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
32 #include "wined3d_private.h"
33 #include "wined3d_gl.h"
35 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
36 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
37 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_perf
);
40 static BOOL
shader_is_pshader_version(enum wined3d_shader_type type
)
42 return type
== WINED3D_SHADER_TYPE_PIXEL
;
45 static BOOL
shader_is_vshader_version(enum wined3d_shader_type type
)
47 return type
== WINED3D_SHADER_TYPE_VERTEX
;
61 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
63 if (shader
!= WINED3D_SHADER_TYPE_VERTEX
&& shader
!= WINED3D_SHADER_TYPE_PIXEL
)
65 ERR("Unsupported shader type '%s'.\n", debug_shader_type(shader
));
69 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
73 case ARB_ZERO
: return "ps_helper_const.x";
74 case ARB_ONE
: return "ps_helper_const.y";
75 case ARB_TWO
: return "coefmul.x";
76 case ARB_0001
: return "ps_helper_const.xxxy";
77 case ARB_EPS
: return "ps_helper_const.z";
85 case ARB_ZERO
: return "helper_const.x";
86 case ARB_ONE
: return "helper_const.y";
87 case ARB_TWO
: return "helper_const.z";
88 case ARB_EPS
: return "helper_const.w";
89 case ARB_0001
: return "helper_const.xxxy";
90 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
93 FIXME("Unmanaged %s shader helper constant requested: %u.\n",
94 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
97 case ARB_ZERO
: return "0.0";
98 case ARB_ONE
: return "1.0";
99 case ARB_TWO
: return "2.0";
100 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
101 case ARB_EPS
: return "1e-8";
102 default: return "bad";
106 static inline BOOL
ffp_clip_emul(const struct wined3d_context
*context
)
108 return context
->lowest_disabled_stage
< 7;
111 /* ARB_program_shader private data */
130 struct wined3d_shader_loop_control loop_control
;
134 struct arb_ps_np2fixup_info
136 struct ps_np2fixup_info super
;
137 /* For ARB we need an offset value:
138 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
139 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
140 * array we need an offset to the index inside the program local parameter array. */
144 struct arb_ps_compile_args
146 struct ps_compile_args super
;
148 WORD clip
; /* only a boolean, use a WORD for alignment */
149 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
152 struct stb_const_desc
154 unsigned char texunit
;
158 struct arb_ps_compiled_shader
160 struct arb_ps_compile_args args
;
161 struct arb_ps_np2fixup_info np2fixup_info
;
162 struct stb_const_desc bumpenvmatconst
[WINED3D_MAX_FFP_TEXTURES
];
163 struct stb_const_desc luminanceconst
[WINED3D_MAX_FFP_TEXTURES
];
164 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
167 unsigned char numbumpenvmatconsts
;
171 struct arb_vs_compile_args
173 struct vs_compile_args super
;
179 unsigned char clip_texcoord
;
180 unsigned char clipplane_mask
;
182 DWORD boolclip_compare
;
184 unsigned int ps_signature
;
187 unsigned char samplers
[4];
188 DWORD samplers_compare
;
190 unsigned char loop_ctrl
[WINED3D_MAX_CONSTS_I
][3];
193 struct arb_vs_compiled_shader
195 struct arb_vs_compile_args args
;
197 UINT int_consts
[WINED3D_MAX_CONSTS_I
];
199 char need_color_unclamp
;
203 struct recorded_instruction
205 struct wined3d_shader_instruction ins
;
209 struct shader_arb_ctx_priv
214 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
216 /* GL_NV_vertex_program2_option or GL_NV_fragment_program_option */
218 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
222 const struct wined3d_gl_info
*gl_info
;
223 const struct arb_vs_compile_args
*cur_vs_args
;
224 const struct arb_ps_compile_args
*cur_ps_args
;
225 const struct arb_ps_compiled_shader
*compiled_fprog
;
226 const struct arb_vs_compiled_shader
*compiled_vprog
;
227 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
228 struct list control_frames
;
232 unsigned int num_loops
, loop_depth
, num_ifcs
;
234 BOOL ps_post_process
;
236 unsigned int vs_clipplanes
;
240 /* For 3.0 vertex shaders */
241 const char *vs_output
[MAX_REG_OUTPUT
];
242 /* For 2.x and earlier vertex shaders */
243 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
245 /* 3.0 pshader input for compatibility with fixed function */
246 const char *ps_input
[MAX_REG_INPUT
];
251 struct wined3d_shader_signature sig
;
253 struct wine_rb_entry entry
;
256 struct arb_pshader_private
{
257 struct arb_ps_compiled_shader
*gl_shaders
;
258 UINT num_gl_shaders
, shader_array_size
;
259 unsigned int input_signature_idx
;
260 unsigned int clipplane_emulation
;
264 struct arb_vshader_private
{
265 struct arb_vs_compiled_shader
*gl_shaders
;
266 UINT num_gl_shaders
, shader_array_size
;
270 struct shader_arb_priv
272 const struct arb_ps_compiled_shader
*compiled_fprog
;
273 const struct arb_vs_compiled_shader
*compiled_vprog
;
274 BOOL use_arbfp_fixed_func
;
275 struct wine_rb_tree fragment_shaders
;
276 BOOL last_ps_const_clamped
;
277 BOOL last_vs_color_unclamp
;
279 struct wine_rb_tree signature_tree
;
282 unsigned int highest_dirty_ps_const
, highest_dirty_vs_const
;
283 char vshader_const_dirty
[WINED3D_MAX_VS_CONSTS_F
];
284 char pshader_const_dirty
[WINED3D_MAX_PS_CONSTS_F
];
285 const struct wined3d_context
*last_context
;
287 const struct wined3d_vertex_pipe_ops
*vertex_pipe
;
288 const struct wined3d_fragment_pipe_ops
*fragment_pipe
;
289 BOOL ffp_proj_control
;
292 /* Context activation for state handlers is done by the caller. */
294 static BOOL
need_rel_addr_const(const struct arb_vshader_private
*shader_data
,
295 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
297 if (shader_data
->rel_offset
) return TRUE
;
298 if (!reg_maps
->usesmova
) return FALSE
;
299 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
302 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
303 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
305 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
306 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
309 static BOOL
need_helper_const(const struct arb_vshader_private
*shader_data
,
310 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
312 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) return TRUE
;
313 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
314 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
315 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
316 if (reg_maps
->usesnrm
) return TRUE
; /* 0.0 */
317 if (reg_maps
->usespow
) return TRUE
; /* EPS, 0.0 and 1.0 */
318 if (reg_maps
->fog
) return TRUE
; /* Clamping fog coord, 0.0 and 1.0 */
322 static unsigned int reserved_vs_const(const struct arb_vshader_private
*shader_data
,
323 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
)
325 unsigned int ret
= 1;
326 /* We use one PARAM for the pos fixup, and in some cases one to load
327 * some immediate values into the shader. */
328 if (need_helper_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
329 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
)) ++ret
;
333 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
334 * When constant_list == NULL, it will load all the constants.
336 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
337 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
339 /* Context activation is done by the caller. */
340 static unsigned int shader_arb_load_constants_f(const struct wined3d_shader
*shader
,
341 const struct wined3d_gl_info
*gl_info
, GLuint target_type
, unsigned int max_constants
,
342 const struct wined3d_vec4
*constants
, char *dirty_consts
)
344 struct wined3d_shader_lconst
*lconst
;
345 unsigned int ret
, i
, j
;
347 if (TRACE_ON(d3d_constants
))
349 for (i
= 0; i
< max_constants
; ++i
)
351 if (!dirty_consts
[i
])
353 TRACE_(d3d_constants
)("Loading constant %u: %s.\n", i
, debug_vec4(&constants
[i
]));
359 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
360 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& shader
->reg_maps
.shader_version
.major
== 1)
363 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
364 * shaders, the first 8 constants are marked dirty for reload
366 for (; i
< min(8, max_constants
); ++i
)
368 if (!dirty_consts
[i
])
372 if (constants
[i
].x
> 1.0f
)
374 else if (constants
[i
].x
< -1.0f
)
375 lcl_const
[0] = -1.0f
;
377 lcl_const
[0] = constants
[i
].x
;
379 if (constants
[i
].y
> 1.0f
)
381 else if (constants
[i
].y
< -1.0f
)
382 lcl_const
[1] = -1.0f
;
384 lcl_const
[1] = constants
[i
].y
;
386 if (constants
[i
].z
> 1.0f
)
388 else if (constants
[i
].z
< -1.0f
)
389 lcl_const
[2] = -1.0f
;
391 lcl_const
[2] = constants
[i
].z
;
393 if (constants
[i
].w
> 1.0f
)
395 else if (constants
[i
].w
< -1.0f
)
396 lcl_const
[3] = -1.0f
;
398 lcl_const
[3] = constants
[i
].w
;
400 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
403 /* If further constants are dirty, reload them without clamping.
405 * The alternative is not to touch them, but then we cannot reset the dirty constant count
406 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
407 * above would always re-check the first 8 constants since max_constant remains at the init
412 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
414 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
415 * or just reloading *all* constants at once
417 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
419 for (; i
< max_constants
; ++i
)
421 if (!dirty_consts
[i
])
424 /* Find the next block of dirty constants */
427 for (++i
; (i
< max_constants
) && dirty_consts
[i
]; ++i
)
432 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, &constants
[j
].x
));
437 for (; i
< max_constants
; ++i
)
442 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, &constants
[i
].x
));
446 checkGLcall("glProgramEnvParameter4fvARB()");
448 /* Load immediate constants */
449 if (shader
->load_local_constsF
)
451 if (TRACE_ON(d3d_shader
))
453 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
455 GLfloat
* values
= (GLfloat
*)lconst
->value
;
456 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
457 values
[0], values
[1], values
[2], values
[3]);
460 /* Immediate constants are clamped for 1.X shaders at loading times */
462 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
464 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
465 ret
= max(ret
, lconst
->idx
+ 1);
466 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
468 checkGLcall("glProgramEnvParameter4fvARB()");
469 return ret
; /* The loaded immediate constants need reloading for the next shader */
471 return 0; /* No constants are dirty now */
475 /* Loads the texture dimensions for NP2 fixup into the currently set
476 * ARB_[vertex/fragment]_programs. */
477 static void shader_arb_load_np2fixup_constants(const struct arb_ps_np2fixup_info
*fixup
,
478 const struct wined3d_gl_info
*gl_info
, const struct wined3d_state
*state
)
480 GLfloat np2fixup_constants
[4 * WINED3D_MAX_FRAGMENT_SAMPLERS
];
481 uint32_t active
= fixup
->super
.active
;
482 const struct wined3d_texture
*tex
;
492 i
= wined3d_bit_scan(&active
);
493 if (!(tex
= state
->textures
[i
]))
495 ERR("Nonexistent texture is flagged for NP2 texcoord fixup.\n");
499 idx
= fixup
->super
.idx
[i
];
500 tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
503 tex_dim
[2] = tex
->pow2_matrix
[0];
504 tex_dim
[3] = tex
->pow2_matrix
[5];
508 tex_dim
[0] = tex
->pow2_matrix
[0];
509 tex_dim
[1] = tex
->pow2_matrix
[5];
513 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
)
515 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
516 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
520 /* Context activation is done by the caller. */
521 static void shader_arb_ps_local_constants(const struct arb_ps_compiled_shader
*gl_shader
,
522 struct wined3d_context_gl
*context_gl
, const struct wined3d_state
*state
, unsigned int rt_height
)
524 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
525 struct wined3d_device
*device
= context_gl
->c
.device
;
526 const struct wined3d_ivec4
*int_consts
;
529 int_consts
= wined3d_buffer_load_sysmem(device
->push_constants
[WINED3D_PUSH_CONSTANTS_PS_I
], &context_gl
->c
);
531 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
533 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
535 /* The state manager takes care that this function is always called if the bump env matrix changes */
536 const float *data
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_MAT00
];
537 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
538 gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
540 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
542 /* WINED3D_TSS_BUMPENVLSCALE and WINED3D_TSS_BUMPENVLOFFSET are next to each other.
543 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
544 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
545 * (they're WINED3D_TSS_TEXTURETRANSFORMFLAGS and WINED3D_TSS_ADDRESSW, so most likely 0 or NaN
547 const float *scale
= (const float *)&state
->texture_states
[texunit
][WINED3D_TSS_BUMPENV_LSCALE
];
548 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
549 gl_shader
->luminanceconst
[i
].const_num
, scale
));
552 checkGLcall("Load bumpmap consts");
554 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
556 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
557 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
562 val
[0] = context_gl
->c
.render_offscreen
? 0.0f
: (float)rt_height
;
563 val
[1] = context_gl
->c
.render_offscreen
? 1.0f
: -1.0f
;
566 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
567 checkGLcall("y correction loading");
570 if (!gl_shader
->num_int_consts
) return;
572 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
574 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
577 val
[0] = (float)int_consts
[i
].x
;
578 val
[1] = (float)int_consts
[i
].y
;
579 val
[2] = (float)int_consts
[i
].z
;
582 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
585 checkGLcall("Load ps int consts");
588 /* Context activation is done by the caller. */
589 static void shader_arb_vs_local_constants(const struct arb_vs_compiled_shader
*gl_shader
,
590 struct wined3d_context_gl
*context_gl
, const struct wined3d_state
*state
)
592 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
593 struct wined3d_device
*device
= context_gl
->c
.device
;
594 const struct wined3d_ivec4
*int_consts
;
595 float position_fixup
[4];
598 /* Upload the position fixup */
599 shader_get_position_fixup(&context_gl
->c
, state
, 1, position_fixup
);
600 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, position_fixup
));
602 if (!gl_shader
->num_int_consts
) return;
604 int_consts
= wined3d_buffer_load_sysmem(device
->push_constants
[WINED3D_PUSH_CONSTANTS_VS_I
], &context_gl
->c
);
606 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
608 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
611 val
[0] = (float)int_consts
[i
].x
;
612 val
[1] = (float)int_consts
[i
].y
;
613 val
[2] = (float)int_consts
[i
].z
;
616 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
619 checkGLcall("Load vs int consts");
622 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
623 const struct wined3d_state
*state
);
626 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
628 * We only support float constants in ARB at the moment, so don't
629 * worry about the Integers or Booleans
631 /* Context activation is done by the caller (state handler). */
632 static void shader_arb_load_constants_internal(struct shader_arb_priv
*priv
, struct wined3d_context_gl
*context_gl
,
633 const struct wined3d_state
*state
, BOOL use_ps
, BOOL use_vs
, BOOL from_shader_select
)
635 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
636 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
637 struct wined3d_device
*device
= context_gl
->c
.device
;
639 if (!from_shader_select
)
641 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
642 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
645 && (vshader
->reg_maps
.boolean_constants
646 || (!gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
647 && (vshader
->reg_maps
.integer_constants
& ~vshader
->reg_maps
.local_int_consts
))))
649 TRACE("bool/integer vertex shader constants potentially modified, forcing shader reselection.\n");
650 shader_arb_select(priv
, &context_gl
->c
, state
);
653 && (pshader
->reg_maps
.boolean_constants
654 || (!gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
]
655 && (pshader
->reg_maps
.integer_constants
& ~pshader
->reg_maps
.local_int_consts
))))
657 TRACE("bool/integer pixel shader constants potentially modified, forcing shader reselection.\n");
658 shader_arb_select(priv
, &context_gl
->c
, state
);
662 if (&context_gl
->c
!= priv
->last_context
)
664 memset(priv
->vshader_const_dirty
, 1,
665 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
666 priv
->highest_dirty_vs_const
= d3d_info
->limits
.vs_uniform_count
;
668 memset(priv
->pshader_const_dirty
, 1,
669 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
670 priv
->highest_dirty_ps_const
= d3d_info
->limits
.ps_uniform_count
;
672 priv
->last_context
= &context_gl
->c
;
677 const struct wined3d_shader
*vshader
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
678 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
680 /* Load DirectX 9 float constants for vertex shader */
681 priv
->highest_dirty_vs_const
= shader_arb_load_constants_f(vshader
,
682 gl_info
, GL_VERTEX_PROGRAM_ARB
, priv
->highest_dirty_vs_const
,
683 wined3d_buffer_load_sysmem(device
->push_constants
[WINED3D_PUSH_CONSTANTS_VS_F
], &context_gl
->c
),
684 priv
->vshader_const_dirty
);
685 shader_arb_vs_local_constants(gl_shader
, context_gl
, state
);
690 const struct wined3d_shader
*pshader
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
691 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
692 UINT rt_height
= state
->fb
.render_targets
[0]->height
;
694 /* Load DirectX 9 float constants for pixel shader */
695 priv
->highest_dirty_ps_const
= shader_arb_load_constants_f(pshader
,
696 gl_info
, GL_FRAGMENT_PROGRAM_ARB
, priv
->highest_dirty_ps_const
,
697 wined3d_buffer_load_sysmem(device
->push_constants
[WINED3D_PUSH_CONSTANTS_PS_F
], &context_gl
->c
),
698 priv
->pshader_const_dirty
);
699 shader_arb_ps_local_constants(gl_shader
, context_gl
, state
, rt_height
);
701 if (context_gl
->c
.constant_update_mask
& WINED3D_SHADER_CONST_PS_NP2_FIXUP
)
702 shader_arb_load_np2fixup_constants(&gl_shader
->np2fixup_info
, gl_info
, state
);
706 static void shader_arb_load_constants(void *shader_priv
, struct wined3d_context
*context
,
707 const struct wined3d_state
*state
)
709 shader_arb_load_constants_internal(shader_priv
, wined3d_context_gl(context
),
710 state
, use_ps(state
), use_vs(state
), FALSE
);
713 static void shader_arb_update_float_vertex_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
715 struct wined3d_context_gl
*context_gl
= wined3d_context_gl_get_current();
716 struct shader_arb_priv
*priv
= device
->shader_priv
;
718 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
719 * context. On a context switch the old context will be fully dirtified */
720 if (!context_gl
|| context_gl
->c
.device
!= device
)
723 memset(priv
->vshader_const_dirty
+ start
, 1, sizeof(*priv
->vshader_const_dirty
) * count
);
724 priv
->highest_dirty_vs_const
= max(priv
->highest_dirty_vs_const
, start
+ count
);
727 static void shader_arb_update_float_pixel_constants(struct wined3d_device
*device
, UINT start
, UINT count
)
729 struct wined3d_context_gl
*context_gl
= wined3d_context_gl_get_current();
730 struct shader_arb_priv
*priv
= device
->shader_priv
;
732 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
733 * context. On a context switch the old context will be fully dirtified */
734 if (!context_gl
|| context_gl
->c
.device
!= device
)
737 memset(priv
->pshader_const_dirty
+ start
, 1, sizeof(*priv
->pshader_const_dirty
) * count
);
738 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, start
+ count
);
741 static void shader_arb_append_imm_vec4(struct wined3d_string_buffer
*buffer
, const float *values
)
745 wined3d_ftoa(values
[0], str
[0]);
746 wined3d_ftoa(values
[1], str
[1]);
747 wined3d_ftoa(values
[2], str
[2]);
748 wined3d_ftoa(values
[3], str
[3]);
749 shader_addline(buffer
, "{%s, %s, %s, %s}", str
[0], str
[1], str
[2], str
[3]);
752 /* Generate the variable & register declarations for the ARB_vertex_program output target */
753 static void shader_generate_arb_declarations(const struct wined3d_shader
*shader
,
754 const struct wined3d_shader_reg_maps
*reg_maps
, struct wined3d_string_buffer
*buffer
,
755 const struct wined3d_gl_info
*gl_info
, unsigned int *num_clipplanes
,
756 const struct shader_arb_ctx_priv
*ctx
)
759 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
760 const struct wined3d_shader_lconst
*lconst
;
761 unsigned max_constantsF
;
764 /* In pixel shaders, all private constants are program local, we don't need anything
765 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
766 * If we need a private constant the GL implementation will squeeze it in somewhere
768 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
769 * immediate values. The posFixup is loaded using program.env for now, so always
770 * subtract one from the number of constants. If the shader uses indirect addressing,
771 * account for the helper const too because we have to declare all available d3d constants
772 * and don't know which are actually used.
776 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
777 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
778 if (max_constantsF
< 24)
779 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
783 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
784 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
785 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
786 * Also prevents max_constantsF from becoming less than 0 and
788 if (max_constantsF
< 96)
789 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
791 if (reg_maps
->usesrelconstF
)
793 DWORD highest_constf
= 0, clip_limit
;
795 max_constantsF
-= reserved_vs_const(shader_data
, reg_maps
, gl_info
);
796 max_constantsF
-= wined3d_popcount(reg_maps
->integer_constants
);
797 max_constantsF
-= gl_info
->reserved_arb_constants
;
799 for (i
= 0; i
< shader
->limits
->constant_float
; ++i
)
801 if (wined3d_bitmap_is_set(reg_maps
->constf
, i
))
805 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
807 if(ctx
->cur_vs_args
->super
.clip_enabled
)
808 clip_limit
= gl_info
->limits
.user_clip_distances
;
814 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
815 clip_limit
= min(wined3d_popcount(mask
), 4);
817 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
818 max_constantsF
-= *num_clipplanes
;
819 if(*num_clipplanes
< clip_limit
)
821 WARN("Only %u clip planes out of %u enabled.\n", *num_clipplanes
,
822 gl_info
->limits
.user_clip_distances
);
827 if (ctx
->target_version
>= NV2
)
828 *num_clipplanes
= gl_info
->limits
.user_clip_distances
;
830 *num_clipplanes
= min(gl_info
->limits
.user_clip_distances
, 4);
834 map
= reg_maps
->temporary
;
837 i
= wined3d_bit_scan(&map
);
838 shader_addline(buffer
, "TEMP R%u;\n", i
);
841 map
= reg_maps
->address
;
844 i
= wined3d_bit_scan(&map
);
845 shader_addline(buffer
, "ADDRESS A%u;\n", i
);
848 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
850 map
= reg_maps
->texcoord
;
853 i
= wined3d_bit_scan(&map
);
854 shader_addline(buffer
, "TEMP T%u;\n", i
);
858 if (!shader
->load_local_constsF
)
860 LIST_FOR_EACH_ENTRY(lconst
, &shader
->constantsF
, struct wined3d_shader_lconst
, entry
)
863 value
= (const float *)lconst
->value
;
864 shader_addline(buffer
, "PARAM C%u = ", lconst
->idx
);
865 shader_arb_append_imm_vec4(buffer
, value
);
866 shader_addline(buffer
, ";\n");
870 /* After subtracting privately used constants from the hardware limit(they are loaded as
871 * local constants), make sure the shader doesn't violate the env constant limit
875 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
879 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
882 /* Avoid declaring more constants than needed */
883 max_constantsF
= min(max_constantsF
, shader
->limits
->constant_float
);
885 /* we use the array-based constants array if the local constants are marked for loading,
886 * because then we use indirect addressing, or when the local constant list is empty,
887 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
888 * local constants do not declare the loaded constants as an array because ARB compilers usually
889 * do not optimize unused constants away
891 if (reg_maps
->usesrelconstF
)
893 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
894 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
895 max_constantsF
, max_constantsF
- 1);
899 for (i
= 0; i
< max_constantsF
; ++i
)
901 if (!shader_constant_is_local(shader
, i
) && wined3d_extract_bits(reg_maps
->constf
, i
, 1))
903 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
909 static const char * const shift_tab
[] = {
910 "dummy", /* 0 (none) */
911 "coefmul.x", /* 1 (x2) */
912 "coefmul.y", /* 2 (x4) */
913 "coefmul.z", /* 3 (x8) */
914 "coefmul.w", /* 4 (x16) */
915 "dummy", /* 5 (x32) */
916 "dummy", /* 6 (x64) */
917 "dummy", /* 7 (x128) */
918 "dummy", /* 8 (d256) */
919 "dummy", /* 9 (d128) */
920 "dummy", /* 10 (d64) */
921 "dummy", /* 11 (d32) */
922 "coefdiv.w", /* 12 (d16) */
923 "coefdiv.z", /* 13 (d8) */
924 "coefdiv.y", /* 14 (d4) */
925 "coefdiv.x" /* 15 (d2) */
928 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
929 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
931 char *ptr
= write_mask
;
933 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
936 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
937 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
938 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
939 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
945 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
947 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
948 * but addressed as "rgba". To fix this we need to swap the register's x
949 * and z components. */
950 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
951 char *ptr
= swizzle_str
;
953 /* swizzle bits fields: wwzzyyxx */
954 DWORD swizzle
= param
->swizzle
;
955 DWORD swizzle_x
= swizzle
& 0x03;
956 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
957 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
958 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
960 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
961 * generate a swizzle string. Unless we need to our own swizzling. */
962 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
965 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
966 *ptr
++ = swizzle_chars
[swizzle_x
];
968 *ptr
++ = swizzle_chars
[swizzle_x
];
969 *ptr
++ = swizzle_chars
[swizzle_y
];
970 *ptr
++ = swizzle_chars
[swizzle_z
];
971 *ptr
++ = swizzle_chars
[swizzle_w
];
978 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
980 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
981 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
983 if (!strcmp(priv
->addr_reg
, src
))
986 snprintf(priv
->addr_reg
, sizeof(priv
->addr_reg
), "%s", src
);
987 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
990 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
991 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
993 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
994 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
996 /* oPos, oFog and oPts in D3D */
997 static const char * const rastout_reg_names
[] = {"TMP_OUT", "TMP_FOGCOORD", "result.pointsize"};
998 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
999 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1000 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1001 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1007 case WINED3DSPR_TEMP
:
1008 sprintf(register_name
, "R%u", reg
->idx
[0].offset
);
1011 case WINED3DSPR_INPUT
:
1014 if (reg_maps
->shader_version
.major
< 3)
1016 if (!reg
->idx
[0].offset
)
1017 strcpy(register_name
, "fragment.color.primary");
1019 strcpy(register_name
, "fragment.color.secondary");
1023 if (reg
->idx
[0].rel_addr
)
1026 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1028 if (!strcmp(rel_reg
, "**aL_emul**"))
1030 unsigned int idx
= ctx
->aL
+ reg
->idx
[0].offset
;
1031 if(idx
< MAX_REG_INPUT
)
1033 strcpy(register_name
, ctx
->ps_input
[idx
]);
1037 ERR("Pixel shader input register out of bounds: %u\n", idx
);
1038 sprintf(register_name
, "out_of_bounds_%u", idx
);
1041 else if (reg_maps
->input_registers
& 0x0300)
1043 /* There are two ways basically:
1045 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
1046 * That means trouble if the loop also contains a breakc or if the control values
1047 * aren't local constants.
