2 * Pixel and vertex shaders implementation using ARB_vertex_program
3 * and ARB_fragment_program GL extensions.
5 * Copyright 2002-2003 Jason Edmeades
6 * Copyright 2002-2003 Raphael Junqueira
7 * Copyright 2004 Christian Costa
8 * Copyright 2005 Oliver Stieber
9 * Copyright 2006 Ivan Gyurdiev
10 * Copyright 2006 Jason Green
11 * Copyright 2006 Henri Verbeet
12 * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
13 * Copyright 2009 Henri Verbeet for CodeWeavers
15 * This library is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU Lesser General Public
17 * License as published by the Free Software Foundation; either
18 * version 2.1 of the License, or (at your option) any later version.
20 * This library is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
23 * Lesser General Public License for more details.
25 * You should have received a copy of the GNU Lesser General Public
26 * License along with this library; if not, write to the Free Software
27 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
35 #include "wined3d_private.h"
37 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader
);
38 WINE_DECLARE_DEBUG_CHANNEL(d3d_constants
);
39 WINE_DECLARE_DEBUG_CHANNEL(d3d_caps
);
40 WINE_DECLARE_DEBUG_CHANNEL(d3d
);
42 /* Extract a line. Note that this modifies the source string. */
43 static char *get_line(char **ptr
)
48 if (!(q
= strstr(p
, "\n")))
60 static void shader_arb_dump_program_source(const char *source
)
63 char *ptr
, *line
, *tmp
;
65 source_size
= strlen(source
) + 1;
66 tmp
= HeapAlloc(GetProcessHeap(), 0, source_size
);
69 ERR("Failed to allocate %u bytes for shader source.\n", source_size
);
72 memcpy(tmp
, source
, source_size
);
75 while ((line
= get_line(&ptr
))) FIXME(" %s\n", line
);
78 HeapFree(GetProcessHeap(), 0, tmp
);
81 /* GL locking for state handlers is done by the caller. */
82 static BOOL
need_rel_addr_const(IWineD3DBaseShaderImpl
*shader
, const struct wined3d_gl_info
*gl_info
)
84 if (shader
->baseShader
.reg_maps
.shader_version
.type
== WINED3D_SHADER_TYPE_VERTEX
)
86 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
) return TRUE
;
88 if (!shader
->baseShader
.reg_maps
.usesmova
) return FALSE
;
89 return !gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
];
92 /* Returns TRUE if result.clip from GL_NV_vertex_program2 should be used and FALSE otherwise */
93 static inline BOOL
use_nv_clip(const struct wined3d_gl_info
*gl_info
)
95 return gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]
96 && !(gl_info
->quirks
& WINED3D_QUIRK_NV_CLIP_BROKEN
);
99 static BOOL
need_helper_const(IWineD3DBaseShaderImpl
*shader
, const struct wined3d_gl_info
*gl_info
)
101 if (need_rel_addr_const(shader
, gl_info
)) return TRUE
;
102 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
]) return TRUE
; /* Need to init colors. */
103 if (gl_info
->quirks
& WINED3D_QUIRK_ARB_VS_OFFSET_LIMIT
) return TRUE
; /* Load the immval offset. */
104 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
) return TRUE
; /* Have to init texcoords. */
105 if (!use_nv_clip(gl_info
)) return TRUE
; /* Init the clip texcoord */
106 if (shader
->baseShader
.reg_maps
.usesnrm
) return TRUE
; /* 0.0 */
107 if (shader
->baseShader
.reg_maps
.usesrcp
) return TRUE
; /* EPS */
111 static unsigned int reserved_vs_const(IWineD3DBaseShaderImpl
*shader
, const struct wined3d_gl_info
*gl_info
)
113 unsigned int ret
= 1;
114 /* We use one PARAM for the pos fixup, and in some cases one to load
115 * some immediate values into the shader
117 if(need_helper_const(shader
, gl_info
)) ret
++;
118 if(need_rel_addr_const(shader
, gl_info
)) ret
++;
122 enum arb_helper_value
133 static const char *arb_get_helper_value(enum wined3d_shader_type shader
, enum arb_helper_value value
)
135 if (shader
== WINED3D_SHADER_TYPE_GEOMETRY
)
137 ERR("Geometry shaders are unsupported\n");
141 if (shader
== WINED3D_SHADER_TYPE_PIXEL
)
145 case ARB_ZERO
: return "ps_helper_const.x";
146 case ARB_ONE
: return "ps_helper_const.y";
147 case ARB_TWO
: return "coefmul.x";
148 case ARB_0001
: return "helper_const.xxxy";
149 case ARB_EPS
: return "ps_helper_const.z";
157 case ARB_ZERO
: return "helper_const.x";
158 case ARB_ONE
: return "helper_const.y";
159 case ARB_TWO
: return "helper_const.z";
160 case ARB_EPS
: return "helper_const.w";
161 case ARB_0001
: return "helper_const.xxxy";
162 case ARB_VS_REL_OFFSET
: return "rel_addr_const.y";
165 FIXME("Unmanaged %s shader helper constant requested: %u\n",
166 shader
== WINED3D_SHADER_TYPE_PIXEL
? "pixel" : "vertex", value
);
169 case ARB_ZERO
: return "0.0";
170 case ARB_ONE
: return "1.0";
171 case ARB_TWO
: return "2.0";
172 case ARB_0001
: return "{0.0, 0.0, 0.0, 1.0}";
173 case ARB_EPS
: return "1e-8";
174 default: return "bad";
178 static inline BOOL
ffp_clip_emul(IWineD3DStateBlockImpl
*stateblock
)
180 return stateblock
->lowest_disabled_stage
< 7;
183 /* ARB_program_shader private data */
202 struct wined3d_shader_loop_control loop_control
;
206 struct arb_ps_np2fixup_info
208 struct ps_np2fixup_info super
;
209 /* For ARB we need a offset value:
210 * With both GLSL and ARB mode the NP2 fixup information (the texture dimensions) are stored in a
211 * consecutive way (GLSL uses a uniform array). Since ARB doesn't know the notion of a "standalone"
212 * array we need an offset to the index inside the program local parameter array. */
216 struct arb_ps_compile_args
218 struct ps_compile_args super
;
220 WORD clip
; /* only a boolean, use a WORD for alignment */
221 unsigned char loop_ctrl
[MAX_CONST_I
][3];
224 struct stb_const_desc
226 unsigned char texunit
;
230 struct arb_ps_compiled_shader
232 struct arb_ps_compile_args args
;
233 struct arb_ps_np2fixup_info np2fixup_info
;
234 struct stb_const_desc bumpenvmatconst
[MAX_TEXTURES
];
235 struct stb_const_desc luminanceconst
[MAX_TEXTURES
];
236 UINT int_consts
[MAX_CONST_I
];
239 unsigned char numbumpenvmatconsts
;
243 struct arb_vs_compile_args
245 struct vs_compile_args super
;
254 DWORD boolclip_compare
;
259 unsigned char samplers
[4];
260 DWORD samplers_compare
;
262 unsigned char loop_ctrl
[MAX_CONST_I
][3];
265 struct arb_vs_compiled_shader
267 struct arb_vs_compile_args args
;
269 UINT int_consts
[MAX_CONST_I
];
271 char need_color_unclamp
;
275 struct recorded_instruction
277 struct wined3d_shader_instruction ins
;
281 struct shader_arb_ctx_priv
286 /* plain GL_ARB_vertex_program or GL_ARB_fragment_program */
288 /* GL_NV_vertex_progam2_option or GL_NV_fragment_program_option */
290 /* GL_NV_vertex_program3 or GL_NV_fragment_program2 */
294 const struct arb_vs_compile_args
*cur_vs_args
;
295 const struct arb_ps_compile_args
*cur_ps_args
;
296 const struct arb_ps_compiled_shader
*compiled_fprog
;
297 const struct arb_vs_compiled_shader
*compiled_vprog
;
298 struct arb_ps_np2fixup_info
*cur_np2fixup_info
;
299 struct list control_frames
;
303 unsigned int num_loops
, loop_depth
, num_ifcs
;
306 unsigned int vs_clipplanes
;
310 /* For 3.0 vertex shaders */
311 const char *vs_output
[MAX_REG_OUTPUT
];
312 /* For 2.x and earlier vertex shaders */
313 const char *texcrd_output
[8], *color_output
[2], *fog_output
;
315 /* 3.0 pshader input for compatibility with fixed function */
316 const char *ps_input
[MAX_REG_INPUT
];
321 struct wined3d_shader_signature_element
*sig
;
323 struct wine_rb_entry entry
;
326 struct arb_pshader_private
{
327 struct arb_ps_compiled_shader
*gl_shaders
;
328 UINT num_gl_shaders
, shader_array_size
;
329 BOOL has_signature_idx
;
330 DWORD input_signature_idx
;
331 DWORD clipplane_emulation
;
335 struct arb_vshader_private
{
336 struct arb_vs_compiled_shader
*gl_shaders
;
337 UINT num_gl_shaders
, shader_array_size
;
340 struct shader_arb_priv
342 GLuint current_vprogram_id
;
343 GLuint current_fprogram_id
;
344 const struct arb_ps_compiled_shader
*compiled_fprog
;
345 const struct arb_vs_compiled_shader
*compiled_vprog
;
346 GLuint depth_blt_vprogram_id
;
347 GLuint depth_blt_fprogram_id_full
[tex_type_count
];
348 GLuint depth_blt_fprogram_id_masked
[tex_type_count
];
349 BOOL use_arbfp_fixed_func
;
350 struct wine_rb_tree fragment_shaders
;
351 BOOL last_ps_const_clamped
;
352 BOOL last_vs_color_unclamp
;
354 struct wine_rb_tree signature_tree
;
358 /********************************************************
359 * ARB_[vertex/fragment]_program helper functions follow
360 ********************************************************/
362 /* Loads floating point constants into the currently set ARB_vertex/fragment_program.
363 * When constant_list == NULL, it will load all the constants.
365 * @target_type should be either GL_VERTEX_PROGRAM_ARB (for vertex shaders)
366 * or GL_FRAGMENT_PROGRAM_ARB (for pixel shaders)
368 /* GL locking is done by the caller */
369 static unsigned int shader_arb_load_constantsF(IWineD3DBaseShaderImpl
*This
, const struct wined3d_gl_info
*gl_info
,
370 GLuint target_type
, unsigned int max_constants
, const float *constants
, char *dirty_consts
)
372 local_constant
* lconst
;
376 if (TRACE_ON(d3d_constants
))
378 for(i
= 0; i
< max_constants
; i
++) {
379 if(!dirty_consts
[i
]) continue;
380 TRACE_(d3d_constants
)("Loading constants %i: %f, %f, %f, %f\n", i
,
381 constants
[i
* 4 + 0], constants
[i
* 4 + 1],
382 constants
[i
* 4 + 2], constants
[i
* 4 + 3]);
388 /* In 1.X pixel shaders constants are implicitly clamped in the range [-1;1] */
389 if (target_type
== GL_FRAGMENT_PROGRAM_ARB
&& This
->baseShader
.reg_maps
.shader_version
.major
== 1)
392 /* ps 1.x supports only 8 constants, clamp only those. When switching between 1.x and higher
393 * shaders, the first 8 constants are marked dirty for reload
395 for(; i
< min(8, max_constants
); i
++) {
396 if(!dirty_consts
[i
]) continue;
400 if (constants
[j
+ 0] > 1.0f
) lcl_const
[0] = 1.0f
;
401 else if (constants
[j
+ 0] < -1.0f
) lcl_const
[0] = -1.0f
;
402 else lcl_const
[0] = constants
[j
+ 0];
404 if (constants
[j
+ 1] > 1.0f
) lcl_const
[1] = 1.0f
;
405 else if (constants
[j
+ 1] < -1.0f
) lcl_const
[1] = -1.0f
;
406 else lcl_const
[1] = constants
[j
+ 1];
408 if (constants
[j
+ 2] > 1.0f
) lcl_const
[2] = 1.0f
;
409 else if (constants
[j
+ 2] < -1.0f
) lcl_const
[2] = -1.0f
;
410 else lcl_const
[2] = constants
[j
+ 2];
412 if (constants
[j
+ 3] > 1.0f
) lcl_const
[3] = 1.0f
;
413 else if (constants
[j
+ 3] < -1.0f
) lcl_const
[3] = -1.0f
;
414 else lcl_const
[3] = constants
[j
+ 3];
416 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, lcl_const
));
419 /* If further constants are dirty, reload them without clamping.
421 * The alternative is not to touch them, but then we cannot reset the dirty constant count
422 * to zero. That's bad for apps that only use PS 1.x shaders, because in that case the code
423 * above would always re-check the first 8 constants since max_constant remains at the init
428 if (gl_info
->supported
[EXT_GPU_PROGRAM_PARAMETERS
])
430 /* TODO: Benchmark if we're better of with finding the dirty constants ourselves,
431 * or just reloading *all* constants at once
433 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type, i, max_constants, constants + (i * 4)));
435 for(; i
< max_constants
; i
++) {
436 if(!dirty_consts
[i
]) continue;
438 /* Find the next block of dirty constants */
441 for(i
++; (i
< max_constants
) && dirty_consts
[i
]; i
++) {
445 GL_EXTCALL(glProgramEnvParameters4fvEXT(target_type
, j
, i
- j
, constants
+ (j
* 4)));
448 for(; i
< max_constants
; i
++) {
449 if(dirty_consts
[i
]) {
451 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, i
, constants
+ (i
* 4)));
455 checkGLcall("glProgramEnvParameter4fvARB()");
457 /* Load immediate constants */
458 if(This
->baseShader
.load_local_constsF
) {
459 if (TRACE_ON(d3d_shader
)) {
460 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
461 GLfloat
* values
= (GLfloat
*)lconst
->value
;
462 TRACE_(d3d_constants
)("Loading local constants %i: %f, %f, %f, %f\n", lconst
->idx
,
463 values
[0], values
[1], values
[2], values
[3]);
466 /* Immediate constants are clamped for 1.X shaders at loading times */
468 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
469 dirty_consts
[lconst
->idx
] = 1; /* Dirtify so the non-immediate constant overwrites it next time */
470 ret
= max(ret
, lconst
->idx
+ 1);
471 GL_EXTCALL(glProgramEnvParameter4fvARB(target_type
, lconst
->idx
, (GLfloat
*)lconst
->value
));
473 checkGLcall("glProgramEnvParameter4fvARB()");
474 return ret
; /* The loaded immediate constants need reloading for the next shader */
476 return 0; /* No constants are dirty now */
481 * Loads the texture dimensions for NP2 fixup into the currently set ARB_[vertex/fragment]_programs.
483 static void shader_arb_load_np2fixup_constants(
484 IWineD3DDevice
* device
,
486 char useVertexShader
) {
488 IWineD3DDeviceImpl
* deviceImpl
= (IWineD3DDeviceImpl
*) device
;
489 const struct shader_arb_priv
* const priv
= (const struct shader_arb_priv
*) deviceImpl
->shader_priv
;
490 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
491 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
493 if (!usePixelShader
) {
494 /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
498 if (priv
->compiled_fprog
&& priv
->compiled_fprog
->np2fixup_info
.super
.active
) {
499 const struct arb_ps_np2fixup_info
* const fixup
= &priv
->compiled_fprog
->np2fixup_info
;
501 WORD active
= fixup
->super
.active
;
502 GLfloat np2fixup_constants
[4 * MAX_FRAGMENT_SAMPLERS
];
504 for (i
= 0; active
; active
>>= 1, ++i
) {
505 const unsigned char idx
= fixup
->super
.idx
[i
];
506 const IWineD3DTextureImpl
* const tex
= (const IWineD3DTextureImpl
*) stateBlock
->textures
[i
];
507 GLfloat
* tex_dim
= &np2fixup_constants
[(idx
>> 1) * 4];
509 if (!(active
& 1)) continue;
512 FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
517 tex_dim
[2] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[3] = tex
->baseTexture
.pow2Matrix
[5];
519 tex_dim
[0] = tex
->baseTexture
.pow2Matrix
[0]; tex_dim
[1] = tex
->baseTexture
.pow2Matrix
[5];
523 for (i
= 0; i
< fixup
->super
.num_consts
; ++i
) {
524 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
,
525 fixup
->offset
+ i
, &np2fixup_constants
[i
* 4]));
530 /* GL locking is done by the caller. */
531 static inline void shader_arb_ps_local_constants(IWineD3DDeviceImpl
* deviceImpl
)
533 const struct wined3d_context
*context
= context_get_current();
534 IWineD3DStateBlockImpl
* stateBlock
= deviceImpl
->stateBlock
;
535 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
537 struct shader_arb_priv
*priv
= deviceImpl
->shader_priv
;
538 const struct arb_ps_compiled_shader
*gl_shader
= priv
->compiled_fprog
;
540 for(i
= 0; i
< gl_shader
->numbumpenvmatconsts
; i
++)
542 int texunit
= gl_shader
->bumpenvmatconst
[i
].texunit
;
544 /* The state manager takes care that this function is always called if the bump env matrix changes */
545 const float *data
= (const float *)&stateBlock
->textureState
[texunit
][WINED3DTSS_BUMPENVMAT00
];
546 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->bumpenvmatconst
[i
].const_num
, data
));
548 if (gl_shader
->luminanceconst
[i
].const_num
!= WINED3D_CONST_NUM_UNUSED
)
550 /* WINED3DTSS_BUMPENVLSCALE and WINED3DTSS_BUMPENVLOFFSET are next to each other.
551 * point gl to the scale, and load 4 floats. x = scale, y = offset, z and w are junk, we
552 * don't care about them. The pointers are valid for sure because the stateblock is bigger.
553 * (they're WINED3DTSS_TEXTURETRANSFORMFLAGS and WINED3DTSS_ADDRESSW, so most likely 0 or NaN
555 const float *scale
= (const float *)&stateBlock
->textureState
[texunit
][WINED3DTSS_BUMPENVLSCALE
];
556 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->luminanceconst
[i
].const_num
, scale
));
559 checkGLcall("Load bumpmap consts");
561 if(gl_shader
->ycorrection
!= WINED3D_CONST_NUM_UNUSED
)
563 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
564 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
569 val
[0] = context
->render_offscreen
? 0.0f
570 : deviceImpl
->render_targets
[0]->currentDesc
.Height
;
571 val
[1] = context
->render_offscreen
? 1.0f
: -1.0f
;
574 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->ycorrection
, val
));
575 checkGLcall("y correction loading");
578 if(gl_shader
->num_int_consts
== 0) return;
580 for(i
= 0; i
< MAX_CONST_I
; i
++)
582 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
585 val
[0] = stateBlock
->pixelShaderConstantI
[4 * i
];
586 val
[1] = stateBlock
->pixelShaderConstantI
[4 * i
+ 1];
587 val
[2] = stateBlock
->pixelShaderConstantI
[4 * i
+ 2];
590 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
593 checkGLcall("Load ps int consts");
596 /* GL locking is done by the caller. */
597 static inline void shader_arb_vs_local_constants(IWineD3DDeviceImpl
* deviceImpl
)
599 IWineD3DStateBlockImpl
* stateBlock
;
600 const struct wined3d_gl_info
*gl_info
= &deviceImpl
->adapter
->gl_info
;
602 struct shader_arb_priv
*priv
= deviceImpl
->shader_priv
;
603 const struct arb_vs_compiled_shader
*gl_shader
= priv
->compiled_vprog
;
605 /* Upload the position fixup */
606 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->pos_fixup
, deviceImpl
->posFixup
));
608 if(gl_shader
->num_int_consts
== 0) return;
610 stateBlock
= deviceImpl
->stateBlock
;
612 for(i
= 0; i
< MAX_CONST_I
; i
++)
614 if(gl_shader
->int_consts
[i
] != WINED3D_CONST_NUM_UNUSED
)
617 val
[0] = stateBlock
->vertexShaderConstantI
[4 * i
];
618 val
[1] = stateBlock
->vertexShaderConstantI
[4 * i
+ 1];
619 val
[2] = stateBlock
->vertexShaderConstantI
[4 * i
+ 2];
622 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, gl_shader
->int_consts
[i
], val
));
625 checkGLcall("Load vs int consts");
629 * Loads the app-supplied constants into the currently set ARB_[vertex/fragment]_programs.
631 * We only support float constants in ARB at the moment, so don't
632 * worry about the Integers or Booleans
634 /* GL locking is done by the caller (state handler) */
635 static void shader_arb_load_constants(const struct wined3d_context
*context
, char usePixelShader
, char useVertexShader
)
637 IWineD3DDeviceImpl
*device
= context
->swapchain
->device
;
638 IWineD3DStateBlockImpl
* stateBlock
= device
->stateBlock
;
639 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
641 if (useVertexShader
) {
642 IWineD3DBaseShaderImpl
* vshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->vertexShader
;
644 /* Load DirectX 9 float constants for vertex shader */
645 device
->highest_dirty_vs_const
= shader_arb_load_constantsF(vshader
, gl_info
, GL_VERTEX_PROGRAM_ARB
,
646 device
->highest_dirty_vs_const
, stateBlock
->vertexShaderConstantF
, context
->vshader_const_dirty
);
647 shader_arb_vs_local_constants(device
);
650 if (usePixelShader
) {
651 IWineD3DBaseShaderImpl
* pshader
= (IWineD3DBaseShaderImpl
*) stateBlock
->pixelShader
;
653 /* Load DirectX 9 float constants for pixel shader */
654 device
->highest_dirty_ps_const
= shader_arb_load_constantsF(pshader
, gl_info
, GL_FRAGMENT_PROGRAM_ARB
,
655 device
->highest_dirty_ps_const
, stateBlock
->pixelShaderConstantF
, context
->pshader_const_dirty
);
656 shader_arb_ps_local_constants(device
);
660 static void shader_arb_update_float_vertex_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
662 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
663 struct wined3d_context
*context
= context_get_current();
665 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
666 * context. On a context switch the old context will be fully dirtified */
667 if (!context
|| context
->swapchain
->device
!= This
) return;
669 memset(context
->vshader_const_dirty
+ start
, 1, sizeof(*context
->vshader_const_dirty
) * count
);
670 This
->highest_dirty_vs_const
= max(This
->highest_dirty_vs_const
, start
+ count
);
673 static void shader_arb_update_float_pixel_constants(IWineD3DDevice
*iface
, UINT start
, UINT count
)
675 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
676 struct wined3d_context
*context
= context_get_current();
678 /* We don't want shader constant dirtification to be an O(contexts), so just dirtify the active
679 * context. On a context switch the old context will be fully dirtified */
680 if (!context
|| context
->swapchain
->device
!= This
) return;
682 memset(context
->pshader_const_dirty
+ start
, 1, sizeof(*context
->pshader_const_dirty
) * count
);
683 This
->highest_dirty_ps_const
= max(This
->highest_dirty_ps_const
, start
+ count
);
686 static DWORD
*local_const_mapping(IWineD3DBaseShaderImpl
*This
)
690 const local_constant
*lconst
;
692 if(This
->baseShader
.load_local_constsF
|| list_empty(&This
->baseShader
.constantsF
)) return NULL
;
694 ret
= HeapAlloc(GetProcessHeap(), 0, sizeof(DWORD
) * This
->baseShader
.limits
.constant_float
);
696 ERR("Out of memory\n");
700 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
701 ret
[lconst
->idx
] = idx
++;
706 /* Generate the variable & register declarations for the ARB_vertex_program output target */
707 static DWORD
shader_generate_arb_declarations(IWineD3DBaseShader
*iface
, const shader_reg_maps
*reg_maps
,
708 struct wined3d_shader_buffer
*buffer
, const struct wined3d_gl_info
*gl_info
, DWORD
*lconst_map
,
709 DWORD
*num_clipplanes
, struct shader_arb_ctx_priv
*ctx
)
711 IWineD3DBaseShaderImpl
* This
= (IWineD3DBaseShaderImpl
*) iface
;
712 DWORD i
, next_local
= 0;
713 char pshader
= shader_is_pshader_version(reg_maps
->shader_version
.type
);
714 unsigned max_constantsF
;
715 const local_constant
*lconst
;
718 /* In pixel shaders, all private constants are program local, we don't need anything
719 * from program.env. Thus we can advertise the full set of constants in pixel shaders.
720 * If we need a private constant the GL implementation will squeeze it in somewhere
722 * With vertex shaders we need the posFixup and on some GL implementations 4 helper
723 * immediate values. The posFixup is loaded using program.env for now, so always
724 * subtract one from the number of constants. If the shader uses indirect addressing,
725 * account for the helper const too because we have to declare all availabke d3d constants
726 * and don't know which are actually used.
730 max_constantsF
= gl_info
->limits
.arb_ps_native_constants
;
731 /* 24 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value. */
732 if (max_constantsF
< 24)
733 max_constantsF
= gl_info
->limits
.arb_ps_float_constants
;
737 max_constantsF
= gl_info
->limits
.arb_vs_native_constants
;
738 /* 96 is the minimum MAX_PROGRAM_ENV_PARAMETERS_ARB value.
739 * Also prevents max_constantsF from becoming less than 0 and
741 if (max_constantsF
< 96)
742 max_constantsF
= gl_info
->limits
.arb_vs_float_constants
;
744 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
745 DWORD highest_constf
= 0, clip_limit
;
747 max_constantsF
-= reserved_vs_const(This
, gl_info
);
748 max_constantsF
-= count_bits(This
->baseShader
.reg_maps
.integer_constants
);
750 for(i
= 0; i
< This
->baseShader
.limits
.constant_float
; i
++)
753 DWORD shift
= i
& 0x1f;
754 if(reg_maps
->constf
[idx
] & (1 << shift
)) highest_constf
= i
;
757 if(use_nv_clip(gl_info
) && ctx
->target_version
>= NV2
)
759 if(ctx
->cur_vs_args
->super
.clip_enabled
)
760 clip_limit
= gl_info
->limits
.clipplanes
;
766 unsigned int mask
= ctx
->cur_vs_args
->clip
.boolclip
.clipplane_mask
;
767 clip_limit
= min(count_bits(mask
), 4);
769 *num_clipplanes
= min(clip_limit
, max_constantsF
- highest_constf
- 1);
770 max_constantsF
-= *num_clipplanes
;
771 if(*num_clipplanes
< clip_limit
)
773 WARN("Only %u clipplanes out of %u enabled\n", *num_clipplanes
, gl_info
->limits
.clipplanes
);
778 if (ctx
->target_version
>= NV2
) *num_clipplanes
= gl_info
->limits
.clipplanes
;
779 else *num_clipplanes
= min(gl_info
->limits
.clipplanes
, 4);
783 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
785 if (map
& 1) shader_addline(buffer
, "TEMP R%u;\n", i
);
788 for (i
= 0, map
= reg_maps
->address
; map
; map
>>= 1, ++i
)
790 if (map
& 1) shader_addline(buffer
, "ADDRESS A%u;\n", i
);
793 if (pshader
&& reg_maps
->shader_version
.major
== 1 && reg_maps
->shader_version
.minor
<= 3)
795 for (i
= 0, map
= reg_maps
->texcoord
; map
; map
>>= 1, ++i
)
797 if (map
& 1) shader_addline(buffer
, "TEMP T%u;\n", i
);
801 /* Load local constants using the program-local space,
802 * this avoids reloading them each time the shader is used
805 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
806 shader_addline(buffer
, "PARAM C%u = program.local[%u];\n", lconst
->idx
,
807 lconst_map
[lconst
->idx
]);
808 next_local
= max(next_local
, lconst_map
[lconst
->idx
] + 1);
812 /* After subtracting privately used constants from the hardware limit(they are loaded as
813 * local constants), make sure the shader doesn't violate the env constant limit
817 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_ps_float_constants
);
821 max_constantsF
= min(max_constantsF
, gl_info
->limits
.arb_vs_float_constants
);
824 /* Avoid declaring more constants than needed */
825 max_constantsF
= min(max_constantsF
, This
->baseShader
.limits
.constant_float
);
827 /* we use the array-based constants array if the local constants are marked for loading,
828 * because then we use indirect addressing, or when the local constant list is empty,
829 * because then we don't know if we're using indirect addressing or not. If we're hardcoding
830 * local constants do not declare the loaded constants as an array because ARB compilers usually
831 * do not optimize unused constants away
833 if(This
->baseShader
.reg_maps
.usesrelconstF
) {
834 /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
835 shader_addline(buffer
, "PARAM C[%d] = { program.env[0..%d] };\n",
836 max_constantsF
, max_constantsF
- 1);
838 for(i
= 0; i
< max_constantsF
; i
++) {
841 mask
= 1 << (i
& 0x1f);
842 if(!shader_constant_is_local(This
, i
) && (This
->baseShader
.reg_maps
.constf
[idx
] & mask
)) {
843 shader_addline(buffer
, "PARAM C%d = program.env[%d];\n",i
, i
);
851 static const char * const shift_tab
[] = {
852 "dummy", /* 0 (none) */
853 "coefmul.x", /* 1 (x2) */
854 "coefmul.y", /* 2 (x4) */
855 "coefmul.z", /* 3 (x8) */
856 "coefmul.w", /* 4 (x16) */
857 "dummy", /* 5 (x32) */
858 "dummy", /* 6 (x64) */
859 "dummy", /* 7 (x128) */
860 "dummy", /* 8 (d256) */
861 "dummy", /* 9 (d128) */
862 "dummy", /* 10 (d64) */
863 "dummy", /* 11 (d32) */
864 "coefdiv.w", /* 12 (d16) */
865 "coefdiv.z", /* 13 (d8) */
866 "coefdiv.y", /* 14 (d4) */
867 "coefdiv.x" /* 15 (d2) */
870 static void shader_arb_get_write_mask(const struct wined3d_shader_instruction
*ins
,
871 const struct wined3d_shader_dst_param
*dst
, char *write_mask
)
873 char *ptr
= write_mask
;
875 if (dst
->write_mask
!= WINED3DSP_WRITEMASK_ALL
)
878 if (dst
->write_mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
879 if (dst
->write_mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
880 if (dst
->write_mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
881 if (dst
->write_mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
887 static void shader_arb_get_swizzle(const struct wined3d_shader_src_param
*param
, BOOL fixup
, char *swizzle_str
)
889 /* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
890 * but addressed as "rgba". To fix this we need to swap the register's x
891 * and z components. */
892 const char *swizzle_chars
= fixup
? "zyxw" : "xyzw";
893 char *ptr
= swizzle_str
;
895 /* swizzle bits fields: wwzzyyxx */
896 DWORD swizzle
= param
->swizzle
;
897 DWORD swizzle_x
= swizzle
& 0x03;
898 DWORD swizzle_y
= (swizzle
>> 2) & 0x03;
899 DWORD swizzle_z
= (swizzle
>> 4) & 0x03;
900 DWORD swizzle_w
= (swizzle
>> 6) & 0x03;
902 /* If the swizzle is the default swizzle (ie, "xyzw"), we don't need to
903 * generate a swizzle string. Unless we need to our own swizzling. */
904 if (swizzle
!= WINED3DSP_NOSWIZZLE
|| fixup
)
907 if (swizzle_x
== swizzle_y
&& swizzle_x
== swizzle_z
&& swizzle_x
== swizzle_w
) {
908 *ptr
++ = swizzle_chars
[swizzle_x
];
910 *ptr
++ = swizzle_chars
[swizzle_x
];
911 *ptr
++ = swizzle_chars
[swizzle_y
];
912 *ptr
++ = swizzle_chars
[swizzle_z
];
913 *ptr
++ = swizzle_chars
[swizzle_w
];
920 static void shader_arb_request_a0(const struct wined3d_shader_instruction
*ins
, const char *src
)
922 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
923 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
925 if(strcmp(priv
->addr_reg
, src
) == 0) return;
927 strcpy(priv
->addr_reg
, src
);
928 shader_addline(buffer
, "ARL A0.x, %s;\n", src
);
931 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
932 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
);
934 static void shader_arb_get_register_name(const struct wined3d_shader_instruction
*ins
,
935 const struct wined3d_shader_register
*reg
, char *register_name
, BOOL
*is_color
)
937 /* oPos, oFog and oPts in D3D */
938 static const char * const rastout_reg_names
[] = {"TMP_OUT", "result.fogcoord", "result.pointsize"};
939 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
940 BOOL pshader
= shader_is_pshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
941 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
947 case WINED3DSPR_TEMP
:
948 sprintf(register_name
, "R%u", reg
->idx
);
951 case WINED3DSPR_INPUT
:
954 if(This
->baseShader
.reg_maps
.shader_version
.major
< 3)
956 if (reg
->idx
== 0) strcpy(register_name
, "fragment.color.primary");
957 else strcpy(register_name
, "fragment.color.secondary");
964 shader_arb_get_src_param(ins
, reg
->rel_addr
, 0, rel_reg
);
966 if(strcmp(rel_reg
, "**aL_emul**") == 0)
968 DWORD idx
= ctx
->aL
+ reg
->idx
;
969 if(idx
< MAX_REG_INPUT
)
971 strcpy(register_name
, ctx
->ps_input
[idx
]);
975 ERR("Pixel shader input register out of bounds: %u\n", idx
);
976 sprintf(register_name
, "out_of_bounds_%u", idx
);
979 else if(This
->baseShader
.reg_maps
.input_registers
& 0x0300)
981 /* There are two ways basically:
983 * 1) Use the unrolling code that is used for loop emulation and unroll the loop.
984 * That means trouble if the loop also contains a breakc or if the control values
985 * aren't local constants.
