Shouldn't be possible, but let's check for it */
1467 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1468 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1469 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1471 else if (flags
& TEX_PROJ
)
1473 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1479 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1480 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1481 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1483 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, dst_str
, sampler_idx
, tex_type
);
1486 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1491 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
1492 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1493 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1494 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
1498 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1499 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1501 /* Generate a line that does the input modifier computation and return the input register to use */
1502 BOOL is_color
= FALSE
;
1506 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1507 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1508 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1509 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1511 /* Assume a new line will be added */
1514 /* Get register name */
1515 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1516 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1518 switch (src
->modifiers
)
1520 case WINED3DSPSM_NONE
:
1521 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1524 case WINED3DSPSM_NEG
:
1525 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1528 case WINED3DSPSM_BIAS
:
1529 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1531 case WINED3DSPSM_BIASNEG
:
1532 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1534 case WINED3DSPSM_SIGN
:
1535 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1537 case WINED3DSPSM_SIGNNEG
:
1538 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1540 case WINED3DSPSM_COMP
:
1541 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1543 case WINED3DSPSM_X2
:
1544 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1546 case WINED3DSPSM_X2NEG
:
1547 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1549 case WINED3DSPSM_DZ
:
1550 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1551 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1553 case WINED3DSPSM_DW
:
1554 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1555 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1557 case WINED3DSPSM_ABS
:
1558 if(ctx
->target_version
>= NV2
) {
1559 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1562 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1565 case WINED3DSPSM_ABSNEG
:
1566 if(ctx
->target_version
>= NV2
) {
1567 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1569 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1570 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1575 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1579 /* Return modified or original register, with swizzle */
1581 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1584 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1586 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1587 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1588 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1590 char src_name
[2][50];
1592 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1594 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1596 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1597 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1600 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1601 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1602 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1603 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1604 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1606 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1607 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1610 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1615 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1616 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1617 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1618 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1619 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1620 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1621 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1622 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1623 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1624 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1625 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1626 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1627 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1629 FIXME("Unknown modifier %u\n", mod
);
1633 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1635 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1636 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1638 char src_name
[3][50];
1639 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1640 ins
->ctx
->reg_maps
->shader_version
.minor
);
1642 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1643 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1645 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1646 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1648 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1652 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1655 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1656 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1658 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1659 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1660 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1661 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1662 dst_name
, src_name
[1], src_name
[2]);
1666 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1668 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1669 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1671 char src_name
[3][50];
1673 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1675 /* Generate input register names (with modifiers) */
1676 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1677 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1678 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1680 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1681 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1684 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1685 * dst = dot2(src0, src1) + src2 */
1686 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1688 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1689 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1691 char src_name
[3][50];
1692 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1694 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1695 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1696 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1698 if(ctx
->target_version
>= NV3
)
1700 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1701 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1702 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1703 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1705 else if(ctx
->target_version
>= NV2
)
1707 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1708 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1709 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1710 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1712 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1714 * .xyxy and other swizzles that we could get with this are not valid in
1715 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1717 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1718 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1719 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1721 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1723 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1724 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1728 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1729 /* 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]
1730 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1732 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1733 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1734 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1735 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1739 /* Map the opcode 1-to-1 to the GL code */
1740 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1742 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1743 const char *instruction
;
1744 char arguments
[256], dst_str
[50];
1746 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1748 switch (ins
->handler_idx
)
1750 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1751 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1752 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1753 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1754 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1755 case WINED3DSIH_DST
: instruction
= "DST"; break;
1756 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1757 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1758 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1759 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1760 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1761 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1762 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1763 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1764 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1765 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1766 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1767 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1768 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1769 default: instruction
= "";
1770 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1774 /* Note that shader_arb_add_dst_param() adds spaces. */
1775 arguments
[0] = '\0';
1776 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1777 for (i
= 0; i
< ins
->src_count
; ++i
)
1780 strcat(arguments
, ", ");
1781 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1782 strcat(arguments
, operand
);
1784 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1787 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1789 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1791 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1792 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1793 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1794 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1795 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1796 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1797 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1799 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1800 char src0_param
[256];
1802 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1804 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1806 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1808 if(ctx
->target_version
>= NV2
) {
1809 shader_hw_map2gl(ins
);
1812 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1813 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1815 /* This implements the mova formula used in GLSL. The first two instructions
1816 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1820 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1822 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1823 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1825 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1826 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1828 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1829 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1830 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1831 if (shader_data
->rel_offset
)
1833 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1835 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1837 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1839 else if (reg_maps
->shader_version
.major
== 1
1840 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1841 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1843 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1844 src0_param
[0] = '\0';
1846 if (shader_data
->rel_offset
)
1848 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1849 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1850 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1851 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1855 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1856 * with more than one component. Thus replicate the first source argument over all
1857 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1858 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1859 tmp_src
.swizzle
= (tmp_src
.swizzle
& 0x3) * 0x55;
1860 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1861 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1864 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1866 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1868 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1871 shader_hw_map2gl(ins
);
1875 shader_hw_map2gl(ins
);
1879 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1881 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1882 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1885 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1886 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1888 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1890 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1892 const char *kilsrc
= "TA";
1895 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1896 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1902 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1903 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1904 * masked out components to 0(won't kill)
1906 char x
= '0', y
= '0', z
= '0', w
= '0';
1907 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1908 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1909 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1910 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1911 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1913 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1917 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1918 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1920 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1921 * or pass in any temporary register(in shader phase 2)
1923 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1924 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1926 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1927 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1928 shader_addline(buffer
, "KIL TA;\n");
1932 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1934 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1935 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1936 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1937 ins
->ctx
->reg_maps
->shader_version
.minor
);
1938 struct wined3d_shader_src_param src
;
1942 DWORD reg_sampler_code
;
1944 BOOL swizzle_coord
= FALSE
;
1946 /* All versions have a destination register */
1947 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1949 /* 1.0-1.4: Use destination register number as texture code.
1950 2.0+: Use provided sampler number as texure code. */
1951 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1952 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1954 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1956 /* 1.0-1.3: Use the texcoord varying.
1957 1.4+: Use provided coordinate source register. */
1958 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1959 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1961 /* TEX is the only instruction that can handle DW and DZ natively */
1963 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1964 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1965 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1969 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
1970 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1971 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1973 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1976 if (reg_sampler_code
< MAX_TEXTURES
)
1977 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
1978 if (flags
& WINED3D_PSARGS_PROJECTED
)
1980 myflags
|= TEX_PROJ
;
1981 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
1982 swizzle_coord
= TRUE
;
1985 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1987 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
1988 if (src_mod
== WINED3DSPSM_DZ
)
1990 swizzle_coord
= TRUE
;
1991 myflags
|= TEX_PROJ
;
1992 } else if(src_mod
== WINED3DSPSM_DW
) {
1993 myflags
|= TEX_PROJ
;
1996 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
1997 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2002 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2003 * reg_coord is a read-only varying register, so we need a temp reg */
2004 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2005 strcpy(reg_coord
, "TA");
2008 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2011 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2013 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2014 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2015 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2016 ins
->ctx
->reg_maps
->shader_version
.minor
);
2019 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2021 DWORD reg
= dst
->reg
.idx
[0].offset
;
2023 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2024 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2028 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2029 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2030 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2034 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2036 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2039 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2043 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2044 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2045 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2046 /* Move .x first in case src_str is "TA" */
2047 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2048 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2049 if (reg1
< MAX_TEXTURES
)
2051 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2052 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2054 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2057 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2059 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2061 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2065 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2066 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2067 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2068 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2069 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2070 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2073 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2075 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2079 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2080 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2081 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2082 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2085 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2087 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2088 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2089 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2090 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2091 DWORD reg_dest_code
;
2093 /* All versions have a destination register. The Tx where the texture coordinates come
2094 * from is the varying incarnation of the texture register
2096 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2097 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2098 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2099 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2101 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2102 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2104 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2105 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2107 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2108 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2111 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2112 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2113 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2114 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2116 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2117 * so we can't let the GL handle this.
2119 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2120 & WINED3D_PSARGS_PROJECTED
)
2122 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2123 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2124 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2126 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2129 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2131 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2133 /* No src swizzles are allowed, so this is ok */
2134 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2135 src_reg
, reg_dest_code
, reg_dest_code
);
2136 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2140 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2142 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2143 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2144 char src0_name
[50], dst_name
[50];
2146 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2148 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2149 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2150 * T<reg+1> register. Use this register to store the calculated vector
2152 tmp_reg
.idx
[0].offset
= reg
+ 1;
2153 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2154 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2157 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2159 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2161 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2162 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2168 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2169 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2171 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2172 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2173 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2174 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2175 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2178 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2180 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2181 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2182 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2183 char src0_name
[50], dst_name
[50];
2184 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2187 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2188 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2189 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2191 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2192 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2194 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2195 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2196 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2197 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2200 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2202 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2203 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2205 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2206 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2208 char src0_name
[50], dst_name
[50];
2211 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2212 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2213 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2215 /* Sample the texture using the calculated coordinates */
2216 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2217 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2218 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2219 tex_mx
->current_row
= 0;
2222 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2224 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2225 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2227 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2228 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2234 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2235 * components for temporary data storage
2237 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2238 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2239 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2241 /* Construct the eye-ray vector from w coordinates */
2242 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2243 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2244 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2246 /* Calculate reflection vector
2248 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2249 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2250 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2251 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2252 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2253 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2254 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2256 /* Sample the texture using the calculated coordinates */
2257 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2258 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2259 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2260 tex_mx
->current_row
= 0;
2263 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2265 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2266 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2268 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2269 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2276 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2277 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2278 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2279 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2280 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2282 /* Calculate reflection vector.
2285 * dst_reg.xyz = 2 * --------- * N - E
2288 * Which normalizes the normal vector
2290 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2291 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2292 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2293 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2294 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2295 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2297 /* Sample the texture using the calculated coordinates */
2298 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2299 flags
= reg
< MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2300 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2301 tex_mx
->current_row
= 0;
2304 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2306 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2307 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2309 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2310 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2312 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2313 * which is essentially an input, is the destination register because it is the first
2314 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2315 * here(writemasks/swizzles are not valid on texdepth)
2317 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2319 /* According to the msdn, the source register(must be r5) is unusable after
2320 * the texdepth instruction, so we're free to modify it
2322 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2324 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2325 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2326 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2328 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2329 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2330 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2331 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2334 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2335 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2336 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2337 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2339 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2340 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2344 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2345 shader_addline(buffer
, "MOV TB, 0.0;\n");
2346 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2348 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2349 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2352 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2353 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2354 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2356 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2359 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2361 /* Handle output register */
2362 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2363 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2364 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2367 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2368 * Perform the 3rd row of a 3x3 matrix multiply */
2369 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2371 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2372 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2373 char dst_str
[50], dst_name
[50];
2377 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2378 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2379 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2380 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2381 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2384 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2385 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2386 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2387 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2389 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2391 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2392 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2393 char src0
[50], dst_name
[50];
2395 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2396 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2398 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2399 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2400 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2402 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2403 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2404 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2406 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2407 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2408 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2409 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2412 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2413 Vertex/Pixel shaders to ARB_vertex_program codes */
2414 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2417 int nComponents
= 0;
2418 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2419 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2420 struct wined3d_shader_instruction tmp_ins
;
2422 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2424 /* Set constants for the temporary argument */
2425 tmp_ins
.ctx
= ins
->ctx
;
2426 tmp_ins
.dst_count
= 1;
2427 tmp_ins
.dst
= &tmp_dst
;
2428 tmp_ins
.src_count
= 2;
2429 tmp_ins
.src
= tmp_src
;
2431 switch(ins
->handler_idx
)
2433 case WINED3DSIH_M4x4
:
2435 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2437 case WINED3DSIH_M4x3
:
2439 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2441 case WINED3DSIH_M3x4
:
2443 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2445 case WINED3DSIH_M3x3
:
2447 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2449 case WINED3DSIH_M3x2
:
2451 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2454 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2458 tmp_dst
= ins
->dst
[0];
2459 tmp_src
[0] = ins
->src
[0];
2460 tmp_src
[1] = ins
->src
[1];
2461 for (i
= 0; i
< nComponents
; ++i
)
2463 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2464 shader_hw_map2gl(&tmp_ins
);
2465 ++tmp_src
[1].reg
.idx
[0].offset
;
2469 static void shader_hw_rcp(const struct wined3d_shader_instruction
*ins
)
2471 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2476 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2477 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2478 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2480 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2486 shader_addline(buffer
, "RCP%s %s, %s;\n", shader_arb_get_modifier(ins
), dst
, src
);
2489 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2491 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2492 const char *instruction
;
2497 switch(ins
->handler_idx
)
2499 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2500 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2501 case WINED3DSIH_EXP
: instruction
= "EX2"; break;
2502 case WINED3DSIH_EXPP
: instruction
= "EXP"; break;
2503 default: instruction
= "";
2504 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2508 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2509 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2510 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2512 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2518 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2521 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2523 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2526 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2527 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2528 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2530 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2531 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2533 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2534 * otherwise NRM or RSQ would return NaN */
2535 if(pshader
&& priv
->target_version
>= NV3
)
2537 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2539 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2541 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2542 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2543 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2545 else if(priv
->target_version
>= NV2
)
2547 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2548 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2549 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2554 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2556 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2557 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2558 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2560 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2561 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2563 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2564 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2565 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2570 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2572 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2574 char src_name
[3][50];
2576 /* ARB_fragment_program has a convenient LRP instruction */
2577 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2578 shader_hw_map2gl(ins
);
2582 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2583 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2584 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2585 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2587 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2588 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2589 dst_name
, src_name
[0], src_name
[2]);
2592 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2594 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2595 * must contain fixed constants. So we need a separate function to filter those constants and
2598 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2599 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2600 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2602 char src_name0
[50], src_name1
[50], src_name2
[50];
2605 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2606 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2607 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2608 /* No modifiers are supported on SCS */
2609 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2611 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2613 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2614 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2616 } else if(priv
->target_version
>= NV2
) {
2617 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2619 /* Sincos writemask must be .x, .y or .xy */
2620 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2621 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2622 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2623 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2625 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2626 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2628 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2629 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2631 * The constants we get are:
2633 * +1 +1, -1 -1 +1 +1 -1 -1
2634 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2635 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2637 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2641 * (x/2)^4 = x^4 / 16
2642 * (x/2)^5 = x^5 / 32
2645 * To get the final result:
2646 * sin(x) = 2 * sin(x/2) * cos(x/2)
2647 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2648 * (from sin(x+y) and cos(x+y) rules)
2650 * As per MSDN, dst.z is undefined after the operation, and so is
2651 * dst.x and dst.y if they're masked out by the writemask. Ie
2652 * sincos dst.y, src1, c0, c1
2653 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2654 * vsa.exe also stops with an error if the dest register is the same register as the source
2655 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2656 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2658 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2659 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2660 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2662 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2663 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2664 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2665 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2666 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2667 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2671 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2672 * properly merge that with MULs in the code above?