1048 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
1049 * source dynamically. The trouble is that we cannot simply read aL.y because it
1050 * is an ADDRESS register. We could however push it, load .zw with a value and use
1051 * ADAC to load the condition code register and pop it again afterwards
1053 FIXME("Relative input register addressing with more than 8 registers\n");
1055 /* This is better than nothing for now */
1056 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1058 else if(ctx
->cur_ps_args
->super
.vp_mode
!= WINED3D_VP_MODE_SHADER
)
1060 /* This is problematic because we'd have to consult the ctx->ps_input strings
1061 * for where to find the varying. Some may be "0.0", others can be texcoords or
1062 * colors. This needs either a pipeline replacement to make the vertex shader feed
1063 * proper varyings, or loop unrolling
1065 * For now use the texcoords and hope for the best
1067 FIXME("Non-vertex shader varying input with indirect addressing\n");
1068 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1072 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1073 * pulls GL_NV_fragment_program2 in
1075 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
[0].offset
);
1080 if (reg
->idx
[0].offset
< MAX_REG_INPUT
)
1082 strcpy(register_name
, ctx
->ps_input
[reg
->idx
[0].offset
]);
1086 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
[0].offset
);
1087 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
[0].offset
);
1094 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1u << reg
->idx
[0].offset
))
1096 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
[0].offset
);
1100 case WINED3DSPR_CONST
:
1101 if (!pshader
&& reg
->idx
[0].rel_addr
)
1103 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1104 UINT rel_offset
= ctx
->target_version
== ARB
? shader_data
->rel_offset
: 0;
1107 if (reg_maps
->shader_version
.major
< 2)
1109 sprintf(rel_reg
, "A0.x");
1113 shader_arb_get_src_param(ins
, reg
->idx
[0].rel_addr
, 0, rel_reg
);
1114 if (ctx
->target_version
== ARB
)
1116 if (!strcmp(rel_reg
, "**aL_emul**"))
1120 shader_arb_request_a0(ins
, rel_reg
);
1121 sprintf(rel_reg
, "A0.x");
1126 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
[0].offset
);
1127 else if (reg
->idx
[0].offset
>= rel_offset
)
1128 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
[0].offset
- rel_offset
);
1130 sprintf(register_name
, "C[%s - %u]", rel_reg
, rel_offset
- reg
->idx
[0].offset
);
1134 if (reg_maps
->usesrelconstF
)
1135 sprintf(register_name
, "C[%u]", reg
->idx
[0].offset
);
1137 sprintf(register_name
, "C%u", reg
->idx
[0].offset
);
1141 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1144 if (reg_maps
->shader_version
.major
== 1
1145 && reg_maps
->shader_version
.minor
<= 3)
1146 /* In ps <= 1.3, Tx is a temporary register as destination
1147 * to all instructions, and as source to most instructions.
1148 * For some instructions it is the texcoord input. Those
1149 * instructions know about the special use. */
1150 sprintf(register_name
, "T%u", reg
->idx
[0].offset
);
1152 /* In ps 1.4 and 2.x Tx is always a (read-only) varying. */
1153 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
[0].offset
);
1157 if (reg_maps
->shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1158 sprintf(register_name
, "A%u", reg
->idx
[0].offset
);
1160 sprintf(register_name
, "A%u_SHADOW", reg
->idx
[0].offset
);
1164 case WINED3DSPR_COLOROUT
:
1165 if (ctx
->ps_post_process
&& !reg
->idx
[0].offset
)
1167 strcpy(register_name
, "TMP_COLOR");
1171 if (ctx
->cur_ps_args
->super
.srgb_correction
)
1172 FIXME("sRGB correction on higher render targets.\n");
1173 if (reg_maps
->rt_mask
> 1)
1174 sprintf(register_name
, "result.color[%u]", reg
->idx
[0].offset
);
1176 strcpy(register_name
, "result.color");
1180 case WINED3DSPR_RASTOUT
:
1181 if (reg
->idx
[0].offset
== 1)
1182 sprintf(register_name
, "%s", ctx
->fog_output
);
1184 sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
[0].offset
]);
1187 case WINED3DSPR_DEPTHOUT
:
1188 strcpy(register_name
, "result.depth");
1191 case WINED3DSPR_ATTROUT
:
1192 /* case WINED3DSPR_OUTPUT: */
1194 sprintf(register_name
, "oD[%u]", reg
->idx
[0].offset
);
1196 strcpy(register_name
, ctx
->color_output
[reg
->idx
[0].offset
]);
1199 case WINED3DSPR_TEXCRDOUT
:
1201 sprintf(register_name
, "oT[%u]", reg
->idx
[0].offset
);
1202 else if (reg_maps
->shader_version
.major
< 3)
1203 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
[0].offset
]);
1205 strcpy(register_name
, ctx
->vs_output
[reg
->idx
[0].offset
]);
1208 case WINED3DSPR_LOOP
:
1209 if(ctx
->target_version
>= NV2
)
1211 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1212 if(pshader
) sprintf(register_name
, "A0.x");
1213 else sprintf(register_name
, "aL.y");
1217 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1218 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1219 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1220 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1223 sprintf(register_name
, "**aL_emul**");
1228 case WINED3DSPR_CONSTINT
:
1229 sprintf(register_name
, "I%u", reg
->idx
[0].offset
);
1232 case WINED3DSPR_MISCTYPE
:
1233 if (!reg
->idx
[0].offset
)
1234 sprintf(register_name
, "vpos");
1235 else if (reg
->idx
[0].offset
== 1)
1236 sprintf(register_name
, "fragment.facing.x");
1238 FIXME("Unknown MISCTYPE register index %u.\n", reg
->idx
[0].offset
);
1242 FIXME("Unhandled register type %#x[%u].\n", reg
->type
, reg
->idx
[0].offset
);
1243 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
[0].offset
);
1248 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1249 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1251 char register_name
[255];
1255 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1256 strcpy(str
, register_name
);
1258 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1259 strcat(str
, write_mask
);
1262 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1264 switch(channel_source
)
1266 case CHANNEL_SOURCE_ZERO
: return "0";
1267 case CHANNEL_SOURCE_ONE
: return "1";
1268 case CHANNEL_SOURCE_X
: return "x";
1269 case CHANNEL_SOURCE_Y
: return "y";
1270 case CHANNEL_SOURCE_Z
: return "z";
1271 case CHANNEL_SOURCE_W
: return "w";
1273 FIXME("Unhandled channel source %#x\n", channel_source
);
1278 struct color_fixup_masks
1284 static struct color_fixup_masks
calc_color_correction(struct color_fixup_desc fixup
, DWORD dst_mask
)
1286 struct color_fixup_masks masks
= {0, 0};
1288 if (is_complex_fixup(fixup
))
1290 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1291 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1295 if (fixup
.x_source
!= CHANNEL_SOURCE_X
)
1296 masks
.source
|= WINED3DSP_WRITEMASK_0
;
1297 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
)
1298 masks
.source
|= WINED3DSP_WRITEMASK_1
;
1299 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
)
1300 masks
.source
|= WINED3DSP_WRITEMASK_2
;
1301 if (fixup
.w_source
!= CHANNEL_SOURCE_W
)
1302 masks
.source
|= WINED3DSP_WRITEMASK_3
;
1303 masks
.source
&= dst_mask
;
1305 if (fixup
.x_sign_fixup
)
1306 masks
.sign
|= WINED3DSP_WRITEMASK_0
;
1307 if (fixup
.y_sign_fixup
)
1308 masks
.sign
|= WINED3DSP_WRITEMASK_1
;
1309 if (fixup
.z_sign_fixup
)
1310 masks
.sign
|= WINED3DSP_WRITEMASK_2
;
1311 if (fixup
.w_sign_fixup
)
1312 masks
.sign
|= WINED3DSP_WRITEMASK_3
;
1313 masks
.sign
&= dst_mask
;
1318 static void gen_color_correction(struct wined3d_string_buffer
*buffer
, const char *dst
,
1319 const char *src
, const char *one
, const char *two
,
1320 struct color_fixup_desc fixup
, struct color_fixup_masks masks
)
1322 const char *sign_fixup_src
= dst
;
1327 sign_fixup_src
= "TA";
1329 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", sign_fixup_src
, src
,
1330 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1331 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1333 else if (masks
.sign
)
1335 sign_fixup_src
= src
;
1341 char *ptr
= reg_mask
;
1343 if (masks
.sign
!= WINED3DSP_WRITEMASK_ALL
)
1346 if (masks
.sign
& WINED3DSP_WRITEMASK_0
)
1348 if (masks
.sign
& WINED3DSP_WRITEMASK_1
)
1350 if (masks
.sign
& WINED3DSP_WRITEMASK_2
)
1352 if (masks
.sign
& WINED3DSP_WRITEMASK_3
)
1357 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", dst
, reg_mask
, sign_fixup_src
, two
, one
);
1361 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1364 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1365 if (!ins
->dst_count
) return "";
1367 mod
= ins
->dst
[0].modifiers
;
1369 /* Silently ignore PARTIALPRECISION if it's not supported */
1370 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1372 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1374 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1375 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1380 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1383 case WINED3DSPDM_SATURATE
:
1386 case WINED3DSPDM_PARTIALPRECISION
:
1393 FIXME("Unknown modifiers 0x%08x\n", mod
);
1398 #define TEX_PROJ 0x1
1399 #define TEX_BIAS 0x2
1401 #define TEX_DERIV 0x10
1403 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, unsigned int sampler_idx
,
1404 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1406 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1407 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1408 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1409 enum wined3d_shader_resource_type resource_type
;
1410 struct color_fixup_masks masks
;
1411 const char *tex_dst
= dst_str
;
1412 BOOL np2_fixup
= FALSE
;
1413 const char *tex_type
;
1418 resource_type
= pixelshader_get_resource_type(ins
->ctx
->reg_maps
, sampler_idx
,
1419 priv
->cur_ps_args
->super
.tex_types
);
1423 resource_type
= ins
->ctx
->reg_maps
->resource_info
[sampler_idx
].type
;
1425 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1426 sampler_idx
+= WINED3D_MAX_FRAGMENT_SAMPLERS
;
1429 switch (resource_type
)
1431 case WINED3D_SHADER_RESOURCE_TEXTURE_1D
:
1435 case WINED3D_SHADER_RESOURCE_TEXTURE_2D
:
1436 if (pshader
&& priv
->cur_ps_args
->super
.np2_fixup
& (1u << sampler_idx
)
1437 && priv
->gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
1444 if (priv
->cur_np2fixup_info
->super
.active
& (1u << sampler_idx
))
1446 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1447 else np2_fixup
= TRUE
;
1452 case WINED3D_SHADER_RESOURCE_TEXTURE_3D
:
1456 case WINED3D_SHADER_RESOURCE_TEXTURE_CUBE
:
1461 ERR("Unexpected resource type %#x.\n", resource_type
);
1465 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1466 * so don't use shader_arb_get_modifier
1468 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1471 /* Fragment samplers always have identity mapping */
1472 if(sampler_idx
>= WINED3D_MAX_FRAGMENT_SAMPLERS
)
1474 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- WINED3D_MAX_FRAGMENT_SAMPLERS
];
1479 masks
= calc_color_correction(priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
],
1480 ins
->dst
[0].write_mask
);
1482 if (masks
.source
|| masks
.sign
)
1486 if (flags
& TEX_DERIV
)
1488 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1489 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1490 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1491 dsx
, dsy
, sampler_idx
, tex_type
);
1493 else if(flags
& TEX_LOD
)
1495 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1496 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1497 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
,
1498 sampler_idx
, tex_type
);
1500 else if (flags
& TEX_BIAS
)
1502 /* Shouldn't be possible, but let's check for it */
1503 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1504 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1505 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1507 else if (flags
& TEX_PROJ
)
1509 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1515 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1516 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1517 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1519 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, tex_dst
, sampler_idx
, tex_type
);
1522 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, tex_dst
, coord_reg
, sampler_idx
, tex_type
);
1527 gen_color_correction(buffer
, dst_str
, tex_dst
,
1528 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1529 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1530 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
], masks
);
1534 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1535 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1537 /* Generate a line that does the input modifier computation and return the input register to use */
1538 BOOL is_color
= FALSE
, insert_line
;
1541 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1542 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1543 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1544 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1546 /* Assume a new line will be added */
1549 /* Get register name */
1550 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1551 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1553 switch (src
->modifiers
)
1555 case WINED3DSPSM_NONE
:
1556 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1557 insert_line
= FALSE
;
1559 case WINED3DSPSM_NEG
:
1560 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1561 insert_line
= FALSE
;
1563 case WINED3DSPSM_BIAS
:
1564 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1566 case WINED3DSPSM_BIASNEG
:
1567 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1569 case WINED3DSPSM_SIGN
:
1570 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1572 case WINED3DSPSM_SIGNNEG
:
1573 shader_addline(buffer
, "MAD T%c, %s, -%s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1575 case WINED3DSPSM_COMP
:
1576 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1578 case WINED3DSPSM_X2
:
1579 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1581 case WINED3DSPSM_X2NEG
:
1582 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1584 case WINED3DSPSM_DZ
:
1585 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1586 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1588 case WINED3DSPSM_DW
:
1589 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1590 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1592 case WINED3DSPSM_ABS
:
1593 if(ctx
->target_version
>= NV2
) {
1594 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1595 insert_line
= FALSE
;
1597 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1600 case WINED3DSPSM_ABSNEG
:
1601 if(ctx
->target_version
>= NV2
) {
1602 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1604 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1605 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1607 insert_line
= FALSE
;
1610 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1611 insert_line
= FALSE
;
1614 /* Return modified or original register, with swizzle */
1616 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1619 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1621 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1622 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1623 unsigned int sampler_code
= dst
->reg
.idx
[0].offset
;
1625 char src_name
[2][50];
1627 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1629 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1631 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1632 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1635 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1636 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1637 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1638 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1639 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1641 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1642 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1645 static enum wined3d_shader_src_modifier
negate_modifiers(enum wined3d_shader_src_modifier mod
, char *extra_char
)
1650 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1651 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1652 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1653 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1654 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1655 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1656 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1657 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1658 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1659 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1660 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1661 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1662 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1663 case WINED3DSPSM_NOT
: FIXME("Unknown modifier %u\n", mod
);
1668 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1670 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1671 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1673 char src_name
[3][50];
1674 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1675 ins
->ctx
->reg_maps
->shader_version
.minor
);
1677 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1678 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1680 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
1681 && ins
->dst
->write_mask
!= WINED3DSP_WRITEMASK_3
)
1683 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1687 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1690 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1691 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1693 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1694 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1695 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1696 shader_addline(buffer
, "CMP%s %s, TA, %s, %s;\n", shader_arb_get_modifier(ins
),
1697 dst_name
, src_name
[1], src_name
[2]);
1701 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1703 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1704 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1706 char src_name
[3][50];
1708 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1710 /* Generate input register names (with modifiers) */
1711 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1712 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1713 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1715 shader_addline(buffer
, "CMP%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1716 dst_name
, src_name
[0], src_name
[2], src_name
[1]);
1719 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1720 * dst = dot2(src0, src1) + src2 */
1721 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1723 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1724 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1726 char src_name
[3][50];
1727 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1729 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1730 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1731 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1733 if(ctx
->target_version
>= NV3
)
1735 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1736 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1737 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1738 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1740 else if(ctx
->target_version
>= NV2
)
1742 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1743 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1744 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1745 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1747 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1749 * .xyxy and other swizzles that we could get with this are not valid in
1750 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1752 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1753 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1754 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1756 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1758 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1759 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1763 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1764 /* 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]
1765 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1767 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1768 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1769 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1770 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1774 /* Map the opcode 1-to-1 to the GL code */
1775 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1777 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1778 const char *instruction
;
1779 char arguments
[256], dst_str
[50];
1781 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1783 switch (ins
->handler_idx
)
1785 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1786 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1787 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1788 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1789 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1790 case WINED3DSIH_DST
: instruction
= "DST"; break;
1791 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1792 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1793 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1794 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1795 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1796 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1797 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1798 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1799 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1800 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1801 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1802 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1803 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1804 default: instruction
= "";
1805 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
1809 /* Note that shader_arb_add_dst_param() adds spaces. */
1810 arguments
[0] = '\0';
1811 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1812 for (i
= 0; i
< ins
->src_count
; ++i
)
1815 strcat(arguments
, ", ");
1816 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1817 strcat(arguments
, operand
);
1819 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1822 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
) {}
1824 static DWORD
shader_arb_select_component(DWORD swizzle
, DWORD component
)
1826 return ((swizzle
>> 2 * component
) & 0x3) * 0x55;
1829 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1831 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
1832 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
1833 BOOL pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
1834 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1835 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
1836 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
1837 const char *two
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_TWO
);
1839 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1840 char src0_param
[256];
1842 if (ins
->handler_idx
== WINED3DSIH_MOVA
)
1844 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1846 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1848 if(ctx
->target_version
>= NV2
) {
1849 shader_hw_map2gl(ins
);
1852 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1853 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1855 /* This implements the mova formula used in GLSL. The first two instructions
1856 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1860 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1862 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1863 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1865 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1866 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1868 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1869 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1870 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1871 if (shader_data
->rel_offset
)
1873 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1875 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1877 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1879 else if (reg_maps
->shader_version
.major
== 1
1880 && !shader_is_pshader_version(reg_maps
->shader_version
.type
)
1881 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1883 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
1884 src0_param
[0] = '\0';
1886 if (shader_data
->rel_offset
&& ctx
->target_version
== ARB
)
1888 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1889 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1890 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1891 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1895 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1896 * with more than one component. Thus replicate the first source argument over all
1897 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1898 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1899 tmp_src
.swizzle
= shader_arb_select_component(tmp_src
.swizzle
, 0);
1900 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1901 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1904 else if (ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& !ins
->dst
[0].reg
.idx
[0].offset
&& pshader
)
1906 if (ctx
->ps_post_process
&& shader
->u
.ps
.color0_mov
)
1908 shader_addline(buffer
, "#mov handled in srgb write or fog code\n");
1911 shader_hw_map2gl(ins
);
1915 shader_hw_map2gl(ins
);
1919 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1921 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1922 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
1925 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1926 * but >= 2.0 honors it (undocumented, but tested by the d3d9 testsuite)
1928 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1930 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1932 const char *kilsrc
= "TA";
1935 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1936 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1942 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1943 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1944 * masked out components to 0(won't kill)
1946 char x
= '0', y
= '0', z
= '0', w
= '0';
1947 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1948 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1949 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1950 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1951 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1953 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1957 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1958 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1960 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1961 * or pass in any temporary register(in shader phase 2)
1963 if (ins
->ctx
->reg_maps
->shader_version
.minor
<= 3)
1964 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
[0].offset
);
1966 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1967 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1968 shader_addline(buffer
, "KIL TA;\n");
1972 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1974 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1975 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1976 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1977 ins
->ctx
->reg_maps
->shader_version
.minor
);
1978 struct wined3d_shader_src_param src
;
1982 unsigned int reg_sampler_code
;
1984 BOOL swizzle_coord
= FALSE
;
1986 /* All versions have a destination register */
1987 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1989 /* 1.0-1.4: Use destination register number as texture code.
1990 2.0+: Use provided sampler number as texture code. */
1991 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1992 reg_sampler_code
= dst
->reg
.idx
[0].offset
;
1994 reg_sampler_code
= ins
->src
[1].reg
.idx
[0].offset
;
1996 /* 1.0-1.3: Use the texcoord varying.
1997 1.4+: Use provided coordinate source register. */
1998 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1999 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
2001 /* TEX is the only instruction that can handle DW and DZ natively */
2003 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
2004 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
2005 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
2009 * 1.1, 1.2, 1.3: Use WINED3D_TSS_TEXTURETRANSFORMFLAGS
2010 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
2011 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
2013 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2016 if (reg_sampler_code
< WINED3D_MAX_FFP_TEXTURES
)
2017 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg_sampler_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2018 if (flags
& WINED3D_PSARGS_PROJECTED
)
2020 myflags
|= TEX_PROJ
;
2021 if ((flags
& ~WINED3D_PSARGS_PROJECTED
) == WINED3D_TTFF_COUNT3
)
2022 swizzle_coord
= TRUE
;
2025 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
2027 enum wined3d_shader_src_modifier src_mod
= ins
->src
[0].modifiers
;
2028 if (src_mod
== WINED3DSPSM_DZ
)
2030 swizzle_coord
= TRUE
;
2031 myflags
|= TEX_PROJ
;
2032 } else if(src_mod
== WINED3DSPSM_DW
) {
2033 myflags
|= TEX_PROJ
;
2036 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
2037 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
2042 /* TXP cannot handle DZ natively, so move the z coordinate to .w.
2043 * reg_coord is a read-only varying register, so we need a temp reg */
2044 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
2045 strcpy(reg_coord
, "TA");
2048 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
2051 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
2053 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2054 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2055 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
2056 ins
->ctx
->reg_maps
->shader_version
.minor
);
2059 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
2061 unsigned int reg
= dst
->reg
.idx
[0].offset
;
2063 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2064 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
2068 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
2069 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2070 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
2074 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
2076 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2079 unsigned int reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2083 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
2084 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2085 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2086 /* Move .x first in case src_str is "TA" */
2087 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
2088 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
2089 if (reg1
< WINED3D_MAX_FFP_TEXTURES
)
2091 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2092 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg1
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2094 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2097 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
2099 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2101 unsigned int reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2105 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2106 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2107 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2108 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2109 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2110 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2113 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2115 unsigned int reg1
= ins
->dst
[0].reg
.idx
[0].offset
;
2119 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2120 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2121 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2122 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2125 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2127 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2128 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2129 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2130 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2131 unsigned int reg_dest_code
;
2133 /* All versions have a destination register. The Tx where the texture coordinates come
2134 * from is the varying incarnation of the texture register
2136 reg_dest_code
= dst
->reg
.idx
[0].offset
;
2137 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2138 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2139 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2141 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2142 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2144 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2145 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2147 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2148 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2151 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2152 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2153 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2154 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2156 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2157 * so we can't let the GL handle this.