986 * 2) Generate an if block that checks if aL.y < 8, == 8 or == 9 and selects the
987 * source dynamically. The trouble is that we cannot simply read aL.y because it
988 * is an ADDRESS register. We could however push it, load .zw with a value and use
989 * ADAC to load the condition code register and pop it again afterwards
991 FIXME("Relative input register addressing with more than 8 registers\n");
993 /* This is better than nothing for now */
994 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
996 else if(ctx
->cur_ps_args
->super
.vp_mode
!= vertexshader
)
998 /* This is problematic because we'd have to consult the ctx->ps_input strings
999 * for where to find the varying. Some may be "0.0", others can be texcoords or
1000 * colors. This needs either a pipeline replacement to make the vertex shader feed
1001 * proper varyings, or loop unrolling
1003 * For now use the texcoords and hope for the best
1005 FIXME("Non-vertex shader varying input with indirect addressing\n");
1006 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
1010 /* D3D supports indirect addressing only with aL in loop registers. The loop instruction
1011 * pulls GL_NV_fragment_program2 in
1013 sprintf(register_name
, "fragment.texcoord[%s + %u]", rel_reg
, reg
->idx
);
1018 if(reg
->idx
< MAX_REG_INPUT
)
1020 strcpy(register_name
, ctx
->ps_input
[reg
->idx
]);
1024 ERR("Pixel shader input register out of bounds: %u\n", reg
->idx
);
1025 sprintf(register_name
, "out_of_bounds_%u", reg
->idx
);
1032 if (ctx
->cur_vs_args
->super
.swizzle_map
& (1 << reg
->idx
)) *is_color
= TRUE
;
1033 sprintf(register_name
, "vertex.attrib[%u]", reg
->idx
);
1037 case WINED3DSPR_CONST
:
1038 if (!pshader
&& reg
->rel_addr
)
1042 UINT rel_offset
= ((IWineD3DVertexShaderImpl
*)This
)->rel_offset
;
1043 if(This
->baseShader
.reg_maps
.shader_version
.major
< 2) {
1044 sprintf(rel_reg
, "A0.x");
1046 shader_arb_get_src_param(ins
, reg
->rel_addr
, 0, rel_reg
);
1047 if(ctx
->target_version
== ARB
) {
1048 if(strcmp(rel_reg
, "**aL_emul**") == 0) {
1051 shader_arb_request_a0(ins
, rel_reg
);
1052 sprintf(rel_reg
, "A0.x");
1057 sprintf(register_name
, "C[%u]", ctx
->aL
+ reg
->idx
);
1058 else if (reg
->idx
>= rel_offset
)
1059 sprintf(register_name
, "C[%s + %u]", rel_reg
, reg
->idx
- rel_offset
);
1061 sprintf(register_name
, "C[%s - %u]", rel_reg
, -reg
->idx
+ rel_offset
);
1065 if (This
->baseShader
.reg_maps
.usesrelconstF
)
1066 sprintf(register_name
, "C[%u]", reg
->idx
);
1068 sprintf(register_name
, "C%u", reg
->idx
);
1072 case WINED3DSPR_TEXTURE
: /* case WINED3DSPR_ADDR: */
1074 if(This
->baseShader
.reg_maps
.shader_version
.major
== 1 &&
1075 This
->baseShader
.reg_maps
.shader_version
.minor
<= 3) {
1076 /* In ps <= 1.3, Tx is a temporary register as destination to all instructions,
1077 * and as source to most instructions. For some instructions it is the texcoord
1078 * input. Those instructions know about the special use
1080 sprintf(register_name
, "T%u", reg
->idx
);
1082 /* in ps 1.4 and 2.x Tx is always a (read-only) varying */
1083 sprintf(register_name
, "fragment.texcoord[%u]", reg
->idx
);
1088 if(This
->baseShader
.reg_maps
.shader_version
.major
== 1 || ctx
->target_version
>= NV2
)
1090 sprintf(register_name
, "A%u", reg
->idx
);
1094 sprintf(register_name
, "A%u_SHADOW", reg
->idx
);
1099 case WINED3DSPR_COLOROUT
:
1100 if(ctx
->cur_ps_args
->super
.srgb_correction
&& reg
->idx
== 0)
1102 strcpy(register_name
, "TMP_COLOR");
1106 if(ctx
->cur_ps_args
->super
.srgb_correction
) FIXME("sRGB correction on higher render targets\n");
1107 if(This
->baseShader
.reg_maps
.highest_render_target
> 0)
1109 sprintf(register_name
, "result.color[%u]", reg
->idx
);
1113 strcpy(register_name
, "result.color");
1118 case WINED3DSPR_RASTOUT
:
1119 if(reg
->idx
== 1) sprintf(register_name
, "%s", ctx
->fog_output
);
1120 else sprintf(register_name
, "%s", rastout_reg_names
[reg
->idx
]);
1123 case WINED3DSPR_DEPTHOUT
:
1124 strcpy(register_name
, "result.depth");
1127 case WINED3DSPR_ATTROUT
:
1128 /* case WINED3DSPR_OUTPUT: */
1129 if (pshader
) sprintf(register_name
, "oD[%u]", reg
->idx
);
1130 else strcpy(register_name
, ctx
->color_output
[reg
->idx
]);
1133 case WINED3DSPR_TEXCRDOUT
:
1136 sprintf(register_name
, "oT[%u]", reg
->idx
);
1140 if(This
->baseShader
.reg_maps
.shader_version
.major
< 3)
1142 strcpy(register_name
, ctx
->texcrd_output
[reg
->idx
]);
1146 strcpy(register_name
, ctx
->vs_output
[reg
->idx
]);
1151 case WINED3DSPR_LOOP
:
1152 if(ctx
->target_version
>= NV2
)
1154 /* Pshader has an implicitly declared loop index counter A0.x that cannot be renamed */
1155 if(pshader
) sprintf(register_name
, "A0.x");
1156 else sprintf(register_name
, "aL.y");
1160 /* Unfortunately this code cannot return the value of ctx->aL here. An immediate value
1161 * would be valid, but if aL is used for indexing(its only use), there's likely an offset,
1162 * thus the result would be something like C[15 + 30], which is not valid in the ARB program
1163 * grammar. So return a marker for the emulated aL and intercept it in constant and varying
1166 sprintf(register_name
, "**aL_emul**");
1171 case WINED3DSPR_CONSTINT
:
1172 sprintf(register_name
, "I%u", reg
->idx
);
1175 case WINED3DSPR_MISCTYPE
:
1178 sprintf(register_name
, "vpos");
1180 else if(reg
->idx
== 1)
1182 sprintf(register_name
, "fragment.facing.x");
1186 FIXME("Unknown MISCTYPE register index %u\n", reg
->idx
);
1191 FIXME("Unhandled register type %#x[%u]\n", reg
->type
, reg
->idx
);
1192 sprintf(register_name
, "unrecognized_register[%u]", reg
->idx
);
1197 static void shader_arb_get_dst_param(const struct wined3d_shader_instruction
*ins
,
1198 const struct wined3d_shader_dst_param
*wined3d_dst
, char *str
)
1200 char register_name
[255];
1204 shader_arb_get_register_name(ins
, &wined3d_dst
->reg
, register_name
, &is_color
);
1205 strcpy(str
, register_name
);
1207 shader_arb_get_write_mask(ins
, wined3d_dst
, write_mask
);
1208 strcat(str
, write_mask
);
1211 static const char *shader_arb_get_fixup_swizzle(enum fixup_channel_source channel_source
)
1213 switch(channel_source
)
1215 case CHANNEL_SOURCE_ZERO
: return "0";
1216 case CHANNEL_SOURCE_ONE
: return "1";
1217 case CHANNEL_SOURCE_X
: return "x";
1218 case CHANNEL_SOURCE_Y
: return "y";
1219 case CHANNEL_SOURCE_Z
: return "z";
1220 case CHANNEL_SOURCE_W
: return "w";
1222 FIXME("Unhandled channel source %#x\n", channel_source
);
1227 static void gen_color_correction(struct wined3d_shader_buffer
*buffer
, const char *reg
,
1228 DWORD dst_mask
, const char *one
, const char *two
, struct color_fixup_desc fixup
)
1232 if (is_complex_fixup(fixup
))
1234 enum complex_fixup complex_fixup
= get_complex_fixup(fixup
);
1235 FIXME("Complex fixup (%#x) not supported\n", complex_fixup
);
1240 if (fixup
.x_source
!= CHANNEL_SOURCE_X
) mask
|= WINED3DSP_WRITEMASK_0
;
1241 if (fixup
.y_source
!= CHANNEL_SOURCE_Y
) mask
|= WINED3DSP_WRITEMASK_1
;
1242 if (fixup
.z_source
!= CHANNEL_SOURCE_Z
) mask
|= WINED3DSP_WRITEMASK_2
;
1243 if (fixup
.w_source
!= CHANNEL_SOURCE_W
) mask
|= WINED3DSP_WRITEMASK_3
;
1248 shader_addline(buffer
, "SWZ %s, %s, %s, %s, %s, %s;\n", reg
, reg
,
1249 shader_arb_get_fixup_swizzle(fixup
.x_source
), shader_arb_get_fixup_swizzle(fixup
.y_source
),
1250 shader_arb_get_fixup_swizzle(fixup
.z_source
), shader_arb_get_fixup_swizzle(fixup
.w_source
));
1254 if (fixup
.x_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_0
;
1255 if (fixup
.y_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_1
;
1256 if (fixup
.z_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_2
;
1257 if (fixup
.w_sign_fixup
) mask
|= WINED3DSP_WRITEMASK_3
;
1263 char *ptr
= reg_mask
;
1265 if (mask
!= WINED3DSP_WRITEMASK_ALL
)
1268 if (mask
& WINED3DSP_WRITEMASK_0
) *ptr
++ = 'x';
1269 if (mask
& WINED3DSP_WRITEMASK_1
) *ptr
++ = 'y';
1270 if (mask
& WINED3DSP_WRITEMASK_2
) *ptr
++ = 'z';
1271 if (mask
& WINED3DSP_WRITEMASK_3
) *ptr
++ = 'w';
1275 shader_addline(buffer
, "MAD %s%s, %s, %s, -%s;\n", reg
, reg_mask
, reg
, two
, one
);
1279 static const char *shader_arb_get_modifier(const struct wined3d_shader_instruction
*ins
)
1282 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1283 if (!ins
->dst_count
) return "";
1285 mod
= ins
->dst
[0].modifiers
;
1287 /* Silently ignore PARTIALPRECISION if its not supported */
1288 if(priv
->target_version
== ARB
) mod
&= ~WINED3DSPDM_PARTIALPRECISION
;
1290 if(mod
& WINED3DSPDM_MSAMPCENTROID
)
1292 FIXME("Unhandled modifier WINED3DSPDM_MSAMPCENTROID\n");
1293 mod
&= ~WINED3DSPDM_MSAMPCENTROID
;
1298 case WINED3DSPDM_SATURATE
| WINED3DSPDM_PARTIALPRECISION
:
1301 case WINED3DSPDM_SATURATE
:
1304 case WINED3DSPDM_PARTIALPRECISION
:
1311 FIXME("Unknown modifiers 0x%08x\n", mod
);
1316 #define TEX_PROJ 0x1
1317 #define TEX_BIAS 0x2
1319 #define TEX_DERIV 0x10
1321 static void shader_hw_sample(const struct wined3d_shader_instruction
*ins
, DWORD sampler_idx
,
1322 const char *dst_str
, const char *coord_reg
, WORD flags
, const char *dsx
, const char *dsy
)
1324 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1325 DWORD sampler_type
= ins
->ctx
->reg_maps
->sampler_type
[sampler_idx
];
1326 const char *tex_type
;
1327 BOOL np2_fixup
= FALSE
;
1328 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1329 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) This
->baseShader
.device
;
1330 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
1332 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
1334 /* D3D vertex shader sampler IDs are vertex samplers(0-3), not global d3d samplers */
1335 if(!pshader
) sampler_idx
+= MAX_FRAGMENT_SAMPLERS
;
1337 switch(sampler_type
) {
1343 if(device
->stateBlock
->textures
[sampler_idx
] &&
1344 IWineD3DBaseTexture_GetTextureDimensions(device
->stateBlock
->textures
[sampler_idx
]) == GL_TEXTURE_RECTANGLE_ARB
) {
1349 if (shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
))
1351 if (priv
->cur_np2fixup_info
->super
.active
& (1 << sampler_idx
))
1353 if (flags
) FIXME("Only ordinary sampling from NP2 textures is supported.\n");
1354 else np2_fixup
= TRUE
;
1359 case WINED3DSTT_VOLUME
:
1363 case WINED3DSTT_CUBE
:
1368 ERR("Unexpected texture type %d\n", sampler_type
);
1372 /* TEX, TXL, TXD and TXP do not support the "H" modifier,
1373 * so don't use shader_arb_get_modifier
1375 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) mod
= "_SAT";
1378 /* Fragment samplers always have indentity mapping */
1379 if(sampler_idx
>= MAX_FRAGMENT_SAMPLERS
)
1381 sampler_idx
= priv
->cur_vs_args
->vertex
.samplers
[sampler_idx
- MAX_FRAGMENT_SAMPLERS
];
1384 if (flags
& TEX_DERIV
)
1386 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with custom derivatives\n");
1387 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with custom derivatives\n");
1388 shader_addline(buffer
, "TXD%s %s, %s, %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1389 dsx
, dsy
,sampler_idx
, tex_type
);
1391 else if(flags
& TEX_LOD
)
1393 if(flags
& TEX_PROJ
) FIXME("Projected texture sampling with explicit lod\n");
1394 if(flags
& TEX_BIAS
) FIXME("Biased texture sampling with explicit lod\n");
1395 shader_addline(buffer
, "TXL%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
,
1396 sampler_idx
, tex_type
);
1398 else if (flags
& TEX_BIAS
)
1400 /* Shouldn't be possible, but let's check for it */
1401 if(flags
& TEX_PROJ
) FIXME("Biased and Projected texture sampling\n");
1402 /* TXB takes the 4th component of the source vector automatically, as d3d. Nothing more to do */
1403 shader_addline(buffer
, "TXB%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1405 else if (flags
& TEX_PROJ
)
1407 shader_addline(buffer
, "TXP%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1413 const unsigned char idx
= priv
->cur_np2fixup_info
->super
.idx
[sampler_idx
];
1414 shader_addline(buffer
, "MUL TA, np2fixup[%u].%s, %s;\n", idx
>> 1,
1415 (idx
% 2) ? "zwxy" : "xyzw", coord_reg
);
1417 shader_addline(buffer
, "TEX%s %s, TA, texture[%u], %s;\n", mod
, dst_str
, sampler_idx
, tex_type
);
1420 shader_addline(buffer
, "TEX%s %s, %s, texture[%u], %s;\n", mod
, dst_str
, coord_reg
, sampler_idx
, tex_type
);
1425 gen_color_correction(buffer
, dst_str
, ins
->dst
[0].write_mask
,
1426 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_ONE
),
1427 arb_get_helper_value(WINED3D_SHADER_TYPE_PIXEL
, ARB_TWO
),
1428 priv
->cur_ps_args
->super
.color_fixup
[sampler_idx
]);
1432 static void shader_arb_get_src_param(const struct wined3d_shader_instruction
*ins
,
1433 const struct wined3d_shader_src_param
*src
, unsigned int tmpreg
, char *outregstr
)
1435 /* Generate a line that does the input modifier computation and return the input register to use */
1436 BOOL is_color
= FALSE
;
1440 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1441 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1442 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
1443 const char *two
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_TWO
);
1445 /* Assume a new line will be added */
1448 /* Get register name */
1449 shader_arb_get_register_name(ins
, &src
->reg
, regstr
, &is_color
);
1450 shader_arb_get_swizzle(src
, is_color
, swzstr
);
1452 switch (src
->modifiers
)
1454 case WINED3DSPSM_NONE
:
1455 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1458 case WINED3DSPSM_NEG
:
1459 sprintf(outregstr
, "-%s%s", regstr
, swzstr
);
1462 case WINED3DSPSM_BIAS
:
1463 shader_addline(buffer
, "ADD T%c, %s, -coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1465 case WINED3DSPSM_BIASNEG
:
1466 shader_addline(buffer
, "ADD T%c, -%s, coefdiv.x;\n", 'A' + tmpreg
, regstr
);
1468 case WINED3DSPSM_SIGN
:
1469 shader_addline(buffer
, "MAD T%c, %s, %s, -%s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1471 case WINED3DSPSM_SIGNNEG
:
1472 shader_addline(buffer
, "MAD T%c, %s, %s, %s;\n", 'A' + tmpreg
, regstr
, two
, one
);
1474 case WINED3DSPSM_COMP
:
1475 shader_addline(buffer
, "SUB T%c, %s, %s;\n", 'A' + tmpreg
, one
, regstr
);
1477 case WINED3DSPSM_X2
:
1478 shader_addline(buffer
, "ADD T%c, %s, %s;\n", 'A' + tmpreg
, regstr
, regstr
);
1480 case WINED3DSPSM_X2NEG
:
1481 shader_addline(buffer
, "ADD T%c, -%s, -%s;\n", 'A' + tmpreg
, regstr
, regstr
);
1483 case WINED3DSPSM_DZ
:
1484 shader_addline(buffer
, "RCP T%c, %s.z;\n", 'A' + tmpreg
, regstr
);
1485 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1487 case WINED3DSPSM_DW
:
1488 shader_addline(buffer
, "RCP T%c, %s.w;\n", 'A' + tmpreg
, regstr
);
1489 shader_addline(buffer
, "MUL T%c, %s, T%c;\n", 'A' + tmpreg
, regstr
, 'A' + tmpreg
);
1491 case WINED3DSPSM_ABS
:
1492 if(ctx
->target_version
>= NV2
) {
1493 sprintf(outregstr
, "|%s%s|", regstr
, swzstr
);
1496 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1499 case WINED3DSPSM_ABSNEG
:
1500 if(ctx
->target_version
>= NV2
) {
1501 sprintf(outregstr
, "-|%s%s|", regstr
, swzstr
);
1503 shader_addline(buffer
, "ABS T%c, %s;\n", 'A' + tmpreg
, regstr
);
1504 sprintf(outregstr
, "-T%c%s", 'A' + tmpreg
, swzstr
);
1509 sprintf(outregstr
, "%s%s", regstr
, swzstr
);
1513 /* Return modified or original register, with swizzle */
1515 sprintf(outregstr
, "T%c%s", 'A' + tmpreg
, swzstr
);
1518 static void pshader_hw_bem(const struct wined3d_shader_instruction
*ins
)
1520 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1521 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1523 char src_name
[2][50];
1524 DWORD sampler_code
= dst
->reg
.idx
;
1526 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1528 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
1530 * Keep in mind that src_name[1] can be "TB" and src_name[0] can be "TA" because modifiers like _x2 are valid
1531 * with bem. So delay loading the first parameter until after the perturbation calculation which needs two
1534 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1535 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, x, z, 0, 0;\n", sampler_code
);
1536 shader_addline(buffer
, "DP3 TC.r, TA, %s;\n", src_name
[1]);
1537 shader_addline(buffer
, "SWZ TA, bumpenvmat%d, y, w, 0, 0;\n", sampler_code
);
1538 shader_addline(buffer
, "DP3 TC.g, TA, %s;\n", src_name
[1]);
1540 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1541 shader_addline(buffer
, "ADD %s, %s, TC;\n", dst_name
, src_name
[0]);
1544 static DWORD
negate_modifiers(DWORD mod
, char *extra_char
)
1549 case WINED3DSPSM_NONE
: return WINED3DSPSM_NEG
;
1550 case WINED3DSPSM_NEG
: return WINED3DSPSM_NONE
;
1551 case WINED3DSPSM_BIAS
: return WINED3DSPSM_BIASNEG
;
1552 case WINED3DSPSM_BIASNEG
: return WINED3DSPSM_BIAS
;
1553 case WINED3DSPSM_SIGN
: return WINED3DSPSM_SIGNNEG
;
1554 case WINED3DSPSM_SIGNNEG
: return WINED3DSPSM_SIGN
;
1555 case WINED3DSPSM_COMP
: *extra_char
= '-'; return WINED3DSPSM_COMP
;
1556 case WINED3DSPSM_X2
: return WINED3DSPSM_X2NEG
;
1557 case WINED3DSPSM_X2NEG
: return WINED3DSPSM_X2
;
1558 case WINED3DSPSM_DZ
: *extra_char
= '-'; return WINED3DSPSM_DZ
;
1559 case WINED3DSPSM_DW
: *extra_char
= '-'; return WINED3DSPSM_DW
;
1560 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABSNEG
;
1561 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
1563 FIXME("Unknown modifier %u\n", mod
);
1567 static void pshader_hw_cnd(const struct wined3d_shader_instruction
*ins
)
1569 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1570 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1572 char src_name
[3][50];
1573 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1574 ins
->ctx
->reg_maps
->shader_version
.minor
);
1577 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1578 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1580 /* The coissue flag changes the semantic of the cnd instruction in <= 1.3 shaders */
1581 if (shader_version
<= WINED3D_SHADER_VERSION(1, 3) && ins
->coissue
)
1583 shader_addline(buffer
, "MOV%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[1]);
1585 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
1588 /* src0 may have a negate srcmod set, so we can't blindly add "-" to the name */
1589 src0_copy
.modifiers
= negate_modifiers(src0_copy
.modifiers
, &extra_neg
);
1591 shader_arb_get_src_param(ins
, &src0_copy
, 0, src_name
[0]);
1592 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1593 shader_addline(buffer
, "ADD TA, %c%s, coefdiv.x;\n", extra_neg
, src_name
[0]);
1594 /* No modifiers supported on CMP */
1595 shader_addline(buffer
, "CMP %s, TA, %s, %s;\n", dst_name
, src_name
[1], src_name
[2]);
1597 /* _SAT on CMP doesn't make much sense, but it is not a pure NOP */
1598 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
1600 shader_arb_get_register_name(ins
, &dst
->reg
, src_name
[0], &is_color
);
1601 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, dst_name
);
1606 static void pshader_hw_cmp(const struct wined3d_shader_instruction
*ins
)
1608 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1609 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1611 char src_name
[3][50];
1614 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1616 /* Generate input register names (with modifiers) */
1617 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1618 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1619 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1621 /* No modifiers are supported on CMP */
1622 shader_addline(buffer
, "CMP %s, %s, %s, %s;\n", dst_name
,
1623 src_name
[0], src_name
[2], src_name
[1]);
1625 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
1627 shader_arb_get_register_name(ins
, &dst
->reg
, src_name
[0], &is_color
);
1628 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name
[0]);
1632 /** Process the WINED3DSIO_DP2ADD instruction in ARB.
1633 * dst = dot2(src0, src1) + src2 */
1634 static void pshader_hw_dp2add(const struct wined3d_shader_instruction
*ins
)
1636 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1637 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1639 char src_name
[3][50];
1640 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1642 shader_arb_get_dst_param(ins
, dst
, dst_name
);
1643 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
1644 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
1646 if(ctx
->target_version
>= NV3
)
1648 /* GL_NV_fragment_program2 has a 1:1 matching instruction */
1649 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1650 shader_addline(buffer
, "DP2A%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1651 dst_name
, src_name
[0], src_name
[1], src_name
[2]);
1653 else if(ctx
->target_version
>= NV2
)
1655 /* dst.x = src2.?, src0.x, src1.x + src0.y * src1.y
1656 * dst.y = src2.?, src0.x, src1.z + src0.y * src1.w
1657 * dst.z = src2.?, src0.x, src1.x + src0.y * src1.y
1658 * dst.z = src2.?, src0.x, src1.z + src0.y * src1.w
1660 * Make sure that src1.zw = src1.xy, then we get a classic dp2add
1662 * .xyxy and other swizzles that we could get with this are not valid in
1663 * plain ARBfp, but luckily the NV extension grammar lifts this limitation.
1665 struct wined3d_shader_src_param tmp_param
= ins
->src
[1];
1666 DWORD swizzle
= tmp_param
.swizzle
& 0xf; /* Selects .xy */
1667 tmp_param
.swizzle
= swizzle
| (swizzle
<< 4); /* Creates .xyxy */
1669 shader_arb_get_src_param(ins
, &tmp_param
, 1, src_name
[1]);
1671 shader_addline(buffer
, "X2D%s %s, %s, %s, %s;\n", shader_arb_get_modifier(ins
),
1672 dst_name
, src_name
[2], src_name
[0], src_name
[1]);
1676 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
1677 /* 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]
1678 * src_name[0] can be TA, but TA is a private temp for modifiers, so it is save to overwrite
1680 shader_addline(buffer
, "MOV TA, %s;\n", src_name
[0]);
1681 shader_addline(buffer
, "MOV TA.z, 0.0;\n");
1682 shader_addline(buffer
, "DP3 TA, TA, %s;\n", src_name
[1]);
1683 shader_addline(buffer
, "ADD%s %s, TA, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
[2]);
1687 /* Map the opcode 1-to-1 to the GL code */
1688 static void shader_hw_map2gl(const struct wined3d_shader_instruction
*ins
)
1690 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1691 const char *instruction
;
1692 char arguments
[256], dst_str
[50];
1694 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1696 switch (ins
->handler_idx
)
1698 case WINED3DSIH_ABS
: instruction
= "ABS"; break;
1699 case WINED3DSIH_ADD
: instruction
= "ADD"; break;
1700 case WINED3DSIH_CRS
: instruction
= "XPD"; break;
1701 case WINED3DSIH_DP3
: instruction
= "DP3"; break;
1702 case WINED3DSIH_DP4
: instruction
= "DP4"; break;
1703 case WINED3DSIH_DST
: instruction
= "DST"; break;
1704 case WINED3DSIH_FRC
: instruction
= "FRC"; break;
1705 case WINED3DSIH_LIT
: instruction
= "LIT"; break;
1706 case WINED3DSIH_LRP
: instruction
= "LRP"; break;
1707 case WINED3DSIH_MAD
: instruction
= "MAD"; break;
1708 case WINED3DSIH_MAX
: instruction
= "MAX"; break;
1709 case WINED3DSIH_MIN
: instruction
= "MIN"; break;
1710 case WINED3DSIH_MOV
: instruction
= "MOV"; break;
1711 case WINED3DSIH_MUL
: instruction
= "MUL"; break;
1712 case WINED3DSIH_SGE
: instruction
= "SGE"; break;
1713 case WINED3DSIH_SLT
: instruction
= "SLT"; break;
1714 case WINED3DSIH_SUB
: instruction
= "SUB"; break;
1715 case WINED3DSIH_MOVA
:instruction
= "ARR"; break;
1716 case WINED3DSIH_DSX
: instruction
= "DDX"; break;
1717 default: instruction
= "";
1718 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
1722 /* Note that shader_arb_add_dst_param() adds spaces. */
1723 arguments
[0] = '\0';
1724 shader_arb_get_dst_param(ins
, dst
, dst_str
);
1725 for (i
= 0; i
< ins
->src_count
; ++i
)
1728 strcat(arguments
, ", ");
1729 shader_arb_get_src_param(ins
, &ins
->src
[i
], i
, operand
);
1730 strcat(arguments
, operand
);
1732 shader_addline(buffer
, "%s%s %s%s;\n", instruction
, shader_arb_get_modifier(ins
), dst_str
, arguments
);
1735 static void shader_hw_nop(const struct wined3d_shader_instruction
*ins
)
1737 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1738 shader_addline(buffer
, "NOP;\n");
1741 static void shader_hw_mov(const struct wined3d_shader_instruction
*ins
)
1743 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1744 BOOL pshader
= shader_is_pshader_version(shader
->baseShader
.reg_maps
.shader_version
.type
);
1745 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
1746 const char *zero
= arb_get_helper_value(shader
->baseShader
.reg_maps
.shader_version
.type
, ARB_ZERO
);
1747 const char *one
= arb_get_helper_value(shader
->baseShader
.reg_maps
.shader_version
.type
, ARB_ONE
);
1748 const char *two
= arb_get_helper_value(shader
->baseShader
.reg_maps
.shader_version
.type
, ARB_TWO
);
1750 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1751 char src0_param
[256];
1753 if(ins
->handler_idx
== WINED3DSIH_MOVA
) {
1755 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1757 if(ctx
->target_version
>= NV2
) {
1758 shader_hw_map2gl(ins
);
1761 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1762 shader_arb_get_write_mask(ins
, &ins
->dst
[0], write_mask
);
1764 /* This implements the mova formula used in GLSL. The first two instructions
1765 * prepare the sign() part. Note that it is fine to have my_sign(0.0) = 1.0
1769 * A0.x = arl(floor(abs(0.0) + 0.5) * 1.0) = floor(0.5) = 0.0 since arl does a floor
1771 * The ARL is performed when A0 is used - the requested component is read from A0_SHADOW into
1772 * A0.x. We can use the overwritten component of A0_shadow as temporary storage for the sign.
1774 shader_addline(buffer
, "SGE A0_SHADOW%s, %s, %s;\n", write_mask
, src0_param
, zero
);
1775 shader_addline(buffer
, "MAD A0_SHADOW%s, A0_SHADOW, %s, -%s;\n", write_mask
, two
, one
);
1777 shader_addline(buffer
, "ABS TA%s, %s;\n", write_mask
, src0_param
);
1778 shader_addline(buffer
, "ADD TA%s, TA, rel_addr_const.x;\n", write_mask
);
1779 shader_addline(buffer
, "FLR TA%s, TA;\n", write_mask
);
1780 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
)
1782 shader_addline(buffer
, "ADD TA%s, TA, %s;\n", write_mask
, offset
);
1784 shader_addline(buffer
, "MUL A0_SHADOW%s, TA, A0_SHADOW;\n", write_mask
);
1786 ((struct shader_arb_ctx_priv
*)ins
->ctx
->backend_data
)->addr_reg
[0] = '\0';
1787 } else if (ins
->ctx
->reg_maps
->shader_version
.major
== 1
1788 && !shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)
1789 && ins
->dst
[0].reg
.type
== WINED3DSPR_ADDR
)
1791 src0_param
[0] = '\0';
1792 if (((IWineD3DVertexShaderImpl
*)shader
)->rel_offset
)
1794 const char *offset
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_VS_REL_OFFSET
);
1795 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_param
);
1796 shader_addline(buffer
, "ADD TA.x, %s, %s;\n", src0_param
, offset
);
1797 shader_addline(buffer
, "ARL A0.x, TA.x;\n");
1801 /* Apple's ARB_vertex_program implementation does not accept an ARL source argument
1802 * with more than one component. Thus replicate the first source argument over all
1803 * 4 components. For example, .xyzw -> .x (or better: .xxxx), .zwxy -> .z, etc) */
1804 struct wined3d_shader_src_param tmp_src
= ins
->src
[0];
1805 tmp_src
.swizzle
= (tmp_src
.swizzle
& 0x3) * 0x55;
1806 shader_arb_get_src_param(ins
, &tmp_src
, 0, src0_param
);
1807 shader_addline(buffer
, "ARL A0.x, %s;\n", src0_param
);
1810 else if(ins
->dst
[0].reg
.type
== WINED3DSPR_COLOROUT
&& ins
->dst
[0].reg
.idx
== 0 && pshader
)
1812 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) shader
;
1813 if(ctx
->cur_ps_args
->super
.srgb_correction
&& ps
->color0_mov
)
1815 shader_addline(buffer
, "#mov handled in srgb write code\n");
1818 shader_hw_map2gl(ins
);
1822 shader_hw_map2gl(ins
);
1826 static void pshader_hw_texkill(const struct wined3d_shader_instruction
*ins
)
1828 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1829 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1832 /* No swizzles are allowed in d3d's texkill. PS 1.x ignores the 4th component as documented,
1833 * but >= 2.0 honors it(undocumented, but tested by the d3d9 testsuit)
1835 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1837 if (ins
->ctx
->reg_maps
->shader_version
.major
>= 2)
1839 const char *kilsrc
= "TA";
1842 shader_arb_get_register_name(ins
, &dst
->reg
, reg_dest
, &is_color
);
1843 if(dst
->write_mask
== WINED3DSP_WRITEMASK_ALL
)
1849 /* Sigh. KIL doesn't support swizzles/writemasks. KIL passes a writemask, but ".xy" for example
1850 * is not valid as a swizzle in ARB (needs ".xyyy"). Use SWZ to load the register properly, and set
1851 * masked out components to 0(won't kill)
1853 char x
= '0', y
= '0', z
= '0', w
= '0';
1854 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) x
= 'x';
1855 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) y
= 'y';
1856 if(dst
->write_mask
& WINED3DSP_WRITEMASK_2
) z
= 'z';
1857 if(dst
->write_mask
& WINED3DSP_WRITEMASK_3
) w
= 'w';
1858 shader_addline(buffer
, "SWZ TA, %s, %c, %c, %c, %c;\n", reg_dest
, x
, y
, z
, w
);
1860 shader_addline(buffer
, "KIL %s;\n", kilsrc
);
1862 /* ARB fp doesn't like swizzles on the parameter of the KIL instruction. To mask the 4th component,
1863 * copy the register into our general purpose TMP variable, overwrite .w and pass TMP to KIL
1865 * ps_1_3 shaders use the texcoord incarnation of the Tx register. ps_1_4 shaders can use the same,
1866 * or pass in any temporary register(in shader phase 2)
1868 if(ins
->ctx
->reg_maps
->shader_version
.minor
<= 3) {
1869 sprintf(reg_dest
, "fragment.texcoord[%u]", dst
->reg
.idx
);
1871 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1873 shader_addline(buffer
, "SWZ TA, %s, x, y, z, 1;\n", reg_dest
);
1874 shader_addline(buffer
, "KIL TA;\n");
1878 static void pshader_hw_tex(const struct wined3d_shader_instruction
*ins
)
1880 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1881 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
1882 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1883 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1884 ins
->ctx
->reg_maps
->shader_version
.minor
);
1885 struct wined3d_shader_src_param src
;
1889 DWORD reg_sampler_code
;
1892 /* All versions have a destination register */
1893 shader_arb_get_dst_param(ins
, dst
, reg_dest
);
1895 /* 1.0-1.4: Use destination register number as texture code.
1896 2.0+: Use provided sampler number as texure code. */
1897 if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1898 reg_sampler_code
= dst
->reg
.idx
;
1900 reg_sampler_code
= ins
->src
[1].reg
.idx
;
1902 /* 1.0-1.3: Use the texcoord varying.
