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, 2013-2014 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"
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf
);
42 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
44 return type
== WINED3D_SHADER_TYPE_PIXEL
;
47 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
49 return type
== WINED3D_SHADER_TYPE_VERTEX
;
52 static const char *get_line(const char **ptr
)
57 if (!(q
= strstr(p
, "\n")))
79 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
81 if (shader
!= WINED3D_SHADER_TYPE_VERTEX
&& shader
!= WINED3D_SHADER_TYPE_PIXEL
)
83 ERR("Unsupported shader type '%s'.\n", debug_shader_type(shader
));
87 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
91 case ARB_ZERO
: return "ps_helper_const.x";
92 case ARB_ONE
: return "ps_helper_const.y";
93 case ARB_TWO
: return "coefmul.x";
94 case ARB_0001
: return "ps_helper_const.xxxy";
95 case ARB_EPS
: return "ps_helper_const.z";
103 case ARB_ZERO
: return "helper_const.x";
104 case ARB_ONE
: return "helper_const.y";
105 case ARB_TWO
: return "helper_const.z";
106 case ARB_EPS
: return "helper_const.w";
107 case ARB_0001
: return "helper_const.xxxy";
108 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
111 FIXME("Unmanaged %s shader helper constant requested: %u.\n",
112 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
115 case ARB_ZERO
: return "0.0";
116 case ARB_ONE
: return "1.0";
117 case ARB_TWO
: return "2.0";
118 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
119 case ARB_EPS
: return "1e-8";
120 default: return "bad";
124 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
126 return context
->lowest_disabled_stage
< 7;
129 /* ARB_program_shader private data */
148 struct wined3d_shader_loop_control loop_control
;
152 struct arb_ps_np2fixup_info
154 struct ps_np2fixup_info super
;
155 /* For ARB we need an offset value:
156 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
157 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
158 * array we need an offset to the index inside the program local parameter array. */
162 struct arb_ps_compile_args
164 struct ps_compile_args super
;
166 WORD clip
; /* only a boolean, use a WORD for alignment */
167 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
170 struct stb_const_desc
172 unsigned char texunit
;
176 struct arb_ps_compiled_shader
178 struct arb_ps_compile_args args
;
179 struct arb_ps_np2fixup_info np2fixup_info
;
180 struct stb_const_desc bumpenvmatconst
[WINED3D_MAX_TEXTURES
];
181 struct stb_const_desc luminanceconst
[WINED3D_MAX_TEXTURES
];
182 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
185 unsigned char numbumpenvmatconsts
;
189 struct arb_vs_compile_args
191 struct vs_compile_args super
;
197 unsigned char clip_texcoord
;
198 unsigned char clipplane_mask
;
200 DWORD boolclip_compare
;
205 unsigned char samplers
[4];
206 DWORD samplers_compare
;
208 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
211 struct arb_vs_compiled_shader
213 struct arb_vs_compile_args args
;
215 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
217 char need_color_unclamp
;
221 struct recorded_instruction
223 struct wined3d_shader_instruction ins
;
227 struct shader_arb_ctx_priv
232 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
234 /* GL_NV_vertex_program2_option or GL_NV_fragment_program_option */
236 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
240 const struct wined3d_gl_info
*gl_info
;
241 const struct arb_vs_compile_args
*cur_vs_args
;
242 const struct arb_ps_compile_args
*cur_ps_args
;
243 const struct arb_ps_compiled_shader
*compiled_fprog
;
244 const struct arb_vs_compiled_shader
*compiled_vprog
;
245 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
246 struct list control_frames
;
250 unsigned int num_loops
, loop_depth
, num_ifcs
;
252 BOOL ps_post_process
;
254 unsigned int vs_clipplanes
;
258 /* For 3.0 vertex shaders */
259 const char *vs_output
[MAX_REG_OUTPUT
];
260 /* For 2.x and earlier vertex shaders */
261 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
263 /* 3.0 pshader input for compatibility with fixed function */
264 const char *ps_input
[MAX_REG_INPUT
];
269 struct wined3d_shader_signature sig
;
271 struct wine_rb_entry entry
;
274 struct arb_pshader_private
{
275 struct arb_ps_compiled_shader
*gl_shaders
;
276 UINT num_gl_shaders
, shader_array_size
;
277 DWORD input_signature_idx
;
278 DWORD clipplane_emulation
;
282 struct arb_vshader_private
{
283 struct arb_vs_compiled_shader
*gl_shaders
;
284 UINT num_gl_shaders
, shader_array_size
;
288 struct shader_arb_priv
290 GLuint current_vprogram_id
;
291 GLuint current_fprogram_id
;
292 const struct arb_ps_compiled_shader
*compiled_fprog
;
293 const struct arb_vs_compiled_shader
*compiled_vprog
;
294 BOOL use_arbfp_fixed_func
;
295 struct wine_rb_tree fragment_shaders
;
296 BOOL last_ps_const_clamped
;
297 BOOL last_vs_color_unclamp
;
299 struct wine_rb_tree signature_tree
;
302 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
303 char vshader_const_dirty
[WINED3D_MAX_VS_CONSTS_F
];
304 char pshader_const_dirty
[WINED3D_MAX_PS_CONSTS_F
];
305 const struct wined3d_context
*last_context
;
307 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
308 const struct wined3d_fragment_pipe_ops
*fragment_pipe
;
309 BOOL ffp_proj_control
;
312 /* Context activation for state handlers is done by the caller. */
314 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
315 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
317 if (shader_data
->rel_offset
) return TRUE
;
318 if (!reg_maps
->usesmova
) return FALSE
;
319 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
322 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
323 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
325 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
326 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
329 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
330 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
332 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
333 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
334 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
335 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
336 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
337 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
338 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
339 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
343 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
344 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
346 unsigned int ret
= 1;
347 /* We use one PARAM for the pos fixup, and in some cases one to load
348 * some immediate values into the shader. */
349 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
350 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
354 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
355 * When constant_list == NULL, it will load all the constants.
357 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
358 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
360 /* Context activation is done by the caller. */
361 static unsigned int shader_arb_load_constants_f(const struct wined3d_shader
*shader
,
362 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
363 const struct wined3d_vec4
*constants
, char *dirty_consts
)
365 struct wined3d_shader_lconst
*lconst
;
366 unsigned int ret
, i
, j
;
368 if (TRACE_ON(d3d_constants
))
370 for (i
= 0; i
< max_constants
; ++i
)
372 if (!dirty_consts
[i
])
374 TRACE_(d3d_constants
)("Loading constant %u: %s.\n", i
, debug_vec4(&constants
[i
]));
380 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
381 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
384 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
385 * shaders, the first 8 constants are marked dirty for reload
387 for (; i
< min(8, max_constants
); ++i
)
389 if (!dirty_consts
[i
])
393 if (constants
[i
].x
> 1.0f
)
395 else if (constants
[i
].x
< -1.0f
)
396 lcl_const
[0] = -1.0f
;
398 lcl_const
[0] = constants
[i
].x
;
400 if (constants
[i
].y
> 1.0f
)
402 else if (constants
[i
].y
< -1.0f
)
403 lcl_const
[1] = -1.0f
;
405 lcl_const
[1] = constants
[i
].y
;
407 if (constants
[i
].z
> 1.0f
)
409 else if (constants
[i
].z
< -1.0f
)
410 lcl_const
[2] = -1.0f
;
412 lcl_const
[2] = constants
[i
].z
;
414 if (constants
[i
].w
> 1.0f
)
416 else if (constants
[i
].w
< -1.0f
)
417 lcl_const
[3] = -1.0f
;
419 lcl_const
[3] = constants
[i
].w
;
421 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
424 /* If further constants are dirty, reload them without clamping.
426 * The alternative is not to touch them, but then we cannot reset the dirty constant count
427 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
428 * above would always re-check the first 8 constants since max_constant remains at the init
433 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
435 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
436 * or just reloading *all* constants at once
438 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
440 for (; i
< max_constants
; ++i
)
442 if (!dirty_consts
[i
])
445 /* Find the next block of dirty constants */
448 for (++i
; (i
< max_constants
) && dirty_consts
[i
]; ++i
)
453 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, &constants
[j
].x
));
458 for (; i
< max_constants
; ++i
)
463 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, &constants
[i
].x
));
467 checkGLcall("glProgramEnvParameter4fvARB()");
469 /* Load immediate constants */
470 if (shader
->load_local_constsF
)
472 if (TRACE_ON(d3d_shader
))
474 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
476 GLfloat
* values
= (GLfloat
*)lconst
->value
;
477 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
478 values
[0], values
[1], values
[2], values
[3]);
481 /* Immediate constants are clamped for 1.X shaders at loading times */
483 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
485 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
486 ret
= max(ret
, lconst
->idx
+ 1);
487 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
489 checkGLcall("glProgramEnvParameter4fvARB()");
490 return ret
; /* The loaded immediate constants need reloading for the next shader */
492 return 0; /* No constants are dirty now */
496 /* Loads the texture dimensions for NP2 fixup into the currently set
497 * ARB_[vertex/fragment]_programs. */
498 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
499 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
501 GLfloat np2fixup_constants
[4 * WINED3D_MAX_FRAGMENT_SAMPLERS
];
502 WORD active
= fixup
->super
.active
;
508 for (i
= 0; active
; active
>>= 1, ++i
)
510 const struct wined3d_texture
*tex
= state
->textures
[i
];
511 unsigned char idx
= fixup
->super
.idx
[i
];
512 GLfloat
*tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
519 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
525 tex_dim
[2] = tex
->pow2_matrix
[0];
526 tex_dim
[3] = tex
->pow2_matrix
[5];
530 tex_dim
[0] = tex
->pow2_matrix
[0];
531 tex_dim
[1] = tex
->pow2_matrix
[5];
535 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
537 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
538 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
542 /* Context activation is done by the caller. */
543 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
544 const struct wined3d_context_gl
*context_gl
, const struct wined3d_state
*state
, unsigned int rt_height
)
546 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
549 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
551 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
553 /* The state manager takes care that this function is always called if the bump env matrix changes */
554 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
555 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
556 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
558 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
560 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
561 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
562 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
563 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
565 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
566 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
567 gl_shader
->luminanceconst
[i
].const_num
, scale
));
570 checkGLcall("Load bumpmap consts");
572 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
574 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
575 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
580 val
[0] = context_gl
->c
.render_offscreen
? 0.0f
: (float)rt_height
;
581 val
[1] = context_gl
->c
.render_offscreen
? 1.0f
: -1.0f
;
584 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
585 checkGLcall("y correction loading");
588 if (!gl_shader
->num_int_consts
) return;
590 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
592 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
595 val
[0] = (float)state
->ps_consts_i
[i
].x
;
596 val
[1] = (float)state
->ps_consts_i
[i
].y
;
597 val
[2] = (float)state
->ps_consts_i
[i
].z
;
600 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
603 checkGLcall("Load ps int consts");
606 /* Context activation is done by the caller. */
607 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
608 const struct wined3d_context_gl
*context_gl
, const struct wined3d_state
*state
)
610 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
611 float position_fixup
[4];
614 /* Upload the position fixup */
615 shader_get_position_fixup(&context_gl
->c
, state
, 1, position_fixup
);
616 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
618 if (!gl_shader
->num_int_consts
) return;
620 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
622 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
625 val
[0] = (float)state
->vs_consts_i
[i
].x
;
626 val
[1] = (float)state
->vs_consts_i
[i
].y
;
627 val
[2] = (float)state
->vs_consts_i
[i
].z
;
630 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
633 checkGLcall("Load vs int consts");
636 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
637 const struct wined3d_state
*state
);
640 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
642 * We only support float constants in ARB at the moment, so don't
643 * worry about the Integers or Booleans
645 /* Context activation is done by the caller (state handler). */
646 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
, struct wined3d_context_gl
*context_gl
,
647 const struct wined3d_state
*state
, BOOL use_ps
, BOOL use_vs
, BOOL from_shader_select
)
649 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
650 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
652 if (!from_shader_select
)
654 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
655 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
658 && (vshader
->reg_maps
.boolean_constants
659 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
660 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
662 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
663 shader_arb_select(priv
, &context_gl
->c
, state
);
666 && (pshader
->reg_maps
.boolean_constants
667 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
668 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
670 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
671 shader_arb_select(priv
, &context_gl
->c
, state
);
675 if (&context_gl
->c
!= priv
->last_context
)
677 memset(priv
->vshader_const_dirty
, 1,
678 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
679 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
681 memset(priv
->pshader_const_dirty
, 1,
682 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
683 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
685 priv
->last_context
= &context_gl
->c
;
690 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
691 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
693 /* Load DirectX 9 float constants for vertex shader */
694 priv
->highest_dirty_vs_const
= shader_arb_load_constants_f(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
695 priv
->highest_dirty_vs_const
, state
->vs_consts_f
, priv
->vshader_const_dirty
);
696 shader_arb_vs_local_constants(gl_shader
, context_gl
, state
);
701 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
702 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
703 UINT rt_height
= state
->fb
.render_targets
[0]->height
;
705 /* Load DirectX 9 float constants for pixel shader */
706 priv
->highest_dirty_ps_const
= shader_arb_load_constants_f(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
707 priv
->highest_dirty_ps_const
, state
->ps_consts_f
, priv
->pshader_const_dirty
);
708 shader_arb_ps_local_constants(gl_shader
, context_gl
, state
, rt_height
);
710 if (context_gl
->c
.constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
711 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
715 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
716 const struct wined3d_state
*state
)
718 shader_arb_load_constants_internal(shader_priv
, wined3d_context_gl(context
),
719 state
, use_ps(state
), use_vs(state
), FALSE
);
722 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
724 struct wined3d_context_gl
*context_gl
= wined3d_context_gl_get_current();
725 struct shader_arb_priv
*priv
= device
->shader_priv
;
727 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
728 * context. On a context switch the old context will be fully dirtified */
729 if (!context_gl
|| context_gl
->c
.device
!= device
)
732 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
733 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
736 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
738 struct wined3d_context_gl
*context_gl
= wined3d_context_gl_get_current();
739 struct shader_arb_priv
*priv
= device
->shader_priv
;
741 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
742 * context. On a context switch the old context will be fully dirtified */
743 if (!context_gl
|| context_gl
->c
.device
!= device
)
746 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
747 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
750 static void shader_arb_append_imm_vec4(struct wined3d_string_buffer
*buffer
, const float *values
)
754 wined3d_ftoa(values
[0], str
[0]);
755 wined3d_ftoa(values
[1], str
[1]);
756 wined3d_ftoa(values
[2], str
[2]);
757 wined3d_ftoa(values
[3], str
[3]);
758 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
761 /* Generate the variable & register declarations for the ARB_vertex_program output target */
762 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
763 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_string_buffer
*buffer
,
764 const struct wined3d_gl_info
*gl_info
, DWORD
*num_clipplanes
,
765 const struct shader_arb_ctx_priv
*ctx
)
768 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
769 const struct wined3d_shader_lconst
*lconst
;
770 unsigned max_constantsF
;
773 /* In pixel shaders, all private constants are program local, we don't need anything
774 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
775 * If we need a private constant the GL implementation will squeeze it in somewhere
777 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
778 * immediate values. The posFixup is loaded using program.env for now, so always
779 * subtract one from the number of constants. If the shader uses indirect addressing,
780 * account for the helper const too because we have to declare all available d3d constants
781 * and don't know which are actually used.
785 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
786 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
787 if (max_constantsF
< 24)
788 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
792 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
793 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
794 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
795 * Also prevents max_constantsF from becoming less than 0 and
797 if (max_constantsF
< 96)
798 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
800 if (reg_maps
->usesrelconstF
)
802 DWORD highest_constf
= 0, clip_limit
;
804 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
805 max_constantsF
-= wined3d_popcount(reg_maps
->integer_constants
);
806 max_constantsF
-= gl_info
->reserved_arb_constants
;
808 for (i
= 0; i
< shader
->limits
->constant_float
; ++i
)
810 if (wined3d_bitmap_is_set(reg_maps
->constf
, i
))
814 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
816 if(ctx
->cur_vs_args
->super
.clip_enabled
)
817 clip_limit
= gl_info
->limits
.user_clip_distances
;
823 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
824 clip_limit
= min(wined3d_popcount(mask
), 4);
826 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
827 max_constantsF
-= *num_clipplanes
;
828 if(*num_clipplanes
< clip_limit
)
830 WARN("Only %u clip planes out of %u enabled.\n", *num_clipplanes
,
831 gl_info
->limits
.user_clip_distances
);
836 if (ctx
->target_version
>= NV2
)
837 *num_clipplanes
= gl_info
->limits
.user_clip_distances
;
839 *num_clipplanes
= min(gl_info
->limits
.user_clip_distances
, 4);
843 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
845 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
848 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
850 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
853 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
855 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
857 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
861 if (!shader
->load_local_constsF
)
863 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
866 value
= (const float *)lconst
->value
;
867 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
868 shader_arb_append_imm_vec4(buffer
, value
);
869 shader_addline(buffer
, ";\n");
873 /* After subtracting privately used constants from the hardware limit(they are loaded as
874 * local constants), make sure the shader doesn't violate the env constant limit
878 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
882 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
885 /* Avoid declaring more constants than needed */
886 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
888 /* we use the array-based constants array if the local constants are marked for loading,
889 * because then we use indirect addressing, or when the local constant list is empty,
890 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
891 * local constants do not declare the loaded constants as an array because ARB compilers usually
892 * do not optimize unused constants away
894 if (reg_maps
->usesrelconstF
)
896 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
897 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
898 max_constantsF
, max_constantsF
- 1);
902 for (i
= 0; i
< max_constantsF
; ++i
)
904 if (!shader_constant_is_local(shader
, i
) && wined3d_extract_bits(reg_maps
->constf
, i
, 1))
906 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
912 static const char * const shift_tab
[] = {
913 "dummy", /* 0 (none) */
914 "coefmul.x", /* 1 (x2) */
915 "coefmul.y", /* 2 (x4) */
916 "coefmul.z", /* 3 (x8) */
917 "coefmul.w", /* 4 (x16) */
918 "dummy", /* 5 (x32) */
919 "dummy", /* 6 (x64) */
920 "dummy", /* 7 (x128) */
921 "dummy", /* 8 (d256) */
922 "dummy", /* 9 (d128) */
923 "dummy", /* 10 (d64) */
924 "dummy", /* 11 (d32) */
925 "coefdiv.w", /* 12 (d16) */
926 "coefdiv.z", /* 13 (d8) */
927 "coefdiv.y", /* 14 (d4) */
928 "coefdiv.x" /* 15 (d2) */
931 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
932 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
934 char *ptr
= write_mask
;
936 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
939 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
940 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
941 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
942 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
948 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
950 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
951 * but addressed as "rgba". To fix this we need to swap the register's x
952 * and z components. */
953 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
954 char *ptr
= swizzle_str
;
956 /* swizzle bits fields: wwzzyyxx */
957 DWORD swizzle
= param
->swizzle
;
958 DWORD swizzle_x
= swizzle
& 0x03;
959 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
960 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
961 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
963 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
964 * generate a swizzle string. Unless we need to our own swizzling. */
965 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
968 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
969 *ptr
++ = swizzle_chars
[swizzle_x
];
971 *ptr
++ = swizzle_chars
[swizzle_x
];
972 *ptr
++ = swizzle_chars
[swizzle_y
];
973 *ptr
++ = swizzle_chars
[swizzle_z
];
974 *ptr
++ = swizzle_chars
[swizzle_w
];
981 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
983 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
984 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
986 if (!strcmp(priv
->addr_reg
, src
))
989 snprintf(priv
->addr_reg
, sizeof(priv
->addr_reg
), "%s", src
);
990 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
993 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
994 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
996 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
997 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
999 /* oPos, oFog and oPts in D3D */
1000 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
1001 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1002 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1003 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1004 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1010 case WINED3DSPR_TEMP
:
1011 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1014 case WINED3DSPR_INPUT
:
1017 if (reg_maps
->shader_version
.major
< 3)
1019 if (!reg
->idx
[0].offset
)
1020 strcpy(register_name
, "fragment.color.primary");
1022 strcpy(register_name
, "fragment.color.secondary");
1026 if (reg
->idx
[0].rel_addr
)
1029 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1031 if (!strcmp(rel_reg
, "**aL_emul**"))
1033 DWORD idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1034 if(idx
< MAX_REG_INPUT
)
1036 strcpy(register_name
, ctx
->ps_input
[idx
]);
1040 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1041 sprintf(register_name
, "out_of_bounds_%u", idx
);
1044 else if (reg_maps
->input_registers
& 0x0300)
1046 /* There are two ways basically:
1048 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1049 * That means trouble if the loop also contains a breakc or if the control values
1050 * aren't local constants.