2673 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2674 * we can merge the sine and cosine MAD rows to calculate them together.
2676 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2677 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2678 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2679 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2682 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2683 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2684 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2686 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2688 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2689 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2691 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2693 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2694 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2699 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2701 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2704 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2706 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2707 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2709 /* SGN is only valid in vertex shaders */
2710 if(ctx
->target_version
>= NV2
) {
2711 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2715 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2716 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2718 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2719 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2721 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2722 * Then use TA, and calculate the final result
2724 * Not reading from TA? Store the first result in TA to avoid overwriting the
2725 * destination if src reg = dst reg
2727 if(strstr(src_name
, "TA"))
2729 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2730 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2731 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2735 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2736 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2737 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2742 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2744 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2750 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2751 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2752 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2754 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2755 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2758 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2764 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2765 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2766 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2767 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2768 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2769 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2770 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2771 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2772 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2773 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2774 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2775 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2776 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2778 FIXME("Unknown modifier %u\n", mod
);
2782 static void shader_hw_log(const struct wined3d_shader_instruction
*ins
)
2784 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2785 char src0
[50], dst
[50];
2786 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2787 BOOL need_abs
= FALSE
;
2790 switch(ins
->handler_idx
)
2792 case WINED3DSIH_LOG
: instr
= "LG2"; break;
2793 case WINED3DSIH_LOGP
: instr
= "LOG"; break;
2795 ERR("Unexpected instruction %d\n", ins
->handler_idx
);
2799 /* LOG and LOGP operate on the absolute value of the input */
2800 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2802 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2803 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2807 shader_addline(buffer
, "ABS TA, %s;\n", src0
);
2808 shader_addline(buffer
, "%s%s %s, TA;\n", instr
, shader_arb_get_modifier(ins
), dst
);
2812 shader_addline(buffer
, "%s%s %s, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src0
);
2816 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2818 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2819 char src0
[50], src1
[50], dst
[50];
2820 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2821 BOOL need_abs
= FALSE
;
2822 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2823 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2825 /* POW operates on the absolute value of the input */
2826 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2828 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2829 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2830 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2833 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2835 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2837 if (priv
->target_version
>= NV2
)
2839 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2840 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2841 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2845 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2846 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2848 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2849 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2850 /* Possibly add flt_eps to avoid getting float special values */
2851 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2852 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2853 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2854 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2858 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2860 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2862 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2865 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2869 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2870 struct list
*e
= list_head(&priv
->control_frames
);
2871 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2873 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2874 /* The constant loader makes sure to load -1 into iX.w */
2875 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2876 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2877 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2881 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2885 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2887 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2889 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2891 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2893 /* The constant loader makes sure to load -1 into iX.w */
2896 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2897 struct list
*e
= list_head(&priv
->control_frames
);
2898 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2900 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2902 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2903 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2904 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2908 shader_addline(buffer
, "REP %s;\n", src_name
);
2912 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2914 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2915 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2919 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2920 struct list
*e
= list_head(&priv
->control_frames
);
2921 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2923 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2924 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2925 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2927 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2931 shader_addline(buffer
, "ENDLOOP;\n");
2935 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2937 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2938 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2942 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2943 struct list
*e
= list_head(&priv
->control_frames
);
2944 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2946 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2947 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2948 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2950 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2954 shader_addline(buffer
, "ENDREP;\n");
2958 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2960 struct control_frame
*control_frame
;
2962 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2964 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2966 ERR("Could not find loop for break\n");
2970 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2972 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2973 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2974 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2978 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2982 shader_addline(buffer
, "BRK;\n");
2986 static const char *get_compare(enum wined3d_shader_rel_op op
)
2990 case WINED3D_SHADER_REL_OP_GT
: return "GT";
2991 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
2992 case WINED3D_SHADER_REL_OP_GE
: return "GE";
2993 case WINED3D_SHADER_REL_OP_LT
: return "LT";
2994 case WINED3D_SHADER_REL_OP_NE
: return "NE";
2995 case WINED3D_SHADER_REL_OP_LE
: return "LE";
2997 FIXME("Unrecognized operator %#x.\n", op
);
3002 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3006 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3007 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3008 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3009 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3010 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3011 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3013 FIXME("Unrecognized operator %#x.\n", op
);
3018 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3020 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3021 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3022 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3025 const char *comp
= get_compare(ins
->flags
);
3027 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3028 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3032 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3033 * away the subtraction result
3035 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3036 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3040 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3041 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3045 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3047 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3048 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3049 struct list
*e
= list_head(&priv
->control_frames
);
3050 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3054 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3056 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3057 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3061 /* Invert the flag. We jump to the else label if the condition is NOT true */
3062 comp
= get_compare(invert_compare(ins
->flags
));
3063 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3064 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3068 comp
= get_compare(ins
->flags
);
3069 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3070 shader_addline(buffer
, "IF %s.x;\n", comp
);
3074 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3076 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3077 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3078 struct list
*e
= list_head(&priv
->control_frames
);
3079 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3080 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3084 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3085 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3086 control_frame
->had_else
= TRUE
;
3090 shader_addline(buffer
, "ELSE;\n");
3094 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3096 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3097 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3098 struct list
*e
= list_head(&priv
->control_frames
);
3099 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3100 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3104 if(control_frame
->had_else
)
3106 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3110 shader_addline(buffer
, "#No else branch. else is endif\n");
3111 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3116 shader_addline(buffer
, "ENDIF;\n");
3120 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3122 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3124 char reg_src
[3][40];
3125 WORD flags
= TEX_DERIV
;
3127 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3128 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3129 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3130 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3132 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3133 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3135 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3138 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3140 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3143 WORD flags
= TEX_LOD
;
3145 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3146 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3148 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3149 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3151 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3154 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3156 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3157 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3159 priv
->in_main_func
= FALSE
;
3160 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3161 * subroutine, don't generate a label that will make GL complain
3163 if(priv
->target_version
== ARB
) return;
3165 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3168 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3169 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3170 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3171 struct wined3d_shader_buffer
*buffer
)
3175 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3176 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3177 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3178 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3180 if (args
->super
.fog_src
== VS_FOG_Z
)
3182 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3188 /* posFixup.x is always 1.0, so we can safely use it */
3189 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3193 /* Clamp fogcoord */
3194 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3195 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3197 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3198 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3202 /* Clipplanes are always stored without y inversion */
3203 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3205 if (args
->super
.clip_enabled
)
3207 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3209 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3213 else if (args
->clip
.boolclip
.clip_texcoord
)
3215 unsigned int cur_clip
= 0;
3216 char component
[4] = {'x', 'y', 'z', 'w'};
3217 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3219 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3221 if (args
->clip
.boolclip
.clipplane_mask
& (1 << i
))
3223 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3224 component
[cur_clip
++], i
);
3230 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3233 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3236 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3239 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3242 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3243 args
->clip
.boolclip
.clip_texcoord
- 1);
3246 /* Write the final position.
3248 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3249 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3250 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3251 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3253 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3254 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3255 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3257 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3258 * and the glsl equivalent
3260 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3262 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3263 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3267 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3268 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3271 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3273 priv_ctx
->footer_written
= TRUE
;
3276 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3278 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3279 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3280 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3281 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3283 if(priv
->target_version
== ARB
) return;
3287 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3288 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, ins
->ctx
->gl_info
, buffer
);
3291 shader_addline(buffer
, "RET;\n");
3294 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3296 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3297 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3300 /* Context activation is done by the caller. */
3301 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3303 GLuint program_id
= 0;
3306 const char *blt_vprogram
=
3308 "PARAM c[1] = { { 1, 0.5 } };\n"
3309 "MOV result.position, vertex.position;\n"
3310 "MOV result.color, c[0].x;\n"
3311 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3314 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3315 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3316 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3317 strlen(blt_vprogram
), blt_vprogram
));
3318 checkGLcall("glProgramStringARB()");
3320 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3323 FIXME("Vertex program error at position %d: %s\n\n", pos
,
3324 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3325 shader_arb_dump_program_source(blt_vprogram
);
3331 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3332 checkGLcall("glGetProgramivARB()");
3333 if (!native
) WARN("Program exceeds native resource limits.\n");
3339 /* Context activation is done by the caller. */
3340 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3341 enum tex_types tex_type
, BOOL masked
)
3343 GLuint program_id
= 0;
3344 const char *fprogram
;
3347 static const char * const blt_fprograms_full
[tex_type_count
] =
3354 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3355 "MOV result.depth.z, R0.x;\n"
3362 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3363 "MOV result.depth.z, R0.x;\n"
3368 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3369 "MOV result.depth.z, R0.x;\n"
3373 static const char * const blt_fprograms_masked
[tex_type_count
] =
3379 "PARAM mask = program.local[0];\n"
3381 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3382 "MUL R0.x, R0.x, R0.y;\n"
3384 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3385 "MOV result.depth.z, R0.x;\n"
3391 "PARAM mask = program.local[0];\n"
3393 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3394 "MUL R0.x, R0.x, R0.y;\n"
3396 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3397 "MOV result.depth.z, R0.x;\n"
3401 "PARAM mask = program.local[0];\n"
3403 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3404 "MUL R0.x, R0.x, R0.y;\n"
3406 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3407 "MOV result.depth.z, R0.x;\n"
3411 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3414 FIXME("tex_type %#x not supported, falling back to tex_2d\n", tex_type
);
3416 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3419 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3420 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3421 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(fprogram
), fprogram
));
3422 checkGLcall("glProgramStringARB()");
3424 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3427 FIXME("Fragment program error at position %d: %s\n\n", pos
,
3428 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3429 shader_arb_dump_program_source(fprogram
);
3435 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3436 checkGLcall("glGetProgramivARB()");
3437 if (!native
) WARN("Program exceeds native resource limits.\n");
3443 static void arbfp_add_sRGB_correction(struct wined3d_shader_buffer
*buffer
, const char *fragcolor
,
3444 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3446 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3450 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3451 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.y;\n", tmp1
, fragcolor
);
3452 /* Calculate the > 0.0031308 case */
3453 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3454 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3455 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3456 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts1.w;\n", fragcolor
, fragcolor
);
3457 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts2.x;\n", fragcolor
, fragcolor
);
3458 /* Calculate the < case */
3459 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts1.x, %s;\n", fragcolor
, fragcolor
);
3463 /* Calculate the > 0.0031308 case */
3464 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts1.z;\n", tmp1
, fragcolor
);
3465 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts1.z;\n", tmp1
, fragcolor
);
3466 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts1.z;\n", tmp1
, fragcolor
);
3467 shader_addline(buffer
, "MUL %s, %s, srgb_consts1.w;\n", tmp1
, tmp1
);
3468 shader_addline(buffer
, "SUB %s, %s, srgb_consts2.x;\n", tmp1
, tmp1
);
3469 /* Calculate the < case */
3470 shader_addline(buffer
, "MUL %s, srgb_consts1.x, %s;\n", tmp2
, fragcolor
);
3471 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3472 shader_addline(buffer
, "SLT %s, srgb_consts1.y, %s;\n", tmp3
, fragcolor
);
3473 shader_addline(buffer
, "SGE %s, srgb_consts1.y, %s;\n", tmp4
, fragcolor
);
3474 /* Store the components > 0.0031308 in the destination */
3475 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3476 /* Add the components that are < 0.0031308 */
3477 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3478 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3479 * result.color writes(.rgb first, then .a), or handle overwriting already written
3480 * components. The assembler uses a temporary register in this case, which is usually
3481 * not allocated from one of our registers that were used earlier.
3484 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3487 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3489 const struct wined3d_shader_lconst
*constant
;
3491 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3493 if (constant
->idx
== idx
)
3495 return constant
->value
;
3501 static void init_ps_input(const struct wined3d_shader
*shader
,
3502 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3504 static const char * const texcoords
[8] =
3506 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3507 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3510 const struct wined3d_shader_signature_element
*sig
= shader
->input_signature
;
3511 const char *semantic_name
;
3514 switch(args
->super
.vp_mode
)
3516 case pretransformed
:
3518 /* The pixelshader has to collect the varyings on its own. In any case properly load
3519 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3520 * other attribs to 0.0.