1903 1.4+: Use provided coordinate source register. */
1904 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1905 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_sampler_code
);
1907 /* TEX is the only instruction that can handle DW and DZ natively */
1909 if(src
.modifiers
== WINED3DSPSM_DW
) src
.modifiers
= WINED3DSPSM_NONE
;
1910 if(src
.modifiers
== WINED3DSPSM_DZ
) src
.modifiers
= WINED3DSPSM_NONE
;
1911 shader_arb_get_src_param(ins
, &src
, 0, reg_coord
);
1915 * 1.1, 1.2, 1.3: Use WINED3DTSS_TEXTURETRANSFORMFLAGS
1916 * 1.4: Use WINED3DSPSM_DZ or WINED3DSPSM_DW on src[0]
1917 * 2.0+: Use WINED3DSI_TEXLD_PROJECT on the opcode
1919 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1922 if(reg_sampler_code
< MAX_TEXTURES
) {
1923 flags
= deviceImpl
->stateBlock
->textureState
[reg_sampler_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
];
1925 if (flags
& WINED3DTTFF_PROJECTED
) {
1926 myflags
|= TEX_PROJ
;
1929 else if (shader_version
< WINED3D_SHADER_VERSION(2,0))
1931 DWORD src_mod
= ins
->src
[0].modifiers
;
1932 if (src_mod
== WINED3DSPSM_DZ
) {
1933 /* TXP cannot handle DZ natively, so move the z coordinate to .w. reg_coord is a read-only
1934 * varying register, so we need a temp reg
1936 shader_addline(ins
->ctx
->buffer
, "SWZ TA, %s, x, y, z, z;\n", reg_coord
);
1937 strcpy(reg_coord
, "TA");
1938 myflags
|= TEX_PROJ
;
1939 } else if(src_mod
== WINED3DSPSM_DW
) {
1940 myflags
|= TEX_PROJ
;
1943 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) myflags
|= TEX_PROJ
;
1944 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) myflags
|= TEX_BIAS
;
1946 shader_hw_sample(ins
, reg_sampler_code
, reg_dest
, reg_coord
, myflags
, NULL
, NULL
);
1949 static void pshader_hw_texcoord(const struct wined3d_shader_instruction
*ins
)
1951 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
1952 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1953 DWORD shader_version
= WINED3D_SHADER_VERSION(ins
->ctx
->reg_maps
->shader_version
.major
,
1954 ins
->ctx
->reg_maps
->shader_version
.minor
);
1957 if (shader_version
< WINED3D_SHADER_VERSION(1,4))
1959 DWORD reg
= dst
->reg
.idx
;
1961 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1962 shader_addline(buffer
, "MOV_SAT %s, fragment.texcoord[%u];\n", dst_str
, reg
);
1966 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
);
1967 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1968 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, reg_src
);
1972 static void pshader_hw_texreg2ar(const struct wined3d_shader_instruction
*ins
)
1974 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1975 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
1976 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
1979 DWORD reg1
= ins
->dst
[0].reg
.idx
;
1983 /* Note that texreg2ar treats Tx as a temporary register, not as a varying */
1984 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
1985 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
1986 /* Move .x first in case src_str is "TA" */
1987 shader_addline(buffer
, "MOV TA.y, %s.x;\n", src_str
);
1988 shader_addline(buffer
, "MOV TA.x, %s.w;\n", src_str
);
1989 flags
= reg1
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg1
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
1990 shader_hw_sample(ins
, reg1
, dst_str
, "TA", flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
1993 static void pshader_hw_texreg2gb(const struct wined3d_shader_instruction
*ins
)
1995 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
1997 DWORD reg1
= ins
->dst
[0].reg
.idx
;
2001 /* Note that texreg2gb treats Tx as a temporary register, not as a varying */
2002 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2003 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2004 shader_addline(buffer
, "MOV TA.x, %s.y;\n", src_str
);
2005 shader_addline(buffer
, "MOV TA.y, %s.z;\n", src_str
);
2006 shader_hw_sample(ins
, reg1
, dst_str
, "TA", 0, NULL
, NULL
);
2009 static void pshader_hw_texreg2rgb(const struct wined3d_shader_instruction
*ins
)
2011 DWORD reg1
= ins
->dst
[0].reg
.idx
;
2015 /* Note that texreg2rg treats Tx as a temporary register, not as a varying */
2016 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2017 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_str
);
2018 shader_hw_sample(ins
, reg1
, dst_str
, src_str
, 0, NULL
, NULL
);
2021 static void pshader_hw_texbem(const struct wined3d_shader_instruction
*ins
)
2023 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2024 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
2025 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2026 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2027 char reg_coord
[40], dst_reg
[50], src_reg
[50];
2028 DWORD reg_dest_code
;
2030 /* All versions have a destination register. The Tx where the texture coordinates come
2031 * from is the varying incarnation of the texture register
2033 reg_dest_code
= dst
->reg
.idx
;
2034 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_reg
);
2035 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_reg
);
2036 sprintf(reg_coord
, "fragment.texcoord[%u]", reg_dest_code
);
2038 /* Sampling the perturbation map in Tsrc was done already, including the signedness correction if needed
2039 * The Tx in which the perturbation map is stored is the tempreg incarnation of the texture register
2041 * GL_NV_fragment_program_option could handle this in one instruction via X2D:
2042 * X2D TA.xy, fragment.texcoord, T%u, bumpenvmat%u.xzyw
2044 * However, the NV extensions are never enabled for <= 2.0 shaders because of the performance penalty that
2045 * comes with it, and texbem is an 1.x only instruction. No 1.x instruction forces us to enable the NV
2048 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, x, z, 0, 0;\n", reg_dest_code
);
2049 shader_addline(buffer
, "DP3 TA.x, TB, %s;\n", src_reg
);
2050 shader_addline(buffer
, "SWZ TB, bumpenvmat%d, y, w, 0, 0;\n", reg_dest_code
);
2051 shader_addline(buffer
, "DP3 TA.y, TB, %s;\n", src_reg
);
2053 /* with projective textures, texbem only divides the static texture coord, not the displacement,
2054 * so we can't let the GL handle this.
2056 if (device
->stateBlock
->textureState
[reg_dest_code
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] & WINED3DTTFF_PROJECTED
)
2058 shader_addline(buffer
, "RCP TB.w, %s.w;\n", reg_coord
);
2059 shader_addline(buffer
, "MUL TB.xy, %s, TB.w;\n", reg_coord
);
2060 shader_addline(buffer
, "ADD TA.xy, TA, TB;\n");
2062 shader_addline(buffer
, "ADD TA.xy, TA, %s;\n", reg_coord
);
2065 shader_hw_sample(ins
, reg_dest_code
, dst_reg
, "TA", 0, NULL
, NULL
);
2067 if (ins
->handler_idx
== WINED3DSIH_TEXBEML
)
2069 /* No src swizzles are allowed, so this is ok */
2070 shader_addline(buffer
, "MAD TA, %s.z, luminance%d.x, luminance%d.y;\n",
2071 src_reg
, reg_dest_code
, reg_dest_code
);
2072 shader_addline(buffer
, "MUL %s, %s, TA;\n", dst_reg
, dst_reg
);
2076 static void pshader_hw_texm3x2pad(const struct wined3d_shader_instruction
*ins
)
2078 DWORD reg
= ins
->dst
[0].reg
.idx
;
2079 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2080 char src0_name
[50], dst_name
[50];
2082 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2084 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2085 /* The next instruction will be a texm3x2tex or texm3x2depth that writes to the uninitialized
2086 * T<reg+1> register. Use this register to store the calculated vector
2088 tmp_reg
.idx
= reg
+ 1;
2089 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2090 shader_addline(buffer
, "DP3 %s.x, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2093 static void pshader_hw_texm3x2tex(const struct wined3d_shader_instruction
*ins
)
2095 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2096 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
2098 DWORD reg
= ins
->dst
[0].reg
.idx
;
2099 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2105 /* We know that we're writing to the uninitialized T<reg> register, so use it for temporary storage */
2106 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2108 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2109 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2110 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2111 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
2112 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2115 static void pshader_hw_texm3x3pad(const struct wined3d_shader_instruction
*ins
)
2117 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2118 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
2119 DWORD reg
= ins
->dst
[0].reg
.idx
;
2120 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2121 char src0_name
[50], dst_name
[50];
2122 struct wined3d_shader_register tmp_reg
= ins
->dst
[0].reg
;
2125 /* There are always 2 texm3x3pad instructions followed by one texm3x3[tex,vspec, ...] instruction, with
2126 * incrementing ins->dst[0].register_idx numbers. So the pad instruction already knows the final destination
2127 * register, and this register is uninitialized(otherwise the assembler complains that it is 'redeclared')
2129 tmp_reg
.idx
= reg
+ 2 - current_state
->current_row
;
2130 shader_arb_get_register_name(ins
, &tmp_reg
, dst_name
, &is_color
);
2132 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2133 shader_addline(buffer
, "DP3 %s.%c, fragment.texcoord[%u], %s;\n",
2134 dst_name
, 'x' + current_state
->current_row
, reg
, src0_name
);
2135 current_state
->texcoord_w
[current_state
->current_row
++] = reg
;
2138 static void pshader_hw_texm3x3tex(const struct wined3d_shader_instruction
*ins
)
2140 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2141 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
2142 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
2144 DWORD reg
= ins
->dst
[0].reg
.idx
;
2145 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2147 char src0_name
[50], dst_name
[50];
2150 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2151 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2152 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, reg
, src0_name
);
2154 /* Sample the texture using the calculated coordinates */
2155 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2156 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
2157 shader_hw_sample(ins
, reg
, dst_str
, dst_name
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2158 current_state
->current_row
= 0;
2161 static void pshader_hw_texm3x3vspec(const struct wined3d_shader_instruction
*ins
)
2163 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2164 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
2165 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
2167 DWORD reg
= ins
->dst
[0].reg
.idx
;
2168 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2174 /* Get the dst reg without writemask strings. We know this register is uninitialized, so we can use all
2175 * components for temporary data storage
2177 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2178 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2179 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2181 /* Construct the eye-ray vector from w coordinates */
2182 shader_addline(buffer
, "MOV TB.x, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[0]);
2183 shader_addline(buffer
, "MOV TB.y, fragment.texcoord[%u].w;\n", current_state
->texcoord_w
[1]);
2184 shader_addline(buffer
, "MOV TB.z, fragment.texcoord[%u].w;\n", reg
);
2186 /* Calculate reflection vector
2188 shader_addline(buffer
, "DP3 %s.w, %s, TB;\n", dst_reg
, dst_reg
);
2189 /* The .w is ignored when sampling, so I can use TB.w to calculate dot(N, N) */
2190 shader_addline(buffer
, "DP3 TB.w, %s, %s;\n", dst_reg
, dst_reg
);
2191 shader_addline(buffer
, "RCP TB.w, TB.w;\n");
2192 shader_addline(buffer
, "MUL %s.w, %s.w, TB.w;\n", dst_reg
, dst_reg
);
2193 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2194 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -TB;\n", dst_reg
, dst_reg
);
2196 /* Sample the texture using the calculated coordinates */
2197 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2198 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
2199 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2200 current_state
->current_row
= 0;
2203 static void pshader_hw_texm3x3spec(const struct wined3d_shader_instruction
*ins
)
2205 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
2206 IWineD3DDeviceImpl
*deviceImpl
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
2207 SHADER_PARSE_STATE
*current_state
= &shader
->baseShader
.parse_state
;
2209 DWORD reg
= ins
->dst
[0].reg
.idx
;
2210 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2217 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0_name
);
2218 shader_arb_get_src_param(ins
, &ins
->src
[0], 1, src1_name
);
2219 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_reg
, &is_color
);
2220 /* Note: dst_reg.xy is input here, generated by two texm3x3pad instructions */
2221 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_reg
, reg
, src0_name
);
2223 /* Calculate reflection vector.
2226 * dst_reg.xyz = 2 * --------- * N - E
2229 * Which normalizes the normal vector
2231 shader_addline(buffer
, "DP3 %s.w, %s, %s;\n", dst_reg
, dst_reg
, src1_name
);
2232 shader_addline(buffer
, "DP3 TC.w, %s, %s;\n", dst_reg
, dst_reg
);
2233 shader_addline(buffer
, "RCP TC.w, TC.w;\n");
2234 shader_addline(buffer
, "MUL %s.w, %s.w, TC.w;\n", dst_reg
, dst_reg
);
2235 shader_addline(buffer
, "MUL %s, %s.w, %s;\n", dst_reg
, dst_reg
, dst_reg
);
2236 shader_addline(buffer
, "MAD %s, coefmul.x, %s, -%s;\n", dst_reg
, dst_reg
, src1_name
);
2238 /* Sample the texture using the calculated coordinates */
2239 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2240 flags
= reg
< MAX_TEXTURES
? deviceImpl
->stateBlock
->textureState
[reg
][WINED3DTSS_TEXTURETRANSFORMFLAGS
] : 0;
2241 shader_hw_sample(ins
, reg
, dst_str
, dst_reg
, flags
& WINED3DTTFF_PROJECTED
? TEX_PROJ
: 0, NULL
, NULL
);
2242 current_state
->current_row
= 0;
2245 static void pshader_hw_texdepth(const struct wined3d_shader_instruction
*ins
)
2247 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2248 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2250 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2251 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2253 /* texdepth has an implicit destination, the fragment depth value. It's only parameter,
2254 * which is essentially an input, is the destination register because it is the first
2255 * parameter. According to the msdn, this must be register r5, but let's keep it more flexible
2256 * here(writemasks/swizzles are not valid on texdepth)
2258 shader_arb_get_dst_param(ins
, dst
, dst_name
);
2260 /* According to the msdn, the source register(must be r5) is unusable after
2261 * the texdepth instruction, so we're free to modify it
2263 shader_addline(buffer
, "MIN %s.y, %s.y, %s;\n", dst_name
, dst_name
, one
);
2265 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2266 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2267 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2269 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2270 shader_addline(buffer
, "MUL TA.x, %s.x, %s.y;\n", dst_name
, dst_name
);
2271 shader_addline(buffer
, "MIN TA.x, TA.x, %s;\n", one
);
2272 shader_addline(buffer
, "MAX result.depth, TA.x, %s;\n", zero
);
2275 /** Process the WINED3DSIO_TEXDP3TEX instruction in ARB:
2276 * Take a 3-component dot product of the TexCoord[dstreg] and src,
2277 * then perform a 1D texture lookup from stage dstregnum, place into dst. */
2278 static void pshader_hw_texdp3tex(const struct wined3d_shader_instruction
*ins
)
2280 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2281 DWORD sampler_idx
= ins
->dst
[0].reg
.idx
;
2285 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2286 shader_addline(buffer
, "MOV TB, 0.0;\n");
2287 shader_addline(buffer
, "DP3 TB.x, fragment.texcoord[%u], %s;\n", sampler_idx
, src0
);
2289 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_str
);
2290 shader_hw_sample(ins
, sampler_idx
, dst_str
, "TB", 0 /* Only one coord, can't be projected */, NULL
, NULL
);
2293 /** Process the WINED3DSIO_TEXDP3 instruction in ARB:
2294 * Take a 3-component dot product of the TexCoord[dstreg] and src. */
2295 static void pshader_hw_texdp3(const struct wined3d_shader_instruction
*ins
)
2297 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2300 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2302 /* Handle output register */
2303 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2304 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2305 shader_addline(buffer
, "DP3 %s, fragment.texcoord[%u], %s;\n", dst_str
, dst
->reg
.idx
, src0
);
2308 /** Process the WINED3DSIO_TEXM3X3 instruction in ARB
2309 * Perform the 3rd row of a 3x3 matrix multiply */
2310 static void pshader_hw_texm3x3(const struct wined3d_shader_instruction
*ins
)
2312 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2313 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2314 char dst_str
[50], dst_name
[50];
2318 shader_arb_get_dst_param(ins
, dst
, dst_str
);
2319 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2320 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2321 shader_addline(buffer
, "DP3 %s.z, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
, src0
);
2322 shader_addline(buffer
, "MOV %s, %s;\n", dst_str
, dst_name
);
2325 /** Process the WINED3DSIO_TEXM3X2DEPTH instruction in ARB:
2326 * Last row of a 3x2 matrix multiply, use the result to calculate the depth:
2327 * Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
2328 * depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
2330 static void pshader_hw_texm3x2depth(const struct wined3d_shader_instruction
*ins
)
2332 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2333 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2334 char src0
[50], dst_name
[50];
2336 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2337 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2339 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src0
);
2340 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2341 shader_addline(buffer
, "DP3 %s.y, fragment.texcoord[%u], %s;\n", dst_name
, dst
->reg
.idx
, src0
);
2343 /* How to deal with the special case dst_name.g == 0? if r != 0, then
2344 * the r * (1 / 0) will give infinity, which is clamped to 1.0, the correct
2345 * result. But if r = 0.0, then 0 * inf = 0, which is incorrect.
2347 shader_addline(buffer
, "RCP %s.y, %s.y;\n", dst_name
, dst_name
);
2348 shader_addline(buffer
, "MUL %s.x, %s.x, %s.y;\n", dst_name
, dst_name
, dst_name
);
2349 shader_addline(buffer
, "MIN %s.x, %s.x, %s;\n", dst_name
, dst_name
, one
);
2350 shader_addline(buffer
, "MAX result.depth, %s.x, %s;\n", dst_name
, zero
);
2353 /** Handles transforming all WINED3DSIO_M?x? opcodes for
2354 Vertex/Pixel shaders to ARB_vertex_program codes */
2355 static void shader_hw_mnxn(const struct wined3d_shader_instruction
*ins
)
2358 int nComponents
= 0;
2359 struct wined3d_shader_dst_param tmp_dst
= {{0}};
2360 struct wined3d_shader_src_param tmp_src
[2] = {{{0}}};
2361 struct wined3d_shader_instruction tmp_ins
;
2363 memset(&tmp_ins
, 0, sizeof(tmp_ins
));
2365 /* Set constants for the temporary argument */
2366 tmp_ins
.ctx
= ins
->ctx
;
2367 tmp_ins
.dst_count
= 1;
2368 tmp_ins
.dst
= &tmp_dst
;
2369 tmp_ins
.src_count
= 2;
2370 tmp_ins
.src
= tmp_src
;
2372 switch(ins
->handler_idx
)
2374 case WINED3DSIH_M4x4
:
2376 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2378 case WINED3DSIH_M4x3
:
2380 tmp_ins
.handler_idx
= WINED3DSIH_DP4
;
2382 case WINED3DSIH_M3x4
:
2384 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2386 case WINED3DSIH_M3x3
:
2388 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2390 case WINED3DSIH_M3x2
:
2392 tmp_ins
.handler_idx
= WINED3DSIH_DP3
;
2395 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2399 tmp_dst
= ins
->dst
[0];
2400 tmp_src
[0] = ins
->src
[0];
2401 tmp_src
[1] = ins
->src
[1];
2402 for (i
= 0; i
< nComponents
; i
++) {
2403 tmp_dst
.write_mask
= WINED3DSP_WRITEMASK_0
<< i
;
2404 shader_hw_map2gl(&tmp_ins
);
2405 ++tmp_src
[1].reg
.idx
;
2409 static void shader_hw_rcp(const struct wined3d_shader_instruction
*ins
)
2411 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2412 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2413 const char *flt_eps
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_EPS
);
2418 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2419 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2420 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2422 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2428 /* TODO: If the destination is readable, and not the same as the source, the destination
2429 * can be used instead of TA
2431 if (priv
->target_version
>= NV2
)
2433 shader_addline(buffer
, "MOVC TA.x, %s;\n", src
);
2434 shader_addline(buffer
, "MOV TA.x (EQ.x), %s;\n", flt_eps
);
2435 shader_addline(buffer
, "RCP%s %s, TA.x;\n", shader_arb_get_modifier(ins
), dst
);
2439 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2440 shader_addline(buffer
, "ABS TA.x, %s;\n", src
);
2441 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2442 shader_addline(buffer
, "MAD TA.x, TA.y, %s, %s;\n", flt_eps
, src
);
2443 shader_addline(buffer
, "RCP%s %s, TA.x;\n", shader_arb_get_modifier(ins
), dst
);
2447 static void shader_hw_scalar_op(const struct wined3d_shader_instruction
*ins
)
2449 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2450 const char *instruction
;
2455 switch(ins
->handler_idx
)
2457 case WINED3DSIH_RSQ
: instruction
= "RSQ"; break;
2458 case WINED3DSIH_RCP
: instruction
= "RCP"; break;
2459 case WINED3DSIH_EXP
: instruction
= "EX2"; break;
2460 case WINED3DSIH_EXPP
: instruction
= "EXP"; break;
2461 default: instruction
= "";
2462 FIXME("Unhandled opcode %#x\n", ins
->handler_idx
);
2466 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
); /* Destination */
2467 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2468 if (ins
->src
[0].swizzle
== WINED3DSP_NOSWIZZLE
)
2470 /* Dx sdk says .x is used if no swizzle is given, but our test shows that
2476 shader_addline(buffer
, "%s%s %s, %s;\n", instruction
, shader_arb_get_modifier(ins
), dst
, src
);
2479 static void shader_hw_nrm(const struct wined3d_shader_instruction
*ins
)
2481 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2484 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2485 BOOL pshader
= shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2486 const char *zero
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ZERO
);
2488 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2489 shader_arb_get_src_param(ins
, &ins
->src
[0], 1 /* Use TB */, src_name
);
2491 /* In D3D, NRM of a vector with length zero returns zero. Catch this situation, as
2492 * otherwise NRM or RSQ would return NaN */
2493 if(pshader
&& priv
->target_version
>= NV3
)
2495 /* GL_NV_fragment_program2's NRM needs protection against length zero vectors too
2497 * TODO: Find out if DP3+NRM+MOV is really faster than DP3+RSQ+MUL
2499 shader_addline(buffer
, "DP3C TA, %s, %s;\n", src_name
, src_name
);
2500 shader_addline(buffer
, "NRM%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2501 shader_addline(buffer
, "MOV %s (EQ), %s;\n", dst_name
, zero
);
2503 else if(priv
->target_version
>= NV2
)
2505 shader_addline(buffer
, "DP3C TA.x, %s, %s;\n", src_name
, src_name
);
2506 shader_addline(buffer
, "RSQ TA.x (NE), TA.x;\n");
2507 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2512 const char *one
= arb_get_helper_value(ins
->ctx
->reg_maps
->shader_version
.type
, ARB_ONE
);
2514 shader_addline(buffer
, "DP3 TA.x, %s, %s;\n", src_name
, src_name
);
2515 /* Pass any non-zero value to RSQ if the input vector has a length of zero. The
2516 * RSQ result doesn't matter, as long as multiplying it by 0 returns 0.
2518 shader_addline(buffer
, "SGE TA.y, -TA.x, %s;\n", zero
);
2519 shader_addline(buffer
, "MAD TA.x, %s, TA.y, TA.x;\n", one
);
2521 shader_addline(buffer
, "RSQ TA.x, TA.x;\n");
2522 /* dst.w = src[0].w * 1 / (src.x^2 + src.y^2 + src.z^2)^(1/2) according to msdn*/
2523 shader_addline(buffer
, "MUL%s %s, %s, TA.x;\n", shader_arb_get_modifier(ins
), dst_name
,
2528 static void shader_hw_lrp(const struct wined3d_shader_instruction
*ins
)
2530 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2532 char src_name
[3][50];
2534 /* ARB_fragment_program has a convenient LRP instruction */
2535 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2536 shader_hw_map2gl(ins
);
2540 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2541 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
[0]);
2542 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name
[1]);
2543 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name
[2]);
2545 shader_addline(buffer
, "SUB TA, %s, %s;\n", src_name
[1], src_name
[2]);
2546 shader_addline(buffer
, "MAD%s %s, %s, TA, %s;\n", shader_arb_get_modifier(ins
),
2547 dst_name
, src_name
[0], src_name
[2]);
2550 static void shader_hw_sincos(const struct wined3d_shader_instruction
*ins
)
2552 /* This instruction exists in ARB, but the d3d instruction takes two extra parameters which
2553 * must contain fixed constants. So we need a separate function to filter those constants and
2556 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2557 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2558 const struct wined3d_shader_dst_param
*dst
= &ins
->dst
[0];
2560 char src_name0
[50], src_name1
[50], src_name2
[50];
2563 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2564 if(shader_is_pshader_version(ins
->ctx
->reg_maps
->shader_version
.type
)) {
2565 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2566 /* No modifiers are supported on SCS */
2567 shader_addline(buffer
, "SCS %s, %s;\n", dst_name
, src_name0
);
2569 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
)
2571 shader_arb_get_register_name(ins
, &dst
->reg
, src_name0
, &is_color
);
2572 shader_addline(buffer
, "MOV_SAT %s, %s;\n", dst_name
, src_name0
);
2574 } else if(priv
->target_version
>= NV2
) {
2575 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2577 /* Sincos writemask must be .x, .y or .xy */
2578 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
)
2579 shader_addline(buffer
, "COS%s %s.x, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2580 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
)
2581 shader_addline(buffer
, "SIN%s %s.y, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name0
);
2583 /* Approximate sine and cosine with a taylor series, as per math textbook. The application passes 8
2584 * helper constants(D3DSINCOSCONST1 and D3DSINCOSCONST2) in src1 and src2.
2586 * sin(x) = x - x^3/3! + x^5/5! - x^7/7! + ...
2587 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6! + ...
2589 * The constants we get are:
2591 * +1 +1, -1 -1 +1 +1 -1 -1
2592 * ---- , ---- , ---- , ----- , ----- , ----- , ------
2593 * 1!*2 2!*4 3!*8 4!*16 5!*32 6!*64 7!*128
2595 * If used with x^2, x^3, x^4 etc they calculate sin(x/2) and cos(x/2):
2599 * (x/2)^4 = x^4 / 16
2600 * (x/2)^5 = x^5 / 32
2603 * To get the final result:
2604 * sin(x) = 2 * sin(x/2) * cos(x/2)
2605 * cos(x) = cos(x/2)^2 - sin(x/2)^2
2606 * (from sin(x+y) and cos(x+y) rules)
2608 * As per MSDN, dst.z is undefined after the operation, and so is
2609 * dst.x and dst.y if they're masked out by the writemask. Ie
2610 * sincos dst.y, src1, c0, c1
2611 * returns the sine in dst.y. dst.x and dst.z are undefined, dst.w is not touched. The assembler
2612 * vsa.exe also stops with an error if the dest register is the same register as the source
2613 * register. This means we can use dest.xyz as temporary storage. The assembler vsa.exe output also
2614 * indicates that sincos consumes 8 instruction slots in vs_2_0(and, strangely, in vs_3_0).
2616 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2617 shader_arb_get_src_param(ins
, &ins
->src
[2], 2, src_name2
);
2618 shader_arb_get_register_name(ins
, &dst
->reg
, dst_name
, &is_color
);
2620 shader_addline(buffer
, "MUL %s.x, %s, %s;\n", dst_name
, src_name0
, src_name0
); /* x ^ 2 */
2621 shader_addline(buffer
, "MUL TA.y, %s.x, %s;\n", dst_name
, src_name0
); /* x ^ 3 */
2622 shader_addline(buffer
, "MUL %s.y, TA.y, %s;\n", dst_name
, src_name0
); /* x ^ 4 */
2623 shader_addline(buffer
, "MUL TA.z, %s.y, %s;\n", dst_name
, src_name0
); /* x ^ 5 */
2624 shader_addline(buffer
, "MUL %s.z, TA.z, %s;\n", dst_name
, src_name0
); /* x ^ 6 */
2625 shader_addline(buffer
, "MUL TA.w, %s.z, %s;\n", dst_name
, src_name0
); /* x ^ 7 */
2629 * Unfortunately we don't get the constants in a DP4-capable form. Is there a way to
2630 * properly merge that with MULs in the code above?
2631 * The swizzles .yz and xw however fit into the .yzxw swizzle added to ps_2_0. Maybe
2632 * we can merge the sine and cosine MAD rows to calculate them together.
2634 shader_addline(buffer
, "MUL TA.x, %s, %s.w;\n", src_name0
, src_name2
); /* x^1, +1/(1!*2) */
2635 shader_addline(buffer
, "MAD TA.x, TA.y, %s.x, TA.x;\n", src_name2
); /* -1/(3!*8) */
2636 shader_addline(buffer
, "MAD TA.x, TA.z, %s.w, TA.x;\n", src_name1
); /* +1/(5!*32) */
2637 shader_addline(buffer
, "MAD TA.x, TA.w, %s.x, TA.x;\n", src_name1
); /* -1/(7!*128) */
2640 shader_addline(buffer
, "MAD TA.y, %s.x, %s.y, %s.z;\n", dst_name
, src_name2
, src_name2
); /* -1/(2!*4), +1.0 */
2641 shader_addline(buffer
, "MAD TA.y, %s.y, %s.z, TA.y;\n", dst_name
, src_name1
); /* +1/(4!*16) */
2642 shader_addline(buffer
, "MAD TA.y, %s.z, %s.y, TA.y;\n", dst_name
, src_name1
); /* -1/(6!*64) */
2644 if(dst
->write_mask
& WINED3DSP_WRITEMASK_0
) {
2646 shader_addline(buffer
, "MUL TA.z, TA.y, TA.y;\n");
2647 shader_addline(buffer
, "MAD %s.x, -TA.x, TA.x, TA.z;\n", dst_name
);
2649 if(dst
->write_mask
& WINED3DSP_WRITEMASK_1
) {
2651 shader_addline(buffer
, "MUL %s.y, TA.x, TA.y;\n", dst_name
);
2652 shader_addline(buffer
, "ADD %s.y, %s.y, %s.y;\n", dst_name
, dst_name
, dst_name
);
2657 static void shader_hw_sgn(const struct wined3d_shader_instruction
*ins
)
2659 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2662 struct shader_arb_ctx_priv
*ctx
= ins
->ctx
->backend_data
;
2664 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst_name
);
2665 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2667 /* SGN is only valid in vertex shaders */
2668 if(ctx
->target_version
>= NV2
) {
2669 shader_addline(buffer
, "SSG%s %s, %s;\n", shader_arb_get_modifier(ins
), dst_name
, src_name
);
2673 /* If SRC > 0.0, -SRC < SRC = TRUE, otherwise false.
2674 * if SRC < 0.0, SRC < -SRC = TRUE. If neither is true, src = 0.0
2676 if(ins
->dst
[0].modifiers
& WINED3DSPDM_SATURATE
) {
2677 shader_addline(buffer
, "SLT %s, -%s, %s;\n", dst_name
, src_name
, src_name
);
2679 /* src contains TA? Write to the dest first. This won't overwrite our destination.
