xref: /aosp_15_r20/external/mesa3d/src/panfrost/compiler/valhall/va_optimize.c (revision 6104692788411f58d303aa86923a9ff6ecaded22)
1 /*
2  * Copyright (C) 2021 Collabora Ltd.
3  *
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:
10  *
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
13  * Software.
14  *
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
21  * SOFTWARE.
22  */
23 
24 #include "va_compiler.h"
25 
26 /* Valhall specific instruction selection optimizations */
27 
28 static enum bi_opcode
va_op_add_imm(enum bi_opcode op)29 va_op_add_imm(enum bi_opcode op)
30 {
31    switch (op) {
32    case BI_OPCODE_FADD_F32:
33       return BI_OPCODE_FADD_IMM_F32;
34    case BI_OPCODE_FADD_V2F16:
35       return BI_OPCODE_FADD_IMM_V2F16;
36    case BI_OPCODE_IADD_S32:
37    case BI_OPCODE_IADD_U32:
38       return BI_OPCODE_IADD_IMM_I32;
39    case BI_OPCODE_IADD_V2S16:
40    case BI_OPCODE_IADD_V2U16:
41       return BI_OPCODE_IADD_IMM_V2I16;
42    case BI_OPCODE_IADD_V4S8:
43    case BI_OPCODE_IADD_V4U8:
44       return BI_OPCODE_IADD_IMM_V4I8;
45    default:
46       return 0;
47    }
48 }
49 
50 static bool
va_is_add_imm(bi_instr * I,unsigned s)51 va_is_add_imm(bi_instr *I, unsigned s)
52 {
53    assert(s < I->nr_srcs);
54 
55    return I->src[s].swizzle == BI_SWIZZLE_H01 && !I->src[s].abs &&
56           !I->src[s].neg && !I->clamp && !I->round;
57 }
58 
59 static unsigned
va_choose_imm(bi_instr * I)60 va_choose_imm(bi_instr *I)
61 {
62    for (unsigned i = 0; i < 2; ++i) {
63       if (I->src[i].type == BI_INDEX_CONSTANT)
64          return i;
65    }
66 
67    return ~0;
68 }
69 
70 /* Lower MOV.i32 #constant --> IADD_IMM.i32 0x0, #constant */
71 static void
va_lower_mov_imm(bi_instr * I)72 va_lower_mov_imm(bi_instr *I)
73 {
74    assert(I->nr_srcs == 1);
75 
76    if (I->src[0].type == BI_INDEX_CONSTANT) {
77       I->op = BI_OPCODE_IADD_IMM_I32;
78       I->index = I->src[0].value;
79       I->src[0] = bi_zero();
80    }
81 }
82 
83 void
va_fuse_add_imm(bi_instr * I)84 va_fuse_add_imm(bi_instr *I)
85 {
86    if (I->op == BI_OPCODE_MOV_I32) {
87       va_lower_mov_imm(I);
88       return;
89    }
90 
91    enum bi_opcode op = va_op_add_imm(I->op);
92    if (!op)
93       return;
94 
95    unsigned s = va_choose_imm(I);
96    if (s > 1)
97       return;
98    if (!va_is_add_imm(I, 1 - s))
99       return;
100 
101    I->op = op;
102    I->index = bi_apply_swizzle(I->src[s].value, I->src[s].swizzle);
103 
104    assert(!I->src[s].abs && "redundant .abs set");
105 
106    /* If the constant is negated, flip the sign bit */
107    if (I->src[s].neg) {
108       if (I->op == BI_OPCODE_FADD_IMM_F32)
109          I->index ^= (1u << 31);
110       else if (I->op == BI_OPCODE_FADD_IMM_V2F16)
111          I->index ^= (1u << 31) | (1u << 15);
112       else
113          unreachable("unexpected .neg");
114    }
115 
116    I->src[0] = I->src[1 - s];
117    bi_drop_srcs(I, 1);
118 }
119 
120 void
va_optimize(bi_context * ctx)121 va_optimize(bi_context *ctx)
122 {
123    bi_foreach_instr_global(ctx, I) {
124       va_fuse_add_imm(I);
125    }
126 }
127