3522 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3523 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3524 * load the texcoord attrib pointers to match the pixel shader signature
3526 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3528 semantic_name
= sig
[i
].semantic_name
;
3529 semantic_idx
= sig
[i
].semantic_idx
;
3530 if (!semantic_name
) continue;
3532 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3534 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.color.primary";
3535 else if(semantic_idx
== 1) priv
->ps_input
[i
] = "fragment.color.secondary";
3536 else priv
->ps_input
[i
] = "0.0";
3538 else if(args
->super
.vp_mode
== fixedfunction
)
3540 priv
->ps_input
[i
] = "0.0";
3542 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3544 if(semantic_idx
< 8) priv
->ps_input
[i
] = texcoords
[semantic_idx
];
3545 else priv
->ps_input
[i
] = "0.0";
3547 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3549 if (!semantic_idx
) priv
->ps_input
[i
] = "fragment.fogcoord";
3550 else priv
->ps_input
[i
] = "0.0";
3554 priv
->ps_input
[i
] = "0.0";
3557 TRACE("v%u, semantic %s%u is %s\n", i
, semantic_name
, semantic_idx
, priv
->ps_input
[i
]);
3562 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3565 for(i
= 0; i
< 8; i
++)
3567 priv
->ps_input
[i
] = texcoords
[i
];
3569 priv
->ps_input
[8] = "fragment.color.primary";
3570 priv
->ps_input
[9] = "fragment.color.secondary";
3575 static void arbfp_add_linear_fog(struct wined3d_shader_buffer
*buffer
,
3576 const char *fragcolor
, const char *tmp
)
3578 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3579 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3580 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3583 /* Context activation is done by the caller. */
3584 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3585 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
3586 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3588 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3589 const DWORD
*function
= shader
->function
;
3592 DWORD next_local
= 0;
3593 struct shader_arb_ctx_priv priv_ctx
;
3594 BOOL dcl_td
= FALSE
;
3595 BOOL want_nv_prog
= FALSE
;
3596 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3599 BOOL custom_linear_fog
= FALSE
;
3602 unsigned int i
, found
= 0;
3604 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3607 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3608 || (reg_maps
->shader_version
.major
< 2 && !i
))
3611 sprintf(srgbtmp
[found
], "R%u", i
);
3613 if (found
== 4) break;
3618 sprintf(srgbtmp
[0], "TA");
3619 sprintf(srgbtmp
[1], "TB");
3620 sprintf(srgbtmp
[2], "TC");
3621 sprintf(srgbtmp
[3], "TD");
3625 sprintf(srgbtmp
[1], "TA");
3626 sprintf(srgbtmp
[2], "TB");
3627 sprintf(srgbtmp
[3], "TC");
3630 sprintf(srgbtmp
[2], "TA");
3631 sprintf(srgbtmp
[3], "TB");
3634 sprintf(srgbtmp
[3], "TA");
3640 /* Create the hw ARB shader */
3641 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3642 priv_ctx
.cur_ps_args
= args
;
3643 priv_ctx
.compiled_fprog
= compiled
;
3644 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3645 init_ps_input(shader
, args
, &priv_ctx
);
3646 list_init(&priv_ctx
.control_frames
);
3647 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3649 /* Avoid enabling NV_fragment_program* if we do not need it.
3651 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3652 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3653 * is faster than what we gain from using higher native instructions. There are some things though
3654 * that cannot be emulated. In that case enable the extensions.
3655 * If the extension is enabled, instruction handlers that support both ways will use it.
3657 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3658 * So enable the best we can get.
3660 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3661 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3663 want_nv_prog
= TRUE
;
3666 shader_addline(buffer
, "!!ARBfp1.0\n");
3667 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3669 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3670 priv_ctx
.target_version
= NV3
;
3672 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3674 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3675 priv_ctx
.target_version
= NV2
;
3679 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3682 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3685 priv_ctx
.target_version
= ARB
;
3688 if (reg_maps
->rt_mask
> 1)
3690 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3693 if (reg_maps
->shader_version
.major
< 3)
3695 switch (args
->super
.fog
)
3697 case WINED3D_FFP_PS_FOG_OFF
:
3699 case WINED3D_FFP_PS_FOG_LINEAR
:
3700 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3702 custom_linear_fog
= TRUE
;
3703 priv_ctx
.ps_post_process
= TRUE
;
3706 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3708 case WINED3D_FFP_PS_FOG_EXP
:
3709 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3711 case WINED3D_FFP_PS_FOG_EXP2
:
3712 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3717 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3718 * unused temps away(but occupies them for the whole shader if they're used once). Always
3719 * declaring them avoids tricky bookkeeping work
3721 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3722 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3723 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3724 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3725 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3726 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3727 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %1.10f, 0.0 };\n", eps
);
3729 if (reg_maps
->shader_version
.major
< 2)
3731 strcpy(fragcolor
, "R0");
3735 if (priv_ctx
.ps_post_process
)
3737 if (shader
->u
.ps
.color0_mov
)
3739 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3743 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3744 strcpy(fragcolor
, "TMP_COLOR");
3747 strcpy(fragcolor
, "result.color");
3751 if(args
->super
.srgb_correction
) {
3752 shader_addline(buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
3753 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
3754 shader_addline(buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
3755 srgb_sub_high
, 0.0, 0.0, 0.0);
3758 /* Base Declarations */
3759 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3761 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3763 unsigned char bump_const
;
3765 if (!(map
& 1)) continue;
3767 bump_const
= compiled
->numbumpenvmatconsts
;
3768 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3769 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3770 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3771 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3773 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3774 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3775 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3776 * textures due to conditional NP2 restrictions)
3778 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3779 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3780 * their location is shader dependent anyway and they cannot be loaded globally.
3782 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3783 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3784 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3785 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3787 if (!(reg_maps
->luminanceparams
& (1 << i
))) continue;
3789 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3790 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3791 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3794 for(i
= 0; i
< MAX_CONST_I
; i
++)
3796 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3797 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3799 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3803 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3804 control_values
[0], control_values
[1], control_values
[2]);
3808 compiled
->int_consts
[i
] = next_local
;
3809 compiled
->num_int_consts
++;
3810 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3815 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3817 compiled
->ycorrection
= next_local
;
3818 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3822 shader_addline(buffer
, "TEMP vpos;\n");
3823 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3824 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3825 * ycorrection.z: 1.0
3826 * ycorrection.w: 0.0
3828 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3829 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3834 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3837 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3838 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3839 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3840 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3841 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3842 * shader compilation errors and the subsequent errors when drawing with this shader. */
3843 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3844 unsigned char cur_fixup_sampler
= 0;
3846 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3847 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3848 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3850 fixup
->offset
= next_local
;
3851 fixup
->super
.active
= 0;
3853 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3854 if (!(map
& (1 << i
))) continue;
3856 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
) {
3857 fixup
->super
.active
|= (1 << i
);
3858 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3860 FIXME("No free constant found to load NP2 fixup data into shader. "
3861 "Sampling from this texture will probably look wrong.\n");
3866 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3867 if (fixup
->super
.num_consts
) {
3868 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3869 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3873 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3875 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3878 /* Base Shader Body */
3879 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
3881 if(args
->super
.srgb_correction
) {
3882 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3883 priv_ctx
.target_version
>= NV2
);
3886 if (custom_linear_fog
)
3887 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3889 if(strcmp(fragcolor
, "result.color")) {
3890 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3892 shader_addline(buffer
, "END\n");
3894 /* TODO: change to resource.glObjectHandle or something like that */
3895 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3897 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3898 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3900 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3901 /* Create the program and check for errors */
3902 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3903 buffer
->content_size
, buffer
->buffer
));
3904 checkGLcall("glProgramStringARB()");
3906 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
3909 FIXME("HW PixelShader Error at position %d: %s\n\n",
3910 errPos
, debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3911 shader_arb_dump_program_source(buffer
->buffer
);
3918 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3919 checkGLcall("glGetProgramivARB()");
3920 if (!native
) WARN("Program exceeds native resource limits.\n");
3926 static int compare_sig(const struct wined3d_shader_signature_element
*sig1
, const struct wined3d_shader_signature_element
*sig2
)
3931 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3933 if (!sig1
[i
].semantic_name
|| !sig2
[i
].semantic_name
)
3935 /* Compare pointers, not contents. One string is NULL(element does not exist), the other one is not NULL */
3936 if(sig1
[i
].semantic_name
!= sig2
[i
].semantic_name
) return sig1
[i
].semantic_name
< sig2
[i
].semantic_name
? -1 : 1;
3940 if ((ret
= strcmp(sig1
[i
].semantic_name
, sig2
[i
].semantic_name
))) return ret
;
3941 if(sig1
[i
].semantic_idx
!= sig2
[i
].semantic_idx
) return sig1
[i
].semantic_idx
< sig2
[i
].semantic_idx
? -1 : 1;
3942 if(sig1
[i
].sysval_semantic
!= sig2
[i
].sysval_semantic
) return sig1
[i
].sysval_semantic
< sig2
[i
].sysval_semantic
? -1 : 1;
3943 if(sig1
[i
].component_type
!= sig2
[i
].component_type
) return sig1
[i
].component_type
< sig2
[i
].component_type
? -1 : 1;
3944 if(sig1
[i
].register_idx
!= sig2
[i
].register_idx
) return sig1
[i
].register_idx
< sig2
[i
].register_idx
? -1 : 1;
3945 if(sig1
[i
].mask
!= sig2
[i
].mask
) return sig1
[i
].mask
< sig2
[i
].mask
? -1 : 1;
3950 static struct wined3d_shader_signature_element
*clone_sig(const struct wined3d_shader_signature_element
*sig
)
3952 struct wined3d_shader_signature_element
*new;
3956 new = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*new) * MAX_REG_INPUT
);
3957 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3959 if (!sig
[i
].semantic_name
) continue;
3962 /* Clone the semantic string */
3963 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
[i
].semantic_name
) + 1);
3964 strcpy(name
, sig
[i
].semantic_name
);
3965 new[i
].semantic_name
= name
;
3970 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature_element
*sig
)
3972 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3973 struct ps_signature
*found_sig
;
3977 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3978 TRACE("Found existing signature %u\n", found_sig
->idx
);
3979 return found_sig
->idx
;
3981 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*sig
));
3982 found_sig
->sig
= clone_sig(sig
);
3983 found_sig
->idx
= priv
->ps_sig_number
++;
3984 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3985 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3987 ERR("Failed to insert program entry.\n");
3989 return found_sig
->idx
;
3992 static void init_output_registers(const struct wined3d_shader
*shader
,
3993 const struct wined3d_shader_signature_element
*ps_input_sig
,
3994 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3997 static const char * const texcoords
[8] =
3999 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4000 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
4002 const char *semantic_name
;
4003 DWORD semantic_idx
, reg_idx
;
4005 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
4006 * and varying 9 to result.color.secondary
4008 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
4010 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
4011 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
4012 "result.color.primary", "result.color.secondary"
4017 TRACE("Pixel shader uses builtin varyings\n");
4018 /* Map builtins to builtins */
4019 for(i
= 0; i
< 8; i
++)
4021 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
4023 priv_ctx
->color_output
[0] = "result.color.primary";
4024 priv_ctx
->color_output
[1] = "result.color.secondary";
4025 priv_ctx
->fog_output
= "TMP_FOGCOORD";
4027 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
4028 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
4030 semantic_name
= shader
->output_signature
[i
].semantic_name
;
4031 if (!semantic_name
) continue;
4033 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
))
4035 TRACE("o%u is TMP_OUT\n", i
);
4036 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "TMP_OUT";
4037 else priv_ctx
->vs_output
[i
] = "TA";
4039 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
4041 TRACE("o%u is result.pointsize\n", i
);
4042 if (!shader
->output_signature
[i
].semantic_idx
) priv_ctx
->vs_output
[i
] = "result.pointsize";
4043 else priv_ctx
->vs_output
[i
] = "TA";
4045 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4047 TRACE("o%u is result.color.?, idx %u\n", i
, shader
->output_signature
[i
].semantic_idx
);
4048 if (!shader
->output_signature
[i
].semantic_idx
)
4049 priv_ctx
->vs_output
[i
] = "result.color.primary";
4050 else if (shader
->output_signature
[i
].semantic_idx
== 1)
4051 priv_ctx
->vs_output
[i
] = "result.color.secondary";
4052 else priv_ctx
->vs_output
[i
] = "TA";
4054 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4056 TRACE("o%u is %s\n", i
, texcoords
[shader
->output_signature
[i
].semantic_idx
]);
4057 if (shader
->output_signature
[i
].semantic_idx
>= 8) priv_ctx
->vs_output
[i
] = "TA";
4058 else priv_ctx
->vs_output
[i
] = texcoords
[shader
->output_signature
[i
].semantic_idx
];
4060 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4062 TRACE("o%u is result.fogcoord\n", i
);
4063 if (shader
->output_signature
[i
].semantic_idx
> 0) priv_ctx
->vs_output
[i
] = "TA";
4064 else priv_ctx
->vs_output
[i
] = "result.fogcoord";
4068 priv_ctx
->vs_output
[i
] = "TA";
4074 TRACE("Pixel shader uses declared varyings\n");
4076 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4077 for(i
= 0; i
< 8; i
++)
4079 priv_ctx
->texcrd_output
[i
] = "TA";
4081 priv_ctx
->color_output
[0] = "TA";
4082 priv_ctx
->color_output
[1] = "TA";
4083 priv_ctx
->fog_output
= "TA";
4085 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4087 semantic_name
= ps_input_sig
[i
].semantic_name
;
4088 semantic_idx
= ps_input_sig
[i
].semantic_idx
;
4089 reg_idx
= ps_input_sig
[i
].register_idx
;
4090 if (!semantic_name
) continue;
4092 /* If a declared input register is not written by builtin arguments, don't write to it.