2680 * Then use TA, and calculate the final result
2682 * Not reading from TA? Store the first result in TA to avoid overwriting the
2683 * destination if src reg = dst reg
2685 if(strstr(src_name
, "TA"))
2687 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2688 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2689 shader_addline(buffer
, "ADD %s, %s, -TA;\n", dst_name
, dst_name
);
2693 shader_addline(buffer
, "SLT TA, -%s, %s;\n", src_name
, src_name
);
2694 shader_addline(buffer
, "SLT %s, %s, -%s;\n", dst_name
, src_name
, src_name
);
2695 shader_addline(buffer
, "ADD %s, TA, -%s;\n", dst_name
, dst_name
);
2700 static void shader_hw_dsy(const struct wined3d_shader_instruction
*ins
)
2702 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2708 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2709 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src
);
2710 shader_arb_get_register_name(ins
, &ins
->dst
[0].reg
, dst_name
, &is_color
);
2712 shader_addline(buffer
, "DDY %s, %s;\n", dst
, src
);
2713 shader_addline(buffer
, "MUL%s %s, %s, ycorrection.y;\n", shader_arb_get_modifier(ins
), dst
, dst_name
);
2716 static DWORD
abs_modifier(DWORD mod
, BOOL
*need_abs
)
2722 case WINED3DSPSM_NONE
: return WINED3DSPSM_ABS
;
2723 case WINED3DSPSM_NEG
: return WINED3DSPSM_ABS
;
2724 case WINED3DSPSM_BIAS
: *need_abs
= TRUE
; return WINED3DSPSM_BIAS
;
2725 case WINED3DSPSM_BIASNEG
: *need_abs
= TRUE
; return WINED3DSPSM_BIASNEG
;
2726 case WINED3DSPSM_SIGN
: *need_abs
= TRUE
; return WINED3DSPSM_SIGN
;
2727 case WINED3DSPSM_SIGNNEG
: *need_abs
= TRUE
; return WINED3DSPSM_SIGNNEG
;
2728 case WINED3DSPSM_COMP
: *need_abs
= TRUE
; return WINED3DSPSM_COMP
;
2729 case WINED3DSPSM_X2
: *need_abs
= TRUE
; return WINED3DSPSM_X2
;
2730 case WINED3DSPSM_X2NEG
: *need_abs
= TRUE
; return WINED3DSPSM_X2NEG
;
2731 case WINED3DSPSM_DZ
: *need_abs
= TRUE
; return WINED3DSPSM_DZ
;
2732 case WINED3DSPSM_DW
: *need_abs
= TRUE
; return WINED3DSPSM_DW
;
2733 case WINED3DSPSM_ABS
: return WINED3DSPSM_ABS
;
2734 case WINED3DSPSM_ABSNEG
: return WINED3DSPSM_ABS
;
2736 FIXME("Unknown modifier %u\n", mod
);
2740 static void shader_hw_log_pow(const struct wined3d_shader_instruction
*ins
)
2742 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2743 char src0
[50], src1
[50], dst
[50];
2744 struct wined3d_shader_src_param src0_copy
= ins
->src
[0];
2745 BOOL need_abs
= FALSE
;
2749 switch(ins
->handler_idx
)
2751 case WINED3DSIH_LOG
: instr
= "LG2"; break;
2752 case WINED3DSIH_LOGP
: instr
= "LOG"; break;
2753 case WINED3DSIH_POW
: instr
= "POW"; arg2
= TRUE
; break;
2755 ERR("Unexpected instruction %d\n", ins
->handler_idx
);
2759 /* LOG, LOGP and POW operate on the absolute value of the input */
2760 src0_copy
.modifiers
= abs_modifier(src0_copy
.modifiers
, &need_abs
);
2762 shader_arb_get_dst_param(ins
, &ins
->dst
[0], dst
);
2763 shader_arb_get_src_param(ins
, &src0_copy
, 0, src0
);
2764 if(arg2
) shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src1
);
2768 shader_addline(buffer
, "ABS TA, %s;\n", src0
);
2771 shader_addline(buffer
, "%s%s %s, TA, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src1
);
2775 shader_addline(buffer
, "%s%s %s, TA;\n", instr
, shader_arb_get_modifier(ins
), dst
);
2780 shader_addline(buffer
, "%s%s %s, %s, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src0
, src1
);
2784 shader_addline(buffer
, "%s%s %s, %s;\n", instr
, shader_arb_get_modifier(ins
), dst
, src0
);
2788 static void shader_hw_loop(const struct wined3d_shader_instruction
*ins
)
2790 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2792 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2795 shader_arb_get_src_param(ins
, &ins
->src
[1], 0, src_name
);
2799 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2800 struct list
*e
= list_head(&priv
->control_frames
);
2801 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2803 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2804 /* The constant loader makes sure to load -1 into iX.w */
2805 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2806 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2807 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2811 shader_addline(buffer
, "LOOP %s;\n", src_name
);
2815 static void shader_hw_rep(const struct wined3d_shader_instruction
*ins
)
2817 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2819 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2821 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name
);
2823 /* The constant loader makes sure to load -1 into iX.w */
2826 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2827 struct list
*e
= list_head(&priv
->control_frames
);
2828 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2830 if(priv
->loop_depth
> 1) shader_addline(buffer
, "PUSHA aL;\n");
2832 shader_addline(buffer
, "ARLC aL, %s.xywz;\n", src_name
);
2833 shader_addline(buffer
, "BRA loop_%u_end (LE.x);\n", control_frame
->no
.loop
);
2834 shader_addline(buffer
, "loop_%u_start:\n", control_frame
->no
.loop
);
2838 shader_addline(buffer
, "REP %s;\n", src_name
);
2842 static void shader_hw_endloop(const struct wined3d_shader_instruction
*ins
)
2844 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2845 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2849 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2850 struct list
*e
= list_head(&priv
->control_frames
);
2851 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2853 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2854 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2855 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2857 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2861 shader_addline(buffer
, "ENDLOOP;\n");
2865 static void shader_hw_endrep(const struct wined3d_shader_instruction
*ins
)
2867 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2868 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2872 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2873 struct list
*e
= list_head(&priv
->control_frames
);
2874 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2876 shader_addline(buffer
, "ARAC aL.xy, aL;\n");
2877 shader_addline(buffer
, "BRA loop_%u_start (GT.x);\n", control_frame
->no
.loop
);
2878 shader_addline(buffer
, "loop_%u_end:\n", control_frame
->no
.loop
);
2880 if(priv
->loop_depth
> 1) shader_addline(buffer
, "POPA aL;\n");
2884 shader_addline(buffer
, "ENDREP;\n");
2888 static const struct control_frame
*find_last_loop(const struct shader_arb_ctx_priv
*priv
)
2890 struct control_frame
*control_frame
;
2892 LIST_FOR_EACH_ENTRY(control_frame
, &priv
->control_frames
, struct control_frame
, entry
)
2894 if(control_frame
->type
== LOOP
|| control_frame
->type
== REP
) return control_frame
;
2896 ERR("Could not find loop for break\n");
2900 static void shader_hw_break(const struct wined3d_shader_instruction
*ins
)
2902 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2903 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2904 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2908 shader_addline(buffer
, "BRA loop_%u_end;\n", control_frame
->no
.loop
);
2912 shader_addline(buffer
, "BRK;\n");
2916 static const char *get_compare(COMPARISON_TYPE flags
)
2920 case COMPARISON_GT
: return "GT";
2921 case COMPARISON_EQ
: return "EQ";
2922 case COMPARISON_GE
: return "GE";
2923 case COMPARISON_LT
: return "LT";
2924 case COMPARISON_NE
: return "NE";
2925 case COMPARISON_LE
: return "LE";
2927 FIXME("Unrecognized comparison value: %u\n", flags
);
2932 static COMPARISON_TYPE
invert_compare(COMPARISON_TYPE flags
)
2936 case COMPARISON_GT
: return COMPARISON_LE
;
2937 case COMPARISON_EQ
: return COMPARISON_NE
;
2938 case COMPARISON_GE
: return COMPARISON_LT
;
2939 case COMPARISON_LT
: return COMPARISON_GE
;
2940 case COMPARISON_NE
: return COMPARISON_EQ
;
2941 case COMPARISON_LE
: return COMPARISON_GT
;
2943 FIXME("Unrecognized comparison value: %u\n", flags
);
2948 static void shader_hw_breakc(const struct wined3d_shader_instruction
*ins
)
2950 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2951 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2952 const struct control_frame
*control_frame
= find_last_loop(ins
->ctx
->backend_data
);
2955 const char *comp
= get_compare(ins
->flags
);
2957 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2958 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2962 /* SUBC CC, src0, src1" works only in pixel shaders, so use TA to throw
2963 * away the subtraction result
2965 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2966 shader_addline(buffer
, "BRA loop_%u_end (%s.x);\n", control_frame
->no
.loop
, comp
);
2970 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2971 shader_addline(buffer
, "BRK (%s.x);\n", comp
);
2975 static void shader_hw_ifc(const struct wined3d_shader_instruction
*ins
)
2977 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
2978 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
2979 struct list
*e
= list_head(&priv
->control_frames
);
2980 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
2984 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
2986 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, src_name0
);
2987 shader_arb_get_src_param(ins
, &ins
->src
[1], 1, src_name1
);
2991 /* Invert the flag. We jump to the else label if the condition is NOT true */
2992 comp
= get_compare(invert_compare(ins
->flags
));
2993 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
2994 shader_addline(buffer
, "BRA ifc_%u_else (%s.x);\n", control_frame
->no
.ifc
, comp
);
2998 comp
= get_compare(ins
->flags
);
2999 shader_addline(buffer
, "SUBC TA, %s, %s;\n", src_name0
, src_name1
);
3000 shader_addline(buffer
, "IF %s.x;\n", comp
);
3004 static void shader_hw_else(const struct wined3d_shader_instruction
*ins
)
3006 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3007 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3008 struct list
*e
= list_head(&priv
->control_frames
);
3009 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3010 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3014 shader_addline(buffer
, "BRA ifc_%u_endif;\n", control_frame
->no
.ifc
);
3015 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3016 control_frame
->had_else
= TRUE
;
3020 shader_addline(buffer
, "ELSE;\n");
3024 static void shader_hw_endif(const struct wined3d_shader_instruction
*ins
)
3026 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3027 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3028 struct list
*e
= list_head(&priv
->control_frames
);
3029 struct control_frame
*control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
3030 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3034 if(control_frame
->had_else
)
3036 shader_addline(buffer
, "ifc_%u_endif:\n", control_frame
->no
.ifc
);
3040 shader_addline(buffer
, "#No else branch. else is endif\n");
3041 shader_addline(buffer
, "ifc_%u_else:\n", control_frame
->no
.ifc
);
3046 shader_addline(buffer
, "ENDIF;\n");
3050 static void shader_hw_texldd(const struct wined3d_shader_instruction
*ins
)
3052 DWORD sampler_idx
= ins
->src
[1].reg
.idx
;
3054 char reg_src
[3][40];
3055 DWORD flags
= TEX_DERIV
;
3057 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3058 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_src
[0]);
3059 shader_arb_get_src_param(ins
, &ins
->src
[2], 1, reg_src
[1]);
3060 shader_arb_get_src_param(ins
, &ins
->src
[3], 2, reg_src
[2]);
3062 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3063 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3065 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_src
[0], flags
, reg_src
[1], reg_src
[2]);
3068 static void shader_hw_texldl(const struct wined3d_shader_instruction
*ins
)
3070 DWORD sampler_idx
= ins
->src
[1].reg
.idx
;
3073 DWORD flags
= TEX_LOD
;
3075 shader_arb_get_dst_param(ins
, &ins
->dst
[0], reg_dest
);
3076 shader_arb_get_src_param(ins
, &ins
->src
[0], 0, reg_coord
);
3078 if (ins
->flags
& WINED3DSI_TEXLD_PROJECT
) flags
|= TEX_PROJ
;
3079 if (ins
->flags
& WINED3DSI_TEXLD_BIAS
) flags
|= TEX_BIAS
;
3081 shader_hw_sample(ins
, sampler_idx
, reg_dest
, reg_coord
, flags
, NULL
, NULL
);
3084 static void shader_hw_label(const struct wined3d_shader_instruction
*ins
)
3086 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3087 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3089 priv
->in_main_func
= FALSE
;
3090 /* Call instructions activate the NV extensions, not labels and rets. If there is an uncalled
3091 * subroutine, don't generate a label that will make GL complain
3093 if(priv
->target_version
== ARB
) return;
3095 shader_addline(buffer
, "l%u:\n", ins
->src
[0].reg
.idx
);
3098 static void vshader_add_footer(IWineD3DVertexShaderImpl
*This
, struct wined3d_shader_buffer
*buffer
,
3099 const struct arb_vs_compile_args
*args
, struct shader_arb_ctx_priv
*priv_ctx
)
3101 const shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
3102 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
3103 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
3106 /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
3107 * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
3108 * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
3109 * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
3111 if(args
->super
.fog_src
== VS_FOG_Z
) {
3112 shader_addline(buffer
, "MOV result.fogcoord, TMP_OUT.z;\n");
3113 } else if (!reg_maps
->fog
) {
3114 /* posFixup.x is always 1.0, so we can savely use it */
3115 shader_addline(buffer
, "ADD result.fogcoord, posFixup.x, -posFixup.x;\n");
3118 /* Write the final position.
3120 * OpenGL coordinates specify the center of the pixel while d3d coords specify
3121 * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
3122 * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
3123 * contains 1.0 to allow a mad, but arb vs swizzles are too restricted for that.
3125 shader_addline(buffer
, "MUL TA, posFixup, TMP_OUT.w;\n");
3126 shader_addline(buffer
, "ADD TMP_OUT.x, TMP_OUT.x, TA.z;\n");
3127 shader_addline(buffer
, "MAD TMP_OUT.y, TMP_OUT.y, posFixup.y, TA.w;\n");
3129 if(use_nv_clip(gl_info
) && priv_ctx
->target_version
>= NV2
)
3131 if(args
->super
.clip_enabled
)
3133 for(i
= 0; i
< priv_ctx
->vs_clipplanes
; i
++)
3135 shader_addline(buffer
, "DP4 result.clip[%u].x, TMP_OUT, state.clip[%u].plane;\n", i
, i
);
3139 else if(args
->clip
.boolclip
.clip_texcoord
)
3141 unsigned int cur_clip
= 0;
3142 char component
[4] = {'x', 'y', 'z', 'w'};
3143 const char *zero
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ZERO
);
3145 for (i
= 0; i
< gl_info
->limits
.clipplanes
; ++i
)
3147 if(args
->clip
.boolclip
.clipplane_mask
& (1 << i
))
3149 shader_addline(buffer
, "DP4 TA.%c, TMP_OUT, state.clip[%u].plane;\n",
3150 component
[cur_clip
++], i
);
3156 shader_addline(buffer
, "MOV TA, %s;\n", zero
);
3159 shader_addline(buffer
, "MOV TA.yzw, %s;\n", zero
);
3162 shader_addline(buffer
, "MOV TA.zw, %s;\n", zero
);
3165 shader_addline(buffer
, "MOV TA.w, %s;\n", zero
);
3168 shader_addline(buffer
, "MOV result.texcoord[%u], TA;\n",
3169 args
->clip
.boolclip
.clip_texcoord
- 1);
3172 /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
3173 * and the glsl equivalent
3175 if(need_helper_const((IWineD3DBaseShaderImpl
*) This
, gl_info
)) {
3176 const char *two
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_TWO
);
3177 shader_addline(buffer
, "MAD TMP_OUT.z, TMP_OUT.z, %s, -TMP_OUT.w;\n", two
);
3179 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, TMP_OUT.z;\n");
3180 shader_addline(buffer
, "ADD TMP_OUT.z, TMP_OUT.z, -TMP_OUT.w;\n");
3183 shader_addline(buffer
, "MOV result.position, TMP_OUT;\n");
3185 priv_ctx
->footer_written
= TRUE
;
3188 static void shader_hw_ret(const struct wined3d_shader_instruction
*ins
)
3190 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3191 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
3192 IWineD3DBaseShaderImpl
*shader
= (IWineD3DBaseShaderImpl
*) ins
->ctx
->shader
;
3193 BOOL vshader
= shader_is_vshader_version(ins
->ctx
->reg_maps
->shader_version
.type
);
3195 if(priv
->target_version
== ARB
) return;
3199 if(priv
->in_main_func
) vshader_add_footer((IWineD3DVertexShaderImpl
*) shader
, buffer
, priv
->cur_vs_args
, priv
);
3202 shader_addline(buffer
, "RET;\n");
3205 static void shader_hw_call(const struct wined3d_shader_instruction
*ins
)
3207 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
3208 shader_addline(buffer
, "CAL l%u;\n", ins
->src
[0].reg
.idx
);
3211 /* GL locking is done by the caller */
3212 static GLuint
create_arb_blt_vertex_program(const struct wined3d_gl_info
*gl_info
)
3214 GLuint program_id
= 0;
3217 const char *blt_vprogram
=
3219 "PARAM c[1] = { { 1, 0.5 } };\n"
3220 "MOV result.position, vertex.position;\n"
3221 "MOV result.color, c[0].x;\n"
3222 "MOV result.texcoord[0], vertex.texcoord[0];\n"
3225 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3226 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, program_id
));
3227 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3228 strlen(blt_vprogram
), blt_vprogram
));
3229 checkGLcall("glProgramStringARB()");
3231 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3234 FIXME("Vertex program error at position %d: %s\n\n", pos
,
3235 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3236 shader_arb_dump_program_source(blt_vprogram
);
3242 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3243 checkGLcall("glGetProgramivARB()");
3244 if (!native
) WARN("Program exceeds native resource limits.\n");
3250 /* GL locking is done by the caller */
3251 static GLuint
create_arb_blt_fragment_program(const struct wined3d_gl_info
*gl_info
,
3252 enum tex_types tex_type
, BOOL masked
)
3254 GLuint program_id
= 0;
3255 const char *fprogram
;
3258 static const char * const blt_fprograms_full
[tex_type_count
] =
3265 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3266 "MOV result.depth.z, R0.x;\n"
3273 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3274 "MOV result.depth.z, R0.x;\n"
3279 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3280 "MOV result.depth.z, R0.x;\n"
3284 static const char * const blt_fprograms_masked
[tex_type_count
] =
3290 "PARAM mask = program.local[0];\n"
3292 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3293 "MUL R0.x, R0.x, R0.y;\n"
3295 "TEX R0.x, fragment.texcoord[0], texture[0], 2D;\n"
3296 "MOV result.depth.z, R0.x;\n"
3302 "PARAM mask = program.local[0];\n"
3304 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3305 "MUL R0.x, R0.x, R0.y;\n"
3307 "TEX R0.x, fragment.texcoord[0], texture[0], CUBE;\n"
3308 "MOV result.depth.z, R0.x;\n"
3312 "PARAM mask = program.local[0];\n"
3314 "SLT R0.xy, fragment.position, mask.zwzw;\n"
3315 "MUL R0.x, R0.x, R0.y;\n"
3317 "TEX R0.x, fragment.texcoord[0], texture[0], RECT;\n"
3318 "MOV result.depth.z, R0.x;\n"
3322 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3325 FIXME("tex_type %#x not supported, falling back to tex_2d\n", tex_type
);
3327 fprogram
= masked
? blt_fprograms_masked
[tex_type
] : blt_fprograms_full
[tex_type
];
3330 GL_EXTCALL(glGenProgramsARB(1, &program_id
));
3331 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, program_id
));
3332 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
, strlen(fprogram
), fprogram
));
3333 checkGLcall("glProgramStringARB()");
3335 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
3338 FIXME("Fragment program error at position %d: %s\n\n", pos
,
3339 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3340 shader_arb_dump_program_source(fprogram
);
3346 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3347 checkGLcall("glGetProgramivARB()");
3348 if (!native
) WARN("Program exceeds native resource limits.\n");
3354 static void arbfp_add_sRGB_correction(struct wined3d_shader_buffer
*buffer
, const char *fragcolor
,
3355 const char *tmp1
, const char *tmp2
, const char *tmp3
, const char *tmp4
, BOOL condcode
)
3357 /* Perform sRGB write correction. See GLX_EXT_framebuffer_sRGB */
3361 /* Sigh. MOVC CC doesn't work, so use one of the temps as dummy dest */
3362 shader_addline(buffer
, "SUBC %s, %s.x, srgb_consts1.y;\n", tmp1
, fragcolor
);
3363 /* Calculate the > 0.0031308 case */
3364 shader_addline(buffer
, "POW %s.x (GE), %s.x, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3365 shader_addline(buffer
, "POW %s.y (GE), %s.y, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3366 shader_addline(buffer
, "POW %s.z (GE), %s.z, srgb_consts1.z;\n", fragcolor
, fragcolor
);
3367 shader_addline(buffer
, "MUL %s.xyz (GE), %s, srgb_consts1.w;\n", fragcolor
, fragcolor
);
3368 shader_addline(buffer
, "SUB %s.xyz (GE), %s, srgb_consts2.x;\n", fragcolor
, fragcolor
);
3369 /* Calculate the < case */
3370 shader_addline(buffer
, "MUL %s.xyz (LT), srgb_consts1.x, %s;\n", fragcolor
, fragcolor
);
3374 /* Calculate the > 0.0031308 case */
3375 shader_addline(buffer
, "POW %s.x, %s.x, srgb_consts1.z;\n", tmp1
, fragcolor
);
3376 shader_addline(buffer
, "POW %s.y, %s.y, srgb_consts1.z;\n", tmp1
, fragcolor
);
3377 shader_addline(buffer
, "POW %s.z, %s.z, srgb_consts1.z;\n", tmp1
, fragcolor
);
3378 shader_addline(buffer
, "MUL %s, %s, srgb_consts1.w;\n", tmp1
, tmp1
);
3379 shader_addline(buffer
, "SUB %s, %s, srgb_consts2.x;\n", tmp1
, tmp1
);
3380 /* Calculate the < case */
3381 shader_addline(buffer
, "MUL %s, srgb_consts1.x, %s;\n", tmp2
, fragcolor
);
3382 /* Get 1.0 / 0.0 masks for > 0.0031308 and < 0.0031308 */
3383 shader_addline(buffer
, "SLT %s, srgb_consts1.y, %s;\n", tmp3
, fragcolor
);
3384 shader_addline(buffer
, "SGE %s, srgb_consts1.y, %s;\n", tmp4
, fragcolor
);
3385 /* Store the components > 0.0031308 in the destination */
3386 shader_addline(buffer
, "MUL %s.xyz, %s, %s;\n", fragcolor
, tmp1
, tmp3
);
3387 /* Add the components that are < 0.0031308 */
3388 shader_addline(buffer
, "MAD %s.xyz, %s, %s, %s;\n", fragcolor
, tmp2
, tmp4
, fragcolor
);
3389 /* Move everything into result.color at once. Nvidia hardware cannot handle partial
3390 * result.color writes(.rgb first, then .a), or handle overwriting already written
3391 * components. The assembler uses a temporary register in this case, which is usually
3392 * not allocated from one of our registers that were used earlier.
3395 /* [0.0;1.0] clamping. Not needed, this is done implicitly */
3398 static const DWORD
*find_loop_control_values(IWineD3DBaseShaderImpl
*This
, DWORD idx
)
3400 const local_constant
*constant
;
3402 LIST_FOR_EACH_ENTRY(constant
, &This
->baseShader
.constantsI
, local_constant
, entry
)
3404 if (constant
->idx
== idx
)
3406 return constant
->value
;
3412 static void init_ps_input(const IWineD3DPixelShaderImpl
*This
, const struct arb_ps_compile_args
*args
,
3413 struct shader_arb_ctx_priv
*priv
)
3415 const char *texcoords
[8] =
3417 "fragment.texcoord[0]", "fragment.texcoord[1]", "fragment.texcoord[2]", "fragment.texcoord[3]",
3418 "fragment.texcoord[4]", "fragment.texcoord[5]", "fragment.texcoord[6]", "fragment.texcoord[7]"
3421 const struct wined3d_shader_signature_element
*sig
= This
->baseShader
.input_signature
;
3422 const char *semantic_name
;
3425 switch(args
->super
.vp_mode
)
3427 case pretransformed
:
3429 /* The pixelshader has to collect the varyings on its own. In any case properly load
3430 * color0 and color1. In the case of pretransformed vertices also load texcoords. Set
3431 * other attribs to 0.0.
3433 * For fixedfunction this behavior is correct, according to the tests. For pretransformed
3434 * we'd either need a replacement shader that can load other attribs like BINORMAL, or
3435 * load the texcoord attrib pointers to match the pixel shader signature
3437 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3439 semantic_name
= sig
[i
].semantic_name
;
3440 semantic_idx
= sig
[i
].semantic_idx
;
3441 if(semantic_name
== NULL
) continue;
3443 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3445 if(semantic_idx
== 0) priv
->ps_input
[i
] = "fragment.color.primary";
3446 else if(semantic_idx
== 1) priv
->ps_input
[i
] = "fragment.color.secondary";
3447 else priv
->ps_input
[i
] = "0.0";
3449 else if(args
->super
.vp_mode
== fixedfunction
)
3451 priv
->ps_input
[i
] = "0.0";
3453 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3455 if(semantic_idx
< 8) priv
->ps_input
[i
] = texcoords
[semantic_idx
];
3456 else priv
->ps_input
[i
] = "0.0";
3458 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3460 if(semantic_idx
== 0) priv
->ps_input
[i
] = "fragment.fogcoord";
3461 else priv
->ps_input
[i
] = "0.0";
3465 priv
->ps_input
[i
] = "0.0";
3468 TRACE("v%u, semantic %s%u is %s\n", i
, semantic_name
, semantic_idx
, priv
->ps_input
[i
]);
3473 /* That one is easy. The vertex shaders provide v0-v7 in fragment.texcoord and v8 and v9 in
3476 for(i
= 0; i
< 8; i
++)
3478 priv
->ps_input
[i
] = texcoords
[i
];
3480 priv
->ps_input
[8] = "fragment.color.primary";
3481 priv
->ps_input
[9] = "fragment.color.secondary";
3486 /* GL locking is done by the caller */
3487 static GLuint
shader_arb_generate_pshader(IWineD3DPixelShaderImpl
*This
, struct wined3d_shader_buffer
*buffer
,
3488 const struct arb_ps_compile_args
*args
, struct arb_ps_compiled_shader
*compiled
)
3490 const shader_reg_maps
* reg_maps
= &This
->baseShader
.reg_maps
;
3491 CONST DWORD
*function
= This
->baseShader
.function
;
3492 const struct wined3d_gl_info
*gl_info
= &((IWineD3DDeviceImpl
*)This
->baseShader
.device
)->adapter
->gl_info
;
3493 const local_constant
*lconst
;
3496 DWORD
*lconst_map
= local_const_mapping((IWineD3DBaseShaderImpl
*) This
), next_local
, cur
;
3497 struct shader_arb_ctx_priv priv_ctx
;
3498 BOOL dcl_td
= FALSE
;
3499 BOOL want_nv_prog
= FALSE
;
3500 struct arb_pshader_private
*shader_priv
= This
->baseShader
.backend_data
;
3505 unsigned int i
, found
= 0;
3507 for (i
= 0, map
= reg_maps
->temporary
; map
; map
>>= 1, ++i
)
3510 || (This
->color0_mov
&& i
== This
->color0_reg
)
3511 || (reg_maps
->shader_version
.major
< 2 && i
== 0))
3514 sprintf(srgbtmp
[found
], "R%u", i
);
3516 if (found
== 4) break;
3521 sprintf(srgbtmp
[0], "TA");
3522 sprintf(srgbtmp
[1], "TB");
3523 sprintf(srgbtmp
[2], "TC");
3524 sprintf(srgbtmp
[3], "TD");
3528 sprintf(srgbtmp
[1], "TA");
3529 sprintf(srgbtmp
[2], "TB");
3530 sprintf(srgbtmp
[3], "TC");
3533 sprintf(srgbtmp
[2], "TA");
3534 sprintf(srgbtmp
[3], "TB");
3537 sprintf(srgbtmp
[3], "TA");
3543 /* Create the hw ARB shader */
3544 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
3545 priv_ctx
.cur_ps_args
= args
;
3546 priv_ctx
.compiled_fprog
= compiled
;
3547 priv_ctx
.cur_np2fixup_info
= &compiled
->np2fixup_info
;
3548 init_ps_input(This
, args
, &priv_ctx
);
3549 list_init(&priv_ctx
.control_frames
);
3551 /* Avoid enabling NV_fragment_program* if we do not need it.
3553 * Enabling GL_NV_fragment_program_option causes the driver to occupy a temporary register,
3554 * and it slows down the shader execution noticeably(about 5%). Usually our instruction emulation
3555 * is faster than what we gain from using higher native instructions. There are some things though
3556 * that cannot be emulated. In that case enable the extensions.
3557 * If the extension is enabled, instruction handlers that support both ways will use it.
3559 * Testing shows no performance difference between OPTION NV_fragment_program2 and NV_fragment_program.
3560 * So enable the best we can get.
3562 if(reg_maps
->usesdsx
|| reg_maps
->usesdsy
|| reg_maps
->loop_depth
> 0 || reg_maps
->usestexldd
||
3563 reg_maps
->usestexldl
|| reg_maps
->usesfacing
|| reg_maps
->usesifc
|| reg_maps
->usescall
)
3565 want_nv_prog
= TRUE
;
3568 shader_addline(buffer
, "!!ARBfp1.0\n");
3569 if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
3571 shader_addline(buffer
, "OPTION NV_fragment_program2;\n");
3572 priv_ctx
.target_version
= NV3
;
3574 else if (want_nv_prog
&& gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
3576 shader_addline(buffer
, "OPTION NV_fragment_program;\n");
3577 priv_ctx
.target_version
= NV2
;
3581 /* This is an error - either we're advertising the wrong shader version, or aren't enforcing some
3584 ERR("The shader requires instructions that are not available in plain GL_ARB_fragment_program\n");
3587 priv_ctx
.target_version
= ARB
;
3590 if(This
->baseShader
.reg_maps
.highest_render_target
> 0)
3592 shader_addline(buffer
, "OPTION ARB_draw_buffers;\n");
3595 if (reg_maps
->shader_version
.major
< 3)
3597 switch(args
->super
.fog
) {
3601 shader_addline(buffer
, "OPTION ARB_fog_linear;\n");
3604 shader_addline(buffer
, "OPTION ARB_fog_exp;\n");
3607 shader_addline(buffer
, "OPTION ARB_fog_exp2;\n");
3612 /* For now always declare the temps. At least the Nvidia assembler optimizes completely
3613 * unused temps away(but occupies them for the whole shader if they're used once). Always
3614 * declaring them avoids tricky bookkeeping work
3616 shader_addline(buffer
, "TEMP TA;\n"); /* Used for modifiers */
3617 shader_addline(buffer
, "TEMP TB;\n"); /* Used for modifiers */
3618 shader_addline(buffer
, "TEMP TC;\n"); /* Used for modifiers */
3619 if(dcl_td
) shader_addline(buffer
, "TEMP TD;\n"); /* Used for sRGB writing */
3620 shader_addline(buffer
, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
3621 shader_addline(buffer
, "PARAM coefmul = { 2, 4, 8, 16 };\n");
3622 shader_addline(buffer
, "PARAM ps_helper_const = { 0.0, 1.0, %1.10f, 0.0 };\n", eps
);
3624 if (reg_maps
->shader_version
.major
< 2)
3626 strcpy(fragcolor
, "R0");
3628 if(args
->super
.srgb_correction
) {
3629 if(This
->color0_mov
) {
3630 sprintf(fragcolor
, "R%u", This
->color0_reg
);
3632 shader_addline(buffer
, "TEMP TMP_COLOR;\n");
3633 strcpy(fragcolor
, "TMP_COLOR");
3636 strcpy(fragcolor
, "result.color");
3640 if(args
->super
.srgb_correction
) {
3641 shader_addline(buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
3642 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
3643 shader_addline(buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
3644 srgb_sub_high
, 0.0, 0.0, 0.0);
3647 /* Base Declarations */
3648 next_local
= shader_generate_arb_declarations((IWineD3DBaseShader
*)This
,
3649 reg_maps
, buffer
, gl_info
, lconst_map
, NULL
, &priv_ctx
);
3651 for (i
= 0, map
= reg_maps
->bumpmat
; map
; map
>>= 1, ++i
)
3653 if (!(map
& 1)) continue;
3655 cur
= compiled
->numbumpenvmatconsts
;
3656 compiled
->bumpenvmatconst
[cur
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3657 compiled
->bumpenvmatconst
[cur
].texunit
= i
;
3658 compiled
->luminanceconst
[cur
].const_num
= WINED3D_CONST_NUM_UNUSED
;
3659 compiled
->luminanceconst
[cur
].texunit
= i
;
3661 /* We can fit the constants into the constant limit for sure because texbem, texbeml, bem and beml are only supported
3662 * in 1.x shaders, and GL_ARB_fragment_program has a constant limit of 24 constants. So in the worst case we're loading
3663 * 8 shader constants, 8 bump matrices and 8 luminance parameters and are perfectly fine. (No NP2 fixup on bumpmapped
3664 * textures due to conditional NP2 restrictions)
3666 * Use local constants to load the bump env parameters, not program.env. This avoids collisions with d3d constants of
3667 * shaders in newer shader models. Since the bump env parameters have to share their space with NP2 fixup constants,
3668 * their location is shader dependent anyway and they cannot be loaded globally.
3670 compiled
->bumpenvmatconst
[cur
].const_num
= next_local
++;
3671 shader_addline(buffer
, "PARAM bumpenvmat%d = program.local[%d];\n",
3672 i
, compiled
->bumpenvmatconst
[cur
].const_num
);
3673 compiled
->numbumpenvmatconsts
= cur
+ 1;
3675 if (!(reg_maps
->luminanceparams
& (1 << i
))) continue;
3677 compiled
->luminanceconst
[cur
].const_num
= next_local
++;
3678 shader_addline(buffer
, "PARAM luminance%d = program.local[%d];\n",
3679 i
, compiled
->luminanceconst
[cur
].const_num
);
3682 for(i
= 0; i
< MAX_CONST_I
; i
++)
3684 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
3685 if (reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
3687 const DWORD
*control_values
= find_loop_control_values((IWineD3DBaseShaderImpl
*) This
, i
);
3691 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
3692 control_values
[0], control_values
[1], control_values
[2]);
3696 compiled
->int_consts
[i
] = next_local
;
3697 compiled
->num_int_consts
++;
3698 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
3703 if(reg_maps
->vpos
|| reg_maps
->usesdsy
)
3705 compiled
->ycorrection
= next_local
;
3706 shader_addline(buffer
, "PARAM ycorrection = program.local[%u];\n", next_local
++);
3710 shader_addline(buffer
, "TEMP vpos;\n");
3711 /* ycorrection.x: Backbuffer height(onscreen) or 0(offscreen).
