1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
|
// Copyright 2021 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <optional>
#include "common/bit_field.h"
#include "common/common_types.h"
#include "shader_recompiler/frontend/ir/modifiers.h"
#include "shader_recompiler/frontend/maxwell/translate/impl/impl.h"
namespace Shader::Maxwell {
namespace {
enum class TextureType : u64 {
_1D,
ARRAY_1D,
_2D,
ARRAY_2D,
_3D,
ARRAY_3D,
CUBE,
ARRAY_CUBE,
};
Shader::TextureType GetType(TextureType type, bool dc) {
switch (type) {
case TextureType::_1D:
return dc ? Shader::TextureType::Shadow1D : Shader::TextureType::Color1D;
case TextureType::ARRAY_1D:
return dc ? Shader::TextureType::ShadowArray1D : Shader::TextureType::ColorArray1D;
case TextureType::_2D:
return dc ? Shader::TextureType::Shadow2D : Shader::TextureType::Color2D;
case TextureType::ARRAY_2D:
return dc ? Shader::TextureType::ShadowArray2D : Shader::TextureType::ColorArray2D;
case TextureType::_3D:
return dc ? Shader::TextureType::Shadow3D : Shader::TextureType::Color3D;
case TextureType::ARRAY_3D:
throw NotImplementedException("3D array texture type");
case TextureType::CUBE:
return dc ? Shader::TextureType::ShadowCube : Shader::TextureType::ColorCube;
case TextureType::ARRAY_CUBE:
return dc ? Shader::TextureType::ShadowArrayCube : Shader::TextureType::ColorArrayCube;
}
throw NotImplementedException("Invalid texture type {}", type);
}
IR::Value MakeOffset(TranslatorVisitor& v, IR::Reg reg, bool has_lod_clamp) {
const IR::U32 value{v.X(reg)};
const u32 base = has_lod_clamp ? 12 : 16;
return v.ir.CompositeConstruct(
v.ir.BitFieldExtract(value, v.ir.Imm32(base), v.ir.Imm32(4), true),
v.ir.BitFieldExtract(value, v.ir.Imm32(base + 4), v.ir.Imm32(4), true));
}
void Impl(TranslatorVisitor& v, u64 insn, bool is_bindless) {
union {
u64 raw;
BitField<49, 1, u64> nodep;
BitField<35, 1, u64> aoffi;
BitField<50, 1, u64> lc;
BitField<51, 3, IR::Pred> sparse_pred;
BitField<0, 8, IR::Reg> dest_reg;
BitField<8, 8, IR::Reg> coord_reg;
BitField<20, 8, IR::Reg> derivate_reg;
BitField<28, 3, TextureType> type;
BitField<31, 4, u64> mask;
BitField<36, 13, u64> cbuf_offset;
} const txd{insn};
const bool has_lod_clamp = txd.lc != 0;
if (has_lod_clamp) {
throw NotImplementedException("TXD.LC - CLAMP is not implemented");
}
IR::Value coords;
u32 num_derivates;
IR::Reg base_reg = txd.coord_reg;
IR::Reg last_reg;
IR::Value handle;
if (!is_bindless) {
handle = v.ir.Imm32(static_cast<u32>(txd.cbuf_offset.Value() * 4));
} else {
handle = v.X(base_reg++);
}
const auto read_array{[&]() -> IR::F32 {
return v.ir.ConvertUToF(32, 16,
v.ir.BitFieldExtract(v.X(last_reg), v.ir.Imm32(0),
v.ir.Imm32(has_lod_clamp ? 12 : 16)));
}};
switch (txd.type) {
case TextureType::_1D: {
coords = v.F(base_reg);
num_derivates = 1;
last_reg = base_reg + 1;
break;
}
case TextureType::ARRAY_1D: {
last_reg = base_reg + 1;
coords = v.ir.CompositeConstruct(v.F(base_reg), read_array());
num_derivates = 1;
break;
}
case TextureType::_2D: {
last_reg = base_reg + 2;
coords = v.ir.CompositeConstruct(v.F(base_reg), v.F(base_reg + 1));
num_derivates = 2;
break;
}
case TextureType::ARRAY_2D: {
last_reg = base_reg + 2;
coords = v.ir.CompositeConstruct(v.F(base_reg), v.F(base_reg + 1), read_array());
num_derivates = 2;
break;
}
default:
throw NotImplementedException("Invalid texture type");
}
const IR::Reg derivate_reg{txd.derivate_reg};
IR::Value derivates;
switch (num_derivates) {
case 1: {
derivates = v.ir.CompositeConstruct(v.F(derivate_reg), v.F(derivate_reg + 1));
break;
}
case 2: {
derivates = v.ir.CompositeConstruct(v.F(derivate_reg), v.F(derivate_reg + 1),
v.F(derivate_reg + 2), v.F(derivate_reg + 3));
break;
}
default:
throw NotImplementedException("Invalid texture type");
}
IR::Value offset;
if (txd.aoffi != 0) {
offset = MakeOffset(v, last_reg, has_lod_clamp);
}
IR::F32 lod_clamp;
if (has_lod_clamp) {
const IR::F32 conv4_8fixp_f = v.ir.Imm32(Common::BitCast<f32>(0x3b800000U));
const IR::F32 tmp = v.ir.ConvertUToF(
32, 16, v.ir.BitFieldExtract(v.X(last_reg), v.ir.Imm32(20), v.ir.Imm32(12)));
lod_clamp = v.ir.FPMul(tmp, conv4_8fixp_f);
}
IR::TextureInstInfo info{};
info.type.Assign(GetType(txd.type, false));
info.num_derivates.Assign(num_derivates);
info.has_lod_clamp.Assign(has_lod_clamp ? 1 : 0);
const IR::Value sample{[&]() -> IR::Value {
return v.ir.ImageGradient(handle, coords, derivates, offset, lod_clamp, info);
}()};
IR::Reg dest_reg{txd.dest_reg};
for (size_t element = 0; element < 4; ++element) {
if (((txd.mask >> element) & 1) == 0) {
continue;
}
v.F(dest_reg, IR::F32{v.ir.CompositeExtract(sample, element)});
++dest_reg;
}
if (txd.sparse_pred != IR::Pred::PT) {
v.ir.SetPred(txd.sparse_pred, v.ir.LogicalNot(v.ir.GetSparseFromOp(sample)));
}
}
} // Anonymous namespace
void TranslatorVisitor::TXD(u64 insn) {
Impl(*this, insn, false);
}
void TranslatorVisitor::TXD_b(u64 insn) {
Impl(*this, insn, true);
}
} // namespace Shader::Maxwell
|