4093 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4095 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4096 * to TMP_OUT in any case
4098 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4100 if (semantic_idx
< 8)
4101 priv_ctx
->texcrd_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4103 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4105 if (semantic_idx
< 2)
4106 priv_ctx
->color_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4108 else if(shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
4111 priv_ctx
->fog_output
= decl_idx_to_string
[reg_idx
];
4118 if (!strcmp(decl_idx_to_string
[reg_idx
], "result.color.primary")
4119 || !strcmp(decl_idx_to_string
[reg_idx
], "result.color.secondary"))
4121 compiled
->need_color_unclamp
= TRUE
;
4125 /* Map declared to declared */
4126 for (i
= 0; i
< (sizeof(shader
->output_signature
) / sizeof(*shader
->output_signature
)); ++i
)
4128 /* Write unread output to TA to throw them away */
4129 priv_ctx
->vs_output
[i
] = "TA";
4130 semantic_name
= shader
->output_signature
[i
].semantic_name
;
4131 if (!semantic_name
) continue;
4133 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_POSITION
)
4134 && !shader
->output_signature
[i
].semantic_idx
)
4136 priv_ctx
->vs_output
[i
] = "TMP_OUT";
4139 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_PSIZE
)
4140 && !shader
->output_signature
[i
].semantic_idx
)
4142 priv_ctx
->vs_output
[i
] = "result.pointsize";
4146 for(j
= 0; j
< MAX_REG_INPUT
; j
++)
4148 if (!ps_input_sig
[j
].semantic_name
) continue;
4150 if (!strcmp(ps_input_sig
[j
].semantic_name
, semantic_name
)
4151 && ps_input_sig
[j
].semantic_idx
== shader
->output_signature
[i
].semantic_idx
)
4153 priv_ctx
->vs_output
[i
] = decl_idx_to_string
[ps_input_sig
[j
].register_idx
];
4155 if (!strcmp(priv_ctx
->vs_output
[i
], "result.color.primary")
4156 || !strcmp(priv_ctx
->vs_output
[i
], "result.color.secondary"))
4158 compiled
->need_color_unclamp
= TRUE
;
4165 /* Context activation is done by the caller. */
4166 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4167 const struct wined3d_gl_info
*gl_info
, struct wined3d_shader_buffer
*buffer
,
4168 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4169 const struct wined3d_shader_signature_element
*ps_input_sig
)
4171 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4172 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4173 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4174 const DWORD
*function
= shader
->function
;
4176 DWORD next_local
= 0;
4177 struct shader_arb_ctx_priv priv_ctx
;
4181 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4182 priv_ctx
.cur_vs_args
= args
;
4183 list_init(&priv_ctx
.control_frames
);
4184 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4186 /* Create the hw ARB shader */
4187 shader_addline(buffer
, "!!ARBvp1.0\n");
4189 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4190 * mesurable performance penalty, and we can always make use of it for clipplanes.
4192 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4194 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4195 priv_ctx
.target_version
= NV3
;
4196 shader_addline(buffer
, "ADDRESS aL;\n");
4198 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4200 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4201 priv_ctx
.target_version
= NV2
;
4202 shader_addline(buffer
, "ADDRESS aL;\n");
4204 priv_ctx
.target_version
= ARB
;
4207 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4209 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4210 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4212 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %1.10f};\n", eps
);
4214 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4216 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4217 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4220 shader_addline(buffer
, "TEMP TA;\n");
4221 shader_addline(buffer
, "TEMP TB;\n");
4223 /* Base Declarations */
4224 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4225 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4227 for(i
= 0; i
< MAX_CONST_I
; i
++)
4229 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4230 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
4232 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4236 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4237 control_values
[0], control_values
[1], control_values
[2]);
4241 compiled
->int_consts
[i
] = next_local
;
4242 compiled
->num_int_consts
++;
4243 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4248 /* We need a constant to fixup the final position */
4249 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4250 compiled
->pos_fixup
= next_local
++;
4252 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4253 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4254 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4255 * a replacement shader depend on the texcoord.w being set properly.
4257 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4258 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4259 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4260 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4261 * this can eat a number of instructions, so skip it unless this cap is set as well
4263 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4265 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4266 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4268 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4271 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4272 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++)
4274 if (reg_maps
->texcoord_mask
[i
] && reg_maps
->texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4275 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4280 /* The shader starts with the main function */
4281 priv_ctx
.in_main_func
= TRUE
;
4282 /* Base Shader Body */
4283 shader_generate_main(shader
, buffer
, reg_maps
, function
, &priv_ctx
);
4285 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4286 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4288 shader_addline(buffer
, "END\n");
4290 /* TODO: change to resource.glObjectHandle or something like that */
4291 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4293 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4294 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4296 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4297 /* Create the program and check for errors */
4298 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
4299 buffer
->content_size
, buffer
->buffer
));
4300 checkGLcall("glProgramStringARB()");
4302 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
4305 FIXME("HW VertexShader Error at position %d: %s\n\n",
4306 errPos
, debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
4307 shader_arb_dump_program_source(buffer
->buffer
);
4314 GL_EXTCALL(glGetProgramivARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
4315 checkGLcall("glGetProgramivARB()");
4316 if (!native
) WARN("Program exceeds native resource limits.\n");
4322 /* Context activation is done by the caller. */
4323 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_shader
*shader
,
4324 const struct arb_ps_compile_args
*args
)
4326 struct wined3d_device
*device
= shader
->device
;
4327 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4328 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4331 struct arb_ps_compiled_shader
*new_array
;
4332 struct wined3d_shader_buffer buffer
;
4333 struct arb_pshader_private
*shader_data
;
4336 if (!shader
->backend_data
)
4338 struct shader_arb_priv
*priv
= device
->shader_priv
;
4340 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4341 shader_data
= shader
->backend_data
;
4342 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4344 if (shader
->reg_maps
.shader_version
.major
< 3)
4345 shader_data
->input_signature_idx
= ~0U;
4347 shader_data
->input_signature_idx
= find_input_signature(priv
, shader
->input_signature
);
4349 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4351 if (!d3d_info
->vs_clipping
)
4352 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4353 d3d_info
->limits
.ffp_blend_stages
- 1);
4355 shader_data
->clipplane_emulation
= ~0U;
4357 shader_data
= shader
->backend_data
;
4359 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4360 * so a linear search is more performant than a hashmap or a binary search
4361 * (cache coherency etc)
4363 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4365 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4366 return &shader_data
->gl_shaders
[i
];
4369 TRACE("No matching GL shader found, compiling a new shader\n");
4370 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4371 if (shader_data
->num_gl_shaders
)
4373 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4374 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4375 new_size
* sizeof(*shader_data
->gl_shaders
));
4377 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4382 ERR("Out of memory\n");
4385 shader_data
->gl_shaders
= new_array
;
4386 shader_data
->shader_array_size
= new_size
;
4389 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4391 pixelshader_update_samplers(shader
, args
->super
.tex_types
);
4393 if (!shader_buffer_init(&buffer
))
4395 ERR("Failed to initialize shader buffer.\n");
4399 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4400 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4401 shader_buffer_free(&buffer
);
4402 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4404 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4407 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4408 const DWORD use_map
, BOOL skip_int
) {
4409 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4410 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4411 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4412 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4413 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4414 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4415 if(skip_int
) return TRUE
;
4417 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4420 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4421 const struct arb_vs_compile_args
*args
,
4422 const struct wined3d_shader_signature_element
*ps_input_sig
)
4424 struct wined3d_device
*device
= shader
->device
;
4425 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4426 DWORD use_map
= device
->stream_info
.use_map
;
4429 struct arb_vs_compiled_shader
*new_array
;
4430 struct wined3d_shader_buffer buffer
;
4431 struct arb_vshader_private
*shader_data
;
4434 if (!shader
->backend_data
)
4436 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4438 shader
->backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4439 shader_data
= shader
->backend_data
;
4441 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4442 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4444 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4446 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4447 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4448 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4450 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4451 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4452 else if (reg_maps
->max_rel_offset
> 63)
4453 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4456 shader_data
= shader
->backend_data
;
4458 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4459 * so a linear search is more performant than a hashmap or a binary search
4460 * (cache coherency etc)
4462 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4463 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4464 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4466 return &shader_data
->gl_shaders
[i
];
4470 TRACE("No matching GL shader found, compiling a new shader\n");
4472 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4473 if (shader_data
->num_gl_shaders
)
4475 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4476 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4477 new_size
* sizeof(*shader_data
->gl_shaders
));
4479 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4484 ERR("Out of memory\n");
4487 shader_data
->gl_shaders
= new_array
;
4488 shader_data
->shader_array_size
= new_size
;
4491 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4493 if (!shader_buffer_init(&buffer
))
4495 ERR("Failed to initialize shader buffer.\n");
4499 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4500 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4502 shader_buffer_free(&buffer
);
4503 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4505 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4508 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4509 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4510 struct arb_ps_compile_args
*args
)
4512 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4513 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4517 find_ps_compile_args(state
, shader
, &args
->super
);
4519 /* This forces all local boolean constants to 1 to make them stateblock independent */
4520 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4522 for(i
= 0; i
< MAX_CONST_B
; i
++)
4524 if (state
->ps_consts_b
[i
])
4525 args
->bools
|= ( 1 << i
);
4528 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4529 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4530 * duplicate the shader than have a no-op KIL instruction in every shader
4532 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4533 && state
->render_states
[WINED3D_RS_CLIPPING
]
4534 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4539 /* Skip if unused or local, or supported natively */
4540 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4541 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4543 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4547 for(i
= 0; i
< MAX_CONST_I
; i
++)
4549 if(int_skip
& (1 << i
))
4551 args
->loop_ctrl
[i
][0] = 0;
4552 args
->loop_ctrl
[i
][1] = 0;
4553 args
->loop_ctrl
[i
][2] = 0;
4557 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
* 4];
4558 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
* 4 + 1];
4559 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
* 4 + 2];
4564 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4565 const struct wined3d_context
*context
, const struct wined3d_shader
*shader
,
4566 struct arb_vs_compile_args
*args
)
4568 struct wined3d_device
*device
= shader
->device
;
4569 const struct wined3d_adapter
*adapter
= device
->adapter
;
4570 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4571 const struct wined3d_d3d_info
*d3d_info
= context
->d3d_info
;
4575 find_vs_compile_args(state
, shader
, &args
->super
);
4577 args
->clip
.boolclip_compare
= 0;
4580 const struct wined3d_shader
*ps
= state
->pixel_shader
;
4581 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4582 args
->ps_signature
= shader_priv
->input_signature_idx
;
4584 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4588 args
->ps_signature
= ~0;
4589 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4591 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(state
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4593 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4596 if (args
->clip
.boolclip
.clip_texcoord
)
4598 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4599 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4600 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4603 /* This forces all local boolean constants to 1 to make them stateblock independent */
4604 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4605 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4606 for(i
= 0; i
< MAX_CONST_B
; i
++)
4608 if (state
->vs_consts_b
[i
])
4609 args
->clip
.boolclip
.bools
|= ( 1 << i
);
4612 args
->vertex
.samplers
[0] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 0];
4613 args
->vertex
.samplers
[1] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 1];
4614 args
->vertex
.samplers
[2] = device
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 2];
4615 args
->vertex
.samplers
[3] = 0;
4617 /* Skip if unused or local */
4618 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4619 /* This is about flow control, not clipping. */
4620 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4622 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4626 for(i
= 0; i
< MAX_CONST_I
; i
++)
4628 if(int_skip
& (1 << i
))
4630 args
->loop_ctrl
[i
][0] = 0;
4631 args
->loop_ctrl
[i
][1] = 0;
4632 args
->loop_ctrl
[i
][2] = 0;
4636 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
* 4];
4637 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
* 4 + 1];
4638 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
* 4 + 2];
4643 /* Context activation is done by the caller. */
4644 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4645 const struct wined3d_state
*state
)
4647 struct shader_arb_priv
*priv
= shader_priv
;
4648 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4651 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4654 struct wined3d_shader
*ps
= state
->pixel_shader
;
4655 struct arb_ps_compile_args compile_args
;
4656 struct arb_ps_compiled_shader
*compiled
;
4658 TRACE("Using pixel shader %p.\n", ps
);
4659 find_arb_ps_compile_args(state
, context
, ps
, &compile_args
);
4660 compiled
= find_arb_pshader(ps
, &compile_args
);
4661 priv
->current_fprogram_id
= compiled
->prgId
;
4662 priv
->compiled_fprog
= compiled
;
4664 /* Bind the fragment program */
4665 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4666 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4668 if (!priv
->use_arbfp_fixed_func
)
4669 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4671 /* Enable OpenGL fragment programs. */
4672 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4673 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4675 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4677 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4678 * a 1.x and newer shader, reload the first 8 constants
4680 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4682 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4683 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4684 for(i
= 0; i
< 8; i
++)
4686 priv
->pshader_const_dirty
[i
] = 1;
4688 /* Also takes care of loading local constants */
4689 shader_arb_load_constants_internal(shader_priv
, context
, state
, TRUE
, FALSE
, TRUE
);
4693 UINT rt_height
= state
->fb
->render_targets
[0]->resource
.height
;
4694 shader_arb_ps_local_constants(compiled
, context
, state
, rt_height
);
4697 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4698 if (compiled
->np2fixup_info
.super
.active
)
4699 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4701 if (ps
->load_local_constsF
)
4702 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4706 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4708 /* Disable only if we're not using arbfp fixed function fragment
4709 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4710 * enabled, and the fixed function pipeline will bind the fixed
4711 * function replacement shader. */
4712 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4713 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4714 priv
->current_fprogram_id
= 0;
4716 priv
->fragment_pipe
->enable_extension(gl_info
, TRUE
);
4721 struct wined3d_shader
*vs
= state
->vertex_shader
;
4722 struct arb_vs_compile_args compile_args
;
4723 struct arb_vs_compiled_shader
*compiled
;
4724 const struct wined3d_shader_signature_element
*ps_input_sig
;
4726 TRACE("Using vertex shader %p\n", vs
);
4727 find_arb_vs_compile_args(state
, context
, vs
, &compile_args
);
4729 /* Instead of searching for the signature in the signature list, read the one from the
4730 * current pixel shader. It's maybe not the shader where the signature came from, but it
4731 * is the same signature and faster to find. */
4732 if (compile_args
.ps_signature
== ~0U)
4733 ps_input_sig
= NULL
;
4735 ps_input_sig
= state
->pixel_shader
->input_signature
;
4737 compiled
= find_arb_vshader(vs
, &compile_args
, ps_input_sig
);
4738 priv
->current_vprogram_id
= compiled
->prgId
;
4739 priv
->compiled_vprog
= compiled
;
4741 /* Bind the vertex program */
4742 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4743 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4745 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4747 /* Enable OpenGL vertex programs */
4748 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4749 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4750 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4751 shader_arb_vs_local_constants(compiled
, context
, state
);
4753 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4754 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4756 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4758 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4759 checkGLcall("glClampColorARB");
4761 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4765 if (vs
->load_local_constsF
)
4766 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4770 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4772 priv
->current_vprogram_id
= 0;
4773 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4774 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4776 priv
->vertex_pipe