3712 * ycorrection.y: -1.0(onscreen), 1.0(offscreen)
3713 * ycorrection.z: 1.0
3714 * ycorrection.w: 0.0
3716 shader_addline(buffer
, "MAD vpos, fragment.position, ycorrection.zyww, ycorrection.wxww;\n");
3717 shader_addline(buffer
, "FLR vpos.xy, vpos;\n");
3722 compiled
->ycorrection
= WINED3D_CONST_NUM_UNUSED
;
3725 /* Load constants to fixup NP2 texcoords if there are still free constants left:
3726 * Constants (texture dimensions) for the NP2 fixup are loaded as local program parameters. This will consume
3727 * at most 8 (MAX_FRAGMENT_SAMPLERS / 2) parameters, which is highly unlikely, since the application had to
3728 * use 16 NP2 textures at the same time. In case that we run out of constants the fixup is simply not
3729 * applied / activated. This will probably result in wrong rendering of the texture, but will save us from
3730 * shader compilation errors and the subsequent errors when drawing with this shader. */
3731 if (priv_ctx
.cur_ps_args
->super
.np2_fixup
) {
3733 struct arb_ps_np2fixup_info
* const fixup
= priv_ctx
.cur_np2fixup_info
;
3734 const WORD map
= priv_ctx
.cur_ps_args
->super
.np2_fixup
;
3735 const UINT max_lconsts
= gl_info
->limits
.arb_ps_local_constants
;
3737 fixup
->offset
= next_local
;
3738 fixup
->super
.active
= 0;
3741 for (i
= 0; i
< MAX_FRAGMENT_SAMPLERS
; ++i
) {
3742 if (!(map
& (1 << i
))) continue;
3744 if (fixup
->offset
+ (cur
>> 1) < max_lconsts
) {
3745 fixup
->super
.active
|= (1 << i
);
3746 fixup
->super
.idx
[i
] = cur
++;
3748 FIXME("No free constant found to load NP2 fixup data into shader. "
3749 "Sampling from this texture will probably look wrong.\n");
3754 fixup
->super
.num_consts
= (cur
+ 1) >> 1;
3755 if (fixup
->super
.num_consts
) {
3756 shader_addline(buffer
, "PARAM np2fixup[%u] = { program.env[%u..%u] };\n",
3757 fixup
->super
.num_consts
, fixup
->offset
, fixup
->super
.num_consts
+ fixup
->offset
- 1);
3760 next_local
+= fixup
->super
.num_consts
;
3763 if (shader_priv
->clipplane_emulation
!= ~0U && args
->clip
)
3765 shader_addline(buffer
, "KIL fragment.texcoord[%u];\n", shader_priv
->clipplane_emulation
);
3768 /* Base Shader Body */
3769 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
3771 if(args
->super
.srgb_correction
) {
3772 arbfp_add_sRGB_correction(buffer
, fragcolor
, srgbtmp
[0], srgbtmp
[1], srgbtmp
[2], srgbtmp
[3],
3773 priv_ctx
.target_version
>= NV2
);
3776 if(strcmp(fragcolor
, "result.color")) {
3777 shader_addline(buffer
, "MOV result.color, %s;\n", fragcolor
);
3779 shader_addline(buffer
, "END\n");
3781 /* TODO: change to resource.glObjectHandle or something like that */
3782 GL_EXTCALL(glGenProgramsARB(1, &retval
));
3784 TRACE("Creating a hw pixel shader, prg=%d\n", retval
);
3785 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, retval
));
3787 TRACE("Created hw pixel shader, prg=%d\n", retval
);
3788 /* Create the program and check for errors */
3789 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
3790 buffer
->bsize
, buffer
->buffer
));
3791 checkGLcall("glProgramStringARB()");
3793 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
3796 FIXME("HW PixelShader Error at position %d: %s\n\n",
3797 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
3798 shader_arb_dump_program_source(buffer
->buffer
);
3805 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
3806 checkGLcall("glGetProgramivARB()");
3807 if (!native
) WARN("Program exceeds native resource limits.\n");
3810 /* Load immediate constants */
3812 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
3813 const float *value
= (const float *)lconst
->value
;
3814 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, lconst_map
[lconst
->idx
], value
));
3815 checkGLcall("glProgramLocalParameter4fvARB");
3817 HeapFree(GetProcessHeap(), 0, lconst_map
);
3823 static int compare_sig(const struct wined3d_shader_signature_element
*sig1
, const struct wined3d_shader_signature_element
*sig2
)
3828 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3830 if(sig1
[i
].semantic_name
== NULL
|| sig2
[i
].semantic_name
== NULL
)
3832 /* Compare pointers, not contents. One string is NULL(element does not exist), the other one is not NULL */
3833 if(sig1
[i
].semantic_name
!= sig2
[i
].semantic_name
) return sig1
[i
].semantic_name
< sig2
[i
].semantic_name
? -1 : 1;
3837 ret
= strcmp(sig1
[i
].semantic_name
, sig2
[i
].semantic_name
);
3838 if(ret
!= 0) return ret
;
3839 if(sig1
[i
].semantic_idx
!= sig2
[i
].semantic_idx
) return sig1
[i
].semantic_idx
< sig2
[i
].semantic_idx
? -1 : 1;
3840 if(sig1
[i
].sysval_semantic
!= sig2
[i
].sysval_semantic
) return sig1
[i
].sysval_semantic
< sig2
[i
].sysval_semantic
? -1 : 1;
3841 if(sig1
[i
].component_type
!= sig2
[i
].component_type
) return sig1
[i
].sysval_semantic
< sig2
[i
].component_type
? -1 : 1;
3842 if(sig1
[i
].register_idx
!= sig2
[i
].register_idx
) return sig1
[i
].register_idx
< sig2
[i
].register_idx
? -1 : 1;
3843 if(sig1
[i
].mask
!= sig2
->mask
) return sig1
[i
].mask
< sig2
[i
].mask
? -1 : 1;
3848 static struct wined3d_shader_signature_element
*clone_sig(const struct wined3d_shader_signature_element
*sig
)
3850 struct wined3d_shader_signature_element
*new;
3854 new = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*new) * MAX_REG_INPUT
);
3855 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3857 if(sig
[i
].semantic_name
== NULL
)
3863 /* Clone the semantic string */
3864 name
= HeapAlloc(GetProcessHeap(), 0, strlen(sig
[i
].semantic_name
) + 1);
3865 strcpy(name
, sig
[i
].semantic_name
);
3866 new[i
].semantic_name
= name
;
3871 static DWORD
find_input_signature(struct shader_arb_priv
*priv
, const struct wined3d_shader_signature_element
*sig
)
3873 struct wine_rb_entry
*entry
= wine_rb_get(&priv
->signature_tree
, sig
);
3874 struct ps_signature
*found_sig
;
3878 found_sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
3879 TRACE("Found existing signature %u\n", found_sig
->idx
);
3880 return found_sig
->idx
;
3882 found_sig
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*sig
));
3883 found_sig
->sig
= clone_sig(sig
);
3884 found_sig
->idx
= priv
->ps_sig_number
++;
3885 TRACE("New signature stored and assigned number %u\n", found_sig
->idx
);
3886 if(wine_rb_put(&priv
->signature_tree
, sig
, &found_sig
->entry
) == -1)
3888 ERR("Failed to insert program entry.\n");
3890 return found_sig
->idx
;
3893 static void init_output_registers(IWineD3DVertexShaderImpl
*shader
, DWORD sig_num
, struct shader_arb_ctx_priv
*priv_ctx
,
3894 struct arb_vs_compiled_shader
*compiled
)
3897 static const char *texcoords
[8] =
3899 "result.texcoord[0]", "result.texcoord[1]", "result.texcoord[2]", "result.texcoord[3]",
3900 "result.texcoord[4]", "result.texcoord[5]", "result.texcoord[6]", "result.texcoord[7]"
3902 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) shader
->baseShader
.device
;
3903 IWineD3DBaseShaderClass
*baseshader
= &shader
->baseShader
;
3904 const struct wined3d_shader_signature_element
*sig
;
3905 const char *semantic_name
;
3906 DWORD semantic_idx
, reg_idx
;
3908 /* Write generic input varyings 0 to 7 to result.texcoord[], varying 8 to result.color.primary
3909 * and varying 9 to result.color.secondary
3911 const char *decl_idx_to_string
[MAX_REG_INPUT
] =
3913 texcoords
[0], texcoords
[1], texcoords
[2], texcoords
[3],
3914 texcoords
[4], texcoords
[5], texcoords
[6], texcoords
[7],
3915 "result.color.primary", "result.color.secondary"
3920 TRACE("Pixel shader uses builtin varyings\n");
3921 /* Map builtins to builtins */
3922 for(i
= 0; i
< 8; i
++)
3924 priv_ctx
->texcrd_output
[i
] = texcoords
[i
];
3926 priv_ctx
->color_output
[0] = "result.color.primary";
3927 priv_ctx
->color_output
[1] = "result.color.secondary";
3928 priv_ctx
->fog_output
= "result.fogcoord";
3930 /* Map declared regs to builtins. Use "TA" to /dev/null unread output */
3931 for (i
= 0; i
< (sizeof(baseshader
->output_signature
) / sizeof(*baseshader
->output_signature
)); ++i
)
3933 semantic_name
= baseshader
->output_signature
[i
].semantic_name
;
3934 if(semantic_name
== NULL
) continue;
3936 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
))
3938 TRACE("o%u is TMP_OUT\n", i
);
3939 if (baseshader
->output_signature
[i
].semantic_idx
== 0) priv_ctx
->vs_output
[i
] = "TMP_OUT";
3940 else priv_ctx
->vs_output
[i
] = "TA";
3942 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
))
3944 TRACE("o%u is result.pointsize\n", i
);
3945 if (baseshader
->output_signature
[i
].semantic_idx
== 0) priv_ctx
->vs_output
[i
] = "result.pointsize";
3946 else priv_ctx
->vs_output
[i
] = "TA";
3948 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
3950 TRACE("o%u is result.color.?, idx %u\n", i
, baseshader
->output_signature
[i
].semantic_idx
);
3951 if (baseshader
->output_signature
[i
].semantic_idx
== 0)
3952 priv_ctx
->vs_output
[i
] = "result.color.primary";
3953 else if (baseshader
->output_signature
[i
].semantic_idx
== 1)
3954 priv_ctx
->vs_output
[i
] = "result.color.secondary";
3955 else priv_ctx
->vs_output
[i
] = "TA";
3957 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
3959 TRACE("o%u is %s\n", i
, texcoords
[baseshader
->output_signature
[i
].semantic_idx
]);
3960 if (baseshader
->output_signature
[i
].semantic_idx
>= 8) priv_ctx
->vs_output
[i
] = "TA";
3961 else priv_ctx
->vs_output
[i
] = texcoords
[baseshader
->output_signature
[i
].semantic_idx
];
3963 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
3965 TRACE("o%u is result.fogcoord\n", i
);
3966 if (baseshader
->output_signature
[i
].semantic_idx
> 0) priv_ctx
->vs_output
[i
] = "TA";
3967 else priv_ctx
->vs_output
[i
] = "result.fogcoord";
3971 priv_ctx
->vs_output
[i
] = "TA";
3977 /* Instead of searching for the signature in the signature list, read the one from the current pixel shader.
3978 * Its maybe not the shader where the signature came from, but it is the same signature and faster to find
3980 sig
= ((IWineD3DPixelShaderImpl
*)device
->stateBlock
->pixelShader
)->baseShader
.input_signature
;
3981 TRACE("Pixel shader uses declared varyings\n");
3983 /* Map builtin to declared. /dev/null the results by default to the TA temp reg */
3984 for(i
= 0; i
< 8; i
++)
3986 priv_ctx
->texcrd_output
[i
] = "TA";
3988 priv_ctx
->color_output
[0] = "TA";
3989 priv_ctx
->color_output
[1] = "TA";
3990 priv_ctx
->fog_output
= "TA";
3992 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
3994 semantic_name
= sig
[i
].semantic_name
;
3995 semantic_idx
= sig
[i
].semantic_idx
;
3996 reg_idx
= sig
[i
].register_idx
;
3997 if(semantic_name
== NULL
) continue;
3999 /* If a declared input register is not written by builtin arguments, don't write to it.
4000 * GL_NV_vertex_program makes sure the input defaults to 0.0, which is correct with D3D
4002 * Don't care about POSITION and PSIZE here - this is a builtin vertex shader, position goes
4003 * to TMP_OUT in any case
4005 if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_TEXCOORD
))
4007 if(semantic_idx
< 8) priv_ctx
->texcrd_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4009 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_COLOR
))
4011 if(semantic_idx
< 2) priv_ctx
->color_output
[semantic_idx
] = decl_idx_to_string
[reg_idx
];
4013 else if(shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_FOG
))
4015 if(semantic_idx
== 0) priv_ctx
->fog_output
= decl_idx_to_string
[reg_idx
];
4022 if(strcmp(decl_idx_to_string
[reg_idx
], "result.color.primary") == 0 ||
4023 strcmp(decl_idx_to_string
[reg_idx
], "result.color.secondary") == 0)
4025 compiled
->need_color_unclamp
= TRUE
;
4029 /* Map declared to declared */
4030 for (i
= 0; i
< (sizeof(baseshader
->output_signature
) / sizeof(*baseshader
->output_signature
)); ++i
)
4032 /* Write unread output to TA to throw them away */
4033 priv_ctx
->vs_output
[i
] = "TA";
4034 semantic_name
= baseshader
->output_signature
[i
].semantic_name
;
4035 if(semantic_name
== NULL
)
4040 if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_POSITION
)
4041 && baseshader
->output_signature
[i
].semantic_idx
== 0)
4043 priv_ctx
->vs_output
[i
] = "TMP_OUT";
4046 else if (shader_match_semantic(semantic_name
, WINED3DDECLUSAGE_PSIZE
)
4047 && baseshader
->output_signature
[i
].semantic_idx
== 0)
4049 priv_ctx
->vs_output
[i
] = "result.pointsize";
4053 for(j
= 0; j
< MAX_REG_INPUT
; j
++)
4055 if(sig
[j
].semantic_name
== NULL
)
4060 if (strcmp(sig
[j
].semantic_name
, semantic_name
) == 0
4061 && sig
[j
].semantic_idx
== baseshader
->output_signature
[i
].semantic_idx
)
4063 priv_ctx
->vs_output
[i
] = decl_idx_to_string
[sig
[j
].register_idx
];
4065 if(strcmp(priv_ctx
->vs_output
[i
], "result.color.primary") == 0 ||
4066 strcmp(priv_ctx
->vs_output
[i
], "result.color.secondary") == 0)
4068 compiled
->need_color_unclamp
= TRUE
;
4075 /* GL locking is done by the caller */
4076 static GLuint
shader_arb_generate_vshader(IWineD3DVertexShaderImpl
*This
, struct wined3d_shader_buffer
*buffer
,
4077 const struct arb_vs_compile_args
*args
, struct arb_vs_compiled_shader
*compiled
)
4079 const shader_reg_maps
*reg_maps
= &This
->baseShader
.reg_maps
;
4080 CONST DWORD
*function
= This
->baseShader
.function
;
4081 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)This
->baseShader
.device
;
4082 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4083 const local_constant
*lconst
;
4085 DWORD next_local
, *lconst_map
= local_const_mapping((IWineD3DBaseShaderImpl
*) This
);
4086 struct shader_arb_ctx_priv priv_ctx
;
4090 memset(&priv_ctx
, 0, sizeof(priv_ctx
));
4091 priv_ctx
.cur_vs_args
= args
;
4092 list_init(&priv_ctx
.control_frames
);
4093 init_output_registers(This
, args
->ps_signature
, &priv_ctx
, compiled
);
4095 /* Create the hw ARB shader */
4096 shader_addline(buffer
, "!!ARBvp1.0\n");
4098 /* Always enable the NV extension if available. Unlike fragment shaders, there is no
4099 * mesurable performance penalty, and we can always make use of it for clipplanes.
4101 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4103 shader_addline(buffer
, "OPTION NV_vertex_program3;\n");
4104 priv_ctx
.target_version
= NV3
;
4105 shader_addline(buffer
, "ADDRESS aL;\n");
4107 else if (gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4109 shader_addline(buffer
, "OPTION NV_vertex_program2;\n");
4110 priv_ctx
.target_version
= NV2
;
4111 shader_addline(buffer
, "ADDRESS aL;\n");
4113 priv_ctx
.target_version
= ARB
;
4116 shader_addline(buffer
, "TEMP TMP_OUT;\n");
4117 if(need_helper_const((IWineD3DBaseShaderImpl
*) This
, gl_info
)) {
4118 shader_addline(buffer
, "PARAM helper_const = { 0.0, 1.0, 2.0, %1.10f};\n", eps
);
4120 if(need_rel_addr_const((IWineD3DBaseShaderImpl
*) This
, gl_info
)) {
4121 shader_addline(buffer
, "PARAM rel_addr_const = { 0.5, %d.0, 0.0, 0.0 };\n", This
->rel_offset
);
4122 shader_addline(buffer
, "TEMP A0_SHADOW;\n");
4125 shader_addline(buffer
, "TEMP TA;\n");
4127 /* Base Declarations */
4128 next_local
= shader_generate_arb_declarations((IWineD3DBaseShader
*)This
,
4129 reg_maps
, buffer
, gl_info
, lconst_map
, &priv_ctx
.vs_clipplanes
, &priv_ctx
);
4131 for(i
= 0; i
< MAX_CONST_I
; i
++)
4133 compiled
->int_consts
[i
] = WINED3D_CONST_NUM_UNUSED
;
4134 if(reg_maps
->integer_constants
& (1 << i
) && priv_ctx
.target_version
>= NV2
)
4136 const DWORD
*control_values
= find_loop_control_values((IWineD3DBaseShaderImpl
*) This
, i
);
4140 shader_addline(buffer
, "PARAM I%u = {%u, %u, %u, -1};\n", i
,
4141 control_values
[0], control_values
[1], control_values
[2]);
4145 compiled
->int_consts
[i
] = next_local
;
4146 compiled
->num_int_consts
++;
4147 shader_addline(buffer
, "PARAM I%u = program.local[%u];\n", i
, next_local
++);
4152 /* We need a constant to fixup the final position */
4153 shader_addline(buffer
, "PARAM posFixup = program.local[%u];\n", next_local
);
4154 compiled
->pos_fixup
= next_local
++;
4156 /* Initialize output parameters. GL_ARB_vertex_program does not require special initialization values
4157 * for output parameters. D3D in theory does not do that either, but some applications depend on a
4158 * proper initialization of the secondary color, and programs using the fixed function pipeline without
4159 * a replacement shader depend on the texcoord.w being set properly.
4161 * GL_NV_vertex_program defines that all output values are initialized to {0.0, 0.0, 0.0, 1.0}. This
4162 * assertion is in effect even when using GL_ARB_vertex_program without any NV specific additions. So
4163 * skip this if NV_vertex_program is supported. Otherwise, initialize the secondary color. For the tex-
4164 * coords, we have a flag in the opengl caps. Many cards do not require the texcoord being set, and
4165 * this can eat a number of instructions, so skip it unless this cap is set as well
4167 if (!gl_info
->supported
[NV_VERTEX_PROGRAM
])
4169 const char *color_init
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_0001
);
4170 shader_addline(buffer
, "MOV result.color.secondary, %s;\n", color_init
);
4172 if (gl_info
->quirks
& WINED3D_QUIRK_SET_TEXCOORD_W
&& !device
->frag_pipe
->ffp_proj_control
)
4175 const char *one
= arb_get_helper_value(WINED3D_SHADER_TYPE_VERTEX
, ARB_ONE
);
4176 for(i
= 0; i
< min(8, MAX_REG_TEXCRD
); i
++)
4178 if (This
->baseShader
.reg_maps
.texcoord_mask
[i
] != 0 &&
4179 This
->baseShader
.reg_maps
.texcoord_mask
[i
] != WINED3DSP_WRITEMASK_ALL
) {
4180 shader_addline(buffer
, "MOV result.texcoord[%u].w, %s\n", i
, one
);
4186 /* The shader starts with the main function */
4187 priv_ctx
.in_main_func
= TRUE
;
4188 /* Base Shader Body */
4189 shader_generate_main((IWineD3DBaseShader
*)This
, buffer
, reg_maps
, function
, &priv_ctx
);
4191 if(!priv_ctx
.footer_written
) vshader_add_footer(This
, buffer
, args
, &priv_ctx
);
4193 shader_addline(buffer
, "END\n");
4195 /* TODO: change to resource.glObjectHandle or something like that */
4196 GL_EXTCALL(glGenProgramsARB(1, &ret
));
4198 TRACE("Creating a hw vertex shader, prg=%d\n", ret
);
4199 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, ret
));
4201 TRACE("Created hw vertex shader, prg=%d\n", ret
);
4202 /* Create the program and check for errors */
4203 GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
4204 buffer
->bsize
, buffer
->buffer
));
4205 checkGLcall("glProgramStringARB()");
4207 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &errPos
);
4210 FIXME("HW VertexShader Error at position %d: %s\n\n",
4211 errPos
, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
4212 shader_arb_dump_program_source(buffer
->buffer
);
4219 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
4220 checkGLcall("glGetProgramivARB()");
4221 if (!native
) WARN("Program exceeds native resource limits.\n");
4223 /* Load immediate constants */
4225 LIST_FOR_EACH_ENTRY(lconst
, &This
->baseShader
.constantsF
, local_constant
, entry
) {
4226 const float *value
= (const float *)lconst
->value
;
4227 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_VERTEX_PROGRAM_ARB
, lconst_map
[lconst
->idx
], value
));
4231 HeapFree(GetProcessHeap(), 0, lconst_map
);
4236 /* GL locking is done by the caller */
4237 static struct arb_ps_compiled_shader
*find_arb_pshader(IWineD3DPixelShaderImpl
*shader
, const struct arb_ps_compile_args
*args
)
4241 struct arb_ps_compiled_shader
*new_array
;
4242 struct wined3d_shader_buffer buffer
;
4243 struct arb_pshader_private
*shader_data
;
4246 if (!shader
->baseShader
.backend_data
)
4248 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) shader
->baseShader
.device
;
4249 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4250 struct shader_arb_priv
*priv
= device
->shader_priv
;
4252 shader
->baseShader
.backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4253 shader_data
= shader
->baseShader
.backend_data
;
4254 shader_data
->clamp_consts
= shader
->baseShader
.reg_maps
.shader_version
.major
== 1;
4256 if(shader
->baseShader
.reg_maps
.shader_version
.major
< 3) shader_data
->input_signature_idx
= ~0;
4257 else shader_data
->input_signature_idx
= find_input_signature(priv
, shader
->baseShader
.input_signature
);
4259 shader_data
->has_signature_idx
= TRUE
;
4260 TRACE("Shader got assigned input signature index %u\n", shader_data
->input_signature_idx
);
4262 if (!device
->vs_clipping
)
4263 shader_data
->clipplane_emulation
= shader_find_free_input_register(&shader
->baseShader
.reg_maps
,
4264 gl_info
->limits
.texture_stages
- 1);
4266 shader_data
->clipplane_emulation
= ~0U;
4268 shader_data
= shader
->baseShader
.backend_data
;
4270 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4271 * so a linear search is more performant than a hashmap or a binary search
4272 * (cache coherency etc)
4274 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4275 if(memcmp(&shader_data
->gl_shaders
[i
].args
, args
, sizeof(*args
)) == 0) {
4276 return &shader_data
->gl_shaders
[i
];
4280 TRACE("No matching GL shader found, compiling a new shader\n");
4281 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4282 if (shader_data
->num_gl_shaders
)
4284 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4285 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4286 new_size
* sizeof(*shader_data
->gl_shaders
));
4288 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4293 ERR("Out of memory\n");
4296 shader_data
->gl_shaders
= new_array
;
4297 shader_data
->shader_array_size
= new_size
;
4300 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4302 pixelshader_update_samplers(&shader
->baseShader
.reg_maps
,
4303 ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->stateBlock
->textures
);
4305 if (!shader_buffer_init(&buffer
))
4307 ERR("Failed to initialize shader buffer.\n");
4311 ret
= shader_arb_generate_pshader(shader
, &buffer
, args
,
4312 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4313 shader_buffer_free(&buffer
);
4314 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4316 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4319 static inline BOOL
vs_args_equal(const struct arb_vs_compile_args
*stored
, const struct arb_vs_compile_args
*new,
4320 const DWORD use_map
, BOOL skip_int
) {
4321 if((stored
->super
.swizzle_map
& use_map
) != new->super
.swizzle_map
) return FALSE
;
4322 if(stored
->super
.clip_enabled
!= new->super
.clip_enabled
) return FALSE
;
4323 if(stored
->super
.fog_src
!= new->super
.fog_src
) return FALSE
;
4324 if(stored
->clip
.boolclip_compare
!= new->clip
.boolclip_compare
) return FALSE
;
4325 if(stored
->ps_signature
!= new->ps_signature
) return FALSE
;
4326 if(stored
->vertex
.samplers_compare
!= new->vertex
.samplers_compare
) return FALSE
;
4327 if(skip_int
) return TRUE
;
4329 return memcmp(stored
->loop_ctrl
, new->loop_ctrl
, sizeof(stored
->loop_ctrl
)) == 0;
4332 static struct arb_vs_compiled_shader
*find_arb_vshader(IWineD3DVertexShaderImpl
*shader
, const struct arb_vs_compile_args
*args
)
4336 struct arb_vs_compiled_shader
*new_array
;
4337 DWORD use_map
= ((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->strided_streams
.use_map
;
4338 struct wined3d_shader_buffer buffer
;
4339 struct arb_vshader_private
*shader_data
;
4341 const struct wined3d_gl_info
*gl_info
= &((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->adapter
->gl_info
;
4343 if (!shader
->baseShader
.backend_data
)
4345 shader
->baseShader
.backend_data
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
));
4347 shader_data
= shader
->baseShader
.backend_data
;
4349 /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
4350 * so a linear search is more performant than a hashmap or a binary search
4351 * (cache coherency etc)
4353 for(i
= 0; i
< shader_data
->num_gl_shaders
; i
++) {
4354 if (vs_args_equal(&shader_data
->gl_shaders
[i
].args
, args
,
4355 use_map
, gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
]))
4357 return &shader_data
->gl_shaders
[i
];
4361 TRACE("No matching GL shader found, compiling a new shader\n");
4363 if(shader_data
->shader_array_size
== shader_data
->num_gl_shaders
) {
4364 if (shader_data
->num_gl_shaders
)
4366 new_size
= shader_data
->shader_array_size
+ max(1, shader_data
->shader_array_size
/ 2);
4367 new_array
= HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, shader_data
->gl_shaders
,
4368 new_size
* sizeof(*shader_data
->gl_shaders
));
4370 new_array
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*shader_data
->gl_shaders
));
4375 ERR("Out of memory\n");
4378 shader_data
->gl_shaders
= new_array
;
4379 shader_data
->shader_array_size
= new_size
;
4382 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].args
= *args
;
4384 if (!shader_buffer_init(&buffer
))
4386 ERR("Failed to initialize shader buffer.\n");
4390 ret
= shader_arb_generate_vshader(shader
, &buffer
, args
,
4391 &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
]);
4392 shader_buffer_free(&buffer
);
4393 shader_data
->gl_shaders
[shader_data
->num_gl_shaders
].prgId
= ret
;
4395 return &shader_data
->gl_shaders
[shader_data
->num_gl_shaders
++];
4398 static inline void find_arb_ps_compile_args(IWineD3DPixelShaderImpl
*shader
, IWineD3DStateBlockImpl
*stateblock
,
4399 struct arb_ps_compile_args
*args
)
4403 const struct wined3d_gl_info
*gl_info
= &((IWineD3DDeviceImpl
*)shader
->baseShader
.device
)->adapter
->gl_info
;
4404 find_ps_compile_args(shader
, stateblock
, &args
->super
);
4406 /* This forces all local boolean constants to 1 to make them stateblock independent */
4407 args
->bools
= shader
->baseShader
.reg_maps
.local_bool_consts
;
4409 for(i
= 0; i
< MAX_CONST_B
; i
++)
4411 if(stateblock
->pixelShaderConstantB
[i
]) args
->bools
|= ( 1 << i
);
4414 /* Only enable the clip plane emulation KIL if at least one clipplane is enabled. The KIL instruction
4415 * is quite expensive because it forces the driver to disable early Z discards. It is cheaper to
4416 * duplicate the shader than have a no-op KIL instruction in every shader
4418 if((!((IWineD3DDeviceImpl
*) shader
->baseShader
.device
)->vs_clipping
) && use_vs(stateblock
) &&
4419 stateblock
->renderState
[WINED3DRS_CLIPPING
] && stateblock
->renderState
[WINED3DRS_CLIPPLANEENABLE
])
4428 /* Skip if unused or local, or supported natively */
4429 int_skip
= ~shader
->baseShader
.reg_maps
.integer_constants
| shader
->baseShader
.reg_maps
.local_int_consts
;
4430 if (int_skip
== 0xffff || gl_info
->supported
[NV_FRAGMENT_PROGRAM_OPTION
])
4432 memset(&args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4436 for(i
= 0; i
< MAX_CONST_I
; i
++)
4438 if(int_skip
& (1 << i
))
4440 args
->loop_ctrl
[i
][0] = 0;
4441 args
->loop_ctrl
[i
][1] = 0;
4442 args
->loop_ctrl
[i
][2] = 0;
4446 args
->loop_ctrl
[i
][0] = stateblock
->pixelShaderConstantI
[i
* 4];
4447 args
->loop_ctrl
[i
][1] = stateblock
->pixelShaderConstantI
[i
* 4 + 1];
4448 args
->loop_ctrl
[i
][2] = stateblock
->pixelShaderConstantI
[i
* 4 + 2];
4453 static inline void find_arb_vs_compile_args(IWineD3DVertexShaderImpl
*shader
, IWineD3DStateBlockImpl
*stateblock
,
4454 struct arb_vs_compile_args
*args
)
4458 IWineD3DDeviceImpl
*dev
= (IWineD3DDeviceImpl
*)shader
->baseShader
.device
;
4459 const struct wined3d_gl_info
*gl_info
= &dev
->adapter
->gl_info
;
4460 find_vs_compile_args(shader
, stateblock
, &args
->super
);
4462 args
->clip
.boolclip_compare
= 0;
4463 if(use_ps(stateblock
))
4465 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) stateblock
->pixelShader
;
4466 struct arb_pshader_private
*shader_priv
= ps
->baseShader
.backend_data
;
4467 args
->ps_signature
= shader_priv
->input_signature_idx
;
4469 args
->clip
.boolclip
.clip_texcoord
= shader_priv
->clipplane_emulation
+ 1;
4473 args
->ps_signature
= ~0;
4474 if(!dev
->vs_clipping
)
4476 args
->clip
.boolclip
.clip_texcoord
= ffp_clip_emul(stateblock
) ? gl_info
->limits
.texture_stages
: 0;
4478 /* Otherwise: Setting boolclip_compare set clip_texcoord to 0 */
4481 if(args
->clip
.boolclip
.clip_texcoord
)
4483 if(stateblock
->renderState
[WINED3DRS_CLIPPING
])
4485 args
->clip
.boolclip
.clipplane_mask
= stateblock
->renderState
[WINED3DRS_CLIPPLANEENABLE
];
4487 /* clipplane_mask was set to 0 by setting boolclip_compare to 0 */
4490 /* This forces all local boolean constants to 1 to make them stateblock independent */
4491 args
->clip
.boolclip
.bools
= shader
->baseShader
.reg_maps
.local_bool_consts
;
4492 /* TODO: Figure out if it would be better to store bool constants as bitmasks in the stateblock */
4493 for(i
= 0; i
< MAX_CONST_B
; i
++)
4495 if(stateblock
->vertexShaderConstantB
[i
]) args
->clip
.boolclip
.bools
|= ( 1 << i
);
4498 args
->vertex
.samplers
[0] = dev
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 0];
4499 args
->vertex
.samplers
[1] = dev
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 1];
4500 args
->vertex
.samplers
[2] = dev
->texUnitMap
[MAX_FRAGMENT_SAMPLERS
+ 2];
4501 args
->vertex
.samplers
[3] = 0;
4503 /* Skip if unused or local */
4504 int_skip
= ~shader
->baseShader
.reg_maps
.integer_constants
| shader
->baseShader
.reg_maps
.local_int_consts
;
4505 /* This is about flow control, not clipping. */
4506 if (int_skip
== 0xffff || gl_info
->supported
[NV_VERTEX_PROGRAM2_OPTION
])
4508 memset(&args
->loop_ctrl
, 0, sizeof(args
->loop_ctrl
));
4512 for(i
= 0; i
< MAX_CONST_I
; i
++)
4514 if(int_skip
& (1 << i
))
4516 args
->loop_ctrl
[i
][0] = 0;
4517 args
->loop_ctrl
[i
][1] = 0;
4518 args
->loop_ctrl
[i
][2] = 0;
4522 args
->loop_ctrl
[i
][0] = stateblock
->vertexShaderConstantI
[i
* 4];
4523 args
->loop_ctrl
[i
][1] = stateblock
->vertexShaderConstantI
[i
* 4 + 1];
4524 args
->loop_ctrl
[i
][2] = stateblock
->vertexShaderConstantI
[i
* 4 + 2];
4529 /* GL locking is done by the caller */
4530 static void shader_arb_select(const struct wined3d_context
*context
, BOOL usePS
, BOOL useVS
)
4532 IWineD3DDeviceImpl
*This
= context
->swapchain
->device
;
4533 struct shader_arb_priv
*priv
= This
->shader_priv
;
