d28595e0fd34c68a73524f0aa5062c7101cfac4c
1 /**************************************************************************
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * 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, sub license, 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:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
29 #include "i915_context.h"
31 #include "util/u_math.h"
34 #define A0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT)
35 #define D0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT)
36 #define T0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT)
37 #define A0_SRC0( reg ) (((reg)&UREG_MASK)>>UREG_A0_SRC0_SHIFT_LEFT)
38 #define A1_SRC0( reg ) (((reg)&UREG_MASK)<<UREG_A1_SRC0_SHIFT_RIGHT)
39 #define A1_SRC1( reg ) (((reg)&UREG_MASK)>>UREG_A1_SRC1_SHIFT_LEFT)
40 #define A2_SRC1( reg ) (((reg)&UREG_MASK)<<UREG_A2_SRC1_SHIFT_RIGHT)
41 #define A2_SRC2( reg ) (((reg)&UREG_MASK)>>UREG_A2_SRC2_SHIFT_LEFT)
43 /* These are special, and don't have swizzle/negate bits.
45 #define T0_SAMPLER( reg ) (GET_UREG_NR(reg)<<T0_SAMPLER_NR_SHIFT)
46 #define T1_ADDRESS_REG( reg ) ((GET_UREG_NR(reg)<<T1_ADDRESS_REG_NR_SHIFT) | \
47 (GET_UREG_TYPE(reg)<<T1_ADDRESS_REG_TYPE_SHIFT))
50 /* Macros for translating UREG's into the various register fields used
51 * by the I915 programmable unit.
53 #define UREG_A0_DEST_SHIFT_LEFT (UREG_TYPE_SHIFT - A0_DEST_TYPE_SHIFT)
54 #define UREG_A0_SRC0_SHIFT_LEFT (UREG_TYPE_SHIFT - A0_SRC0_TYPE_SHIFT)
55 #define UREG_A1_SRC0_SHIFT_RIGHT (A1_SRC0_CHANNEL_W_SHIFT - UREG_CHANNEL_W_SHIFT)
56 #define UREG_A1_SRC1_SHIFT_LEFT (UREG_TYPE_SHIFT - A1_SRC1_TYPE_SHIFT)
57 #define UREG_A2_SRC1_SHIFT_RIGHT (A2_SRC1_CHANNEL_W_SHIFT - UREG_CHANNEL_W_SHIFT)
58 #define UREG_A2_SRC2_SHIFT_LEFT (UREG_TYPE_SHIFT - A2_SRC2_TYPE_SHIFT)
60 #define UREG_MASK 0xffffff00
61 #define UREG_TYPE_NR_MASK ((REG_TYPE_MASK << UREG_TYPE_SHIFT) | \
62 (REG_NR_MASK << UREG_NR_SHIFT))
66 i915_get_temp(struct i915_fp_compile
*p
)
68 int bit
= ffs(~p
->temp_flag
);
70 i915_program_error(p
, "i915_get_temp: out of temporaries");
74 p
->temp_flag
|= 1 << (bit
- 1);
80 i915_release_temp(struct i915_fp_compile
*p
, int reg
)
82 p
->temp_flag
&= ~(1 << reg
);
87 * Get unpreserved temporary, a temp whose value is not preserved between
91 i915_get_utemp(struct i915_fp_compile
* p
)
93 int bit
= ffs(~p
->utemp_flag
);
95 i915_program_error(p
, "i915_get_utemp: out of temporaries");
99 p
->utemp_flag
|= 1 << (bit
- 1);
100 return UREG(REG_TYPE_U
, (bit
- 1));
104 i915_release_utemps(struct i915_fp_compile
*p
)
106 p
->utemp_flag
= ~0x7;
111 i915_emit_decl(struct i915_fp_compile
*p
,
112 uint type
, uint nr
, uint d0_flags
)
114 uint reg
= UREG(type
, nr
);
116 if (type
== REG_TYPE_T
) {
117 if (p
->decl_t
& (1 << nr
))
120 p
->decl_t
|= (1 << nr
);
122 else if (type
== REG_TYPE_S
) {
123 if (p
->decl_s
& (1 << nr
))
126 p
->decl_s
|= (1 << nr
);
131 if (p
->decl
< p
->declarations
+ I915_PROGRAM_SIZE
) {
132 *(p
->decl
++) = (D0_DCL
| D0_DEST(reg
) | d0_flags
);
