Merge remote branch 'origin/mesa_7_7_branch'
[mesa.git] / src / mesa / drivers / dri / r300 / r300_fragprog_common.c
1 /*
2 * Copyright (C) 2009 Maciej Cencora <m.cencora@gmail.com>
3 *
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining
7 * a copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sublicense, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial
16 * portions of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
21 * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
22 * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
23 * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
24 * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 */
27
28 /**
29 * \file
30 *
31 * Fragment program compiler. Perform transformations on the intermediate
32 * representation until the program is in a form where we can translate
33 * it more or less directly into machine-readable form.
34 *
35 * \author Ben Skeggs <darktama@iinet.net.au>
36 * \author Jerome Glisse <j.glisse@gmail.com>
37 */
38
39 #include "r300_fragprog_common.h"
40
41 #include "shader/program.h"
42 #include "shader/prog_parameter.h"
43 #include "shader/prog_print.h"
44
45 #include "compiler/radeon_compiler.h"
46
47 #include "radeon_mesa_to_rc.h"
48 #include "r300_state.h"
49
50
51 static GLuint build_dtm(GLuint depthmode)
52 {
53 switch(depthmode) {
54 default:
55 case GL_LUMINANCE: return 0;
56 case GL_INTENSITY: return 1;
57 case GL_ALPHA: return 2;
58 }
59 }
60
61 static GLuint build_func(GLuint comparefunc)
62 {
63 return comparefunc - GL_NEVER;
64 }
65
66 /**
67 * Collect all external state that is relevant for compiling the given
68 * fragment program.
69 */
70 static void build_state(
71 r300ContextPtr r300,
72 struct gl_fragment_program *fp,
73 struct r300_fragment_program_external_state *state)
74 {
75 int unit;
76
77 _mesa_bzero(state, sizeof(*state));
78
79 for(unit = 0; unit < 16; ++unit) {
80 if (fp->Base.ShadowSamplers & (1 << unit)) {
81 struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
82
83 state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
84 state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
85 }
86 }
87 }
88
89
90 /**
91 * Transform the program to support fragment.position.
92 *
93 * Introduce a small fragment at the start of the program that will be
94 * the only code that directly reads the FRAG_ATTRIB_WPOS input.
95 * All other code pieces that reference that input will be rewritten
96 * to read from a newly allocated temporary.
97 *
98 */
99 static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler, struct r300_fragment_program * fp)
100 {
101 int i;
102
103 fp->wpos_attr = FRAG_ATTRIB_MAX;
104 if (!(compiler->Base.Program.InputsRead & FRAG_BIT_WPOS)) {
105 return;
106 }
107
108 for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7; ++i)
109 {
110 if (!(compiler->Base.Program.InputsRead & (1 << i))) {
111 fp->wpos_attr = i;
112 break;
113 }
114 }
115
116 /* No free texcoord found, fall-back to software rendering */
117 if (fp->wpos_attr == FRAG_ATTRIB_MAX)
118 {
119 compiler->Base.Error = 1;
120 return;
121 }
122
123 rc_transform_fragment_wpos(&compiler->Base, FRAG_ATTRIB_WPOS, fp->wpos_attr);
124 }
125
126 /**
127 * Rewrite fragment.fogcoord to use a texture coordinate slot.
128 * Note that fogcoord is forced into an X001 pattern, and this enforcement
129 * is done here.
130 *
131 * See also the counterpart rewriting for vertex programs.
132 */
133 static void rewriteFog(struct r300_fragment_program_compiler *compiler, struct r300_fragment_program * fp)
134 {
135 struct rc_src_register src;
136 int i;
137
138 fp->fog_attr = FRAG_ATTRIB_MAX;
139 if (!(compiler->Base.Program.InputsRead & FRAG_BIT_FOGC)) {
140 return;
141 }
142
143 for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7; ++i)
144 {
145 if (!(compiler->Base.Program.InputsRead & (1 << i))) {
146 fp->fog_attr = i;
147 break;
148 }
149 }
150
151 /* No free texcoord found, fall-back to software rendering */
152 if (fp->fog_attr == FRAG_ATTRIB_MAX)
153 {
154 compiler->Base.Error = 1;
155 return;
156 }
157
158 memset(&src, 0, sizeof(src));
159 src.File = RC_FILE_INPUT;
160 src.Index = fp->fog_attr;
161 src.Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
162 rc_move_input(&compiler->Base, FRAG_ATTRIB_FOGC, src);
163 }
164
165
166 /**
167 * Reserve hardware temporary registers for the program inputs.
168 *
169 * @note This allocation is performed explicitly, because the order of inputs
170 * is determined by the RS hardware.