1051 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1052 * source dynamically. The trouble is that we cannot simply read aL.y because it
1053 * is an ADDRESS register. We could however push it, load .zw with a value and use
1054 * ADAC to load the condition code register and pop it again afterwards
1056 FIXME("Relative input register addressing with more than 8 registers\n");
1058 /* This is better than nothing for now */
1059 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1061 else if(ctx
->cur_ps_args
->super
.vp_mode
!= WINED3D_VP_MODE_SHADER
)
1063 /* This is problematic because we'd have to consult the ctx->ps_input strings
1064 * for where to find the varying. Some may be "0.0", others can be texcoords or
1065 * colors. This needs either a pipeline replacement to make the vertex shader feed
1066 * proper varyings, or loop unrolling
1068 * For now use the texcoords and hope for the best
1070 FIXME("Non-vertex shader varying input with indirect addressing\n");
1071 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1075 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1076 * pulls GL_NV_fragment_program2 in
1078 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1083 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1085 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1089 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1090 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1097 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1u << reg
->idx
[0].offset
))
1099 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1103 case WINED3DSPR_CONST
:
1104 if (!pshader
&& reg
->idx
[0].rel_addr
)
1106 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1107 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1110 if (reg_maps
->shader_version
.major
< 2)
1112 sprintf(rel_reg
, "A0.x");
1116 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1117 if (ctx
->target_version
== ARB
)
1119 if (!strcmp(rel_reg
, "**aL_emul**"))
1123 shader_arb_request_a0(ins
, rel_reg
);
1124 sprintf(rel_reg
, "A0.x");
1129 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1130 else if (reg
->idx
[0].offset
>= rel_offset
)
1131 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1133 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1137 if (reg_maps
->usesrelconstF
)
1138 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1140 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1144 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1147 if (reg_maps
->shader_version
.major
== 1
1148 && reg_maps
->shader_version
.minor
<= 3)
1149 /* In ps <= 1.3, Tx is a temporary register as destination
1150 * to all instructions, and as source to most instructions.
1151 * For some instructions it is the texcoord input. Those
1152 * instructions know about the special use. */
1153 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1155 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1156 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1160 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1161 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1163 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1167 case WINED3DSPR_COLOROUT
:
1168 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1170 strcpy(register_name
, "TMP_COLOR");
1174 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1175 FIXME("sRGB correction on higher render targets.\n");
1176 if (reg_maps
->rt_mask
> 1)
1177 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1179 strcpy(register_name
, "result.color");
1183 case WINED3DSPR_RASTOUT
:
1184 if (reg
->idx
[0].offset
== 1)
1185 sprintf(register_name
, "%s", ctx
->fog_output
);
1187 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1190 case WINED3DSPR_DEPTHOUT
:
1191 strcpy(register_name
, "result.depth");
1194 case WINED3DSPR_ATTROUT
:
1195 /* case WINED3DSPR_OUTPUT: */
1197 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1199 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1202 case WINED3DSPR_TEXCRDOUT
:
1204 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1205 else if (reg_maps
->shader_version
.major
< 3)
1206 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1208 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1211 case WINED3DSPR_LOOP
:
1212 if(ctx
->target_version
>= NV2
)
1214 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1215 if(pshader
) sprintf(register_name
, "A0.x");
1216 else sprintf(register_name
, "aL.y");
1220 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1221 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1222 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1223 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1226 sprintf(register_name
, "**aL_emul**");
1231 case WINED3DSPR_CONSTINT
:
1232 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1235 case WINED3DSPR_MISCTYPE
:
1236 if (!reg
->idx
[0].offset
)
1237 sprintf(register_name
, "vpos");
1238 else if (reg
->idx
[0].offset
== 1)
1239 sprintf(register_name
, "fragment.facing.x");
1241 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1245 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1246 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1251 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1252 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1254 char register_name
[255];
1258 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1259 strcpy(str
, register_name
);
1261 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1262 strcat(str
, write_mask
);
1265 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1267 switch(channel_source
)
1269 case CHANNEL_SOURCE_ZERO
: return "0";
1270 case CHANNEL_SOURCE_ONE
: return "1";
1271 case CHANNEL_SOURCE_X
: return "x";
1272 case CHANNEL_SOURCE_Y
: return "y";
1273 case CHANNEL_SOURCE_Z
: return "z";
1274 case CHANNEL_SOURCE_W
: return "w";
1276 FIXME("Unhandled channel source %#x\n", channel_source
);
1281 struct color_fixup_masks
1287 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1289 struct color_fixup_masks masks
= {0, 0};
1291 if (is_complex_fixup(fixup
))
1293 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1294 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1298 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1299 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1300 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1301 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1302 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1303 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1304 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1305 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1306 masks
.source
&= dst_mask
;
1308 if (fixup
.x_sign_fixup
)
1309 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1310 if (fixup
.y_sign_fixup
)
1311 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1312 if (fixup
.z_sign_fixup
)
1313 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1314 if (fixup
.w_sign_fixup
)
1315 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1316 masks
.sign
&= dst_mask
;
1321 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1322 const char *src
, const char *one
, const char *two
,
1323 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1325 const char *sign_fixup_src
= dst
;
1330 sign_fixup_src
= "TA";
1332 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1333 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1334 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1336 else if (masks
.sign
)
1338 sign_fixup_src
= src
;
1344 char *ptr
= reg_mask
;
1346 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1349 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1351 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1353 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1355 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1360 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1364 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1367 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1368 if (!ins
->dst_count
) return "";
1370 mod
= ins
->dst
[0].modifiers
;
1372 /* Silently ignore PARTIALPRECISION if it's not supported */
1373 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1375 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1377 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1378 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1383 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1386 case WINED3DSPDM_SATURATE
:
1389 case WINED3DSPDM_PARTIALPRECISION
:
1396 FIXME("Unknown modifiers 0x%08x\n", mod
);
1401 #define TEX_PROJ 0x1
1402 #define TEX_BIAS 0x2
1404 #define TEX_DERIV 0x10
1406 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1407 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1409 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1410 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1411 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1412 enum wined3d_shader_resource_type resource_type
;
1413 struct color_fixup_masks masks
;
1414 const char *tex_dst
= dst_str
;
1415 BOOL np2_fixup
= FALSE
;
1416 const char *tex_type
;
1421 resource_type
= pixelshader_get_resource_type(ins
->ctx
->reg_maps
, sampler_idx
,
1422 priv
->cur_ps_args
->super
.tex_types
);
1426 resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1428 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1429 sampler_idx
+= WINED3D_MAX_FRAGMENT_SAMPLERS
;
1432 switch (resource_type
)
1434 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1438 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1439 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1u << sampler_idx
)
1440 && priv
->gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1447 if (priv
->cur_np2fixup_info
->super
.active
& (1u << sampler_idx
))
1449 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1450 else np2_fixup
= TRUE
;
1455 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1459 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1464 ERR("Unexpected resource type %#x.\n", resource_type
);
1468 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1469 * so don't use shader_arb_get_modifier
1471 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1474 /* Fragment samplers always have identity mapping */
1475 if(sampler_idx
>= WINED3D_MAX_FRAGMENT_SAMPLERS
)
1477 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- WINED3D_MAX_FRAGMENT_SAMPLERS
];
1482 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1483 ins
->dst
[0].write_mask
);
1485 if (masks
.source
|| masks
.sign
)
1489 if (flags
& TEX_DERIV
)
1491 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1492 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1493 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1494 dsx
, dsy
, sampler_idx
, tex_type
);
1496 else if(flags
& TEX_LOD
)
1498 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1499 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1500 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1501 sampler_idx
, tex_type
);
1503 else if (flags
& TEX_BIAS
)
1505 /* Shouldn't be possible, but let's check for it */
1506 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1507 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1508 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1510 else if (flags
& TEX_PROJ
)
1512 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1518 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1519 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1520 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1522 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1525 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1530 gen_color_correction(buffer
, dst_str
, tex_dst
,
1531 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1532 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1533 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1537 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1538 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1540 /* Generate a line that does the input modifier computation and return the input register to use */
1541 BOOL is_color
= FALSE
, insert_line
;
1544 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1545 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1546 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1547 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1549 /* Assume a new line will be added */
1552 /* Get register name */
1553 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1554 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1556 switch (src
->modifiers
)
1558 case WINED3DSPSM_NONE
:
1559 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1560 insert_line
= FALSE
;
1562 case WINED3DSPSM_NEG
:
1563 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1564 insert_line
= FALSE
;
1566 case WINED3DSPSM_BIAS
:
1567 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1569 case WINED3DSPSM_BIASNEG
:
1570 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1572 case WINED3DSPSM_SIGN
:
1573 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1575 case WINED3DSPSM_SIGNNEG
:
1576 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1578 case WINED3DSPSM_COMP
:
1579 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1581 case WINED3DSPSM_X2
:
1582 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1584 case WINED3DSPSM_X2NEG
:
1585 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1587 case WINED3DSPSM_DZ
:
1588 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1589 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1591 case WINED3DSPSM_DW
:
1592 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1593 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1595 case WINED3DSPSM_ABS
:
1596 if(ctx
->target_version
>= NV2
) {
1597 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1598 insert_line
= FALSE
;
1600 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1603 case WINED3DSPSM_ABSNEG
:
1604 if(ctx
->target_version
>= NV2
) {
1605 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1607 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1608 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1610 insert_line
= FALSE
;
1613 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1614 insert_line
= FALSE
;
1617 /* Return modified or original register, with swizzle */
1619 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1622 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1624 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1625 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1626 DWORD sampler_code
= dst
->reg
.idx
[0].offset
;
1628 char src_name
[2][50];
1630 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1632 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1634 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1635 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1638 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1639 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1640 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1641 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1642 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1644 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1645 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1648 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1653 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1654 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1655 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1656 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1657 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1658 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1659 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1660 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1661 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1662 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1663 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1664 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1665 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1667 FIXME("Unknown modifier %u\n", mod
);
1671 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1673 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1674 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1676 char src_name
[3][50];
1677 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1678 ins
->ctx
->reg_maps
->shader_version
.minor
);
1680 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1681 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1683 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1684 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1686 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1690 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1693 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1694 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1696 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1697 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1698 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1699 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1700 dst_name
, src_name
[1], src_name
[2]);
1704 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1706 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1707 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1709 char src_name
[3][50];
1711 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1713 /* Generate input register names (with modifiers) */
1714 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1715 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1716 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1718 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1719 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1722 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1723 * dst = dot2(src0, src1) + src2 */
1724 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1726 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1727 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1729 char src_name
[3][50];
1730 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1732 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1733 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1734 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1736 if(ctx
->target_version
>= NV3
)
1738 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1739 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1740 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1741 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1743 else if(ctx
->target_version
>= NV2
)
1745 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1746 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1747 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1748 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1750 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1752 * .xyxy and other swizzles that we could get with this are not valid in
1753 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1755 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1756 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1757 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1759 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1761 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1762 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1766 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1767 /* 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]
1768 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1770 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1771 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1772 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1773 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1777 /* Map the opcode 1-to-1 to the GL code */
1778 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1780 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1781 const char *instruction
;
1782 char arguments
[256], dst_str
[50];
1784 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1786 switch (ins
->handler_idx
)
1788 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1789 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1790 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1791 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1792 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1793 case WINED3DSIH_DST
: instruction
= "DST"; break;
1794 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1795 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1796 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1797 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1798 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1799 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1800 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1801 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1802 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1803 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1804 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1805 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1806 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1807 default: instruction
= "";
1808 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
1812 /* Note that shader_arb_add_dst_param() adds spaces. */
1813 arguments
[0] = '\0';
1814 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1815 for (i
= 0; i
< ins
->src_count
; ++i
)
1818 strcat(arguments
, ", ");
1819 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1820 strcat(arguments
, operand
);
1822 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1825 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1827 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1829 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1832 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1834 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1835 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1836 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1837 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1838 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1839 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1840 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1842 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1843 char src0_param
[256];
1845 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1847 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1849 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1851 if(ctx
->target_version
>= NV2
) {
1852 shader_hw_map2gl(ins
);
1855 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1856 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1858 /* This implements the mova formula used in GLSL. The first two instructions
1859 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1863 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1865 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1866 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1868 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1869 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1871 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1872 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1873 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1874 if (shader_data
->rel_offset
)
1876 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1878 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1880 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1882 else if (reg_maps
->shader_version
.major
== 1
1883 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1884 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1886 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1887 src0_param
[0] = '\0';
1889 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1891 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1892 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1893 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1894 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1898 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1899 * with more than one component. Thus replicate the first source argument over all
1900 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1901 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1902 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1903 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1904 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1907 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1909 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1911 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1914 shader_hw_map2gl(ins
);
1918 shader_hw_map2gl(ins
);
1922 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1924 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1925 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1928 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1929 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1931 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1933 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1935 const char *kilsrc
= "TA";
1938 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1939 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1945 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1946 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1947 * masked out components to 0(won't kill)
1949 char x
= '0', y
= '0', z
= '0', w
= '0';
1950 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1951 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1952 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1953 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1954 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1956 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1960 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1961 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1963 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1964 * or pass in any temporary register(in shader phase 2)
1966 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1967 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1969 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1970 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1971 shader_addline(buffer
, "KIL TA;\n");
1975 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1977 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1978 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1979 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1980 ins
->ctx
->reg_maps
->shader_version
.minor
);
1981 struct wined3d_shader_src_param src
;
1985 DWORD reg_sampler_code
;
1987 BOOL swizzle_coord
= FALSE
;
1989 /* All versions have a destination register */
1990 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1992 /* 1.0-1.4: Use destination register number as texture code.
1993 2.0+: Use provided sampler number as texture code. */
1994 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1995 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1997 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1999 /* 1.0-1.3: Use the texcoord varying.
2000 1.4+: Use provided coordinate source register. */
2001 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2002 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
2004 /* TEX is the only instruction that can handle DW and DZ natively */
2006 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
2007 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
2008 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
2012 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
2013 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
2014 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2016 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2019 if (reg_sampler_code
< WINED3D_MAX_TEXTURES
)
2020 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2021 if (flags
& WINED3D_PSARGS_PROJECTED
)
2023 myflags
|= TEX_PROJ
;
2024 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2025 swizzle_coord
= TRUE
;
2028 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2030 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2031 if (src_mod
== WINED3DSPSM_DZ
)
2033 swizzle_coord
= TRUE
;
2034 myflags
|= TEX_PROJ
;
2035 } else if(src_mod
== WINED3DSPSM_DW
) {
2036 myflags
|= TEX_PROJ
;
2039 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2040 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2045 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2046 * reg_coord is a read-only varying register, so we need a temp reg */
2047 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2048 strcpy(reg_coord
, "TA");
2051 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2054 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2056 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2057 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2058 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2059 ins
->ctx
->reg_maps
->shader_version
.minor
);
2062 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2064 DWORD reg
= dst
->reg
.idx
[0].offset
;
2066 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2067 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2071 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2072 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2073 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2077 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2079 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2082 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2086 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2087 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2088 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2089 /* Move .x first in case src_str is "TA" */
2090 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2091 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2092 if (reg1
< WINED3D_MAX_TEXTURES
)
2094 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2095 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2097 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2100 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2102 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2104 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2108 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2109 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2110 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2111 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2112 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2113 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2116 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2118 DWORD reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2122 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2123 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2124 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2125 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2128 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2130 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2131 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2132 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2133 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2134 DWORD reg_dest_code
;
2136 /* All versions have a destination register. The Tx where the texture coordinates come
2137 * from is the varying incarnation of the texture register
2139 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2140 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2141 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2142 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2144 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2145 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2147 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2148 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2150 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2151 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2154 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2155 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2156 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2157 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2159 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2160 * so we can't let the GL handle this.
2162 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2163 & WINED3D_PSARGS_PROJECTED
)
2165 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2166 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2167 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2169 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2172 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2174 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2176 /* No src swizzles are allowed, so this is ok */
2177 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2178 src_reg
, reg_dest_code
, reg_dest_code
);
2179 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2183 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2185 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2186 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2187 char src0_name
[50], dst_name
[50];
2189 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2191 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2192 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2193 * T<reg+1> register. Use this register to store the calculated vector
2195 tmp_reg
.idx
[0].offset
= reg
+ 1;
2196 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2197 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2200 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2202 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2204 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2205 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2211 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2212 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2214 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2215 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2216 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2217 flags
= reg
< WINED3D_MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2218 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2221 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2223 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2224 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2225 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2226 char src0_name
[50], dst_name
[50];
2227 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2230 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2231 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2232 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2234 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2235 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2237 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2238 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2239 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2240 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2243 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2245 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2246 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2248 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2249 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2251 char src0_name
[50], dst_name
[50];
2254 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2255 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2256 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2258 /* Sample the texture using the calculated coordinates */
2259 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2260 flags
= reg
< WINED3D_MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2261 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2262 tex_mx
->current_row
= 0;
2265 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2267 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2268 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2270 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2271 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2277 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2278 * components for temporary data storage
2280 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2281 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2282 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2284 /* Construct the eye-ray vector from w coordinates */
2285 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2286 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2287 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2289 /* Calculate reflection vector
2291 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2292 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2293 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2294 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2295 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2296 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2297 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2299 /* Sample the texture using the calculated coordinates */
2300 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2301 flags
= reg
< WINED3D_MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2302 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2303 tex_mx
->current_row
= 0;
2306 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2308 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2309 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2311 DWORD reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2312 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2319 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2320 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2321 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2322 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2323 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2325 /* Calculate reflection vector.