2159 if ((priv
->cur_ps_args
->super
.tex_transform
>> reg_dest_code
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
)
2160 & WINED3D_PSARGS_PROJECTED
)
2162 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2163 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2164 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2166 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2169 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2171 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2173 /* No src swizzles are allowed, so this is ok */
2174 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2175 src_reg
, reg_dest_code
, reg_dest_code
);
2176 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2180 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2182 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2183 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2184 char src0_name
[50], dst_name
[50];
2186 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2188 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2189 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2190 * T<reg+1> register. Use this register to store the calculated vector
2192 tmp_reg
.idx
[0].offset
= reg
+ 1;
2193 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2194 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2197 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2199 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2200 unsigned int flags
= 0;
2201 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2202 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2208 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2209 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2211 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2212 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2213 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2214 if (reg
< WINED3D_MAX_FFP_TEXTURES
)
2215 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2216 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2219 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2221 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2222 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2223 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2224 char src0_name
[50], dst_name
[50];
2225 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2228 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2229 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2230 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2232 tmp_reg
.idx
[0].offset
= reg
+ 2 - tex_mx
->current_row
;
2233 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2235 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2236 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2237 dst_name
, 'x' + tex_mx
->current_row
, reg
, src0_name
);
2238 tex_mx
->texcoord_w
[tex_mx
->current_row
++] = reg
;
2241 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2243 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2244 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2245 unsigned int flags
= 0;
2246 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2247 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2249 char src0_name
[50], dst_name
[50];
2252 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2253 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2254 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2256 /* Sample the texture using the calculated coordinates */
2257 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2258 if (reg
< WINED3D_MAX_FFP_TEXTURES
)
2259 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2260 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2261 tex_mx
->current_row
= 0;
2264 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2266 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2267 struct wined3d_shader_tex_mx
*tex_mx
= ins
->ctx
->tex_mx
;
2268 unsigned int flags
= 0;
2269 unsigned int reg
= ins
->dst
[0].reg
.idx
[0].offset
;
2270 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2276 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2277 * components for temporary data storage
2279 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2280 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2281 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2283 /* Construct the eye-ray vector from w coordinates */
2284 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[0]);
2285 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", tex_mx
->texcoord_w
[1]);
2286 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2288 /* Calculate reflection vector
2290 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2291 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2292 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2293 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2294 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2295 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2296 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2298 /* Sample the texture using the calculated coordinates */
2299 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2300 if (reg
< WINED3D_MAX_FFP_TEXTURES
)
2301 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
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
;
2310 unsigned int flags
= 0;
2311 unsigned int 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 if (reg
< WINED3D_MAX_FFP_TEXTURES
)
2343 flags
= priv
->cur_ps_args
->super
.tex_transform
>> reg
* WINED3D_PSARGS_TEXTRANSFORM_SHIFT
;
2344 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3D_PSARGS_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2345 tex_mx
->current_row
= 0;
2348 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2350 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2351 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2353 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2354 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2356 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2357 * which is essentially an input, is the destination register because it is the first
2358 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2359 * here(writemasks/swizzles are not valid on texdepth)
2361 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2363 /* According to the msdn, the source register(must be r5) is unusable after
2364 * the texdepth instruction, so we're free to modify it
2366 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2368 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2369 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2370 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2372 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2373 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2374 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2375 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2378 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2379 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2380 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2381 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2383 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2384 unsigned int sampler_idx
= ins
->dst
[0].reg
.idx
[0].offset
;
2388 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2389 shader_addline(buffer
, "MOV TB, 0.0;\n");
2390 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2392 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2393 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2396 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2397 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2398 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2400 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2403 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2405 /* Handle output register */
2406 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2407 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2408 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
[0].offset
, src0
);
2411 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2412 * Perform the 3rd row of a 3x3 matrix multiply */
2413 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2415 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2416 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2417 char dst_str
[50], dst_name
[50];
2421 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2422 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2423 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2424 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2425 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2428 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2429 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2430 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2431 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2433 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2435 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2436 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2437 char src0
[50], dst_name
[50];
2439 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2440 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2442 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2443 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2444 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
[0].offset
, src0
);
2446 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2447 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2448 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2450 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2451 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2452 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2453 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2456 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2457 Vertex/Pixel shaders to ARB_vertex_program codes */
2458 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2461 int nComponents
= 0;
2462 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2463 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2464 struct wined3d_shader_instruction tmp_ins
;
2466 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2468 /* Set constants for the temporary argument */
2469 tmp_ins
.ctx
= ins
->ctx
;
2470 tmp_ins
.dst_count
= 1;
2471 tmp_ins
.dst
= &tmp_dst
;
2472 tmp_ins
.src_count
= 2;
2473 tmp_ins
.src
= tmp_src
;
2475 switch(ins
->handler_idx
)
2477 case WINED3DSIH_M4x4
:
2479 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2481 case WINED3DSIH_M4x3
:
2483 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2485 case WINED3DSIH_M3x4
:
2487 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2489 case WINED3DSIH_M3x3
:
2491 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2493 case WINED3DSIH_M3x2
:
2495 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2498 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2502 tmp_dst
= ins
->dst
[0];
2503 tmp_src
[0] = ins
->src
[0];
2504 tmp_src
[1] = ins
->src
[1];
2505 for (i
= 0; i
< nComponents
; ++i
)
2507 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2508 shader_hw_map2gl(&tmp_ins
);
2509 ++tmp_src
[1].reg
.idx
[0].offset
;
2513 static enum wined3d_shader_src_modifier
abs_modifier(enum wined3d_shader_src_modifier mod
, BOOL
*need_abs
)
2519 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2520 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2521 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2522 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2523 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2524 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2525 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2526 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2527 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2528 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2529 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2530 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2531 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2532 case WINED3DSPSM_NOT
: FIXME("Unknown modifier %u\n", mod
);
2537 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2539 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2540 const char *instruction
;
2541 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2542 BOOL need_abs
= FALSE
;
2547 switch(ins
->handler_idx
)
2549 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2550 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2551 case WINED3DSIH_EXPP
:
2552 if (ins
->ctx
->reg_maps
->shader_version
.major
< 2)
2554 instruction
= "EXP";
2558 case WINED3DSIH_EXP
:
2559 instruction
= "EX2";
2561 case WINED3DSIH_LOG
:
2562 case WINED3DSIH_LOGP
:
2563 /* The precision requirements suggest that LOGP matches ARBvp's LOG
2564 * instruction, but notice that the output of those instructions is
2566 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2567 instruction
= "LG2";
2569 default: instruction
= "";
2570 FIXME("Unhandled opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
2574 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2576 src0_copy
.swizzle
= shader_arb_select_component(src0_copy
.swizzle
, 3);
2578 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2579 shader_arb_get_src_param(ins
, &src0_copy
, 0, src
);
2583 shader_addline(buffer
, "ABS TA.w, %s;\n", src
);
2584 shader_addline(buffer
, "%s%s %s, TA.w;\n", instruction
, shader_arb_get_modifier(ins
), dst
);
2588 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2593 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2595 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2598 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2599 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2600 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2602 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2603 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2605 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2606 * otherwise NRM or RSQ would return NaN */
2607 if(pshader
&& priv
->target_version
>= NV3
)
2609 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2611 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2613 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2614 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2615 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2617 else if(priv
->target_version
>= NV2
)
2619 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2620 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2621 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2626 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2628 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2629 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2630 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2632 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2633 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2635 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2636 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2637 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2642 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2644 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2646 char src_name
[3][50];
2648 /* ARB_fragment_program has a convenient LRP instruction */
2649 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2650 shader_hw_map2gl(ins
);
2654 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2655 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2656 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2657 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2659 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2660 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2661 dst_name
, src_name
[0], src_name
[2]);
2664 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2666 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2667 * must contain fixed constants. So we need a separate function to filter those constants and
2670 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2671 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2672 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2674 char src_name0
[50], src_name1
[50], src_name2
[50];
2677 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2678 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2679 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2680 /* No modifiers are supported on SCS */
2681 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2683 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2685 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2686 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2688 } else if(priv
->target_version
>= NV2
) {
2689 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2691 /* Sincos writemask must be .x, .y or .xy */
2692 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2693 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2694 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2695 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2697 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2698 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2700 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2701 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2703 * The constants we get are:
2705 * +1 +1, -1 -1 +1 +1 -1 -1
2706 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2707 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2709 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2713 * (x/2)^4 = x^4 / 16
2714 * (x/2)^5 = x^5 / 32
2717 * To get the final result:
2718 * sin(x) = 2 * sin(x/2) * cos(x/2)
2719 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2720 * (from sin(x+y) and cos(x+y) rules)
2722 * As per MSDN, dst.z is undefined after the operation, and so is
2723 * dst.x and dst.y if they're masked out by the writemask. Ie
2724 * sincos dst.y, src1, c0, c1
2725 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2726 * vsa.exe also stops with an error if the dest register is the same register as the source
2727 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2728 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2730 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2731 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2732 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2734 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2735 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2736 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2737 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2738 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2739 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2743 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2744 * properly merge that with MULs in the code above?
2745 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2746 * we can merge the sine and cosine MAD rows to calculate them together.
2748 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2749 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2750 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2751 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2754 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2755 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2756 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2758 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2760 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2761 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2763 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2765 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2766 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2771 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2773 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2776 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2778 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2779 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2781 /* SGN is only valid in vertex shaders */
2782 if(ctx
->target_version
>= NV2
) {
2783 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2787 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2788 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2790 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2791 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2793 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2794 * Then use TA, and calculate the final result
2796 * Not reading from TA? Store the first result in TA to avoid overwriting the
2797 * destination if src reg = dst reg
2799 if(strstr(src_name
, "TA"))
2801 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2802 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2803 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2807 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2808 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2809 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2814 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2816 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2822 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2823 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2824 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2826 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2827 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2830 static void shader_hw_pow(const struct wined3d_shader_instruction
*ins
)
2832 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2833 char src0
[50], src1
[50], dst
[50];
2834 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2835 BOOL need_abs
= FALSE
;
2836 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2837 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2839 /* POW operates on the absolute value of the input */
2840 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2842 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2843 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2844 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2847 shader_addline(buffer
, "ABS TA.x, %s;\n", src0
);
2849 shader_addline(buffer
, "MOV TA.x, %s;\n", src0
);
2851 if (priv
->target_version
>= NV2
)
2853 shader_addline(buffer
, "MOVC TA.y, %s;\n", src1
);
2854 shader_addline(buffer
, "POW%s %s, TA.x, TA.y;\n", shader_arb_get_modifier(ins
), dst
);
2855 shader_addline(buffer
, "MOV %s (EQ.y), %s;\n", dst
, one
);
2859 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2860 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2862 shader_addline(buffer
, "ABS TA.y, %s;\n", src1
);
2863 shader_addline(buffer
, "SGE TA.y, -TA.y, %s;\n", zero
);
2864 /* Possibly add flt_eps to avoid getting float special values */
2865 shader_addline(buffer
, "MAD TA.z, TA.y, %s, %s;\n", flt_eps
, src1
);
2866 shader_addline(buffer
, "POW%s TA.x, TA.x, TA.z;\n", shader_arb_get_modifier(ins
));
2867 shader_addline(buffer
, "MAD TA.x, -TA.x, TA.y, TA.x;\n");
2868 shader_addline(buffer
, "MAD %s, TA.y, %s, TA.x;\n", dst
, one
);
2872 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2874 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2876 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2879 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2883 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2884 struct list
*e
= list_head(&priv
->control_frames
);
2885 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2887 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2888 /* The constant loader makes sure to load -1 into iX.w */
2889 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2890 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2891 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2895 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2899 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2901 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2903 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2905 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2907 /* The constant loader makes sure to load -1 into iX.w */
2910 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2911 struct list
*e
= list_head(&priv
->control_frames
);
2912 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2914 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2916 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2917 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2918 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2922 shader_addline(buffer
, "REP %s;\n", src_name
);
2926 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2928 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2929 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2933 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2934 struct list
*e
= list_head(&priv
->control_frames
);
2935 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2937 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2938 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2939 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2941 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2945 shader_addline(buffer
, "ENDLOOP;\n");
2949 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2951 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2952 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2956 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2957 struct list
*e
= list_head(&priv
->control_frames
);
2958 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2960 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2961 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2962 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2964 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2968 shader_addline(buffer
, "ENDREP;\n");
2972 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2974 struct control_frame
*control_frame
;
2976 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2978 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2980 ERR("Could not find loop for break\n");
2984 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2986 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
2987 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2988 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2992 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2996 shader_addline(buffer
, "BRK;\n");
3000 static const char *get_compare(enum wined3d_shader_rel_op op
)
3004 case WINED3D_SHADER_REL_OP_GT
: return "GT";
3005 case WINED3D_SHADER_REL_OP_EQ
: return "EQ";
3006 case WINED3D_SHADER_REL_OP_GE
: return "GE";
3007 case WINED3D_SHADER_REL_OP_LT
: return "LT";
3008 case WINED3D_SHADER_REL_OP_NE
: return "NE";
3009 case WINED3D_SHADER_REL_OP_LE
: return "LE";
3011 FIXME("Unrecognized operator %#x.\n", op
);
3016 static enum wined3d_shader_rel_op
invert_compare(enum wined3d_shader_rel_op op
)
3020 case WINED3D_SHADER_REL_OP_GT
: return WINED3D_SHADER_REL_OP_LE
;
3021 case WINED3D_SHADER_REL_OP_EQ
: return WINED3D_SHADER_REL_OP_NE
;
3022 case WINED3D_SHADER_REL_OP_GE
: return WINED3D_SHADER_REL_OP_LT
;
3023 case WINED3D_SHADER_REL_OP_LT
: return WINED3D_SHADER_REL_OP_GE
;
3024 case WINED3D_SHADER_REL_OP_NE
: return WINED3D_SHADER_REL_OP_EQ
;
3025 case WINED3D_SHADER_REL_OP_LE
: return WINED3D_SHADER_REL_OP_GT
;
3027 FIXME("Unrecognized operator %#x.\n", op
);
3032 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
3034 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3035 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3036 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
3039 const char *comp
= get_compare(ins
->flags
);
3041 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3042 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3046 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
3047 * away the subtraction result
3049 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3050 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
3054 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3055 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
3059 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
3061 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3062 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3063 struct list
*e
= list_head(&priv
->control_frames
);
3064 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3068 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3070 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
3071 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
3075 /* Invert the flag. We jump to the else label if the condition is NOT true */
3076 comp
= get_compare(invert_compare(ins
->flags
));
3077 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3078 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
3082 comp
= get_compare(ins
->flags
);
3083 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3084 shader_addline(buffer
, "IF %s.x;\n", comp
);
3088 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3090 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3091 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3092 struct list
*e
= list_head(&priv
->control_frames
);
3093 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3094 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3098 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3099 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3100 control_frame
->had_else
= TRUE
;
3104 shader_addline(buffer
, "ELSE;\n");
3108 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3110 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3111 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3112 struct list
*e
= list_head(&priv
->control_frames
);
3113 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3114 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3118 if(control_frame
->had_else
)
3120 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3124 shader_addline(buffer
, "#No else branch. else is endif\n");
3125 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3130 shader_addline(buffer
, "ENDIF;\n");
3134 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3136 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3138 char reg_src
[3][40];
3139 WORD flags
= TEX_DERIV
;
3141 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3142 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3143 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3144 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3146 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3147 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3149 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3152 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3154 DWORD sampler_idx
= ins
->src
[1].reg
.idx
[0].offset
;
3157 WORD flags
= TEX_LOD
;
3159 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3160 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3162 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3163 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3165 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3168 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3170 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3171 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3173 priv
->in_main_func
= FALSE
;
3174 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3175 * subroutine, don't generate a label that will make GL complain
3177 if(priv
->target_version
== ARB
) return;
3179 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
[0].offset
);
3182 static void vshader_add_footer(struct shader_arb_ctx_priv
*priv_ctx
,
3183 const struct arb_vshader_private
*shader_data
, const struct arb_vs_compile_args
*args
,
3184 const struct wined3d_shader_reg_maps
*reg_maps
, const struct wined3d_gl_info
*gl_info
,
3185 struct wined3d_string_buffer
*buffer
)
3189 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3190 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3191 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3192 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3194 if (args
->super
.fog_src
== VS_FOG_Z
)
3196 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3202 /* posFixup.x is always 1.0, so we can safely use it */
3203 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3207 /* Clamp fogcoord */
3208 const char *zero
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ZERO
);
3209 const char *one
= arb_get_helper_value(reg_maps
->shader_version
.type
, ARB_ONE
);
3211 shader_addline(buffer
, "MIN TMP_FOGCOORD.x, TMP_FOGCOORD.x, %s;\n", one
);
3212 shader_addline(buffer
, "MAX result.fogcoord.x, TMP_FOGCOORD.x, %s;\n", zero
);
3216 /* Clipplanes are always stored without y inversion */
3217 if (use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3219 if (args
->super
.clip_enabled
)
3221 for (i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3223 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3227 else if (args
->clip
.boolclip
.clip_texcoord
)
3229 static const char component
[4] = {'x', 'y', 'z', 'w'};
3230 unsigned int cur_clip
= 0;
3231 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3233 for (i
= 0; i
< gl_info
->limits
.user_clip_distances
; ++i
)
3235 if (args
->clip
.boolclip
.clipplane_mask
& (1u << i
))
3237 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3238 component
[cur_clip
++], i
);
3244 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3247 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3250 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3253 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3256 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3257 args
->clip
.boolclip
.clip_texcoord
- 1);
3260 /* Write the final position.
3262 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3263 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3264 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3265 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3267 if (!gl_info
->supported
[ARB_CLIP_CONTROL
])
3269 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3270 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3271 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3273 /* Z coord [0;1]->[-1;1] mapping, see comment in
3274 * get_projection_matrix() in utils.c. */
3275 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
3277 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3278 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3282 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3283 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3287 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3289 priv_ctx
->footer_written
= TRUE
;
3292 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3294 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3295 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3296 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
3297 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3299 if(priv
->target_version
== ARB
) return;
3303 if (priv
->in_main_func
) vshader_add_footer(priv
, shader
->backend_data
,
3304 priv
->cur_vs_args
, ins
->ctx
->reg_maps
, priv
->gl_info
, buffer
);
3307 shader_addline(buffer
, "RET;\n");
3310 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3312 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
3313 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
[0].offset
);
3316 static BOOL
shader_arb_compile(const struct wined3d_gl_info
*gl_info
, GLenum target
, const char *src
)
3318 const char *ptr
, *end
, *line
;
3321 if (TRACE_ON(d3d_shader
))
3324 end
= ptr
+ strlen(ptr
);
3325 while ((line
= wined3d_get_line(&ptr
, end
)))
3327 TRACE_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3331 GL_EXTCALL(glProgramStringARB(target
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(src
), src
));
3332 checkGLcall("glProgramStringARB()");
3334 if (FIXME_ON(d3d_shader
))
3336 gl_info
->gl_ops
.gl
.p_glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3339 FIXME_(d3d_shader
)("Program error at position %d: %s\n\n", pos
,
3340 debugstr_a((const char *)gl_info
->gl_ops
.gl
.p_glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3342 end
= ptr
+ strlen(ptr
);
3343 while ((line
= wined3d_get_line(&ptr
, end
)))
3345 FIXME_(d3d_shader
)(" %.*s", (int)(ptr
- line
), line
);
3347 FIXME_(d3d_shader
)("\n");
3353 if (WARN_ON(d3d_perf
))
3355 GL_EXTCALL(glGetProgramivARB(target
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3356 checkGLcall("glGetProgramivARB()");
3358 WARN_(d3d_perf
)("Program exceeds native resource limits.\n");
3364 static void arbfp_add_sRGB_correction(struct wined3d_string_buffer
*buffer
, const char *fragcolor
,
3365 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3367 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3371 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3372 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.x;\n", tmp1
, fragcolor
);
3373 /* Calculate the > 0.0031308 case */
3374 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3375 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3376 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts0.x;\n", fragcolor
, fragcolor
);
3377 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts0.y;\n", fragcolor
, fragcolor
);
3378 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts0.z;\n", fragcolor
, fragcolor
);
3379 /* Calculate the < case */
3380 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts0.w, %s;\n", fragcolor
, fragcolor
);
3384 /* Calculate the > 0.0031308 case */
3385 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts0.x;\n", tmp1
, fragcolor
);
3386 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts0.x;\n", tmp1
, fragcolor
);
3387 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts0.x;\n", tmp1
, fragcolor
);
3388 shader_addline(buffer
, "MUL %s, %s, srgb_consts0.y;\n", tmp1
, tmp1
);
3389 shader_addline(buffer
, "SUB %s, %s, srgb_consts0.z;\n", tmp1
, tmp1
);
3390 /* Calculate the < case */
3391 shader_addline(buffer
, "MUL %s, srgb_consts0.w, %s;\n", tmp2
, fragcolor
);
3392 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3393 shader_addline(buffer
, "SLT %s, srgb_consts1.x, %s;\n", tmp3
, fragcolor
);
3394 shader_addline(buffer
, "SGE %s, srgb_consts1.x, %s;\n", tmp4
, fragcolor
);
3395 /* Store the components > 0.0031308 in the destination */
3396 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3397 /* Add the components that are < 0.0031308 */
3398 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3399 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3400 * result.color writes(.rgb first, then .a), or handle overwriting already written
3401 * components. The assembler uses a temporary register in this case, which is usually
3402 * not allocated from one of our registers that were used earlier.
3405 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3408 static const unsigned int *find_loop_control_values(const struct wined3d_shader
*shader
, DWORD idx
)
3410 const struct wined3d_shader_lconst
*constant
;
3412 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
3414 if (constant
->idx
== idx
)
3416 return constant
->value
;
3422 static void init_ps_input(const struct wined3d_shader
*shader
,
3423 const struct arb_ps_compile_args
*args
, struct shader_arb_ctx_priv
*priv
)
3425 static const char * const texcoords
[8] =
3427 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3428 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3431 const struct wined3d_shader_signature_element
*input
;
3432 const char *semantic_name
;
3433 unsigned int semantic_idx
;
3435 if (args
->super
.vp_mode
== WINED3D_VP_MODE_SHADER
)
3437 /* That one is easy. The vertex shaders provide v0-v7 in
3438 * fragment.texcoord and v8 and v9 in fragment.color. */
3439 for (i
= 0; i
< 8; ++i
)
3441 priv
->ps_input
[i
] = texcoords
[i
];
3443 priv
->ps_input
[8] = "fragment.color.primary";
3444 priv
->ps_input
[9] = "fragment.color.secondary";
3448 /* The fragment shader has to collect the varyings on its own. In any case
3449 * properly load color0 and color1. In the case of pre-transformed
3450 * vertices also load texture coordinates. Set other attributes to 0.0.
3452 * For fixed-function this behavior is correct, according to the tests.
3453 * For pre-transformed we'd either need a replacement shader that can load
3454 * other attributes like BINORMAL, or load the texture coordinate
3455 * attribute pointers to match the fragment shader signature. */
3456 for (i
= 0; i
< shader
->input_signature
.element_count
; ++i
)
3458 input
= &shader
->input_signature
.elements
[i
];
3459 if (!(semantic_name
= input
->semantic_name
))
3461 semantic_idx
= input
->semantic_idx
;
3463 if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3466 priv
->ps_input
[input
->register_idx
] = "fragment.color.primary";
3467 else if (semantic_idx
== 1)
3468 priv
->ps_input
[input
->register_idx
] = "fragment.color.secondary";
3470 priv
->ps_input
[input
->register_idx
] = "0.0";
3472 else if (args
->super
.vp_mode
== WINED3D_VP_MODE_FF
)
3474 priv
->ps_input
[input
->register_idx
] = "0.0";
3476 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3478 if (semantic_idx
< 8)
3479 priv
->ps_input
[input
->register_idx
] = texcoords
[semantic_idx
];
3481 priv
->ps_input
[input
->register_idx
] = "0.0";
3483 else if (shader_match_semantic(semantic_name
, WINED3D_DECL_USAGE_FOG
))
3486 priv
->ps_input
[input
->register_idx
] = "fragment.fogcoord";
3488 priv
->ps_input
[input
->register_idx
] = "0.0";
3492 priv
->ps_input
[input
->register_idx
] = "0.0";
3495 TRACE("v%u, semantic %s%u is %s\n", input
->register_idx
,
3496 semantic_name
, semantic_idx
, priv
->ps_input
[input
->register_idx
]);
3500 static void arbfp_add_linear_fog(struct wined3d_string_buffer
*buffer
,
3501 const char *fragcolor
, const char *tmp
)
3503 shader_addline(buffer
, "SUB %s.x, state.fog.params.z, fragment.fogcoord.x;\n", tmp
);
3504 shader_addline(buffer
, "MUL_SAT %s.x, %s.x, state.fog.params.w;\n", tmp
, tmp
);
3505 shader_addline(buffer
, "LRP %s.rgb, %s.x, %s, state.fog.color;\n", fragcolor
, tmp
, fragcolor
);
3508 /* Context activation is done by the caller. */
3509 static GLuint
shader_arb_generate_pshader(const struct wined3d_shader
*shader
,
3510 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
3511 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3513 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
3516 unsigned int next_local
= 0;
3517 struct shader_arb_ctx_priv priv_ctx
;
3518 BOOL dcl_td
= FALSE
;
3519 BOOL want_nv_prog
= FALSE
;
3520 struct arb_pshader_private
*shader_priv
= shader
->backend_data
;
3521 BOOL custom_linear_fog
= FALSE
;
3526 unsigned int i
, found
= 0;
3528 map
= reg_maps
->temporary
;
3531 i
= wined3d_bit_scan(&map
);
3532 if ((shader
->u
.ps
.color0_mov
&& i
== shader
->u
.ps
.color0_reg
)
3533 || (reg_maps
->shader_version
.major
< 2 && !i
))
3536 sprintf(srgbtmp
[found
], "R%u", i
);
3538 if (found
== 4) break;
3543 sprintf(srgbtmp
[0], "TA");
3544 sprintf(srgbtmp
[1], "TB");
3545 sprintf(srgbtmp
[2], "TC");
3546 sprintf(srgbtmp
[3], "TD");
3550 sprintf(srgbtmp
[1], "TA");
3551 sprintf(srgbtmp
[2], "TB");
3552 sprintf(srgbtmp
[3], "TC");
3555 sprintf(srgbtmp
[2], "TA");
3556 sprintf(srgbtmp
[3], "TB");
3559 sprintf(srgbtmp
[3], "TA");
3565 /* Create the hw ARB shader */
3566 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3567 priv_ctx
.gl_info
= gl_info
;
3568 priv_ctx
.cur_ps_args
= args
;
3569 priv_ctx
.compiled_fprog
= compiled
;
3570 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3571 init_ps_input(shader
, args
, &priv_ctx
);
3572 list_init(&priv_ctx
.control_frames
);
3573 priv_ctx
.ps_post_process
= args
->super
.srgb_correction
;
3575 /* Avoid enabling NV_fragment_program* if we do not need it.
3577 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3578 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3579 * is faster than what we gain from using higher native instructions. There are some things though
3580 * that cannot be emulated. In that case enable the extensions.
3581 * If the extension is enabled, instruction handlers that support both ways will use it.
3583 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3584 * So enable the best we can get.
3586 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3587 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3589 want_nv_prog
= TRUE
;
3592 shader_addline(buffer
, "!!ARBfp1.0\n");
3593 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3595 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3596 priv_ctx
.target_version
= NV3
;
3598 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3600 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3601 priv_ctx
.target_version
= NV2
;
3605 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3608 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3611 priv_ctx
.target_version
= ARB
;
3614 if (reg_maps
->rt_mask
> 1)
3616 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3619 if (reg_maps
->shader_version
.major
< 3)
3621 switch (args
->super
.fog
)
3623 case WINED3D_FFP_PS_FOG_OFF
:
3625 case WINED3D_FFP_PS_FOG_LINEAR
:
3626 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
3628 custom_linear_fog
= TRUE
;
3629 priv_ctx
.ps_post_process
= TRUE
;
3632 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3634 case WINED3D_FFP_PS_FOG_EXP
:
3635 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3637 case WINED3D_FFP_PS_FOG_EXP2
:
3638 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3643 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3644 * unused temps away(but occupies them for the whole shader if they're used once). Always
3645 * declaring them avoids tricky bookkeeping work
3647 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3648 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3649 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3650 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3651 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3652 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3653 wined3d_ftoa(eps
, ftoa_tmp
);
3654 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %s, 0.0 };\n", ftoa_tmp
);
3656 if (reg_maps
->shader_version
.major
< 2)
3658 strcpy(fragcolor
, "R0");
3662 if (priv_ctx
.ps_post_process
)
3664 if (shader
->u
.ps
.color0_mov
)
3666 sprintf(fragcolor
, "R%u", shader
->u
.ps
.color0_reg
);
3670 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3671 strcpy(fragcolor
, "TMP_COLOR");
3674 strcpy(fragcolor
, "result.color");
3678 if (args
->super
.srgb_correction
)
3680 shader_addline(buffer
, "PARAM srgb_consts0 = ");
3681 shader_arb_append_imm_vec4(buffer
, &wined3d_srgb_const
[0].x
);
3682 shader_addline(buffer
, ";\n");
3683 shader_addline(buffer
, "PARAM srgb_consts1 = ");
3684 shader_arb_append_imm_vec4(buffer
, &wined3d_srgb_const
[1].x
);
3685 shader_addline(buffer
, ";\n");
3688 /* Base Declarations */
3689 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
, NULL
, &priv_ctx
);
3691 map
= reg_maps
->bumpmat
;
3694 unsigned char bump_const
;
3696 i
= wined3d_bit_scan(&map
);
3697 bump_const
= compiled
->numbumpenvmatconsts
;
3698 compiled
->bumpenvmatconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3699 compiled
->bumpenvmatconst
[bump_const
].texunit
= i
;
3700 compiled
->luminanceconst
[bump_const
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3701 compiled
->luminanceconst
[bump_const
].texunit
= i
;
3703 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3704 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3705 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3706 * textures due to conditional NP2 restrictions)
3708 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3709 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3710 * their location is shader dependent anyway and they cannot be loaded globally.