4534 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
4537 /* Deal with pixel shaders first so the vertex shader arg function has the input signature ready */
4539 struct arb_ps_compile_args compile_args
;
4540 struct arb_ps_compiled_shader
*compiled
;
4541 IWineD3DPixelShaderImpl
*ps
= (IWineD3DPixelShaderImpl
*) This
->stateBlock
->pixelShader
;
4543 TRACE("Using pixel shader %p\n", This
->stateBlock
->pixelShader
);
4544 find_arb_ps_compile_args(ps
, This
->stateBlock
, &compile_args
);
4545 compiled
= find_arb_pshader(ps
, &compile_args
);
4546 priv
->current_fprogram_id
= compiled
->prgId
;
4547 priv
->compiled_fprog
= compiled
;
4549 /* Bind the fragment program */
4550 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4551 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, priv->current_fprogram_id);");
4553 if(!priv
->use_arbfp_fixed_func
) {
4554 /* Enable OpenGL fragment programs */
4555 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4556 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB);");
4558 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
4560 /* Pixel Shader 1.x constants are clamped to [-1;1], Pixel Shader 2.0 constants are not. If switching between
4561 * a 1.x and newer shader, reload the first 8 constants
4563 if(priv
->last_ps_const_clamped
!= ((struct arb_pshader_private
*)ps
->baseShader
.backend_data
)->clamp_consts
)
4565 priv
->last_ps_const_clamped
= ((struct arb_pshader_private
*)ps
->baseShader
.backend_data
)->clamp_consts
;
4566 This
->highest_dirty_ps_const
= max(This
->highest_dirty_ps_const
, 8);
4567 for(i
= 0; i
< 8; i
++)
4569 context
->pshader_const_dirty
[i
] = 1;
4571 /* Also takes care of loading local constants */
4572 shader_arb_load_constants(context
, TRUE
, FALSE
);
4576 shader_arb_ps_local_constants(This
);
4579 /* Force constant reloading for the NP2 fixup (see comment in shader_glsl_select for more info) */
4580 if (compiled
->np2fixup_info
.super
.active
)
4581 shader_arb_load_np2fixup_constants((IWineD3DDevice
*)This
, usePS
, useVS
);
4583 else if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
] && !priv
->use_arbfp_fixed_func
)
4585 /* Disable only if we're not using arbfp fixed function fragment processing. If this is used,
4586 * keep GL_FRAGMENT_PROGRAM_ARB enabled, and the fixed function pipeline will bind the fixed function
4587 * replacement shader
4589 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4590 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4591 priv
->current_fprogram_id
= 0;
4595 struct arb_vs_compile_args compile_args
;
4596 struct arb_vs_compiled_shader
*compiled
;
4597 IWineD3DVertexShaderImpl
*vs
= (IWineD3DVertexShaderImpl
*) This
->stateBlock
->vertexShader
;
4599 TRACE("Using vertex shader %p\n", This
->stateBlock
->vertexShader
);
4600 find_arb_vs_compile_args(vs
, This
->stateBlock
, &compile_args
);
4601 compiled
= find_arb_vshader(vs
, &compile_args
);
4602 priv
->current_vprogram_id
= compiled
->prgId
;
4603 priv
->compiled_vprog
= compiled
;
4605 /* Bind the vertex program */
4606 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4607 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, priv->current_vprogram_id);");
4609 /* Enable OpenGL vertex programs */
4610 glEnable(GL_VERTEX_PROGRAM_ARB
);
4611 checkGLcall("glEnable(GL_VERTEX_PROGRAM_ARB);");
4612 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
4613 shader_arb_vs_local_constants(This
);
4615 if(priv
->last_vs_color_unclamp
!= compiled
->need_color_unclamp
) {
4616 priv
->last_vs_color_unclamp
= compiled
->need_color_unclamp
;
4618 if (gl_info
->supported
[ARB_COLOR_BUFFER_FLOAT
])
4620 GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB
, !compiled
->need_color_unclamp
));
4621 checkGLcall("glClampColorARB");
4623 FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
4627 else if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4629 priv
->current_vprogram_id
= 0;
4630 glDisable(GL_VERTEX_PROGRAM_ARB
);
4631 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4635 /* GL locking is done by the caller */
4636 static void shader_arb_select_depth_blt(IWineD3DDevice
*iface
, enum tex_types tex_type
, const SIZE
*ds_mask_size
)
4638 const float mask
[] = {0.0f
, 0.0f
, (float)ds_mask_size
->cx
, (float)ds_mask_size
->cy
};
4639 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4640 BOOL masked
= ds_mask_size
->cx
&& ds_mask_size
->cy
;
4641 struct shader_arb_priv
*priv
= This
->shader_priv
;
4642 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4643 GLuint
*blt_fprogram
;
4645 if (!priv
->depth_blt_vprogram_id
) priv
->depth_blt_vprogram_id
= create_arb_blt_vertex_program(gl_info
);
4646 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->depth_blt_vprogram_id
));
4647 glEnable(GL_VERTEX_PROGRAM_ARB
);
4649 blt_fprogram
= masked
? &priv
->depth_blt_fprogram_id_masked
[tex_type
] : &priv
->depth_blt_fprogram_id_full
[tex_type
];
4650 if (!*blt_fprogram
) *blt_fprogram
= create_arb_blt_fragment_program(gl_info
, tex_type
, masked
);
4651 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, *blt_fprogram
));
4652 if (masked
) GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, mask
));
4653 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
4656 /* GL locking is done by the caller */
4657 static void shader_arb_deselect_depth_blt(IWineD3DDevice
*iface
) {
4658 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4659 struct shader_arb_priv
*priv
= This
->shader_priv
;
4660 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4662 if (priv
->current_vprogram_id
) {
4663 GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB
, priv
->current_vprogram_id
));
4664 checkGLcall("glBindProgramARB(GL_VERTEX_PROGRAM_ARB, vertexShader->prgId);");
4666 TRACE("(%p) : Bound vertex program %u and enabled GL_VERTEX_PROGRAM_ARB\n", This
, priv
->current_vprogram_id
);
4668 glDisable(GL_VERTEX_PROGRAM_ARB
);
4669 checkGLcall("glDisable(GL_VERTEX_PROGRAM_ARB)");
4672 if (priv
->current_fprogram_id
) {
4673 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, priv
->current_fprogram_id
));
4674 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, pixelShader->prgId);");
4676 TRACE("(%p) : Bound fragment program %u and enabled GL_FRAGMENT_PROGRAM_ARB\n", This
, priv
->current_fprogram_id
);
4677 } else if(!priv
->use_arbfp_fixed_func
) {
4678 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
4679 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
4683 static void shader_arb_destroy(IWineD3DBaseShader
*iface
) {
4684 IWineD3DBaseShaderImpl
*baseShader
= (IWineD3DBaseShaderImpl
*) iface
;
4685 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*)baseShader
->baseShader
.device
;
4686 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
4688 if (shader_is_pshader_version(baseShader
->baseShader
.reg_maps
.shader_version
.type
))
4690 struct arb_pshader_private
*shader_data
= baseShader
->baseShader
.backend_data
;
4693 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4695 if (shader_data
->num_gl_shaders
)
4697 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4700 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4702 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4703 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4707 context_release(context
);
4710 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4711 HeapFree(GetProcessHeap(), 0, shader_data
);
4712 baseShader
->baseShader
.backend_data
= NULL
;
4716 struct arb_vshader_private
*shader_data
= baseShader
->baseShader
.backend_data
;
4719 if(!shader_data
) return; /* This can happen if a shader was never compiled */
4721 if (shader_data
->num_gl_shaders
)
4723 struct wined3d_context
*context
= context_acquire(device
, NULL
);
4726 for (i
= 0; i
< shader_data
->num_gl_shaders
; ++i
)
4728 GL_EXTCALL(glDeleteProgramsARB(1, &shader_data
->gl_shaders
[i
].prgId
));
4729 checkGLcall("GL_EXTCALL(glDeleteProgramsARB(1, &shader_data->gl_shaders[i].prgId))");
4733 context_release(context
);
4736 HeapFree(GetProcessHeap(), 0, shader_data
->gl_shaders
);
4737 HeapFree(GetProcessHeap(), 0, shader_data
);
4738 baseShader
->baseShader
.backend_data
= NULL
;
4742 static int sig_tree_compare(const void *key
, const struct wine_rb_entry
*entry
)
4744 struct ps_signature
*e
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4745 return compare_sig(key
, e
->sig
);
4748 static const struct wine_rb_functions sig_tree_functions
=
4756 static HRESULT
shader_arb_alloc(IWineD3DDevice
*iface
) {
4757 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4758 struct shader_arb_priv
*priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*priv
));
4759 if(wine_rb_init(&priv
->signature_tree
, &sig_tree_functions
) == -1)
4761 ERR("RB tree init failed\n");
4762 HeapFree(GetProcessHeap(), 0, priv
);
4763 return E_OUTOFMEMORY
;
4765 This
->shader_priv
= priv
;
4769 static void release_signature(struct wine_rb_entry
*entry
, void *context
)
4771 struct ps_signature
*sig
= WINE_RB_ENTRY_VALUE(entry
, struct ps_signature
, entry
);
4773 for(i
= 0; i
< MAX_REG_INPUT
; i
++)
4775 HeapFree(GetProcessHeap(), 0, (char *) sig
->sig
[i
].semantic_name
);
4777 HeapFree(GetProcessHeap(), 0, sig
->sig
);
4778 HeapFree(GetProcessHeap(), 0, sig
);
4781 /* Context activation is done by the caller. */
4782 static void shader_arb_free(IWineD3DDevice
*iface
) {
4783 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*)iface
;
4784 const struct wined3d_gl_info
*gl_info
= &This
->adapter
->gl_info
;
4785 struct shader_arb_priv
*priv
= This
->shader_priv
;
4789 if(priv
->depth_blt_vprogram_id
) {
4790 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_vprogram_id
));
4792 for (i
= 0; i
< tex_type_count
; ++i
)
4794 if (priv
->depth_blt_fprogram_id_full
[i
])
4796 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_full
[i
]));
4798 if (priv
->depth_blt_fprogram_id_masked
[i
])
4800 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->depth_blt_fprogram_id_masked
[i
]));
4805 wine_rb_destroy(&priv
->signature_tree
, release_signature
, NULL
);
4806 HeapFree(GetProcessHeap(), 0, This
->shader_priv
);
4809 static BOOL
shader_arb_dirty_const(IWineD3DDevice
*iface
) {
4813 static void shader_arb_get_caps(const struct wined3d_gl_info
*gl_info
, struct shader_caps
*pCaps
)
4815 if (gl_info
->supported
[ARB_VERTEX_PROGRAM
])
4819 /* 96 is the minimum allowed value of MAX_PROGRAM_ENV_PARAMETERS_ARB
4820 * for vertex programs. If the native limit is less than that it's
4821 * not very useful, and e.g. Mesa swrast returns 0, probably to
4822 * indicate it's a software implementation. */
4823 if (gl_info
->limits
.arb_vs_native_constants
< 96)
4824 vs_consts
= gl_info
->limits
.arb_vs_float_constants
;
4826 vs_consts
= min(gl_info
->limits
.arb_vs_float_constants
, gl_info
->limits
.arb_vs_native_constants
);
4828 if (gl_info
->supported
[NV_VERTEX_PROGRAM3
])
4830 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(3,0);
4831 TRACE_(d3d_caps
)("Hardware vertex shader version 3.0 enabled (NV_VERTEX_PROGRAM3)\n");
4833 else if (vs_consts
>= 256)
4835 /* Shader Model 2.0 requires at least 256 vertex shader constants */
4836 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(2,0);
4837 TRACE_(d3d_caps
)("Hardware vertex shader version 2.0 enabled (ARB_PROGRAM)\n");
4841 pCaps
->VertexShaderVersion
= WINED3DVS_VERSION(1,1);
4842 TRACE_(d3d_caps
)("Hardware vertex shader version 1.1 enabled (ARB_PROGRAM)\n");
4844 pCaps
->MaxVertexShaderConst
= vs_consts
;
4848 pCaps
->VertexShaderVersion
= 0;
4849 pCaps
->MaxVertexShaderConst
= 0;
4852 if (gl_info
->supported
[ARB_FRAGMENT_PROGRAM
])
4856 /* Similar as above for vertex programs, but the minimum for fragment
4857 * programs is 24. */
4858 if (gl_info
->limits
.arb_ps_native_constants
< 24)
4859 ps_consts
= gl_info
->limits
.arb_ps_float_constants
;
4861 ps_consts
= min(gl_info
->limits
.arb_ps_float_constants
, gl_info
->limits
.arb_ps_native_constants
);
4863 if (gl_info
->supported
[NV_FRAGMENT_PROGRAM2
])
4865 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(3,0);
4866 TRACE_(d3d_caps
)("Hardware pixel shader version 3.0 enabled (NV_FRAGMENT_PROGRAM2)\n");
4868 else if (ps_consts
>= 32)
4870 /* Shader Model 2.0 requires at least 32 pixel shader constants */
4871 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(2,0);
4872 TRACE_(d3d_caps
)("Hardware pixel shader version 2.0 enabled (ARB_PROGRAM)\n");
4876 pCaps
->PixelShaderVersion
= WINED3DPS_VERSION(1,4);
4877 TRACE_(d3d_caps
)("Hardware pixel shader version 1.4 enabled (ARB_PROGRAM)\n");
4879 pCaps
->PixelShader1xMaxValue
= 8.0f
;
4880 pCaps
->MaxPixelShaderConst
= ps_consts
;
4884 pCaps
->PixelShaderVersion
= 0;
4885 pCaps
->PixelShader1xMaxValue
= 0.0f
;
4886 pCaps
->MaxPixelShaderConst
= 0;
4889 pCaps
->VSClipping
= use_nv_clip(gl_info
);
4892 static BOOL
shader_arb_color_fixup_supported(struct color_fixup_desc fixup
)
4894 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
4896 TRACE("Checking support for color_fixup:\n");
4897 dump_color_fixup_desc(fixup
);
4900 /* We support everything except complex conversions. */
4901 if (!is_complex_fixup(fixup
))
4907 TRACE("[FAILED]\n");
4911 static void shader_arb_add_instruction_modifiers(const struct wined3d_shader_instruction
*ins
) {
4913 char write_mask
[20], regstr
[50];
4914 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
4915 BOOL is_color
= FALSE
;
4916 const struct wined3d_shader_dst_param
*dst
;
4918 if (!ins
->dst_count
) return;
4922 if(shift
== 0) return; /* Saturate alone is handled by the instructions */
4924 shader_arb_get_write_mask(ins
, dst
, write_mask
);
4925 shader_arb_get_register_name(ins
, &dst
->reg
, regstr
, &is_color
);
4927 /* Generate a line that does the output modifier computation
4928 * FIXME: _SAT vs shift? _SAT alone is already handled in the instructions, if this
4929 * maps problems in e.g. _d4_sat modify shader_arb_get_modifier
4931 shader_addline(buffer
, "MUL%s %s%s, %s, %s;\n", shader_arb_get_modifier(ins
),
4932 regstr
, write_mask
, regstr
, shift_tab
[shift
]);
4935 static const SHADER_HANDLER shader_arb_instruction_handler_table
[WINED3DSIH_TABLE_SIZE
] =
4937 /* WINED3DSIH_ABS */ shader_hw_map2gl
,
4938 /* WINED3DSIH_ADD */ shader_hw_map2gl
,
4939 /* WINED3DSIH_BEM */ pshader_hw_bem
,
4940 /* WINED3DSIH_BREAK */ shader_hw_break
,
4941 /* WINED3DSIH_BREAKC */ shader_hw_breakc
,
4942 /* WINED3DSIH_BREAKP */ NULL
,
4943 /* WINED3DSIH_CALL */ shader_hw_call
,
4944 /* WINED3DSIH_CALLNZ */ NULL
,
4945 /* WINED3DSIH_CMP */ pshader_hw_cmp
,
4946 /* WINED3DSIH_CND */ pshader_hw_cnd
,
4947 /* WINED3DSIH_CRS */ shader_hw_map2gl
,
4948 /* WINED3DSIH_CUT */ NULL
,
4949 /* WINED3DSIH_DCL */ NULL
,
4950 /* WINED3DSIH_DEF */ NULL
,
4951 /* WINED3DSIH_DEFB */ NULL
,
4952 /* WINED3DSIH_DEFI */ NULL
,
4953 /* WINED3DSIH_DP2ADD */ pshader_hw_dp2add
,
4954 /* WINED3DSIH_DP3 */ shader_hw_map2gl
,
4955 /* WINED3DSIH_DP4 */ shader_hw_map2gl
,
4956 /* WINED3DSIH_DST */ shader_hw_map2gl
,
4957 /* WINED3DSIH_DSX */ shader_hw_map2gl
,
4958 /* WINED3DSIH_DSY */ shader_hw_dsy
,
4959 /* WINED3DSIH_ELSE */ shader_hw_else
,
4960 /* WINED3DSIH_EMIT */ NULL
,
4961 /* WINED3DSIH_ENDIF */ shader_hw_endif
,
4962 /* WINED3DSIH_ENDLOOP */ shader_hw_endloop
,
4963 /* WINED3DSIH_ENDREP */ shader_hw_endrep
,
4964 /* WINED3DSIH_EXP */ shader_hw_scalar_op
,
4965 /* WINED3DSIH_EXPP */ shader_hw_scalar_op
,
4966 /* WINED3DSIH_FRC */ shader_hw_map2gl
,
4967 /* WINED3DSIH_IADD */ NULL
,
4968 /* WINED3DSIH_IF */ NULL
/* Hardcoded into the shader */,
4969 /* WINED3DSIH_IFC */ shader_hw_ifc
,
4970 /* WINED3DSIH_IGE */ NULL
,
4971 /* WINED3DSIH_LABEL */ shader_hw_label
,
4972 /* WINED3DSIH_LIT */ shader_hw_map2gl
,
4973 /* WINED3DSIH_LOG */ shader_hw_log_pow
,
4974 /* WINED3DSIH_LOGP */ shader_hw_log_pow
,
4975 /* WINED3DSIH_LOOP */ shader_hw_loop
,
4976 /* WINED3DSIH_LRP */ shader_hw_lrp
,
4977 /* WINED3DSIH_LT */ NULL
,
4978 /* WINED3DSIH_M3x2 */ shader_hw_mnxn
,
4979 /* WINED3DSIH_M3x3 */ shader_hw_mnxn
,
4980 /* WINED3DSIH_M3x4 */ shader_hw_mnxn
,
4981 /* WINED3DSIH_M4x3 */ shader_hw_mnxn
,
4982 /* WINED3DSIH_M4x4 */ shader_hw_mnxn
,
4983 /* WINED3DSIH_MAD */ shader_hw_map2gl
,
4984 /* WINED3DSIH_MAX */ shader_hw_map2gl
,
4985 /* WINED3DSIH_MIN */ shader_hw_map2gl
,
4986 /* WINED3DSIH_MOV */ shader_hw_mov
,
4987 /* WINED3DSIH_MOVA */ shader_hw_mov
,
4988 /* WINED3DSIH_MUL */ shader_hw_map2gl
,
4989 /* WINED3DSIH_NOP */ shader_hw_nop
,
4990 /* WINED3DSIH_NRM */ shader_hw_nrm
,
4991 /* WINED3DSIH_PHASE */ NULL
,
4992 /* WINED3DSIH_POW */ shader_hw_log_pow
,
4993 /* WINED3DSIH_RCP */ shader_hw_rcp
,
4994 /* WINED3DSIH_REP */ shader_hw_rep
,
4995 /* WINED3DSIH_RET */ shader_hw_ret
,
4996 /* WINED3DSIH_RSQ */ shader_hw_scalar_op
,
4997 /* WINED3DSIH_SETP */ NULL
,
4998 /* WINED3DSIH_SGE */ shader_hw_map2gl
,
4999 /* WINED3DSIH_SGN */ shader_hw_sgn
,
5000 /* WINED3DSIH_SINCOS */ shader_hw_sincos
,
5001 /* WINED3DSIH_SLT */ shader_hw_map2gl
,
5002 /* WINED3DSIH_SUB */ shader_hw_map2gl
,
5003 /* WINED3DSIH_TEX */ pshader_hw_tex
,
5004 /* WINED3DSIH_TEXBEM */ pshader_hw_texbem
,
5005 /* WINED3DSIH_TEXBEML */ pshader_hw_texbem
,
5006 /* WINED3DSIH_TEXCOORD */ pshader_hw_texcoord
,
5007 /* WINED3DSIH_TEXDEPTH */ pshader_hw_texdepth
,
5008 /* WINED3DSIH_TEXDP3 */ pshader_hw_texdp3
,
5009 /* WINED3DSIH_TEXDP3TEX */ pshader_hw_texdp3tex
,
5010 /* WINED3DSIH_TEXKILL */ pshader_hw_texkill
,
5011 /* WINED3DSIH_TEXLDD */ shader_hw_texldd
,
5012 /* WINED3DSIH_TEXLDL */ shader_hw_texldl
,
5013 /* WINED3DSIH_TEXM3x2DEPTH */ pshader_hw_texm3x2depth
,
5014 /* WINED3DSIH_TEXM3x2PAD */ pshader_hw_texm3x2pad
,
5015 /* WINED3DSIH_TEXM3x2TEX */ pshader_hw_texm3x2tex
,
5016 /* WINED3DSIH_TEXM3x3 */ pshader_hw_texm3x3
,
5017 /* WINED3DSIH_TEXM3x3DIFF */ NULL
,
5018 /* WINED3DSIH_TEXM3x3PAD */ pshader_hw_texm3x3pad
,
5019 /* WINED3DSIH_TEXM3x3SPEC */ pshader_hw_texm3x3spec
,
5020 /* WINED3DSIH_TEXM3x3TEX */ pshader_hw_texm3x3tex
,
5021 /* WINED3DSIH_TEXM3x3VSPEC */ pshader_hw_texm3x3vspec
,
5022 /* WINED3DSIH_TEXREG2AR */ pshader_hw_texreg2ar
,
5023 /* WINED3DSIH_TEXREG2GB */ pshader_hw_texreg2gb
,
5024 /* WINED3DSIH_TEXREG2RGB */ pshader_hw_texreg2rgb
,
5027 static inline BOOL
get_bool_const(const struct wined3d_shader_instruction
*ins
, IWineD3DBaseShaderImpl
*This
, DWORD idx
)
5029 BOOL vshader
= shader_is_vshader_version(This
->baseShader
.reg_maps
.shader_version
.type
);
5031 WORD flag
= (1 << idx
);
5032 const local_constant
*constant
;
5033 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5035 if(This
->baseShader
.reg_maps
.local_bool_consts
& flag
)
5037 /* What good is a if(bool) with a hardcoded local constant? I don't know, but handle it */
5038 LIST_FOR_EACH_ENTRY(constant
, &This
->baseShader
.constantsB
, local_constant
, entry
)
5040 if (constant
->idx
== idx
)
5042 return constant
->value
[0];
5045 ERR("Local constant not found\n");
5050 if(vshader
) bools
= priv
->cur_vs_args
->clip
.boolclip
.bools
;
5051 else bools
= priv
->cur_ps_args
->bools
;
5052 return bools
& flag
;
5056 static void get_loop_control_const(const struct wined3d_shader_instruction
*ins
,
5057 IWineD3DBaseShaderImpl
*This
, UINT idx
, struct wined3d_shader_loop_control
*loop_control
)
5059 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5061 /* Integer constants can either be a local constant, or they can be stored in the shader
5062 * type specific compile args. */
5063 if (This
->baseShader
.reg_maps
.local_int_consts
& (1 << idx
))
5065 const local_constant
*constant
;
5067 LIST_FOR_EACH_ENTRY(constant
, &This
->baseShader
.constantsI
, local_constant
, entry
)
5069 if (constant
->idx
== idx
)
5071 loop_control
->count
= constant
->value
[0];
5072 loop_control
->start
= constant
->value
[1];
5073 /* Step is signed. */
5074 loop_control
->step
= (int)constant
->value
[2];
5078 /* If this happens the flag was set incorrectly */
5079 ERR("Local constant not found\n");
5080 loop_control
->count
= 0;
5081 loop_control
->start
= 0;
5082 loop_control
->step
= 0;
5086 switch (This
->baseShader
.reg_maps
.shader_version
.type
)
5088 case WINED3D_SHADER_TYPE_VERTEX
:
5089 /* Count and aL start value are unsigned */
5090 loop_control
->count
= priv
->cur_vs_args
->loop_ctrl
[idx
][0];
5091 loop_control
->start
= priv
->cur_vs_args
->loop_ctrl
[idx
][1];
5092 /* Step is signed. */
5093 loop_control
->step
= ((char)priv
->cur_vs_args
->loop_ctrl
[idx
][2]);
5096 case WINED3D_SHADER_TYPE_PIXEL
:
5097 loop_control
->count
= priv
->cur_ps_args
->loop_ctrl
[idx
][0];
5098 loop_control
->start
= priv
->cur_ps_args
->loop_ctrl
[idx
][1];
5099 loop_control
->step
= ((char)priv
->cur_ps_args
->loop_ctrl
[idx
][2]);
5103 FIXME("Unhandled shader type %#x.\n", This
->baseShader
.reg_maps
.shader_version
.type
);
5108 static void record_instruction(struct list
*list
, const struct wined3d_shader_instruction
*ins
)
5111 struct wined3d_shader_dst_param
*dst_param
= NULL
;
5112 struct wined3d_shader_src_param
*src_param
= NULL
, *rel_addr
= NULL
;
5113 struct recorded_instruction
*rec
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*rec
));
5116 ERR("Out of memory\n");
5121 dst_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
));
5122 if(!dst_param
) goto free
;
5123 *dst_param
= *ins
->dst
;
5124 if(ins
->dst
->reg
.rel_addr
)
5126 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*dst_param
->reg
.rel_addr
));
5127 if(!rel_addr
) goto free
;
5128 *rel_addr
= *ins
->dst
->reg
.rel_addr
;
5129 dst_param
->reg
.rel_addr
= rel_addr
;
5131 rec
->ins
.dst
= dst_param
;
5133 src_param
= HeapAlloc(GetProcessHeap(), 0, sizeof(*src_param
) * ins
->src_count
);
5134 if(!src_param
) goto free
;
5135 for(i
= 0; i
< ins
->src_count
; i
++)
5137 src_param
[i
] = ins
->src
[i
];
5138 if(ins
->src
[i
].reg
.rel_addr
)
5140 rel_addr
= HeapAlloc(GetProcessHeap(), 0, sizeof(*rel_addr
));
5141 if(!rel_addr
) goto free
;
5142 *rel_addr
= *ins
->src
[i
].reg
.rel_addr
;
5143 src_param
[i
].reg
.rel_addr
= rel_addr
;
5146 rec
->ins
.src
= src_param
;
5147 list_add_tail(list
, &rec
->entry
);
5151 ERR("Out of memory\n");
5154 HeapFree(GetProcessHeap(), 0, (void *) dst_param
->reg
.rel_addr
);
5155 HeapFree(GetProcessHeap(), 0, dst_param
);
5159 for(i
= 0; i
< ins
->src_count
; i
++)
5161 HeapFree(GetProcessHeap(), 0, (void *) src_param
[i
].reg
.rel_addr
);
5163 HeapFree(GetProcessHeap(), 0, src_param
);
5165 HeapFree(GetProcessHeap(), 0, rec
);
5168 static void free_recorded_instruction(struct list
*list
)
5170 struct recorded_instruction
*rec_ins
, *entry2
;
5173 LIST_FOR_EACH_ENTRY_SAFE(rec_ins
, entry2
, list
, struct recorded_instruction
, entry
)
5175 list_remove(&rec_ins
->entry
);
5176 if(rec_ins
->ins
.dst
)
5178 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.dst
->reg
.rel_addr
);
5179 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.dst
);
5181 if(rec_ins
->ins
.src
)
5183 for(i
= 0; i
< rec_ins
->ins
.src_count
; i
++)
5185 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.src
[i
].reg
.rel_addr
);
5187 HeapFree(GetProcessHeap(), 0, (void *) rec_ins
->ins
.src
);
5189 HeapFree(GetProcessHeap(), 0, rec_ins
);
5193 static void shader_arb_handle_instruction(const struct wined3d_shader_instruction
*ins
) {
5194 SHADER_HANDLER hw_fct
;
5195 struct shader_arb_ctx_priv
*priv
= ins
->ctx
->backend_data
;
5196 IWineD3DBaseShaderImpl
*This
= (IWineD3DBaseShaderImpl
*)ins
->ctx
->shader
;
5197 struct control_frame
*control_frame
;
5198 struct wined3d_shader_buffer
*buffer
= ins
->ctx
->buffer
;
5201 if(ins
->handler_idx
== WINED3DSIH_LOOP
|| ins
->handler_idx
== WINED3DSIH_REP
)
5203 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5204 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5206 if(ins
->handler_idx
== WINED3DSIH_LOOP
) control_frame
->type
= LOOP
;
5207 if(ins
->handler_idx
== WINED3DSIH_REP
) control_frame
->type
= REP
;
5209 if(priv
->target_version
>= NV2
)
5211 control_frame
->no
.loop
= priv
->num_loops
++;
5216 /* Don't bother recording when we're in a not used if branch */
5222 if(!priv
->recording
)
5224 list_init(&priv
->record
);
5225 priv
->recording
= TRUE
;
5226 control_frame
->outer_loop
= TRUE
;
5227 get_loop_control_const(ins
, This
, ins
->src
[0].reg
.idx
, &control_frame
->loop_control
);
5228 return; /* Instruction is handled */
5230 /* Record this loop in the outer loop's recording */
5233 else if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5235 if(priv
->target_version
>= NV2
)
5237 /* Nothing to do. The control frame is popped after the HW instr handler */
5241 struct list
*e
= list_head(&priv
->control_frames
);
5242 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5243 list_remove(&control_frame
->entry
);
5245 if(control_frame
->outer_loop
)
5247 int iteration
, aL
= 0;
5250 /* Turn off recording before playback */
5251 priv
->recording
= FALSE
;
5253 /* Move the recorded instructions to a separate list and get them out of the private data
5254 * structure. If there are nested loops, the shader_arb_handle_instruction below will
5255 * be recorded again, thus priv->record might be overwritten
5258 list_move_tail(©
, &priv
->record
);
5259 list_init(&priv
->record
);
5261 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5263 shader_addline(buffer
, "#unrolling loop: %u iterations, aL=%u, inc %d\n",
5264 control_frame
->loop_control
.count
, control_frame
->loop_control
.start
,
5265 control_frame
->loop_control
.step
);
5266 aL
= control_frame
->loop_control
.start
;
5270 shader_addline(buffer
, "#unrolling rep: %u iterations\n", control_frame
->loop_control
.count
);
5273 for (iteration
= 0; iteration
< control_frame
->loop_control
.count
; ++iteration
)
5275 struct recorded_instruction
*rec_ins
;
5276 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5279 shader_addline(buffer
, "#Iteration %d, aL=%d\n", iteration
, aL
);
5283 shader_addline(buffer
, "#Iteration %d\n", iteration
);
5286 LIST_FOR_EACH_ENTRY(rec_ins
, ©
, struct recorded_instruction
, entry
)
5288 shader_arb_handle_instruction(&rec_ins
->ins
);
5291 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
)
5293 aL
+= control_frame
->loop_control
.step
;
5296 shader_addline(buffer
, "#end loop/rep\n");
5298 free_recorded_instruction(©
);
5299 HeapFree(GetProcessHeap(), 0, control_frame
);
5300 return; /* Instruction is handled */
5304 /* This is a nested loop. Proceed to the normal recording function */
5305 HeapFree(GetProcessHeap(), 0, control_frame
);
5312 record_instruction(&priv
->record
, ins
);
5317 if(ins
->handler_idx
== WINED3DSIH_IF
)
5319 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5320 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5321 control_frame
->type
= IF
;
5323 bool_const
= get_bool_const(ins
, This
, ins
->src
[0].reg
.idx
);
5324 if(ins
->src
[0].modifiers
== WINED3DSPSM_NOT
) bool_const
= !bool_const
;
5325 if(!priv
->muted
&& bool_const
== FALSE
)
5327 shader_addline(buffer
, "#if(FALSE){\n");
5329 control_frame
->muting
= TRUE
;
5331 else shader_addline(buffer
, "#if(TRUE) {\n");
5333 return; /* Instruction is handled */
5335 else if(ins
->handler_idx
== WINED3DSIH_IFC
)
5337 /* IF(bool) and if_cond(a, b) use the same ELSE and ENDIF tokens */
5338 control_frame
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(*control_frame
));
5339 control_frame
->type
= IFC
;
5340 control_frame
->no
.ifc
= priv
->num_ifcs
++;
5341 list_add_head(&priv
->control_frames
, &control_frame
->entry
);
5343 else if(ins
->handler_idx
== WINED3DSIH_ELSE
)
5345 struct list
*e
= list_head(&priv
->control_frames
);
5346 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5348 if(control_frame
->type
== IF
)
5350 shader_addline(buffer
, "#} else {\n");
5351 if(!priv
->muted
&& !control_frame
->muting
)
5354 control_frame
->muting
= TRUE
;
5356 else if(control_frame
->muting
) priv
->muted
= FALSE
;
5357 return; /* Instruction is handled. */
5359 /* In case of an ifc, generate a HW shader instruction */
5361 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5363 struct list
*e
= list_head(&priv
->control_frames
);
5364 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5366 if(control_frame
->type
== IF
)
5368 shader_addline(buffer
, "#} endif\n");
5369 if(control_frame
->muting
) priv
->muted
= FALSE
;
5370 list_remove(&control_frame
->entry
);
5371 HeapFree(GetProcessHeap(), 0, control_frame
);
5372 return; /* Instruction is handled */
5376 if(priv
->muted
) return;
5378 /* Select handler */
5379 hw_fct
= shader_arb_instruction_handler_table
[ins
->handler_idx
];
5381 /* Unhandled opcode */
5384 FIXME("Backend can't handle opcode %#x\n", ins
->handler_idx
);
5389 if(ins
->handler_idx
== WINED3DSIH_ENDLOOP
|| ins
->handler_idx
== WINED3DSIH_ENDREP
)
5391 struct list
*e
= list_head(&priv
->control_frames
);
5392 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5393 list_remove(&control_frame
->entry
);
5394 HeapFree(GetProcessHeap(), 0, control_frame
);
5397 else if(ins
->handler_idx
== WINED3DSIH_ENDIF
)
5399 /* Non-ifc ENDIFs don't reach that place because of the return in the if block above */
5400 struct list
*e
= list_head(&priv
->control_frames
);
5401 control_frame
= LIST_ENTRY(e
, struct control_frame
, entry
);
5402 list_remove(&control_frame
->entry
);
5403 HeapFree(GetProcessHeap(), 0, control_frame
);
5407 shader_arb_add_instruction_modifiers(ins
);
5410 const shader_backend_t arb_program_shader_backend
= {
5411 shader_arb_handle_instruction
,
5413 shader_arb_select_depth_blt
,
5414 shader_arb_deselect_depth_blt
,
5415 shader_arb_update_float_vertex_constants
,
5416 shader_arb_update_float_pixel_constants
,
5417 shader_arb_load_constants
,
5418 shader_arb_load_np2fixup_constants
,
5422 shader_arb_dirty_const
,
5423 shader_arb_get_caps
,
5424 shader_arb_color_fixup_supported
,
5427 /* ARB_fragment_program fixed function pipeline replacement definitions */
5428 #define ARB_FFP_CONST_TFACTOR 0