2328 * dst_reg.xyz = 2 * --------- * N - E
2331 * Which normalizes the normal vector
2333 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2334 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2335 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2336 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2337 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2338 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2340 /* Sample the texture using the calculated coordinates */
2341 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2342 flags
= reg
< WINED3D_MAX_TEXTURES
? priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
: 0;
2343 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2344 tex_mx
->current_row
= 0;
2347 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2349 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2350 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2352 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2353 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2355 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2356 * which is essentially an input, is the destination register because it is the first
2357 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2358 * here(writemasks/swizzles are not valid on texdepth)
2360 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2362 /* According to the msdn, the source register(must be r5) is unusable after
2363 * the texdepth instruction, so we're free to modify it
2365 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2367 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2368 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2369 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2371 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2372 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2373 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2374 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2377 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2378 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2379 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2380 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2382 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2383 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2387 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2388 shader_addline(buffer
, "MOV TB, 0.0;\n");
2389 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2391 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2392 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2395 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2396 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2397 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2399 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2402 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2404 /* Handle output register */
2405 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2406 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2407 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2410 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2411 * Perform the 3rd row of a 3x3 matrix multiply */
2412 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2414 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2415 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2416 char dst_str
[50], dst_name
[50];
2420 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2421 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2422 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2423 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2424 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2427 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2428 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2429 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2430 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2432 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2434 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2435 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2436 char src0
[50], dst_name
[50];
2438 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2439 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2441 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2442 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2443 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2445 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2446 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2447 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2449 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2450 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2451 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2452 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2455 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2456 Vertex/Pixel shaders to ARB_vertex_program codes */
2457 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2460 int nComponents
= 0;
2461 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2462 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2463 struct wined3d_shader_instruction tmp_ins
;
2465 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2467 /* Set constants for the temporary argument */
2468 tmp_ins
.ctx
= ins
->ctx
;
2469 tmp_ins
.dst_count
= 1;
2470 tmp_ins
.dst
= &tmp_dst
;
2471 tmp_ins
.src_count
= 2;
2472 tmp_ins
.src
= tmp_src
;
2474 switch(ins
->handler_idx
)
2476 case WINED3DSIH_M4x4
:
2478 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2480 case WINED3DSIH_M4x3
:
2482 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2484 case WINED3DSIH_M3x4
:
2486 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2488 case WINED3DSIH_M3x3
:
2490 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2492 case WINED3DSIH_M3x2
:
2494 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2497 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2501 tmp_dst
= ins
->dst
[0];
2502 tmp_src
[0] = ins
->src
[0];
2503 tmp_src
[1] = ins
->src
[1];
2504 for (i
= 0; i
< nComponents
; ++i
)
2506 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2507 shader_hw_map2gl(&tmp_ins
);
2508 ++tmp_src
[1].reg
.idx
[0].offset
;
2512 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2518 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2519 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2520 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2521 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2522 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2523 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2524 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2525 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2526 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2527 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2528 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2529 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2530 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2532 FIXME("Unknown modifier %u\n", mod
);
2536 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2538 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2539 const char *instruction
;
2540 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2541 BOOL need_abs
= FALSE
;
2546 switch(ins
->handler_idx
)
2548 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2549 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2550 case WINED3DSIH_EXPP
:
2551 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2553 instruction
= "EXP";
2557 case WINED3DSIH_EXP
:
2558 instruction
= "EX2";
2560 case WINED3DSIH_LOG
:
2561 case WINED3DSIH_LOGP
:
2562 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2563 * instruction, but notice that the output of those instructions is
2565 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2566 instruction
= "LG2";
2568 default: instruction
= "";
2569 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2573 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2575 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2577 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2578 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2582 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2583 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2587 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2592 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2594 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2597 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2598 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2599 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2601 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2602 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2604 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2605 * otherwise NRM or RSQ would return NaN */
2606 if(pshader
&& priv
->target_version
>= NV3
)
2608 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2610 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2612 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2613 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2614 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2616 else if(priv
->target_version
>= NV2
)
2618 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2619 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2620 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2625 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2627 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2628 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2629 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2631 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2632 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2634 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2635 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2636 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2641 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2643 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2645 char src_name
[3][50];
2647 /* ARB_fragment_program has a convenient LRP instruction */
2648 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2649 shader_hw_map2gl(ins
);
2653 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2654 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2655 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2656 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2658 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2659 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2660 dst_name
, src_name
[0], src_name
[2]);
2663 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2665 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2666 * must contain fixed constants. So we need a separate function to filter those constants and
2669 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2670 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2671 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2673 char src_name0
[50], src_name1
[50], src_name2
[50];
2676 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2677 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2678 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2679 /* No modifiers are supported on SCS */
2680 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2682 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2684 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2685 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2687 } else if(priv
->target_version
>= NV2
) {
2688 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2690 /* Sincos writemask must be .x, .y or .xy */
2691 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2692 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2693 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2694 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2696 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2697 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2699 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2700 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2702 * The constants we get are:
2704 * +1 +1, -1 -1 +1 +1 -1 -1
2705 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2706 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2708 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2712 * (x/2)^4 = x^4 / 16
2713 * (x/2)^5 = x^5 / 32
2716 * To get the final result:
2717 * sin(x) = 2 * sin(x/2) * cos(x/2)
2718 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2719 * (from sin(x+y) and cos(x+y) rules)
2721 * As per MSDN, dst.z is undefined after the operation, and so is
2722 * dst.x and dst.y if they're masked out by the writemask. Ie
2723 * sincos dst.y, src1, c0, c1
2724 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2725 * vsa.exe also stops with an error if the dest register is the same register as the source
2726 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2727 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2729 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2730 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2731 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2733 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2734 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2735 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2736 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2737 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2738 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2742 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2743 * properly merge that with MULs in the code above?
2744 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2745 * we can merge the sine and cosine MAD rows to calculate them together.
2747 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2748 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2749 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2750 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2753 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2754 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2755 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2757 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2759 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2760 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2762 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2764 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2765 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2770 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2772 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2775 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2777 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2778 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2780 /* SGN is only valid in vertex shaders */
2781 if(ctx
->target_version
>= NV2
) {
2782 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2786 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2787 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2789 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2790 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2792 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2793 * Then use TA, and calculate the final result
2795 * Not reading from TA? Store the first result in TA to avoid overwriting the
2796 * destination if src reg = dst reg
2798 if(strstr(src_name
, "TA"))
2800 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2801 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2802 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2806 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2807 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2808 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2813 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2815 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2821 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2822 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2823 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2825 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2826 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2829 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2831 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2832 char src0
[50], src1
[50], dst
[50];
2833 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2834 BOOL need_abs
= FALSE
;
2835 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2836 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2838 /* POW operates on the absolute value of the input */
2839 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2841 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2842 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2843 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2846 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2848 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2850 if (priv
->target_version
>= NV2
)
2852 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2853 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2854 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2858 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2859 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2861 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2862 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2863 /* Possibly add flt_eps to avoid getting float special values */
2864 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2865 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2866 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2867 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2871 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2873 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2875 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2878 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2882 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2883 struct list
*e
= list_head(&priv
->control_frames
);
2884 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2886 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2887 /* The constant loader makes sure to load -1 into iX.w */
2888 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2889 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2890 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2894 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2898 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2900 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2902 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2904 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2906 /* The constant loader makes sure to load -1 into iX.w */
2909 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2910 struct list
*e
= list_head(&priv
->control_frames
);
2911 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2913 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2915 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2916 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2917 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2921 shader_addline(buffer
, "REP %s;\n", src_name
);
2925 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2927 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2928 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2932 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2933 struct list
*e
= list_head(&priv
->control_frames
);
2934 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2936 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2937 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2938 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2940 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2944 shader_addline(buffer
, "ENDLOOP;\n");
2948 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2950 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2951 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2955 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2956 struct list
*e
= list_head(&priv
->control_frames
);
2957 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2959 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2960 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2961 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2963 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2967 shader_addline(buffer
, "ENDREP;\n");
2971 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2973 struct control_frame
*control_frame
;
2975 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2977 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2979 ERR("Could not find loop for break\n");
2983 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2985 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2986 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2987 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2991 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2995 shader_addline(buffer
, "BRK;\n");
2999 static const char *get_compare(enum wined3d_shader_rel_op op
)
3003 case WINED3D_SHADER_REL_OP_GT
: return "GT";
3004 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
3005 case WINED3D_SHADER_REL_OP_GE
: return "GE";
3006 case WINED3D_SHADER_REL_OP_LT
: return "LT";
3007 case WINED3D_SHADER_REL_OP_NE
: return "NE";
3008 case WINED3D_SHADER_REL_OP_LE
: return "LE";
3010 FIXME("Unrecognized operator %#x.\n", op
);
3015 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3019 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3020 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3021 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3022 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3023 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3024 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3026 FIXME("Unrecognized operator %#x.\n", op
);
3031 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3033 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3034 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3035 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3038 const char *comp
= get_compare(ins
->flags
);
3040 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3041 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3045 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3046 * away the subtraction result
3048 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3049 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3053 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3054 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3058 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3060 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3061 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3062 struct list
*e
= list_head(&priv
->control_frames
);
3063 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3067 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3069 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3070 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3074 /* Invert the flag. We jump to the else label if the condition is NOT true */
3075 comp
= get_compare(invert_compare(ins
->flags
));
3076 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3077 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3081 comp
= get_compare(ins
->flags
);
3082 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3083 shader_addline(buffer
, "IF %s.x;\n", comp
);
3087 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3089 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3090 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3091 struct list
*e
= list_head(&priv
->control_frames
);
3092 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3093 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3097 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3098 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3099 control_frame
->had_else
= TRUE
;
3103 shader_addline(buffer
, "ELSE;\n");
3107 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3109 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3110 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3111 struct list
*e
= list_head(&priv
->control_frames
);
3112 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3113 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3117 if(control_frame
->had_else
)
3119 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3123 shader_addline(buffer
, "#No else branch. else is endif\n");
3124 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3129 shader_addline(buffer
, "ENDIF;\n");
3133 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3135 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3137 char reg_src
[3][40];
3138 WORD flags
= TEX_DERIV
;
3140 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3141 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3142 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3143 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3145 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3146 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3148 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3151 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3153 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3156 WORD flags
= TEX_LOD
;
3158 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3159 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3161 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3162 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3164 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3167 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3169 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3170 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3172 priv
->in_main_func
= FALSE
;
3173 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3174 * subroutine, don't generate a label that will make GL complain
3176 if(priv
->target_version
== ARB
) return;
3178 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3181 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3182 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3183 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3184 struct wined3d_string_buffer
*buffer
)
3188 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3189 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3190 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3191 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3193 if (args
->super
.fog_src
== VS_FOG_Z
)
3195 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3201 /* posFixup.x is always 1.0, so we can safely use it */
3202 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3206 /* Clamp fogcoord */
3207 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3208 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3210 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3211 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3215 /* Clipplanes are always stored without y inversion */
3216 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3218 if (args
->super
.clip_enabled
)
3220 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3222 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3226 else if (args
->clip
.boolclip
.clip_texcoord
)
3228 static const char component
[4] = {'x', 'y', 'z', 'w'};
3229 unsigned int cur_clip
= 0;
3230 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3232 for (i
= 0; i
< gl_info
->limits
.user_clip_distances
; ++i
)
3234 if (args
->clip
.boolclip
.clipplane_mask
& (1u << i
))
3236 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3237 component
[cur_clip
++], i
);
3243 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3246 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3249 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3252 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3255 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3256 args
->clip
.boolclip
.clip_texcoord
- 1);
3259 /* Write the final position.
3261 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3262 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3263 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3264 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3266 if (!gl_info
->supported
[ARB_CLIP_CONTROL
])
3268 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3269 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3270 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3272 /* Z coord [0;1]->[-1;1] mapping, see comment in
3273 * get_projection_matrix() in utils.c. */
3274 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3276 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3277 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3281 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3282 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3286 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3288 priv_ctx
->footer_written
= TRUE
;
3291 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3293 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3294 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3295 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3296 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3298 if(priv
->target_version
== ARB
) return;
3302 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3303 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, priv
->gl_info
, buffer
);
3306 shader_addline(buffer
, "RET;\n");
3309 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3311 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3312 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3315 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3317 const char *ptr
, *line
;
3320 if (TRACE_ON(d3d_shader
))
3323 while ((line
= get_line(&ptr
))) TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3326 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3327 checkGLcall("glProgramStringARB()");
3329 if (FIXME_ON(d3d_shader
))
3331 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3334 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3335 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3337 while ((line
= get_line(&ptr
))) FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3338 FIXME_(d3d_shader
)("\n");
3344 if (WARN_ON(d3d_perf
))
3346 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3347 checkGLcall("glGetProgramivARB()");
3349 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3355 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3356 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3358 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3362 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3363 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3364 /* Calculate the > 0.0031308 case */
3365 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3366 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3367 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3368 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3369 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3370 /* Calculate the < case */
3371 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3375 /* Calculate the > 0.0031308 case */
3376 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3377 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3378 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3379 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3380 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3381 /* Calculate the < case */
3382 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3383 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3384 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3385 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3386 /* Store the components > 0.0031308 in the destination */
3387 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3388 /* Add the components that are < 0.0031308 */
3389 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3390 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3391 * result.color writes(.rgb first, then .a), or handle overwriting already written
3392 * components. The assembler uses a temporary register in this case, which is usually
3393 * not allocated from one of our registers that were used earlier.
3396 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3399 static const DWORD
*find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3401 const struct wined3d_shader_lconst
*constant
;
3403 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3405 if (constant
->idx
== idx
)
3407 return constant
->value
;
3413 static void init_ps_input(const struct wined3d_shader
*shader
,
3414 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3416 static const char * const texcoords
[8] =
3418 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3419 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3422 const struct wined3d_shader_signature_element
*input
;
3423 const char *semantic_name
;
3426 if (args
->super
.vp_mode
== WINED3D_VP_MODE_SHADER
)
3428 /* That one is easy. The vertex shaders provide v0-v7 in
3429 * fragment.texcoord and v8 and v9 in fragment.color. */
3430 for (i
= 0; i
< 8; ++i
)
3432 priv
->ps_input
[i
] = texcoords
[i
];
3434 priv
->ps_input
[8] = "fragment.color.primary";
3435 priv
->ps_input
[9] = "fragment.color.secondary";
3439 /* The fragment shader has to collect the varyings on its own. In any case
3440 * properly load color0 and color1. In the case of pre-transformed
3441 * vertices also load texture coordinates. Set other attributes to 0.0.
3443 * For fixed-function this behavior is correct, according to the tests.
3444 * For pre-transformed we'd either need a replacement shader that can load
3445 * other attributes like BINORMAL, or load the texture coordinate
3446 * attribute pointers to match the fragment shader signature. */
3447 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3449 input
= &shader
->input_signature
.elements
[i
];
3450 if (!(semantic_name
= input
->semantic_name
))
3452 semantic_idx
= input
->semantic_idx
;
3454 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3457 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3458 else if (semantic_idx
== 1)
3459 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3461 priv
->ps_input
[input
->register_idx
] = "0.0";
3463 else if (args
->super
.vp_mode
== WINED3D_VP_MODE_FF
)
3465 priv
->ps_input
[input
->register_idx
] = "0.0";
3467 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3469 if (semantic_idx
< 8)
3470 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3472 priv
->ps_input
[input
->register_idx
] = "0.0";
3474 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3477 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3479 priv
->ps_input
[input
->register_idx
] = "0.0";
3483 priv
->ps_input
[input
->register_idx
] = "0.0";
3486 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3487 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3491 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3492 const char *fragcolor
, const char *tmp
)
3494 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3495 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3496 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3499 /* Context activation is done by the caller. */
3500 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3501 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3502 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3504 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3507 DWORD next_local
= 0;
3508 struct shader_arb_ctx_priv priv_ctx
;
3509 BOOL dcl_td
= FALSE
;
3510 BOOL want_nv_prog
= FALSE
;
3511 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3513 BOOL custom_linear_fog
= FALSE
;
3517 unsigned int i
, found
= 0;
3519 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3522 || (shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3523 || (reg_maps
->shader_version
.major
< 2 && !i
))
3526 sprintf(srgbtmp
[found
], "R%u", i
);
3528 if (found
== 4) break;
3533 sprintf(srgbtmp
[0], "TA");
3534 sprintf(srgbtmp
[1], "TB");
3535 sprintf(srgbtmp
[2], "TC");
3536 sprintf(srgbtmp
[3], "TD");
3540 sprintf(srgbtmp
[1], "TA");
3541 sprintf(srgbtmp
[2], "TB");
3542 sprintf(srgbtmp
[3], "TC");
3545 sprintf(srgbtmp
[2], "TA");
3546 sprintf(srgbtmp
[3], "TB");
3549 sprintf(srgbtmp
[3], "TA");
3555 /* Create the hw ARB shader */
3556 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3557 priv_ctx
.gl_info
= gl_info
;
3558 priv_ctx
.cur_ps_args
= args
;
3559 priv_ctx
.compiled_fprog
= compiled
;
3560 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3561 init_ps_input(shader
, args
, &priv_ctx
);
3562 list_init(&priv_ctx
.control_frames
);
3563 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3565 /* Avoid enabling NV_fragment_program* if we do not need it.
3567 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3568 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3569 * is faster than what we gain from using higher native instructions. There are some things though
3570 * that cannot be emulated. In that case enable the extensions.
3571 * If the extension is enabled, instruction handlers that support both ways will use it.
3573 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3574 * So enable the best we can get.
3576 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3577 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3579 want_nv_prog
= TRUE
;
3582 shader_addline(buffer
, "!!ARBfp1.0\n");
3583 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3585 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3586 priv_ctx
.target_version
= NV3
;
3588 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3590 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3591 priv_ctx
.target_version
= NV2
;
3595 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3598 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3601 priv_ctx
.target_version
= ARB
;
3604 if (reg_maps
->rt_mask
> 1)
3606 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3609 if (reg_maps
->shader_version
.major
< 3)
3611 switch (args
->super
.fog
)
3613 case WINED3D_FFP_PS_FOG_OFF
:
3615 case WINED3D_FFP_PS_FOG_LINEAR
:
3616 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3618 custom_linear_fog
= TRUE
;
3619 priv_ctx
.ps_post_process
= TRUE
;
3622 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3624 case WINED3D_FFP_PS_FOG_EXP
:
3625 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3627 case WINED3D_FFP_PS_FOG_EXP2
:
3628 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3633 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3634 * unused temps away(but occupies them for the whole shader if they're used once). Always
3635 * declaring them avoids tricky bookkeeping work
3637 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3638 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3639 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3640 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3641 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3642 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3643 wined3d_ftoa(eps
, ftoa_tmp
);
3644 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3646 if (reg_maps
->shader_version
.major
< 2)
3648 strcpy(fragcolor
, "R0");
3652 if (priv_ctx
.ps_post_process
)
3654 if (shader
->u
.ps
.color0_mov
)
3656 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3660 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3661 strcpy(fragcolor
, "TMP_COLOR");
3664 strcpy(fragcolor
, "result.color");
3668 if (args
->super
.srgb_correction
)
3670 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3671 shader_arb_append_imm_vec4(buffer
, &wined3d_srgb_const
[0].x
);
3672 shader_addline(buffer
, ";\n");
3673 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3674 shader_arb_append_imm_vec4(buffer
, &wined3d_srgb_const
[1].x
);
3675 shader_addline(buffer
, ";\n");
3678 /* Base Declarations */
3679 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3681 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3683 unsigned char bump_const
;
3685 if (!(map
& 1)) continue;
3687 bump_const
= compiled
->numbumpenvmatconsts
;
3688 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3689 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3690 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3691 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3693 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3694 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3695 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3696 * textures due to conditional NP2 restrictions)
3698 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3699 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3700 * their location is shader dependent anyway and they cannot be loaded globally.
