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
|
// Copyright 2018 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "core/memory.h"
#include "video_core/engines/fermi_2d.h"
#include "video_core/rasterizer_interface.h"
#include "video_core/textures/decoders.h"
namespace Tegra::Engines {
Fermi2D::Fermi2D(VideoCore::RasterizerInterface& rasterizer, MemoryManager& memory_manager)
: memory_manager(memory_manager), rasterizer{rasterizer} {}
void Fermi2D::WriteReg(u32 method, u32 value) {
ASSERT_MSG(method < Regs::NUM_REGS,
"Invalid Fermi2D register, increase the size of the Regs structure");
regs.reg_array[method] = value;
switch (method) {
case FERMI2D_REG_INDEX(trigger): {
HandleSurfaceCopy();
break;
}
}
}
void Fermi2D::HandleSurfaceCopy() {
LOG_WARNING(HW_GPU, "Requested a surface copy with operation {}",
static_cast<u32>(regs.operation));
const GPUVAddr source = regs.src.Address();
const GPUVAddr dest = regs.dst.Address();
// TODO(Subv): Only same-format and same-size copies are allowed for now.
ASSERT(regs.src.format == regs.dst.format);
ASSERT(regs.src.width * regs.src.height == regs.dst.width * regs.dst.height);
// TODO(Subv): Only raw copies are implemented.
ASSERT(regs.operation == Regs::Operation::SrcCopy);
const VAddr source_cpu = *memory_manager.GpuToCpuAddress(source);
const VAddr dest_cpu = *memory_manager.GpuToCpuAddress(dest);
u32 src_bytes_per_pixel = RenderTargetBytesPerPixel(regs.src.format);
u32 dst_bytes_per_pixel = RenderTargetBytesPerPixel(regs.dst.format);
if (!rasterizer.AccelerateSurfaceCopy(regs.src, regs.dst)) {
rasterizer.FlushRegion(source_cpu, src_bytes_per_pixel * regs.src.width * regs.src.height);
// We have to invalidate the destination region to evict any outdated surfaces from the
// cache. We do this before actually writing the new data because the destination address
// might contain a dirty surface that will have to be written back to memory.
rasterizer.InvalidateRegion(dest_cpu,
dst_bytes_per_pixel * regs.dst.width * regs.dst.height);
if (regs.src.linear == regs.dst.linear) {
// If the input layout and the output layout are the same, just perform a raw copy.
ASSERT(regs.src.BlockHeight() == regs.dst.BlockHeight());
Memory::CopyBlock(dest_cpu, source_cpu,
src_bytes_per_pixel * regs.dst.width * regs.dst.height);
return;
}
u8* src_buffer = Memory::GetPointer(source_cpu);
u8* dst_buffer = Memory::GetPointer(dest_cpu);
if (!regs.src.linear && regs.dst.linear) {
// If the input is tiled and the output is linear, deswizzle the input and copy it over.
Texture::CopySwizzledData(regs.src.width, regs.src.height, regs.src.depth,
src_bytes_per_pixel, dst_bytes_per_pixel, src_buffer,
dst_buffer, true, regs.src.BlockHeight(),
regs.src.BlockDepth());
} else {
// If the input is linear and the output is tiled, swizzle the input and copy it over.
Texture::CopySwizzledData(regs.src.width, regs.src.height, regs.src.depth,
src_bytes_per_pixel, dst_bytes_per_pixel, dst_buffer,
src_buffer, false, regs.dst.BlockHeight(),
regs.dst.BlockDepth());
}
}
}
} // namespace Tegra::Engines
|