3712 compiled
->bumpenvmatconst
[bump_const
].const_num
= next_local
++;
3713 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3714 i
, compiled
->bumpenvmatconst
[bump_const
].const_num
);
3715 compiled
->numbumpenvmatconsts
= bump_const
+ 1;
3717 if (!(reg_maps
->luminanceparams
& (1u << i
)))
3720 compiled
->luminanceconst
[bump_const
].const_num
= next_local
++;
3721 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3722 i
, compiled
->luminanceconst
[bump_const
].const_num
);
3725 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
3727 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3728 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
3730 const unsigned int *control_values
= find_loop_control_values(shader
, i
);
3734 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3735 control_values
[0], control_values
[1], control_values
[2]);
3739 compiled
->int_consts
[i
] = next_local
;
3740 compiled
->num_int_consts
++;
3741 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3746 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3748 compiled
->ycorrection
= next_local
;
3749 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3753 shader_addline(buffer
, "TEMP vpos;\n");
3754 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3755 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3756 * ycorrection.z: 1.0
3757 * ycorrection.w: 0.0
3759 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3760 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3765 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3768 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3769 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3770 * at most 8 (WINED3D_MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3771 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3772 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3773 * shader compilation errors and the subsequent errors when drawing with this shader. */
3774 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3775 unsigned char cur_fixup_sampler
= 0;
3777 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3778 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3779 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3781 fixup
->offset
= next_local
;
3782 fixup
->super
.active
= 0;
3784 for (i
= 0; i
< WINED3D_MAX_FRAGMENT_SAMPLERS
; ++i
)
3786 if (!(map
& (1u << i
)))
3789 if (fixup
->offset
+ (cur_fixup_sampler
>> 1) < max_lconsts
)
3791 fixup
->super
.active
|= (1u << i
);
3792 fixup
->super
.idx
[i
] = cur_fixup_sampler
++;
3796 FIXME("No free constant found to load NP2 fixup data into shader. "
3797 "Sampling from this texture will probably look wrong.\n");
3802 fixup
->super
.num_consts
= (cur_fixup_sampler
+ 1) >> 1;
3803 if (fixup
->super
.num_consts
) {
3804 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3805 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3809 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3811 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3814 /* Base Shader Body */
3815 if (FAILED(shader_generate_code(shader
, buffer
, reg_maps
, &priv_ctx
, NULL
, NULL
)))
3818 if(args
->super
.srgb_correction
) {
3819 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3820 priv_ctx
.target_version
>= NV2
);
3823 if (custom_linear_fog
)
3824 arbfp_add_linear_fog(buffer
, fragcolor
, "TA");
3826 if(strcmp(fragcolor
, "result.color")) {
3827 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3829 shader_addline(buffer
, "END\n");
3831 /* TODO: change to resource.glObjectHandle or something like that */
3832 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3834 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3835 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3837 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3838 if (!shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
->buffer
))
3844 static int compare_sig(const struct wined3d_shader_signature
*sig1
, const struct wined3d_shader_signature
*sig2
)
3849 if (sig1
->element_count
!= sig2
->element_count
)
3850 return sig1
->element_count
< sig2
->element_count
? -1 : 1;
3852 for (i
= 0; i
< sig1
->element_count
; ++i
)
3854 const struct wined3d_shader_signature_element
*e1
, *e2
;
3856 e1
= &sig1
->elements
[i
];
3857 e2
= &sig2
->elements
[i
];
3859 if (!e1
->semantic_name
|| !e2
->semantic_name
)
3861 /* Compare pointers, not contents. One string is NULL (element
3862 * does not exist), the other one is not NULL. */
3863 if (e1
->semantic_name
!= e2
->semantic_name
)
3864 return e1
->semantic_name
< e2
->semantic_name
? -1 : 1;
3868 if ((ret
= strcmp(e1
->semantic_name
, e2
->semantic_name
)))
3870 if ((ret
= wined3d_uint32_compare(e1
->semantic_idx
, e2
->semantic_idx
)))
3872 if ((ret
= wined3d_uint32_compare(e1
->sysval_semantic
, e2
->sysval_semantic
)))
3874 if ((ret
= wined3d_uint32_compare(e1
->component_type
, e2
->component_type
)))
3876 if ((ret
= wined3d_uint32_compare(e1
->register_idx
, e2
->register_idx
)))
3878 if ((ret
= wined3d_uint32_compare(e1
->mask
, e2
->mask
)))
3884 static void clone_sig(struct wined3d_shader_signature
*new, const struct wined3d_shader_signature
*sig
)
3889 new->element_count
= sig
->element_count
;
3890 new->elements
= heap_calloc(new->element_count
, sizeof(*new->elements
));
3891 for (i
= 0; i
< sig
->element_count
; ++i
)
3893 new->elements
[i
] = sig
->elements
[i
];
3895 if (!new->elements
[i
].semantic_name
)
3898 /* Clone the semantic string */
3899 name
= heap_alloc(strlen(sig
->elements
[i
].semantic_name
) + 1);
3900 strcpy(name
, sig
->elements
[i
].semantic_name
);
3901 new->elements
[i
].semantic_name
= name
;
3905 static unsigned int find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature
*sig
)
3907 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3908 struct ps_signature
*found_sig
;
3912 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3913 TRACE("Found existing signature %u\n", found_sig
->idx
);
3914 return found_sig
->idx
;
3916 found_sig
= heap_alloc_zero(sizeof(*found_sig
));
3917 clone_sig(&found_sig
->sig
, sig
);
3918 found_sig
->idx
= priv
->ps_sig_number
++;
3919 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3920 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3922 ERR("Failed to insert program entry.\n");
3924 return found_sig
->idx
;
3927 static void init_output_registers(const struct wined3d_shader
*shader
,
3928 const struct wined3d_shader_signature
*ps_input_sig
,
3929 struct shader_arb_ctx_priv
*priv_ctx
, struct arb_vs_compiled_shader
*compiled
)
3932 static const char * const texcoords
[8] =
3934 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3935 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3937 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3938 * and varying 9 to result.color.secondary
3940 static const char * const decl_idx_to_string
[MAX_REG_INPUT
] =
3942 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3943 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]",
3944 "result.color.primary", "result.color.secondary"
3949 TRACE("Pixel shader uses builtin varyings\n");
3950 /* Map builtins to builtins */
3951 for(i
= 0; i
< 8; i
++)
3953 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3955 priv_ctx
->color_output
[0] = "result.color.primary";
3956 priv_ctx
->color_output
[1] = "result.color.secondary";
3957 priv_ctx
->fog_output
= "TMP_FOGCOORD";
3959 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3960 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
3962 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
3964 if (!output
->semantic_name
)
3967 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
))
3969 TRACE("o%u is TMP_OUT\n", output
->register_idx
);
3970 if (!output
->semantic_idx
)
3971 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
3973 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3975 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
))
3977 TRACE("o%u is result.pointsize\n", output
->register_idx
);
3978 if (!output
->semantic_idx
)
3979 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
3981 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3983 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
3985 TRACE("o%u is result.color.?, idx %u\n", output
->register_idx
, output
->semantic_idx
);
3986 if (!output
->semantic_idx
)
3987 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.primary";
3988 else if (output
->semantic_idx
== 1)
3989 priv_ctx
->vs_output
[output
->register_idx
] = "result.color.secondary";
3990 else priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3992 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
3994 TRACE("o%u is result.texcoord[%u]\n", output
->register_idx
, output
->semantic_idx
);
3995 if (output
->semantic_idx
>= 8)
3996 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
3998 priv_ctx
->vs_output
[output
->register_idx
] = texcoords
[output
->semantic_idx
];
4000 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4002 TRACE("o%u is result.fogcoord\n", output
->register_idx
);
4003 if (output
->semantic_idx
> 0)
4004 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4006 priv_ctx
->vs_output
[output
->register_idx
] = "result.fogcoord";
4010 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4016 TRACE("Pixel shader uses declared varyings\n");
4018 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
4019 for(i
= 0; i
< 8; i
++)
4021 priv_ctx
->texcrd_output
[i
] = "TA";
4023 priv_ctx
->color_output
[0] = "TA";
4024 priv_ctx
->color_output
[1] = "TA";
4025 priv_ctx
->fog_output
= "TA";
4027 for (i
= 0; i
< ps_input_sig
->element_count
; ++i
)
4029 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[i
];
4031 if (!input
->semantic_name
)
4034 /* If a declared input register is not written by builtin arguments, don't write to it.
4035 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4037 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4038 * to TMP_OUT in any case
4040 if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_TEXCOORD
))
4042 if (input
->semantic_idx
< 8)
4043 priv_ctx
->texcrd_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4045 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_COLOR
))
4047 if (input
->semantic_idx
< 2)
4048 priv_ctx
->color_output
[input
->semantic_idx
] = decl_idx_to_string
[input
->register_idx
];
4050 else if (shader_match_semantic(input
->semantic_name
, WINED3D_DECL_USAGE_FOG
))
4052 if (!input
->semantic_idx
)
4053 priv_ctx
->fog_output
= decl_idx_to_string
[input
->register_idx
];
4060 if (!strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.primary")
4061 || !strcmp(decl_idx_to_string
[input
->register_idx
], "result.color.secondary"))
4063 compiled
->need_color_unclamp
= TRUE
;
4067 /* Map declared to declared */
4068 for (i
= 0; i
< shader
->output_signature
.element_count
; ++i
)
4070 const struct wined3d_shader_signature_element
*output
= &shader
->output_signature
.elements
[i
];
4072 /* Write unread output to TA to throw them away */
4073 priv_ctx
->vs_output
[output
->register_idx
] = "TA";
4075 if (!output
->semantic_name
)
4078 if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_POSITION
) && !output
->semantic_idx
)
4080 priv_ctx
->vs_output
[output
->register_idx
] = "TMP_OUT";
4083 else if (shader_match_semantic(output
->semantic_name
, WINED3D_DECL_USAGE_PSIZE
) && !output
->semantic_idx
)
4085 priv_ctx
->vs_output
[output
->register_idx
] = "result.pointsize";
4089 for (j
= 0; j
< ps_input_sig
->element_count
; ++j
)
4091 const struct wined3d_shader_signature_element
*input
= &ps_input_sig
->elements
[j
];
4093 if (!input
->semantic_name
)
4096 if (!strcmp(input
->semantic_name
, output
->semantic_name
)
4097 && input
->semantic_idx
== output
->semantic_idx
)
4099 priv_ctx
->vs_output
[output
->register_idx
] = decl_idx_to_string
[input
->register_idx
];
4101 if (!strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.primary")
4102 || !strcmp(priv_ctx
->vs_output
[output
->register_idx
], "result.color.secondary"))
4104 compiled
->need_color_unclamp
= TRUE
;
4111 /* Context activation is done by the caller. */
4112 static GLuint
shader_arb_generate_vshader(const struct wined3d_shader
*shader
,
4113 const struct wined3d_gl_info
*gl_info
, struct wined3d_string_buffer
*buffer
,
4114 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
,
4115 const struct wined3d_shader_signature
*ps_input_sig
)
4117 const struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4118 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4120 unsigned int next_local
= 0;
4121 struct shader_arb_ctx_priv priv_ctx
;
4124 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4125 priv_ctx
.gl_info
= gl_info
;
4126 priv_ctx
.cur_vs_args
= args
;
4127 list_init(&priv_ctx
.control_frames
);
4128 init_output_registers(shader
, ps_input_sig
, &priv_ctx
, compiled
);
4130 /* Create the hw ARB shader */
4131 shader_addline(buffer
, "!!ARBvp1.0\n");
4133 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4134 * mesurable performance penalty, and we can always make use of it for clipplanes.
4136 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4138 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4139 priv_ctx
.target_version
= NV3
;
4140 shader_addline(buffer
, "ADDRESS aL;\n");
4142 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4144 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4145 priv_ctx
.target_version
= NV2
;
4146 shader_addline(buffer
, "ADDRESS aL;\n");
4148 priv_ctx
.target_version
= ARB
;
4151 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4153 shader_addline(buffer
, "TEMP TMP_FOGCOORD;\n");
4154 if (need_helper_const(shader_data
, reg_maps
, gl_info
))
4157 wined3d_ftoa(eps
, ftoa_tmp
);
4158 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %s};\n", ftoa_tmp
);
4160 if (need_rel_addr_const(shader_data
, reg_maps
, gl_info
))
4162 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", shader_data
->rel_offset
);
4163 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4166 shader_addline(buffer
, "TEMP TA;\n");
4167 shader_addline(buffer
, "TEMP TB;\n");
4169 /* Base Declarations */
4170 shader_generate_arb_declarations(shader
, reg_maps
, buffer
, gl_info
,
4171 &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4173 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4175 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4176 if (reg_maps
->integer_constants
& (1u << i
) && priv_ctx
.target_version
>= NV2
)
4178 const unsigned int *control_values
= find_loop_control_values(shader
, i
);
4182 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4183 control_values
[0], control_values
[1], control_values
[2]);
4187 compiled
->int_consts
[i
] = next_local
;
4188 compiled
->num_int_consts
++;
4189 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4194 /* We need a constant to fixup the final position */
4195 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4196 compiled
->pos_fixup
= next_local
++;
4198 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4199 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4200 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4201 * a replacement shader depend on the texcoord.w being set properly.
4203 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4204 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4205 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4206 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4207 * this can eat a number of instructions, so skip it unless this cap is set as well
4209 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4211 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4212 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4215 /* The shader starts with the main function */
4216 priv_ctx
.in_main_func
= TRUE
;
4217 /* Base Shader Body */
4218 if (FAILED(shader_generate_code(shader
, buffer
, reg_maps
, &priv_ctx
, NULL
, NULL
)))
4221 if (!priv_ctx
.footer_written
) vshader_add_footer(&priv_ctx
,
4222 shader_data
, args
, reg_maps
, gl_info
, buffer
);
4224 shader_addline(buffer
, "END\n");
4226 /* TODO: change to resource.glObjectHandle or something like that */
4227 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4229 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4230 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4232 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4233 if (!shader_arb_compile(gl_info
, GL_VERTEX_PROGRAM_ARB
, buffer
->buffer
))
4239 /* Context activation is done by the caller. */
4240 static struct arb_ps_compiled_shader
*find_arb_pshader(struct wined3d_context_gl
*context_gl
,
4241 struct wined3d_shader
*shader
, const struct arb_ps_compile_args
*args
)
4243 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4244 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4245 struct wined3d_device
*device
= shader
->device
;
4248 struct arb_ps_compiled_shader
*new_array
;
4249 struct wined3d_string_buffer buffer
;
4250 struct arb_pshader_private
*shader_data
;
4253 if (!shader
->backend_data
)
4255 struct shader_arb_priv
*priv
= device
->shader_priv
;
4257 shader
->backend_data
= heap_alloc_zero(sizeof(*shader_data
));
4258 shader_data
= shader
->backend_data
;
4259 shader_data
->clamp_consts
= shader
->reg_maps
.shader_version
.major
== 1;
4261 if (shader
->reg_maps
.shader_version
.major
< 3)
4262 shader_data
->input_signature_idx
= ~0U;
4264 shader_data
->input_signature_idx
= find_input_signature(priv
, &shader
->input_signature
);
4266 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4268 if (!d3d_info
->vs_clipping
)
4269 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->reg_maps
,
4270 d3d_info
->limits
.ffp_blend_stages
- 1);
4272 shader_data
->clipplane_emulation
= ~0U;
4274 shader_data
= shader
->backend_data
;
4276 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4277 * so a linear search is more performant than a hashmap or a binary search
4278 * (cache coherency etc)
4280 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4282 if (!memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)))
4283 return &shader_data
->gl_shaders
[i
];
4286 TRACE("No matching GL shader found, compiling a new shader\n");
4287 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4288 if (shader_data
->num_gl_shaders
)
4290 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4291 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4292 new_size
* sizeof(*shader_data
->gl_shaders
));
4296 new_array
= heap_alloc_zero(sizeof(*shader_data
->gl_shaders
));
4301 ERR("Out of memory\n");
4304 shader_data
->gl_shaders
= new_array
;
4305 shader_data
->shader_array_size
= new_size
;
4308 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4310 if (!string_buffer_init(&buffer
))
4312 ERR("Failed to initialize shader buffer.\n");
4316 ret
= shader_arb_generate_pshader(shader
, gl_info
, &buffer
, args
,
4317 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4318 string_buffer_free(&buffer
);
4319 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4321 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4324 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4325 const DWORD use_map
, BOOL skip_int
) {
4326 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4327 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4328 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4329 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4330 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4331 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4332 if(skip_int
) return TRUE
;
4334 return !memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
));
4337 static struct arb_vs_compiled_shader
*find_arb_vshader(struct wined3d_shader
*shader
,
4338 const struct wined3d_gl_info
*gl_info
, DWORD use_map
, const struct arb_vs_compile_args
*args
,
4339 const struct wined3d_shader_signature
*ps_input_sig
)
4343 struct arb_vs_compiled_shader
*new_array
;
4344 struct wined3d_string_buffer buffer
;
4345 struct arb_vshader_private
*shader_data
;
4348 if (!shader
->backend_data
)
4350 const struct wined3d_shader_reg_maps
*reg_maps
= &shader
->reg_maps
;
4352 shader
->backend_data
= heap_alloc_zero(sizeof(*shader_data
));
4353 shader_data
= shader
->backend_data
;
4355 if ((gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
)
4356 && reg_maps
->min_rel_offset
<= reg_maps
->max_rel_offset
)
4358 if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 127)
4360 FIXME("The difference between the minimum and maximum relative offset is > 127.\n");
4361 FIXME("Which this OpenGL implementation does not support. Try using GLSL.\n");
4362 FIXME("Min: %u, Max: %u.\n", reg_maps
->min_rel_offset
, reg_maps
->max_rel_offset
);
4364 else if (reg_maps
->max_rel_offset
- reg_maps
->min_rel_offset
> 63)
4365 shader_data
->rel_offset
= reg_maps
->min_rel_offset
+ 63;
4366 else if (reg_maps
->max_rel_offset
> 63)
4367 shader_data
->rel_offset
= reg_maps
->min_rel_offset
;
4370 shader_data
= shader
->backend_data
;
4372 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4373 * so a linear search is more performant than a hashmap or a binary search
4374 * (cache coherency etc)
4376 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4377 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4378 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4380 return &shader_data
->gl_shaders
[i
];
4384 TRACE("No matching GL shader found, compiling a new shader\n");
4386 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4387 if (shader_data
->num_gl_shaders
)
4389 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4390 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4391 new_size
* sizeof(*shader_data
->gl_shaders
));
4395 new_array
= heap_alloc_zero(sizeof(*shader_data
->gl_shaders
));
4400 ERR("Out of memory\n");
4403 shader_data
->gl_shaders
= new_array
;
4404 shader_data
->shader_array_size
= new_size
;
4407 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4409 if (!string_buffer_init(&buffer
))
4411 ERR("Failed to initialize shader buffer.\n");
4415 ret
= shader_arb_generate_vshader(shader
, gl_info
, &buffer
, args
,
4416 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
],
4418 string_buffer_free(&buffer
);
4419 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4421 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4424 static void find_arb_ps_compile_args(const struct wined3d_state
*state
,
4425 struct wined3d_context_gl
*context_gl
, const struct wined3d_shader
*shader
,
4426 struct arb_ps_compile_args
*args
)
4428 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4429 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4430 struct wined3d_device
*device
= context_gl
->c
.device
;
4431 const struct wined3d_ivec4
*int_consts
;
4432 const BOOL
*bool_consts
;
4436 bool_consts
= wined3d_buffer_load_sysmem(device
->push_constants
[WINED3D_PUSH_CONSTANTS_PS_B
], &context_gl
->c
);
4437 int_consts
= wined3d_buffer_load_sysmem(device
->push_constants
[WINED3D_PUSH_CONSTANTS_PS_I
], &context_gl
->c
);
4439 find_ps_compile_args(state
, shader
, context_gl
->c
.stream_info
.position_transformed
, &args
->super
, &context_gl
->c
);
4441 /* This forces all local boolean constants to 1 to make them stateblock independent */
4442 args
->bools
= shader
->reg_maps
.local_bool_consts
;
4444 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4447 args
->bools
|= ( 1u << i
);
4450 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4451 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4452 * duplicate the shader than have a no-op KIL instruction in every shader
4454 if (!d3d_info
->vs_clipping
&& use_vs(state
)
4455 && state
->render_states
[WINED3D_RS_CLIPPING
]
4456 && state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
])
4461 /* Skip if unused or local, or supported natively */
4462 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4463 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4465 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4469 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4471 if (int_skip
& (1u << i
))
4473 args
->loop_ctrl
[i
][0] = 0;
4474 args
->loop_ctrl
[i
][1] = 0;
4475 args
->loop_ctrl
[i
][2] = 0;
4479 args
->loop_ctrl
[i
][0] = int_consts
[i
].x
;
4480 args
->loop_ctrl
[i
][1] = int_consts
[i
].y
;
4481 args
->loop_ctrl
[i
][2] = int_consts
[i
].z
;
4486 static void find_arb_vs_compile_args(const struct wined3d_state
*state
,
4487 struct wined3d_context_gl
*context_gl
, const struct wined3d_shader
*shader
,
4488 struct arb_vs_compile_args
*args
)
4490 const struct wined3d_d3d_info
*d3d_info
= context_gl
->c
.d3d_info
;
4491 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4492 const struct wined3d_device
*device
= shader
->device
;
4493 const struct wined3d_adapter
*adapter
= device
->adapter
;
4494 const struct wined3d_ivec4
*int_consts
;
4495 const BOOL
*bool_consts
;
4499 bool_consts
= wined3d_buffer_load_sysmem(device
->push_constants
[WINED3D_PUSH_CONSTANTS_VS_B
], &context_gl
->c
);
4500 int_consts
= wined3d_buffer_load_sysmem(device
->push_constants
[WINED3D_PUSH_CONSTANTS_VS_I
], &context_gl
->c
);
4502 find_vs_compile_args(state
, shader
, &args
->super
, &context_gl
->c
);
4504 args
->clip
.boolclip_compare
= 0;
4507 const struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4508 const struct arb_pshader_private
*shader_priv
= ps
->backend_data
;
4509 args
->ps_signature
= shader_priv
->input_signature_idx
;
4511 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4515 args
->ps_signature
= ~0;
4516 if (!d3d_info
->vs_clipping
&& adapter
->fragment_pipe
== &arbfp_fragment_pipeline
)
4517 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(&context_gl
->c
) ? d3d_info
->limits
.ffp_blend_stages
: 0;
4518 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4521 if (args
->clip
.boolclip
.clip_texcoord
)
4523 if (state
->render_states
[WINED3D_RS_CLIPPING
])
4524 args
->clip
.boolclip
.clipplane_mask
= (unsigned char)state
->render_states
[WINED3D_RS_CLIPPLANEENABLE
];
4525 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4528 /* This forces all local boolean constants to 1 to make them stateblock independent */
4529 args
->clip
.boolclip
.bools
= shader
->reg_maps
.local_bool_consts
;
4530 for (i
= 0; i
< WINED3D_MAX_CONSTS_B
; ++i
)
4533 args
->clip
.boolclip
.bools
|= (1u << i
);
4536 args
->vertex
.samplers
[0] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 0];
4537 args
->vertex
.samplers
[1] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 1];
4538 args
->vertex
.samplers
[2] = context_gl
->tex_unit_map
[WINED3D_MAX_FRAGMENT_SAMPLERS
+ 2];
4539 args
->vertex
.samplers
[3] = 0;
4541 /* Skip if unused or local */
4542 int_skip
= ~shader
->reg_maps
.integer_constants
| shader
->reg_maps
.local_int_consts
;
4543 /* This is about flow control, not clipping. */
4544 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4546 memset(args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4550 for (i
= 0; i
< WINED3D_MAX_CONSTS_I
; ++i
)
4552 if (int_skip
& (1u << i
))
4554 args
->loop_ctrl
[i
][0] = 0;
4555 args
->loop_ctrl
[i
][1] = 0;
4556 args
->loop_ctrl
[i
][2] = 0;
4560 args
->loop_ctrl
[i
][0] = int_consts
[i
].x
;
4561 args
->loop_ctrl
[i
][1] = int_consts
[i
].y
;
4562 args
->loop_ctrl
[i
][2] = int_consts
[i
].z
;
4567 /* Context activation is done by the caller. */
4568 static void shader_arb_select(void *shader_priv
, struct wined3d_context
*context
,
4569 const struct wined3d_state
*state
)
4571 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
4572 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4573 struct shader_arb_priv
*priv
= shader_priv
;
4576 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4579 struct wined3d_shader
*ps
= state
->shader
[WINED3D_SHADER_TYPE_PIXEL
];
4580 struct arb_ps_compile_args compile_args
;
4581 struct arb_ps_compiled_shader
*compiled
;
4583 TRACE("Using pixel shader %p.\n", ps
);
4584 find_arb_ps_compile_args(state
, context_gl
, ps
, &compile_args
);
4585 compiled
= find_arb_pshader(context_gl
, ps
, &compile_args
);
4586 priv
->compiled_fprog
= compiled
;
4588 /* Bind the fragment program */
4589 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, compiled
->prgId
));
4590 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, compiled->prgId);");
4592 if (!priv
->use_arbfp_fixed_func
)
4593 priv
->fragment_pipe
->fp_enable(context
, FALSE
);
4595 /* Enable OpenGL fragment programs. */
4596 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4597 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4599 TRACE("Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB.\n", compiled
->prgId
);
4601 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4602 * a 1.x and newer shader, reload the first 8 constants
4604 if (priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
)
4606 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->backend_data
)->clamp_consts
;
4607 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, 8);
4608 for(i
= 0; i
< 8; i
++)
4610 priv
->pshader_const_dirty
[i
] = 1;
4612 /* Also takes care of loading local constants */
4613 shader_arb_load_constants_internal(shader_priv
, context_gl
, state
, TRUE
, FALSE
, TRUE
);
4617 UINT rt_height
= state
->fb
.render_targets
[0]->height
;
4618 shader_arb_ps_local_constants(compiled
, context_gl
, state
, rt_height
);
4621 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4622 if (compiled
->np2fixup_info
.super
.active
)
4623 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_NP2_FIXUP
;
4625 if (ps
->load_local_constsF
)
4626 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
4630 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4632 /* Disable only if we're not using arbfp fixed function fragment
4633 * processing. If this is used, keep GL_FRAGMENT_PROGRAM_ARB
4634 * enabled, and the fixed function pipeline will bind the fixed
4635 * function replacement shader. */
