scl/strato
1
0
Fork 0
mirror of https://github.com/Takiiiiiiii/strato.git synced 2025-07-17 08:46:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

Rework Texture & Buffer for Context and FenceCycle Chaining

Browse Source 
GPU resources have been designed with locking by fences in mind, fences were treated as implicit locks on a GPU, design paradigms such as `GraphicsContext` simply unlocking the texture mutex after attaching it which would set the fence cycle were considered fine prior but are unoptimal as it enforces that a `FenceCycle` effectively ensures exclusivity. This conflates the function of a mutex which is mutual exclusion and that of the fence which is to track GPU-side completion and led to tying if it was acceptable to use a GPU resource to GPU completion rather than simply if it was not currently being used by the CPU which is the function of the mutex.

This rework fixes this with the groundwork that has been laid with previous commits, as `Context` semantics are utilized to move back to using mutexes for locking of resources and tracking the usage on the GPU in a cleaner way rather than arbitrary fence comparisons. This also leads to cleaning up a lot of methods that involved usage of fences that no longer require it and therefore can be entirely removed, further cleaning up the codebase. It also opens the door for future improvements such as the removal of `hostImmutableCycle` and replacing them with better solutions, the implementation of which is broken at the moment regardless.

While moving to `Context`-based locking the question of multiple GPU workloads being in-flight while using overlapping resources came up which brought a fundamental limitation of `FenceCycle` to light which was that only one resource could be concurrently attached to a cycle and it could not adequately represent multi-cycle dependencies. `FenceCycle` chaining was designed to fix this inadequacy and allows for several different GPU workloads to be in-flight concurrently while utilizing the same resources as long as they can ensure GPU-GPU synchronization.
This commit is contained in:
PixelyIon
2022-06-26 14:40:46 +05:30
parent 07d45ee504
commit 1239907ce8
6 changed files with 83 additions and 142 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
app/src/main/cpp/skyline/gpu/texture/texture.cpp
View File

@ -185,7 +185,7 @@ namespace skyline::gpu {
});
}
std::shared_ptr<memory::StagingBuffer> Texture::SynchronizeHostImpl(const std::shared_ptr<FenceCycle> &pCycle) {
std::shared_ptr<memory::StagingBuffer> Texture::SynchronizeHostImpl() {
if (!guest)
throw exception("Synchronization of host textures requires a valid guest texture to synchronize from");
else if (guest->dimensions != dimensions)
@ -207,8 +207,7 @@ namespace skyline::gpu {
if (layout == vk::ImageLayout::eUndefined)
TransitionLayout(vk::ImageLayout::eGeneral);
bufferData = std::get<memory::Image>(backing).data();
if (cycle.lock() != pCycle)
WaitOnFence();
WaitOnFence();
return nullptr;
} else {
throw exception("Guest -> Host synchronization of images tiled as '{}' isn't implemented", vk::to_string(tiling));
@ -313,9 +312,6 @@ namespace skyline::gpu {
}
}
if (stagingBuffer && cycle.lock() != pCycle)
WaitOnFence();
return stagingBuffer;
}
@ -632,10 +628,9 @@ namespace skyline::gpu {
void Texture::WaitOnFence() {
TRACE_EVENT("gpu", "Texture::WaitOnFence");
auto lCycle{cycle.lock()};
if (lCycle) {
lCycle->Wait();
cycle.reset();
if (cycle) {
cycle->Wait();
cycle = nullptr;
}
}
@ -682,12 +677,13 @@ namespace skyline::gpu {
TRACE_EVENT("gpu", "Texture::SynchronizeHost");
auto stagingBuffer{SynchronizeHostImpl(nullptr)};
auto stagingBuffer{SynchronizeHostImpl()};
if (stagingBuffer) {
auto lCycle{gpu.scheduler.Submit([&](vk::raii::CommandBuffer &commandBuffer) {
CopyFromStagingBuffer(commandBuffer, stagingBuffer);
})};
lCycle->AttachObjects(stagingBuffer, shared_from_this());
lCycle->ChainCycle(cycle);
cycle = lCycle;
}
@ -706,10 +702,11 @@ namespace skyline::gpu {
TRACE_EVENT("gpu", "Texture::SynchronizeHostWithBuffer");
auto stagingBuffer{SynchronizeHostImpl(pCycle)};
auto stagingBuffer{SynchronizeHostImpl()};
if (stagingBuffer) {
CopyFromStagingBuffer(commandBuffer, stagingBuffer);
pCycle->AttachObjects(stagingBuffer, shared_from_this());
pCycle->ChainCycle(cycle);
cycle = pCycle;
}
@ -743,7 +740,6 @@ namespace skyline::gpu {
TRACE_EVENT("gpu", "Texture::SynchronizeGuest");
WaitOnBacking();
WaitOnFence();
if (tiling == vk::ImageTiling::eOptimal || !std::holds_alternative<memory::Image>(backing)) {
auto stagingBuffer{gpu.memory.AllocateStagingBuffer(surfaceSize)};
@ -752,9 +748,11 @@ namespace skyline::gpu {
CopyIntoStagingBuffer(commandBuffer, stagingBuffer);
})};
lCycle->AttachObject(std::make_shared<TextureBufferCopy>(shared_from_this(), stagingBuffer));
lCycle->ChainCycle(cycle);
cycle = lCycle;
} else if (tiling == vk::ImageTiling::eLinear) {
// We can optimize linear texture sync on a UMA by mapping the texture onto the CPU and copying directly from it rather than using a staging buffer
WaitOnFence();
CopyToGuest(std::get<memory::Image>(backing).data());
} else {
throw exception("Host -> Guest synchronization of images tiled as '{}' isn't implemented", vk::to_string(tiling));
@ -779,8 +777,7 @@ namespace skyline::gpu {
TRACE_EVENT("gpu", "Texture::SynchronizeGuestWithBuffer");
WaitOnBacking();
if (cycle.lock() != pCycle)
WaitOnFence();
pCycle->ChainCycle(cycle);
if (tiling == vk::ImageTiling::eOptimal || !std::holds_alternative<memory::Image>(backing)) {
auto stagingBuffer{gpu.memory.AllocateStagingBuffer(surfaceSize)};
@ -812,10 +809,7 @@ namespace skyline::gpu {
void Texture::CopyFrom(std::shared_ptr<Texture> source, const vk::ImageSubresourceRange &subresource) {
WaitOnBacking();
WaitOnFence();
source->WaitOnBacking();
source->WaitOnFence();
if (source->layout == vk::ImageLayout::eUndefined)
throw exception("Cannot copy from image with undefined layout");
@ -896,6 +890,7 @@ namespace skyline::gpu {
});
})};
lCycle->AttachObjects(std::move(source), shared_from_this());
lCycle->ChainCycle(cycle);
cycle = lCycle;
}
}