171 */
172 static void allocate_hw_inputs(
173 struct r300_fragment_program_compiler * c,
174 void (*allocate)(void * data, unsigned input, unsigned hwreg),
175 void * mydata)
176 {
177 GLuint InputsRead = c->Base.Program.InputsRead;
178 int i;
179 GLuint hwindex = 0;
180
181 /* Primary colour */
182 if (InputsRead & FRAG_BIT_COL0)
183 allocate(mydata, FRAG_ATTRIB_COL0, hwindex++);
184 InputsRead &= ~FRAG_BIT_COL0;
185
186 /* Secondary color */
187 if (InputsRead & FRAG_BIT_COL1)
188 allocate(mydata, FRAG_ATTRIB_COL1, hwindex++);
189 InputsRead &= ~FRAG_BIT_COL1;
190
191 /* Texcoords */
192 for (i = 0; i < 8; i++) {
193 if (InputsRead & (FRAG_BIT_TEX0 << i))
194 allocate(mydata, FRAG_ATTRIB_TEX0+i, hwindex++);
195 }
196 InputsRead &= ~FRAG_BITS_TEX_ANY;
197
198 /* Fogcoords treated as a texcoord */
199 if (InputsRead & FRAG_BIT_FOGC)
200 allocate(mydata, FRAG_ATTRIB_FOGC, hwindex++);
201 InputsRead &= ~FRAG_BIT_FOGC;
202
203 /* fragment position treated as a texcoord */
204 if (InputsRead & FRAG_BIT_WPOS)
205 allocate(mydata, FRAG_ATTRIB_WPOS, hwindex++);
206 InputsRead &= ~FRAG_BIT_WPOS;
207
208 /* Anything else */
209 if (InputsRead)
210 rc_error(&c->Base, "Don't know how to handle inputs 0x%x\n", InputsRead);
211 }
212
213
214 static void translate_fragment_program(GLcontext *ctx, struct r300_fragment_program_cont *cont, struct r300_fragment_program *fp)
215 {
216 r300ContextPtr r300 = R300_CONTEXT(ctx);
217 struct r300_fragment_program_compiler compiler;
218
219 rc_init(&compiler.Base);
220 compiler.Base.Debug = (RADEON_DEBUG & RADEON_PIXEL) ? GL_TRUE : GL_FALSE;
221
222 compiler.code = &fp->code;
223 compiler.state = fp->state;
224 compiler.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) ? GL_TRUE : GL_FALSE;
225 compiler.OutputDepth = FRAG_RESULT_DEPTH;
226 compiler.OutputColor = FRAG_RESULT_COLOR;
227 compiler.AllocateHwInputs = &allocate_hw_inputs;
228
229 if (compiler.Base.Debug) {
230 fflush(stderr);
231 _mesa_printf("Fragment Program: Initial program:\n");
232 _mesa_print_program(&cont->Base.Base);
233 fflush(stderr);
234 }
235
236 radeon_mesa_to_rc_program(&compiler.Base, &cont->Base.Base);
237
238 insert_WPOS_trailer(&compiler, fp);
239
240 rewriteFog(&compiler, fp);
241
242 r3xx_compile_fragment_program(&compiler);
243
244 if (compiler.is_r500) {
245 /* We need to support the non-KMS DRM interface, which
246 * artificially limits the number of instructions and
247 * constants which are available to us.
248 *
249 * See also the comment in r300_context.c where we
250 * set the MAX_NATIVE_xxx values.
251 */
252 if (fp->code.code.r500.inst_end >= 255 || fp->code.constants.Count > 255)
253 rc_error(&compiler.Base, "Program is too big (upgrade to r300g to avoid this limitation).\n");
254 }
255
256 fp->error = compiler.Base.Error;
257
258 fp->InputsRead = compiler.Base.Program.InputsRead;
259
260 rc_destroy(&compiler.Base);
261 }
262
263 struct r300_fragment_program *r300SelectAndTranslateFragmentShader(GLcontext *ctx)
264 {
265 r300ContextPtr r300 = R300_CONTEXT(ctx);
266 struct r300_fragment_program_cont *fp_list;
267 struct r300_fragment_program *fp;
268 struct r300_fragment_program_external_state state;
269
270 fp_list = (struct r300_fragment_program_cont *)ctx->FragmentProgram._Current;
271 build_state(r300, ctx->FragmentProgram._Current, &state);
272
273 fp = fp_list->progs;
274 while (fp) {
275 if (_mesa_memcmp(&fp->state, &state, sizeof(state)) == 0) {
276 return r300->selected_fp = fp;
277 }
278 fp = fp->next;
279 }
280
281 fp = _mesa_calloc(sizeof(struct r300_fragment_program));
282
283 fp->state = state;
284
285 fp->next = fp_list->progs;
286 fp_list->progs = fp;
287
288 translate_fragment_program(ctx, fp_list, fp);
289
290 return r300->selected_fp = fp;
291 }