4636 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4637 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4639 priv
->fragment_pipe
->fp_enable(context
, TRUE
);
4644 struct wined3d_shader
*vs
= state
->shader
[WINED3D_SHADER_TYPE_VERTEX
];
4645 struct arb_vs_compile_args compile_args
;
4646 struct arb_vs_compiled_shader
*compiled
;
4647 const struct wined3d_shader_signature
*ps_input_sig
;
4649 TRACE("Using vertex shader %p\n", vs
);
4650 find_arb_vs_compile_args(state
, context_gl
, vs
, &compile_args
);
4652 /* Instead of searching for the signature in the signature list, read the one from the
4653 * current pixel shader. It's maybe not the shader where the signature came from, but it
4654 * is the same signature and faster to find. */
4655 if (compile_args
.ps_signature
== ~0U)
4656 ps_input_sig
= NULL
;
4658 ps_input_sig
= &state
->shader
[WINED3D_SHADER_TYPE_PIXEL
]->input_signature
;
4660 compiled
= find_arb_vshader(vs
, gl_info
, context
->stream_info
.use_map
,
4661 &compile_args
, ps_input_sig
);
4662 priv
->compiled_vprog
= compiled
;
4664 /* Bind the vertex program */
4665 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, compiled
->prgId
));
4666 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, compiled->prgId);");
4668 priv
->vertex_pipe
->vp_enable(context
, FALSE
);
4670 /* Enable OpenGL vertex programs */
4671 gl_info
->gl_ops
.gl
.p_glEnable(GL_VERTEX_PROGRAM_ARB
);
4672 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4673 TRACE("Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", compiled
->prgId
);
4674 shader_arb_vs_local_constants(compiled
, context_gl
, state
);
4676 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4677 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4679 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4681 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4682 checkGLcall("glClampColorARB");
4684 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4688 if (vs
->load_local_constsF
)
4689 context
->constant_update_mask
|= WINED3D_SHADER_CONST_VS_F
;
4693 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4695 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4696 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4698 priv
->vertex_pipe
->vp_enable(context
, TRUE
);
4702 static void shader_arb_select_compute(void *shader_priv
, struct wined3d_context
*context
,
4703 const struct wined3d_state
*state
)
4705 ERR("Compute pipeline not supported by the ARB shader backend.\n");
4708 /* Context activation is done by the caller. */
4709 static void shader_arb_disable(void *shader_priv
, struct wined3d_context
*context
)
4711 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
4712 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
4713 struct shader_arb_priv
*priv
= shader_priv
;
4715 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4717 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4718 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4720 priv
->fragment_pipe
->fp_enable(context
, FALSE
);
4722 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4724 gl_info
->gl_ops
.gl
.p_glDisable(GL_VERTEX_PROGRAM_ARB
);
4725 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4727 priv
->vertex_pipe
->vp_enable(context
, FALSE
);
4729 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
] && priv
->last_vs_color_unclamp
)
4731 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, GL_FIXED_ONLY_ARB
));
4732 checkGLcall("glClampColorARB");
4733 priv
->last_vs_color_unclamp
= FALSE
;
4736 context
->shader_update_mask
= (1u << WINED3D_SHADER_TYPE_PIXEL
)
4737 | (1u << WINED3D_SHADER_TYPE_VERTEX
)
4738 | (1u << WINED3D_SHADER_TYPE_GEOMETRY
)
4739 | (1u << WINED3D_SHADER_TYPE_HULL
)
4740 | (1u << WINED3D_SHADER_TYPE_DOMAIN
)
4741 | (1u << WINED3D_SHADER_TYPE_COMPUTE
);
4744 static void shader_arb_destroy(struct wined3d_shader
*shader
)
4746 struct wined3d_device
*device
= shader
->device
;
4747 const struct wined3d_gl_info
*gl_info
;
4748 struct wined3d_context
*context
;
4751 /* This can happen if a shader was never compiled */
4752 if (!shader
->backend_data
)
4755 context
= context_acquire(device
, NULL
, 0);
4756 gl_info
= wined3d_context_gl(context
)->gl_info
;
4758 if (shader_is_pshader_version(shader
->reg_maps
.shader_version
.type
))
4760 struct arb_pshader_private
*shader_data
= shader
->backend_data
;
4762 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4763 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4765 heap_free(shader_data
->gl_shaders
);
4769 struct arb_vshader_private
*shader_data
= shader
->backend_data
;
4771 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4772 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4774 heap_free(shader_data
->gl_shaders
);
4777 checkGLcall("delete programs");
4779 context_release(context
);
4781 heap_free(shader
->backend_data
);
4782 shader
->backend_data
= NULL
;
4785 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4787 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4788 return compare_sig(key
, &e
->sig
);
4791 static HRESULT
shader_arb_alloc(struct wined3d_device
*device
, const struct wined3d_vertex_pipe_ops
*vertex_pipe
,
4792 const struct wined3d_fragment_pipe_ops
*fragment_pipe
)
4794 const struct wined3d_d3d_info
*d3d_info
= &device
->adapter
->d3d_info
;
4795 struct fragment_caps fragment_caps
;
4796 void *vertex_priv
, *fragment_priv
;
4797 struct shader_arb_priv
*priv
;
4799 if (!(priv
= heap_alloc_zero(sizeof(*priv
))))
4800 return E_OUTOFMEMORY
;
4802 if (!(vertex_priv
= vertex_pipe
->vp_alloc(&arb_program_shader_backend
, priv
)))
4804 ERR("Failed to initialize vertex pipe.\n");
4809 if (!(fragment_priv
= fragment_pipe
->alloc_private(&arb_program_shader_backend
, priv
)))
4811 ERR("Failed to initialize fragment pipe.\n");
4812 vertex_pipe
->vp_free(device
, NULL
);
4817 memset(priv
->vshader_const_dirty
, 1,
4818 sizeof(*priv
->vshader_const_dirty
) * d3d_info
->limits
.vs_uniform_count
);
4819 memset(priv
->pshader_const_dirty
, 1,
4820 sizeof(*priv
->pshader_const_dirty
) * d3d_info
->limits
.ps_uniform_count
);
4822 wine_rb_init(&priv
->signature_tree
, sig_tree_compare
);
4824 priv
->vertex_pipe
= vertex_pipe
;
4825 priv
->fragment_pipe
= fragment_pipe
;
4826 fragment_pipe
->get_caps(device
->adapter
, &fragment_caps
);
4827 priv
->ffp_proj_control
= fragment_caps
.wined3d_caps
& WINED3D_FRAGMENT_CAP_PROJ_CONTROL
;
4829 device
->vertex_priv
= vertex_priv
;
4830 device
->fragment_priv
= fragment_priv
;
4831 device
->shader_priv
= priv
;
4836 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4838 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4841 for (i
= 0; i
< sig
->sig
.element_count
; ++i
)
4843 heap_free((char *)sig
->sig
.elements
[i
].semantic_name
);
4845 heap_free(sig
->sig
.elements
);
4849 /* Context activation is done by the caller. */
4850 static void shader_arb_free(struct wined3d_device
*device
, struct wined3d_context
*context
)
4852 struct shader_arb_priv
*priv
= device
->shader_priv
;
4854 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4855 priv
->fragment_pipe
->free_private(device
, context
);
4856 priv
->vertex_pipe
->vp_free(device
, context
);
4857 heap_free(device
->shader_priv
);
4860 static BOOL
shader_arb_allocate_context_data(struct wined3d_context
*context
)
4865 static void shader_arb_free_context_data(struct wined3d_context
*context
)
4867 struct shader_arb_priv
*priv
;
4869 priv
= context
->device
->shader_priv
;
4870 if (priv
->last_context
== context
)
4871 priv
->last_context
= NULL
;
4874 static void shader_arb_init_context_state(struct wined3d_context
*context
) {}
4876 static void shader_arb_get_caps(const struct wined3d_adapter
*adapter
, struct shader_caps
*caps
)
4878 const struct wined3d_gl_info
*gl_info
= &wined3d_adapter_gl_const(adapter
)->gl_info
;
4880 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4885 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
4886 * for vertex programs. If the native limit is less than that it's
4887 * not very useful, and e.g. Mesa swrast returns 0, probably to
4888 * indicate it's a software implementation. */
4889 if (gl_info
->limits
.arb_vs_native_constants
< 96)
4890 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
4892 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
4894 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4897 TRACE("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
4899 else if (vs_consts
>= 256)
4901 /* Shader Model 2.0 requires at least 256 vertex shader constants */
4903 TRACE("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
4908 TRACE("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
4910 caps
->vs_version
= min(wined3d_settings
.max_sm_vs
, vs_version
);
4911 caps
->vs_uniform_count
= min(WINED3D_MAX_VS_CONSTS_F
, vs_consts
);
4915 caps
->vs_version
= 0;
4916 caps
->vs_uniform_count
= 0;
4919 caps
->hs_version
= 0;
4920 caps
->ds_version
= 0;
4921 caps
->gs_version
= 0;
4922 caps
->cs_version
= 0;
4924 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4929 /* Similar as above for vertex programs, but the minimum for fragment
4930 * programs is 24. */
4931 if (gl_info
->limits
.arb_ps_native_constants
< 24)
4932 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
4934 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
4936 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4939 TRACE("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
4941 else if (ps_consts
>= 32)
4943 /* Shader Model 2.0 requires at least 32 pixel shader constants */
4945 TRACE("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
4950 TRACE("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
4952 caps
->ps_version
= min(wined3d_settings
.max_sm_ps
, ps_version
);
4953 caps
->ps_uniform_count
= min(WINED3D_MAX_PS_CONSTS_F
, ps_consts
);
4954 caps
->ps_1x_max_value
= 8.0f
;
4958 caps
->ps_version
= 0;
4959 caps
->ps_uniform_count
= 0;
4960 caps
->ps_1x_max_value
= 0.0f
;
4963 caps
->varying_count
= 0;
4964 caps
->wined3d_caps
= WINED3D_SHADER_CAP_SRGB_WRITE
;
4965 if (use_nv_clip(gl_info
))
4966 caps
->wined3d_caps
|= WINED3D_SHADER_CAP_VS_CLIPPING
;
4969 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
4971 /* We support everything except complex conversions. */
4972 return !is_complex_fixup(fixup
);
4975 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
4977 char write_mask
[20], regstr
[50];
4978 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
4979 BOOL is_color
= FALSE
;
4980 const struct wined3d_shader_dst_param
*dst
;
4982 if (!ins
->dst_count
) return;
4986 if (!shift
) return; /* Saturate alone is handled by the instructions */
4988 shader_arb_get_write_mask(ins
, dst
, write_mask
);
4989 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
4991 /* Generate a line that does the output modifier computation
4992 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
4993 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
4995 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
4996 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
4999 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
5001 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
5002 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
5003 /* WINED3DSIH_AND */ NULL
,
5004 /* WINED3DSIH_ATOMIC_AND */ NULL
,
5005 /* WINED3DSIH_ATOMIC_CMP_STORE */ NULL
,
5006 /* WINED3DSIH_ATOMIC_IADD */ NULL
,
5007 /* WINED3DSIH_ATOMIC_IMAX */ NULL
,
5008 /* WINED3DSIH_ATOMIC_IMIN */ NULL
,
5009 /* WINED3DSIH_ATOMIC_OR */ NULL
,
5010 /* WINED3DSIH_ATOMIC_UMAX */ NULL
,
5011 /* WINED3DSIH_ATOMIC_UMIN */ NULL
,
5012 /* WINED3DSIH_ATOMIC_XOR */ NULL
,
5013 /* WINED3DSIH_BEM */ pshader_hw_bem
,
5014 /* WINED3DSIH_BFI */ NULL
,
5015 /* WINED3DSIH_BFREV */ NULL
,
5016 /* WINED3DSIH_BREAK */ shader_hw_break
,
5017 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
5018 /* WINED3DSIH_BREAKP */ NULL
,
5019 /* WINED3DSIH_BUFINFO */ NULL
,
5020 /* WINED3DSIH_CALL */ shader_hw_call
,
5021 /* WINED3DSIH_CALLNZ */ NULL
,
5022 /* WINED3DSIH_CASE */ NULL
,
5023 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
5024 /* WINED3DSIH_CND */ pshader_hw_cnd
,
5025 /* WINED3DSIH_CONTINUE */ NULL
,
5026 /* WINED3DSIH_CONTINUEP */ NULL
,
5027 /* WINED3DSIH_COUNTBITS */ NULL
,
5028 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
5029 /* WINED3DSIH_CUT */ NULL
,
5030 /* WINED3DSIH_CUT_STREAM */ NULL
,
5031 /* WINED3DSIH_DCL */ shader_hw_nop
,
5032 /* WINED3DSIH_DCL_CONSTANT_BUFFER */ shader_hw_nop
,
5033 /* WINED3DSIH_DCL_FUNCTION_BODY */ NULL
,
5034 /* WINED3DSIH_DCL_FUNCTION_TABLE */ NULL
,
5035 /* WINED3DSIH_DCL_GLOBAL_FLAGS */ NULL
,
5036 /* WINED3DSIH_DCL_GS_INSTANCES */ NULL
,
5037 /* WINED3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT */ NULL
,
5038 /* WINED3DSIH_DCL_HS_JOIN_PHASE_INSTANCE_COUNT */ NULL
,
5039 /* WINED3DSIH_DCL_HS_MAX_TESSFACTOR */ NULL
,
5040 /* WINED3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER */ NULL
,
5041 /* WINED3DSIH_DCL_INDEX_RANGE */ NULL
,
5042 /* WINED3DSIH_DCL_INDEXABLE_TEMP */ NULL
,
5043 /* WINED3DSIH_DCL_INPUT */ NULL
,
5044 /* WINED3DSIH_DCL_INPUT_CONTROL_POINT_COUNT */ NULL
,
5045 /* WINED3DSIH_DCL_INPUT_PRIMITIVE */ shader_hw_nop
,
5046 /* WINED3DSIH_DCL_INPUT_PS */ NULL
,
5047 /* WINED3DSIH_DCL_INPUT_PS_SGV */ NULL
,
5048 /* WINED3DSIH_DCL_INPUT_PS_SIV */ NULL
,
5049 /* WINED3DSIH_DCL_INPUT_SGV */ NULL
,
5050 /* WINED3DSIH_DCL_INPUT_SIV */ NULL
,
5051 /* WINED3DSIH_DCL_INTERFACE */ NULL
,
5052 /* WINED3DSIH_DCL_OUTPUT */ NULL
,
5053 /* WINED3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT */ NULL
,
5054 /* WINED3DSIH_DCL_OUTPUT_SIV */ NULL
,
5055 /* WINED3DSIH_DCL_OUTPUT_TOPOLOGY */ shader_hw_nop
,
5056 /* WINED3DSIH_DCL_RESOURCE_RAW */ NULL
,
5057 /* WINED3DSIH_DCL_RESOURCE_STRUCTURED */ NULL
,
5058 /* WINED3DSIH_DCL_SAMPLER */ NULL
,
5059 /* WINED3DSIH_DCL_STREAM */ NULL
,
5060 /* WINED3DSIH_DCL_TEMPS */ NULL
,
5061 /* WINED3DSIH_DCL_TESSELLATOR_DOMAIN */ NULL
,
5062 /* WINED3DSIH_DCL_TESSELLATOR_OUTPUT_PRIMITIVE */ NULL
,
5063 /* WINED3DSIH_DCL_TESSELLATOR_PARTITIONING */ NULL
,
5064 /* WINED3DSIH_DCL_TGSM_RAW */ NULL
,
5065 /* WINED3DSIH_DCL_TGSM_STRUCTURED */ NULL
,
5066 /* WINED3DSIH_DCL_THREAD_GROUP */ NULL
,
5067 /* WINED3DSIH_DCL_UAV_RAW */ NULL
,
5068 /* WINED3DSIH_DCL_UAV_STRUCTURED */ NULL
,
5069 /* WINED3DSIH_DCL_UAV_TYPED */ NULL
,
5070 /* WINED3DSIH_DCL_VERTICES_OUT */ shader_hw_nop
,
5071 /* WINED3DSIH_DEF */ shader_hw_nop
,
5072 /* WINED3DSIH_DEFAULT */ NULL
,
5073 /* WINED3DSIH_DEFB */ shader_hw_nop
,
5074 /* WINED3DSIH_DEFI */ shader_hw_nop
,
5075 /* WINED3DSIH_DIV */ NULL
,
5076 /* WINED3DSIH_DP2 */ NULL
,
5077 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
5078 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
5079 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
5080 /* WINED3DSIH_DST */ shader_hw_map2gl
,
5081 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
5082 /* WINED3DSIH_DSX_COARSE */ NULL
,
5083 /* WINED3DSIH_DSX_FINE */ NULL
,
5084 /* WINED3DSIH_DSY */ shader_hw_dsy
,
5085 /* WINED3DSIH_DSY_COARSE */ NULL
,
5086 /* WINED3DSIH_DSY_FINE */ NULL
,
5087 /* WINED3DSIH_ELSE */ shader_hw_else
,
5088 /* WINED3DSIH_EMIT */ NULL
,
5089 /* WINED3DSIH_EMIT_STREAM */ NULL
,
5090 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
5091 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
5092 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
5093 /* WINED3DSIH_ENDSWITCH */ NULL
,
5094 /* WINED3DSIH_EQ */ NULL
,
5095 /* WINED3DSIH_EVAL_CENTROID */ NULL
,
5096 /* WINED3DSIH_EVAL_SAMPLE_INDEX */ NULL
,
5097 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
5098 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
5099 /* WINED3DSIH_F16TOF32 */ NULL
,
5100 /* WINED3DSIH_F32TOF16 */ NULL
,
5101 /* WINED3DSIH_FCALL */ NULL
,
5102 /* WINED3DSIH_FIRSTBIT_HI */ NULL
,
5103 /* WINED3DSIH_FIRSTBIT_LO */ NULL
,
5104 /* WINED3DSIH_FIRSTBIT_SHI */ NULL
,
5105 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
5106 /* WINED3DSIH_FTOI */ NULL
,
5107 /* WINED3DSIH_FTOU */ NULL
,
5108 /* WINED3DSIH_GATHER4 */ NULL
,
5109 /* WINED3DSIH_GATHER4_C */ NULL
,
5110 /* WINED3DSIH_GATHER4_PO */ NULL
,
5111 /* WINED3DSIH_GATHER4_PO_C */ NULL
,
5112 /* WINED3DSIH_GE */ NULL
,
5113 /* WINED3DSIH_HS_CONTROL_POINT_PHASE */ NULL
,
5114 /* WINED3DSIH_HS_DECLS */ NULL
,
5115 /* WINED3DSIH_HS_FORK_PHASE */ NULL
,
5116 /* WINED3DSIH_HS_JOIN_PHASE */ NULL
,
5117 /* WINED3DSIH_IADD */ NULL
,
5118 /* WINED3DSIH_IBFE */ NULL
,
5119 /* WINED3DSIH_IEQ */ NULL
,
5120 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
5121 /* WINED3DSIH_IFC */ shader_hw_ifc
,
5122 /* WINED3DSIH_IGE */ NULL
,
5123 /* WINED3DSIH_ILT */ NULL
,
5124 /* WINED3DSIH_IMAD */ NULL
,
5125 /* WINED3DSIH_IMAX */ NULL
,
5126 /* WINED3DSIH_IMIN */ NULL
,
5127 /* WINED3DSIH_IMM_ATOMIC_ALLOC */ NULL
,
5128 /* WINED3DSIH_IMM_ATOMIC_AND */ NULL
,
5129 /* WINED3DSIH_IMM_ATOMIC_CMP_EXCH */ NULL
,
5130 /* WINED3DSIH_IMM_ATOMIC_CONSUME */ NULL
,
5131 /* WINED3DSIH_IMM_ATOMIC_EXCH */ NULL
,
5132 /* WINED3DSIH_IMM_ATOMIC_IADD */ NULL
,
5133 /* WINED3DSIH_IMM_ATOMIC_IMAX */ NULL
,
5134 /* WINED3DSIH_IMM_ATOMIC_IMIN */ NULL
,
5135 /* WINED3DSIH_IMM_ATOMIC_OR */ NULL
,
5136 /* WINED3DSIH_IMM_ATOMIC_UMAX */ NULL
,
5137 /* WINED3DSIH_IMM_ATOMIC_UMIN */ NULL
,
5138 /* WINED3DSIH_IMM_ATOMIC_XOR */ NULL
,
5139 /* WINED3DSIH_IMUL */ NULL
,
5140 /* WINED3DSIH_INE */ NULL
,
5141 /* WINED3DSIH_INEG */ NULL
,
5142 /* WINED3DSIH_ISHL */ NULL
,
5143 /* WINED3DSIH_ISHR */ NULL
,
5144 /* WINED3DSIH_ITOF */ NULL
,
5145 /* WINED3DSIH_LABEL */ shader_hw_label
,
5146 /* WINED3DSIH_LD */ NULL
,
5147 /* WINED3DSIH_LD2DMS */ NULL
,
5148 /* WINED3DSIH_LD_RAW */ NULL
,
5149 /* WINED3DSIH_LD_STRUCTURED */ NULL
,
5150 /* WINED3DSIH_LD_UAV_TYPED */ NULL
,
5151 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
5152 /* WINED3DSIH_LOD */ NULL
,
5153 /* WINED3DSIH_LOG */ shader_hw_scalar_op
,
5154 /* WINED3DSIH_LOGP */ shader_hw_scalar_op
,
5155 /* WINED3DSIH_LOOP */ shader_hw_loop
,
5156 /* WINED3DSIH_LRP */ shader_hw_lrp
,
5157 /* WINED3DSIH_LT */ NULL
,
5158 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
5159 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
5160 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
5161 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
5162 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
5163 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
5164 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
5165 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
5166 /* WINED3DSIH_MOV */ shader_hw_mov
,
5167 /* WINED3DSIH_MOVA */ shader_hw_mov
,
5168 /* WINED3DSIH_MOVC */ NULL
,
5169 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
5170 /* WINED3DSIH_NE */ NULL
,
5171 /* WINED3DSIH_NOP */ shader_hw_nop
,
5172 /* WINED3DSIH_NOT */ NULL
,
5173 /* WINED3DSIH_NRM */ shader_hw_nrm
,
5174 /* WINED3DSIH_OR */ NULL
,
5175 /* WINED3DSIH_PHASE */ shader_hw_nop
,
5176 /* WINED3DSIH_POW */ shader_hw_pow
,
5177 /* WINED3DSIH_RCP */ shader_hw_scalar_op
,
5178 /* WINED3DSIH_REP */ shader_hw_rep
,
5179 /* WINED3DSIH_RESINFO */ NULL
,
5180 /* WINED3DSIH_RET */ shader_hw_ret
,
5181 /* WINED3DSIH_RETP */ NULL
,
5182 /* WINED3DSIH_ROUND_NE */ NULL
,
5183 /* WINED3DSIH_ROUND_NI */ NULL
,
5184 /* WINED3DSIH_ROUND_PI */ NULL
,
5185 /* WINED3DSIH_ROUND_Z */ NULL
,
5186 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
5187 /* WINED3DSIH_SAMPLE */ NULL
,
5188 /* WINED3DSIH_SAMPLE_B */ NULL
,
5189 /* WINED3DSIH_SAMPLE_C */ NULL
,
5190 /* WINED3DSIH_SAMPLE_C_LZ */ NULL
,
5191 /* WINED3DSIH_SAMPLE_GRAD */ NULL
,
5192 /* WINED3DSIH_SAMPLE_INFO */ NULL
,
5193 /* WINED3DSIH_SAMPLE_LOD */ NULL
,
5194 /* WINED3DSIH_SAMPLE_POS */ NULL
,
5195 /* WINED3DSIH_SETP */ NULL
,
5196 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
5197 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5198 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5199 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5200 /* WINED3DSIH_SQRT */ NULL
,
5201 /* WINED3DSIH_STORE_RAW */ NULL
,
5202 /* WINED3DSIH_STORE_STRUCTURED */ NULL
,
5203 /* WINED3DSIH_STORE_UAV_TYPED */ NULL
,
5204 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5205 /* WINED3DSIH_SWAPC */ NULL
,
5206 /* WINED3DSIH_SWITCH */ NULL
,
5207 /* WINED3DSIH_SYNC */ NULL
,
5208 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5209 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5210 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5211 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5212 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5213 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5214 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5215 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5216 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5217 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5218 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5219 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5220 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5221 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5222 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5223 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5224 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5225 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5226 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5227 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5228 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5229 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5230 /* WINED3DSIH_UBFE */ NULL
,
5231 /* WINED3DSIH_UDIV */ NULL
,
5232 /* WINED3DSIH_UGE */ NULL
,
5233 /* WINED3DSIH_ULT */ NULL
,
5234 /* WINED3DSIH_UMAX */ NULL
,
5235 /* WINED3DSIH_UMIN */ NULL
,
5236 /* WINED3DSIH_UMUL */ NULL
,
5237 /* WINED3DSIH_USHR */ NULL
,
5238 /* WINED3DSIH_UTOF */ NULL
,
5239 /* WINED3DSIH_XOR */ NULL
,
5242 static BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
,
5243 const struct wined3d_shader
*shader
, DWORD idx
)
5245 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5246 BOOL vshader
= shader_is_vshader_version(reg_maps
->shader_version
.type
);
5247 const struct wined3d_shader_lconst
*constant
;
5249 WORD flag
= (1u << idx
);
5250 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5252 if (reg_maps
->local_bool_consts
& flag
)
5254 /* What good is an if(bool) with a hardcoded local constant? I don't know, but handle it */
5255 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsB
, struct wined3d_shader_lconst
, entry
)
5257 if (constant
->idx
== idx
)
5259 return constant
->value
[0];
5262 ERR("Local constant not found\n");
5267 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5268 else bools
= priv
->cur_ps_args
->bools
;
5269 return bools
& flag
;
5273 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5274 const struct wined3d_shader
*shader
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5276 const struct wined3d_shader_reg_maps
*reg_maps
= ins
->ctx
->reg_maps
;
5277 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5279 /* Integer constants can either be a local constant, or they can be stored in the shader
5280 * type specific compile args. */
5281 if (reg_maps
->local_int_consts
& (1u << idx
))
5283 const struct wined3d_shader_lconst
*constant
;
5285 LIST_FOR_EACH_ENTRY(constant
, &shader
->constantsI
, struct wined3d_shader_lconst
, entry
)
5287 if (constant
->idx
== idx
)
5289 loop_control
->count
= constant
->value
[0];
5290 loop_control
->start
= constant
->value
[1];
5291 /* Step is signed. */
5292 loop_control
->step
= (int)constant
->value
[2];
5296 /* If this happens the flag was set incorrectly */
5297 ERR("Local constant not found\n");
5298 loop_control
->count
= 0;
5299 loop_control
->start
= 0;
5300 loop_control
->step
= 0;
5304 switch (reg_maps
->shader_version
.type
)
5306 case WINED3D_SHADER_TYPE_VERTEX
:
5307 /* Count and aL start value are unsigned */
5308 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5309 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5310 /* Step is signed. */
5311 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5314 case WINED3D_SHADER_TYPE_PIXEL
:
5315 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5316 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5317 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5321 FIXME("Unhandled shader type %#x.\n", reg_maps
->shader_version
.type
);
5326 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5328 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
;
5329 struct wined3d_shader_dst_param
*dst_param
;
5330 struct recorded_instruction
*rec
;
5333 if (!(rec
= heap_alloc_zero(sizeof(*rec
))))
5335 ERR("Out of memory\n");
5340 if (!(dst_param
= heap_alloc(sizeof(*dst_param
))))
5342 *dst_param
= *ins
->dst
;
5343 if (ins
->dst
->reg
.idx
[0].rel_addr
)
5345 if (!(rel_addr
= heap_alloc(sizeof(*rel_addr
))))
5347 *rel_addr
= *ins
->dst
->reg
.idx
[0].rel_addr
;
5348 dst_param
->reg
.idx
[0].rel_addr
= rel_addr
;
5350 rec
->ins
.dst
= dst_param
;
5352 if (!(src_param
= heap_calloc(ins
->src_count
, sizeof(*src_param
))))
5354 for (i
= 0; i
< ins
->src_count
; ++i
)
5356 src_param
[i
] = ins
->src
[i
];
5357 if (ins
->src
[i
].reg
.idx
[0].rel_addr
)
5359 if (!(rel_addr
= heap_alloc(sizeof(*rel_addr
))))
5361 *rel_addr
= *ins
->src
[i
].reg
.idx
[0].rel_addr
;
5362 src_param
[i
].reg
.idx
[0].rel_addr
= rel_addr
;
5365 rec
->ins
.src
= src_param
;
5366 list_add_tail(list
, &rec
->entry
);
5370 ERR("Out of memory\n");
5373 heap_free((void *)dst_param
->reg
.idx
[0].rel_addr
);
5374 heap_free(dst_param
);
5378 for (i
= 0; i
< ins
->src_count
; ++i
)
5380 heap_free((void *)src_param
[i
].reg
.idx
[0].rel_addr
);
5382 heap_free(src_param
);
5387 static void free_recorded_instruction(struct list
*list
)
5389 struct recorded_instruction
*rec_ins
, *entry2
;
5392 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5394 list_remove(&rec_ins
->entry
);
5395 if (rec_ins
->ins
.dst
)
5397 heap_free((void *)rec_ins
->ins
.dst
->reg
.idx
[0].rel_addr
);
5398 heap_free((void *)rec_ins
->ins
.dst
);
5400 if (rec_ins
->ins
.src
)
5402 for (i
= 0; i
< rec_ins
->ins
.src_count
; ++i
)
5404 heap_free((void *)rec_ins
->ins
.src
[i
].reg
.idx
[0].rel_addr
);
5406 heap_free((void *)rec_ins
->ins
.src
);
5412 static void pop_control_frame(const struct wined3d_shader_instruction
*ins
)
5414 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5415 struct control_frame
*control_frame
;
5417 if (ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5419 struct list
*e
= list_head(&priv
->control_frames
);
5420 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5421 list_remove(&control_frame
->entry
);
5422 heap_free(control_frame
);
5425 else if (ins
->handler_idx
== WINED3DSIH_ENDIF
)
5427 /* Non-ifc ENDIFs were already handled previously. */
5428 struct list
*e
= list_head(&priv
->control_frames
);
5429 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5430 list_remove(&control_frame
->entry
);
5431 heap_free(control_frame
);
5435 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5436 SHADER_HANDLER hw_fct
;
5437 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5438 const struct wined3d_shader
*shader
= ins
->ctx
->shader
;
5439 struct control_frame
*control_frame
;
5440 struct wined3d_string_buffer
*buffer
= ins
->ctx
->buffer
;
5443 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5445 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5446 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5448 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5449 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5451 if(priv
->target_version
>= NV2
)
5453 control_frame
->no
.loop
= priv
->num_loops
++;
5458 /* Don't bother recording when we're in a not used if branch */
5464 if(!priv
->recording
)
5466 list_init(&priv
->record
);
5467 priv
->recording
= TRUE
;
5468 control_frame
->outer_loop
= TRUE
;
5469 get_loop_control_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
, &control_frame
->loop_control
);
5470 return; /* Instruction is handled */
5472 /* Record this loop in the outer loop's recording */
5475 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5477 if(priv
->target_version
>= NV2
)
5479 /* Nothing to do. The control frame is popped after the HW instr handler */
5483 struct list
*e
= list_head(&priv
->control_frames
);
5484 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5485 list_remove(&control_frame
->entry
);
5487 if(control_frame
->outer_loop
)
5489 unsigned int iteration
;
5493 /* Turn off recording before playback */
5494 priv
->recording
= FALSE
;
5496 /* Move the recorded instructions to a separate list and get them out of the private data
5497 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5498 * be recorded again, thus priv->record might be overwritten
5501 list_move_tail(©
, &priv
->record
);
5502 list_init(&priv
->record
);
5504 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5506 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5507 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5508 control_frame
->loop_control
.step
);
5509 aL
= control_frame
->loop_control
.start
;
5513 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5516 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5518 struct recorded_instruction
*rec_ins
;
5519 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5522 shader_addline(buffer
, "#Iteration %u, aL=%d\n", iteration
, aL
);
5526 shader_addline(buffer
, "#Iteration %u\n", iteration
);
5529 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5531 shader_arb_handle_instruction(&rec_ins
->ins
);
5534 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5536 aL
+= control_frame
->loop_control
.step
;
5539 shader_addline(buffer
, "#end loop/rep\n");
5541 free_recorded_instruction(©
);
5542 heap_free(control_frame
);
5543 return; /* Instruction is handled */
5547 /* This is a nested loop. Proceed to the normal recording function */
5548 heap_free(control_frame