->vp_enable(gl_info
, TRUE
);
4781 /* Context activation is done by the caller. */
4782 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4784 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4785 struct shader_arb_priv
*priv
= shader_priv
;
4787 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4789 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4790 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4791 priv
->current_fprogram_id
= 0;
4793 priv
->fragment_pipe
->enable_extension(gl_info
, FALSE
);
4795 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4797 priv
->current_vprogram_id
= 0;
4798 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4799 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4801 priv
->vertex_pipe
->vp_enable(gl_info
, FALSE
);
4803 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4805 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4806 checkGLcall("glClampColorARB");
4807 priv
->last_vs_color_unclamp
= FALSE
;
4810 context
->shader_update_mask
= (1 << WINED3D_SHADER_TYPE_PIXEL
)
4811 | (1 << WINED3D_SHADER_TYPE_VERTEX
)
4812 | (1 << WINED3D_SHADER_TYPE_GEOMETRY
);
4815 /* Context activation is done by the caller. */
4816 static void shader_arb_select_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
,
4817 enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4819 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4820 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4821 struct shader_arb_priv
*priv
= shader_priv
;
4822 GLuint
*blt_fprogram
;
4824 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4825 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4826 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4828 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4829 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4830 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4831 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4832 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4835 /* Context activation is done by the caller. */
4836 static void shader_arb_deselect_depth_blt(void *shader_priv
, const struct wined3d_gl_info
*gl_info
)
4838 struct shader_arb_priv
*priv
= shader_priv
;
4840 if (priv
->current_vprogram_id
) {
4841 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4842 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4844 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB.\n", priv
->current_vprogram_id
);
4848 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4849 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4852 if (priv
->current_fprogram_id
) {
4853 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4854 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4856 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", priv
->current_fprogram_id
);
4858 else if(!priv
->use_arbfp_fixed_func
)
4860 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4861 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4865 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4867 struct wined3d_device
*device
= shader
->device
;
4868 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4870 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4872 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4875 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4877 if (shader_data
->num_gl_shaders
)
4879 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4881 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4883 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4884 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4887 context_release(context
);
4890 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4891 HeapFree(GetProcessHeap(), 0, shader_data
);
4892 shader
->backend_data
= NULL
;
4896 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4899 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4901 if (shader_data
->num_gl_shaders
)
4903 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4905 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4907 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4908 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4911 context_release(context
);
4914 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4915 HeapFree(GetProcessHeap(), 0, shader_data
);
4916 shader
->backend_data
= NULL
;
4920 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4922 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4923 return compare_sig(key
, e
->sig
);
4926 static const struct wine_rb_functions sig_tree_functions
=
4934 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4935 const struct fragment_pipeline
*fragment_pipe
)
4937 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4938 struct fragment_caps fragment_caps
;
4939 void *vertex_priv
, *fragment_priv
;
4940 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4942 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4944 ERR("Failed to initialize vertex pipe.\n");
4945 HeapFree(GetProcessHeap(), 0, priv
);
4949 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4951 ERR("Failed to initialize fragment pipe.\n");
4952 vertex_pipe
->vp_free(device
);
4953 HeapFree(GetProcessHeap(), 0, priv
);
4957 priv
->vshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4958 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4959 if (!priv
->vshader_const_dirty
)
4961 memset(priv
->vshader_const_dirty
, 1,
4962 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4964 priv
->pshader_const_dirty
= HeapAlloc(GetProcessHeap(), 0,
4965 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4966 if (!priv
->pshader_const_dirty
)
4968 memset(priv
->pshader_const_dirty
, 1,
4969 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4971 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4973 ERR("RB tree init failed\n");
4977 priv
->vertex_pipe
= vertex_pipe
;
4978 priv
->fragment_pipe
= fragment_pipe
;
4979 fragment_pipe
->get_caps(&device
->adapter
->gl_info
, &fragment_caps
);
4980 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4982 device
->vertex_priv
= vertex_priv
;
4983 device
->fragment_priv
= fragment_priv
;
4984 device
->shader_priv
= priv
;
4989 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
4990 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
4991 fragment_pipe
->free_private(device
);
4992 vertex_pipe
->vp_free(device
);
4993 HeapFree(GetProcessHeap(), 0, priv
);
4994 return E_OUTOFMEMORY
;
4997 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4999 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
5001 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
5003 HeapFree(GetProcessHeap(), 0, (char *) sig
->sig
[i
].semantic_name
);
5005 HeapFree(GetProcessHeap(), 0, sig
->sig
);
5006 HeapFree(GetProcessHeap(), 0, sig
);
5009 /* Context activation is done by the caller. */
5010 static void shader_arb_free(struct wined3d_device
*device
)
5012 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
5013 struct shader_arb_priv
*priv
= device
->shader_priv
;
5016 if (priv
->depth_blt_vprogram_id
)
5017 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
5019 for (i
= 0; i
< tex_type_count
; ++i
)
5021 if (priv
->depth_blt_fprogram_id_full
[i
])
5023 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
5025 if (priv
->depth_blt_fprogram_id_masked
[i
])
5027 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
5031 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
5032 HeapFree(GetProcessHeap(), 0, priv
->pshader_const_dirty
);
5033 HeapFree(GetProcessHeap(), 0, priv
->vshader_const_dirty
);
5034 priv
->fragment_pipe
->free_private(device
);
5035 priv
->vertex_pipe
->vp_free(device
);
5036 HeapFree(GetProcessHeap(), 0, device
->shader_priv
);
5039 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
5044 static void shader_arb_free_context_data(struct wined3d_context
*context
)
5046 struct shader_arb_priv
*priv
= context
->swapchain
->device
->shader_priv
;
5048 if (priv
->last_context
== context
)
5049 priv
->last_context
= NULL
;
5052 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*caps
)
5054 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
5059 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
5060 * for vertex programs. If the native limit is less than that it's
5061 * not very useful, and e.g. Mesa swrast returns 0, probably to
5062 * indicate it's a software implementation. */
5063 if (gl_info
->limits
.arb_vs_native_constants
< 96)
5064 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
5066 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
5068 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
5071 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
5073 else if (vs_consts
>= 256)
5075 /* Shader Model 2.0 requires at least 256 vertex shader constants */
5077 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
5082 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
5084 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
5085 caps
->vs_uniform_count
= vs_consts
;
5089 caps
->vs_version
= 0;
5090 caps
->vs_uniform_count
= 0;
5093 caps
->gs_version
= 0;
5095 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
5100 /* Similar as above for vertex programs, but the minimum for fragment
5101 * programs is 24. */
5102 if (gl_info
->limits
.arb_ps_native_constants
< 24)
5103 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
5105 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
5107 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
5110 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
5112 else if (ps_consts
>= 32)
5114 /* Shader Model 2.0 requires at least 32 pixel shader constants */
5116 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
5121 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
5123 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
5124 caps
->ps_uniform_count
= ps_consts
;
5125 caps
->ps_1x_max_value
= 8.0f
;
5129 caps
->ps_version
= 0;
5130 caps
->ps_uniform_count
= 0;
5131 caps
->ps_1x_max_value
= 0.0f
;
5134 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
5135 if (use_nv_clip(gl_info
))
5136 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
5139 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
5141 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
5143 TRACE("Checking support for color_fixup:\n");
5144 dump_color_fixup_desc(fixup
);
5147 /* We support everything except complex conversions. */
5148 if (!is_complex_fixup(fixup
))
5154 TRACE("[FAILED]\n");
5158 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
5160 char write_mask
[20], regstr
[50];
5161 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5162 BOOL is_color
= FALSE
;
5163 const struct wined3d_shader_dst_param
*dst
;
5165 if (!ins
->dst_count
) return;
5169 if (!shift
) return; /* Saturate alone is handled by the instructions */
5171 shader_arb_get_write_mask(ins
, dst
, write_mask
);
5172 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
5174 /* Generate a line that does the output modifier computation
5175 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
5176 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
5178 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
5179 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
5182 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5184 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5185 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5186 /* WINED3DSIH_AND */ NULL
,
5187 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5188 /* WINED3DSIH_BREAK */ shader_hw_break
,
5189 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5190 /* WINED3DSIH_BREAKP */ NULL
,
5191 /* WINED3DSIH_CALL */ shader_hw_call
,
5192 /* WINED3DSIH_CALLNZ */ NULL
,
5193 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5194 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5195 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5196 /* WINED3DSIH_CUT */ NULL
,
5197 /* WINED3DSIH_DCL */ shader_hw_nop
,
5198 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5199 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5200 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5201 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5202 /* WINED3DSIH_DEF */ shader_hw_nop
,
5203 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5204 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5205 /* WINED3DSIH_DIV */ NULL
,
5206 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5207 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5208 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5209 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5210 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5211 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5212 /* WINED3DSIH_ELSE */ shader_hw_else
,
5213 /* WINED3DSIH_EMIT */ NULL
,
5214 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5215 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5216 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5217 /* WINED3DSIH_EQ */ NULL
,
5218 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5219 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5220 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5221 /* WINED3DSIH_FTOI */ NULL
,
5222 /* WINED3DSIH_GE */ NULL
,
5223 /* WINED3DSIH_IADD */ NULL
,
5224 /* WINED3DSIH_IEQ */ NULL
,
5225 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5226 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5227 /* WINED3DSIH_IGE */ NULL
,
5228 /* WINED3DSIH_IMUL */ NULL
,
5229 /* WINED3DSIH_ITOF */ NULL
,
5230 /* WINED3DSIH_LABEL */ shader_hw_label
,
5231 /* WINED3DSIH_LD */ NULL
,
5232 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5233 /* WINED3DSIH_LOG */ shader_hw_log
,
5234 /* WINED3DSIH_LOGP */ shader_hw_log
,
5235 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5236 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5237 /* WINED3DSIH_LT */ NULL
,
5238 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5239 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5240 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5241 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5242 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5243 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5244 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5245 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5246 /* WINED3DSIH_MOV */ shader_hw_mov
,
5247 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5248 /* WINED3DSIH_MOVC */ NULL
,
5249 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5250 /* WINED3DSIH_NOP */ shader_hw_nop
,
5251 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5252 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5253 /* WINED3DSIH_POW */ shader_hw_pow
,
5254 /* WINED3DSIH_RCP */ shader_hw_rcp
,
5255 /* WINED3DSIH_REP */ shader_hw_rep
,
5256 /* WINED3DSIH_RET */ shader_hw_ret
,
5257 /* WINED3DSIH_ROUND_NI */ NULL
,
5258 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5259 /* WINED3DSIH_SAMPLE */ NULL
,
5260 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5261 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5262 /* WINED3DSIH_SETP */ NULL
,
5263 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5264 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5265 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5266 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5267 /* WINED3DSIH_SQRT */ NULL
,
5268 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5269 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5270 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5271 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5272 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5273 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5274 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5275 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5276 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5277 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5278 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5279 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5280 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5281 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5282 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5283 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5284 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5285 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5286 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5287 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5288 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5289 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5290 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5291 /* WINED3DSIH_UDIV */ NULL
,
5292 /* WINED3DSIH_USHR */ NULL
,
5293 /* WINED3DSIH_UTOF */ NULL
,
5294 /* WINED3DSIH_XOR */ NULL
,
5297 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5298 const struct wined3d_shader
*shader
, DWORD idx
)
5300 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5301 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5302 const struct wined3d_shader_lconst
*constant
;
5304 WORD flag
= (1 << idx
);
5305 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5307 if (reg_maps
->local_bool_consts
& flag
)
5309 /* What good is a if(bool) with a hardcoded local constant? I don't know, but handle it */
5310 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5312 if (constant
->idx
== idx
)
5314 return constant
->value
[0];
5317 ERR("Local constant not found\n");
5322 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5323 else bools
= priv
->cur_ps_args
->bools
;
5324 return bools
& flag
;
5328 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5329 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5331 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5332 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5334 /* Integer constants can either be a local constant, or they can be stored in the shader
5335 * type specific compile args. */
5336 if (reg_maps
->local_int_consts
& (1 << idx
))
5338 const struct wined3d_shader_lconst
*constant
;
5340 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5342 if (constant
->idx
== idx
)
5344 loop_control
->count
= constant
->value
[0];
5345 loop_control
->start
= constant
->value
[1];
5346 /* Step is signed. */
5347 loop_control
->step
= (int)constant
->value
[2];
5351 /* If this happens the flag was set incorrectly */
5352 ERR("Local constant not found\n");
5353 loop_control
->count
= 0;
5354 loop_control
->start
= 0;
5355 loop_control
->step
= 0;
5359 switch (reg_maps
->shader_version
.type
)
5361 case WINED3D_SHADER_TYPE_VERTEX
:
5362 /* Count and aL start value are unsigned */
5363 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5364 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5365 /* Step is signed. */
5366 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5369 case WINED3D_SHADER_TYPE_PIXEL
:
5370 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5371 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5372 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5376 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5381 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5384 struct wined3d_shader_dst_param
*dst_param
= NULL
;
5385 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
5386 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5389 ERR("Out of memory\n");
5394 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5395 if(!dst_param
) goto free
;
5396 *dst_param
= *ins
->dst
;
5397 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5399 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5402 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5403 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5405 rec
->ins
.dst
= dst_param
;
5407 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5410 for (i
= 0; i
< ins
->src_count
; ++i
)
5412 src_param
[i
] = ins
->src
[i
];
5413 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5415 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5418 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5419 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5422 rec
->ins
.src
= src_param
;
5423 list_add_tail(list
, &rec
->entry
);
5427 ERR("Out of memory\n");
5430 HeapFree(GetProcessHeap(), 0, (void *)dst_param
->reg
.idx
[0].rel_addr
);
5431 HeapFree(GetProcessHeap(), 0, dst_param
);
5435 for(i
= 0; i
< ins
->src_count
; i
++)
5437 HeapFree(GetProcessHeap(), 0, (void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5439 HeapFree(GetProcessHeap(), 0, src_param
);
5441 HeapFree(GetProcessHeap(), 0, rec
);
5444 static void free_recorded_instruction(struct list
*list
)
5446 struct recorded_instruction
*rec_ins
, *entry2
;
5449 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5451 list_remove(&rec_ins
->entry
);
5452 if (rec_ins
->ins
.dst
)
5454 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5455 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.dst
);
5457 if (rec_ins
->ins
.src
)
5459 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5461 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5463 HeapFree(GetProcessHeap(), 0, (void *)rec_ins
->ins
.src
);
5465 HeapFree(GetProcessHeap(), 0, rec_ins
);
5469 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5471 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5472 struct control_frame
*control_frame
;
