0
0
Fork 0
This repository has been archived on 2024-10-12. You can view files and clone it, but cannot push or open issues or pull requests.
ryujinx-final/Ryujinx.Graphics.Gpu/Memory/BufferManager.cs

1001 lines
38 KiB
C#
Raw Normal View History

using Ryujinx.Common;
2019-10-13 06:02:07 +00:00
using Ryujinx.Graphics.GAL;
using Ryujinx.Graphics.Gpu.State;
using Ryujinx.Graphics.Shader;
Memory Read/Write Tracking using Region Handles (#1272) * WIP Range Tracking - Texture invalidation seems to have large problems - Buffer/Pool invalidation may have problems - Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution. - Native project is in the messiest possible location. - [HACK] JIT memory access always uses native "fast" path - [HACK] Trying some things with texture invalidation and views. It works :) Still a few hacks, messy things, slow things More work in progress stuff (also move to memory project) Quite a bit faster now. - Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former. - The Virtual range list is now non-overlapping like the physical one. - Fixed some bugs where regions could leak. - Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road) Move some stuff. I think we'll eventually just put the dll and so for this in a nuget package. Fix rebase. [WIP] MultiRegionHandle variable size ranges - Avoid reprotecting regions that change often (needs some tweaking) - There's still a bug in buffers, somehow. - Might want different api for minimum granularity Fix rebase issue Commit everything needed for software only tracking. Remove native components. Remove more native stuff. Cleanup Use a separate window for the background context, update opentk. (fixes linux) Some experimental changes Should get things working up to scratch - still need to try some things with flush/modification and res scale. Include address with the region action. Initial work to make range tracking work Still a ton of bugs Fix some issues with the new stuff. * Fix texture flush instability There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it) * Find the destination texture for Buffer->Texture full copy Greatly improves performance for nvdec videos (with range tracking) * Further improve texture tracking * Disable Memory Tracking for view parents This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice) The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future. * Introduce some tracking tests. WIP * Complete base tests. * Add more tests for multiregion, fix existing test. * Cleanup Part 1 * Remove unnecessary code from memory tracking * Fix some inconsistencies with 3D texture rule. * Add dispose tests. * Use a background thread for the background context. Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster. Also nerf the multithreading test a bit. * Copy to texture with matching alignment This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size. * Track reads for buffer copies. Synchronize new buffers before copying overlaps. * Remove old texture flushing mechanisms. Range tracking all the way, baby. * Wake the background thread when disposing. Avoids a deadlock when games are closed. * Address Feedback 1 * Separate TextureCopy instance for background thread Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread. * Add missing XML docs. * Address Feedback * Maybe I should start drinking coffee. * Some more feedback. * Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
using Ryujinx.Memory.Range;
2019-10-13 06:02:07 +00:00
using System;
using System.Collections.ObjectModel;
using System.Linq;
2019-10-13 06:02:07 +00:00
namespace Ryujinx.Graphics.Gpu.Memory
{
/// <summary>
/// Buffer manager.
/// </summary>
2019-10-13 06:02:07 +00:00
class BufferManager
{
private const int StackToHeapThreshold = 16;
private const int OverlapsBufferInitialCapacity = 10;
private const int OverlapsBufferMaxCapacity = 10000;
2019-10-13 06:02:07 +00:00
private const ulong BufferAlignmentSize = 0x1000;
private const ulong BufferAlignmentMask = BufferAlignmentSize - 1;
private GpuContext _context;
private RangeList<Buffer> _buffers;
private Buffer[] _bufferOverlaps;
2019-10-13 06:02:07 +00:00
private IndexBuffer _indexBuffer;
private VertexBuffer[] _vertexBuffers;
private BufferBounds[] _transformFeedbackBuffers;
2019-10-13 06:02:07 +00:00
/// <summary>
/// Holds shader stage buffer state and binding information.
/// </summary>
2019-10-13 06:02:07 +00:00
private class BuffersPerStage
{
/// <summary>
/// Shader buffer binding information.
/// </summary>
public BufferDescriptor[] Bindings { get; }
/// <summary>
/// Buffer regions.
/// </summary>
2019-10-13 06:02:07 +00:00
public BufferBounds[] Buffers { get; }
/// <summary>
/// Total amount of buffers used on the shader.
/// </summary>
public int Count { get; private set; }
/// <summary>
/// Creates a new instance of the shader stage buffer information.
/// </summary>
/// <param name="count">Maximum amount of buffers that the shader stage can use</param>
2019-10-13 06:02:07 +00:00
public BuffersPerStage(int count)
{
Bindings = new BufferDescriptor[count];
2019-10-13 06:02:07 +00:00
Buffers = new BufferBounds[count];
}
/// <summary>
/// Sets the region of a buffer at a given slot.
