Linuxydable/suyu
1
0
Fork 0
forked from suyu/suyu
Code Pull requests Activity
suyu/src/video_core/renderer_vulkan/vk_buffer_cache.h

73 lines
2.1 KiB
C++
Raw Normal View History

vk_buffer_cache: Implement a buffer cache This buffer cache is just like OpenGL's buffer cache with some minor style changes. It uses VKStreamBuffer.
2019-02-28 04:49:28 +01:00
// Copyright 2019 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
#include <memory>
#include "common/common_types.h"
#include "video_core/buffer_cache/buffer_cache.h"
#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
#include "video_core/renderer_vulkan/vk_stream_buffer.h"
vulkan_common: Move allocator to the common directory Allow using the abstraction from the OpenGL backend.
2021-01-03 22:17:57 +01:00
#include "video_core/vulkan_common/vulkan_memory_allocator.h"
vulkan_common: Rename renderer_vulkan/wrapper.h to vulkan_common/vulkan_wrapper.h Allows sharing Vulkan wrapper code between different rendering backends.
2020-12-25 01:30:11 +01:00
#include "video_core/vulkan_common/vulkan_wrapper.h"
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
namespace Vulkan {
renderer_vulkan: Rename VKDevice to Device The "VK" prefix predates the "Vulkan" namespace. It was carried around the codebase for consistency. "VKDevice" currently is a bad alias with "VkDevice" (only an upcase character of difference) that can cause confusion. Rename all instances of it.
2020-12-26 05:10:53 +01:00
class Device;
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
class VKScheduler;
buffer_cache: Avoid passing references of shared pointers and misc style changes Instead of using as template argument a shared pointer, use the underlying type and manage shared pointers explicitly. This can make removing shared pointers from the cache more easy. While we are at it, make some misc style changes and general improvements (like insert_or_assign instead of operator[] + operator=).
2020-06-09 23:27:59 +02:00
class Buffer final : public VideoCommon::BufferBlock {
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
public:
renderer_vulkan: Rename Vulkan memory manager to memory allocator "Memory manager" collides with the guest GPU memory manager, and a memory allocator sounds closer to what the abstraction aims to be.
2021-01-03 22:11:01 +01:00
explicit Buffer(const Device& device, MemoryAllocator& memory_allocator, VKScheduler& scheduler,
vk_memory_manager: Improve memory manager and its API Fix a bug where the memory allocator could leave gaps between commits. To fix this the allocation algorithm was reworked, although it's still short in number of lines of code. Rework the allocation API to self-contained movable objects instead of naively using an unique_ptr to do the job for us. Remove the VK prefix.
2020-12-31 02:58:05 +01:00
StagingBufferPool& staging_pool, VAddr cpu_addr_, std::size_t size_);
buffer_cache: Avoid passing references of shared pointers and misc style changes Instead of using as template argument a shared pointer, use the underlying type and manage shared pointers explicitly. This can make removing shared pointers from the cache more easy. While we are at it, make some misc style changes and general improvements (like insert_or_assign instead of operator[] + operator=).
2020-06-09 23:27:59 +02:00
~Buffer();
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
video_core: Resolve more variable shadowing scenarios pt.3 Cleans out the rest of the occurrences of variable shadowing and makes any further occurrences of shadowing compiler errors.
2020-12-05 17:40:14 +01:00
void Upload(std::size_t offset, std::size_t data_size, const u8* data);
buffer_cache: Use buffer methods instead of cache virtual methods
2020-06-20 01:47:48 +02:00
video_core: Resolve more variable shadowing scenarios pt.3 Cleans out the rest of the occurrences of variable shadowing and makes any further occurrences of shadowing compiler errors.
2020-12-05 17:40:14 +01:00
void Download(std::size_t offset, std::size_t data_size, u8* data);
buffer_cache: Use buffer methods instead of cache virtual methods
2020-06-20 01:47:48 +02:00
void CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst_offset,
video_core: Resolve more variable shadowing scenarios pt.3 Cleans out the rest of the occurrences of variable shadowing and makes any further occurrences of shadowing compiler errors.
2020-12-05 17:40:14 +01:00
std::size_t copy_size);
buffer_cache: Use buffer methods instead of cache virtual methods
2020-06-20 01:47:48 +02:00
buffer_cache: Avoid passing references of shared pointers and misc style changes Instead of using as template argument a shared pointer, use the underlying type and manage shared pointers explicitly. This can make removing shared pointers from the cache more easy. While we are at it, make some misc style changes and general improvements (like insert_or_assign instead of operator[] + operator=).
2020-06-09 23:27:59 +02:00
VkBuffer Handle() const {
vk_memory_manager: Improve memory manager and its API Fix a bug where the memory allocator could leave gaps between commits. To fix this the allocation algorithm was reworked, although it's still short in number of lines of code. Rework the allocation API to self-contained movable objects instead of naively using an unique_ptr to do the job for us. Remove the VK prefix.
2020-12-31 02:58:05 +01:00
return *buffer;
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
}
gl_buffer_cache: Mark buffers as resident Make stream buffer and cached buffers as resident and query their address. This allows us to use GPU addresses for several proprietary Nvidia extensions.
2020-05-11 21:35:04 +02:00
u64 Address() const {
return 0;
}
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
private:
renderer_vulkan: Rename VKDevice to Device The "VK" prefix predates the "Vulkan" namespace. It was carried around the codebase for consistency. "VKDevice" currently is a bad alias with "VkDevice" (only an upcase character of difference) that can cause confusion. Rename all instances of it.