5429 #define ARB_FFP_CONST_SPECULAR_ENABLE ((ARB_FFP_CONST_TFACTOR) + 1)
5430 #define ARB_FFP_CONST_CONSTANT(i) ((ARB_FFP_CONST_SPECULAR_ENABLE) + 1 + i)
5431 #define ARB_FFP_CONST_BUMPMAT(i) ((ARB_FFP_CONST_CONSTANT(7)) + 1 + i)
5432 #define ARB_FFP_CONST_LUMINANCE(i) ((ARB_FFP_CONST_BUMPMAT(7)) + 1 + i)
5434 struct arbfp_ffp_desc
5436 struct ffp_frag_desc parent
;
5438 unsigned int num_textures_used
;
5441 /* Context activation is done by the caller. */
5442 static void arbfp_enable(IWineD3DDevice
*iface
, BOOL enable
) {
5445 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
5446 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
5448 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
5449 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
5454 static HRESULT
arbfp_alloc(IWineD3DDevice
*iface
) {
5455 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
5456 struct shader_arb_priv
*priv
;
5457 /* Share private data between the shader backend and the pipeline replacement, if both
5458 * are the arb implementation. This is needed to figure out whether ARBfp should be disabled
5459 * if no pixel shader is bound or not
5461 if(This
->shader_backend
== &arb_program_shader_backend
) {
5462 This
->fragment_priv
= This
->shader_priv
;
5464 This
->fragment_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct shader_arb_priv
));
5465 if(!This
->fragment_priv
) return E_OUTOFMEMORY
;
5467 priv
= This
->fragment_priv
;
5468 if (wine_rb_init(&priv
->fragment_shaders
, &wined3d_ffp_frag_program_rb_functions
) == -1)
5470 ERR("Failed to initialize rbtree.\n");
5471 HeapFree(GetProcessHeap(), 0, This
->fragment_priv
);
5472 return E_OUTOFMEMORY
;
5474 priv
->use_arbfp_fixed_func
= TRUE
;
5478 /* Context activation is done by the caller. */
5479 static void arbfp_free_ffpshader(struct wine_rb_entry
*entry
, void *context
)
5481 const struct wined3d_gl_info
*gl_info
= context
;
5482 struct arbfp_ffp_desc
*entry_arb
= WINE_RB_ENTRY_VALUE(entry
, struct arbfp_ffp_desc
, parent
.entry
);
5485 GL_EXTCALL(glDeleteProgramsARB(1, &entry_arb
->shader
));
5486 checkGLcall("glDeleteProgramsARB(1, &entry_arb->shader)");
5487 HeapFree(GetProcessHeap(), 0, entry_arb
);
5491 /* Context activation is done by the caller. */
5492 static void arbfp_free(IWineD3DDevice
*iface
) {
5493 IWineD3DDeviceImpl
*This
= (IWineD3DDeviceImpl
*) iface
;
5494 struct shader_arb_priv
*priv
= This
->fragment_priv
;
5496 wine_rb_destroy(&priv
->fragment_shaders
, arbfp_free_ffpshader
, &This
->adapter
->gl_info
);
5497 priv
->use_arbfp_fixed_func
= FALSE
;
5499 if(This
->shader_backend
!= &arb_program_shader_backend
) {
5500 HeapFree(GetProcessHeap(), 0, This
->fragment_priv
);
5504 static void arbfp_get_caps(const struct wined3d_gl_info
*gl_info
, struct fragment_caps
*caps
)
5506 caps
->PrimitiveMiscCaps
= WINED3DPMISCCAPS_TSSARGTEMP
;
5507 caps
->TextureOpCaps
= WINED3DTEXOPCAPS_DISABLE
|
5508 WINED3DTEXOPCAPS_SELECTARG1
|
5509 WINED3DTEXOPCAPS_SELECTARG2
|
5510 WINED3DTEXOPCAPS_MODULATE4X
|
5511 WINED3DTEXOPCAPS_MODULATE2X
|
5512 WINED3DTEXOPCAPS_MODULATE
|
5513 WINED3DTEXOPCAPS_ADDSIGNED2X
|
5514 WINED3DTEXOPCAPS_ADDSIGNED
|
5515 WINED3DTEXOPCAPS_ADD
|
5516 WINED3DTEXOPCAPS_SUBTRACT
|
5517 WINED3DTEXOPCAPS_ADDSMOOTH
|
5518 WINED3DTEXOPCAPS_BLENDCURRENTALPHA
|
5519 WINED3DTEXOPCAPS_BLENDFACTORALPHA
|
5520 WINED3DTEXOPCAPS_BLENDTEXTUREALPHA
|
5521 WINED3DTEXOPCAPS_BLENDDIFFUSEALPHA
|
5522 WINED3DTEXOPCAPS_BLENDTEXTUREALPHAPM
|
5523 WINED3DTEXOPCAPS_MODULATEALPHA_ADDCOLOR
|
5524 WINED3DTEXOPCAPS_MODULATECOLOR_ADDALPHA
|
5525 WINED3DTEXOPCAPS_MODULATEINVCOLOR_ADDALPHA
|
5526 WINED3DTEXOPCAPS_MODULATEINVALPHA_ADDCOLOR
|
5527 WINED3DTEXOPCAPS_DOTPRODUCT3
|
5528 WINED3DTEXOPCAPS_MULTIPLYADD
|
5529 WINED3DTEXOPCAPS_LERP
|
5530 WINED3DTEXOPCAPS_BUMPENVMAP
|
5531 WINED3DTEXOPCAPS_BUMPENVMAPLUMINANCE
;
5533 /* TODO: Implement WINED3DTEXOPCAPS_PREMODULATE */
5535 caps
->MaxTextureBlendStages
= 8;
5536 caps
->MaxSimultaneousTextures
= min(gl_info
->limits
.fragment_samplers
, 8);
5539 static void state_texfactor_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5541 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5542 IWineD3DDeviceImpl
*device
= stateblock
->device
;
5545 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5546 * application provided constants
5548 if(device
->shader_backend
== &arb_program_shader_backend
) {
5549 if (use_ps(stateblock
)) return;
5551 context
->pshader_const_dirty
[ARB_FFP_CONST_TFACTOR
] = 1;
5552 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_TFACTOR
+ 1);
5555 D3DCOLORTOGLFLOAT4(stateblock
->renderState
[WINED3DRS_TEXTUREFACTOR
], col
);
5556 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_TFACTOR
, col
));
5557 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_TFACTOR, col)");
5561 static void state_arb_specularenable(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5563 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5564 IWineD3DDeviceImpl
*device
= stateblock
->device
;
5567 /* Don't load the parameter if we're using an arbfp pixel shader, otherwise we'll overwrite
5568 * application provided constants
5570 if(device
->shader_backend
== &arb_program_shader_backend
) {
5571 if (use_ps(stateblock
)) return;
5573 context
->pshader_const_dirty
[ARB_FFP_CONST_SPECULAR_ENABLE
] = 1;
5574 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_SPECULAR_ENABLE
+ 1);
5577 if(stateblock
->renderState
[WINED3DRS_SPECULARENABLE
]) {
5578 /* The specular color has no alpha */
5579 col
[0] = 1.0f
; col
[1] = 1.0f
;
5580 col
[2] = 1.0f
; col
[3] = 0.0f
;
5582 col
[0] = 0.0f
; col
[1] = 0.0f
;
5583 col
[2] = 0.0f
; col
[3] = 0.0f
;
5585 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_SPECULAR_ENABLE
, col
));
5586 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_SPECULAR_ENABLE, col)");
5589 static void set_bumpmat_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5591 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5592 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5593 IWineD3DDeviceImpl
*device
= stateblock
->device
;
5596 if (use_ps(stateblock
))
5599 && (((IWineD3DPixelShaderImpl
*)stateblock
->pixelShader
)->baseShader
.reg_maps
.bumpmat
& (1 << stage
)))
5601 /* The pixel shader has to know the bump env matrix. Do a constants update if it isn't scheduled
5604 if (!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
))
5605 stateblock_apply_state(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
5608 if(device
->shader_backend
== &arb_program_shader_backend
) {
5609 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
5612 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
5613 context
->pshader_const_dirty
[ARB_FFP_CONST_BUMPMAT(stage
)] = 1;
5614 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_BUMPMAT(stage
) + 1);
5617 mat
[0][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT00
]);
5618 mat
[0][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT01
]);
5619 mat
[1][0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT10
]);
5620 mat
[1][1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVMAT11
]);
5622 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_BUMPMAT(stage
), &mat
[0][0]));
5623 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_BUMPMAT(stage), &mat[0][0])");
5626 static void tex_bumpenvlum_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
5628 DWORD stage
= (state
- STATE_TEXTURESTAGE(0, 0)) / (WINED3D_HIGHEST_TEXTURE_STATE
+ 1);
5629 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
5630 IWineD3DDeviceImpl
*device
= stateblock
->device
;
5633 if (use_ps(stateblock
))
5636 && (((IWineD3DPixelShaderImpl
*)stateblock
->pixelShader
)->baseShader
.reg_maps
.luminanceparams
& (1 << stage
)))
5638 /* The pixel shader has to know the luminance offset. Do a constants update if it
5639 * isn't scheduled anyway
5641 if (!isStateDirty(context
, STATE_PIXELSHADERCONSTANT
))
5642 stateblock_apply_state(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
5645 if(device
->shader_backend
== &arb_program_shader_backend
) {
5646 /* Exit now, don't set the bumpmat below, otherwise we may overwrite pixel shader constants */
5649 } else if(device
->shader_backend
== &arb_program_shader_backend
) {
5650 context
->pshader_const_dirty
[ARB_FFP_CONST_LUMINANCE(stage
)] = 1;
5651 device
->highest_dirty_ps_const
= max(device
->highest_dirty_ps_const
, ARB_FFP_CONST_LUMINANCE(stage
) + 1);
5654 param
[0] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLSCALE
]);
5655 param
[1] = *((float *) &stateblock
->textureState
[stage
][WINED3DTSS_BUMPENVLOFFSET
]);
5659 GL_EXTCALL(glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, ARB_FFP_CONST_LUMINANCE(stage
), param
));
5660 checkGLcall("glProgramEnvParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, ARB_FFP_CONST_LUMINANCE(stage), param)");
5663 static const char *get_argreg(struct wined3d_shader_buffer
*buffer
, DWORD argnum
, unsigned int stage
, DWORD arg
)
5667 if(arg
== ARG_UNUSED
) return "unused"; /* This is the marker for unused registers */
5669 switch(arg
& WINED3DTA_SELECTMASK
) {
5670 case WINED3DTA_DIFFUSE
:
5671 ret
= "fragment.color.primary"; break;
5673 case WINED3DTA_CURRENT
:
5674 if(stage
== 0) ret
= "fragment.color.primary";
5678 case WINED3DTA_TEXTURE
:
5680 case 0: ret
= "tex0"; break;
5681 case 1: ret
= "tex1"; break;
5682 case 2: ret
= "tex2"; break;
5683 case 3: ret
= "tex3"; break;
5684 case 4: ret
= "tex4"; break;
5685 case 5: ret
= "tex5"; break;
5686 case 6: ret
= "tex6"; break;
5687 case 7: ret
= "tex7"; break;
5688 default: ret
= "unknown texture";
5692 case WINED3DTA_TFACTOR
:
5693 ret
= "tfactor"; break;
5695 case WINED3DTA_SPECULAR
:
5696 ret
= "fragment.color.secondary"; break;
5698 case WINED3DTA_TEMP
:
5699 ret
= "tempreg"; break;
5701 case WINED3DTA_CONSTANT
:
5702 FIXME("Implement perstage constants\n");
5704 case 0: ret
= "const0"; break;
5705 case 1: ret
= "const1"; break;
5706 case 2: ret
= "const2"; break;
5707 case 3: ret
= "const3"; break;
5708 case 4: ret
= "const4"; break;
5709 case 5: ret
= "const5"; break;
5710 case 6: ret
= "const6"; break;
5711 case 7: ret
= "const7"; break;
5712 default: ret
= "unknown constant";
5720 if(arg
& WINED3DTA_COMPLEMENT
) {
5721 shader_addline(buffer
, "SUB arg%u, const.x, %s;\n", argnum
, ret
);
5722 if(argnum
== 0) ret
= "arg0";
5723 if(argnum
== 1) ret
= "arg1";
5724 if(argnum
== 2) ret
= "arg2";
5726 if(arg
& WINED3DTA_ALPHAREPLICATE
) {
5727 shader_addline(buffer
, "MOV arg%u, %s.w;\n", argnum
, ret
);
5728 if(argnum
== 0) ret
= "arg0";
5729 if(argnum
== 1) ret
= "arg1";
5730 if(argnum
== 2) ret
= "arg2";
5735 static void gen_ffp_instr(struct wined3d_shader_buffer
*buffer
, unsigned int stage
, BOOL color
,
5736 BOOL alpha
, DWORD dst
, DWORD op
, DWORD dw_arg0
, DWORD dw_arg1
, DWORD dw_arg2
)
5738 const char *dstmask
, *dstreg
, *arg0
, *arg1
, *arg2
;
5739 unsigned int mul
= 1;
5740 BOOL mul_final_dest
= FALSE
;
5742 if(color
&& alpha
) dstmask
= "";
5743 else if(color
) dstmask
= ".xyz";
5744 else dstmask
= ".w";
5746 if(dst
== tempreg
) dstreg
= "tempreg";
5747 else dstreg
= "ret";
5749 arg0
= get_argreg(buffer
, 0, stage
, dw_arg0
);
5750 arg1
= get_argreg(buffer
, 1, stage
, dw_arg1
);
5751 arg2
= get_argreg(buffer
, 2, stage
, dw_arg2
);
5754 case WINED3DTOP_DISABLE
:
5755 if(stage
== 0) shader_addline(buffer
, "MOV %s%s, fragment.color.primary;\n", dstreg
, dstmask
);
5758 case WINED3DTOP_SELECTARG2
:
5760 case WINED3DTOP_SELECTARG1
:
5761 shader_addline(buffer
, "MOV %s%s, %s;\n", dstreg
, dstmask
, arg1
);
5764 case WINED3DTOP_MODULATE4X
:
5766 case WINED3DTOP_MODULATE2X
:
5768 if(strcmp(dstreg
, "result.color") == 0) {
5770 mul_final_dest
= TRUE
;
5772 case WINED3DTOP_MODULATE
:
5773 shader_addline(buffer
, "MUL %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5776 case WINED3DTOP_ADDSIGNED2X
:
5778 if(strcmp(dstreg
, "result.color") == 0) {
5780 mul_final_dest
= TRUE
;
5782 case WINED3DTOP_ADDSIGNED
:
5783 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
5785 case WINED3DTOP_ADD
:
5786 shader_addline(buffer
, "ADD_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5789 case WINED3DTOP_SUBTRACT
:
5790 shader_addline(buffer
, "SUB_SAT %s%s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
);
5793 case WINED3DTOP_ADDSMOOTH
:
5794 shader_addline(buffer
, "SUB arg1, const.x, %s;\n", arg1
);
5795 shader_addline(buffer
, "MAD_SAT %s%s, arg1, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5798 case WINED3DTOP_BLENDCURRENTALPHA
:
5799 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_CURRENT
);
5800 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5802 case WINED3DTOP_BLENDFACTORALPHA
:
5803 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TFACTOR
);
5804 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5806 case WINED3DTOP_BLENDTEXTUREALPHA
:
5807 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
5808 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5810 case WINED3DTOP_BLENDDIFFUSEALPHA
:
5811 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_DIFFUSE
);
5812 shader_addline(buffer
, "LRP %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5815 case WINED3DTOP_BLENDTEXTUREALPHAPM
:
5816 arg0
= get_argreg(buffer
, 0, stage
, WINED3DTA_TEXTURE
);
5817 shader_addline(buffer
, "SUB arg0.w, const.x, %s.w;\n", arg0
);
5818 shader_addline(buffer
, "MAD_SAT %s%s, %s, arg0.w, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5821 /* D3DTOP_PREMODULATE ???? */
5823 case WINED3DTOP_MODULATEINVALPHA_ADDCOLOR
:
5824 shader_addline(buffer
, "SUB arg0.w, const.x, %s;\n", arg1
);
5825 shader_addline(buffer
, "MAD_SAT %s%s, arg0.w, %s, %s;\n", dstreg
, dstmask
, arg2
, arg1
);
5827 case WINED3DTOP_MODULATEALPHA_ADDCOLOR
:
5828 shader_addline(buffer
, "MAD_SAT %s%s, %s.w, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
5830 case WINED3DTOP_MODULATEINVCOLOR_ADDALPHA
:
5831 shader_addline(buffer
, "SUB arg0, const.x, %s;\n", arg1
);
5832 shader_addline(buffer
, "MAD_SAT %s%s, arg0, %s, %s.w;\n", dstreg
, dstmask
, arg2
, arg1
);
5834 case WINED3DTOP_MODULATECOLOR_ADDALPHA
:
5835 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s.w;\n", dstreg
, dstmask
, arg1
, arg2
, arg1
);
5838 case WINED3DTOP_DOTPRODUCT3
:
5840 if(strcmp(dstreg
, "result.color") == 0) {
5842 mul_final_dest
= TRUE
;
5844 shader_addline(buffer
, "SUB arg1, %s, const.w;\n", arg1
);
5845 shader_addline(buffer
, "SUB arg2, %s, const.w;\n", arg2
);
5846 shader_addline(buffer
, "DP3_SAT %s%s, arg1, arg2;\n", dstreg
, dstmask
);
5849 case WINED3DTOP_MULTIPLYADD
:
5850 shader_addline(buffer
, "MAD_SAT %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg1
, arg2
, arg0
);
5853 case WINED3DTOP_LERP
:
5854 /* The msdn is not quite right here */
5855 shader_addline(buffer
, "LRP %s%s, %s, %s, %s;\n", dstreg
, dstmask
, arg0
, arg1
, arg2
);
5858 case WINED3DTOP_BUMPENVMAP
:
5859 case WINED3DTOP_BUMPENVMAPLUMINANCE
:
5860 /* Those are handled in the first pass of the shader(generation pass 1 and 2) already */
5864 FIXME("Unhandled texture op %08x\n", op
);
5868 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.y;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
5869 } else if(mul
== 4) {
5870 shader_addline(buffer
, "MUL_SAT %s%s, %s, const.z;\n", mul_final_dest
? "result.color" : dstreg
, dstmask
, dstreg
);
5874 static GLuint
gen_arbfp_ffp_shader(const struct ffp_frag_settings
*settings
, IWineD3DStateBlockImpl
*stateblock
)
5876 const struct wined3d_gl_info
*gl_info
= &stateblock
->device
->adapter
->gl_info
;
5878 struct wined3d_shader_buffer buffer
;
5879 BOOL tex_read
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5880 BOOL bump_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5881 BOOL luminance_used
[MAX_TEXTURES
] = {FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
, FALSE
};
5882 const char *textype
;
5883 const char *instr
, *sat
;
5884 char colorcor_dst
[8];
5886 DWORD arg0
, arg1
, arg2
;
5887 BOOL tempreg_used
= FALSE
, tfactor_used
= FALSE
;
5889 const char *final_combiner_src
= "ret";
5892 /* Find out which textures are read */
5893 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
5894 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) break;
5895 arg0
= settings
->op
[stage
].carg0
& WINED3DTA_SELECTMASK
;
5896 arg1
= settings
->op
[stage
].carg1
& WINED3DTA_SELECTMASK
;
5897 arg2
= settings
->op
[stage
].carg2
& WINED3DTA_SELECTMASK
;
5898 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5899 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5900 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5902 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHA
) tex_read
[stage
] = TRUE
;
5903 if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDTEXTUREALPHAPM
) tex_read
[stage
] = TRUE
;
5904 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
) {
5905 bump_used
[stage
] = TRUE
;
5906 tex_read
[stage
] = TRUE
;
5908 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
5909 bump_used
[stage
] = TRUE
;
5910 tex_read
[stage
] = TRUE
;
5911 luminance_used
[stage
] = TRUE
;
5912 } else if(settings
->op
[stage
].cop
== WINED3DTOP_BLENDFACTORALPHA
) {
5913 tfactor_used
= TRUE
;
5916 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
5917 tfactor_used
= TRUE
;
5920 if(settings
->op
[stage
].dst
== tempreg
) tempreg_used
= TRUE
;
5921 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
5922 tempreg_used
= TRUE
;
5925 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) continue;
5926 arg0
= settings
->op
[stage
].aarg0
& WINED3DTA_SELECTMASK
;
5927 arg1
= settings
->op
[stage
].aarg1
& WINED3DTA_SELECTMASK
;
5928 arg2
= settings
->op
[stage
].aarg2
& WINED3DTA_SELECTMASK
;
5929 if(arg0
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5930 if(arg1
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5931 if(arg2
== WINED3DTA_TEXTURE
) tex_read
[stage
] = TRUE
;
5933 if(arg0
== WINED3DTA_TEMP
|| arg1
== WINED3DTA_TEMP
|| arg2
== WINED3DTA_TEMP
) {
5934 tempreg_used
= TRUE
;
5936 if(arg0
== WINED3DTA_TFACTOR
|| arg1
== WINED3DTA_TFACTOR
|| arg2
== WINED3DTA_TFACTOR
) {
5937 tfactor_used
= TRUE
;
5942 if (!shader_buffer_init(&buffer
))
5944 ERR("Failed to initialize shader buffer.\n");
5948 shader_addline(&buffer
, "!!ARBfp1.0\n");
5950 switch(settings
->fog
) {
5951 case FOG_OFF
: break;
5952 case FOG_LINEAR
: shader_addline(&buffer
, "OPTION ARB_fog_linear;\n"); break;
5953 case FOG_EXP
: shader_addline(&buffer
, "OPTION ARB_fog_exp;\n"); break;
5954 case FOG_EXP2
: shader_addline(&buffer
, "OPTION ARB_fog_exp2;\n"); break;
5955 default: FIXME("Unexpected fog setting %d\n", settings
->fog
);
5958 shader_addline(&buffer
, "PARAM const = {1, 2, 4, 0.5};\n");
5959 shader_addline(&buffer
, "TEMP TMP;\n");
5960 shader_addline(&buffer
, "TEMP ret;\n");
5961 if(tempreg_used
|| settings
->sRGB_write
) shader_addline(&buffer
, "TEMP tempreg;\n");
5962 shader_addline(&buffer
, "TEMP arg0;\n");
5963 shader_addline(&buffer
, "TEMP arg1;\n");
5964 shader_addline(&buffer
, "TEMP arg2;\n");
5965 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
5966 if(!tex_read
[stage
]) continue;
5967 shader_addline(&buffer
, "TEMP tex%u;\n", stage
);
5968 if(!bump_used
[stage
]) continue;
5969 shader_addline(&buffer
, "PARAM bumpmat%u = program.env[%u];\n", stage
, ARB_FFP_CONST_BUMPMAT(stage
));
5970 if(!luminance_used
[stage
]) continue;
5971 shader_addline(&buffer
, "PARAM luminance%u = program.env[%u];\n", stage
, ARB_FFP_CONST_LUMINANCE(stage
));
5974 shader_addline(&buffer
, "PARAM tfactor = program.env[%u];\n", ARB_FFP_CONST_TFACTOR
);
5976 shader_addline(&buffer
, "PARAM specular_enable = program.env[%u];\n", ARB_FFP_CONST_SPECULAR_ENABLE
);
5978 if(settings
->sRGB_write
) {
5979 shader_addline(&buffer
, "PARAM srgb_consts1 = {%f, %f, %f, %f};\n",
5980 srgb_mul_low
, srgb_cmp
, srgb_pow
, srgb_mul_high
);
5981 shader_addline(&buffer
, "PARAM srgb_consts2 = {%f, %f, %f, %f};\n",
5982 srgb_sub_high
, 0.0, 0.0, 0.0);
5985 if(ffp_clip_emul(stateblock
) && settings
->emul_clipplanes
) shader_addline(&buffer
, "KIL fragment.texcoord[7];\n");
5987 /* Generate texture sampling instructions) */
5988 for(stage
= 0; stage
< MAX_TEXTURES
&& settings
->op
[stage
].cop
!= WINED3DTOP_DISABLE
; stage
++) {
5989 if(!tex_read
[stage
]) continue;
5991 switch(settings
->op
[stage
].tex_type
) {
5992 case tex_1d
: textype
= "1D"; break;
5993 case tex_2d
: textype
= "2D"; break;
5994 case tex_3d
: textype
= "3D"; break;
5995 case tex_cube
: textype
= "CUBE"; break;
5996 case tex_rect
: textype
= "RECT"; break;
5997 default: textype
= "unexpected_textype"; break;
6000 if(settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAP
||
6001 settings
->op
[stage
].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
6007 if(settings
->op
[stage
].projected
== proj_none
) {
6009 } else if(settings
->op
[stage
].projected
== proj_count4
||
6010 settings
->op
[stage
].projected
== proj_count3
) {
6013 FIXME("Unexpected projection mode %d\n", settings
->op
[stage
].projected
);
6018 (settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAP
||
6019 settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
)) {
6020 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, x, z, 0, 0;\n", stage
- 1);
6021 shader_addline(&buffer
, "DP3 ret.x, arg1, tex%u;\n", stage
- 1);
6022 shader_addline(&buffer
, "SWZ arg1, bumpmat%u, y, w, 0, 0;\n", stage
- 1);
6023 shader_addline(&buffer
, "DP3 ret.y, arg1, tex%u;\n", stage
- 1);
6025 /* with projective textures, texbem only divides the static texture coord, not the displacement,
6026 * so multiply the displacement with the dividing parameter before passing it to TXP
6028 if (settings
->op
[stage
].projected
!= proj_none
) {
6029 if(settings
->op
[stage
].projected
== proj_count4
) {
6030 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].w;\n", stage
);
6031 shader_addline(&buffer
, "MUL ret.xyz, ret, fragment.texcoord[%u].w, fragment.texcoord[%u];\n", stage
, stage
);
6033 shader_addline(&buffer
, "MOV ret.w, fragment.texcoord[%u].z;\n", stage
);
6034 shader_addline(&buffer
, "MAD ret.xyz, ret, fragment.texcoord[%u].z, fragment.texcoord[%u];\n", stage
, stage
);
6037 shader_addline(&buffer
, "ADD ret, ret, fragment.texcoord[%u];\n", stage
);
6040 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
6041 instr
, sat
, stage
, stage
, textype
);
6042 if(settings
->op
[stage
- 1].cop
== WINED3DTOP_BUMPENVMAPLUMINANCE
) {
6043 shader_addline(&buffer
, "MAD_SAT ret.x, tex%u.z, luminance%u.x, luminance%u.y;\n",
6044 stage
- 1, stage
- 1, stage
- 1);
6045 shader_addline(&buffer
, "MUL tex%u, tex%u, ret.x;\n", stage
, stage
);
6047 } else if(settings
->op
[stage
].projected
== proj_count3
) {
6048 shader_addline(&buffer
, "MOV ret, fragment.texcoord[%u];\n", stage
);
6049 shader_addline(&buffer
, "MOV ret.w, ret.z;\n");
6050 shader_addline(&buffer
, "%s%s tex%u, ret, texture[%u], %s;\n",
6051 instr
, sat
, stage
, stage
, textype
);
6053 shader_addline(&buffer
, "%s%s tex%u, fragment.texcoord[%u], texture[%u], %s;\n",
6054 instr
, sat
, stage
, stage
, stage
, textype
);
6057 sprintf(colorcor_dst
, "tex%u", stage
);
6058 gen_color_correction(&buffer
, colorcor_dst
, WINED3DSP_WRITEMASK_ALL
, "const.x", "const.y",
6059 settings
->op
[stage
].color_fixup
);
6062 /* Generate the main shader */
6063 for(stage
= 0; stage
< MAX_TEXTURES
; stage
++) {
6064 if(settings
->op
[stage
].cop
== WINED3DTOP_DISABLE
) {
6066 final_combiner_src
= "fragment.color.primary";
6071 if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
6072 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
6073 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
;
6074 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG1
&&
6075 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
6076 op_equal
= settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg2
;
6077 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
6078 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG1
) {
6079 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg1
;
6080 } else if(settings
->op
[stage
].cop
== WINED3DTOP_SELECTARG2
&&
6081 settings
->op
[stage
].aop
== WINED3DTOP_SELECTARG2
) {
6082 op_equal
= settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6084 op_equal
= settings
->op
[stage
].aop
== settings
->op
[stage
].cop
&&
6085 settings
->op
[stage
].carg0
== settings
->op
[stage
].aarg0
&&
6086 settings
->op
[stage
].carg1
== settings
->op
[stage
].aarg1
&&
6087 settings
->op
[stage
].carg2
== settings
->op
[stage
].aarg2
;
6090 if(settings
->op
[stage
].aop
== WINED3DTOP_DISABLE
) {
6091 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6092 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6093 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6095 shader_addline(&buffer
, "MOV ret.w, fragment.color.primary.w;\n");
6097 } else if(op_equal
) {
6098 gen_ffp_instr(&buffer
, stage
, TRUE
, TRUE
, settings
->op
[stage
].dst
,
6099 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6100 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6102 gen_ffp_instr(&buffer
, stage
, TRUE
, FALSE
, settings
->op
[stage
].dst
,
6103 settings
->op
[stage
].cop
, settings
->op
[stage
].carg0
,
6104 settings
->op
[stage
].carg1
, settings
->op
[stage
].carg2
);
6105 gen_ffp_instr(&buffer
, stage
, FALSE
, TRUE
, settings
->op
[stage
].dst
,
6106 settings
->op
[stage
].aop
, settings
->op
[stage
].aarg0
,
6107 settings
->op
[stage
].aarg1
, settings
->op
[stage
].aarg2
);
6111 if(settings
->sRGB_write
) {
6112 shader_addline(&buffer
, "MAD ret, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6113 arbfp_add_sRGB_correction(&buffer
, "ret", "arg0", "arg1", "arg2", "tempreg", FALSE
);
6114 shader_addline(&buffer
, "MOV result.color, ret;\n");
6116 shader_addline(&buffer
, "MAD result.color, fragment.color.secondary, specular_enable, %s;\n", final_combiner_src
);
6120 shader_addline(&buffer
, "END\n");
6122 /* Generate the shader */
6123 GL_EXTCALL(glGenProgramsARB(1, &ret
));
6124 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, ret
));
6125 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6126 strlen(buffer
.buffer
), buffer
.buffer
));
6127 checkGLcall("glProgramStringARB()");
6129 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6132 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6133 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6134 shader_arb_dump_program_source(buffer
.buffer
);
6140 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
6141 checkGLcall("glGetProgramivARB()");
6142 if (!native
) WARN("Program exceeds native resource limits.\n");
6145 shader_buffer_free(&buffer
);
6149 static void fragment_prog_arbfp(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
6151 const struct wined3d_gl_info
*gl_info
= context
->gl_info
;
6152 IWineD3DDeviceImpl
*device
= stateblock
->device
;
6153 struct shader_arb_priv
*priv
= device
->fragment_priv
;
6154 BOOL use_pshader
= use_ps(stateblock
);
6155 BOOL use_vshader
= use_vs(stateblock
);
6156 struct ffp_frag_settings settings
;
6157 const struct arbfp_ffp_desc
*desc
;
6160 TRACE("state %#x, stateblock %p, context %p\n", state
, stateblock
, context
);
6162 if(isStateDirty(context
, STATE_RENDER(WINED3DRS_FOGENABLE
))) {
6163 if(!use_pshader
&& device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
6164 /* Reload fixed function constants since they collide with the pixel shader constants */
6165 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
6166 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
6168 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
6169 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
6170 } else if(use_pshader
&& !isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
6171 device
->shader_backend
->shader_select(context
, use_pshader
, use_vshader
);
6177 /* Find or create a shader implementing the fixed function pipeline settings, then activate it */
6178 gen_ffp_frag_op(stateblock
, &settings
, FALSE
);
6179 desc
= (const struct arbfp_ffp_desc
*)find_ffp_frag_shader(&priv
->fragment_shaders
, &settings
);
6181 struct arbfp_ffp_desc
*new_desc
= HeapAlloc(GetProcessHeap(), 0, sizeof(*new_desc
));
6184 ERR("Out of memory\n");
6187 new_desc
->num_textures_used
= 0;
6188 for (i
= 0; i
< gl_info
->limits
.texture_stages
; ++i
)
6190 if(settings
.op
[i
].cop
== WINED3DTOP_DISABLE
) break;
6191 new_desc
->num_textures_used
= i
;
6194 memcpy(&new_desc
->parent
.settings
, &settings
, sizeof(settings
));
6195 new_desc
->shader
= gen_arbfp_ffp_shader(&settings
, stateblock
);
6196 add_ffp_frag_shader(&priv
->fragment_shaders
, &new_desc
->parent
);
6197 TRACE("Allocated fixed function replacement shader descriptor %p\n", new_desc
);
6201 /* Now activate the replacement program. GL_FRAGMENT_PROGRAM_ARB is already active(however, note the
6202 * comment above the shader_select call below). If e.g. GLSL is active, the shader_select call will
6205 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, desc
->shader
));
6206 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, desc->shader)");
6207 priv
->current_fprogram_id
= desc
->shader
;
6209 if(device
->shader_backend
== &arb_program_shader_backend
&& context
->last_was_pshader
) {
6210 /* Reload fixed function constants since they collide with the pixel shader constants */
6211 for(i
= 0; i
< MAX_TEXTURES
; i
++) {
6212 set_bumpmat_arbfp(STATE_TEXTURESTAGE(i
, WINED3DTSS_BUMPENVMAT00
), stateblock
, context
);
6214 state_texfactor_arbfp(STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), stateblock
, context
);
6215 state_arb_specularenable(STATE_RENDER(WINED3DRS_SPECULARENABLE
), stateblock
, context
);
6217 context
->last_was_pshader
= FALSE
;
6219 context
->last_was_pshader
= TRUE
;
6222 /* Finally, select the shader. If a pixel shader is used, it will be set and enabled by the shader backend.