3702 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3703 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3704 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3705 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3707 if (!(reg_maps
->luminanceparams
& (1u << i
)))
3710 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3711 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3712 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3715 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
3717 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3718 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
3720 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
3724 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3725 control_values
[0], control_values
[1], control_values
[2]);
3729 compiled
->int_consts
[i
] = next_local
;
3730 compiled
->num_int_consts
++;
3731 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3736 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3738 compiled
->ycorrection
= next_local
;
3739 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3743 shader_addline(buffer
, "TEMP vpos;\n");
3744 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3745 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3746 * ycorrection.z: 1.0
3747 * ycorrection.w: 0.0
3749 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3750 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3755 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3758 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3759 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3760 * at most 8 (WINED3D_MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3761 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3762 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3763 * shader compilation errors and the subsequent errors when drawing with this shader. */
3764 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3765 unsigned char cur_fixup_sampler
= 0;
3767 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3768 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3769 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3771 fixup
->offset
= next_local
;
3772 fixup
->super
.active
= 0;
3774 for (i
= 0; i
< WINED3D_MAX_FRAGMENT_SAMPLERS
; ++i
)
3776 if (!(map
& (1u << i
)))
3779 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
)
3781 fixup
->super
.active
|= (1u << i
);
3782 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3786 FIXME("No free constant found to load NP2 fixup data into shader. "
3787 "Sampling from this texture will probably look wrong.\n");
3792 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3793 if (fixup
->super
.num_consts
) {
3794 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3795 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3799 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3801 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3804 /* Base Shader Body */
3805 if (FAILED(shader_generate_code(shader
, buffer
, reg_maps
, &priv_ctx
, NULL
, NULL
)))
3808 if(args
->super
.srgb_correction
) {
3809 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3810 priv_ctx
.target_version
>= NV2
);
3813 if (custom_linear_fog
)
3814 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3816 if(strcmp(fragcolor
, "result.color")) {
3817 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3819 shader_addline(buffer
, "END\n");
3821 /* TODO: change to resource.glObjectHandle or something like that */
3822 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3824 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3825 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3827 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3828 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3834 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3839 if (sig1
->element_count
!= sig2
->element_count
)
3840 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3842 for (i
= 0; i
< sig1
->element_count
; ++i
)
3844 const struct wined3d_shader_signature_element
*e1
, *e2
;
3846 e1
= &sig1
->elements
[i
];
3847 e2
= &sig2
->elements
[i
];
3849 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3851 /* Compare pointers, not contents. One string is NULL (element
3852 * does not exist), the other one is not NULL. */
3853 if (e1
->semantic_name
!= e2
->semantic_name
)
3854 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3858 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3860 if (e1
->semantic_idx
!= e2
->semantic_idx
)
3861 return e1
->semantic_idx
< e2
->semantic_idx
? -1 : 1;
3862 if (e1
->sysval_semantic
!= e2
->sysval_semantic
)
3863 return e1
->sysval_semantic
< e2
->sysval_semantic
? -1 : 1;
3864 if (e1
->component_type
!= e2
->component_type
)
3865 return e1
->component_type
< e2
->component_type
? -1 : 1;
3866 if (e1
->register_idx
!= e2
->register_idx
)
3867 return e1
->register_idx
< e2
->register_idx
? -1 : 1;
3868 if (e1
->mask
!= e2
->mask
)
3869 return e1
->mask
< e2
->mask
? -1 : 1;
3874 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
3879 new->element_count
= sig
->element_count
;
3880 new->elements
= heap_calloc(new->element_count
, sizeof(*new->elements
));
3881 for (i
= 0; i
< sig
->element_count
; ++i
)
3883 new->elements
[i
] = sig
->elements
[i
];
3885 if (!new->elements
[i
].semantic_name
)
3888 /* Clone the semantic string */
3889 name
= heap_alloc(strlen(sig
->elements
[i
].semantic_name
) + 1);
3890 strcpy(name
, sig
->elements
[i
].semantic_name
);
3891 new->elements
[i
].semantic_name
= name
;
3895 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
3897 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3898 struct ps_signature
*found_sig
;
3902 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3903 TRACE("Found existing signature %u\n", found_sig
->idx
);
3904 return found_sig
->idx
;
3906 found_sig
= heap_alloc_zero(sizeof(*found_sig
));
3907 clone_sig(&found_sig
->sig
, sig
);
3908 found_sig
->idx
= priv
->ps_sig_number
++;
3909 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3910 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3912 ERR("Failed to insert program entry.\n");
3914 return found_sig
->idx
;
3917 static void init_output_registers(const struct wined3d_shader
*shader
,
3918 const struct wined3d_shader_signature
*ps_input_sig
,
3919 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3922 static const char * const texcoords
[8] =
3924 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3925 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3927 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3928 * and varying 9 to result.color.secondary
3930 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3932 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3933 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3934 "result.color.primary", "result.color.secondary"
3939 TRACE("Pixel shader uses builtin varyings\n");
3940 /* Map builtins to builtins */
3941 for(i
= 0; i
< 8; i
++)
3943 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3945 priv_ctx
->color_output
[0] = "result.color.primary";
3946 priv_ctx
->color_output
[1] = "result.color.secondary";
3947 priv_ctx
->fog_output
= "TMP_FOGCOORD";
3949 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3950 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
3952 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
3954 if (!output
->semantic_name
)
3957 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
3959 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
3960 if (!output
->semantic_idx
)
3961 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
3963 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3965 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
3967 TRACE("o%u is result.pointsize\n", output
->register_idx
);
3968 if (!output
->semantic_idx
)
3969 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
3971 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3973 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3975 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
3976 if (!output
->semantic_idx
)
3977 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
3978 else if (output
->semantic_idx
== 1)
3979 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
3980 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3982 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3984 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
3985 if (output
->semantic_idx
>= 8)
3986 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3988 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
3990 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
3992 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
3993 if (output
->semantic_idx
> 0)
3994 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3996 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
4000 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4006 TRACE("Pixel shader uses declared varyings\n");
4008 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4009 for(i
= 0; i
< 8; i
++)
4011 priv_ctx
->texcrd_output
[i
] = "TA";
4013 priv_ctx
->color_output
[0] = "TA";
4014 priv_ctx
->color_output
[1] = "TA";
4015 priv_ctx
->fog_output
= "TA";
4017 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4019 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4021 if (!input
->semantic_name
)
4024 /* If a declared input register is not written by builtin arguments, don't write to it.
4025 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4027 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4028 * to TMP_OUT in any case
4030 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4032 if (input
->semantic_idx
< 8)
4033 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4035 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4037 if (input
->semantic_idx
< 2)
4038 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4040 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4042 if (!input
->semantic_idx
)
4043 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4050 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4051 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4053 compiled
->need_color_unclamp
= TRUE
;
4057 /* Map declared to declared */
4058 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4060 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4062 /* Write unread output to TA to throw them away */
4063 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4065 if (!output
->semantic_name
)
4068 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4070 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4073 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4075 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4079 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4081 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4083 if (!input
->semantic_name
)
4086 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4087 && input
->semantic_idx
== output
->semantic_idx
)
4089 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4091 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4092 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4094 compiled
->need_color_unclamp
= TRUE
;
4101 /* Context activation is done by the caller. */
4102 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4103 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4104 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4105 const struct wined3d_shader_signature
*ps_input_sig
)
4107 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4108 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4109 struct shader_arb_priv
*priv
= shader
->device
->shader_priv
;
4111 DWORD next_local
= 0;
4112 struct shader_arb_ctx_priv priv_ctx
;
4115 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4116 priv_ctx
.gl_info
= gl_info
;
4117 priv_ctx
.cur_vs_args
= args
;
4118 list_init(&priv_ctx
.control_frames
);
4119 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4121 /* Create the hw ARB shader */
4122 shader_addline(buffer
, "!!ARBvp1.0\n");
4124 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4125 * mesurable performance penalty, and we can always make use of it for clipplanes.
4127 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4129 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4130 priv_ctx
.target_version
= NV3
;
4131 shader_addline(buffer
, "ADDRESS aL;\n");
4133 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4135 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4136 priv_ctx
.target_version
= NV2
;
4137 shader_addline(buffer
, "ADDRESS aL;\n");
4139 priv_ctx
.target_version
= ARB
;
4142 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4144 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4145 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4148 wined3d_ftoa(eps
, ftoa_tmp
);
4149 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4151 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4153 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4154 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4157 shader_addline(buffer
, "TEMP TA;\n");
4158 shader_addline(buffer
, "TEMP TB;\n");
4160 /* Base Declarations */
4161 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4162 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4164 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4166 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4167 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
4169 const DWORD
*control_values
= find_loop_control_values(shader
, i
);
4173 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4174 control_values
[0], control_values
[1], control_values
[2]);
4178 compiled
->int_consts
[i
] = next_local
;
4179 compiled
->num_int_consts
++;
4180 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4185 /* We need a constant to fixup the final position */
4186 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4187 compiled
->pos_fixup
= next_local
++;
4189 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4190 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4191 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4192 * a replacement shader depend on the texcoord.w being set properly.
4194 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4195 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4196 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4197 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4198 * this can eat a number of instructions, so skip it unless this cap is set as well
4200 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4202 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4203 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4205 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !priv
->ffp_proj_control
)
4208 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4209 for(i
= 0; i
< MAX_REG_TEXCRD
; i
++)
4211 if (reg_maps
->u
.texcoord_mask
[i
] && reg_maps
->u
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
)
4212 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4217 /* The shader starts with the main function */
4218 priv_ctx
.in_main_func
= TRUE
;
4219 /* Base Shader Body */
4220 if (FAILED(shader_generate_code(shader
, buffer
, reg_maps
, &priv_ctx
, NULL
, NULL
)))
4223 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4224 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4226 shader_addline(buffer
, "END\n");
4228 /* TODO: change to resource.glObjectHandle or something like that */
4229 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4231 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4232 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4234 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4235 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4241 /* Context activation is done by the caller. */
4242 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_context_gl
*context_gl
,
4243 struct wined3d_shader
*shader
, const struct arb_ps_compile_args
*args
)
4245 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4246 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4247 struct wined3d_device
*device
= shader
->device
;
4250 struct arb_ps_compiled_shader
*new_array
;
4251 struct wined3d_string_buffer buffer
;
4252 struct arb_pshader_private
*shader_data
;
4255 if (!shader
->backend_data
)
4257 struct shader_arb_priv
*priv
= device
->shader_priv
;
4259 shader
->backend_data
= heap_alloc_zero(sizeof(*shader_data
));
4260 shader_data
= shader
->backend_data
;
4261 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4263 if (shader
->reg_maps
.shader_version
.major
< 3)
4264 shader_data
->input_signature_idx
= ~0U;
4266 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4268 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4270 if (!d3d_info
->vs_clipping
)
4271 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4272 d3d_info
->limits
.ffp_blend_stages
- 1);
4274 shader_data
->clipplane_emulation
= ~0U;
4276 shader_data
= shader
->backend_data
;
4278 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4279 * so a linear search is more performant than a hashmap or a binary search
4280 * (cache coherency etc)
4282 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4284 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4285 return &shader_data
->gl_shaders
[i
];
4288 TRACE("No matching GL shader found, compiling a new shader\n");
4289 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4290 if (shader_data
->num_gl_shaders
)
4292 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4293 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4294 new_size
* sizeof(*shader_data
->gl_shaders
));
4298 new_array
= heap_alloc_zero(sizeof(*shader_data
->gl_shaders
));
4303 ERR("Out of memory\n");
4306 shader_data
->gl_shaders
= new_array
;
4307 shader_data
->shader_array_size
= new_size
;
4310 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4312 if (!string_buffer_init(&buffer
))
4314 ERR("Failed to initialize shader buffer.\n");
4318 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4319 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4320 string_buffer_free(&buffer
);
4321 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4323 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4326 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4327 const DWORD use_map
, BOOL skip_int
) {
4328 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4329 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4330 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4331 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4332 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4333 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4334 if(skip_int
) return TRUE
;
4336 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4339 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4340 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4341 const struct wined3d_shader_signature
*ps_input_sig
)
4345 struct arb_vs_compiled_shader
*new_array
;
4346 struct wined3d_string_buffer buffer
;
4347 struct arb_vshader_private
*shader_data
;
4350 if (!shader
->backend_data
)
4352 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4354 shader
->backend_data
= heap_alloc_zero(sizeof(*shader_data
));
4355 shader_data
= shader
->backend_data
;
4357 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4358 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4360 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4362 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4363 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4364 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4366 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4367 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4368 else if (reg_maps
->max_rel_offset
> 63)
4369 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4372 shader_data
= shader
->backend_data
;
4374 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4375 * so a linear search is more performant than a hashmap or a binary search
4376 * (cache coherency etc)
4378 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4379 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4380 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4382 return &shader_data
->gl_shaders
[i
];
4386 TRACE("No matching GL shader found, compiling a new shader\n");
4388 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4389 if (shader_data
->num_gl_shaders
)
4391 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4392 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4393 new_size
* sizeof(*shader_data
->gl_shaders
));
4397 new_array
= heap_alloc_zero(sizeof(*shader_data
->gl_shaders
));
4402 ERR("Out of memory\n");
4405 shader_data
->gl_shaders
= new_array
;
4406 shader_data
->shader_array_size
= new_size
;
4409 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4411 if (!string_buffer_init(&buffer
))
4413 ERR("Failed to initialize shader buffer.\n");
4417 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4418 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4420 string_buffer_free(&buffer
);
4421 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4423 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4426 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4427 const struct wined3d_context_gl
*context_gl
, const struct wined3d_shader
*shader
,
4428 struct arb_ps_compile_args
*args
)
4430 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4431 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4435 find_ps_compile_args(state
, shader
, context_gl
->c
.stream_info
.position_transformed
, &args
->super
, &context_gl
->c
);
4437 /* This forces all local boolean constants to 1 to make them stateblock independent */
4438 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4440 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4442 if (state
->ps_consts_b
[i
])
4443 args
->bools
|= ( 1u << i
);
4446 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4447 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4448 * duplicate the shader than have a no-op KIL instruction in every shader
4450 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4451 && state
->render_states
[WINED3D_RS_CLIPPING
]
4452 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4457 /* Skip if unused or local, or supported natively */
4458 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4459 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4461 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4465 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4467 if (int_skip
& (1u << i
))
4469 args
->loop_ctrl
[i
][0] = 0;
4470 args
->loop_ctrl
[i
][1] = 0;
4471 args
->loop_ctrl
[i
][2] = 0;
4475 args
->loop_ctrl
[i
][0] = state
->ps_consts_i
[i
].x
;
4476 args
->loop_ctrl
[i
][1] = state
->ps_consts_i
[i
].y
;
4477 args
->loop_ctrl
[i
][2] = state
->ps_consts_i
[i
].z
;
4482 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4483 const struct wined3d_context_gl
*context_gl
, const struct wined3d_shader
*shader
,
4484 struct arb_vs_compile_args
*args
)
4486 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4487 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4488 const struct wined3d_device
*device
= shader
->device
;
4489 const struct wined3d_adapter
*adapter
= device
->adapter
;
4493 find_vs_compile_args(state
, shader
, context_gl
->c
.stream_info
.swizzle_map
, &args
->super
, &context_gl
->c
);
4495 args
->clip
.boolclip_compare
= 0;
4498 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4499 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4500 args
->ps_signature
= shader_priv
->input_signature_idx
;
4502 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4506 args
->ps_signature
= ~0;
4507 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4508 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(&context_gl
->c
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4509 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4512 if (args
->clip
.boolclip
.clip_texcoord
)
4514 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4515 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4516 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4519 /* This forces all local boolean constants to 1 to make them stateblock independent */
4520 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4521 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4522 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4524 if (state
->vs_consts_b
[i
])
4525 args
->clip
.boolclip
.bools
|= (1u << i
);
4528 args
->vertex
.samplers
[0] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 0];
4529 args
->vertex
.samplers
[1] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 1];
4530 args
->vertex
.samplers
[2] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 2];
4531 args
->vertex
.samplers
[3] = 0;
4533 /* Skip if unused or local */
4534 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4535 /* This is about flow control, not clipping. */
4536 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4538 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4542 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4544 if (int_skip
& (1u << i
))
4546 args
->loop_ctrl
[i
][0] = 0;
4547 args
->loop_ctrl
[i
][1] = 0;
4548 args
->loop_ctrl
[i
][2] = 0;
4552 args
->loop_ctrl
[i
][0] = state
->vs_consts_i
[i
].x
;
4553 args
->loop_ctrl
[i
][1] = state
->vs_consts_i
[i
].y
;
4554 args
->loop_ctrl
[i
][2] = state
->vs_consts_i
[i
].z
;
4559 /* Context activation is done by the caller. */
4560 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4561 const struct wined3d_state
*state
)
4563 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
4564 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4565 struct shader_arb_priv
*priv
= shader_priv
;
4568 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4571 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4572 struct arb_ps_compile_args compile_args
;
4573 struct arb_ps_compiled_shader
*compiled
;
4575 TRACE("Using pixel shader %p.\n", ps
);
4576 find_arb_ps_compile_args(state
, context_gl
, ps
, &compile_args
);
4577 compiled
= find_arb_pshader(context_gl
, ps
, &compile_args
);
4578 priv
->current_fprogram_id
= compiled
->prgId
;
4579 priv
->compiled_fprog
= compiled
;
4581 /* Bind the fragment program */
4582 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4583 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4585 if (!priv
->use_arbfp_fixed_func
)
4586 priv
->fragment_pipe
->fp_enable(context
, FALSE
);
4588 /* Enable OpenGL fragment programs. */
4589 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4590 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4592 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", priv
->current_fprogram_id
);
4594 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4595 * a 1.x and newer shader, reload the first 8 constants
4597 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4599 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4600 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4601 for(i
= 0; i
< 8; i
++)
4603 priv
->pshader_const_dirty
[i
] = 1;
4605 /* Also takes care of loading local constants */
4606 shader_arb_load_constants_internal(shader_priv
, context_gl
, state
, TRUE
, FALSE
, TRUE
);
4610 UINT rt_height
= state
->fb
.render_targets
[0]->height
;
4611 shader_arb_ps_local_constants(compiled
, context_gl
, state
, rt_height
);
4614 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4615 if (compiled
->np2fixup_info
.super
.active
)
4616 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4618 if (ps
->load_local_constsF
)
4619 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4623 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4625 /* Disable only if we're not using arbfp fixed function fragment
4626 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4627 * enabled, and the fixed function pipeline will bind the fixed
4628 * function replacement shader. */
4629 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4630 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4631 priv
->current_fprogram_id
= 0;
4633 priv
->fragment_pipe
->fp_enable(context
, TRUE
);
4638 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4639 struct arb_vs_compile_args compile_args
;
4640 struct arb_vs_compiled_shader
*compiled
;
4641 const struct wined3d_shader_signature
*ps_input_sig
;
4643 TRACE("Using vertex shader %p\n", vs
);
4644 find_arb_vs_compile_args(state
, context_gl
, vs
, &compile_args
);
4646 /* Instead of searching for the signature in the signature list, read the one from the
4647 * current pixel shader. It's maybe not the shader where the signature came from, but it
4648 * is the same signature and faster to find. */
4649 if (compile_args
.ps_signature
== ~0U)
4650 ps_input_sig
= NULL
;
4652 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4654 compiled
= find_arb_vshader(vs
, gl_info
, context
->stream_info
.use_map
,
4655 &compile_args
, ps_input_sig
);
4656 priv
->current_vprogram_id
= compiled
->prgId
;
4657 priv
->compiled_vprog
= compiled
;
4659 /* Bind the vertex program */
4660 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4661 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4663 priv
->vertex_pipe
->vp_enable(context
, FALSE
);