2020-12-26 05:10:53 +01:00
const Device& device;
buffer_cache: Use buffer methods instead of cache virtual methods
2020-06-20 01:47:48 +02:00
VKScheduler& scheduler;
vk_memory_manager: Improve memory manager and its API Fix a bug where the memory allocator could leave gaps between commits. To fix this the allocation algorithm was reworked, although it's still short in number of lines of code. Rework the allocation API to self-contained movable objects instead of naively using an unique_ptr to do the job for us. Remove the VK prefix.
2020-12-31 02:58:05 +01:00
StagingBufferPool& staging_pool;
buffer_cache: Use buffer methods instead of cache virtual methods
2020-06-20 01:47:48 +02:00
vk_memory_manager: Improve memory manager and its API Fix a bug where the memory allocator could leave gaps between commits. To fix this the allocation algorithm was reworked, although it's still short in number of lines of code. Rework the allocation API to self-contained movable objects instead of naively using an unique_ptr to do the job for us. Remove the VK prefix.
2020-12-31 02:58:05 +01:00
vk::Buffer buffer;
MemoryCommit commit;
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
};
renderer_vulkan: Drop Vulkan-Hpp
2020-03-27 05:33:21 +01:00
class VKBufferCache final : public VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer> {
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
public:
video_core: Rewrite the texture cache The current texture cache has several points that hurt maintainability and performance. It's easy to break unrelated parts of the cache when doing minor changes. The cache can easily forget valuable information about the cached textures by CPU writes or simply by its normal usage.The current texture cache has several points that hurt maintainability and performance. It's easy to break unrelated parts of the cache when doing minor changes. The cache can easily forget valuable information about the cached textures by CPU writes or simply by its normal usage. This commit aims to address those issues.
2020-12-30 06:25:23 +01:00
explicit VKBufferCache(VideoCore::RasterizerInterface& rasterizer,
Tegra::MemoryManager& gpu_memory, Core::Memory::Memory& cpu_memory,
renderer_vulkan: Rename Vulkan memory manager to memory allocator "Memory manager" collides with the guest GPU memory manager, and a memory allocator sounds closer to what the abstraction aims to be.
2021-01-03 22:11:01 +01:00
const Device& device, MemoryAllocator& memory_allocator,
video_core: Rewrite the texture cache The current texture cache has several points that hurt maintainability and performance. It's easy to break unrelated parts of the cache when doing minor changes. The cache can easily forget valuable information about the cached textures by CPU writes or simply by its normal usage.The current texture cache has several points that hurt maintainability and performance. It's easy to break unrelated parts of the cache when doing minor changes. The cache can easily forget valuable information about the cached textures by CPU writes or simply by its normal usage. This commit aims to address those issues.
2020-12-30 06:25:23 +01:00
VKScheduler& scheduler, VKStreamBuffer& stream_buffer,
vk_memory_manager: Improve memory manager and its API Fix a bug where the memory allocator could leave gaps between commits. To fix this the allocation algorithm was reworked, although it's still short in number of lines of code. Rework the allocation API to self-contained movable objects instead of naively using an unique_ptr to do the job for us. Remove the VK prefix.
2020-12-31 02:58:05 +01:00
StagingBufferPool& staging_pool);
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
~VKBufferCache();
gl_buffer_cache: Mark buffers as resident Make stream buffer and cached buffers as resident and query their address. This allows us to use GPU addresses for several proprietary Nvidia extensions.
2020-05-11 21:35:04 +02:00
BufferInfo GetEmptyBuffer(std::size_t size) override;
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
protected:
buffer_cache: Avoid passing references of shared pointers and misc style changes Instead of using as template argument a shared pointer, use the underlying type and manage shared pointers explicitly. This can make removing shared pointers from the cache more easy. While we are at it, make some misc style changes and general improvements (like insert_or_assign instead of operator[] + operator=).
2020-06-09 23:27:59 +02:00
std::shared_ptr<Buffer> CreateBlock(VAddr cpu_addr, std::size_t size) override;
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
private:
renderer_vulkan: Rename VKDevice to Device The "VK" prefix predates the "Vulkan" namespace. It was carried around the codebase for consistency. "VKDevice" currently is a bad alias with "VkDevice" (only an upcase character of difference) that can cause confusion. Rename all instances of it.
2020-12-26 05:10:53 +01:00
const Device& device;
renderer_vulkan: Rename Vulkan memory manager to memory allocator "Memory manager" collides with the guest GPU memory manager, and a memory allocator sounds closer to what the abstraction aims to be.
2021-01-03 22:11:01 +01:00
MemoryAllocator& memory_allocator;
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
VKScheduler& scheduler;
vk_memory_manager: Improve memory manager and its API Fix a bug where the memory allocator could leave gaps between commits. To fix this the allocation algorithm was reworked, although it's still short in number of lines of code. Rework the allocation API to self-contained movable objects instead of naively using an unique_ptr to do the job for us. Remove the VK prefix.
2020-12-31 02:58:05 +01:00
StagingBufferPool& staging_pool;
vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache The stream buffer before this commit once it was full (no more bytes to write before looping) waiting for all previous operations to finish. This was a temporary solution and had a noticeable performance penalty in performance (from what a profiler showed). To avoid this mark with fences usages of the stream buffer and once it loops wait for them to be signaled. On average this will never wait. Each fence knows where its usage finishes, resulting in a non-paged stream buffer. On the other side, the buffer cache is reimplemented using the generic buffer cache. It makes use of the staging buffer pool and the new stream buffer.
2020-01-06 21:59:20 +01:00
};
} // namespace Vulkan
Copy permalink