5474 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5476 struct list
*e
= list_head(&priv
->control_frames
);
5477 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5478 list_remove(&control_frame
->entry
);
5479 HeapFree(GetProcessHeap(), 0, control_frame
);
5482 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5484 /* Non-ifc ENDIFs were already handled previously. */
5485 struct list
*e
= list_head(&priv
->control_frames
);
5486 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5487 list_remove(&control_frame
->entry
);
5488 HeapFree(GetProcessHeap(), 0, control_frame
);
5492 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5493 SHADER_HANDLER hw_fct
;
5494 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5495 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5496 struct control_frame
*control_frame
;
5497 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5500 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5502 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5503 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5505 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5506 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5508 if(priv
->target_version
>= NV2
)
5510 control_frame
->no
.loop
= priv
->num_loops
++;
5515 /* Don't bother recording when we're in a not used if branch */
5521 if(!priv
->recording
)
5523 list_init(&priv
->record
);
5524 priv
->recording
= TRUE
;
5525 control_frame
->outer_loop
= TRUE
;
5526 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5527 return; /* Instruction is handled */
5529 /* Record this loop in the outer loop's recording */
5532 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5534 if(priv
->target_version
>= NV2
)
5536 /* Nothing to do. The control frame is popped after the HW instr handler */
5540 struct list
*e
= list_head(&priv
->control_frames
);
5541 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5542 list_remove(&control_frame
->entry
);
5544 if(control_frame
->outer_loop
)
5546 unsigned int iteration
;
5550 /* Turn off recording before playback */
5551 priv
->recording
= FALSE
;
5553 /* Move the recorded instructions to a separate list and get them out of the private data
5554 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5555 * be recorded again, thus priv->record might be overwritten
5558 list_move_tail(©
, &priv
->record
);
5559 list_init(&priv
->record
);
5561 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5563 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5564 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5565 control_frame
->loop_control
.step
);
5566 aL
= control_frame
->loop_control
.start
;
5570 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5573 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5575 struct recorded_instruction
*rec_ins
;
5576 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5579 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5583 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5586 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5588 shader_arb_handle_instruction(&rec_ins
->ins
);
5591 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5593 aL
+= control_frame
->loop_control
.step
;
5596 shader_addline(buffer
, "#end loop/rep\n");
5598 free_recorded_instruction(©
);
5599 HeapFree(GetProcessHeap(), 0, control_frame
);
5600 return; /* Instruction is handled */
5604 /* This is a nested loop. Proceed to the normal recording function */
5605 HeapFree(GetProcessHeap(), 0, control_frame
);
5612 record_instruction(&priv
->record
, ins
);
5617 if(ins
->handler_idx
== WINED3DSIH_IF
)
5619 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5620 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5621 control_frame
->type
= IF
;
5623 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5624 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5625 bool_const
= !bool_const
;
5626 if (!priv
->muted
&& !bool_const
)
5628 shader_addline(buffer
, "#if(FALSE){\n");
5630 control_frame
->muting
= TRUE
;
5632 else shader_addline(buffer
, "#if(TRUE) {\n");
5634 return; /* Instruction is handled */
5636 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5638 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5639 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5640 control_frame
->type
= IFC
;
5641 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5642 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5644 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5646 struct list
*e
= list_head(&priv
->control_frames
);
5647 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5649 if(control_frame
->type
== IF
)
5651 shader_addline(buffer
, "#} else {\n");
5652 if(!priv
->muted
&& !control_frame
->muting
)
5655 control_frame
->muting
= TRUE
;
5657 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5658 return; /* Instruction is handled. */
5660 /* In case of an ifc, generate a HW shader instruction */
5661 if (control_frame
->type
!= IFC
)
5662 ERR("Control frame does not match.\n");
5664 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5666 struct list
*e
= list_head(&priv
->control_frames
);
5667 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5669 if(control_frame
->type
== IF
)
5671 shader_addline(buffer
, "#} endif\n");
5672 if(control_frame
->muting
) priv
->muted
= FALSE
;
5673 list_remove(&control_frame
->entry
);
5674 HeapFree(GetProcessHeap(), 0, control_frame
);
5675 return; /* Instruction is handled */
5677 /* In case of an ifc, generate a HW shader instruction */
5678 if (control_frame
->type
!= IFC
)
5679 ERR("Control frame does not match.\n");
5684 pop_control_frame(ins
);
5688 /* Select handler */
5689 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5691 /* Unhandled opcode */
5694 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5699 pop_control_frame(ins
);
5701 shader_arb_add_instruction_modifiers(ins
);
5704 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5706 struct shader_arb_priv
*priv
= shader_priv
;
5708 return priv
->ffp_proj_control
;
5711 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5713 shader_arb_handle_instruction
,
5716 shader_arb_select_depth_blt
,
5717 shader_arb_deselect_depth_blt
,
5718 shader_arb_update_float_vertex_constants
,
5719 shader_arb_update_float_pixel_constants
,
5720 shader_arb_load_constants
,
5724 shader_arb_allocate_context_data
,
5725 shader_arb_free_context_data
,
5726 shader_arb_get_caps
,
5727 shader_arb_color_fixup_supported
,
5728 shader_arb_has_ffp_proj_control
,
5731 /* ARB_fragment_program fixed function pipeline replacement definitions */
5732 #define ARB_FFP_CONST_TFACTOR 0
5733 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
5734 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5735 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5736 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5738 struct arbfp_ffp_desc
5740 struct ffp_frag_desc parent
;
5744 /* Context activation is done by the caller. */
5745 static void arbfp_enable(const struct wined3d_gl_info
*gl_info
, BOOL enable
)
5749 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5750 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5754 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5755 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5759 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5761 struct shader_arb_priv
*priv
;
5763 /* Share private data between the shader backend and the pipeline
5764 * replacement, if both are the arb implementation. This is needed to
5765 * figure out whether ARBfp should be disabled if no pixel shader is bound
5767 if (shader_backend
== &arb_program_shader_backend
)
5769 else if (!(priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
))))
5772 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5774 ERR("Failed to initialize rbtree.\n");
5775 if (priv
!= shader_priv
)
5776 HeapFree(GetProcessHeap(), 0, priv
);
5779 priv
->use_arbfp_fixed_func
= TRUE
;
5784 /* Context activation is done by the caller. */
5785 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5787 const struct wined3d_gl_info
*gl_info
= context
;
5788 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5790 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5791 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5792 HeapFree(GetProcessHeap(), 0, entry_arb
);
5795 /* Context activation is done by the caller. */
5796 static void arbfp_free(struct wined3d_device
*device
)
5798 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5800 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &device
->adapter
->gl_info
);
5801 priv
->use_arbfp_fixed_func
= FALSE
;
5803 if (device
->shader_backend
!= &arb_program_shader_backend
)
5805 HeapFree(GetProcessHeap(), 0, device
->fragment_priv
);
5809 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5811 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5812 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
;
5813 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5814 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5815 WINED3DTEXOPCAPS_SELECTARG1
|
5816 WINED3DTEXOPCAPS_SELECTARG2
|
5817 WINED3DTEXOPCAPS_MODULATE4X
|
5818 WINED3DTEXOPCAPS_MODULATE2X
|
5819 WINED3DTEXOPCAPS_MODULATE
|
5820 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5821 WINED3DTEXOPCAPS_ADDSIGNED
|
5822 WINED3DTEXOPCAPS_ADD
|
5823 WINED3DTEXOPCAPS_SUBTRACT
|
5824 WINED3DTEXOPCAPS_ADDSMOOTH
|
5825 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5826 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5827 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5828 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5829 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5830 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5831 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5832 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5833 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5834 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5835 WINED3DTEXOPCAPS_MULTIPLYADD
|
5836 WINED3DTEXOPCAPS_LERP
|
5837 WINED3DTEXOPCAPS_BUMPENVMAP
|
5838 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5840 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5842 caps
->MaxTextureBlendStages
= 8;
5843 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5846 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5847 const struct wined3d_state
*state
, DWORD state_id
)
5849 struct wined3d_device
*device
= context
->swapchain
->device
;
5850 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5853 /* Don't load the parameter if we're using an arbfp pixel shader,
5854 * otherwise we'll overwrite application provided constants. */
5855 if (device
->shader_backend
== &arb_program_shader_backend
)
5857 struct shader_arb_priv
*priv
;
5859 if (use_ps(state
)) return;
5861 priv
= device
->shader_priv
;
5862 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5863 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5866 D3DCOLORTOGLFLOAT4(state
->render_states
[WINED3D_RS_TEXTUREFACTOR
], col
);
5867 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5868 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5871 static void state_arb_specularenable(struct wined3d_context
*context
,
5872 const struct wined3d_state
*state
, DWORD state_id
)
5874 struct wined3d_device
*device
= context
->swapchain
->device
;
5875 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5878 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5879 * application provided constants
5881 if (device
->shader_backend
== &arb_program_shader_backend
)
5883 struct shader_arb_priv
*priv
;
5885 if (use_ps(state
)) return;
5887 priv
= device
->shader_priv
;
5888 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5889 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5892 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5894 /* The specular color has no alpha */
5895 col
[0] = 1.0f
; col
[1] = 1.0f
;
5896 col
[2] = 1.0f
; col
[3] = 0.0f
;
5898 col
[0] = 0.0f
; col
[1] = 0.0f
;
5899 col
[2] = 0.0f
; col
[3] = 0.0f
;
5901 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5902 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5905 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5907 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5908 struct wined3d_device
*device
= context
->swapchain
->device
;
5909 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5912 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5914 if (device
->shader_backend
== &arb_program_shader_backend
)
5916 struct shader_arb_priv
*priv
= device
->shader_priv
;
5918 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5922 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5923 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5926 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5927 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5928 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5929 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5931 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5932 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5935 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5936 const struct wined3d_state
*state
, DWORD state_id
)
5938 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5939 struct wined3d_device
*device
= context
->swapchain
->device
;
5940 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5943 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5945 if (device
->shader_backend
== &arb_program_shader_backend
)
5947 struct shader_arb_priv
*priv
= device
->shader_priv
;
5949 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5953 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5954 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5957 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5958 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5962 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5963 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5966 static const char *get_argreg(struct wined3d_shader_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
5970 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
5972 switch(arg
& WINED3DTA_SELECTMASK
) {
5973 case WINED3DTA_DIFFUSE
:
5974 ret
= "fragment.color.primary"; break;
5976 case WINED3DTA_CURRENT
:
5977 if (!stage
) ret
= "fragment.color.primary";
5981 case WINED3DTA_TEXTURE
:
5983 case 0: ret
= "tex0"; break;
5984 case 1: ret
= "tex1"; break;
5985 case 2: ret
= "tex2"; break;
5986 case 3: ret
= "tex3"; break;
5987 case 4: ret
= "tex4"; break;
5988 case 5: ret
= "tex5"; break;
5989 case 6: ret
= "tex6"; break;
5990 case 7: ret
= "tex7"; break;
5991 default: ret
= "unknown texture";
5995 case WINED3DTA_TFACTOR
:
5996 ret
= "tfactor"; break;
5998 case WINED3DTA_SPECULAR
:
5999 ret
= "fragment.color.secondary"; break;
6001 case WINED3DTA_TEMP
:
6002 ret
= "tempreg"; break;
6004 case WINED3DTA_CONSTANT
:
6005 FIXME("Implement perstage constants\n");
6007 case 0: ret
= "const0"; break;
6008 case 1: ret
= "const1"; break;
6009 case 2: ret
= "const2"; break;
6010 case 3: ret
= "const3"; break;
6011 case 4: ret
= "const4"; break;
6012 case 5: ret
= "const5"; break;
6013 case 6: ret
= "const6"; break;
6014 case 7: ret
= "const7"; break;
6015 default: ret
= "unknown constant";
6023 if(arg
& WINED3DTA_COMPLEMENT
) {
6024 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6025 if(argnum
== 0) ret
= "arg0";
6026 if(argnum
== 1) ret
= "arg1";
6027 if(argnum
== 2) ret
= "arg2";
6029 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6030 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6031 if(argnum
== 0) ret
= "arg0";
6032 if(argnum
== 1) ret
= "arg1";
6033 if(argnum
== 2) ret
= "arg2";
6038 static void gen_ffp_instr(struct wined3d_shader_buffer
*buffer
, unsigned int stage
, BOOL color
,
6039 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6041 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6042 unsigned int mul
= 1;
6044 if(color
&& alpha
) dstmask
= "";
6045 else if(color
) dstmask
= ".xyz";
6046 else dstmask
= ".w";
6048 if(dst
== tempreg
) dstreg
= "tempreg";
6049 else dstreg
= "ret";
6051 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6052 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6053 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6057 case WINED3D_TOP_DISABLE
:
6059 shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
6062 case WINED3D_TOP_SELECT_ARG2
:
6065 case WINED3D_TOP_SELECT_ARG1
:
6066 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6069 case WINED3D_TOP_MODULATE_4X
:
6072 case WINED3D_TOP_MODULATE_2X
:
6075 case WINED3D_TOP_MODULATE
:
6076 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6079 case WINED3D_TOP_ADD_SIGNED_2X
:
6082 case WINED3D_TOP_ADD_SIGNED
:
6083 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6086 case WINED3D_TOP_ADD
:
6087 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6090 case WINED3D_TOP_SUBTRACT
:
6091 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6094 case WINED3D_TOP_ADD_SMOOTH
:
6095 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6096 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6099 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6100 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6101 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6103 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6104 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6105 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6107 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6108 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6109 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6111 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6112 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6113 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6116 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6117 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6118 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6119 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6122 /* D3DTOP_PREMODULATE ???? */
6124 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6125 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6126 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6128 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6129 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6131 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6132 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6133 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6135 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6136 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6139 case WINED3D_TOP_DOTPRODUCT3
:
6141 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6142 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6143 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6146 case WINED3D_TOP_MULTIPLY_ADD
:
6147 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6150 case WINED3D_TOP_LERP
:
6151 /* The msdn is not quite right here */
6152 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6155 case WINED3D_TOP_BUMPENVMAP
:
6156 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6157 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6161 FIXME("Unhandled texture op %08x\n", op
);
6165 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6167 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6170 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6173 struct wined3d_shader_buffer buffer
;
6174 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6175 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6176 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
6177 UINT lowest_disabled_stage
;
6178 const char *textype
;
6180 char colorcor_dst
[8];
6182 DWORD arg0
, arg1
, arg2
;
6183 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6185 const char *final_combiner_src
= "ret";
6187 BOOL custom_linear_fog
= FALSE
;
6189 /* Find out which textures are read */
6190 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6192 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6194 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6195 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6196 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6197 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6198 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6199 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6201 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA
)
6202 tex_read
[stage
] = TRUE
;
6203 if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
)
6204 tex_read
[stage
] = TRUE
;
6205 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP
)
6207 bump_used
[stage
] = TRUE
;
6208 tex_read
[stage
] = TRUE
;
6210 if (settings
->op
[stage
].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6212 bump_used
[stage
] = TRUE
;
6213 tex_read
[stage
] = TRUE
;
6214 luminance_used
[stage
] = TRUE
;
6216 else if (settings
->op
[stage
].cop
== WINED3D_TOP_BLEND_FACTOR_ALPHA
)
6218 tfactor_used
= TRUE
;
6221 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6222 tfactor_used
= TRUE
;
6225 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
6226 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6227 tempreg_used
= TRUE
;