);
5555 record_instruction(&priv
->record
, ins
);
5560 if(ins
->handler_idx
== WINED3DSIH_IF
)
5562 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5563 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5564 control_frame
->type
= IF
;
5566 bool_const
= get_bool_const(ins
, shader
, ins
->src
[0].reg
.idx
[0].offset
);
5567 if (ins
->src
[0].modifiers
== WINED3DSPSM_NOT
)
5568 bool_const
= !bool_const
;
5569 if (!priv
->muted
&& !bool_const
)
5571 shader_addline(buffer
, "#if(FALSE){\n");
5573 control_frame
->muting
= TRUE
;
5575 else shader_addline(buffer
, "#if(TRUE) {\n");
5577 return; /* Instruction is handled */
5579 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5581 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5582 control_frame
= heap_alloc_zero(sizeof(*control_frame
));
5583 control_frame
->type
= IFC
;
5584 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5585 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5587 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5589 struct list
*e
= list_head(&priv
->control_frames
);
5590 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5592 if(control_frame
->type
== IF
)
5594 shader_addline(buffer
, "#} else {\n");
5595 if(!priv
->muted
&& !control_frame
->muting
)
5598 control_frame
->muting
= TRUE
;
5600 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5601 return; /* Instruction is handled. */
5603 /* In case of an ifc, generate a HW shader instruction */
5604 if (control_frame
->type
!= IFC
)
5605 ERR("Control frame does not match.\n");
5607 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5609 struct list
*e
= list_head(&priv
->control_frames
);
5610 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5612 if(control_frame
->type
== IF
)
5614 shader_addline(buffer
, "#} endif\n");
5615 if(control_frame
->muting
) priv
->muted
= FALSE
;
5616 list_remove(&control_frame
->entry
);
5617 heap_free(control_frame
);
5618 return; /* Instruction is handled */
5620 /* In case of an ifc, generate a HW shader instruction */
5621 if (control_frame
->type
!= IFC
)
5622 ERR("Control frame does not match.\n");
5627 pop_control_frame(ins
);
5631 /* Select handler */
5632 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5634 /* Unhandled opcode */
5637 FIXME("Backend can't handle opcode %s.\n", debug_d3dshaderinstructionhandler(ins
->handler_idx
));
5642 pop_control_frame(ins
);
5644 shader_arb_add_instruction_modifiers(ins
);
5647 static BOOL
shader_arb_has_ffp_proj_control(void *shader_priv
)
5649 struct shader_arb_priv
*priv
= shader_priv
;
5651 return priv
->ffp_proj_control
;
5654 static void shader_arb_precompile(void *shader_priv
, struct wined3d_shader
*shader
) {}
5656 static uint64_t shader_arb_shader_compile(struct wined3d_context
*context
, const struct wined3d_shader_desc
*shader_desc
,
5657 enum wined3d_shader_type shader_type
)
5659 ERR("Not implemented.\n");
5663 const struct wined3d_shader_backend_ops arb_program_shader_backend
=
5665 shader_arb_handle_instruction
,
5666 shader_arb_precompile
,
5668 shader_arb_select_compute
,
5670 shader_arb_update_float_vertex_constants
,
5671 shader_arb_update_float_pixel_constants
,
5672 shader_arb_load_constants
,
5676 shader_arb_allocate_context_data
,
5677 shader_arb_free_context_data
,
5678 shader_arb_init_context_state
,
5679 shader_arb_get_caps
,
5680 shader_arb_color_fixup_supported
,
5681 shader_arb_has_ffp_proj_control
,
5682 shader_arb_shader_compile
,
5685 /* ARB_fragment_program fixed function pipeline replacement definitions */
5686 #define ARB_FFP_CONST_TFACTOR 0
5687 #define ARB_FFP_CONST_COLOR_KEY_LOW ((ARB_FFP_CONST_TFACTOR) + 1)
5688 #define ARB_FFP_CONST_COLOR_KEY_HIGH ((ARB_FFP_CONST_COLOR_KEY_LOW) + 1)
5689 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_COLOR_KEY_HIGH) + 1)
5690 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5691 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5692 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5694 struct arbfp_ffp_desc
5696 struct ffp_frag_desc parent
;
5700 /* Context activation is done by the caller. */
5701 static void arbfp_enable(const struct wined3d_context
*context
, BOOL enable
)
5703 const struct wined3d_gl_info
*gl_info
= wined3d_context_gl_const(context
)->gl_info
;
5707 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5708 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5712 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5713 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5717 static void *arbfp_alloc(const struct wined3d_shader_backend_ops
*shader_backend
, void *shader_priv
)
5719 struct shader_arb_priv
*priv
;
5721 /* Share private data between the shader backend and the pipeline
5722 * replacement, if both are the arb implementation. This is needed to
5723 * figure out whether ARBfp should be disabled if no pixel shader is bound
5725 if (shader_backend
== &arb_program_shader_backend
)
5727 else if (!(priv
= heap_alloc_zero(sizeof(*priv
))))
5730 wine_rb_init(&priv
->fragment_shaders
, wined3d_ffp_frag_program_key_compare
);
5731 priv
->use_arbfp_fixed_func
= TRUE
;
5736 /* Context activation is done by the caller. */
5737 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *param
)
5739 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5740 struct wined3d_context_gl
*context_gl
= param
;
5741 const struct wined3d_gl_info
*gl_info
;
5743 gl_info
= context_gl
->gl_info
;
5744 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5745 checkGLcall("delete ffp program");
5746 heap_free(entry_arb
);
5749 /* Context activation is done by the caller. */
5750 static void arbfp_free(struct wined3d_device
*device
, struct wined3d_context
*context
)
5752 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5753 struct shader_arb_priv
*priv
= device
->fragment_priv
;
5755 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, context_gl
);
5756 priv
->use_arbfp_fixed_func
= FALSE
;
5758 if (device
->shader_backend
!= &arb_program_shader_backend
)
5759 heap_free(device
->fragment_priv
);
5762 static void arbfp_get_caps(const struct wined3d_adapter
*adapter
, struct fragment_caps
*caps
)
5764 const struct wined3d_gl_info
*gl_info
= &wined3d_adapter_gl_const(adapter
)->gl_info
;
5766 caps
->wined3d_caps
= WINED3D_FRAGMENT_CAP_PROJ_CONTROL
5767 | WINED3D_FRAGMENT_CAP_SRGB_WRITE
5768 | WINED3D_FRAGMENT_CAP_COLOR_KEY
;
5769 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5770 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5771 WINED3DTEXOPCAPS_SELECTARG1
|
5772 WINED3DTEXOPCAPS_SELECTARG2
|
5773 WINED3DTEXOPCAPS_MODULATE4X
|
5774 WINED3DTEXOPCAPS_MODULATE2X
|
5775 WINED3DTEXOPCAPS_MODULATE
|
5776 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5777 WINED3DTEXOPCAPS_ADDSIGNED
|
5778 WINED3DTEXOPCAPS_ADD
|
5779 WINED3DTEXOPCAPS_SUBTRACT
|
5780 WINED3DTEXOPCAPS_ADDSMOOTH
|
5781 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5782 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5783 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5784 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5785 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5786 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5787 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5788 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5789 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5790 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5791 WINED3DTEXOPCAPS_MULTIPLYADD
|
5792 WINED3DTEXOPCAPS_LERP
|
5793 WINED3DTEXOPCAPS_BUMPENVMAP
|
5794 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5796 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5798 caps
->MaxTextureBlendStages
= WINED3D_MAX_FFP_TEXTURES
;
5799 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.samplers
[WINED3D_SHADER_TYPE_PIXEL
], WINED3D_MAX_FFP_TEXTURES
);
5802 static unsigned int arbfp_get_emul_mask(const struct wined3d_adapter
*adapter
)
5804 return GL_EXT_EMUL_ARB_MULTITEXTURE
| GL_EXT_EMUL_EXT_FOG_COORD
;
5807 static void state_texfactor_arbfp(struct wined3d_context
*context
,
5808 const struct wined3d_state
*state
, DWORD state_id
)
5810 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5811 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5812 struct wined3d_device
*device
= context
->device
;
5813 struct wined3d_color color
;
5815 if (device
->shader_backend
== &arb_program_shader_backend
)
5817 struct shader_arb_priv
*priv
;
5819 /* Don't load the parameter if we're using an arbfp pixel shader,
5820 * otherwise we'll overwrite application provided constants. */
5824 priv
= device
->shader_priv
;
5825 priv
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5826 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5829 wined3d_color_from_d3dcolor(&color
, state
->render_states
[WINED3D_RS_TEXTUREFACTOR
]);
5830 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, &color
.r
));
5831 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, &color.r)");
5834 static void state_tss_constant_arbfp(struct wined3d_context
*context
,
5835 const struct wined3d_state
*state
, DWORD state_id
)
5837 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5838 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5839 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5840 struct wined3d_device
*device
= context
->device
;
5841 struct wined3d_color color
;
5843 if (device
->shader_backend
== &arb_program_shader_backend
)
5845 struct shader_arb_priv
*priv
;
5847 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5848 * application provided constants.
5853 priv
= device
->shader_priv
;
5854 priv
->pshader_const_dirty
[ARB_FFP_CONST_CONSTANT(stage
)] = 1;
5855 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_CONSTANT(stage
) + 1);
5858 wined3d_color_from_d3dcolor(&color
, state
->texture_states
[stage
][WINED3D_TSS_CONSTANT
]);
5859 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_CONSTANT(stage
), &color
.r
));
5860 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_CONSTANT(stage), &color.r)");
5863 static void state_arb_specularenable(struct wined3d_context
*context
,
5864 const struct wined3d_state
*state
, DWORD state_id
)
5866 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5867 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5868 struct wined3d_device
*device
= context
->device
;
5871 if (device
->shader_backend
== &arb_program_shader_backend
)
5873 struct shader_arb_priv
*priv
;
5875 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5876 * application provided constants.
5881 priv
= device
->shader_priv
;
5882 priv
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5883 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5886 if (state
->render_states
[WINED3D_RS_SPECULARENABLE
])
5888 /* The specular color has no alpha */
5889 col
[0] = 1.0f
; col
[1] = 1.0f
;
5890 col
[2] = 1.0f
; col
[3] = 0.0f
;
5892 col
[0] = 0.0f
; col
[1] = 0.0f
;
5893 col
[2] = 0.0f
; col
[3] = 0.0f
;
5895 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5896 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5899 static void set_bumpmat_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5901 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5902 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5903 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5904 struct wined3d_device
*device
= context
->device
;
5907 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5909 if (device
->shader_backend
== &arb_program_shader_backend
)
5911 struct shader_arb_priv
*priv
= device
->shader_priv
;
5913 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants. */
5917 priv
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5918 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5921 mat
[0][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT00
]);
5922 mat
[0][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT01
]);
5923 mat
[1][0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT10
]);
5924 mat
[1][1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_MAT11
]);
5926 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5927 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5930 static void tex_bumpenvlum_arbfp(struct wined3d_context
*context
,
5931 const struct wined3d_state
*state
, DWORD state_id
)
5933 DWORD stage
= (state_id
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5934 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5935 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5936 struct wined3d_device
*device
= context
->device
;
5939 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_BUMP_ENV
;
5941 if (device
->shader_backend
== &arb_program_shader_backend
)
5943 struct shader_arb_priv
*priv
= device
->shader_priv
;
5945 /* Exit now, don't set the luminance below, otherwise we may overwrite pixel shader constants. */
5949 priv
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5950 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5953 param
[0] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LSCALE
]);
5954 param
[1] = *((float *)&state
->texture_states
[stage
][WINED3D_TSS_BUMPENV_LOFFSET
]);
5958 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5959 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5962 static void alpha_test_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5964 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5965 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5969 TRACE("context %p, state %p, state_id %#lx.\n", context
, state
, state_id
);
5971 if (state
->render_states
[WINED3D_RS_ALPHATESTENABLE
])
5973 gl_info
->gl_ops
.gl
.p_glEnable(GL_ALPHA_TEST
);
5974 checkGLcall("glEnable GL_ALPHA_TEST");
5978 gl_info
->gl_ops
.gl
.p_glDisable(GL_ALPHA_TEST
);
5979 checkGLcall("glDisable GL_ALPHA_TEST");
5983 ref
= wined3d_alpha_ref(state
);
5984 glParm
= wined3d_gl_compare_func(state
->render_states
[WINED3D_RS_ALPHAFUNC
]);
5988 gl_info
->gl_ops
.gl
.p_glAlphaFunc(glParm
, ref
);
5989 checkGLcall("glAlphaFunc");
5993 static void color_key_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
5995 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
5996 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
5997 const struct wined3d_texture
*texture
= state
->textures
[0];
5998 struct wined3d_device
*device
= context
->device
;
5999 struct wined3d_color float_key
[2];
6004 if (device
->shader_backend
== &arb_program_shader_backend
)
6006 struct shader_arb_priv
*priv
;
6008 /* Don't load the parameter if we're using an arbfp pixel shader,
6009 * otherwise we'll overwrite application provided constants. */
6013 priv
= device
->shader_priv
;
6014 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_LOW
] = 1;
6015 priv
->pshader_const_dirty
[ARB_FFP_CONST_COLOR_KEY_HIGH
] = 1;
6016 priv
->highest_dirty_ps_const
= max(priv
->highest_dirty_ps_const
, ARB_FFP_CONST_COLOR_KEY_HIGH
+ 1);
6019 wined3d_format_get_float_color_key(texture
->resource
.format
, &texture
->async
.src_blt_color_key
, float_key
);
6021 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_LOW
, &float_key
[0].r
));
6022 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_LOW, &float_key[0].r)");
6023 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_COLOR_KEY_HIGH
, &float_key
[1].r
));
6024 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_COLOR_KEY_HIGH, &float_key[1].r)");
6027 static const char *get_argreg(struct wined3d_string_buffer
*buffer
, unsigned int argnum
, unsigned int stage
, DWORD arg
)
6031 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
6033 switch(arg
& WINED3DTA_SELECTMASK
) {
6034 case WINED3DTA_DIFFUSE
:
6035 ret
= "fragment.color.primary"; break;
6037 case WINED3DTA_CURRENT
:
6041 case WINED3DTA_TEXTURE
:
6043 case 0: ret
= "tex0"; break;
6044 case 1: ret
= "tex1"; break;
6045 case 2: ret
= "tex2"; break;
6046 case 3: ret
= "tex3"; break;
6047 case 4: ret
= "tex4"; break;
6048 case 5: ret
= "tex5"; break;
6049 case 6: ret
= "tex6"; break;
6050 case 7: ret
= "tex7"; break;
6051 default: ret
= "unknown texture";
6055 case WINED3DTA_TFACTOR
:
6056 ret
= "tfactor"; break;
6058 case WINED3DTA_SPECULAR
:
6059 ret
= "fragment.color.secondary"; break;
6061 case WINED3DTA_TEMP
:
6062 ret
= "tempreg"; break;
6064 case WINED3DTA_CONSTANT
:
6066 case 0: ret
= "const0"; break;
6067 case 1: ret
= "const1"; break;
6068 case 2: ret
= "const2"; break;
6069 case 3: ret
= "const3"; break;
6070 case 4: ret
= "const4"; break;
6071 case 5: ret
= "const5"; break;
6072 case 6: ret
= "const6"; break;
6073 case 7: ret
= "const7"; break;
6074 default: ret
= "unknown constant";
6082 if(arg
& WINED3DTA_COMPLEMENT
) {
6083 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
6084 if(argnum
== 0) ret
= "arg0";
6085 if(argnum
== 1) ret
= "arg1";
6086 if(argnum
== 2) ret
= "arg2";
6088 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
6089 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
6090 if(argnum
== 0) ret
= "arg0";
6091 if(argnum
== 1) ret
= "arg1";
6092 if(argnum
== 2) ret
= "arg2";
6097 static void gen_ffp_instr(struct wined3d_string_buffer
*buffer
, unsigned int stage
, BOOL color
,
6098 BOOL alpha
, BOOL tmp_dst
, unsigned int op
, unsigned int dw_arg0
, unsigned int dw_arg1
, unsigned int dw_arg2
)
6100 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
6101 unsigned int mul
= 1;
6110 dstreg
= tmp_dst
? "tempreg" : "ret";
6112 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
6113 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
6114 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
6118 case WINED3D_TOP_DISABLE
:
6121 case WINED3D_TOP_SELECT_ARG2
:
6124 case WINED3D_TOP_SELECT_ARG1
:
6125 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
6128 case WINED3D_TOP_MODULATE_4X
:
6131 case WINED3D_TOP_MODULATE_2X
:
6134 case WINED3D_TOP_MODULATE
:
6135 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6138 case WINED3D_TOP_ADD_SIGNED_2X
:
6141 case WINED3D_TOP_ADD_SIGNED
:
6142 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6145 case WINED3D_TOP_ADD
:
6146 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6149 case WINED3D_TOP_SUBTRACT
:
6150 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
6153 case WINED3D_TOP_ADD_SMOOTH
:
6154 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
6155 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6158 case WINED3D_TOP_BLEND_CURRENT_ALPHA
:
6159 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
6160 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6162 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6163 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
6164 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6166 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6167 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6168 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6170 case WINED3D_TOP_BLEND_DIFFUSE_ALPHA
:
6171 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
6172 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6175 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6176 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
6177 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
6178 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6181 /* D3DTOP_PREMODULATE ???? */
6183 case WINED3D_TOP_MODULATE_INVALPHA_ADD_COLOR
:
6184 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
6185 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
6187 case WINED3D_TOP_MODULATE_ALPHA_ADD_COLOR
:
6188 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6190 case WINED3D_TOP_MODULATE_INVCOLOR_ADD_ALPHA
:
6191 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
6192 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
6194 case WINED3D_TOP_MODULATE_COLOR_ADD_ALPHA
:
6195 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
6198 case WINED3D_TOP_DOTPRODUCT3
:
6200 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
6201 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
6202 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
6205 case WINED3D_TOP_MULTIPLY_ADD
:
6206 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
6209 case WINED3D_TOP_LERP
:
6210 /* The msdn is not quite right here */
6211 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
6214 case WINED3D_TOP_BUMPENVMAP
:
6215 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6216 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
6220 FIXME("Unhandled texture op %08x\n", op
);
6224 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", dstreg
, dstmask
, dstreg
);
6226 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", dstreg
, dstmask
, dstreg
);
6229 static const char *arbfp_texture_target(enum wined3d_gl_resource_type type
)
6233 case WINED3D_GL_RES_TYPE_TEX_1D
:
6235 case WINED3D_GL_RES_TYPE_TEX_2D
:
6237 case WINED3D_GL_RES_TYPE_TEX_3D
:
6239 case WINED3D_GL_RES_TYPE_TEX_CUBE
:
6241 case WINED3D_GL_RES_TYPE_TEX_RECT
:
6244 return "unexpected_resource_type";
6248 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, const struct wined3d_gl_info
*gl_info
)
6250 BYTE tex_read
= 0, bump_used
= 0, luminance_used
= 0, constant_used
= 0;
6251 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
6252 unsigned int stage
, lowest_disabled_stage
;
6253 struct wined3d_string_buffer buffer
;
6254 struct color_fixup_masks masks
;
6255 BOOL custom_linear_fog
= FALSE
;
6256 const char *textype
, *instr
;
6257 DWORD arg0
, arg1
, arg2
;
6258 char colorcor_dst
[8];
6262 if (!string_buffer_init(&buffer
))
6264 ERR("Failed to initialize shader buffer.\n");
6268 shader_addline(&buffer
, "!!ARBfp1.0\n");
6270 if (settings
->color_key_enabled
)
6272 shader_addline(&buffer
, "PARAM color_key_low = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_LOW
);
6273 shader_addline(&buffer
, "PARAM color_key_high = program.env[%u];\n", ARB_FFP_CONST_COLOR_KEY_HIGH
);
6277 /* Find out which textures are read */
6278 for (stage
= 0; stage
< WINED3D_MAX_FFP_TEXTURES
; ++stage
)
6280 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6283 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
6284 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
6285 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
6287 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6288 tex_read
|= 1u << stage
;
6289 if (settings
->op
[stage
].tmp_dst
)
6290 tempreg_used
= TRUE
;
6291 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6292 tempreg_used
= TRUE
;
6293 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6294 tfactor_used
= TRUE
;
6295 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6296 constant_used
|= 1u << stage
;
6298 switch (settings
->op
[stage
].cop
)
6300 case WINED3D_TOP_BUMPENVMAP_LUMINANCE
:
6301 luminance_used
|= 1u << stage
;
6303 case WINED3D_TOP_BUMPENVMAP
:
6304 bump_used
|= 1u << stage
;
6306 case WINED3D_TOP_BLEND_TEXTURE_ALPHA
:
6307 case WINED3D_TOP_BLEND_TEXTURE_ALPHA_PM
:
6308 tex_read
|= 1u << stage
;
6311 case WINED3D_TOP_BLEND_FACTOR_ALPHA
:
6312 tfactor_used
= TRUE
;
6319 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6322 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
6323 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
6324 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
6326 if (arg0
== WINED3DTA_TEXTURE
|| arg1
== WINED3DTA_TEXTURE
|| arg2
== WINED3DTA_TEXTURE
)
6327 tex_read
|= 1u << stage
;
6328 if (arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
)
6329 tempreg_used
= TRUE
;
6330 if (arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
)
6331 tfactor_used
= TRUE
;
6332 if (arg0
== WINED3DTA_CONSTANT
|| arg1
== WINED3DTA_CONSTANT
|| arg2
== WINED3DTA_CONSTANT
)
6333 constant_used
|= 1u << stage
;
6335 lowest_disabled_stage
= stage
;
6337 switch (settings
->fog
)
6339 case WINED3D_FFP_PS_FOG_OFF
: break;
6340 case WINED3D_FFP_PS_FOG_LINEAR
:
6341 if (gl_info
->quirks
& WINED3D_QUIRK_BROKEN_ARB_FOG
)
6343 custom_linear_fog
= TRUE
;
6346 shader_addline(&buffer
, "OPTION ARB_fog_linear;\n");
6349 case WINED3D_FFP_PS_FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
6350 case WINED3D_FFP_PS_FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
6351 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
6354 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
6355 shader_addline(&buffer
, "TEMP TMP;\n");
6356 shader_addline(&buffer
, "TEMP ret;\n");
6357 if (tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
6358 shader_addline(&buffer
, "TEMP arg0;\n");
6359 shader_addline(&buffer
, "TEMP arg1;\n");
6360 shader_addline(&buffer
, "TEMP arg2;\n");
6361 for (stage
= 0; stage
< WINED3D_MAX_FFP_TEXTURES
; ++stage
)
6363 if (constant_used
& (1u << stage
))
6364 shader_addline(&buffer
, "PARAM const%u = program.env[%u];\n", stage
, ARB_FFP_CONST_CONSTANT(stage
));
6366 if (!(tex_read
& (1u << stage
)))
6369 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
6371 if (!(bump_used
& (1u << stage
)))
6373 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
6375 if (!(luminance_used
& (1u << stage
)))
6377 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
6380 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
6381 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
6383 if (settings
->sRGB_write
)
6385 shader_addline(&buffer
, "PARAM srgb_consts0 = ");
6386 shader_arb_append_imm_vec4(&buffer
, &wined3d_srgb_const
[0].x
);
6387 shader_addline(&buffer
, ";\n");
6388 shader_addline(&buffer
, "PARAM srgb_consts1 = ");
6389 shader_arb_append_imm_vec4(&buffer
, &wined3d_srgb_const
[1].x
);
6390 shader_addline(&buffer
, ";\n");
6393 if (lowest_disabled_stage
< 7 && settings
->emul_clipplanes
)
6394 shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
6396 if (tempreg_used
|| settings
->sRGB_write
)
6397 shader_addline(&buffer
, "MOV tempreg, 0.0;\n");
6399 /* Generate texture sampling instructions */
6400 for (stage
= 0; stage
< WINED3D_MAX_FFP_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3D_TOP_DISABLE
; ++stage
)
6402 if (!(tex_read
& (1u << stage
)))
6405 textype
= arbfp_texture_target(settings
->op
[stage
].tex_type
);
6407 if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_NONE
)
6411 else if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT4
6412 || settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT3
)
6418 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6423 && (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP
6424 || settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
))
6426 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6427 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6428 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6429 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6431 /* With projective textures, texbem only divides the static
6432 * texture coordinate, not the displacement, so multiply the
6433 * displacement with the dividing parameter before passing it to
6435 if (settings
->op
[stage
].projected
!= WINED3D_PROJECTION_NONE
)
6437 if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT4
)
6439 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6440 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n",
6445 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6446 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n",
6452 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6455 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6456 instr
, stage
, stage
, textype
);
6457 if (settings
->op
[stage
- 1].cop
== WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6459 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6460 stage
- 1, stage
- 1, stage
- 1);
6461 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6464 else if (settings
->op
[stage
].projected
== WINED3D_PROJECTION_COUNT3
)
6466 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6467 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6468 shader_addline(&buffer
, "%s tex%u, ret, texture[%u], %s;\n",
6469 instr
, stage
, stage
, textype
);
6473 shader_addline(&buffer
, "%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6474 instr
, stage
, stage
, stage
, textype
);
6477 sprintf(colorcor_dst
, "tex%u", stage
);
6478 masks
= calc_color_correction(settings
->op
[stage
].color_fixup
, WINED3DSP_WRITEMASK_ALL
);
6479 gen_color_correction(&buffer
, colorcor_dst
, colorcor_dst
, "const.x", "const.y",
6480 settings
->op
[stage
].color_fixup
, masks
);
6483 if (settings
->color_key_enabled
)
6485 shader_addline(&buffer
, "SLT TMP, tex0, color_key_low;\n"); /* below low key */
6486 shader_addline(&buffer
, "SGE ret, tex0, color_key_high;\n"); /* above high key */
6487 shader_addline(&buffer
, "ADD TMP, TMP, ret;\n"); /* or */
6488 shader_addline(&buffer
, "DP4 TMP.b, TMP, TMP;\n"); /* on any channel */
6489 shader_addline(&buffer
, "SGE TMP, -TMP.b, 0.0;\n"); /* logical not */
6490 shader_addline(&buffer
, "KIL -TMP;\n"); /* discard if true */
6493 shader_addline(&buffer
, "MOV ret, fragment.color.primary;\n");
6495 /* Generate the main shader */
6496 for (stage
= 0; stage
< WINED3D_MAX_FFP_TEXTURES
; ++stage
)
6498 if (settings
->op
[stage
].cop
== WINED3D_TOP_DISABLE
)
6501 if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6502 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6503 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6504 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG1
6505 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6506 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6507 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6508 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG1
)
6509 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6510 else if (settings
->op
[stage
].cop
== WINED3D_TOP_SELECT_ARG2
6511 && settings
->op
[stage
].aop
== WINED3D_TOP_SELECT_ARG2
)
6512 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6514 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
6515 && settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
6516 && settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
