2 * Copyright (C) 2019 Google.
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 FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
24 #include "util/ralloc.h"
29 is_fp16_conv(struct ir3_instruction
*instr
)
31 if (instr
->opc
!= OPC_MOV
)
34 struct ir3_register
*dst
= instr
->regs
[0];
35 struct ir3_register
*src
= instr
->regs
[1];
37 /* disallow conversions that cannot be folded into
40 if (dst
->flags
& (IR3_REG_EVEN
| IR3_REG_POS_INF
))
43 if (dst
->flags
& (IR3_REG_RELATIV
| IR3_REG_ARRAY
))
45 if (src
->flags
& (IR3_REG_RELATIV
| IR3_REG_ARRAY
))
48 if (instr
->cat1
.src_type
== TYPE_F32
&&
49 instr
->cat1
.dst_type
== TYPE_F16
)
52 if (instr
->cat1
.src_type
== TYPE_F16
&&
53 instr
->cat1
.dst_type
== TYPE_F32
)
60 all_uses_fp16_conv(struct ir3_instruction
*conv_src
)
62 foreach_ssa_use (use
, conv_src
)
63 if (!is_fp16_conv(use
))
68 /* For an instruction which has a conversion folded in, re-write the
69 * uses of *all* conv's that used that src to be a simple mov that
70 * cp can eliminate. This avoids invalidating the SSA uses, it just
71 * shifts the use to a simple mov.
74 rewrite_src_uses(struct ir3_instruction
*src
)
76 foreach_ssa_use (use
, src
) {
77 assert(is_fp16_conv(use
));
80 use
->regs
[1]->flags
|= IR3_REG_HALF
;
82 use
->regs
[1]->flags
&= ~IR3_REG_HALF
;
85 use
->cat1
.src_type
= use
->cat1
.dst_type
;
90 try_conversion_folding(struct ir3_instruction
*conv
)
92 struct ir3_instruction
*src
;
94 if (!is_fp16_conv(conv
))
97 /* NOTE: we can have non-ssa srcs after copy propagation: */
98 src
= ssa(conv
->regs
[1]);
105 /* avoid folding f2f32(f2f16) together, in cases where this is legal to
106 * do (glsl) nir should have handled that for us already:
108 if (is_fp16_conv(src
))
120 /* if src is a "cov" and type doesn't match, then it can't be folded
121 * for example cov.u32u16+cov.f16f32 can't be folded to cov.u32f32
123 if (src
->cat1
.dst_type
!= src
->cat1
.src_type
&&
124 conv
->cat1
.src_type
!= src
->cat1
.dst_type
)
131 if (!all_uses_fp16_conv(src
))
134 if (src
->opc
== OPC_MOV
) {
135 if (src
->cat1
.dst_type
== src
->cat1
.src_type
) {
136 /* If we're folding a conversion into a bitwise move, we need to
137 * change the dst type to F32 to get the right behavior, since we
138 * could be moving a float with a u32.u32 move.
140 src
->cat1
.dst_type
= conv
->cat1
.dst_type
;
141 src
->cat1
.src_type
= conv
->cat1
.src_type
;
143 /* Otherwise, for typechanging movs, we can just change the dst
144 * type to F16 to collaps the two conversions. For example
145 * cov.s32f32 follwed by cov.f32f16 becomes cov.s32f16.
147 src
->cat1
.dst_type
= conv
->cat1
.dst_type
;
151 ir3_set_dst_type(src
, is_half(conv
));
152 rewrite_src_uses(src
);
158 ir3_cf(struct ir3
*ir
)
160 void *mem_ctx
= ralloc_context(NULL
);
161 bool progress
= false;
163 ir3_find_ssa_uses(ir
, mem_ctx
, false);
165 foreach_block (block
, &ir
->block_list
) {
166 foreach_instr (instr
, &block
->instr_list
) {
167 progress
|= try_conversion_folding(instr
);
171 ralloc_free(mem_ctx
);