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
181
182
183
184
185
186
187
188
189
190
|
// Copyright 2019 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "common/assert.h"
#include "common/common_types.h"
#include "common/microprofile.h"
#include "video_core/memory_manager.h"
#include "video_core/texture_cache/surface_base.h"
#include "video_core/texture_cache/surface_params.h"
#include "video_core/textures/convert.h"
namespace VideoCommon {
MICROPROFILE_DEFINE(GPU_Load_Texture, "GPU", "Texture Load", MP_RGB(128, 192, 128));
MICROPROFILE_DEFINE(GPU_Flush_Texture, "GPU", "Texture Flush", MP_RGB(128, 192, 128));
using Tegra::Texture::ConvertFromGuestToHost;
using VideoCore::MortonSwizzleMode;
using VideoCore::Surface::SurfaceCompression;
SurfaceBaseImpl::SurfaceBaseImpl(GPUVAddr gpu_addr, const SurfaceParams& params)
: params{params}, mipmap_sizes(params.num_levels),
mipmap_offsets(params.num_levels), gpu_addr{gpu_addr}, host_memory_size{
params.GetHostSizeInBytes()} {
std::size_t offset = 0;
for (u32 level = 0; level < params.num_levels; ++level) {
const std::size_t mipmap_size{params.GetGuestMipmapSize(level)};
mipmap_sizes[level] = mipmap_size;
mipmap_offsets[level] = offset;
offset += mipmap_size;
}
layer_size = offset;
if (params.is_layered) {
if (params.is_tiled) {
layer_size =
SurfaceParams::AlignLayered(layer_size, params.block_height, params.block_depth);
}
guest_memory_size = layer_size * params.depth;
} else {
guest_memory_size = layer_size;
}
}
void SurfaceBaseImpl::SwizzleFunc(MortonSwizzleMode mode, u8* memory, const SurfaceParams& params,
u8* buffer, u32 level) {
const u32 width{params.GetMipWidth(level)};
const u32 height{params.GetMipHeight(level)};
const u32 block_height{params.GetMipBlockHeight(level)};
const u32 block_depth{params.GetMipBlockDepth(level)};
std::size_t guest_offset{mipmap_offsets[level]};
if (params.is_layered) {
std::size_t host_offset{0};
const std::size_t guest_stride = layer_size;
const std::size_t host_stride = params.GetHostLayerSize(level);
for (u32 layer = 0; layer < params.depth; ++layer) {
MortonSwizzle(mode, params.pixel_format, width, block_height, height, block_depth, 1,
params.tile_width_spacing, buffer + host_offset, memory + guest_offset);
guest_offset += guest_stride;
host_offset += host_stride;
}
} else {
MortonSwizzle(mode, params.pixel_format, width, block_height, height, block_depth,
params.GetMipDepth(level), params.tile_width_spacing, buffer,
memory + guest_offset);
}
}
void SurfaceBaseImpl::LoadBuffer(Tegra::MemoryManager& memory_manager,
StagingCache& staging_cache) {
MICROPROFILE_SCOPE(GPU_Load_Texture);
auto& staging_buffer = staging_cache.GetBuffer(0);
u8* host_ptr;
is_continuous = memory_manager.IsBlockContinuous(gpu_addr, guest_memory_size);
// Handle continuouty
if (is_continuous) {
// Use physical memory directly
host_ptr = memory_manager.GetPointer(gpu_addr);
if (!host_ptr) {
return;
}
} else {
// Use an extra temporal buffer
auto& tmp_buffer = staging_cache.GetBuffer(1);
tmp_buffer.resize(guest_memory_size);
host_ptr = tmp_buffer.data();
memory_manager.ReadBlockUnsafe(gpu_addr, host_ptr, guest_memory_size);
}
if (params.is_tiled) {
ASSERT_MSG(params.block_width == 0, "Block width is defined as {} on texture target {}",
params.block_width, static_cast<u32>(params.target));
for (u32 level = 0; level < params.num_levels; ++level) {
const std::size_t host_offset{params.GetHostMipmapLevelOffset(level)};
SwizzleFunc(MortonSwizzleMode::MortonToLinear, host_ptr, params,
staging_buffer.data() + host_offset, level);
}
} else {
ASSERT_MSG(params.num_levels == 1, "Linear mipmap loading is not implemented");
const u32 bpp{params.GetBytesPerPixel()};
const u32 block_width{params.GetDefaultBlockWidth()};
const u32 block_height{params.GetDefaultBlockHeight()};
const u32 width{(params.width + block_width - 1) / block_width};
const u32 height{(params.height + block_height - 1) / block_height};
const u32 copy_size{width * bpp};
if (params.pitch == copy_size) {
std::memcpy(staging_buffer.data(), host_ptr, params.GetHostSizeInBytes());
} else {
const u8* start{host_ptr};
u8* write_to{staging_buffer.data()};
for (u32 h = height; h > 0; --h) {
std::memcpy(write_to, start, copy_size);
start += params.pitch;
write_to += copy_size;
}
}
}
auto compression_type = params.GetCompressionType();
if (compression_type == SurfaceCompression::None ||
compression_type == SurfaceCompression::Compressed)
return;
for (u32 level_up = params.num_levels; level_up > 0; --level_up) {
const u32 level = level_up - 1;
const std::size_t in_host_offset{params.GetHostMipmapLevelOffset(level)};
const std::size_t out_host_offset = compression_type == SurfaceCompression::Rearranged
? in_host_offset
: params.GetConvertedMipmapOffset(level);
u8* in_buffer = staging_buffer.data() + in_host_offset;
u8* out_buffer = staging_buffer.data() + out_host_offset;
ConvertFromGuestToHost(in_buffer, out_buffer, params.pixel_format,
params.GetMipWidth(level), params.GetMipHeight(level),
params.GetMipDepth(level), true, true);
}
}
void SurfaceBaseImpl::FlushBuffer(Tegra::MemoryManager& memory_manager,
StagingCache& staging_cache) {
MICROPROFILE_SCOPE(GPU_Flush_Texture);
auto& staging_buffer = staging_cache.GetBuffer(0);
u8* host_ptr;
// Handle continuouty
if (is_continuous) {
// Use physical memory directly
host_ptr = memory_manager.GetPointer(gpu_addr);
if (!host_ptr) {
return;
}
} else {
// Use an extra temporal buffer
auto& tmp_buffer = staging_cache.GetBuffer(1);
tmp_buffer.resize(guest_memory_size);
host_ptr = tmp_buffer.data();
}
if (params.is_tiled) {
ASSERT_MSG(params.block_width == 0, "Block width is defined as {}", params.block_width);
for (u32 level = 0; level < params.num_levels; ++level) {
const std::size_t host_offset{params.GetHostMipmapLevelOffset(level)};
SwizzleFunc(MortonSwizzleMode::LinearToMorton, host_ptr, params,
staging_buffer.data() + host_offset, level);
}
} else {
ASSERT(params.target == SurfaceTarget::Texture2D);
ASSERT(params.num_levels == 1);
const u32 bpp{params.GetBytesPerPixel()};
const u32 copy_size{params.width * bpp};
if (params.pitch == copy_size) {
std::memcpy(host_ptr, staging_buffer.data(), guest_memory_size);
} else {
u8* start{host_ptr};
const u8* read_to{staging_buffer.data()};
for (u32 h = params.height; h > 0; --h) {
std::memcpy(start, read_to, copy_size);
start += params.pitch;
read_to += copy_size;
}
}
}
if (!is_continuous) {
memory_manager.WriteBlockUnsafe(gpu_addr, host_ptr, guest_memory_size);
}
}
} // namespace VideoCommon
|