6223 * If this shader backend is arbfp(most likely), then it will simply overwrite the last fixed function replace-
6224 * ment shader. If the shader backend is not ARB, it currently is important that the opengl implementation
6225 * type overwrites GL_ARB_fragment_program. This is currently the case with GLSL. If we really want to use
6226 * atifs or nvrc pixel shaders with arb fragment programs we'd have to disable GL_FRAGMENT_PROGRAM_ARB here
6228 * Don't call shader_select if the vertex shader is dirty, because it will be called later on by the vertex
6231 if(!isStateDirty(context
, device
->StateTable
[STATE_VSHADER
].representative
)) {
6232 device
->shader_backend
->shader_select(context
, use_pshader
, use_vshader
);
6234 if (!isStateDirty(context
, STATE_VERTEXSHADERCONSTANT
) && (use_vshader
|| use_pshader
))
6235 stateblock_apply_state(STATE_VERTEXSHADERCONSTANT
, stateblock
, context
);
6237 if (use_pshader
) stateblock_apply_state(STATE_PIXELSHADERCONSTANT
, stateblock
, context
);
6240 /* We can't link the fog states to the fragment state directly since the vertex pipeline links them
6241 * to FOGENABLE. A different linking in different pipeline parts can't be expressed in the combined
6242 * state table, so we need to handle that with a forwarding function. The other invisible side effect
6243 * is that changing the fog start and fog end(which links to FOGENABLE in vertex) results in the
6244 * fragment_prog_arbfp function being called because FOGENABLE is dirty, which calls this function here
6246 static void state_arbfp_fog(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
6248 enum fogsource new_source
;
6250 TRACE("state %#x, stateblock %p, context %p\n", state
, stateblock
, context
);
6252 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
6253 fragment_prog_arbfp(state
, stateblock
, context
);
6256 if(!stateblock
->renderState
[WINED3DRS_FOGENABLE
]) return;
6258 if(stateblock
->renderState
[WINED3DRS_FOGTABLEMODE
] == WINED3DFOG_NONE
) {
6259 if(use_vs(stateblock
)) {
6260 new_source
= FOGSOURCE_VS
;
6262 if(stateblock
->renderState
[WINED3DRS_FOGVERTEXMODE
] == WINED3DFOG_NONE
|| context
->last_was_rhw
) {
6263 new_source
= FOGSOURCE_COORD
;
6265 new_source
= FOGSOURCE_FFP
;
6269 new_source
= FOGSOURCE_FFP
;
6271 if(new_source
!= context
->fog_source
) {
6272 context
->fog_source
= new_source
;
6273 state_fogstartend(STATE_RENDER(WINED3DRS_FOGSTART
), stateblock
, context
);
6277 static void textransform(DWORD state
, IWineD3DStateBlockImpl
*stateblock
, struct wined3d_context
*context
)
6279 if(!isStateDirty(context
, STATE_PIXELSHADER
)) {
6280 fragment_prog_arbfp(state
, stateblock
, context
);
6284 static const struct StateEntryTemplate arbfp_fragmentstate_template
[] = {
6285 {STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), { STATE_RENDER(WINED3DRS_TEXTUREFACTOR
), state_texfactor_arbfp
}, WINED3D_GL_EXT_NONE
},
6286 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6287 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6288 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6289 {STATE_TEXTURESTAGE(0, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6290 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6291 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6292 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6293 {STATE_TEXTURESTAGE(0, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6294 {STATE_TEXTURESTAGE(0, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6295 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6296 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6297 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6298 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6299 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6300 {STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(0, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6301 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6302 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6303 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6304 {STATE_TEXTURESTAGE(1, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6305 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6306 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6307 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6308 {STATE_TEXTURESTAGE(1, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6309 {STATE_TEXTURESTAGE(1, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6310 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6311 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6312 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6313 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6314 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6315 {STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(1, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6316 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6317 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6318 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6319 {STATE_TEXTURESTAGE(2, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6320 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6321 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6322 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6323 {STATE_TEXTURESTAGE(2, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6324 {STATE_TEXTURESTAGE(2, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6325 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6326 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6327 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6328 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6329 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6330 {STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(2, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6331 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6332 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6333 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6334 {STATE_TEXTURESTAGE(3, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6335 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6336 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6337 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6338 {STATE_TEXTURESTAGE(3, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6339 {STATE_TEXTURESTAGE(3, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6340 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6341 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6342 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6343 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6344 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6345 {STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(3, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6346 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6347 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6348 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6349 {STATE_TEXTURESTAGE(4, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6350 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6351 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6352 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6353 {STATE_TEXTURESTAGE(4, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6354 {STATE_TEXTURESTAGE(4, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6355 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6356 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6357 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6358 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6359 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6360 {STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(4, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6361 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6362 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6363 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6364 {STATE_TEXTURESTAGE(5, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6365 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6366 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6367 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6368 {STATE_TEXTURESTAGE(5, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6369 {STATE_TEXTURESTAGE(5, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6370 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6371 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6372 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6373 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6374 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6375 {STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(5, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6376 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6377 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6378 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6379 {STATE_TEXTURESTAGE(6, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6380 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6381 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6382 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6383 {STATE_TEXTURESTAGE(6, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6384 {STATE_TEXTURESTAGE(6, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6385 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6386 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6387 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6388 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6389 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6390 {STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(6, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6391 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLOROP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6392 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6393 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6394 {STATE_TEXTURESTAGE(7, WINED3DTSS_COLORARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6395 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAOP
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6396 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG1
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6397 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG2
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6398 {STATE_TEXTURESTAGE(7, WINED3DTSS_ALPHAARG0
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6399 {STATE_TEXTURESTAGE(7, WINED3DTSS_RESULTARG
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6400 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), set_bumpmat_arbfp
}, WINED3D_GL_EXT_NONE
},
6401 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT01
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6402 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT10
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6403 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT11
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVMAT00
), NULL
}, WINED3D_GL_EXT_NONE
},
6404 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), tex_bumpenvlum_arbfp
}, WINED3D_GL_EXT_NONE
},
6405 {STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLOFFSET
), { STATE_TEXTURESTAGE(7, WINED3DTSS_BUMPENVLSCALE
), NULL
}, WINED3D_GL_EXT_NONE
},
6406 {STATE_SAMPLER(0), { STATE_SAMPLER(0), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6407 {STATE_SAMPLER(1), { STATE_SAMPLER(1), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6408 {STATE_SAMPLER(2), { STATE_SAMPLER(2), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6409 {STATE_SAMPLER(3), { STATE_SAMPLER(3), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6410 {STATE_SAMPLER(4), { STATE_SAMPLER(4), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6411 {STATE_SAMPLER(5), { STATE_SAMPLER(5), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6412 {STATE_SAMPLER(6), { STATE_SAMPLER(6), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6413 {STATE_SAMPLER(7), { STATE_SAMPLER(7), sampler_texdim
}, WINED3D_GL_EXT_NONE
},
6414 {STATE_PIXELSHADER
, { STATE_PIXELSHADER
, fragment_prog_arbfp
}, WINED3D_GL_EXT_NONE
},
6415 {STATE_RENDER(WINED3DRS_FOGENABLE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), state_arbfp_fog
}, WINED3D_GL_EXT_NONE
},
6416 {STATE_RENDER(WINED3DRS_FOGTABLEMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6417 {STATE_RENDER(WINED3DRS_FOGVERTEXMODE
), { STATE_RENDER(WINED3DRS_FOGENABLE
), NULL
}, WINED3D_GL_EXT_NONE
},
6418 {STATE_RENDER(WINED3DRS_FOGSTART
), { STATE_RENDER(WINED3DRS_FOGSTART
), state_fogstartend
}, WINED3D_GL_EXT_NONE
},
6419 {STATE_RENDER(WINED3DRS_FOGEND
), { STATE_RENDER(WINED3DRS_FOGSTART
), NULL
}, WINED3D_GL_EXT_NONE
},
6420 {STATE_RENDER(WINED3DRS_SRGBWRITEENABLE
), { STATE_PIXELSHADER
, NULL
}, WINED3D_GL_EXT_NONE
},
6421 {STATE_RENDER(WINED3DRS_FOGCOLOR
), { STATE_RENDER(WINED3DRS_FOGCOLOR
), state_fogcolor
}, WINED3D_GL_EXT_NONE
},
6422 {STATE_RENDER(WINED3DRS_FOGDENSITY
), { STATE_RENDER(WINED3DRS_FOGDENSITY
), state_fogdensity
}, WINED3D_GL_EXT_NONE
},
6423 {STATE_TEXTURESTAGE(0,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(0, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6424 {STATE_TEXTURESTAGE(1,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(1, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6425 {STATE_TEXTURESTAGE(2,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(2, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6426 {STATE_TEXTURESTAGE(3,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(3, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6427 {STATE_TEXTURESTAGE(4,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(4, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6428 {STATE_TEXTURESTAGE(5,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(5, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6429 {STATE_TEXTURESTAGE(6,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(6, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6430 {STATE_TEXTURESTAGE(7,WINED3DTSS_TEXTURETRANSFORMFLAGS
),{STATE_TEXTURESTAGE(7, WINED3DTSS_TEXTURETRANSFORMFLAGS
), textransform
}, WINED3D_GL_EXT_NONE
},
6431 {STATE_RENDER(WINED3DRS_SPECULARENABLE
), { STATE_RENDER(WINED3DRS_SPECULARENABLE
), state_arb_specularenable
}, WINED3D_GL_EXT_NONE
},
6432 {0 /* Terminate */, { 0, 0 }, WINED3D_GL_EXT_NONE
},
6435 const struct fragment_pipeline arbfp_fragment_pipeline
= {
6440 shader_arb_color_fixup_supported
,
6441 arbfp_fragmentstate_template
,
6442 TRUE
/* We can disable projected textures */
6445 struct arbfp_blit_priv
{
6446 GLenum yuy2_rect_shader
, yuy2_2d_shader
;
6447 GLenum uyvy_rect_shader
, uyvy_2d_shader
;
6448 GLenum yv12_rect_shader
, yv12_2d_shader
;
6449 GLenum p8_rect_shader
, p8_2d_shader
;
6450 GLuint palette_texture
;
6453 static HRESULT
arbfp_blit_alloc(IWineD3DDevice
*iface
) {
6454 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
6455 device
->blit_priv
= HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY
, sizeof(struct arbfp_blit_priv
));
6456 if(!device
->blit_priv
) {
6457 ERR("Out of memory\n");
6458 return E_OUTOFMEMORY
;
6463 /* Context activation is done by the caller. */
6464 static void arbfp_blit_free(IWineD3DDevice
*iface
) {
6465 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
6466 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6467 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6470 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_rect_shader
));
6471 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yuy2_2d_shader
));
6472 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_rect_shader
));
6473 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->uyvy_2d_shader
));
6474 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_rect_shader
));
6475 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->yv12_2d_shader
));
6476 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->p8_rect_shader
));
6477 GL_EXTCALL(glDeleteProgramsARB(1, &priv
->p8_2d_shader
));
6478 checkGLcall("Delete yuv and p8 programs");
6480 if(priv
->palette_texture
) glDeleteTextures(1, &priv
->palette_texture
);
6483 HeapFree(GetProcessHeap(), 0, device
->blit_priv
);
6484 device
->blit_priv
= NULL
;
6487 static BOOL
gen_planar_yuv_read(struct wined3d_shader_buffer
*buffer
, enum complex_fixup fixup
,
6488 GLenum textype
, char *luminance
)
6491 const char *tex
, *texinstr
;
6493 if (fixup
== COMPLEX_FIXUP_UYVY
) {
6501 case GL_TEXTURE_2D
: tex
= "2D"; texinstr
= "TXP"; break;
6502 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; texinstr
= "TEX"; break;
6504 /* This is more tricky than just replacing the texture type - we have to navigate
6505 * properly in the texture to find the correct chroma values
6507 FIXME("Implement yuv correction for non-2d, non-rect textures\n");
6511 /* First we have to read the chroma values. This means we need at least two pixels(no filtering),
6512 * or 4 pixels(with filtering). To get the unmodified chromas, we have to rid ourselves of the
6513 * filtering when we sample the texture.
6515 * These are the rules for reading the chroma:
6521 * So we have to get the sampling x position in non-normalized coordinates in integers
6523 if(textype
!= GL_TEXTURE_RECTANGLE_ARB
) {
6524 shader_addline(buffer
, "MUL texcrd.xy, fragment.texcoord[0], size.x;\n");
6525 shader_addline(buffer
, "MOV texcrd.w, size.x;\n");
6527 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6529 /* We must not allow filtering between pixel x and x+1, this would mix U and V
6530 * Vertical filtering is ok. However, bear in mind that the pixel center is at
6533 shader_addline(buffer
, "FLR texcrd.x, texcrd.x;\n");
6534 shader_addline(buffer
, "ADD texcrd.x, texcrd.x, coef.y;\n");
6536 /* Divide the x coordinate by 0.5 and get the fraction. This gives 0.25 and 0.75 for the
6537 * even and odd pixels respectively
6539 shader_addline(buffer
, "MUL texcrd2, texcrd, coef.y;\n");
6540 shader_addline(buffer
, "FRC texcrd2, texcrd2;\n");
6542 /* Sample Pixel 1 */
6543 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6545 /* Put the value into either of the chroma values */
6546 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6547 shader_addline(buffer
, "MUL chroma.x, luminance.%c, temp.x;\n", chroma
);
6548 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6549 shader_addline(buffer
, "MUL chroma.y, luminance.%c, temp.x;\n", chroma
);
6551 /* Sample pixel 2. If we read an even pixel(SLT above returned 1), sample
6552 * the pixel right to the current one. Otherwise, sample the left pixel.
6553 * Bias and scale the SLT result to -1;1 and add it to the texcrd.x.
6555 shader_addline(buffer
, "MAD temp.x, temp.x, coef.z, -coef.x;\n");
6556 shader_addline(buffer
, "ADD texcrd.x, texcrd, temp.x;\n");
6557 shader_addline(buffer
, "%s luminance, texcrd, texture[0], %s;\n", texinstr
, tex
);
6559 /* Put the value into the other chroma */
6560 shader_addline(buffer
, "SGE temp.x, texcrd2.x, coef.y;\n");
6561 shader_addline(buffer
, "MAD chroma.y, luminance.%c, temp.x, chroma.y;\n", chroma
);
6562 shader_addline(buffer
, "SLT temp.x, texcrd2.x, coef.y;\n");
6563 shader_addline(buffer
, "MAD chroma.x, luminance.%c, temp.x, chroma.x;\n", chroma
);
6565 /* TODO: If filtering is enabled, sample a 2nd pair of pixels left or right of
6566 * the current one and lerp the two U and V values
6569 /* This gives the correctly filtered luminance value */
6570 shader_addline(buffer
, "TEX luminance, fragment.texcoord[0], texture[0], %s;\n", tex
);
6575 static BOOL
gen_yv12_read(struct wined3d_shader_buffer
*buffer
, GLenum textype
, char *luminance
)
6580 case GL_TEXTURE_2D
: tex
= "2D"; break;
6581 case GL_TEXTURE_RECTANGLE_ARB
: tex
= "RECT"; break;
6583 FIXME("Implement yv12 correction for non-2d, non-rect textures\n");
6587 /* YV12 surfaces contain a WxH sized luminance plane, followed by a (W/2)x(H/2)
6588 * V and a (W/2)x(H/2) U plane, each with 8 bit per pixel. So the effective
6589 * bitdepth is 12 bits per pixel. Since the U and V planes have only half the
6590 * pitch of the luminance plane, the packing into the gl texture is a bit
6591 * unfortunate. If the whole texture is interpreted as luminance data it looks
6592 * approximately like this:
6594 * +----------------------------------+----
6606 * +----------------+-----------------+----
6608 * | U even rows | U odd rows |
6610 * +----------------+------------------ -
6612 * | V even rows | V odd rows |
6614 * +----------------+-----------------+----
6618 * So it appears as if there are 4 chroma images, but in fact the odd rows
6619 * in the chroma images are in the same row as the even ones. So its is
6620 * kinda tricky to read
6622 * When reading from rectangle textures, keep in mind that the input y coordinates
6623 * go from 0 to d3d_height, whereas the opengl texture height is 1.5 * d3d_height
6625 shader_addline(buffer
, "PARAM yv12_coef = {%f, %f, %f, %f};\n",
6626 2.0f
/ 3.0f
, 1.0f
/ 6.0f
, (2.0f
/ 3.0f
) + (1.0f
/ 6.0f
), 1.0f
/ 3.0f
);
6628 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6629 /* the chroma planes have only half the width */
6630 shader_addline(buffer
, "MUL texcrd.x, texcrd.x, coef.y;\n");
6632 /* The first value is between 2/3 and 5/6th of the texture's height, so scale+bias
6633 * the coordinate. Also read the right side of the image when reading odd lines
6635 * Don't forget to clamp the y values in into the range, otherwise we'll get filtering
6638 if(textype
== GL_TEXTURE_2D
) {
6640 shader_addline(buffer
, "RCP chroma.w, size.y;\n");
6642 shader_addline(buffer
, "MUL texcrd2.y, texcrd.y, size.y;\n");
6644 shader_addline(buffer
, "FLR texcrd2.y, texcrd2.y;\n");
6645 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, yv12_coef.y, yv12_coef.x;\n");
6647 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6648 shader_addline(buffer
, "ADD texcrd2.x, texcrd2.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6649 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6650 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6651 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6653 /* clamp, keep the half pixel origin in mind */
6654 shader_addline(buffer
, "MAD temp.y, coef.y, chroma.w, yv12_coef.x;\n");
6655 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6656 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.z;\n");
6657 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6659 /* Read from [size - size+size/4] */
6660 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6661 shader_addline(buffer
, "MAD texcrd.y, texcrd.y, coef.w, size.y;\n");
6663 /* Read odd lines from the right side(add size * 0.5 to the x coordinate */
6664 shader_addline(buffer
, "ADD texcrd2.x, texcrd.y, yv12_coef.y;\n"); /* To avoid 0.5 == 0.5 comparisons */
6665 shader_addline(buffer
, "FRC texcrd2.x, texcrd2.x;\n");
6666 shader_addline(buffer
, "SGE texcrd2.x, texcrd2.x, coef.y;\n");
6667 shader_addline(buffer
, "MUL texcrd2.x, texcrd2.x, size.x;\n");
6668 shader_addline(buffer
, "MAD texcrd.x, texcrd2.x, coef.y, texcrd.x;\n");
6670 /* Make sure to read exactly from the pixel center */
6671 shader_addline(buffer
, "FLR texcrd.y, texcrd.y;\n");
6672 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, coef.y;\n");
6675 shader_addline(buffer
, "MAD temp.y, size.y, coef.w, size.y;\n");
6676 shader_addline(buffer
, "ADD temp.y, temp.y, -coef.y;\n");
6677 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6678 shader_addline(buffer
, "ADD temp.y, size.y, -coef.y;\n");
6679 shader_addline(buffer
, "MAX texcrd.y, temp.y, texcrd.y;\n");
6681 /* Read the texture, put the result into the output register */
6682 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6683 shader_addline(buffer
, "MOV chroma.x, temp.w;\n");
6685 /* The other chroma value is 1/6th of the texture lower, from 5/6th to 6/6th
6686 * No need to clamp because we're just reusing the already clamped value from above
6688 if(textype
== GL_TEXTURE_2D
) {
6689 shader_addline(buffer
, "ADD texcrd.y, texcrd.y, yv12_coef.y;\n");
6691 shader_addline(buffer
, "MAD texcrd.y, size.y, coef.w, texcrd.y;\n");
6693 shader_addline(buffer
, "TEX temp, texcrd, texture[0], %s;\n", tex
);
6694 shader_addline(buffer
, "MOV chroma.y, temp.w;\n");
6696 /* Sample the luminance value. It is in the top 2/3rd of the texture, so scale the y coordinate.
6697 * Clamp the y coordinate to prevent the chroma values from bleeding into the sampled luminance
6698 * values due to filtering
6700 shader_addline(buffer
, "MOV texcrd, fragment.texcoord[0];\n");
6701 if(textype
== GL_TEXTURE_2D
) {
6702 /* Multiply the y coordinate by 2/3 and clamp it */
6703 shader_addline(buffer
, "MUL texcrd.y, texcrd.y, yv12_coef.x;\n");
6704 shader_addline(buffer
, "MAD temp.y, -coef.y, chroma.w, yv12_coef.x;\n");
6705 shader_addline(buffer
, "MIN texcrd.y, temp.y, texcrd.y;\n");
6706 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
6708 /* Reading from texture_rectangles is pretty straightforward, just use the unmodified
6709 * texture coordinate. It is still a good idea to clamp it though, since the opengl texture
6712 shader_addline(buffer
, "ADD temp.x, size.y, -coef.y;\n");
6713 shader_addline(buffer
, "MIN texcrd.y, texcrd.y, size.x;\n");
6714 shader_addline(buffer
, "TEX luminance, texcrd, texture[0], %s;\n", tex
);
6721 static GLuint
gen_p8_shader(IWineD3DDeviceImpl
*device
, GLenum textype
)
6723 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6725 struct wined3d_shader_buffer buffer
;
6726 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6730 if (!shader_buffer_init(&buffer
))
6732 ERR("Failed to initialize shader buffer.\n");
6737 GL_EXTCALL(glGenProgramsARB(1, &shader
));
6738 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
6741 shader_buffer_free(&buffer
);
6745 shader_addline(&buffer
, "!!ARBfp1.0\n");
6746 shader_addline(&buffer
, "TEMP index;\n");
6748 /* { 255/256, 0.5/255*255/256, 0, 0 } */
6749 shader_addline(&buffer
, "PARAM constants = { 0.996, 0.00195, 0, 0 };\n");
6751 /* The alpha-component contains the palette index */
6752 if(textype
== GL_TEXTURE_RECTANGLE_ARB
)
6753 shader_addline(&buffer
, "TXP index, fragment.texcoord[0], texture[0], RECT;\n");
6755 shader_addline(&buffer
, "TEX index, fragment.texcoord[0], texture[0], 2D;\n");
6757 /* Scale the index by 255/256 and add a bias of '0.5' in order to sample in the middle */
6758 shader_addline(&buffer
, "MAD index.a, index.a, constants.x, constants.y;\n");
6760 /* Use the alpha-component as an index in the palette to get the final color */
6761 shader_addline(&buffer
, "TEX result.color, index.a, texture[1], 1D;\n");
6762 shader_addline(&buffer
, "END\n");
6765 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6766 strlen(buffer
.buffer
), buffer
.buffer
));
6767 checkGLcall("glProgramStringARB()");
6769 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6772 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6773 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6774 shader_arb_dump_program_source(buffer
.buffer
);
6777 if (textype
== GL_TEXTURE_RECTANGLE_ARB
)
6778 priv
->p8_rect_shader
= shader
;
6780 priv
->p8_2d_shader
= shader
;
6782 shader_buffer_free(&buffer
);
6788 /* Context activation is done by the caller. */
6789 static void upload_palette(IWineD3DSurfaceImpl
*surface
)
6792 IWineD3DDeviceImpl
*device
= surface
->resource
.device
;
6793 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6794 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6795 BOOL colorkey
= (surface
->CKeyFlags
& WINEDDSD_CKSRCBLT
) ? TRUE
: FALSE
;
6797 d3dfmt_p8_init_palette(surface
, table
, colorkey
);
6800 if (!priv
->palette_texture
)
6801 glGenTextures(1, &priv
->palette_texture
);
6803 GL_EXTCALL(glActiveTextureARB(GL_TEXTURE1
));
6804 glBindTexture(GL_TEXTURE_1D
, priv
->palette_texture
);
6806 glTexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
6808 glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MIN_FILTER
, GL_NEAREST
);
6809 /* Make sure we have discrete color levels. */
6810 glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_MAG_FILTER
, GL_NEAREST
);
6811 glTexParameteri(GL_TEXTURE_1D
, GL_TEXTURE_WRAP_S
, GL_CLAMP_TO_EDGE
);
6812 /* Upload the palette */
6813 /* TODO: avoid unneeed uploads in the future by adding some SFLAG_PALETTE_DIRTY mechanism */
6814 glTexImage1D(GL_TEXTURE_1D
, 0, GL_RGBA
, 256, 0, GL_RGBA
, GL_UNSIGNED_BYTE
, table
);
6816 /* Switch back to unit 0 in which the 2D texture will be stored. */
6817 GL_EXTCALL(glActiveTextureARB(GL_TEXTURE0
));
6821 /* Context activation is done by the caller. */
6822 static GLuint
gen_yuv_shader(IWineD3DDeviceImpl
*device
, enum complex_fixup yuv_fixup
, GLenum textype
)
6824 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6826 struct wined3d_shader_buffer buffer
;
6827 char luminance_component
;
6828 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6832 if (!shader_buffer_init(&buffer
))
6834 ERR("Failed to initialize shader buffer.\n");
6839 GL_EXTCALL(glGenProgramsARB(1, &shader
));
6840 checkGLcall("GL_EXTCALL(glGenProgramsARB(1, &shader))");
6841 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
6842 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
6845 shader_buffer_free(&buffer
);
6849 /* The YUY2 and UYVY formats contain two pixels packed into a 32 bit macropixel,
6850 * giving effectively 16 bit per pixel. The color consists of a luminance(Y) and
6851 * two chroma(U and V) values. Each macropixel has two luminance values, one for
6852 * each single pixel it contains, and one U and one V value shared between both
6855 * The data is loaded into an A8L8 texture. With YUY2, the luminance component
6856 * contains the luminance and alpha the chroma. With UYVY it is vice versa. Thus
6857 * take the format into account when generating the read swizzles
6859 * Reading the Y value is straightforward - just sample the texture. The hardware
6860 * takes care of filtering in the horizontal and vertical direction.
6862 * Reading the U and V values is harder. We have to avoid filtering horizontally,
6863 * because that would mix the U and V values of one pixel or two adjacent pixels.
6864 * Thus floor the texture coordinate and add 0.5 to get an unfiltered read,
6865 * regardless of the filtering setting. Vertical filtering works automatically
6866 * though - the U and V values of two rows are mixed nicely.
6868 * Appart of avoiding filtering issues, the code has to know which value it just
6869 * read, and where it can find the other one. To determine this, it checks if
6870 * it sampled an even or odd pixel, and shifts the 2nd read accordingly.
6872 * Handling horizontal filtering of U and V values requires reading a 2nd pair
6873 * of pixels, extracting U and V and mixing them. This is not implemented yet.
6875 * An alternative implementation idea is to load the texture as A8R8G8B8 texture,
6876 * with width / 2. This way one read gives all 3 values, finding U and V is easy
6877 * in an unfiltered situation. Finding the luminance on the other hand requires
6878 * finding out if it is an odd or even pixel. The real drawback of this approach
6879 * is filtering. This would have to be emulated completely in the shader, reading
6880 * up two 2 packed pixels in up to 2 rows and interpolating both horizontally and
6881 * vertically. Beyond that it would require adjustments to the texture handling
6882 * code to deal with the width scaling
6884 shader_addline(&buffer
, "!!ARBfp1.0\n");
6885 shader_addline(&buffer
, "TEMP luminance;\n");
6886 shader_addline(&buffer
, "TEMP temp;\n");
6887 shader_addline(&buffer
, "TEMP chroma;\n");
6888 shader_addline(&buffer
, "TEMP texcrd;\n");
6889 shader_addline(&buffer
, "TEMP texcrd2;\n");
6890 shader_addline(&buffer
, "PARAM coef = {1.0, 0.5, 2.0, 0.25};\n");
6891 shader_addline(&buffer
, "PARAM yuv_coef = {1.403, 0.344, 0.714, 1.770};\n");
6892 shader_addline(&buffer
, "PARAM size = program.local[0];\n");
6896 case COMPLEX_FIXUP_UYVY
:
6897 case COMPLEX_FIXUP_YUY2
:
6898 if (!gen_planar_yuv_read(&buffer
, yuv_fixup
, textype
, &luminance_component
))
6900 shader_buffer_free(&buffer
);
6905 case COMPLEX_FIXUP_YV12
:
6906 if (!gen_yv12_read(&buffer
, textype
, &luminance_component
))
6908 shader_buffer_free(&buffer
);
6914 FIXME("Unsupported YUV fixup %#x\n", yuv_fixup
);
6915 shader_buffer_free(&buffer
);
6919 /* Calculate the final result. Formula is taken from
6920 * http://www.fourcc.org/fccyvrgb.php. Note that the chroma
6921 * ranges from -0.5 to 0.5
6923 shader_addline(&buffer
, "SUB chroma.xy, chroma, coef.y;\n");
6925 shader_addline(&buffer
, "MAD result.color.x, chroma.x, yuv_coef.x, luminance.%c;\n", luminance_component
);
6926 shader_addline(&buffer
, "MAD temp.x, -chroma.y, yuv_coef.y, luminance.%c;\n", luminance_component
);
6927 shader_addline(&buffer
, "MAD result.color.y, -chroma.x, yuv_coef.z, temp.x;\n");
6928 shader_addline(&buffer
, "MAD result.color.z, chroma.y, yuv_coef.w, luminance.%c;\n", luminance_component
);
6929 shader_addline(&buffer
, "END\n");
6932 GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_FORMAT_ASCII_ARB
,
6933 strlen(buffer
.buffer
), buffer
.buffer
));
6934 checkGLcall("glProgramStringARB()");
6936 glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB
, &pos
);
6939 FIXME("Fragment program error at position %d: %s\n\n", pos
,
6940 debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB
)));
6941 shader_arb_dump_program_source(buffer
.buffer
);
6947 GL_EXTCALL(glGetProgramivARB(GL_FRAGMENT_PROGRAM_ARB
, GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB
, &native
));
6948 checkGLcall("glGetProgramivARB()");
6949 if (!native
) WARN("Program exceeds native resource limits.\n");
6952 shader_buffer_free(&buffer
);
6957 case COMPLEX_FIXUP_YUY2
:
6958 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yuy2_rect_shader
= shader
;
6959 else priv
->yuy2_2d_shader
= shader
;
6962 case COMPLEX_FIXUP_UYVY
:
6963 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->uyvy_rect_shader
= shader
;
6964 else priv
->uyvy_2d_shader
= shader
;
6967 case COMPLEX_FIXUP_YV12
:
6968 if (textype
== GL_TEXTURE_RECTANGLE_ARB
) priv
->yv12_rect_shader
= shader
;
6969 else priv
->yv12_2d_shader
= shader
;
6972 ERR("Unsupported complex fixup: %d\n", yuv_fixup
);
6978 /* Context activation is done by the caller. */
6979 static HRESULT
arbfp_blit_set(IWineD3DDevice
*iface
, IWineD3DSurfaceImpl
*surface
)
6982 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
6983 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
6984 float size
[4] = {surface
->pow2Width
, surface
->pow2Height
, 1, 1};
6985 struct arbfp_blit_priv
*priv
= device
->blit_priv
;
6986 enum complex_fixup fixup
;
6987 GLenum textype
= surface
->texture_target
;
6989 if (!is_complex_fixup(surface
->resource
.format_desc
->color_fixup
))
6992 dump_color_fixup_desc(surface
->resource
.format_desc
->color_fixup
);
6993 /* Don't bother setting up a shader for unconverted formats */
6996 checkGLcall("glEnable(textype)");
7001 fixup
= get_complex_fixup(surface
->resource
.format_desc
->color_fixup
);
7005 case COMPLEX_FIXUP_YUY2
:
7006 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yuy2_rect_shader
: priv
->yuy2_2d_shader
;
7009 case COMPLEX_FIXUP_UYVY
:
7010 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->uyvy_rect_shader
: priv
->uyvy_2d_shader
;
7013 case COMPLEX_FIXUP_YV12
:
7014 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->yv12_rect_shader
: priv
->yv12_2d_shader
;
7017 case COMPLEX_FIXUP_P8
:
7018 shader
= textype
== GL_TEXTURE_RECTANGLE_ARB
? priv
->p8_rect_shader
: priv
->p8_2d_shader
;
7019 if (!shader
) shader
= gen_p8_shader(device
, textype
);
7021 upload_palette(surface
);
7025 FIXME("Unsupported complex fixup %#x, not setting a shader\n", fixup
);
7028 checkGLcall("glEnable(textype)");
7033 if (!shader
) shader
= gen_yuv_shader(device
, fixup
, textype
);
7036 glEnable(GL_FRAGMENT_PROGRAM_ARB
);
7037 checkGLcall("glEnable(GL_FRAGMENT_PROGRAM_ARB)");
7038 GL_EXTCALL(glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB
, shader
));
7039 checkGLcall("glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, shader)");
7040 GL_EXTCALL(glProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB
, 0, size
));
7041 checkGLcall("glProgramLocalParameter4fvARB");
7047 /* Context activation is done by the caller. */
7048 static void arbfp_blit_unset(IWineD3DDevice
*iface
) {
7049 IWineD3DDeviceImpl
*device
= (IWineD3DDeviceImpl
*) iface
;
7050 const struct wined3d_gl_info
*gl_info
= &device
->adapter
->gl_info
;
7053 glDisable(GL_FRAGMENT_PROGRAM_ARB
);
7054 checkGLcall("glDisable(GL_FRAGMENT_PROGRAM_ARB)");
7055 glDisable(GL_TEXTURE_2D
);
7056 checkGLcall("glDisable(GL_TEXTURE_2D)");
7057 if (gl_info
->supported
[ARB_TEXTURE_CUBE_MAP
])
7059 glDisable(GL_TEXTURE_CUBE_MAP_ARB
);
7060 checkGLcall("glDisable(GL_TEXTURE_CUBE_MAP_ARB)");
7062 if (gl_info
->supported
[ARB_TEXTURE_RECTANGLE
])
7064 glDisable(GL_TEXTURE_RECTANGLE_ARB
);
7065 checkGLcall("glDisable(GL_TEXTURE_RECTANGLE_ARB)");
7070 static BOOL
arbfp_blit_supported(const struct wined3d_gl_info
*gl_info
, enum blit_operation blit_op
,
7071 const RECT
*src_rect
, DWORD src_usage
, WINED3DPOOL src_pool
,
7072 const struct wined3d_format_desc
*src_format_desc
,
7073 const RECT
*dst_rect
, DWORD dst_usage
, WINED3DPOOL dst_pool
,
7074 const struct wined3d_format_desc
*dst_format_desc
)
7076 enum complex_fixup src_fixup
;
7078 if (blit_op
!= BLIT_OP_BLIT
)
7080 TRACE("Unsupported blit_op=%d\n", blit_op
);
7084 src_fixup
= get_complex_fixup(src_format_desc
->color_fixup
);
7085 if (TRACE_ON(d3d_shader
) && TRACE_ON(d3d
))
7087 TRACE("Checking support for fixup:\n");
7088 dump_color_fixup_desc(src_format_desc
->color_fixup
);
7091 if (!is_identity_fixup(dst_format_desc
->color_fixup
))
7093 TRACE("Destination fixups are not supported\n");
7097 if (is_identity_fixup(src_format_desc
->color_fixup
))
7103 /* We only support YUV conversions. */
7104 if (!is_complex_fixup(src_format_desc
->color_fixup
))
7106 TRACE("[FAILED]\n");
7112 case COMPLEX_FIXUP_YUY2
:
7113 case COMPLEX_FIXUP_UYVY
:
7114 case COMPLEX_FIXUP_YV12
:
7115 case COMPLEX_FIXUP_P8
:
7120 FIXME("Unsupported YUV fixup %#x\n", src_fixup
);
7121 TRACE("[FAILED]\n");
7126 HRESULT
arbfp_blit_surface(IWineD3DDeviceImpl
*device
, IWineD3DSurfaceImpl
*src_surface
, const RECT
*src_rect
,
7127 IWineD3DSurfaceImpl
*dst_surface
, const RECT
*dst_rect_in
, enum blit_operation blit_op
,
7130 IWineD3DSwapChainImpl
*dst_swapchain
;
7131 struct wined3d_context
*context
;
7132 RECT dst_rect
= *dst_rect_in
;
7134 /* Now load the surface */
7135 surface_internal_preload(src_surface
, SRGB_RGB
);
7137 /* Activate the destination context, set it up for blitting */
7138 context
= context_acquire(device
, dst_surface
);
7139 context_apply_blit_state(context
, device
);
7141 /* The coordinates of the ddraw front buffer are always fullscreen ('screen coordinates',
7142 * while OpenGL coordinates are window relative.
7143 * Also beware of the origin difference(top left vs bottom left).
7144 * Also beware that the front buffer's surface size is screen width x screen height,
7145 * whereas the real gl drawable size is the size of the window. */
7146 dst_swapchain
= (dst_surface
->Flags
& SFLAG_SWAPCHAIN
) ? (IWineD3DSwapChainImpl
*)dst_surface
->container
: NULL
;
7147 if (dst_swapchain
&& dst_surface
== dst_swapchain
->front_buffer
)
7148 surface_translate_frontbuffer_coords(dst_surface
, context
->win_handle
, &dst_rect
);
7150 arbfp_blit_set((IWineD3DDevice
*)device
, src_surface
);
7154 /* Draw a textured quad */
7155 draw_textured_quad(src_surface
, src_rect
, &dst_rect
, Filter
);
7159 /* Leave the opengl state valid for blitting */
7160 arbfp_blit_unset((IWineD3DDevice
*)device
);
7162 if (wined3d_settings
.strict_draw_ordering
|| (dst_swapchain
7163 && (dst_surface
== dst_swapchain
->front_buffer
7164 || dst_swapchain
->num_contexts
> 1)))
7165 wglFlush(); /* Flush to ensure ordering across contexts. */
7167 context_release(context
);
7169 IWineD3DSurface_ModifyLocation((IWineD3DSurface
*)dst_surface
, SFLAG_INDRAWABLE
, TRUE
);
7173 static HRESULT
arbfp_blit_color_fill(IWineD3DDeviceImpl
*device
, IWineD3DSurfaceImpl
*dst_surface
, const RECT
*dst_rect
, DWORD fill_color
)
7175 FIXME("Color filling not implemented by arbfp_blit\n");
7176 return WINED3DERR_INVALIDCALL
;
7179 const struct blit_shader arbfp_blit
= {
7184 arbfp_blit_supported
,
7185 arbfp_blit_color_fill