4665 /* Enable OpenGL vertex programs */
4666 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4667 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4668 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", priv
->current_vprogram_id
);
4669 shader_arb_vs_local_constants(compiled
, context_gl
, state
);
4671 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4672 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4674 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4676 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4677 checkGLcall("glClampColorARB");
4679 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4683 if (vs
->load_local_constsF
)
4684 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4688 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4690 priv
->current_vprogram_id
= 0;
4691 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4692 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4694 priv
->vertex_pipe
->vp_enable(context
, TRUE
);
4698 static void shader_arb_select_compute(void *shader_priv
, struct wined3d_context
*context
,
4699 const struct wined3d_state
*state
)
4701 ERR("Compute pipeline not supported by the ARB shader backend.\n");
4704 /* Context activation is done by the caller. */
4705 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4707 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
4708 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4709 struct shader_arb_priv
*priv
= shader_priv
;
4711 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4713 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4714 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4715 priv
->current_fprogram_id
= 0;
4717 priv
->fragment_pipe
->fp_enable(context
, FALSE
);
4719 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4721 priv
->current_vprogram_id
= 0;
4722 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4723 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4725 priv
->vertex_pipe
->vp_enable(context
, FALSE
);
4727 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4729 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4730 checkGLcall("glClampColorARB");
4731 priv
->last_vs_color_unclamp
= FALSE
;
4734 context
->shader_update_mask
= (1u << WINED3D_SHADER_TYPE_PIXEL
)
4735 | (1u << WINED3D_SHADER_TYPE_VERTEX
)
4736 | (1u << WINED3D_SHADER_TYPE_GEOMETRY
)
4737 | (1u << WINED3D_SHADER_TYPE_HULL
)
4738 | (1u << WINED3D_SHADER_TYPE_DOMAIN
)
4739 | (1u << WINED3D_SHADER_TYPE_COMPUTE
);
4742 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4744 struct wined3d_device
*device
= shader
->device
;
4745 const struct wined3d_gl_info
*gl_info
;
4746 struct wined3d_context
*context
;
4749 /* This can happen if a shader was never compiled */
4750 if (!shader
->backend_data
)
4753 context
= context_acquire(device
, NULL
, 0);
4754 gl_info
= wined3d_context_gl(context
)->gl_info
;
4756 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4758 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4760 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4761 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4763 heap_free(shader_data
->gl_shaders
);
4767 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4769 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4770 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4772 heap_free(shader_data
->gl_shaders
);
4775 checkGLcall("delete programs");
4777 context_release(context
);
4779 heap_free(shader
->backend_data
);
4780 shader
->backend_data
= NULL
;
4783 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4785 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4786 return compare_sig(key
, &e
->sig
);
4789 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4790 const struct wined3d_fragment_pipe_ops
*fragment_pipe
)
4792 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4793 struct fragment_caps fragment_caps
;
4794 void *vertex_priv
, *fragment_priv
;
4795 struct shader_arb_priv
*priv
;
4797 if (!(priv
= heap_alloc_zero(sizeof(*priv
))))
4798 return E_OUTOFMEMORY
;
4800 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4802 ERR("Failed to initialize vertex pipe.\n");
4807 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4809 ERR("Failed to initialize fragment pipe.\n");
4810 vertex_pipe
->vp_free(device
, NULL
);
4815 memset(priv
->vshader_const_dirty
, 1,
4816 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4817 memset(priv
->pshader_const_dirty
, 1,
4818 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4820 wine_rb_init(&priv
->signature_tree
, sig_tree_compare
);
4822 priv
->vertex_pipe
= vertex_pipe
;
4823 priv
->fragment_pipe
= fragment_pipe
;
4824 fragment_pipe
->get_caps(device
->adapter
, &fragment_caps
);
4825 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4827 device
->vertex_priv
= vertex_priv
;
4828 device
->fragment_priv
= fragment_priv
;
4829 device
->shader_priv
= priv
;
4834 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4836 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4839 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
4841 heap_free((char *)sig
->sig
.elements
[i
].semantic_name
);
4843 heap_free(sig
->sig
.elements
);
4847 /* Context activation is done by the caller. */
4848 static void shader_arb_free(struct wined3d_device
*device
, struct wined3d_context
*context
)
4850 struct shader_arb_priv
*priv
= device
->shader_priv
;
4852 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4853 priv
->fragment_pipe
->free_private(device
, context
);
4854 priv
->vertex_pipe
->vp_free(device
, context
);
4855 heap_free(device
->shader_priv
);
4858 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
4863 static void shader_arb_free_context_data(struct wined3d_context
*context
)
4865 struct shader_arb_priv
*priv
;
4867 priv
= context
->device
->shader_priv
;
4868 if (priv
->last_context
== context
)
4869 priv
->last_context
= NULL
;
4872 static void shader_arb_init_context_state(struct wined3d_context
*context
) {}
4874 static void shader_arb_get_caps(const struct wined3d_adapter
*adapter
, struct shader_caps
*caps
)
4876 const struct wined3d_gl_info
*gl_info
= &adapter
->gl_info
;
4878 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4883 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
4884 * for vertex programs. If the native limit is less than that it's
4885 * not very useful, and e.g. Mesa swrast returns 0, probably to
4886 * indicate it's a software implementation. */
4887 if (gl_info
->limits
.arb_vs_native_constants
< 96)
4888 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
4890 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
4892 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4895 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
4897 else if (vs_consts
>= 256)
4899 /* Shader Model 2.0 requires at least 256 vertex shader constants */
4901 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
4906 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
4908 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
4909 caps
->vs_uniform_count
= min(WINED3D_MAX_VS_CONSTS_F
, vs_consts
);
4913 caps
->vs_version
= 0;
4914 caps
->vs_uniform_count
= 0;
4917 caps
->hs_version
= 0;
4918 caps
->ds_version
= 0;
4919 caps
->gs_version
= 0;
4920 caps
->cs_version
= 0;
4922 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4927 /* Similar as above for vertex programs, but the minimum for fragment
4928 * programs is 24. */
4929 if (gl_info
->limits
.arb_ps_native_constants
< 24)
4930 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
4932 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
4934 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4937 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
4939 else if (ps_consts
>= 32)
4941 /* Shader Model 2.0 requires at least 32 pixel shader constants */
4943 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
4948 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
4950 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
4951 caps
->ps_uniform_count
= min(WINED3D_MAX_PS_CONSTS_F
, ps_consts
);
4952 caps
->ps_1x_max_value
= 8.0f
;
4956 caps
->ps_version
= 0;
4957 caps
->ps_uniform_count
= 0;
4958 caps
->ps_1x_max_value
= 0.0f
;
4961 caps
->varying_count
= 0;
4962 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
4963 if (use_nv_clip(gl_info
))
4964 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
4967 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
4969 /* We support everything except complex conversions. */
4970 return !is_complex_fixup(fixup
);
4973 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
4975 char write_mask
[20], regstr
[50];
4976 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
4977 BOOL is_color
= FALSE
;
4978 const struct wined3d_shader_dst_param
*dst
;
4980 if (!ins
->dst_count
) return;
4984 if (!shift
) return; /* Saturate alone is handled by the instructions */
4986 shader_arb_get_write_mask(ins
, dst
, write_mask
);
4987 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
4989 /* Generate a line that does the output modifier computation
4990 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
4991 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
4993 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
4994 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
4997 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4999 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5000 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5001 /* WINED3DSIH_AND */ NULL
,
5002 /* WINED3DSIH_ATOMIC_AND */ NULL
,
5003 /* WINED3DSIH_ATOMIC_CMP_STORE */ NULL
,
5004 /* WINED3DSIH_ATOMIC_IADD */ NULL
,
5005 /* WINED3DSIH_ATOMIC_IMAX */ NULL
,
5006 /* WINED3DSIH_ATOMIC_IMIN */ NULL
,
5007 /* WINED3DSIH_ATOMIC_OR */ NULL
,
5008 /* WINED3DSIH_ATOMIC_UMAX */ NULL
,
5009 /* WINED3DSIH_ATOMIC_UMIN */ NULL
,
5010 /* WINED3DSIH_ATOMIC_XOR */ NULL
,
5011 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5012 /* WINED3DSIH_BFI */ NULL
,
5013 /* WINED3DSIH_BFREV */ NULL
,
5014 /* WINED3DSIH_BREAK */ shader_hw_break
,
5015 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5016 /* WINED3DSIH_BREAKP */ NULL
,
5017 /* WINED3DSIH_BUFINFO */ NULL
,
5018 /* WINED3DSIH_CALL */ shader_hw_call
,
5019 /* WINED3DSIH_CALLNZ */ NULL
,
5020 /* WINED3DSIH_CASE */ NULL
,
5021 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5022 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5023 /* WINED3DSIH_CONTINUE */ NULL
,
5024 /* WINED3DSIH_CONTINUEP */ NULL
,
5025 /* WINED3DSIH_COUNTBITS */ NULL
,
5026 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5027 /* WINED3DSIH_CUT */ NULL
,
5028 /* WINED3DSIH_CUT_STREAM */ NULL
,
5029 /* WINED3DSIH_DCL */ shader_hw_nop
,
5030 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5031 /* WINED3DSIH_DCL_FUNCTION_BODY */ NULL
,
5032 /* WINED3DSIH_DCL_FUNCTION_TABLE */ NULL
,
5033 /* WINED3DSIH_DCL_GLOBAL_FLAGS */ NULL
,
5034 /* WINED3DSIH_DCL_GS_INSTANCES */ NULL
,
5035 /* WINED3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT */ NULL
,
5036 /* WINED3DSIH_DCL_HS_JOIN_PHASE_INSTANCE_COUNT */ NULL
,
5037 /* WINED3DSIH_DCL_HS_MAX_TESSFACTOR */ NULL
,
5038 /* WINED3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER */ NULL
,
5039 /* WINED3DSIH_DCL_INDEX_RANGE */ NULL
,
5040 /* WINED3DSIH_DCL_INDEXABLE_TEMP */ NULL
,
5041 /* WINED3DSIH_DCL_INPUT */ NULL
,
5042 /* WINED3DSIH_DCL_INPUT_CONTROL_POINT_COUNT */ NULL
,
5043 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5044 /* WINED3DSIH_DCL_INPUT_PS */ NULL
,
5045 /* WINED3DSIH_DCL_INPUT_PS_SGV */ NULL
,
5046 /* WINED3DSIH_DCL_INPUT_PS_SIV */ NULL
,
5047 /* WINED3DSIH_DCL_INPUT_SGV */ NULL
,
5048 /* WINED3DSIH_DCL_INPUT_SIV */ NULL
,
5049 /* WINED3DSIH_DCL_INTERFACE */ NULL
,
5050 /* WINED3DSIH_DCL_OUTPUT */ NULL
,
5051 /* WINED3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT */ NULL
,
5052 /* WINED3DSIH_DCL_OUTPUT_SIV */ NULL
,
5053 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5054 /* WINED3DSIH_DCL_RESOURCE_RAW */ NULL
,
5055 /* WINED3DSIH_DCL_RESOURCE_STRUCTURED */ NULL
,
5056 /* WINED3DSIH_DCL_SAMPLER */ NULL
,
5057 /* WINED3DSIH_DCL_STREAM */ NULL
,
5058 /* WINED3DSIH_DCL_TEMPS */ NULL
,
5059 /* WINED3DSIH_DCL_TESSELLATOR_DOMAIN */ NULL
,
5060 /* WINED3DSIH_DCL_TESSELLATOR_OUTPUT_PRIMITIVE */ NULL
,
5061 /* WINED3DSIH_DCL_TESSELLATOR_PARTITIONING */ NULL
,
5062 /* WINED3DSIH_DCL_TGSM_RAW */ NULL
,
5063 /* WINED3DSIH_DCL_TGSM_STRUCTURED */ NULL
,
5064 /* WINED3DSIH_DCL_THREAD_GROUP */ NULL
,
5065 /* WINED3DSIH_DCL_UAV_RAW */ NULL
,
5066 /* WINED3DSIH_DCL_UAV_STRUCTURED */ NULL
,
5067 /* WINED3DSIH_DCL_UAV_TYPED */ NULL
,
5068 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5069 /* WINED3DSIH_DEF */ shader_hw_nop
,
5070 /* WINED3DSIH_DEFAULT */ NULL
,
5071 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5072 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5073 /* WINED3DSIH_DIV */ NULL
,
5074 /* WINED3DSIH_DP2 */ NULL
,
5075 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5076 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5077 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5078 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5079 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5080 /* WINED3DSIH_DSX_COARSE */ NULL
,
5081 /* WINED3DSIH_DSX_FINE */ NULL
,
5082 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5083 /* WINED3DSIH_DSY_COARSE */ NULL
,
5084 /* WINED3DSIH_DSY_FINE */ NULL
,
5085 /* WINED3DSIH_ELSE */ shader_hw_else
,
5086 /* WINED3DSIH_EMIT */ NULL
,
5087 /* WINED3DSIH_EMIT_STREAM */ NULL
,
5088 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5089 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5090 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5091 /* WINED3DSIH_ENDSWITCH */ NULL
,
5092 /* WINED3DSIH_EQ */ NULL
,
5093 /* WINED3DSIH_EVAL_SAMPLE_INDEX */ NULL
,
5094 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5095 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5096 /* WINED3DSIH_F16TOF32 */ NULL
,
5097 /* WINED3DSIH_F32TOF16 */ NULL
,
5098 /* WINED3DSIH_FCALL */ NULL
,
5099 /* WINED3DSIH_FIRSTBIT_HI */ NULL
,
5100 /* WINED3DSIH_FIRSTBIT_LO */ NULL
,
5101 /* WINED3DSIH_FIRSTBIT_SHI */ NULL
,
5102 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5103 /* WINED3DSIH_FTOI */ NULL
,
5104 /* WINED3DSIH_FTOU */ NULL
,
5105 /* WINED3DSIH_GATHER4 */ NULL
,
5106 /* WINED3DSIH_GATHER4_C */ NULL
,
5107 /* WINED3DSIH_GATHER4_PO */ NULL
,
5108 /* WINED3DSIH_GATHER4_PO_C */ NULL
,
5109 /* WINED3DSIH_GE */ NULL
,
5110 /* WINED3DSIH_HS_CONTROL_POINT_PHASE */ NULL
,
5111 /* WINED3DSIH_HS_DECLS */ NULL
,
5112 /* WINED3DSIH_HS_FORK_PHASE */ NULL
,
5113 /* WINED3DSIH_HS_JOIN_PHASE */ NULL
,
5114 /* WINED3DSIH_IADD */ NULL
,
5115 /* WINED3DSIH_IBFE */ NULL
,
5116 /* WINED3DSIH_IEQ */ NULL
,
5117 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5118 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5119 /* WINED3DSIH_IGE */ NULL
,
5120 /* WINED3DSIH_ILT */ NULL
,
5121 /* WINED3DSIH_IMAD */ NULL
,
5122 /* WINED3DSIH_IMAX */ NULL
,
5123 /* WINED3DSIH_IMIN */ NULL
,
5124 /* WINED3DSIH_IMM_ATOMIC_ALLOC */ NULL
,
5125 /* WINED3DSIH_IMM_ATOMIC_AND */ NULL
,
5126 /* WINED3DSIH_IMM_ATOMIC_CMP_EXCH */ NULL
,
5127 /* WINED3DSIH_IMM_ATOMIC_CONSUME */ NULL
,
5128 /* WINED3DSIH_IMM_ATOMIC_EXCH */ NULL
,
5129 /* WINED3DSIH_IMM_ATOMIC_IADD */ NULL
,
5130 /* WINED3DSIH_IMM_ATOMIC_IMAX */ NULL
,
5131 /* WINED3DSIH_IMM_ATOMIC_IMIN */ NULL
,
5132 /* WINED3DSIH_IMM_ATOMIC_OR */ NULL
,
5133 /* WINED3DSIH_IMM_ATOMIC_UMAX */ NULL
,
5134 /* WINED3DSIH_IMM_ATOMIC_UMIN */ NULL
,
5135 /* WINED3DSIH_IMM_ATOMIC_XOR */ NULL
,
5136 /* WINED3DSIH_IMUL */ NULL
,
5137 /* WINED3DSIH_INE */ NULL
,
5138 /* WINED3DSIH_INEG */ NULL
,
5139 /* WINED3DSIH_ISHL */ NULL
,
5140 /* WINED3DSIH_ISHR */ NULL
,
5141 /* WINED3DSIH_ITOF */ NULL
,
5142 /* WINED3DSIH_LABEL */ shader_hw_label
,
5143 /* WINED3DSIH_LD */ NULL
,
5144 /* WINED3DSIH_LD2DMS */ NULL
,
5145 /* WINED3DSIH_LD_RAW */ NULL
,
5146 /* WINED3DSIH_LD_STRUCTURED */ NULL
,
5147 /* WINED3DSIH_LD_UAV_TYPED */ NULL
,
5148 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5149 /* WINED3DSIH_LOD */ NULL
,
5150 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5151 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5152 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5153 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5154 /* WINED3DSIH_LT */ NULL
,
5155 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5156 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5157 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5158 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5159 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5160 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5161 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5162 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5163 /* WINED3DSIH_MOV */ shader_hw_mov
,
5164 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5165 /* WINED3DSIH_MOVC */ NULL
,
5166 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5167 /* WINED3DSIH_NE */ NULL
,
5168 /* WINED3DSIH_NOP */ shader_hw_nop
,
5169 /* WINED3DSIH_NOT */ NULL
,
5170 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5171 /* WINED3DSIH_OR */ NULL
,
5172 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5173 /* WINED3DSIH_POW */ shader_hw_pow
,
5174 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5175 /* WINED3DSIH_REP */ shader_hw_rep
,
5176 /* WINED3DSIH_RESINFO */ NULL
,
5177 /* WINED3DSIH_RET */ shader_hw_ret
,
5178 /* WINED3DSIH_RETP */ NULL
,
5179 /* WINED3DSIH_ROUND_NE */ NULL
,
5180 /* WINED3DSIH_ROUND_NI */ NULL
,
5181 /* WINED3DSIH_ROUND_PI */ NULL
,
5182 /* WINED3DSIH_ROUND_Z */ NULL
,
5183 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5184 /* WINED3DSIH_SAMPLE */ NULL
,
5185 /* WINED3DSIH_SAMPLE_B */ NULL
,
5186 /* WINED3DSIH_SAMPLE_C */ NULL
,
5187 /* WINED3DSIH_SAMPLE_C_LZ */ NULL
,
5188 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5189 /* WINED3DSIH_SAMPLE_INFO */ NULL
,
5190 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5191 /* WINED3DSIH_SAMPLE_POS */ NULL
,
5192 /* WINED3DSIH_SETP */ NULL
,
5193 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5194 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5195 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5196 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5197 /* WINED3DSIH_SQRT */ NULL
,
5198 /* WINED3DSIH_STORE_RAW */ NULL
,
5199 /* WINED3DSIH_STORE_STRUCTURED */ NULL
,
5200 /* WINED3DSIH_STORE_UAV_TYPED */ NULL
,
5201 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5202 /* WINED3DSIH_SWAPC */ NULL
,
5203 /* WINED3DSIH_SWITCH */ NULL
,
5204 /* WINED3DSIH_SYNC */ NULL
,
5205 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5206 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5207 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5208 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5209 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5210 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5211 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5212 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5213 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5214 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5215 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5216 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5217 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5218 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5219 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5220 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5221 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5222 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5223 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5224 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5225 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5226 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5227 /* WINED3DSIH_UBFE */ NULL
,
5228 /* WINED3DSIH_UDIV */ NULL
,
5229 /* WINED3DSIH_UGE */ NULL
,
5230 /* WINED3DSIH_ULT */ NULL
,
5231 /* WINED3DSIH_UMAX */ NULL
,
5232 /* WINED3DSIH_UMIN */ NULL
,
5233 /* WINED3DSIH_UMUL */ NULL
,
5234 /* WINED3DSIH_USHR */ NULL
,
5235 /* WINED3DSIH_UTOF */ NULL
,
5236 /* WINED3DSIH_XOR */ NULL
,
5239 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5240 const struct wined3d_shader
*shader
, DWORD idx
)
5242 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5243 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5244 const struct wined3d_shader_lconst
*constant
;
5246 WORD flag
= (1u << idx
);
5247 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5249 if (reg_maps
->local_bool_consts
& flag
)
5251 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5252 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5254 if (constant
->idx
== idx
)
5256 return constant
->value
[0];
5259 ERR("Local constant not found\n");
5264 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5265 else bools
= priv
->cur_ps_args
->bools
;
5266 return bools
& flag
;
5270 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5271 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5273 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5274 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5276 /* Integer constants can either be a local constant, or they can be stored in the shader
5277 * type specific compile args. */
5278 if (reg_maps
->local_int_consts
& (1u << idx
))
5280 const struct wined3d_shader_lconst
*constant
;
5282 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5284 if (constant
->idx
== idx
)
5286 loop_control
->count
= constant
->value
[0];
5287 loop_control
->start
= constant
->value
[1];
5288 /* Step is signed. */
5289 loop_control
->step
= (int)constant
->value
[2];
5293 /* If this happens the flag was set incorrectly */
5294 ERR("Local constant not found\n");
5295 loop_control
->count
= 0;
5296 loop_control
->start
= 0;
5297 loop_control
->step
= 0;
5301 switch (reg_maps
->shader_version
.type
)
5303 case WINED3D_SHADER_TYPE_VERTEX
:
5304 /* Count and aL start value are unsigned */
5305 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5306 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5307 /* Step is signed. */
5308 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5311 case WINED3D_SHADER_TYPE_PIXEL
:
5312 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5313 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5314 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5318 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5323 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5325 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
;
5326 struct wined3d_shader_dst_param
*dst_param
;
5327 struct recorded_instruction
*rec
;
5330 if (!(rec
= heap_alloc_zero(sizeof(*rec
))))
5332 ERR("Out of memory\n");
5337 if (!(dst_param
= heap_alloc(sizeof(*dst_param
))))
5339 *dst_param
= *ins
->dst
;
5340 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5342 if (!(rel_addr
= heap_alloc(sizeof(*rel_addr
))))
5344 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5345 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5347 rec
->ins
.dst
= dst_param
;
5349 if (!(src_param
= heap_calloc(ins
->src_count
, sizeof(*src_param
))))
5351 for (i
= 0; i
< ins
->src_count
; ++i
)
5353 src_param
[i
] = ins
->src
[i
];
5354 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5356 if (!(rel_addr
= heap_alloc(sizeof(*rel_addr
))))
5358 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5359 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5362 rec
->ins
.src
= src_param
;
5363 list_add_tail(list
, &rec
->entry
);
5367 ERR("Out of memory\n");
5370 heap_free((void *)dst_param
->reg
.idx
[0].rel_addr
);
5371 heap_free(dst_param
);
5375 for (i
= 0; i
< ins
->src_count
; ++i
)
5377 heap_free((void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5379 heap_free(src_param
);
5384 static void free_recorded_instruction(struct list
*list
)
5386 struct recorded_instruction
*rec_ins
, *entry2
;
5389 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5391 list_remove(&rec_ins
->entry
);
5392 if (rec_ins
->ins
.dst
)
5394 heap_free((void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5395 heap_free((void *)rec_ins
->ins
.dst
);
5397 if (rec_ins
->ins
.src
)
5399 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5401 heap_free((void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5403 heap_free((void *)rec_ins
->ins
.src
);
5409 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5411 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5412 struct control_frame
*control_frame
;
5414 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5416 struct list
*e
= list_head(&priv
->control_frames
);
5417 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5418 list_remove(&control_frame
->entry
);
5419 heap_free(control_frame
);
5422 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5424 /* Non-ifc ENDIFs were already handled previously. */
5425 struct list
*e
= list_head(&priv
->control_frames
);
5426 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5427 list_remove(&control_frame
->entry
);
5428 heap_free(control_frame
);
5432 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5433 SHADER_HANDLER hw_fct
;
5434 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5435 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5436 struct control_frame
*control_frame
;
5437 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5440 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5442 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5443 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5445 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5446 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5448 if(priv
->target_version
>= NV2
)
5450 control_frame
->no
.loop
= priv
->num_loops
++;
5455 /* Don't bother recording when we're in a not used if branch */
5461 if(!priv
->recording
)
5463 list_init(&priv
->record
);
5464 priv
->recording
= TRUE
;
5465 control_frame
->outer_loop
= TRUE
;
5466 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5467 return; /* Instruction is handled */
5469 /* Record this loop in the outer loop's recording */
5472 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5474 if(priv
->target_version
>= NV2
)
5476 /* Nothing to do. The control frame is popped after the HW instr handler */
5480 struct list
*e
= list_head(&priv
->control_frames
);
5481 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5482 list_remove(&control_frame
->entry
);
5484 if(control_frame
->outer_loop
)
5486 unsigned int iteration
;
5490 /* Turn off recording before playback */
5491 priv
->recording
= FALSE
;
5493 /* Move the recorded instructions to a separate list and get them out of the private data
5494 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5495 * be recorded again, thus priv->record might be overwritten
5498 list_move_tail(©
, &priv
->record
);
5499 list_init(&priv
->record
);
5501 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5503 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5504 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5505 control_frame
->loop_control
.step
);
5506 aL
= control_frame
->loop_control
.start
;
5510 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5513 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5515 struct recorded_instruction
*rec_ins
;
5516 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5519 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5523 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5526 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5528 shader_arb_handle_instruction(&rec_ins
->ins
);
5531 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5533 aL
+= control_frame
->loop_control
.step
;
5536 shader_addline(buffer
, "#end loop/rep\n");
5538 free_recorded_instruction(©
);
5539 heap_free(control_frame
);
5540 return; /* Instruction is handled */
5544 /* This is a nested loop. Proceed to the normal recording function */
5545 heap_free(control_frame
);
5552 record_instruction(&priv
->record
, ins
);
5557 if(ins
->handler_idx
== WINED3DSIH_IF
)
5559 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5560 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5561 control_frame
->type
= IF
;
5563 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5564 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5565 bool_const
= !bool_const
;
5566 if (!priv
->muted
&& !bool_const
)
5568 shader_addline(buffer
, "#if(FALSE){\n");
5570 control_frame
->muting
= TRUE
;
5572 else shader_addline(buffer
, "#if(TRUE) {\n");
5574 return; /* Instruction is handled */
5576 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5578 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5579 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5580 control_frame
->type
= IFC
;
5581 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5582 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5584 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5586 struct list
*e
= list_head(&priv
->control_frames
);
5587 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5589 if(control_frame
->type
== IF
)
5591 shader_addline(buffer
, "#} else {\n");
5592 if(!priv
->muted
&& !control_frame
->muting
)
5595 control_frame
->muting
= TRUE
;
5597 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5598 return; /* Instruction is handled. */