6230 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6232 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6233 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6234 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6235 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6236 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6237 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
6239 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
6240 tempreg_used
= TRUE
;
6242 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
6243 tfactor_used
= TRUE
;
6246 lowest_disabled_stage
= stage
;
6249 if (!shader_buffer_init(&buffer
))
6251 ERR("Failed to initialize shader buffer.\n");
6255 shader_addline(&buffer
, "!!ARBfp1.0\n");
6257 switch (settings
->fog
)
6259 case WINED3D_FFP_PS_FOG_OFF
: break;
6260 case WINED3D_FFP_PS_FOG_LINEAR
:
6261 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6263 custom_linear_fog
= TRUE
;
6266 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6269 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6270 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6271 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6274 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6275 shader_addline(&buffer
, "TEMP TMP;\n");
6276 shader_addline(&buffer
, "TEMP ret;\n");
6277 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6278 shader_addline(&buffer
, "TEMP arg0;\n");
6279 shader_addline(&buffer
, "TEMP arg1;\n");
6280 shader_addline(&buffer
, "TEMP arg2;\n");
6281 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6282 if(!tex_read
[stage
]) continue;
6283 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6284 if(!bump_used
[stage
]) continue;
6285 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6286 if(!luminance_used
[stage
]) continue;
6287 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6290 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6292 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6294 if(settings
->sRGB_write
) {
6295 shader_addline(&buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
6296 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
6297 shader_addline(&buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
6298 srgb_sub_high
, 0.0, 0.0, 0.0);
6301 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6302 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6304 /* Generate texture sampling instructions) */
6305 for (stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6307 if (!tex_read
[stage
])
6310 switch(settings
->op
[stage
].tex_type
) {
6311 case tex_1d
: textype
= "1D"; break;
6312 case tex_2d
: textype
= "2D"; break;
6313 case tex_3d
: textype
= "3D"; break;
6314 case tex_cube
: textype
= "CUBE"; break;
6315 case tex_rect
: textype
= "RECT"; break;
6316 default: textype
= "unexpected_textype"; break;
6319 if(settings
->op
[stage
].projected
== proj_none
) {
6321 } else if(settings
->op
[stage
].projected
== proj_count4
||
6322 settings
->op
[stage
].projected
== proj_count3
) {
6325 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6330 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6331 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6333 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6334 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6335 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6336 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6338 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6339 * so multiply the displacement with the dividing parameter before passing it to TXP
6341 if (settings
->op
[stage
].projected
!= proj_none
) {
6342 if(settings
->op
[stage
].projected
== proj_count4
) {
6343 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6344 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6346 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6347 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6350 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6353 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6354 instr
, stage
, stage
, textype
);
6355 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6357 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6358 stage
- 1, stage
- 1, stage
- 1);
6359 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6361 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6362 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6363 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6364 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6365 instr
, stage
, stage
, textype
);
6367 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6368 instr
, stage
, stage
, stage
, textype
);
6371 sprintf(colorcor_dst
, "tex%u", stage
);
6372 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
6373 settings
->op
[stage
].color_fixup
);
6376 /* Generate the main shader */
6377 for (stage
= 0; stage
< MAX_TEXTURES
; ++stage
)
6379 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6382 final_combiner_src
= "fragment.color.primary";
6386 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6387 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6388 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6389 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6390 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6391 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6392 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6393 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6394 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6395 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6396 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6397 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6399 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6400 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6401 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6402 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6404 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6406 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6407 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6408 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6410 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
6414 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6415 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6416 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6418 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6419 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6420 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6421 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6422 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6423 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6427 if (settings
->sRGB_write
|| custom_linear_fog
)
6429 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6430 if (settings
->sRGB_write
)
6431 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6432 if (custom_linear_fog
)
6433 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6434 shader_addline(&buffer
, "MOV result.color, ret;\n");
6438 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n",
6439 final_combiner_src
);
6443 shader_addline(&buffer
, "END\n");
6445 /* Generate the shader */
6446 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6447 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6448 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6449 strlen(buffer
.buffer
), buffer
.buffer
));
6450 checkGLcall("glProgramStringARB()");
6452 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6455 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6456 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6457 shader_arb_dump_program_source(buffer
.buffer
);
6463 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
6464 checkGLcall("glGetProgramivARB()");
6465 if (!native
) WARN("Program exceeds native resource limits.\n");
6468 shader_buffer_free(&buffer
);
6472 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6474 const struct wined3d_device
*device
= context
->swapchain
->device
;
6475 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6476 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6477 BOOL use_pshader
= use_ps(state
);
6478 struct ffp_frag_settings settings
;
6479 const struct arbfp_ffp_desc
*desc
;
6482 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6484 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6486 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6488 /* Reload fixed function constants since they collide with the
6489 * pixel shader constants. */
6490 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6492 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6494 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6495 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6497 else if (use_pshader
)
6499 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6506 /* Find or create a shader implementing the fixed function pipeline
6507 * settings, then activate it. */
6508 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6509 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6511 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6514 ERR("Out of memory\n");
6518 new_desc
->parent
.settings
= settings
;
6519 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6520 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6521 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6525 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6526 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6529 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6530 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6531 priv
->current_fprogram_id
= desc
->shader
;
6533 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6535 /* Reload fixed function constants since they collide with the
6536 * pixel shader constants. */
6537 for (i
= 0; i
< MAX_TEXTURES
; ++i
)
6539 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6541 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6542 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6544 context
->last_was_pshader
= FALSE
;
6546 context
->last_was_pshader
= TRUE
;
6549 context
->shader_update_mask
|= 1 << WINED3D_SHADER_TYPE_PIXEL
;
6552 /* We can't link the fog states to the fragment state directly since the
6553 * vertex pipeline links them to FOGENABLE. A different linking in different
6554 * pipeline parts can't be expressed in the combined state table, so we need
6555 * to handle that with a forwarding function. The other invisible side effect
6556 * is that changing the fog start and fog end (which links to FOGENABLE in
6557 * vertex) results in the fragment_prog_arbfp function being called because
6558 * FOGENABLE is dirty, which calls this function here. */
6559 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6561 enum fogsource new_source
;
6562 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6563 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6565 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6567 if (!isStateDirty(context
, STATE_PIXELSHADER
))
6568 fragment_prog_arbfp(context
, state
, state_id
);
6570 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6573 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6577 new_source
= FOGSOURCE_VS
;
6581 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6582 new_source
= FOGSOURCE_COORD
;
6584 new_source
= FOGSOURCE_FFP
;
6589 new_source
= FOGSOURCE_FFP
;
6592 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6594 context
->fog_source
= new_source
;
6595 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6599 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6601 if (!isStateDirty(context
, STATE_PIXELSHADER
))
6602 fragment_prog_arbfp(context
, state
, state_id
);
6605 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] =
6607 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6608 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6609 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6610 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6611 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6612 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6613 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6614 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6615 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6616 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6617 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6618 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6619 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6620 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6621 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6622 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6623 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6624 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6625 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6626 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6627 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6628 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6629 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6630 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6631 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6632 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6633 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6634 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6635 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6636 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6637 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6638 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6639 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6640 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6641 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6642 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6643 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6644 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6645 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6646 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6647 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6648 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6649 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6650 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6651 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6652 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6653 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6654 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6655 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6656 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6657 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6658 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6659 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6660 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6661 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6662 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6663 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6664 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6665 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6666 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6667 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6668 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6669 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6670 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6671 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6672 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6673 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6674 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6675 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6676 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6677 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6678 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6679 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6680 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6681 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6682 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6683 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6684 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6685 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6686 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6687 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6688 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6689 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6690 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6691 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6692 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6693 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6694 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6695 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6696 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6697 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6698 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6699 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6700 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6701 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6702 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6703 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6735 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6740 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6741 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6742 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6743 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6744 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6745 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7, WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6746 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6747 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6750 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6755 shader_arb_color_fixup_supported
,
6756 arbfp_fragmentstate_template
,
6759 struct arbfp_blit_priv
{
6760 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
6761 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
6762 GLenum yv12_rect_shader
, yv12_2d_shader
;
6763 GLenum p8_rect_shader
, p8_2d_shader
;
6764 GLuint palette_texture
;
6767 static HRESULT
arbfp_blit_alloc(struct wined3d_device
*device
)
6769 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
6770 if(!device
->blit_priv
) {
6771 ERR("Out of memory\n");
6772 return E_OUTOFMEMORY
;
6777 /* Context activation is done by the caller. */
6778 static void arbfp_blit_free(struct wined3d_device
*device
)
6780 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6781 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6783 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
6784 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
6785 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
6786 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
6787 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
6788 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
6789 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->p8_rect_shader
));
6790 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->p8_2d_shader
));
6791 checkGLcall("Delete yuv and p8 programs");
6793 if (priv
->palette_texture
)
6794 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &priv
->palette_texture
);
6796 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6797 device
->blit_priv
= NULL
;
6800 static BOOL
gen_planar_yuv_read(struct wined3d_shader_buffer
*buffer
, enum complex_fixup fixup
,
6801 GLenum textype
, char *luminance
)
6804 const char *tex
, *texinstr
;
6806 if (fixup
== COMPLEX_FIXUP_UYVY
) {
6814 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
6815 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
6817 /* This is more tricky than just replacing the texture type - we have to navigate
6818 * properly in the texture to find the correct chroma values
6820 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
6824 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6825 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6826 * filtering when we sample the texture.
6828 * These are the rules for reading the chroma:
6834 * So we have to get the sampling x position in non-normalized coordinates in integers
6836 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
6837 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6838 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6840 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6842 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6843 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6846 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6847 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6849 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6850 * even and odd pixels respectively
6852 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6853 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6855 /* Sample Pixel 1 */
6856 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6858 /* Put the value into either of the chroma values */
6859 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6860 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6861 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6862 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6864 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6865 * the pixel right to the current one. Otherwise, sample the left pixel.