6517 && settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6519 if (settings
->op
[stage
].aop
== WINED3D_TOP_DISABLE
)
6521 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].tmp_dst
,
6522 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6523 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6527 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].tmp_dst
,
6528 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6529 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6531 else if (settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP
6532 && settings
->op
[stage
].cop
!= WINED3D_TOP_BUMPENVMAP_LUMINANCE
)
6534 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].tmp_dst
,
6535 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6536 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6537 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].tmp_dst
,
6538 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6539 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6543 if (settings
->sRGB_write
|| custom_linear_fog
)
6545 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, ret;\n");
6546 if (settings
->sRGB_write
)
6547 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6548 if (custom_linear_fog
)
6549 arbfp_add_linear_fog(&buffer
, "ret", "arg0");
6550 shader_addline(&buffer
, "MOV result.color, ret;\n");
6554 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, ret;\n");
6558 shader_addline(&buffer
, "END\n");
6560 /* Generate the shader */
6561 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6562 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6563 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
6565 string_buffer_free(&buffer
);
6569 static void fragment_prog_arbfp(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6571 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
6572 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
6573 const struct wined3d_device
*device
= context
->device
;
6574 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6575 BOOL use_pshader
= use_ps(state
);
6576 struct ffp_frag_settings settings
;
6577 const struct arbfp_ffp_desc
*desc
;
6580 TRACE("context %p, state %p, state_id %#lx.\n", context
, state
, state_id
);
6584 /* Find or create a shader implementing the fixed function pipeline
6585 * settings, then activate it. */
6586 wined3d_ffp_get_fs_settings(context
, state
, &settings
, FALSE
);
6587 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6590 struct arbfp_ffp_desc
*new_desc
;
6592 if (!(new_desc
= heap_alloc(sizeof(*new_desc
))))
6594 ERR("Out of memory\n");
6598 new_desc
->parent
.settings
= settings
;
6599 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, gl_info
);
6600 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6601 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6605 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active (however, note the
6606 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6609 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6610 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6612 if (device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
)
6614 /* Reload fixed function constants since they collide with the
6615 * pixel shader constants. */
6616 for (i
= 0; i
< WINED3D_MAX_FFP_TEXTURES
; ++i
)
6618 set_bumpmat_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_BUMPENV_MAT00
));
6619 state_tss_constant_arbfp(context
, state
, STATE_TEXTURESTAGE(i
, WINED3D_TSS_CONSTANT
));
6621 state_texfactor_arbfp(context
, state
, STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
));
6622 state_arb_specularenable(context
, state
, STATE_RENDER(WINED3D_RS_SPECULARENABLE
));
6623 color_key_arbfp(context
, state
, STATE_COLOR_KEY
);
6625 context
->last_was_pshader
= FALSE
;
6627 else if (!context
->last_was_pshader
)
6629 if (device
->shader_backend
== &arb_program_shader_backend
)
6630 context
->constant_update_mask
|= WINED3D_SHADER_CONST_PS_F
;
6631 context
->last_was_pshader
= TRUE
;
6634 context
->shader_update_mask
|= 1u << WINED3D_SHADER_TYPE_PIXEL
;
6637 /* We can't link the fog states to the fragment state directly since the
6638 * vertex pipeline links them to FOGENABLE. A different linking in different
6639 * pipeline parts can't be expressed in the combined state table, so we need
6640 * to handle that with a forwarding function. The other invisible side effect
6641 * is that changing the fog start and fog end (which links to FOGENABLE in
6642 * vertex) results in the fragment_prog_arbfp function being called because
6643 * FOGENABLE is dirty, which calls this function here. */
6644 static void state_arbfp_fog(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6646 enum fogsource new_source
;
6647 DWORD fogstart
= state
->render_states
[WINED3D_RS_FOGSTART
];
6648 DWORD fogend
= state
->render_states
[WINED3D_RS_FOGEND
];
6650 TRACE("context %p, state %p, state_id %#lx.\n", context
, state
, state_id
);
6652 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6653 fragment_prog_arbfp(context
, state
, state_id
);
6655 if (!state
->render_states
[WINED3D_RS_FOGENABLE
])
6658 if (state
->render_states
[WINED3D_RS_FOGTABLEMODE
] == WINED3D_FOG_NONE
)
6662 new_source
= FOGSOURCE_VS
;
6666 if (state
->render_states
[WINED3D_RS_FOGVERTEXMODE
] == WINED3D_FOG_NONE
|| context
->last_was_rhw
)
6667 new_source
= FOGSOURCE_COORD
;
6669 new_source
= FOGSOURCE_FFP
;
6674 new_source
= FOGSOURCE_FFP
;
6677 if (new_source
!= context
->fog_source
|| fogstart
== fogend
)
6679 context
->fog_source
= new_source
;
6680 state_fogstartend(context
, state
, STATE_RENDER(WINED3D_RS_FOGSTART
));
6684 static void textransform(struct wined3d_context
*context
, const struct wined3d_state
*state
, DWORD state_id
)
6686 if (!isStateDirty(context
, STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
)))
6687 fragment_prog_arbfp(context
, state
, state_id
);
6690 static const struct wined3d_state_entry_template arbfp_fragmentstate_template
[] =
6692 {STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), { STATE_RENDER(WINED3D_RS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6693 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6694 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6695 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6696 {STATE_TEXTURESTAGE(0, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6697 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6698 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6699 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6700 {STATE_TEXTURESTAGE(0, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6701 {STATE_TEXTURESTAGE(0, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6702 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6703 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6704 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6705 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6706 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6707 {STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6708 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6709 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6710 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6711 {STATE_TEXTURESTAGE(1, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6712 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6713 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6714 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6715 {STATE_TEXTURESTAGE(1, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6716 {STATE_TEXTURESTAGE(1, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6717 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6718 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6719 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6720 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6721 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6722 {STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6723 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6724 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6725 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6726 {STATE_TEXTURESTAGE(2, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6727 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6728 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6729 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6730 {STATE_TEXTURESTAGE(2, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6731 {STATE_TEXTURESTAGE(2, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6732 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6733 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6734 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6735 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6736 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6737 {STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6738 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6739 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6740 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6741 {STATE_TEXTURESTAGE(3, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6742 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6743 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6744 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6745 {STATE_TEXTURESTAGE(3, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6746 {STATE_TEXTURESTAGE(3, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6747 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6748 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6749 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6750 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6751 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6752 {STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6753 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6754 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6755 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6756 {STATE_TEXTURESTAGE(4, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6757 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6758 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6759 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6760 {STATE_TEXTURESTAGE(4, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6761 {STATE_TEXTURESTAGE(4, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6762 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6763 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6764 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6765 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6766 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6767 {STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6768 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6769 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6770 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6771 {STATE_TEXTURESTAGE(5, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6772 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6773 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6774 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6775 {STATE_TEXTURESTAGE(5, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6776 {STATE_TEXTURESTAGE(5, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6777 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6778 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6779 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6780 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6781 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6782 {STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6783 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6784 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6785 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6786 {STATE_TEXTURESTAGE(6, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6787 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6788 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6789 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6790 {STATE_TEXTURESTAGE(6, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6791 {STATE_TEXTURESTAGE(6, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6792 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6793 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6794 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6795 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6796 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6797 {STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6798 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6799 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6800 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6801 {STATE_TEXTURESTAGE(7, WINED3D_TSS_COLOR_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6802 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_OP
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6803 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG1
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6804 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG2
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6805 {STATE_TEXTURESTAGE(7, WINED3D_TSS_ALPHA_ARG0
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6806 {STATE_TEXTURESTAGE(7, WINED3D_TSS_RESULT_ARG
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6807 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6808 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT01
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6809 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT10
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6810 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT11
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_MAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6811 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6812 {STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LOFFSET
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_BUMPENV_LSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6813 {STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6814 {STATE_RENDER(WINED3D_RS_ALPHAFUNC
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6815 {STATE_RENDER(WINED3D_RS_ALPHAREF
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6816 {STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), { STATE_RENDER(WINED3D_RS_ALPHATESTENABLE
), alpha_test_arbfp
}, WINED3D_GL_EXT_NONE
},
6817 {STATE_RENDER(WINED3D_RS_COLORKEYENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6818 {STATE_COLOR_KEY
, { STATE_COLOR_KEY
, color_key_arbfp
}, WINED3D_GL_EXT_NONE
},
6819 {STATE_RENDER(WINED3D_RS_FOGENABLE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6820 {STATE_RENDER(WINED3D_RS_FOGTABLEMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6821 {STATE_RENDER(WINED3D_RS_FOGVERTEXMODE
), { STATE_RENDER(WINED3D_RS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6822 {STATE_RENDER(WINED3D_RS_FOGSTART
), { STATE_RENDER(WINED3D_RS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6823 {STATE_RENDER(WINED3D_RS_FOGEND
), { STATE_RENDER(WINED3D_RS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6824 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), state_srgbwrite
}, ARB_FRAMEBUFFER_SRGB
},
6825 {STATE_RENDER(WINED3D_RS_SRGBWRITEENABLE
), { STATE_SHADER(WINED3D_SHADER_TYPE_PIXEL
), NULL
}, WINED3D_GL_EXT_NONE
},
6826 {STATE_RENDER(WINED3D_RS_FOGCOLOR
), { STATE_RENDER(WINED3D_RS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6827 {STATE_RENDER(WINED3D_RS_FOGDENSITY
), { STATE_RENDER(WINED3D_RS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6828 {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(0,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6829 {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(1,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6830 {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(2,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6831 {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(3,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6832 {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(4,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6833 {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(5,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6834 {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(6,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6835 {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), {STATE_TEXTURESTAGE(7,WINED3D_TSS_TEXTURE_TRANSFORM_FLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6836 {STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(0, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6837 {STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(1, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6838 {STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(2, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6839 {STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(3, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6840 {STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(4, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6841 {STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(5, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6842 {STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(6, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6843 {STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), { STATE_TEXTURESTAGE(7, WINED3D_TSS_CONSTANT
), state_tss_constant_arbfp
}, WINED3D_GL_EXT_NONE
},
6844 {STATE_RENDER(WINED3D_RS_SPECULARENABLE
), { STATE_RENDER(WINED3D_RS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6845 {STATE_RENDER(WINED3D_RS_SHADEMODE
), { STATE_RENDER(WINED3D_RS_SHADEMODE
), state_shademode
}, WINED3D_GL_EXT_NONE
},
6846 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6849 static BOOL
arbfp_alloc_context_data(struct wined3d_context
*context
)
6854 static void arbfp_free_context_data(struct wined3d_context
*context
)
6858 const struct wined3d_fragment_pipe_ops arbfp_fragment_pipeline
=
6862 arbfp_get_emul_mask
,
6865 arbfp_alloc_context_data
,
6866 arbfp_free_context_data
,
6867 shader_arb_color_fixup_supported
,
6868 arbfp_fragmentstate_template
,
6871 struct arbfp_blit_type
6873 enum complex_fixup fixup
: 4;
6874 enum wined3d_gl_resource_type res_type
: 3;
6875 DWORD use_color_key
: 1;
6879 struct arbfp_blit_desc
6882 struct arbfp_blit_type type
;
6883 struct wine_rb_entry entry
;
6886 #define ARBFP_BLIT_PARAM_SIZE 0
6887 #define ARBFP_BLIT_PARAM_COLOR_KEY_LOW 1
6888 #define ARBFP_BLIT_PARAM_COLOR_KEY_HIGH 2
6890 struct wined3d_arbfp_blitter
6892 struct wined3d_blitter blitter
;
6893 struct wine_rb_tree shaders
;
6894 GLuint palette_texture
;
6897 static int arbfp_blit_type_compare(const void *key
, const struct wine_rb_entry
*entry
)
6899 const struct arbfp_blit_type
*ka
= key
;
6900 const struct arbfp_blit_type
*kb
= &WINE_RB_ENTRY_VALUE(entry
, const struct arbfp_blit_desc
, entry
)->type
;
6902 return memcmp(ka
, kb
, sizeof(*ka
));
6905 /* Context activation is done by the caller. */
6906 static void arbfp_free_blit_shader(struct wine_rb_entry
*entry
, void *ctx
)
6908 struct arbfp_blit_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
6909 const struct wined3d_gl_info
*gl_info
;
6910 struct wined3d_context_gl
*context_gl
;
6913 gl_info
= context_gl
->gl_info
;
6915 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
6916 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
6917 heap_free(entry_arb
);
6920 /* Context activation is done by the caller. */
6921 static void arbfp_blitter_destroy(struct wined3d_blitter
*blitter
, struct wined3d_context
*context
)
6923 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
6924 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
6925 struct wined3d_arbfp_blitter
*arbfp_blitter
;
6926 struct wined3d_blitter
*next
;
6928 if ((next
= blitter
->next
))
6929 next
->ops
->blitter_destroy(next
, &context_gl
->c
);
6931 arbfp_blitter
= CONTAINING_RECORD(blitter
, struct wined3d_arbfp_blitter
, blitter
);
6933 wine_rb_destroy(&arbfp_blitter
->shaders
, arbfp_free_blit_shader
, context_gl
);
6934 checkGLcall("Delete blit programs");
6936 if (arbfp_blitter
->palette_texture
)
6937 gl_info
->gl_ops
.gl
.p_glDeleteTextures(1, &arbfp_blitter
->palette_texture
);
6939 heap_free(arbfp_blitter
);
6942 static void gen_packed_yuv_read(struct wined3d_string_buffer
*buffer
,
6943 const struct arbfp_blit_type
*type
, char *luminance
)
6946 const char *tex
, *texinstr
= "TXP";
6948 if (type
->fixup
== COMPLEX_FIXUP_UYVY
)
6959 tex
= arbfp_texture_target(type
->res_type
);
6960 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_RECT
)
6963 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6964 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6965 * filtering when we sample the texture.
6967 * These are the rules for reading the chroma:
6973 * So we have to get the sampling x position in non-normalized coordinates in integers
6975 if (type
->res_type
!= WINED3D_GL_RES_TYPE_TEX_RECT
)
6977 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6978 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6982 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6984 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6985 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6988 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6989 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6991 /* Multiply the x coordinate by 0.5 and get the fraction. This gives 0.25
6992 * and 0.75 for the even and odd pixels respectively. */
6993 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6994 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6996 /* Sample Pixel 1. */
6997 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6999 /* Put the value into either of the chroma values */
7000 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7001 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
7002 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7003 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
7005 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
7006 * the pixel right to the current one. Otherwise, sample the left pixel.
7007 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
7009 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
7010 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
7011 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
7013 /* Put the value into the other chroma */
7014 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
7015 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
7016 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
7017 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
7019 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
7020 * the current one and lerp the two U and V values
7023 /* This gives the correctly filtered luminance value */
7024 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
7027 static void gen_yv12_read(struct wined3d_string_buffer
*buffer
,
7028 const struct arbfp_blit_type
*type
, char *luminance
)
7031 static const float yv12_coef
[]
7032 = {2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
};
7034 tex
= arbfp_texture_target(type
->res_type
);
7036 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7037 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
7038 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
7039 * pitch of the luminance plane, the packing into the gl texture is a bit
7040 * unfortunate. If the whole texture is interpreted as luminance data it looks
7041 * approximately like this:
7043 * +----------------------------------+----
7055 * +----------------+-----------------+----
7057 * | V even rows | V odd rows |
7059 * +----------------+------------------ -
7061 * | U even rows | U odd rows |
7063 * +----------------+-----------------+----
7067 * So it appears as if there are 4 chroma images, but in fact the odd rows
7068 * in the chroma images are in the same row as the even ones. So it is
7069 * kinda tricky to read
7071 * When reading from rectangle textures, keep in mind that the input y coordinates
7072 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
7074 shader_addline(buffer
, "PARAM yv12_coef = ");
7075 shader_arb_append_imm_vec4(buffer
, yv12_coef
);
7076 shader_addline(buffer
, ";\n");
7078 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7079 /* the chroma planes have only half the width */
7080 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7082 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
7083 * the coordinate. Also read the right side of the image when reading odd lines
7085 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
7088 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7090 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
7092 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
7094 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
7095 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
7097 /* Read odd lines from the right side (add size * 0.5 to the x coordinate). */
7098 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7099 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7100 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7101 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7103 /* clamp, keep the half pixel origin in mind */
7104 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
7105 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7106 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
7107 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7111 /* The y coordinate for V is in the range [size, size + size / 4). */
7112 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7113 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
7115 /* Read odd lines from the right side (add size * 0.5 to the x coordinate). */
7116 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
7117 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
7118 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
7119 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
7120 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
7122 /* Make sure to read exactly from the pixel center */
7123 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
7124 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
7127 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
7128 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7129 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7130 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7131 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7133 /* Read the texture, put the result into the output register */
7134 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7135 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7137 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
7138 * No need to clamp because we're just reusing the already clamped value from above
7140 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7141 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
7143 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
7144 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7145 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7147 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7148 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7149 * values due to filtering
7151 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7152 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7154 /* Multiply the y coordinate by 2/3 and clamp it */
7155 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
7156 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
7157 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7158 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7162 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7163 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7166 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7167 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7168 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7173 static void gen_nv12_read(struct wined3d_string_buffer
*buffer
,
7174 const struct arbfp_blit_type
*type
, char *luminance
)
7177 static const float nv12_coef
[]
7178 = {2.0f
/ 3.0f
, 1.0f
/ 3.0f
, 1.0f
, 1.0f
};
7180 tex
= arbfp_texture_target(type
->res_type
);
7182 /* NV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
7183 * sized plane where each component is an UV pair. So the effective
7184 * bitdepth is 12 bits per pixel If the whole texture is interpreted as luminance
7185 * data it looks approximately like this:
7187 * +----------------------------------+----
7199 * +----------------------------------+----
7200 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7201 * |UVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUVUV|
7207 * +----------------------------------+----
7209 * When reading from rectangle textures, keep in mind that the input y coordinates
7210 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height. */
7212 shader_addline(buffer
, "PARAM nv12_coef = ");
7213 shader_arb_append_imm_vec4(buffer
, nv12_coef
);
7214 shader_addline(buffer
, ";\n");
7216 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7217 /* We only have half the number of chroma pixels. */
7218 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
7220 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7222 shader_addline(buffer
, "RCP chroma.w, size.x;\n");
7223 shader_addline(buffer
, "RCP chroma.z, size.y;\n");
7225 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, nv12_coef.y, nv12_coef.x;\n");
7227 /* We must not allow filtering horizontally, this would mix U and V.
