2 * Copyright © 2014 Broadcom
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
25 * @file vc4_opt_algebraic.c
27 * This is the optimization pass for miscellaneous changes to instructions
28 * where we can simplify the operation by some knowledge about the specific
31 * Mostly this will be a matter of turning things into MOVs so that they can
32 * later be copy-propagated out.
36 #include "util/u_math.h"
41 dump_from(struct vc4_compile
*c
, struct qinst
*inst
)
46 fprintf(stderr
, "optimizing: ");
47 qir_dump_inst(c
, inst
);
48 fprintf(stderr
, "\n");
52 dump_to(struct vc4_compile
*c
, struct qinst
*inst
)
57 fprintf(stderr
, "to: ");
58 qir_dump_inst(c
, inst
);
59 fprintf(stderr
, "\n");
63 follow_movs(struct qinst
**defs
, struct qreg reg
)
65 while (reg
.file
== QFILE_TEMP
&& defs
[reg
.index
]->op
== QOP_MOV
)
66 reg
= defs
[reg
.index
]->src
[0];
72 is_zero(struct vc4_compile
*c
, struct qinst
**defs
, struct qreg reg
)
74 reg
= follow_movs(defs
, reg
);
76 return (reg
.file
== QFILE_UNIF
&&
77 c
->uniform_contents
[reg
.index
] == QUNIFORM_CONSTANT
&&
78 c
->uniform_data
[reg
.index
] == 0);
82 is_1f(struct vc4_compile
*c
, struct qinst
**defs
, struct qreg reg
)
84 reg
= follow_movs(defs
, reg
);
86 return (reg
.file
== QFILE_UNIF
&&
87 c
->uniform_contents
[reg
.index
] == QUNIFORM_CONSTANT
&&
88 c
->uniform_data
[reg
.index
] == fui(1.0));
92 replace_with_mov(struct vc4_compile
*c
, struct qinst
*inst
, struct qreg arg
)
97 inst
->src
[1] = c
->undef
;
102 add_replace_zero(struct vc4_compile
*c
,
107 if (!is_zero(c
, defs
, inst
->src
[arg
]))
109 replace_with_mov(c
, inst
, inst
->src
[1 - arg
]);
114 fmul_replace_zero(struct vc4_compile
*c
,
119 if (!is_zero(c
, defs
, inst
->src
[arg
]))
121 replace_with_mov(c
, inst
, inst
->src
[arg
]);
126 fmul_replace_one(struct vc4_compile
*c
,
131 if (!is_1f(c
, defs
, inst
->src
[arg
]))
133 replace_with_mov(c
, inst
, inst
->src
[1 - arg
]);
138 qir_opt_algebraic(struct vc4_compile
*c
)
140 bool progress
= false;
141 struct simple_node
*node
;
142 struct qinst
*defs
[c
->num_temps
];
144 foreach(node
, &c
->instructions
) {
145 struct qinst
*inst
= (struct qinst
*)node
;
147 if (inst
->dst
.file
== QFILE_TEMP
)
148 defs
[inst
->dst
.index
] = inst
;
152 /* SF just looks at the sign bit, or whether all the
153 * bits are 0. This is preserved across an itof
156 if (inst
->src
[0].file
== QFILE_TEMP
&&
157 defs
[inst
->src
[0].index
]->op
== QOP_ITOF
) {
160 defs
[inst
->src
[0].index
]->src
[0];
171 if (qir_reg_equals(inst
->src
[0], inst
->src
[1])) {
172 /* Turn "dst = (sf == x) ? a : a)" into
175 replace_with_mov(c
, inst
, inst
->src
[1]);
180 if (is_zero(c
, defs
, inst
->src
[1])) {
181 /* Replace references to a 0 uniform value
182 * with the SEL_X_0 equivalent.
185 inst
->op
-= (QOP_SEL_X_Y_ZS
- QOP_SEL_X_0_ZS
);
186 inst
->src
[1] = c
->undef
;
192 if (is_zero(c
, defs
, inst
->src
[0])) {
193 /* Replace references to a 0 uniform value
194 * with the SEL_X_0 equivalent, flipping the
195 * condition being evaluated since the operand
199 inst
->op
-= QOP_SEL_X_Y_ZS
;
201 inst
->op
+= QOP_SEL_X_0_ZS
;
202 inst
->src
[0] = inst
->src
[1];
203 inst
->src
[1] = c
->undef
;
213 if (is_zero(c
, defs
, inst
->src
[1])) {
214 replace_with_mov(c
, inst
, inst
->src
[0]);
219 if (add_replace_zero(c
, defs
, inst
, 0) ||
220 add_replace_zero(c
, defs
, inst
, 1)) {
227 if (add_replace_zero(c
, defs
, inst
, 0) ||
228 add_replace_zero(c
, defs
, inst
, 1)) {
233 /* FADD(a, FSUB(0, b)) -> FSUB(a, b) */
234 if (inst
->src
[1].file
== QFILE_TEMP
&&
235 defs
[inst
->src
[1].index
]->op
== QOP_FSUB
) {
236 struct qinst
*fsub
= defs
[inst
->src
[1].index
];
237 if (is_zero(c
, defs
, fsub
->src
[0])) {
240 inst
->src
[1] = fsub
->src
[1];
247 /* FADD(FSUB(0, b), a) -> FSUB(a, b) */
248 if (inst
->src
[0].file
== QFILE_TEMP
&&
249 defs
[inst
->src
[0].index
]->op
== QOP_FSUB
) {
250 struct qinst
*fsub
= defs
[inst
->src
[0].index
];
251 if (is_zero(c
, defs
, fsub
->src
[0])) {
254 inst
->src
[0] = inst
->src
[1];
255 inst
->src
[1] = fsub
->src
[1];
264 if (fmul_replace_zero(c
, defs
, inst
, 0) ||
265 fmul_replace_zero(c
, defs
, inst
, 1) ||
266 fmul_replace_one(c
, defs
, inst
, 0) ||
267 fmul_replace_one(c
, defs
, inst
, 1)) {