6866 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6868 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6869 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6870 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6872 /* Put the value into the other chroma */
6873 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6874 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
6875 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6876 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
6878 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
6879 * the current one and lerp the two U and V values
6882 /* This gives the correctly filtered luminance value */
6883 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
6888 static BOOL
gen_yv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
, char *luminance
)
6893 case GL_TEXTURE_2D
: tex
= "2D"; break;
6894 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
6896 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
6900 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
6901 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
6902 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
6903 * pitch of the luminance plane, the packing into the gl texture is a bit
6904 * unfortunate. If the whole texture is interpreted as luminance data it looks
6905 * approximately like this:
6907 * +----------------------------------+----
6919 * +----------------+-----------------+----
6921 * | U even rows | U odd rows |
6923 * +----------------+------------------ -
6925 * | V even rows | V odd rows |
6927 * +----------------+-----------------+----
6931 * So it appears as if there are 4 chroma images, but in fact the odd rows
6932 * in the chroma images are in the same row as the even ones. So its is
6933 * kinda tricky to read
6935 * When reading from rectangle textures, keep in mind that the input y coordinates
6936 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
6938 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
6939 2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
);
6941 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6942 /* the chroma planes have only half the width */
6943 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
6945 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
6946 * the coordinate. Also read the right side of the image when reading odd lines
6948 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
6951 if(textype
== GL_TEXTURE_2D
) {
6953 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
6955 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
6957 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
6958 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
6960 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6961 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6962 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6963 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6964 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6966 /* clamp, keep the half pixel origin in mind */
6967 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
6968 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6969 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
6970 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6972 /* Read from [size - size+size/4] */
6973 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6974 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
6976 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6977 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6978 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6979 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6980 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
6981 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6983 /* Make sure to read exactly from the pixel center */
6984 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6985 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
6988 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
6989 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
6990 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6991 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
6992 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6994 /* Read the texture, put the result into the output register */
6995 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6996 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
6998 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
6999 * No need to clamp because we're just reusing the already clamped value from above
7001 if(textype
== GL_TEXTURE_2D
) {
7002 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7004 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7006 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7007 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7009 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7010 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7011 * values due to filtering
7013 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7014 if(textype
== GL_TEXTURE_2D
) {
7015 /* Multiply the y coordinate by 2/3 and clamp it */
7016 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7017 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7018 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7019 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7021 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7022 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7025 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7026 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7027 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7034 static GLuint
gen_p8_shader(struct arbfp_blit_priv
*priv
,
7035 const struct wined3d_gl_info
*gl_info
, GLenum textype
)
7038 struct wined3d_shader_buffer buffer
;
7042 if (!shader_buffer_init(&buffer
))
7044 ERR("Failed to initialize shader buffer.\n");
7048 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7049 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7052 shader_buffer_free(&buffer
);
7056 shader_addline(&buffer
, "!!ARBfp1.0\n");
7057 shader_addline(&buffer
, "TEMP index;\n");
7059 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7060 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7062 /* The alpha-component contains the palette index */
7063 if(textype
== GL_TEXTURE_RECTANGLE_ARB
)
7064 shader_addline(&buffer
, "TXP index, fragment.texcoord[0], texture[0], RECT;\n");
7066 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], 2D;\n");
7068 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7069 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7071 /* Use the alpha-component as an index in the palette to get the final color */
7072 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7073 shader_addline(&buffer
, "END\n");
7075 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7076 strlen(buffer
.buffer
), buffer
.buffer
));
7077 checkGLcall("glProgramStringARB()");
7079 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7082 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7083 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7084 shader_arb_dump_program_source(buffer
.buffer
);
7087 if (textype
== GL_TEXTURE_RECTANGLE_ARB
)
7088 priv
->p8_rect_shader
= shader
;
7090 priv
->p8_2d_shader
= shader
;
7092 shader_buffer_free(&buffer
);
7097 /* Context activation is done by the caller. */
7098 static void upload_palette(const struct wined3d_surface
*surface
, struct wined3d_context
*context
)
7101 struct wined3d_device
*device
= surface
->resource
.device
;
7102 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7103 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
7104 BOOL colorkey
= (surface
->CKeyFlags
& WINEDDSD_CKSRCBLT
) != 0;
7106 d3dfmt_p8_init_palette(surface
, table
, colorkey
);
7108 if (!priv
->palette_texture
)
7109 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &priv
->palette_texture
);
7111 GL_EXTCALL(glActiveTextureARB(GL_TEXTURE1
));
7112 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
7114 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7116 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7117 /* Make sure we have discrete color levels. */
7118 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7119 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7120 /* Upload the palette */
7121 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7122 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGBA
, 256, 0, GL_RGBA
, GL_UNSIGNED_BYTE
, table
);
7124 /* Switch back to unit 0 in which the 2D texture will be stored. */
7125 context_active_texture(context
, gl_info
, 0);
7128 /* Context activation is done by the caller. */
7129 static GLuint
gen_yuv_shader(struct arbfp_blit_priv
*priv
, const struct wined3d_gl_info
*gl_info
,
7130 enum complex_fixup yuv_fixup
, GLenum textype
)
7133 struct wined3d_shader_buffer buffer
;
7134 char luminance_component
;
7138 if (!shader_buffer_init(&buffer
))
7140 ERR("Failed to initialize shader buffer.\n");
7144 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7145 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7146 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7147 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7150 shader_buffer_free(&buffer
);
7154 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7155 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7156 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7157 * each single pixel it contains, and one U and one V value shared between both
7160 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7161 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7162 * take the format into account when generating the read swizzles
7164 * Reading the Y value is straightforward - just sample the texture. The hardware
7165 * takes care of filtering in the horizontal and vertical direction.
7167 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7168 * because that would mix the U and V values of one pixel or two adjacent pixels.
7169 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7170 * regardless of the filtering setting. Vertical filtering works automatically
7171 * though - the U and V values of two rows are mixed nicely.
7173 * Apart of avoiding filtering issues, the code has to know which value it just
7174 * read, and where it can find the other one. To determine this, it checks if
7175 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7177 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7178 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7180 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7181 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7182 * in an unfiltered situation. Finding the luminance on the other hand requires
7183 * finding out if it is an odd or even pixel. The real drawback of this approach
7184 * is filtering. This would have to be emulated completely in the shader, reading
7185 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7186 * vertically. Beyond that it would require adjustments to the texture handling
7187 * code to deal with the width scaling
7189 shader_addline(&buffer
, "!!ARBfp1.0\n");
7190 shader_addline(&buffer
, "TEMP luminance;\n");
7191 shader_addline(&buffer
, "TEMP temp;\n");
7192 shader_addline(&buffer
, "TEMP chroma;\n");
7193 shader_addline(&buffer
, "TEMP texcrd;\n");
7194 shader_addline(&buffer
, "TEMP texcrd2;\n");
7195 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7196 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7197 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
7201 case COMPLEX_FIXUP_UYVY
:
7202 case COMPLEX_FIXUP_YUY2
:
7203 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
7205 shader_buffer_free(&buffer
);
7210 case COMPLEX_FIXUP_YV12
:
7211 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
7213 shader_buffer_free(&buffer
);
7219 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
7220 shader_buffer_free(&buffer
);
7224 /* Calculate the final result. Formula is taken from
7225 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7226 * ranges from -0.5 to 0.5
7228 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7230 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7231 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7232 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7233 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7234 shader_addline(&buffer
, "END\n");
7236 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
7237 strlen(buffer
.buffer
), buffer
.buffer
));
7238 checkGLcall("glProgramStringARB()");
7240 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
7243 FIXME("Fragment program error at position %d: %s\n\n", pos
,
7244 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
7245 shader_arb_dump_program_source(buffer
.buffer
);
7251 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
7252 checkGLcall("glGetProgramivARB()");
7253 if (!native
) WARN("Program exceeds native resource limits.\n");
7256 shader_buffer_free(&buffer
);
7260 case COMPLEX_FIXUP_YUY2
:
7261 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yuy2_rect_shader
= shader
;
7262 else priv
->yuy2_2d_shader
= shader
;
7265 case COMPLEX_FIXUP_UYVY
:
7266 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->uyvy_rect_shader
= shader
;
7267 else priv
->uyvy_2d_shader
= shader
;
7270 case COMPLEX_FIXUP_YV12
:
7271 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yv12_rect_shader
= shader
;
7272 else priv
->yv12_2d_shader
= shader
;
7275 ERR("Unsupported complex fixup: %d\n", yuv_fixup
);
7281 /* Context activation is done by the caller. */
7282 static HRESULT
arbfp_blit_set(void *blit_priv
, struct wined3d_context
*context
, const struct wined3d_surface
*surface
)
7285 float size
[4] = {(float) surface
->pow2Width
, (float) surface
->pow2Height
, 1.0f
, 1.0f
};
7286 struct arbfp_blit_priv
*priv
= blit_priv
;
7287 enum complex_fixup fixup
;
7288 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
7291 if (surface
->container
)
7292 textype
= surface
->container
->target
;
7294 textype
= surface
->texture_target
;
7296 if (surface
->flags
& SFLAG_CONVERTED
)
7298 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7299 checkGLcall("glEnable(textype)");
7303 if (!is_complex_fixup(surface
->resource
.format
->color_fixup
))
7306 dump_color_fixup_desc(surface
->resource
.format
->color_fixup
);
7307 /* Don't bother setting up a shader for unconverted formats */
7308 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7309 checkGLcall("glEnable(textype)");
7313 fixup
= get_complex_fixup(surface
->resource
.format
->color_fixup
);
7317 case COMPLEX_FIXUP_YUY2
:
7318 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yuy2_rect_shader
: priv
->yuy2_2d_shader
;
7321 case COMPLEX_FIXUP_UYVY
:
7322 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->uyvy_rect_shader
: priv
->uyvy_2d_shader
;
7325 case COMPLEX_FIXUP_YV12
:
7326 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yv12_rect_shader
: priv
->yv12_2d_shader
;
7329 case COMPLEX_FIXUP_P8
:
7330 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->p8_rect_shader
: priv
->p8_2d_shader
;
7331 if (!shader
) shader
= gen_p8_shader(priv
, gl_info
, textype
);
7333 upload_palette(surface
, context
);
7337 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7338 gl_info
->gl_ops
.gl
.p_glEnable(textype
);
7339 checkGLcall("glEnable(textype)");
7343 if (!shader
) shader
= gen_yuv_shader(priv
, gl_info
, fixup
, textype
);
7345 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7346 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7347 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7348 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7349 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
7350 checkGLcall("glProgramLocalParameter4fvARB");
7355 /* Context activation is done by the caller. */
7356 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7358 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7359 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7360 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_2D
);
7361 checkGLcall("glDisable(GL_TEXTURE_2D)");
7362 if (gl_info
->supported
[ARB_TEXTURE_CUBE_MAP
])
7364 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
7365 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
7367 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
7369 gl_info
->gl_ops
.gl
.p_glDisable(GL_TEXTURE_RECTANGLE_ARB
);
7370 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
7374 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
, enum wined3d_blit_op blit_op
,
7375 const RECT
*src_rect
, DWORD src_usage
, enum wined3d_pool src_pool
, const struct wined3d_format
*src_format
,
7376 const RECT
*dst_rect
, DWORD dst_usage
, enum wined3d_pool dst_pool
, const struct wined3d_format
*dst_format
)
7378 enum complex_fixup src_fixup
;
7380 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7383 if (blit_op
!= WINED3D_BLIT_OP_COLOR_BLIT
)
7385 TRACE("Unsupported blit_op=%d\n", blit_op
);
7389 if (src_pool
== WINED3D_POOL_SYSTEM_MEM
|| dst_pool
== WINED3D_POOL_SYSTEM_MEM
)
7392 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7393 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7395 TRACE("Checking support for fixup:\n");
7396 dump_color_fixup_desc(src_format
->color_fixup
);
7399 if (!is_identity_fixup(dst_format
->color_fixup
))
7401 TRACE("Destination fixups are not supported\n");
7405 if (is_identity_fixup(src_format
->color_fixup
))
7411 /* We only support YUV conversions. */
7412 if (!is_complex_fixup(src_format
->color_fixup
))
7414 TRACE("[FAILED]\n");
7420 case COMPLEX_FIXUP_YUY2
:
7421 case COMPLEX_FIXUP_UYVY
:
7422 case COMPLEX_FIXUP_YV12
:
7423 case COMPLEX_FIXUP_P8
:
7428 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7429 TRACE("[FAILED]\n");
7434 HRESULT
arbfp_blit_surface(struct wined3d_device
*device
, DWORD filter
,
7435 struct wined3d_surface
*src_surface
, const RECT
*src_rect_in
,
7436 struct wined3d_surface
*dst_surface
, const RECT
*dst_rect_in
)
7438 struct wined3d_context
*context
;
7439 RECT src_rect
= *src_rect_in
;
7440 RECT dst_rect
= *dst_rect_in
;
7442 /* Now load the surface */
7443 if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7444 && (src_surface
->flags
& (SFLAG_INTEXTURE
| SFLAG_INDRAWABLE
)) == SFLAG_INDRAWABLE
)
7446 /* Without FBO blits transferring from the drawable to the texture is
7447 * expensive, because we have to flip the data in sysmem. Since we can
7448 * flip in the blitter, we don't actually need that flip anyway. So we
7449 * use the surface's texture as scratch texture, and flip the source
7450 * rectangle instead. */
7451 surface_load_fb_texture(src_surface
, FALSE
);
7453 src_rect
.top
= src_surface
->resource
.height
- src_rect
.top
;
7454 src_rect
.bottom
= src_surface
->resource
.height
- src_rect
.bottom
;
7457 surface_internal_preload(src_surface
, SRGB_RGB
);
7459 /* Activate the destination context, set it up for blitting */
7460 context
= context_acquire(device
, dst_surface
);
7461 context_apply_blit_state(context
, device
);
7463 if (!surface_is_offscreen(dst_surface
))
7464 surface_translate_drawable_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7466 arbfp_blit_set(device
->blit_priv
, context
, src_surface
);
7468 /* Draw a textured quad */
7469 draw_textured_quad(src_surface
, context
, &src_rect
, &dst_rect
, filter
);
7471 /* Leave the opengl state valid for blitting */
7472 arbfp_blit_unset(context
->gl_info
);
7474 if (wined3d_settings
.strict_draw_ordering
7475 || (dst_surface
->swapchain
&& (dst_surface
->swapchain
->front_buffer
== dst_surface
)))
7476 context
->gl_info
->gl_ops
.gl
.p_glFlush(); /* Flush to ensure ordering across contexts. */
7478 context_release(context
);
7480 surface_modify_location(dst_surface
, dst_surface
->draw_binding
, TRUE
);
7484 static HRESULT
arbfp_blit_color_fill(struct wined3d_device
*device
, struct wined3d_surface
*dst_surface
,
7485 const RECT
*dst_rect
, const struct wined3d_color
*color
)
7487 FIXME("Color filling not implemented by arbfp_blit\n");
7488 return WINED3DERR_INVALIDCALL
;
7491 static HRESULT
arbfp_blit_depth_fill(struct wined3d_device
*device
,
7492 struct wined3d_surface
*surface
, const RECT
*rect
, float depth
)
7494 FIXME("Depth filling not implemented by arbfp_blit.\n");
7495 return WINED3DERR_INVALIDCALL
;
7498 const struct blit_shader arbfp_blit
= {
7503 arbfp_blit_supported
,
7504 arbfp_blit_color_fill
,
7505 arbfp_blit_depth_fill
,