133 *(p
->decl
++) = D1_MBZ
;
134 *(p
->decl
++) = D2_MBZ
;
137 i915_program_error(p
, "Out of declarations");
144 i915_emit_arith(struct i915_fp_compile
* p
,
148 uint saturate
, uint src0
, uint src1
, uint src2
)
153 assert(GET_UREG_TYPE(dest
) != REG_TYPE_CONST
);
154 dest
= UREG(GET_UREG_TYPE(dest
), GET_UREG_NR(dest
));
157 if (GET_UREG_TYPE(src0
) == REG_TYPE_CONST
)
159 if (GET_UREG_TYPE(src1
) == REG_TYPE_CONST
)
161 if (GET_UREG_TYPE(src2
) == REG_TYPE_CONST
)
164 /* Recursively call this function to MOV additional const values
165 * into temporary registers. Use utemp registers for this -
166 * currently shouldn't be possible to run out, but keep an eye on
170 uint s
[3], first
, i
, old_utemp_flag
;
175 old_utemp_flag
= p
->utemp_flag
;
177 first
= GET_UREG_NR(s
[c
[0]]);
178 for (i
= 1; i
< nr_const
; i
++) {
179 if (GET_UREG_NR(s
[c
[i
]]) != first
) {
180 uint tmp
= i915_get_utemp(p
);
182 i915_emit_arith(p
, A0_MOV
, tmp
, A0_DEST_CHANNEL_ALL
, 0,
191 p
->utemp_flag
= old_utemp_flag
; /* restore */
194 if (p
->csr
< p
->program
+ I915_PROGRAM_SIZE
) {
195 *(p
->csr
++) = (op
| A0_DEST(dest
) | mask
| saturate
| A0_SRC0(src0
));
196 *(p
->csr
++) = (A1_SRC0(src0
) | A1_SRC1(src1
));
197 *(p
->csr
++) = (A2_SRC1(src1
) | A2_SRC2(src2
));
200 i915_program_error(p
, "Out of instructions");
202 if (GET_UREG_TYPE(dest
) == REG_TYPE_R
)
203 p
->register_phases
[GET_UREG_NR(dest
)] = p
->nr_tex_indirect
;
211 * Emit a texture load or texkill instruction.
212 * \param dest the dest i915 register
213 * \param destmask the dest register writemask
214 * \param sampler the i915 sampler register
215 * \param coord the i915 source texcoord operand
216 * \param opcode the instruction opcode
218 uint
i915_emit_texld( struct i915_fp_compile
*p
,
225 const uint k
= UREG(GET_UREG_TYPE(coord
), GET_UREG_NR(coord
));
229 /* texcoord is swizzled or negated. Need to allocate a new temporary
230 * register (a utemp / unpreserved temp) won't do.
234 temp
= i915_get_temp(p
); /* get temp reg index */
235 tempReg
= UREG(REG_TYPE_R
, temp
); /* make i915 register */
237 i915_emit_arith( p
, A0_MOV
,
238 tempReg
, A0_DEST_CHANNEL_ALL
, /* dest reg, writemask */
240 coord
, 0, 0 ); /* src0, src1, src2 */
242 /* new src texcoord is tempReg */
246 /* Don't worry about saturate as we only support
248 if (destmask
!= A0_DEST_CHANNEL_ALL
) {
249 /* if not writing to XYZW... */
250 uint tmp
= i915_get_utemp(p
);
251 i915_emit_texld( p
, tmp
, A0_DEST_CHANNEL_ALL
, sampler
, coord
, opcode
);
252 i915_emit_arith( p
, A0_MOV
, dest
, destmask
, 0, tmp
, 0, 0 );
253 /* XXX release utemp here? */
256 assert(GET_UREG_TYPE(dest
) != REG_TYPE_CONST
);
257 assert(dest
== UREG(GET_UREG_TYPE(dest
), GET_UREG_NR(dest
)));
259 /* Output register being oC or oD defines a phase boundary */
260 if (GET_UREG_TYPE(dest
) == REG_TYPE_OC
||
261 GET_UREG_TYPE(dest
) == REG_TYPE_OD
)
262 p
->nr_tex_indirect
++;
264 /* Reading from an r# register whose contents depend on output of the
265 * current phase defines a phase boundary.