5600 /* In case of an ifc, generate a HW shader instruction */
5601 if (control_frame
->type
!= IFC
)
5602 ERR("Control frame does not match.\n");
5604 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5606 struct list
*e
= list_head(&priv
->control_frames
);
5607 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5609 if(control_frame
->type
== IF
)
5611 shader_addline(buffer
, "#} endif\n");
5612 if(control_frame
->muting
) priv
->muted
= FALSE
;
5613 list_remove(&control_frame
->entry
);
5614 heap_free(control_frame
);
5615 return; /* Instruction is handled */
5617 /* In case of an ifc, generate a HW shader instruction */
5618 if (control_frame
->type
!= IFC
)
5619 ERR("Control frame does not match.\n");
5624 pop_control_frame(ins
);
5628 /* Select handler */
5629 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5631 /* Unhandled opcode */
5634 FIXME("Backend can't handle opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
5639 pop_control_frame(ins
);
5641 shader_arb_add_instruction_modifiers(ins
);
5644 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5646 struct shader_arb_priv
*priv
= shader_priv
;
5648 return priv
->ffp_proj_control
;
5651 static void shader_arb_precompile(void *shader_priv
, struct wined3d_shader
*shader
) {}
5653 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5655 shader_arb_handle_instruction
,
5656 shader_arb_precompile
,
5658 shader_arb_select_compute
,
5660 shader_arb_update_float_vertex_constants
,
5661 shader_arb_update_float_pixel_constants
,
5662 shader_arb_load_constants
,
5666 shader_arb_allocate_context_data
,
5667 shader_arb_free_context_data
,
5668 shader_arb_init_context_state
,
5669 shader_arb_get_caps
,
5670 shader_arb_color_fixup_supported
,
5671 shader_arb_has_ffp_proj_control
,
5674 /* ARB_fragment_program fixed function pipeline replacement definitions */
5675 #define ARB_FFP_CONST_TFACTOR 0
5676 #define ARB_FFP_CONST_COLOR_KEY_LOW ((ARB_FFP_CONST_TFACTOR) + 1)
5677 #define ARB_FFP_CONST_COLOR_KEY_HIGH ((ARB_FFP_CONST_COLOR_KEY_LOW) + 1)
5678 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY_HIGH) + 1)
5679 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5680 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5681 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5683 struct arbfp_ffp_desc
5685 struct ffp_frag_desc parent
;
5689 /* Context activation is done by the caller. */
5690 static void arbfp_enable(const struct wined3d_context
*context
, BOOL enable
)
5692 const struct wined3d_gl_info
*gl_info
= wined3d_context_gl_const(context
)->gl_info
;
5696 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5697 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5701 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5702 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5706 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5708 struct shader_arb_priv
*priv
;
5710 /* Share private data between the shader backend and the pipeline
5711 * replacement, if both are the arb implementation. This is needed to
5712 * figure out whether ARBfp should be disabled if no pixel shader is bound
5714 if (shader_backend
== &arb_program_shader_backend
)
5716 else if (!(priv
= heap_alloc_zero(sizeof(*priv
))))
5719 wine_rb_init(&priv
->fragment_shaders
, wined3d_ffp_frag_program_key_compare
);
5720 priv
->use_arbfp_fixed_func
= TRUE
;
5725 /* Context activation is done by the caller. */
5726 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *param
)
5728 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5729 struct wined3d_context_gl
*context_gl
= param
;
5730 const struct wined3d_gl_info
*gl_info
;
5732 gl_info
= context_gl
->gl_info
;
5733 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5734 checkGLcall("delete ffp program");
5735 heap_free(entry_arb
);
5738 /* Context activation is done by the caller. */
5739 static void arbfp_free(struct wined3d_device
*device
, struct wined3d_context
*context
)
5741 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5742 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5744 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, context_gl
);
5745 priv
->use_arbfp_fixed_func
= FALSE
;
5747 if (device
->shader_backend
!= &arb_program_shader_backend
)
5748 heap_free(device
->fragment_priv
);
5751 static void arbfp_get_caps(const struct wined3d_adapter
*adapter
, struct fragment_caps
*caps
)
5753 const struct wined3d_gl_info
*gl_info
= &adapter
->gl_info
;
5755 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5756 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5757 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5758 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5759 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5760 WINED3DTEXOPCAPS_SELECTARG1
|
5761 WINED3DTEXOPCAPS_SELECTARG2
|
5762 WINED3DTEXOPCAPS_MODULATE4X
|
5763 WINED3DTEXOPCAPS_MODULATE2X
|
5764 WINED3DTEXOPCAPS_MODULATE
|
5765 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5766 WINED3DTEXOPCAPS_ADDSIGNED
|
5767 WINED3DTEXOPCAPS_ADD
|
5768 WINED3DTEXOPCAPS_SUBTRACT
|
5769 WINED3DTEXOPCAPS_ADDSMOOTH
|
5770 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5771 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5772 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5773 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5774 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5775 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5776 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5777 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5778 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5779 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5780 WINED3DTEXOPCAPS_MULTIPLYADD
|
5781 WINED3DTEXOPCAPS_LERP
|
5782 WINED3DTEXOPCAPS_BUMPENVMAP
|
5783 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5785 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5787 caps
->MaxTextureBlendStages
= WINED3D_MAX_TEXTURES
;
5788 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.samplers
[WINED3D_SHADER_TYPE_PIXEL
], WINED3D_MAX_TEXTURES
);
5791 static DWORD
arbfp_get_emul_mask(const struct wined3d_gl_info
*gl_info
)
5793 return GL_EXT_EMUL_ARB_MULTITEXTURE
| GL_EXT_EMUL_EXT_FOG_COORD
;
5796 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5797 const struct wined3d_state
*state
, DWORD state_id
)
5799 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5800 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5801 struct wined3d_device
*device
= context
->device
;
5802 struct wined3d_color color
;
5804 if (device
->shader_backend
== &arb_program_shader_backend
)
5806 struct shader_arb_priv
*priv
;
5808 /* Don't load the parameter if we're using an arbfp pixel shader,
5809 * otherwise we'll overwrite application provided constants. */
5813 priv
= device
->shader_priv
;
5814 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5815 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5818 wined3d_color_from_d3dcolor(&color
, state
->render_states
[WINED3D_RS_TEXTUREFACTOR
]);
5819 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, &color
.r
));
5820 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, &color.r)");
5823 static void state_tss_constant_arbfp(struct wined3d_context
*context
,
5824 const struct wined3d_state
*state
, DWORD state_id
)
5826 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5827 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5828 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5829 struct wined3d_device
*device
= context
->device
;
5830 struct wined3d_color color
;
5832 if (device
->shader_backend
== &arb_program_shader_backend
)
5834 struct shader_arb_priv
*priv
;
5836 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5837 * application provided constants.
5842 priv
= device
->shader_priv
;
5843 priv
->pshader_const_dirty
[ARB_FFP_CONST_CONSTANT(stage
)] = 1;
5844 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_CONSTANT(stage
) + 1);
5847 wined3d_color_from_d3dcolor(&color
, state
->texture_states
[stage
][WINED3D_TSS_CONSTANT
]);
5848 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_CONSTANT(stage
), &color
.r
));
5849 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_CONSTANT(stage), &color.r)");
5852 static void state_arb_specularenable(struct wined3d_context
*context
,
5853 const struct wined3d_state
*state
, DWORD state_id
)
5855 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5856 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5857 struct wined3d_device
*device
= context
->device
;
5860 if (device
->shader_backend
== &arb_program_shader_backend
)
5862 struct shader_arb_priv
*priv
;
5864 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5865 * application provided constants.
5870 priv
= device
->shader_priv
;
5871 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5872 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5875 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5877 /* The specular color has no alpha */
5878 col
[0] = 1.0f
; col
[1] = 1.0f
;
5879 col
[2] = 1.0f
; col
[3] = 0.0f
;
5881 col
[0] = 0.0f
; col
[1] = 0.0f
;
5882 col
[2] = 0.0f
; col
[3] = 0.0f
;
5884 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5885 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5888 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5890 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5891 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5892 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5893 struct wined3d_device
*device
= context
->device
;
5896 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5898 if (device
->shader_backend
== &arb_program_shader_backend
)
5900 struct shader_arb_priv
*priv
= device
->shader_priv
;
5902 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5906 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5907 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5910 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5911 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5912 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5913 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5915 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5916 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5919 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5920 const struct wined3d_state
*state
, DWORD state_id
)
5922 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5923 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5924 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5925 struct wined3d_device
*device
= context
->device
;
5928 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5930 if (device
->shader_backend
== &arb_program_shader_backend
)
5932 struct shader_arb_priv
*priv
= device
->shader_priv
;
5934 /* Exit now, don't set the luminance below, otherwise we may overwrite pixel shader constants. */
5938 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5939 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5942 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5943 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5947 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5948 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5951 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5953 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5954 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5958 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
5960 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
5962 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
5963 checkGLcall("glEnable GL_ALPHA_TEST");
5967 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
5968 checkGLcall("glDisable GL_ALPHA_TEST");
5972 ref
= wined3d_alpha_ref(state
);
5973 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
5977 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
5978 checkGLcall("glAlphaFunc");
5982 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5984 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5985 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5986 const struct wined3d_texture
*texture
= state
->textures
[0];
5987 struct wined3d_device
*device
= context
->device
;
5988 struct wined3d_color float_key
[2];
5993 if (device
->shader_backend
== &arb_program_shader_backend
)
5995 struct shader_arb_priv
*priv
;
5997 /* Don't load the parameter if we're using an arbfp pixel shader,
5998 * otherwise we'll overwrite application provided constants. */
6002 priv
= device
->shader_priv
;
6003 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_LOW
] = 1;
6004 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_HIGH
] = 1;
6005 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY_HIGH
+ 1);
6008 wined3d_format_get_float_color_key(texture
->resource
.format
, &texture
->async
.src_blt_color_key
, float_key
);
6010 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_LOW
, &float_key
[0].r
));
6011 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_LOW, &float_key[0].r)");
6012 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_HIGH
, &float_key
[1].r
));
6013 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_HIGH, &float_key[1].r)");
6016 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
6020 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6022 switch(arg
& WINED3DTA_SELECTMASK
) {
6023 case WINED3DTA_DIFFUSE
:
6024 ret
= "fragment.color.primary"; break;
6026 case WINED3DTA_CURRENT
:
6030 case WINED3DTA_TEXTURE
:
6032 case 0: ret
= "tex0"; break;
6033 case 1: ret
= "tex1"; break;
6034 case 2: ret
= "tex2"; break;
6035 case 3: ret
= "tex3"; break;
6036 case 4: ret
= "tex4"; break;
6037 case 5: ret
= "tex5"; break;
6038 case 6: ret
= "tex6"; break;
6039 case 7: ret
= "tex7"; break;
6040 default: ret
= "unknown texture";
6044 case WINED3DTA_TFACTOR
:
6045 ret
= "tfactor"; break;
6047 case WINED3DTA_SPECULAR
:
6048 ret
= "fragment.color.secondary"; break;
6050 case WINED3DTA_TEMP
:
6051 ret
= "tempreg"; break;
6053 case WINED3DTA_CONSTANT
:
6055 case 0: ret
= "const0"; break;
6056 case 1: ret
= "const1"; break;
6057 case 2: ret
= "const2"; break;
6058 case 3: ret
= "const3"; break;
6059 case 4: ret
= "const4"; break;
6060 case 5: ret
= "const5"; break;
6061 case 6: ret
= "const6"; break;
6062 case 7: ret
= "const7"; break;
6063 default: ret
= "unknown constant";
6071 if(arg
& WINED3DTA_COMPLEMENT
) {
6072 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6073 if(argnum
== 0) ret
= "arg0";
6074 if(argnum
== 1) ret
= "arg1";
6075 if(argnum
== 2) ret
= "arg2";
6077 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6078 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6079 if(argnum
== 0) ret
= "arg0";
6080 if(argnum
== 1) ret
= "arg1";
6081 if(argnum
== 2) ret
= "arg2";
6086 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6087 BOOL alpha
, BOOL tmp_dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
6089 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6090 unsigned int mul
= 1;
6099 dstreg
= tmp_dst
? "tempreg" : "ret";
6101 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6102 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6103 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6107 case WINED3D_TOP_DISABLE
:
6110 case WINED3D_TOP_SELECT_ARG2
:
6113 case WINED3D_TOP_SELECT_ARG1
:
6114 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6117 case WINED3D_TOP_MODULATE_4X
:
6120 case WINED3D_TOP_MODULATE_2X
:
6123 case WINED3D_TOP_MODULATE
:
6124 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6127 case WINED3D_TOP_ADD_SIGNED_2X
:
6130 case WINED3D_TOP_ADD_SIGNED
:
6131 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6134 case WINED3D_TOP_ADD
:
6135 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6138 case WINED3D_TOP_SUBTRACT
:
6139 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6142 case WINED3D_TOP_ADD_SMOOTH
:
6143 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6144 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6147 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6148 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6149 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6151 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6152 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6153 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6155 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6156 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6157 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6159 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6160 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6161 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6164 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6165 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6166 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6167 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6170 /* D3DTOP_PREMODULATE ???? */
6172 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6173 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6174 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6176 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6177 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6179 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6180 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6181 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6183 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6184 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6187 case WINED3D_TOP_DOTPRODUCT3
:
6189 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6190 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6191 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6194 case WINED3D_TOP_MULTIPLY_ADD
:
6195 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6198 case WINED3D_TOP_LERP
:
6199 /* The msdn is not quite right here */
6200 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6203 case WINED3D_TOP_BUMPENVMAP
:
6204 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6205 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6209 FIXME("Unhandled texture op %08x\n", op
);
6213 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6215 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6218 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6222 case WINED3D_GL_RES_TYPE_TEX_1D
:
6224 case WINED3D_GL_RES_TYPE_TEX_2D
:
6226 case WINED3D_GL_RES_TYPE_TEX_3D
:
6228 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6230 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6233 return "unexpected_resource_type";
6237 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6239 BYTE tex_read
= 0, bump_used
= 0, luminance_used
= 0, constant_used
= 0;
6240 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6241 unsigned int stage
, lowest_disabled_stage
;
6242 struct wined3d_string_buffer buffer
;
6243 struct color_fixup_masks masks
;
6244 BOOL custom_linear_fog
= FALSE
;
6245 const char *textype
, *instr
;
6246 DWORD arg0
, arg1
, arg2
;
6247 char colorcor_dst
[8];
6251 if (!string_buffer_init(&buffer
))
6253 ERR("Failed to initialize shader buffer.\n");
6257 shader_addline(&buffer
, "!!ARBfp1.0\n");
6259 if (settings
->color_key_enabled
)
6261 shader_addline(&buffer
, "PARAM color_key_low = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_LOW
);
6262 shader_addline(&buffer
, "PARAM color_key_high = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_HIGH
);
6266 /* Find out which textures are read */
6267 for (stage
= 0; stage
< WINED3D_MAX_TEXTURES
; ++stage
)
6269 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6272 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6273 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6274 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6276 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6277 tex_read
|= 1u << stage
;
6278 if (settings
->op
[stage
].tmp_dst
)
6279 tempreg_used
= TRUE
;
6280 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6281 tempreg_used
= TRUE
;
6282 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6283 tfactor_used
= TRUE
;
6284 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6285 constant_used
|= 1u << stage
;
6287 switch (settings
->op
[stage
].cop
)
6289 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6290 luminance_used
|= 1u << stage
;
6292 case WINED3D_TOP_BUMPENVMAP
:
6293 bump_used
|= 1u << stage
;
6295 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6296 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6297 tex_read
|= 1u << stage
;
6300 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6301 tfactor_used
= TRUE
;
6308 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6311 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6312 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6313 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6315 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6316 tex_read
|= 1u << stage
;
6317 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6318 tempreg_used
= TRUE
;
6319 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6320 tfactor_used
= TRUE
;
6321 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6322 constant_used
|= 1u << stage
;
6324 lowest_disabled_stage
= stage
;
6326 switch (settings
->fog
)
6328 case WINED3D_FFP_PS_FOG_OFF
: break;
6329 case WINED3D_FFP_PS_FOG_LINEAR
:
6330 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6332 custom_linear_fog
= TRUE
;
6335 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6338 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6339 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6340 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6343 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6344 shader_addline(&buffer
, "TEMP TMP;\n");
6345 shader_addline(&buffer
, "TEMP ret;\n");
6346 if (tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6347 shader_addline(&buffer
, "TEMP arg0;\n");
6348 shader_addline(&buffer
, "TEMP arg1;\n");
6349 shader_addline(&buffer
, "TEMP arg2;\n");
6350 for (stage
= 0; stage
< WINED3D_MAX_TEXTURES
; ++stage
)
6352 if (constant_used
& (1u << stage
))
6353 shader_addline(&buffer
, "PARAM const%u = program.env[%u];\n", stage
, ARB_FFP_CONST_CONSTANT(stage
));
6355 if (!(tex_read
& (1u << stage
)))
6358 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6360 if (!(bump_used
& (1u << stage
)))
6362 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6364 if (!(luminance_used
& (1u << stage
)))
6366 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6369 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6370 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6372 if (settings
->sRGB_write
)
6374 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6375 shader_arb_append_imm_vec4(&buffer
, &wined3d_srgb_const
[0].x
);
6376 shader_addline(&buffer
, ";\n");
6377 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6378 shader_arb_append_imm_vec4(&buffer
, &wined3d_srgb_const
[1].x
);
6379 shader_addline(&buffer
, ";\n");
6382 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6383 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6385 if (tempreg_used
|| settings
->sRGB_write
)
6386 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6388 /* Generate texture sampling instructions */
6389 for (stage
= 0; stage
< WINED3D_MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6391 if (!(tex_read
& (1u << stage
)))
6394 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6396 if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_NONE
)
6400 else if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT4
6401 || settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT3
)
6407 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6412 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6413 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6415 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6416 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6417 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6418 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6420 /* With projective textures, texbem only divides the static
6421 * texture coordinate, not the displacement, so multiply the
6422 * displacement with the dividing parameter before passing it to
6424 if (settings
->op
[stage
].projected
!= WINED3D_PROJECTION_NONE
)
6426 if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT4
)
6428 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6429 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n",
6434 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6435 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n",
6441 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6444 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6445 instr
, stage
, stage
, textype
);
6446 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6448 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6449 stage
- 1, stage
- 1, stage
- 1);
6450 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6453 else if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT3
)
6455 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6456 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6457 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6458 instr
, stage
, stage
, textype
);
6462 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6463 instr
, stage
, stage
, stage
, textype
);
6466 sprintf(colorcor_dst
, "tex%u", stage
);
6467 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6468 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6469 settings
->op
[stage
].color_fixup
, masks
);
6472 if (settings
->color_key_enabled
)
6474 shader_addline(&buffer
, "SLT TMP, tex0, color_key_low;\n"); /* below low key */
6475 shader_addline(&buffer
, "SGE ret, tex0, color_key_high;\n"); /* above high key */
6476 shader_addline(&buffer
, "ADD TMP, TMP, ret;\n"); /* or */
6477 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n"); /* on any channel */
6478 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n"); /* logical not */
6479 shader_addline(&buffer
, "KIL -TMP;\n"); /* discard if true */
6482 shader_addline(&buffer
, "MOV ret, fragment.color.primary;\n");
6484 /* Generate the main shader */
6485 for (stage
= 0; stage
< WINED3D_MAX_TEXTURES
; ++stage
)
6487 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6490 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6491 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6492 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6493 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6494 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6495 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6496 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6497 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6498 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6499 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6500 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6501 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6503 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6504 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6505 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6506 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6508 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6510 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].tmp_dst
,
6511 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6512 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6516 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].tmp_dst
,
6517 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6518 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6520 else if (settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP
6521 && settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6523 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].tmp_dst
,
6524 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6525 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6526 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].tmp_dst
,
6527 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6528 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6532 if (settings
->sRGB_write
|| custom_linear_fog
)
6534 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, ret;\n");
6535 if (settings
->sRGB_write
)
6536 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6537 if (custom_linear_fog
)
6538 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6539 shader_addline(&buffer
, "MOV result.color, ret;\n");
6543 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, ret;\n");
6547 shader_addline(&buffer
, "END\n");
6549 /* Generate the shader */
6550 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6551 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6552 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6554 string_buffer_free(&buffer
);
6558 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6560 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
6561 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
6562 const struct wined3d_device
*device
= context
->device
;
6563 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6564 BOOL use_pshader
= use_ps(state
);
6565 struct ffp_frag_settings settings
;
6566 const struct arbfp_ffp_desc
*desc
;
6569 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6571 if (isStateDirty(context