7228 * Vertical filtering is ok. However, bear in mind that the pixel center is at
7229 * 0.5, so add 0.5. */
7231 /* Convert to non-normalized coordinates so we can find the
7232 * individual pixel. */
7233 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, size.x;\n");
7234 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7235 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7236 * add 0.5 to hit the center of the pixel. */
7237 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7239 /* Convert back to normalized coordinates. */
7240 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, chroma.w;\n");
7242 /* Clamp, keep the half pixel origin in mind. */
7243 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.z, nv12_coef.x;\n");
7244 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7245 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.z, nv12_coef.z;\n");
7246 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7250 /* The y coordinate for chroma is in the range [size, size + size / 2). */
7251 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.y, size.y;\n");
7253 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
7254 /* Multiply by 2 since chroma components are stored in UV pixel pairs,
7255 * add 0.5 to hit the center of the pixel. */
7256 shader_addline(buffer
, "MAD texcrd.x, texcrd.x, coef.z, coef.y;\n");
7259 shader_addline(buffer
, "MAD temp.y, size.y, coef.y, size.y;\n");
7260 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
7261 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7262 shader_addline(buffer
, "ADD temp.y, size.y, coef.y;\n");
7263 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
7265 /* Read the texture, put the result into the output register. */
7266 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7267 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
7269 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7272 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, chroma.w;\n");
7277 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.x;\n");
7279 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
7280 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
7282 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
7283 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
7284 * values due to filtering. */
7285 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
7286 if (type
->res_type
== WINED3D_GL_RES_TYPE_TEX_2D
)
7288 /* Multiply the y coordinate by 2/3 and clamp it */
7289 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, nv12_coef.x;\n");
7290 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, nv12_coef.x;\n");
7291 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
7292 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7296 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
7297 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
7300 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
7301 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
7302 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
7307 /* Context activation is done by the caller. */
7308 static GLuint
gen_p8_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7311 struct wined3d_string_buffer buffer
;
7312 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7314 /* This should not happen because we only use this conversion for
7315 * present blits which don't use color keying. */
7316 if (type
->use_color_key
)
7317 FIXME("Implement P8 color keying.\n");
7320 if (!string_buffer_init(&buffer
))
7322 ERR("Failed to initialize shader buffer.\n");
7326 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7327 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7330 string_buffer_free(&buffer
);
7334 shader_addline(&buffer
, "!!ARBfp1.0\n");
7335 shader_addline(&buffer
, "TEMP index;\n");
7337 /* { 255/256, 0.5/255*255/256, 0, 0 } */
7338 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
7340 /* The alpha-component contains the palette index */
7341 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7343 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
7344 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
7346 /* Use the alpha-component as an index in the palette to get the final color */
7347 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
7348 shader_addline(&buffer
, "END\n");
7350 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7352 string_buffer_free(&buffer
);
7357 /* Context activation is done by the caller. */
7358 static void arbfp_blitter_upload_palette(struct wined3d_arbfp_blitter
*blitter
,
7359 const struct wined3d_texture_gl
*texture_gl
, struct wined3d_context_gl
*context_gl
)
7361 const struct wined3d_palette
*palette
= texture_gl
->t
.swapchain
? texture_gl
->t
.swapchain
->palette
: NULL
;
7362 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
7364 if (!blitter
->palette_texture
)
7365 gl_info
->gl_ops
.gl
.p_glGenTextures(1, &blitter
->palette_texture
);
7367 GL_EXTCALL(glActiveTexture(GL_TEXTURE1
));
7368 gl_info
->gl_ops
.gl
.p_glBindTexture(GL_TEXTURE_1D
, blitter
->palette_texture
);
7370 gl_info
->gl_ops
.gl
.p_glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
7372 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
7373 /* Make sure we have discrete color levels. */
7374 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
7375 gl_info
->gl_ops
.gl
.p_glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
7376 /* TODO: avoid unneeded uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
7379 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 256, 0, GL_BGRA
,
7380 GL_UNSIGNED_INT_8_8_8_8_REV
, palette
->colors
);
7384 static const DWORD black
;
7385 FIXME("P8 surface loaded without a palette.\n");
7386 gl_info
->gl_ops
.gl
.p_glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGB
, 1, 0, GL_BGRA
,
7387 GL_UNSIGNED_INT_8_8_8_8_REV
, &black
);
7390 /* Switch back to unit 0 in which the 2D texture will be stored. */
7391 wined3d_context_gl_active_texture(context_gl
, gl_info
, 0);
7394 /* Context activation is done by the caller. */
7395 static GLuint
gen_yuv_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7398 struct wined3d_string_buffer buffer
;
7399 char luminance_component
;
7401 if (type
->use_color_key
)
7402 FIXME("Implement YUV color keying.\n");
7405 if (!string_buffer_init(&buffer
))
7407 ERR("Failed to initialize shader buffer.\n");
7411 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7412 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
7413 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7414 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7417 string_buffer_free(&buffer
);
7421 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
7422 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
7423 * two chroma(U and V) values. Each macropixel has two luminance values, one for
7424 * each single pixel it contains, and one U and one V value shared between both
7427 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
7428 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
7429 * take the format into account when generating the read swizzles
7431 * Reading the Y value is straightforward - just sample the texture. The hardware
7432 * takes care of filtering in the horizontal and vertical direction.
7434 * Reading the U and V values is harder. We have to avoid filtering horizontally,
7435 * because that would mix the U and V values of one pixel or two adjacent pixels.
7436 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
7437 * regardless of the filtering setting. Vertical filtering works automatically
7438 * though - the U and V values of two rows are mixed nicely.
7440 * Apart of avoiding filtering issues, the code has to know which value it just
7441 * read, and where it can find the other one. To determine this, it checks if
7442 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
7444 * Handling horizontal filtering of U and V values requires reading a 2nd pair
7445 * of pixels, extracting U and V and mixing them. This is not implemented yet.
7447 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
7448 * with width / 2. This way one read gives all 3 values, finding U and V is easy
7449 * in an unfiltered situation. Finding the luminance on the other hand requires
7450 * finding out if it is an odd or even pixel. The real drawback of this approach
7451 * is filtering. This would have to be emulated completely in the shader, reading
7452 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
7453 * vertically. Beyond that it would require adjustments to the texture handling
7454 * code to deal with the width scaling
7456 shader_addline(&buffer
, "!!ARBfp1.0\n");
7457 shader_addline(&buffer
, "TEMP luminance;\n");
7458 shader_addline(&buffer
, "TEMP temp;\n");
7459 shader_addline(&buffer
, "TEMP chroma;\n");
7460 shader_addline(&buffer
, "TEMP texcrd;\n");
7461 shader_addline(&buffer
, "TEMP texcrd2;\n");
7462 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
7463 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
7464 shader_addline(&buffer
, "PARAM size = program.local[%u];\n", ARBFP_BLIT_PARAM_SIZE
);
7466 switch (type
->fixup
)
7468 case COMPLEX_FIXUP_UYVY
:
7469 case COMPLEX_FIXUP_YUY2
:
7470 gen_packed_yuv_read(&buffer
, type
, &luminance_component
);
7473 case COMPLEX_FIXUP_YV12
:
7474 gen_yv12_read(&buffer
, type
, &luminance_component
);
7477 case COMPLEX_FIXUP_NV12
:
7478 gen_nv12_read(&buffer
, type
, &luminance_component
);
7482 FIXME("Unsupported YUV fixup %#x\n", type
->fixup
);
7483 string_buffer_free(&buffer
);
7487 /* Calculate the final result. Formula is taken from
7488 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
7489 * ranges from -0.5 to 0.5
7491 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
7493 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
7494 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
7495 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
7496 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
7497 shader_addline(&buffer
, "END\n");
7499 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7501 string_buffer_free(&buffer
);
7506 /* Context activation is done by the caller. */
7507 static GLuint
arbfp_gen_plain_shader(const struct wined3d_gl_info
*gl_info
, const struct arbfp_blit_type
*type
)
7510 struct wined3d_string_buffer buffer
;
7511 const char *tex_target
= arbfp_texture_target(type
->res_type
);
7514 if (!string_buffer_init(&buffer
))
7516 ERR("Failed to initialize shader buffer.\n");
7520 GL_EXTCALL(glGenProgramsARB(1, &shader
));
7523 string_buffer_free(&buffer
);
7526 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7528 shader_addline(&buffer
, "!!ARBfp1.0\n");
7530 if (type
->use_color_key
)
7532 shader_addline(&buffer
, "TEMP color;\n");
7533 shader_addline(&buffer
, "TEMP less, greater;\n");
7534 shader_addline(&buffer
, "PARAM color_key_low = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_LOW
);
7535 shader_addline(&buffer
, "PARAM color_key_high = program.local[%u];\n", ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
);
7536 shader_addline(&buffer
, "TEX color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7537 shader_addline(&buffer
, "SLT less, color, color_key_low;\n"); /* below low key */
7538 shader_addline(&buffer
, "SGE greater, color, color_key_high;\n"); /* above high key */
7539 shader_addline(&buffer
, "ADD less, less, greater;\n"); /* or */
7540 shader_addline(&buffer
, "DP4 less.b, less, less;\n"); /* on any channel */
7541 shader_addline(&buffer
, "SGE less, -less.b, 0.0;\n"); /* logical not */
7542 shader_addline(&buffer
, "KIL -less;\n"); /* discard if true */
7543 shader_addline(&buffer
, "MOV result.color, color;\n");
7547 shader_addline(&buffer
, "TEX result.color, fragment.texcoord[0], texture[0], %s;\n", tex_target
);
7550 shader_addline(&buffer
, "END\n");
7552 shader_arb_compile(gl_info
, GL_FRAGMENT_PROGRAM_ARB
, buffer
.buffer
);
7554 string_buffer_free(&buffer
);
7559 /* Context activation is done by the caller. */
7560 static HRESULT
arbfp_blit_set(struct wined3d_arbfp_blitter
*blitter
, struct wined3d_context_gl
*context_gl
,
7561 const struct wined3d_texture_gl
*texture_gl
, unsigned int sub_resource_idx
,
7562 const struct wined3d_color_key
*color_key
)
7564 const struct wined3d_gl_info
*gl_info
= context_gl
->gl_info
;
7565 enum complex_fixup fixup
;
7566 struct wine_rb_entry
*entry
;
7567 struct arbfp_blit_type type
;
7568 struct arbfp_blit_desc
*desc
;
7569 struct wined3d_color float_color_key
[2];
7570 struct wined3d_vec4 size
;
7574 level
= sub_resource_idx
% texture_gl
->t
.level_count
;
7575 size
.x
= wined3d_texture_get_level_pow2_width(&texture_gl
->t
, level
);
7576 size
.y
= wined3d_texture_get_level_pow2_height(&texture_gl
->t
, level
);
7580 if (is_complex_fixup(texture_gl
->t
.resource
.format
->color_fixup
))
7581 fixup
= get_complex_fixup(texture_gl
->t
.resource
.format
->color_fixup
);
7583 fixup
= COMPLEX_FIXUP_NONE
;
7585 switch (texture_gl
->target
)
7588 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_1D
;
7592 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7596 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_3D
;
7599 case GL_TEXTURE_CUBE_MAP_ARB
:
7600 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_CUBE
;
7603 case GL_TEXTURE_RECTANGLE_ARB
:
7604 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_RECT
;
7608 ERR("Unexpected GL texture type %#x.\n", texture_gl
->target
);
7609 type
.res_type
= WINED3D_GL_RES_TYPE_TEX_2D
;
7612 type
.use_color_key
= !!color_key
;
7615 if ((entry
= wine_rb_get(&blitter
->shaders
, &type
)))
7617 desc
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_blit_desc
, entry
);
7618 shader
= desc
->shader
;
7624 case COMPLEX_FIXUP_NONE
:
7625 if (!is_identity_fixup(texture_gl
->t
.resource
.format
->color_fixup
))
7626 FIXME("Implement support for sign or swizzle fixups.\n");
7627 shader
= arbfp_gen_plain_shader(gl_info
, &type
);
7630 case COMPLEX_FIXUP_P8
:
7631 shader
= gen_p8_shader(gl_info
, &type
);
7634 case COMPLEX_FIXUP_YUY2
:
7635 case COMPLEX_FIXUP_UYVY
:
7636 case COMPLEX_FIXUP_YV12
:
7637 case COMPLEX_FIXUP_NV12
:
7638 shader
= gen_yuv_shader(gl_info
, &type
);
7642 FIXME("Unsupported fixup %#x.\n", fixup
);
7648 ERR("Failed to get shader for fixup %#x.\n", fixup
);
7652 if (!(desc
= heap_alloc(sizeof(*desc
))))
7656 desc
->shader
= shader
;
7657 if (wine_rb_put(&blitter
->shaders
, &desc
->type
, &desc
->entry
) == -1)
7660 ERR("Out of memory\n");
7661 GL_EXTCALL(glDeleteProgramsARB(1, &shader
));
7662 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader))");
7663 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, 0));
7664 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, 0)");
7666 return E_OUTOFMEMORY
;
7670 if (fixup
== COMPLEX_FIXUP_P8
)
7671 arbfp_blitter_upload_palette(blitter
, texture_gl
, context_gl
);
7673 gl_info
->gl_ops
.gl
.p_glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7674 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7675 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7676 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7677 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARBFP_BLIT_PARAM_SIZE
, &size
.x
));
7678 checkGLcall("glProgramLocalParameter4fvARB");
7679 if (type
.use_color_key
)
7681 wined3d_format_get_float_color_key(texture_gl
->t
.resource
.format
, color_key
, float_color_key
);
7682 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7683 ARBFP_BLIT_PARAM_COLOR_KEY_LOW
, &float_color_key
[0].r
));
7684 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
7685 ARBFP_BLIT_PARAM_COLOR_KEY_HIGH
, &float_color_key
[1].r
));
7686 checkGLcall("glProgramLocalParameter4fvARB");
7692 /* Context activation is done by the caller. */
7693 static void arbfp_blit_unset(const struct wined3d_gl_info
*gl_info
)
7695 gl_info
->gl_ops
.gl
.p_glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7696 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7699 static BOOL
arbfp_blit_supported(enum wined3d_blit_op blit_op
, const struct wined3d_context
*context
,
7700 const struct wined3d_resource
*src_resource
, DWORD src_location
,
7701 const struct wined3d_resource
*dst_resource
, DWORD dst_location
)
7703 const struct wined3d_format
*src_format
= src_resource
->format
;
7704 const struct wined3d_format
*dst_format
= dst_resource
->format
;
7705 enum complex_fixup src_fixup
;
7708 if (src_resource
->type
!= WINED3D_RTYPE_TEXTURE_2D
)
7711 if (blit_op
== WINED3D_BLIT_OP_RAW_BLIT
&& dst_format
->id
== src_format
->id
)
7713 if (dst_format
->depth_size
|| dst_format
->stencil_size
)
7714 blit_op
= WINED3D_BLIT_OP_DEPTH_BLIT
;
7716 blit_op
= WINED3D_BLIT_OP_COLOR_BLIT
;
7721 case WINED3D_BLIT_OP_COLOR_BLIT_CKEY
:
7722 if (!context
->d3d_info
->shader_color_key
)
7724 /* The conversion modifies the alpha channel so the color key might no longer match. */
7725 TRACE("Color keying not supported with converted textures.\n");
7728 case WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
:
7729 case WINED3D_BLIT_OP_COLOR_BLIT
:
7733 TRACE("Unsupported blit_op=%d\n", blit_op
);
7737 decompress
= (src_format
->attrs
& WINED3D_FORMAT_ATTR_COMPRESSED
)
7738 && !(dst_format
->attrs
& WINED3D_FORMAT_ATTR_COMPRESSED
);
7739 if (!decompress
&& !(src_resource
->access
& dst_resource
->access
& WINED3D_RESOURCE_ACCESS_GPU
))
7742 src_fixup
= get_complex_fixup(src_format
->color_fixup
);
7743 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7745 TRACE("Checking support for fixup:\n");
7746 dump_color_fixup_desc(src_format
->color_fixup
);
7749 if (!is_identity_fixup(dst_format
->color_fixup
)
7750 && (dst_format
->id
!= src_format
->id
|| dst_location
!= WINED3D_LOCATION_DRAWABLE
))
7752 TRACE("Destination fixups are not supported\n");
7756 if (is_identity_fixup(src_format
->color_fixup
))
7762 /* We only support YUV conversions. */
7763 if (!is_complex_fixup(src_format
->color_fixup
))
7765 if (wined3d_settings
.offscreen_rendering_mode
== ORM_BACKBUFFER
)
7767 WARN("Claiming fixup support because of ORM_BACKBUFFER.\n");
7771 TRACE("[FAILED]\n");
7777 case COMPLEX_FIXUP_YUY2
:
7778 case COMPLEX_FIXUP_UYVY
:
7779 case COMPLEX_FIXUP_YV12
:
7780 case COMPLEX_FIXUP_NV12
:
7781 case COMPLEX_FIXUP_P8
:
7786 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7787 TRACE("[FAILED]\n");
7792 static DWORD
arbfp_blitter_blit(struct wined3d_blitter
*blitter
, enum wined3d_blit_op op
,
7793 struct wined3d_context
*context
, struct wined3d_texture
*src_texture
, unsigned int src_sub_resource_idx
,
7794 DWORD src_location
, const RECT
*src_rect
, struct wined3d_texture
*dst_texture
,
7795 unsigned int dst_sub_resource_idx
, DWORD dst_location
, const RECT
*dst_rect
,
7796 const struct wined3d_color_key
*color_key
, enum wined3d_texture_filter_type filter
,
7797 const struct wined3d_format
*resolve_format
)
7799 struct wined3d_texture_gl
*src_texture_gl
= wined3d_texture_gl(src_texture
);
7800 struct wined3d_context_gl
*context_gl
= wined3d_context_gl(context
);
7801 struct wined3d_device
*device
= dst_texture
->resource
.device
;
7802 struct wined3d_texture
*staging_texture
= NULL
;
7803 struct wined3d_arbfp_blitter
*arbfp_blitter
;
7804 struct wined3d_color_key alpha_test_key
;
7805 struct wined3d_blitter
*next
;
7806 unsigned int src_level
;
7809 TRACE("blitter %p, op %#x, context %p, src_texture %p, src_sub_resource_idx %u, src_location %s, "
7810 "src_rect %s, dst_texture %p, dst_sub_resource_idx %u, dst_location %s, dst_rect %s, "
7811 "colour_key %p, filter %s, resolve format %p.\n",
7812 blitter
, op
, context
, src_texture
, src_sub_resource_idx
, wined3d_debug_location(src_location
),
7813 wine_dbgstr_rect(src_rect
), dst_texture
, dst_sub_resource_idx
, wined3d_debug_location(dst_location
),
7814 wine_dbgstr_rect(dst_rect
), color_key
, debug_d3dtexturefiltertype(filter
), resolve_format
);
7816 if (!arbfp_blit_supported(op
, context
, &src_texture
->resource
, src_location
,
7817 &dst_texture
->resource
, dst_location
))
7819 if (!(next
= blitter
->next
))
7821 ERR("No blitter to handle blit op %#x.\n", op
);
7822 return dst_location
;
7825 TRACE("Forwarding to blitter %p.\n", next
);
7826 return next
->ops
->blitter_blit(next
, op
, context
, src_texture
, src_sub_resource_idx
, src_location
,
7827 src_rect
, dst_texture
, dst_sub_resource_idx
, dst_location
, dst_rect
, color_key
, filter
, resolve_format
);
7830 arbfp_blitter
= CONTAINING_RECORD(blitter
, struct wined3d_arbfp_blitter
, blitter
);
7832 if (!(src_texture
->resource
.access
& WINED3D_RESOURCE_ACCESS_GPU
))
7834 struct wined3d_resource_desc desc
;
7835 struct wined3d_box upload_box
;
7838 TRACE("Source texture is not GPU accessible, creating a staging texture.\n");
7840 src_level
= src_sub_resource_idx
% src_texture
->level_count
;
7841 desc
.resource_type
= WINED3D_RTYPE_TEXTURE_2D
;
7842 desc
.format
= src_texture
->resource
.format
->id
;
7843 desc
.multisample_type
= src_texture
->resource
.multisample_type
;
7844 desc
.multisample_quality
= src_texture
->resource
.multisample_quality
;
7845 desc
.usage
= WINED3DUSAGE_PRIVATE
;
7846 desc
.bind_flags
= 0;
7847 desc
.access
= WINED3D_RESOURCE_ACCESS_GPU
;
7848 desc
.width
= wined3d_texture_get_level_width(src_texture
, src_level
);
7849 desc
.height
= wined3d_texture_get_level_height(src_texture
, src_level
);
7853 if (FAILED(hr
= wined3d_texture_create(device
, &desc
, 1, 1, 0,
7854 NULL
, NULL
, &wined3d_null_parent_ops
, &staging_texture
)))
7856 ERR("Failed to create staging texture, hr %#lx.\n", hr
);
7857 return dst_location
;
7860 wined3d_box_set(&upload_box
, 0, 0, desc
.width
, desc
.height
, 0, desc
.depth
);
7861 wined3d_texture_upload_from_texture(staging_texture
, 0, 0, 0, 0,
7862 src_texture
, src_sub_resource_idx
, &upload_box
);
7864 src_texture
= staging_texture
;
7865 src_texture_gl
= wined3d_texture_gl(src_texture
);
7866 src_sub_resource_idx
= 0;
7868 else if (wined3d_settings
.offscreen_rendering_mode
!= ORM_FBO
7869 && (src_texture
->sub_resources
[src_sub_resource_idx
].locations
7870 & (WINED3D_LOCATION_TEXTURE_RGB
| WINED3D_LOCATION_DRAWABLE
)) == WINED3D_LOCATION_DRAWABLE
7871 && !wined3d_resource_is_offscreen(&src_texture
->resource
))
7874 /* Without FBO blits transferring from the drawable to the texture is
7875 * expensive, because we have to flip the data in sysmem. Since we can
7876 * flip in the blitter, we don't actually need that flip anyway. So we
7877 * use the surface's texture as scratch texture, and flip the source
7878 * rectangle instead. */
7879 texture2d_load_fb_texture(src_texture_gl
, src_sub_resource_idx
, FALSE
, context
);
7882 src_level
= src_sub_resource_idx
% src_texture
->level_count
;
7883 s
.top
= wined3d_texture_get_level_height(src_texture
, src_level
) - s
.top
;
7884 s
.bottom
= wined3d_texture_get_level_height(src_texture
, src_level
) - s
.bottom
;
7889 wined3d_texture_load(src_texture
, context
, FALSE
);
7892 if (wined3d_texture_is_full_rect(dst_texture
, dst_sub_resource_idx
% dst_texture
->level_count
, dst_rect
))
7893 wined3d_texture_prepare_location(dst_texture
, dst_sub_resource_idx
, context
, dst_location
);
7895 wined3d_texture_load_location(dst_texture
, dst_sub_resource_idx
, context
, dst_location
);
7897 context_gl_apply_texture_draw_state(context_gl
, dst_texture
, dst_sub_resource_idx
, dst_location
);
7898 wined3d_context_gl_apply_ffp_blit_state(context_gl
, device
);
7900 if (dst_location
== WINED3D_LOCATION_DRAWABLE
)
7903 wined3d_texture_translate_drawable_coords(dst_texture
, context_gl
->window
, &d
);
7907 if (op
== WINED3D_BLIT_OP_COLOR_BLIT_ALPHATEST
)
7909 const struct wined3d_format
*fmt
= src_texture
->resource
.format
;
7910 alpha_test_key
.color_space_low_value
= 0;
7911 alpha_test_key
.color_space_high_value
= ~(wined3d_mask_from_size(fmt
->alpha_size
) << fmt
->alpha_offset
);
7912 color_key
= &alpha_test_key
;
7915 arbfp_blit_set(arbfp_blitter
, context_gl
, src_texture_gl
, src_sub_resource_idx
, color_key
);
7917 /* Draw a textured quad */
7918 wined3d_context_gl_draw_textured_quad(context_gl
, src_texture_gl
,
7919 src_sub_resource_idx
, src_rect
, dst_rect
, filter
);
7921 /* Leave the opengl state valid for blitting */
7922 arbfp_blit_unset(context_gl
->gl_info
);
7924 if (dst_texture
->swapchain
&& (dst_texture
->swapchain
->front_buffer
== dst_texture
))
7925 context_gl
->gl_info
->gl_ops
.gl
.p_glFlush();
7927 if (staging_texture
)
7928 wined3d_texture_decref(staging_texture
);
7930 return dst_location
;
7933 static void arbfp_blitter_clear(struct wined3d_blitter
*blitter
, struct wined3d_device
*device
,
7934 unsigned int rt_count
, const struct wined3d_fb_state
*fb
, unsigned int rect_count
, const RECT
*clear_rects
,
7935 const RECT
*draw_rect
, uint32_t flags
, const struct wined3d_color
*colour
, float depth
, unsigned int stencil
)
7937 struct wined3d_blitter
*next
;
7939 if ((next
= blitter
->next
))
7940 next
->ops
->blitter_clear(next
, device
, rt_count
, fb
, rect_count
,
7941 clear_rects
, draw_rect
, flags
, colour
, depth
, stencil
);
7944 static const struct wined3d_blitter_ops arbfp_blitter_ops
=
7946 arbfp_blitter_destroy
,
7947 arbfp_blitter_clear
,
7951 void wined3d_arbfp_blitter_create(struct wined3d_blitter
**next
, const struct wined3d_device
*device
)
7953 const struct wined3d_gl_info
*gl_info
= &wined3d_adapter_gl(device
->adapter
)->gl_info
;
7954 struct wined3d_arbfp_blitter
*blitter
;
7956 if (device
->shader_backend
!= &arb_program_shader_backend
7957 && device
->shader_backend
!= &glsl_shader_backend
)
7960 if (!gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
7963 if (!gl_info
->supported
[WINED3D_GL_LEGACY_CONTEXT
])
7966 if (!(blitter
= heap_alloc(sizeof(*blitter
))))
7968 ERR("Failed to allocate blitter.\n");
7972 TRACE("Created blitter %p.\n", blitter
);
7974 blitter
->blitter
.ops
= &arbfp_blitter_ops
;
7975 blitter
->blitter
.next
= *next
;
7976 wine_rb_init(&blitter
->shaders
, arbfp_blit_type_compare
);
7977 blitter
->palette_texture
= 0;
7978 *next
= &blitter
->blitter
;