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
|
// Copyright 2019 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <algorithm>
#include <vector>
#include <fmt/format.h>
#include "common/assert.h"
#include "common/bit_field.h"
#include "common/common_types.h"
#include "common/logging/log.h"
#include "video_core/engines/shader_bytecode.h"
#include "video_core/shader/node_helper.h"
#include "video_core/shader/shader_ir.h"
namespace VideoCommon::Shader {
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
namespace {
std::size_t GetImageTypeNumCoordinates(Tegra::Shader::ImageType image_type) {
switch (image_type) {
case Tegra::Shader::ImageType::Texture1D:
case Tegra::Shader::ImageType::TextureBuffer:
return 1;
case Tegra::Shader::ImageType::Texture1DArray:
case Tegra::Shader::ImageType::Texture2D:
return 2;
case Tegra::Shader::ImageType::Texture2DArray:
case Tegra::Shader::ImageType::Texture3D:
return 3;
}
UNREACHABLE();
return 1;
}
} // Anonymous namespace
u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
const auto opcode = OpCode::Decode(instr);
switch (opcode->get().GetId()) {
case OpCode::Id::SUST: {
UNIMPLEMENTED_IF(instr.sust.mode != Tegra::Shader::SurfaceDataMode::P);
UNIMPLEMENTED_IF(instr.sust.out_of_bounds_store != Tegra::Shader::OutOfBoundsStore::Ignore);
UNIMPLEMENTED_IF(instr.sust.component_mask_selector != 0xf); // Ensure we have an RGBA store
std::vector<Node> values;
constexpr std::size_t hardcoded_size{4};
for (std::size_t i = 0; i < hardcoded_size; ++i) {
values.push_back(GetRegister(instr.gpr0.Value() + i));
}
std::vector<Node> coords;
const std::size_t num_coords{GetImageTypeNumCoordinates(instr.sust.image_type)};
for (std::size_t i = 0; i < num_coords; ++i) {
coords.push_back(GetRegister(instr.gpr8.Value() + i));
}
const auto type{instr.sust.image_type};
auto& image{instr.sust.is_immediate ? GetImage(instr.image, type)
: GetBindlessImage(instr.gpr39, type)};
image.MarkWrite();
MetaImage meta{image, values};
bb.push_back(Operation(OperationCode::ImageStore, meta, std::move(coords)));
break;
}
case OpCode::Id::SUATOM: {
UNIMPLEMENTED_IF(instr.suatom_d.is_ba != 0);
Node value = GetRegister(instr.gpr0);
std::vector<Node> coords;
const std::size_t num_coords{GetImageTypeNumCoordinates(instr.sust.image_type)};
for (std::size_t i = 0; i < num_coords; ++i) {
coords.push_back(GetRegister(instr.gpr8.Value() + i));
}
const OperationCode operation_code = [instr] {
switch (instr.suatom_d.operation) {
case Tegra::Shader::ImageAtomicOperation::Add:
return OperationCode::AtomicImageAdd;
case Tegra::Shader::ImageAtomicOperation::Min:
return OperationCode::AtomicImageMin;
case Tegra::Shader::ImageAtomicOperation::Max:
return OperationCode::AtomicImageMax;
case Tegra::Shader::ImageAtomicOperation::And:
return OperationCode::AtomicImageAnd;
case Tegra::Shader::ImageAtomicOperation::Or:
return OperationCode::AtomicImageOr;
case Tegra::Shader::ImageAtomicOperation::Xor:
return OperationCode::AtomicImageXor;
case Tegra::Shader::ImageAtomicOperation::Exch:
return OperationCode::AtomicImageExchange;
default:
UNIMPLEMENTED_MSG("Unimplemented operation={}",
static_cast<u32>(instr.suatom_d.operation.Value()));
return OperationCode::AtomicImageAdd;
}
}();
const auto& image{GetImage(instr.image, instr.suatom_d.image_type, instr.suatom_d.size)};
MetaImage meta{image, {std::move(value)}};
SetRegister(bb, instr.gpr0, Operation(operation_code, meta, std::move(coords)));
break;
}
default:
UNIMPLEMENTED_MSG("Unhandled image instruction: {}", opcode->get().GetName());
}
return pc;
}
Image& ShaderIR::GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type,
std::optional<Tegra::Shader::ImageAtomicSize> size) {
const auto offset{static_cast<std::size_t>(image.index.Value())};
if (const auto image = TryUseExistingImage(offset, type, size)) {
return *image;
}
const std::size_t next_index{used_images.size()};
return used_images.emplace(offset, Image{offset, next_index, type, size}).first->second;
}
Image& ShaderIR::GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type,
std::optional<Tegra::Shader::ImageAtomicSize> size) {
const Node image_register{GetRegister(reg)};
const auto [base_image, cbuf_index, cbuf_offset]{
TrackCbuf(image_register, global_code, static_cast<s64>(global_code.size()))};
const auto cbuf_key{(static_cast<u64>(cbuf_index) << 32) | static_cast<u64>(cbuf_offset)};
if (const auto image = TryUseExistingImage(cbuf_key, type, size)) {
return *image;
}
const std::size_t next_index{used_images.size()};
return used_images.emplace(cbuf_key, Image{cbuf_index, cbuf_offset, next_index, type, size})
.first->second;
}
Image* ShaderIR::TryUseExistingImage(u64 offset, Tegra::Shader::ImageType type,
std::optional<Tegra::Shader::ImageAtomicSize> size) {
auto it = used_images.find(offset);
if (it == used_images.end()) {
return nullptr;
}
auto& image = it->second;
ASSERT(image.GetType() == type);
if (size) {
// We know the size, if it's known it has to be the same as before, otherwise we can set it.
if (image.IsSizeKnown()) {
ASSERT(image.GetSize() == size);
} else {
image.SetSize(*size);
}
}
return ℑ
}
} // namespace VideoCommon::Shader
|