, STATE_RENDER(WINED3D_RS_FOGENABLE
)))
6573 if (!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6575 /* Reload fixed function constants since they collide with the
6576 * pixel shader constants. */
6577 for (i
= 0; i
< WINED3D_MAX_TEXTURES
; ++i
)
6579 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6580 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6582 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6583 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6584 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6586 else if (use_pshader
)
6588 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6595 /* Find or create a shader implementing the fixed function pipeline
6596 * settings, then activate it. */
6597 gen_ffp_frag_op(context
, state
, &settings
, FALSE
);
6598 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6601 struct arbfp_ffp_desc
*new_desc
;
6603 if (!(new_desc
= heap_alloc(sizeof(*new_desc
))))
6605 ERR("Out of memory\n");
6609 new_desc
->parent
.settings
= settings
;
6610 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6611 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6612 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6616 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6617 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6620 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6621 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6622 priv
->current_fprogram_id
= desc
->shader
;
6624 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6626 /* Reload fixed function constants since they collide with the
6627 * pixel shader constants. */
6628 for (i
= 0; i
< WINED3D_MAX_TEXTURES
; ++i
)
6630 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6631 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6633 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6634 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6635 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6637 context
->last_was_pshader
= FALSE
;
6639 else if (!context
->last_was_pshader
)
6641 if (device
->shader_backend
== &arb_program_shader_backend
)
6642 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6643 context
->last_was_pshader
= TRUE
;
6646 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6649 /* We can't link the fog states to the fragment state directly since the
6650 * vertex pipeline links them to FOGENABLE. A different linking in different
6651 * pipeline parts can't be expressed in the combined state table, so we need
6652 * to handle that with a forwarding function. The other invisible side effect
6653 * is that changing the fog start and fog end (which links to FOGENABLE in
6654 * vertex) results in the fragment_prog_arbfp function being called because
6655 * FOGENABLE is dirty, which calls this function here. */
6656 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6658 enum fogsource new_source
;
6659 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6660 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6662 TRACE("context %p, state %p, state_id %#x.\n", context
, state
, state_id
);
6664 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6665 fragment_prog_arbfp(context
, state
, state_id
);
6667 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6670 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6674 new_source
= FOGSOURCE_VS
;
6678 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6679 new_source
= FOGSOURCE_COORD
;
6681 new_source
= FOGSOURCE_FFP
;
6686 new_source
= FOGSOURCE_FFP
;
6689 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6691 context
->fog_source
= new_source
;
6692 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6696 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6698 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6699 fragment_prog_arbfp(context
, state
, state_id
);
6702 static const struct wined3d_state_entry_template arbfp_fragmentstate_template
[] =
6704 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6735 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6740 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6741 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6742 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6743 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6744 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6745 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6746 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6747 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6748 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6749 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6750 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6751 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6752 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6753 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6754 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6755 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6756 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6757 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6758 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6759 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6760 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6761 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6762 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6763 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6764 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6765 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6766 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6767 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6768 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6769 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6770 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6771 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6772 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6773 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6774 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6775 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6776 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6777 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6778 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6779 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6780 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6781 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6782 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6783 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6784 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6785 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6786 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6787 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6788 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6789 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6790 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6791 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6792 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6793 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6794 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6795 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6796 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6797 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6798 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6799 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6800 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6801 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6804 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6805 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6806 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6807 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6808 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6809 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6810 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6811 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6812 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6813 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6814 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6815 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6816 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6817 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6818 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6819 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6820 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6821 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6822 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6823 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6824 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6825 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6826 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6827 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6828 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6829 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6830 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6831 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6832 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6833 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6834 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6835 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6836 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6837 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6838 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6839 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6840 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6841 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6842 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6843 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6844 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6845 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6846 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6847 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6848 {STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6849 {STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6850 {STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6851 {STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6852 {STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6853 {STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6854 {STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6855 {STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6856 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6857 {STATE_RENDER(WINED3D_RS_SHADEMODE
), { STATE_RENDER(WINED3D_RS_SHADEMODE
), state_shademode
}, WINED3D_GL_EXT_NONE
},
6858 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6861 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6866 static void arbfp_free_context_data(struct wined3d_context
*context
)
6870 const struct wined3d_fragment_pipe_ops arbfp_fragment_pipeline
=
6874 arbfp_get_emul_mask
,
6877 arbfp_alloc_context_data
,
6878 arbfp_free_context_data
,
6879 shader_arb_color_fixup_supported
,
6880 arbfp_fragmentstate_template
,
6883 struct arbfp_blit_type
6885 enum complex_fixup fixup
: 4;
6886 enum wined3d_gl_resource_type res_type
: 3;
6887 DWORD use_color_key
: 1;
6891 struct arbfp_blit_desc
6894 struct arbfp_blit_type type
;
6895 struct wine_rb_entry entry
;
6898 #define ARBFP_BLIT_PARAM_SIZE 0
6899 #define ARBFP_BLIT_PARAM_COLOR_KEY_LOW 1
6900 #define ARBFP_BLIT_PARAM_COLOR_KEY_HIGH 2
6902 struct wined3d_arbfp_blitter
6904 struct wined3d_blitter blitter
;
6905 struct wine_rb_tree shaders
;
6906 GLuint palette_texture
;
6909 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6911 const struct arbfp_blit_type
*ka
= key
;
6912 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6914 return memcmp(ka
, kb
, sizeof(*ka
));
6917 /* Context activation is done by the caller. */
6918 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *ctx
)
6920 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6921 const struct wined3d_gl_info
*gl_info
;
6922 struct wined3d_context_gl
*context_gl
;
6925 gl_info
= context_gl
->gl_info
;
6927 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6928 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6929 heap_free(entry_arb
);
6932 /* Context activation is done by the caller. */
6933 static void arbfp_blitter_destroy(struct wined3d_blitter
*blitter
, struct wined3d_context
*context
)
6935 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
6936 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
6937 struct wined3d_arbfp_blitter
*arbfp_blitter
;
6938 struct wined3d_blitter
*next
;
6940 if ((next
= blitter
->next
))
6941 next
->ops
->blitter_destroy(next
, &context_gl
->c
);
6943 arbfp_blitter
= CONTAINING_RECORD(blitter
, struct wined3d_arbfp_blitter
, blitter
);
6945 wine_rb_destroy(&arbfp_blitter
->shaders
, arbfp_free_blit_shader
, context_gl
);
6946 checkGLcall("Delete blit programs");
6948 if (arbfp_blitter
->palette_texture
)
6949 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &arbfp_blitter
->palette_texture
);
6951 heap_free(arbfp_blitter
);
6954 static void gen_packed_yuv_read(struct wined3d_string_buffer
*buffer
,
6955 const struct arbfp_blit_type
*type
, char *luminance
)
6958 const char *tex
, *texinstr
= "TXP";
6960 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
6971 tex
= arbfp_texture_target(type
->res_type
);
6972 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
6975 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6976 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6977 * filtering when we sample the texture.
6979 * These are the rules for reading the chroma:
6985 * So we have to get the sampling x position in non-normalized coordinates in integers
6987 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
6989 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6990 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6994 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6996 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6997 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7000 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7001 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
7003 /* Multiply the x coordinate by 0.5 and get the fraction. This gives 0.25
7004 * and 0.75 for the even and odd pixels respectively. */
7005 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
7006 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
7008 /* Sample Pixel 1. */
7009 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7011 /* Put the value into either of the chroma values */
7012 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7013 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
7014 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7015 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
7017 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
7018 * the pixel right to the current one. Otherwise, sample the left pixel.
7019 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
7021 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
7022 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
7023 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7025 /* Put the value into the other chroma */
7026 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7027 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7028 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7029 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7031 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7032 * the current one and lerp the two U and V values
7035 /* This gives the correctly filtered luminance value */
7036 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7039 static void gen_yv12_read(struct wined3d_string_buffer
*buffer
,
7040 const struct arbfp_blit_type
*type
, char *luminance
)
7043 static const float yv12_coef
[]
7044 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7046 tex
= arbfp_texture_target(type
->res_type
);
7048 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7049 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7050 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7051 * pitch of the luminance plane, the packing into the gl texture is a bit
7052 * unfortunate. If the whole texture is interpreted as luminance data it looks
7053 * approximately like this:
7055 * +----------------------------------+----
7067 * +----------------+-----------------+----
7069 * | V even rows | V odd rows |
7071 * +----------------+------------------ -
7073 * | U even rows | U odd rows |
7075 * +----------------+-----------------+----
7079 * So it appears as if there are 4 chroma images, but in fact the odd rows
7080 * in the chroma images are in the same row as the even ones. So it is
7081 * kinda tricky to read
7083 * When reading from rectangle textures, keep in mind that the input y coordinates
7084 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7086 shader_addline(buffer
, "PARAM yv12_coef = ");
7087 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7088 shader_addline(buffer
, ";\n");
7090 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7091 /* the chroma planes have only half the width */
7092 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7094 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7095 * the coordinate. Also read the right side of the image when reading odd lines
7097 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7100 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7102 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7104 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7106 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7107 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7109 /* Read odd lines from the right side (add size * 0.5 to the x coordinate). */
7110 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7111 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7112 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7113 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7115 /* clamp, keep the half pixel origin in mind */
7116 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7117 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7118 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7119 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7123 /* The y coordinate for V is in the range [size, size + size / 4). */
7124 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7125 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7127 /* Read odd lines from the right side (add size * 0.5 to the x coordinate). */
7128 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7129 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7130 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7131 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7132 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7134 /* Make sure to read exactly from the pixel center */
7135 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7136 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7139 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7140 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7141 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7142 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7143 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7145 /* Read the texture, put the result into the output register */
7146 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7147 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7149 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7150 * No need to clamp because we're just reusing the already clamped value from above
7152 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7153 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7155 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7156 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7157 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7159 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7160 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7161 * values due to filtering
7163 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7164 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7166 /* Multiply the y coordinate by 2/3 and clamp it */
7167 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7168 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7169 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7170 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7174 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7175 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7178 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7179 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7180 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7185 static void gen_nv12_read(struct wined3d_string_buffer
*buffer
,
7186 const struct arbfp_blit_type
*type
, char *luminance
)
7189 static const float nv12_coef
[]
7190 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7192 tex
= arbfp_texture_target(type
->res_type
);
7194 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7195 * sized plane where each component is an UV pair. So the effective
7196 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7197 * data it looks approximately like this:
7199 * +----------------------------------+----
7211 * +----------------------------------+----
7212 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7213 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7219 * +----------------------------------+----
7221 * When reading from rectangle textures, keep in mind that the input y coordinates
7222 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7224 shader_addline(buffer
, "PARAM nv12_coef = ");
7225 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7226 shader_addline(buffer
, ";\n");
7228 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7229 /* We only have half the number of chroma pixels. */
7230 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7232 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7234 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7235 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7237 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7239 /* We must not allow filtering horizontally, this would mix U and V.
7240 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7241 * 0.5, so add 0.5. */
7243 /* Convert to non-normalized coordinates so we can find the
7244 * individual pixel. */
7245 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7246 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7247 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7248 * add 0.5 to hit the center of the pixel. */
7249 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7251 /* Convert back to normalized coordinates. */
7252 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7254 /* Clamp, keep the half pixel origin in mind. */
7255 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7256 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7257 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7258 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7262 /* The y coordinate for chroma is in the range [size, size + size / 2). */
7263 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7265 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7266 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7267 * add 0.5 to hit the center of the pixel. */
7268 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7271 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7272 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7273 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7274 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7275 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7277 /* Read the texture, put the result into the output register. */
7278 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7279 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7281 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7284 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7289 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7291 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7292 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7294 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7295 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7296 * values due to filtering. */
7297 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7298 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7300 /* Multiply the y coordinate by 2/3 and clamp it */
7301 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7302 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7303 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7304 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7308 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7309 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7312 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7313 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7314 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7319 /* Context activation is done by the caller. */
7320 static GLuint
gen_p8_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7323 struct wined3d_string_buffer buffer
;
7324 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7326 /* This should not happen because we only use this conversion for
7327 * present blits which don't use color keying. */
7328 if (type
->use_color_key
)
7329 FIXME("Implement P8 color keying.\n");
7332 if (!string_buffer_init(&buffer
))
7334 ERR("Failed to initialize shader buffer.\n");
7338 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7339 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7342 string_buffer_free(&buffer
);
7346 shader_addline(&buffer
, "!!ARBfp1.0\n");
7347 shader_addline(&buffer
, "TEMP index;\n");
7349 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7350 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7352 /* The alpha-component contains the palette index */
7353 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7355 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7356 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7358 /* Use the alpha-component as an index in the palette to get the final color */
7359 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7360 shader_addline(&buffer
, "END\n");
7362 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7364 string_buffer_free(&buffer
);
7369 /* Context activation is done by the caller. */
7370 static void arbfp_blitter_upload_palette(struct wined3d_arbfp_blitter
*blitter
,
7371 const struct wined3d_texture_gl
*texture_gl
, struct wined3d_context_gl
*context_gl
)
7373 const struct wined3d_palette
*palette
= texture_gl
->t
.swapchain
? texture_gl
->t
.swapchain
->palette
: NULL
;
7374 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
7376 if (!blitter
->palette_texture
)
7377 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &blitter
->palette_texture
);
7379 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7380 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, blitter
->palette_texture
);
7382 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7384 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7385 /* Make sure we have discrete color levels. */
7386 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7387 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7388 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7391 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7392 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7396 static const DWORD black
;
7397 FIXME("P8 surface loaded without a palette.\n");
7398 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7399 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7402 /* Switch back to unit 0 in which the 2D texture will be stored. */
7403 wined3d_context_gl_active_texture(context_gl
, gl_info
, 0);
7406 /* Context activation is done by the caller. */
7407 static GLuint
gen_yuv_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7410 struct wined3d_string_buffer buffer
;
7411 char luminance_component
;
7413 if (type
->use_color_key
)
7414 FIXME("Implement YUV color keying.\n");
7417 if (!string_buffer_init(&buffer
))
7419 ERR("Failed to initialize shader buffer.\n");
7423 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7424 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7425 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7426 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7429 string_buffer_free(&buffer
);
7433 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7434 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7435 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7436 * each single pixel it contains, and one U and one V value shared between both
7439 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7440 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7441 * take the format into account when generating the read swizzles
7443 * Reading the Y value is straightforward - just sample the texture. The hardware
7444 * takes care of filtering in the horizontal and vertical direction.