267 if (GET_UREG_TYPE(coord
) == REG_TYPE_R
&&
268 p
->register_phases
[GET_UREG_NR(coord
)] == p
->nr_tex_indirect
)
269 p
->nr_tex_indirect
++;
271 if (p
->csr
< p
->program
+ I915_PROGRAM_SIZE
) {
272 *(p
->csr
++) = (opcode
|
274 T0_SAMPLER( sampler
));
276 *(p
->csr
++) = T1_ADDRESS_REG( coord
);
277 *(p
->csr
++) = T2_MBZ
;
280 i915_program_error(p
, "Out of instructions");
282 if (GET_UREG_TYPE(dest
) == REG_TYPE_R
)
283 p
->register_phases
[GET_UREG_NR(dest
)] = p
->nr_tex_indirect
;
289 i915_release_temp(p
, temp
);
296 i915_emit_const1f(struct i915_fp_compile
* p
, float c0
)
298 struct i915_fragment_shader
*ifs
= p
->shader
;
302 return swizzle(UREG(REG_TYPE_R
, 0), ZERO
, ZERO
, ZERO
, ZERO
);
304 return swizzle(UREG(REG_TYPE_R
, 0), ONE
, ONE
, ONE
, ONE
);
306 for (reg
= 0; reg
< I915_MAX_CONSTANT
; reg
++) {
307 if (ifs
->constant_flags
[reg
] == I915_CONSTFLAG_USER
)
309 for (idx
= 0; idx
< 4; idx
++) {
310 if (!(ifs
->constant_flags
[reg
] & (1 << idx
)) ||
311 ifs
->constants
[reg
][idx
] == c0
) {
312 ifs
->constants
[reg
][idx
] = c0
;
313 ifs
->constant_flags
[reg
] |= 1 << idx
;
314 if (reg
+ 1 > ifs
->num_constants
)
315 ifs
->num_constants
= reg
+ 1;
316 return swizzle(UREG(REG_TYPE_CONST
, reg
), idx
, ZERO
, ZERO
, ONE
);
321 i915_program_error(p
, "i915_emit_const1f: out of constants");
326 i915_emit_const2f(struct i915_fp_compile
* p
, float c0
, float c1
)
328 struct i915_fragment_shader
*ifs
= p
->shader
;
332 return swizzle(i915_emit_const1f(p
, c1
), ZERO
, X
, Z
, W
);
334 return swizzle(i915_emit_const1f(p
, c1
), ONE
, X
, Z
, W
);
337 return swizzle(i915_emit_const1f(p
, c0
), X
, ZERO
, Z
, W
);
339 return swizzle(i915_emit_const1f(p
, c0
), X
, ONE
, Z
, W
);
341 // XXX emit swizzle here for 0, 1, -1 and any combination thereof
342 // we can use swizzle + neg for that
343 for (reg
= 0; reg
< I915_MAX_CONSTANT
; reg
++) {
344 if (ifs
->constant_flags
[reg
] == 0xf ||
345 ifs
->constant_flags
[reg
] == I915_CONSTFLAG_USER
)
347 for (idx
= 0; idx
< 3; idx
++) {
348 if (!(ifs
->constant_flags
[reg
] & (3 << idx
))) {
349 ifs
->constants
[reg
][idx
+ 0] = c0
;
350 ifs
->constants
[reg
][idx
+ 1] = c1
;
351 ifs
->constant_flags
[reg
] |= 3 << idx
;
352 if (reg
+ 1 > ifs
->num_constants
)
353 ifs
->num_constants
= reg
+ 1;
354 return swizzle(UREG(REG_TYPE_CONST
, reg
), idx
, idx
+ 1, ZERO
, ONE
);
359 i915_program_error(p
, "i915_emit_const2f: out of constants");
364 i915_emit_const4f(struct i915_fp_compile
* p
,
365 float c0
, float c1
, float c2
, float c3
)
367 struct i915_fragment_shader
*ifs
= p
->shader
;
370 // XXX emit swizzle here for 0, 1, -1 and any combination thereof
371 // we can use swizzle + neg for that
372 printf("const %f %f %f %f\n",c0
,c1
,c2
,c3
);
373 for (reg
= 0; reg
< I915_MAX_CONSTANT
; reg
++) {
374 if (ifs
->constant_flags
[reg
] == 0xf &&
375 ifs
->constants
[reg
][0] == c0
&&
376 ifs
->constants
[reg
][1] == c1
&&
377 ifs
->constants
[reg
][2] == c2
&&
378 ifs
->constants
[reg
][3] == c3
) {
379 return UREG(REG_TYPE_CONST
, reg
);
381 else if (ifs
->constant_flags
[reg
] == 0) {
383 ifs
->constants
[reg
][0] = c0
;
384 ifs
->constants
[reg
][1] = c1
;
385 ifs
->constants
[reg
][2] = c2
;
386 ifs
->constants
[reg
][3] = c3
;
387 ifs
->constant_flags
[reg
] = 0xf;
388 if (reg
+ 1 > ifs
->num_constants
)
389 ifs
->num_constants
= reg
+ 1;
390 return UREG(REG_TYPE_CONST
, reg
);
394 i915_program_error(p
, "i915_emit_const4f: out of constants");
400 i915_emit_const4fv(struct i915_fp_compile
* p
, const float * c
)
402 return i915_emit_const4f(p
, c
[0], c
[1], c
[2], c
[3]);