7446 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7447 * because that would mix the U and V values of one pixel or two adjacent pixels.
7448 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7449 * regardless of the filtering setting. Vertical filtering works automatically
7450 * though - the U and V values of two rows are mixed nicely.
7452 * Apart of avoiding filtering issues, the code has to know which value it just
7453 * read, and where it can find the other one. To determine this, it checks if
7454 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7456 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7457 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7459 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7460 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7461 * in an unfiltered situation. Finding the luminance on the other hand requires
7462 * finding out if it is an odd or even pixel. The real drawback of this approach
7463 * is filtering. This would have to be emulated completely in the shader, reading
7464 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7465 * vertically. Beyond that it would require adjustments to the texture handling
7466 * code to deal with the width scaling
7468 shader_addline(&buffer
, "!!ARBfp1.0\n");
7469 shader_addline(&buffer
, "TEMP luminance;\n");
7470 shader_addline(&buffer
, "TEMP temp;\n");
7471 shader_addline(&buffer
, "TEMP chroma;\n");
7472 shader_addline(&buffer
, "TEMP texcrd;\n");
7473 shader_addline(&buffer
, "TEMP texcrd2;\n");
7474 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7475 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7476 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7478 switch (type
->fixup
)
7480 case COMPLEX_FIXUP_UYVY
:
7481 case COMPLEX_FIXUP_YUY2
:
7482 gen_packed_yuv_read(&buffer
, type
, &luminance_component
);
7485 case COMPLEX_FIXUP_YV12
:
7486 gen_yv12_read(&buffer
, type
, &luminance_component
);
7489 case COMPLEX_FIXUP_NV12
:
7490 gen_nv12_read(&buffer
, type
, &luminance_component
);
7494 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7495 string_buffer_free(&buffer
);
7499 /* Calculate the final result. Formula is taken from
7500 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7501 * ranges from -0.5 to 0.5
7503 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7505 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7506 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7507 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7508 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7509 shader_addline(&buffer
, "END\n");
7511 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7513 string_buffer_free(&buffer
);
7518 /* Context activation is done by the caller. */
7519 static GLuint
arbfp_gen_plain_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7522 struct wined3d_string_buffer buffer
;
7523 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7526 if (!string_buffer_init(&buffer
))
7528 ERR("Failed to initialize shader buffer.\n");
7532 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7535 string_buffer_free(&buffer
);
7538 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7540 shader_addline(&buffer
, "!!ARBfp1.0\n");
7542 if (type
->use_color_key
)
7544 shader_addline(&buffer
, "TEMP color;\n");
7545 shader_addline(&buffer
, "TEMP less, greater;\n");
7546 shader_addline(&buffer
, "PARAM color_key_low = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_LOW
);
7547 shader_addline(&buffer
, "PARAM color_key_high = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
);
7548 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7549 shader_addline(&buffer
, "SLT less, color, color_key_low;\n"); /* below low key */
7550 shader_addline(&buffer
, "SGE greater, color, color_key_high;\n"); /* above high key */
7551 shader_addline(&buffer
, "ADD less, less, greater;\n"); /* or */
7552 shader_addline(&buffer
, "DP4 less.b, less, less;\n"); /* on any channel */
7553 shader_addline(&buffer
, "SGE less, -less.b, 0.0;\n"); /* logical not */
7554 shader_addline(&buffer
, "KIL -less;\n"); /* discard if true */
7555 shader_addline(&buffer
, "MOV result.color, color;\n");
7559 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7562 shader_addline(&buffer
, "END\n");
7564 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7566 string_buffer_free(&buffer
);
7571 /* Context activation is done by the caller. */
7572 static HRESULT
arbfp_blit_set(struct wined3d_arbfp_blitter
*blitter
, struct wined3d_context_gl
*context_gl
,
7573 const struct wined3d_texture_gl
*texture_gl
, unsigned int sub_resource_idx
,
7574 const struct wined3d_color_key
*color_key
)
7576 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
7577 enum complex_fixup fixup
;
7578 struct wine_rb_entry
*entry
;
7579 struct arbfp_blit_type type
;
7580 struct arbfp_blit_desc
*desc
;
7581 struct wined3d_color float_color_key
[2];
7582 struct wined3d_vec4 size
;
7586 level
= sub_resource_idx
% texture_gl
->t
.level_count
;
7587 size
.x
= wined3d_texture_get_level_pow2_width(&texture_gl
->t
, level
);
7588 size
.y
= wined3d_texture_get_level_pow2_height(&texture_gl
->t
, level
);
7592 if (is_complex_fixup(texture_gl
->t
.resource
.format
->color_fixup
))
7593 fixup
= get_complex_fixup(texture_gl
->t
.resource
.format
->color_fixup
);
7595 fixup
= COMPLEX_FIXUP_NONE
;
7597 switch (texture_gl
->target
)
7600 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7604 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7608 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7611 case GL_TEXTURE_CUBE_MAP_ARB
:
7612 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7615 case GL_TEXTURE_RECTANGLE_ARB
:
7616 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7620 ERR("Unexpected GL texture type %#x.\n", texture_gl
->target
);
7621 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7624 type
.use_color_key
= !!color_key
;
7627 if ((entry
= wine_rb_get(&blitter
->shaders
, &type
)))
7629 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7630 shader
= desc
->shader
;
7636 case COMPLEX_FIXUP_NONE
:
7637 if (!is_identity_fixup(texture_gl
->t
.resource
.format
->color_fixup
))
7638 FIXME("Implement support for sign or swizzle fixups.\n");
7639 shader
= arbfp_gen_plain_shader(gl_info
, &type
);
7642 case COMPLEX_FIXUP_P8
:
7643 shader
= gen_p8_shader(gl_info
, &type
);
7646 case COMPLEX_FIXUP_YUY2
:
7647 case COMPLEX_FIXUP_UYVY
:
7648 case COMPLEX_FIXUP_YV12
:
7649 case COMPLEX_FIXUP_NV12
:
7650 shader
= gen_yuv_shader(gl_info
, &type
);
7654 FIXME("Unsupported fixup %#x.\n", fixup
);
7660 ERR("Failed to get shader for fixup %#x.\n", fixup
);
7664 if (!(desc
= heap_alloc(sizeof(*desc
))))
7668 desc
->shader
= shader
;
7669 if (wine_rb_put(&blitter
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7672 ERR("Out of memory\n");
7673 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7674 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7675 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7676 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7678 return E_OUTOFMEMORY
;
7682 if (fixup
== COMPLEX_FIXUP_P8
)
7683 arbfp_blitter_upload_palette(blitter
, texture_gl
, context_gl
);
7685 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7686 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7687 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7688 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7689 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, &size
.x
));
7690 checkGLcall("glProgramLocalParameter4fvARB");
7691 if (type
.use_color_key
)
7693 wined3d_format_get_float_color_key(texture_gl
->t
.resource
.format
, color_key
, float_color_key
);
7694 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7695 ARBFP_BLIT_PARAM_COLOR_KEY_LOW
, &float_color_key
[0].r
));
7696 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7697 ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
, &float_color_key
[1].r
));
7698 checkGLcall("glProgramLocalParameter4fvARB");
7704 /* Context activation is done by the caller. */
7705 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7707 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7708 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7711 static BOOL
arbfp_blit_supported(enum wined3d_blit_op blit_op
, const struct wined3d_context
*context
,
7712 const struct wined3d_resource
*src_resource
, DWORD src_location
,
7713 const struct wined3d_resource
*dst_resource
, DWORD dst_location
)
7715 const struct wined3d_format
*src_format
= src_resource
->format
;
7716 const struct wined3d_format
*dst_format
= dst_resource
->format
;
7717 enum complex_fixup src_fixup
;
7720 if (src_resource
->type
!= WINED3D_RTYPE_TEXTURE_2D
)
7723 if (blit_op
== WINED3D_BLIT_OP_RAW_BLIT
&& dst_format
->id
== src_format
->id
)
7725 if (dst_format
->flags
[WINED3D_GL_RES_TYPE_TEX_2D
] & (WINED3DFMT_FLAG_DEPTH
| WINED3DFMT_FLAG_STENCIL
))
7726 blit_op
= WINED3D_BLIT_OP_DEPTH_BLIT
;
7728 blit_op
= WINED3D_BLIT_OP_COLOR_BLIT
;
7733 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7734 if (!context
->d3d_info
->shader_color_key
)
7736 /* The conversion modifies the alpha channel so the color key might no longer match. */
7737 TRACE("Color keying not supported with converted textures.\n");
7740 case WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
:
7741 case WINED3D_BLIT_OP_COLOR_BLIT
:
7745 TRACE("Unsupported blit_op=%d\n", blit_op
);
7749 decompress
= (src_format
->flags
[WINED3D_GL_RES_TYPE_TEX_2D
] & WINED3DFMT_FLAG_COMPRESSED
)
7750 && !(dst_format
->flags
[WINED3D_GL_RES_TYPE_TEX_2D
] & WINED3DFMT_FLAG_COMPRESSED
);
7751 if (!decompress
&& !(src_resource
->access
& dst_resource
->access
& WINED3D_RESOURCE_ACCESS_GPU
))
7754 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7755 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7757 TRACE("Checking support for fixup:\n");
7758 dump_color_fixup_desc(src_format
->color_fixup
);
7761 if (!is_identity_fixup(dst_format
->color_fixup
)
7762 && (dst_format
->id
!= src_format
->id
|| dst_location
!= WINED3D_LOCATION_DRAWABLE
))
7764 TRACE("Destination fixups are not supported\n");
7768 if (is_identity_fixup(src_format
->color_fixup
))
7774 /* We only support YUV conversions. */
7775 if (!is_complex_fixup(src_format
->color_fixup
))
7777 if (wined3d_settings
.offscreen_rendering_mode
== ORM_BACKBUFFER
)
7779 WARN("Claiming fixup support because of ORM_BACKBUFFER.\n");
7783 TRACE("[FAILED]\n");
7789 case COMPLEX_FIXUP_YUY2
:
7790 case COMPLEX_FIXUP_UYVY
:
7791 case COMPLEX_FIXUP_YV12
:
7792 case COMPLEX_FIXUP_NV12
:
7793 case COMPLEX_FIXUP_P8
:
7798 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7799 TRACE("[FAILED]\n");
7804 static DWORD
arbfp_blitter_blit(struct wined3d_blitter
*blitter
, enum wined3d_blit_op op
,
7805 struct wined3d_context
*context
, struct wined3d_texture
*src_texture
, unsigned int src_sub_resource_idx
,
7806 DWORD src_location
, const RECT
*src_rect
, struct wined3d_texture
*dst_texture
,
7807 unsigned int dst_sub_resource_idx
, DWORD dst_location
, const RECT
*dst_rect
,
7808 const struct wined3d_color_key
*color_key
, enum wined3d_texture_filter_type filter
)
7810 struct wined3d_texture_gl
*src_texture_gl
= wined3d_texture_gl(src_texture
);
7811 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
7812 struct wined3d_device
*device
= dst_texture
->resource
.device
;
7813 struct wined3d_texture
*staging_texture
= NULL
;
7814 struct wined3d_arbfp_blitter
*arbfp_blitter
;
7815 struct wined3d_color_key alpha_test_key
;
7816 struct wined3d_blitter
*next
;
7817 unsigned int src_level
;
7820 TRACE("blitter %p, op %#x, context %p, src_texture %p, src_sub_resource_idx %u, src_location %s, src_rect %s, "
7821 "dst_texture %p, dst_sub_resource_idx %u, dst_location %s, dst_rect %s, colour_key %p, filter %s.\n",
7822 blitter
, op
, context
, src_texture
, src_sub_resource_idx
, wined3d_debug_location(src_location
),
7823 wine_dbgstr_rect(src_rect
), dst_texture
, dst_sub_resource_idx
, wined3d_debug_location(dst_location
),
7824 wine_dbgstr_rect(dst_rect
), color_key
, debug_d3dtexturefiltertype(filter
));
7826 if (!arbfp_blit_supported(op
, context
, &src_texture
->resource
, src_location
,
7827 &dst_texture
->resource
, dst_location
))
7829 if (!(next
= blitter
->next
))
7831 ERR("No blitter to handle blit op %#x.\n", op
);
7832 return dst_location
;
7835 TRACE("Forwarding to blitter %p.\n", next
);
7836 return next
->ops
->blitter_blit(next
, op
, context
, src_texture
, src_sub_resource_idx
, src_location
,
7837 src_rect
, dst_texture
, dst_sub_resource_idx
, dst_location
, dst_rect
, color_key
, filter
);
7840 arbfp_blitter
= CONTAINING_RECORD(blitter
, struct wined3d_arbfp_blitter
, blitter
);
7842 if (!(src_texture
->resource
.access
& WINED3D_RESOURCE_ACCESS_GPU
))
7844 struct wined3d_resource_desc desc
;
7845 struct wined3d_box upload_box
;
7848 TRACE("Source texture is not GPU accessible, creating a staging texture.\n");
7850 src_level
= src_sub_resource_idx
% src_texture
->level_count
;
7851 desc
.resource_type
= WINED3D_RTYPE_TEXTURE_2D
;
7852 desc
.format
= src_texture
->resource
.format
->id
;
7853 desc
.multisample_type
= src_texture
->resource
.multisample_type
;
7854 desc
.multisample_quality
= src_texture
->resource
.multisample_quality
;
7855 desc
.usage
= WINED3DUSAGE_PRIVATE
;
7856 desc
.bind_flags
= 0;
7857 desc
.access
= WINED3D_RESOURCE_ACCESS_GPU
;
7858 desc
.width
= wined3d_texture_get_level_width(src_texture
, src_level
);
7859 desc
.height
= wined3d_texture_get_level_height(src_texture
, src_level
);
7863 if (FAILED(hr
= wined3d_texture_create(device
, &desc
, 1, 1, 0,
7864 NULL
, NULL
, &wined3d_null_parent_ops
, &staging_texture
)))
7866 ERR("Failed to create staging texture, hr %#x.\n", hr
);
7867 return dst_location
;
7870 wined3d_box_set(&upload_box
, 0, 0, desc
.width
, desc
.height
, 0, desc
.depth
);
7871 wined3d_texture_upload_from_texture(staging_texture
, 0, 0, 0, 0,
7872 src_texture
, src_sub_resource_idx
, &upload_box
);
7874 src_texture
= staging_texture
;
7875 src_texture_gl
= wined3d_texture_gl(src_texture
);
7876 src_sub_resource_idx
= 0;
7878 else if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7879 && (src_texture
->sub_resources
[src_sub_resource_idx
].locations
7880 & (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
)) == WINED3D_LOCATION_DRAWABLE
7881 && !wined3d_resource_is_offscreen(&src_texture
->resource
))
7884 /* Without FBO blits transferring from the drawable to the texture is
7885 * expensive, because we have to flip the data in sysmem. Since we can
7886 * flip in the blitter, we don't actually need that flip anyway. So we
7887 * use the surface's texture as scratch texture, and flip the source
7888 * rectangle instead. */
7889 texture2d_load_fb_texture(src_texture_gl
, src_sub_resource_idx
, FALSE
, context
);
7892 src_level
= src_sub_resource_idx
% src_texture
->level_count
;
7893 s
.top
= wined3d_texture_get_level_height(src_texture
, src_level
) - s
.top
;
7894 s
.bottom
= wined3d_texture_get_level_height(src_texture
, src_level
) - s
.bottom
;
7899 wined3d_texture_load(src_texture
, context
, FALSE
);
7902 wined3d_context_gl_apply_ffp_blit_state(context_gl
, device
);
7904 if (dst_location
== WINED3D_LOCATION_DRAWABLE
)
7907 wined3d_texture_translate_drawable_coords(dst_texture
, context_gl
->window
, &d
);
7911 if (wined3d_settings
.offscreen_rendering_mode
== ORM_FBO
)
7915 if (dst_location
== WINED3D_LOCATION_DRAWABLE
)
7917 TRACE("Destination texture %p is onscreen.\n", dst_texture
);
7918 buffer
= wined3d_texture_get_gl_buffer(dst_texture
);
7922 TRACE("Destination texture %p is offscreen.\n", dst_texture
);
7923 buffer
= GL_COLOR_ATTACHMENT0
;
7925 wined3d_context_gl_apply_fbo_state_blit(context_gl
, GL_DRAW_FRAMEBUFFER
,
7926 &dst_texture
->resource
, dst_sub_resource_idx
, NULL
, 0, dst_location
);
7927 wined3d_context_gl_set_draw_buffer(context_gl
, buffer
);
7928 wined3d_context_gl_check_fbo_status(context_gl
, GL_DRAW_FRAMEBUFFER
);
7929 context_invalidate_state(context
, STATE_FRAMEBUFFER
);
7932 if (op
== WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
)
7934 const struct wined3d_format
*fmt
= src_texture
->resource
.format
;
7935 alpha_test_key
.color_space_low_value
= 0;
7936 alpha_test_key
.color_space_high_value
= ~(((1u << fmt
->alpha_size
) - 1) << fmt
->alpha_offset
);
7937 color_key
= &alpha_test_key
;
7940 arbfp_blit_set(arbfp_blitter
, context_gl
, src_texture_gl
, src_sub_resource_idx
, color_key
);
7942 /* Draw a textured quad */
7943 wined3d_context_gl_draw_textured_quad(context_gl
, src_texture_gl
,
7944 src_sub_resource_idx
, src_rect
, dst_rect
, filter
);
7946 /* Leave the opengl state valid for blitting */
7947 arbfp_blit_unset(context_gl
->gl_info
);
7949 if (dst_texture
->swapchain
&& (dst_texture
->swapchain
->front_buffer
== dst_texture
))
7950 context_gl
->gl_info
->gl_ops
.gl
.p_glFlush();
7952 if (staging_texture
)
7953 wined3d_texture_decref(staging_texture
);
7955 return dst_location
;
7958 static void arbfp_blitter_clear(struct wined3d_blitter
*blitter
, struct wined3d_device
*device
,
7959 unsigned int rt_count
, const struct wined3d_fb_state
*fb
, unsigned int rect_count
, const RECT
*clear_rects
,
7960 const RECT
*draw_rect
, DWORD flags
, const struct wined3d_color
*colour
, float depth
, DWORD stencil
)
7962 struct wined3d_blitter
*next
;
7964 if ((next
= blitter
->next
))
7965 next
->ops
->blitter_clear(next
, device
, rt_count
, fb
, rect_count
,
7966 clear_rects
, draw_rect
, flags
, colour
, depth
, stencil
);
7969 static const struct wined3d_blitter_ops arbfp_blitter_ops
=
7971 arbfp_blitter_destroy
,
7972 arbfp_blitter_clear
,
7976 void wined3d_arbfp_blitter_create(struct wined3d_blitter
**next
, const struct wined3d_device
*device
)
7978 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
7979 struct wined3d_arbfp_blitter
*blitter
;
7981 if (device
->shader_backend
!= &arb_program_shader_backend
7982 && device
->shader_backend
!= &glsl_shader_backend
)
7985 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7988 if (!gl_info
->supported
[WINED3D_GL_LEGACY_CONTEXT
])
7991 if (!(blitter
= heap_alloc(sizeof(*blitter
))))
7993 ERR("Failed to allocate blitter.\n");
7997 TRACE("Created blitter %p.\n", blitter
);
7999 blitter
->blitter
.ops
= &arbfp_blitter_ops
;
8000 blitter
->blitter
.next
= *next
;
8001 wine_rb_init(&blitter
->shaders
, arbfp_blit_type_compare
);
8002 blitter
->palette_texture
= 0;
8003 *next
= &blitter
->blitter
;