/// </summary>
/// <param name="index">Buffer slot</param>
/// <param name="address">Region virtual address</param>
/// <param name="size">Region size in bytes</param>
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
/// <param name="flags">Buffer usage flags</param>
public void SetBounds(int index, ulong address, ulong size, BufferUsageFlags flags = BufferUsageFlags.None)
2019-10-13 06:02:07 +00:00
{
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
Buffers[index] = new BufferBounds(address, size, flags);
}
/// <summary>
/// Sets shader buffer binding information.
/// </summary>
/// <param name="descriptors">Buffer binding information</param>
public void SetBindings(ReadOnlyCollection<BufferDescriptor> descriptors)
{
if (descriptors == null)
{
Count = 0;
return;
}
descriptors.CopyTo(Bindings, 0);
Count = descriptors.Count;
2019-10-13 06:02:07 +00:00
}
}
private BuffersPerStage _cpStorageBuffers;
private BuffersPerStage _cpUniformBuffers;
private BuffersPerStage[] _gpStorageBuffers;
private BuffersPerStage[] _gpUniformBuffers;
private int _cpStorageBufferBindings;
private int _cpUniformBufferBindings;
private int _gpStorageBufferBindings;
private int _gpUniformBufferBindings;
2019-10-13 06:02:07 +00:00
private bool _gpStorageBuffersDirty;
private bool _gpUniformBuffersDirty;
private bool _indexBufferDirty;
private bool _vertexBuffersDirty;
private uint _vertexBuffersEnableMask;
private bool _transformFeedbackBuffersDirty;
2019-10-13 06:02:07 +00:00
private bool _rebind;
/// <summary>
/// Creates a new instance of the buffer manager.
/// </summary>
/// <param name="context">The GPU context that the buffer manager belongs to</param>
2019-10-13 06:02:07 +00:00
public BufferManager(GpuContext context)
{
_context = context;
_buffers = new RangeList<Buffer>();
_bufferOverlaps = new Buffer[OverlapsBufferInitialCapacity];
2019-10-13 06:02:07 +00:00
_vertexBuffers = new VertexBuffer[Constants.TotalVertexBuffers];
_transformFeedbackBuffers = new BufferBounds[Constants.TotalTransformFeedbackBuffers];
2019-10-13 06:02:07 +00:00
_cpStorageBuffers = new BuffersPerStage(Constants.TotalCpStorageBuffers);
_cpUniformBuffers = new BuffersPerStage(Constants.TotalCpUniformBuffers);
2020-01-01 15:39:09 +00:00
_gpStorageBuffers = new BuffersPerStage[Constants.ShaderStages];
_gpUniformBuffers = new BuffersPerStage[Constants.ShaderStages];
2019-10-13 06:02:07 +00:00
2020-01-01 15:39:09 +00:00
for (int index = 0; index < Constants.ShaderStages; index++)
2019-10-13 06:02:07 +00:00
{
_gpStorageBuffers[index] = new BuffersPerStage(Constants.TotalGpStorageBuffers);
_gpUniformBuffers[index] = new BuffersPerStage(Constants.TotalGpUniformBuffers);
}
}
/// <summary>
/// Sets the memory range with the index buffer data, to be used for subsequent draw calls.
/// </summary>
/// <param name="gpuVa">Start GPU virtual address of the index buffer</param>
/// <param name="size">Size, in bytes, of the index buffer</param>
/// <param name="type">Type of each index buffer element</param>
2019-10-13 06:02:07 +00:00
public void SetIndexBuffer(ulong gpuVa, ulong size, IndexType type)
{
ulong address = TranslateAndCreateBuffer(gpuVa, size);
_indexBuffer.Address = address;
_indexBuffer.Size = size;
_indexBuffer.Type = type;
_indexBufferDirty = true;
}
/// <summary>
/// Sets a new index buffer that overrides the one set on the call to <see cref="CommitGraphicsBindings"/>.
/// </summary>
/// <param name="buffer">Buffer to be used as index buffer</param>
/// <param name="type">Type of each index buffer element</param>
public void SetIndexBuffer(BufferRange buffer, IndexType type)
{
_context.Renderer.Pipeline.SetIndexBuffer(buffer, type);
_indexBufferDirty = true;
}
/// <summary>
/// Sets the memory range with vertex buffer data, to be used for subsequent draw calls.
/// </summary>
/// <param name="index">Index of the vertex buffer (up to 16)</param>
/// <param name="gpuVa">GPU virtual address of the buffer</param>
/// <param name="size">Size in bytes of the buffer</param>
/// <param name="stride">Stride of the buffer, defined as the number of bytes of each vertex</param>
/// <param name="divisor">Vertex divisor of the buffer, for instanced draws</param>
2019-10-13 06:02:07 +00:00
public void SetVertexBuffer(int index, ulong gpuVa, ulong size, int stride, int divisor)
{
ulong address = TranslateAndCreateBuffer(gpuVa, size);
_vertexBuffers[index].Address = address;
_vertexBuffers[index].Size = size;
_vertexBuffers[index].Stride = stride;
_vertexBuffers[index].Divisor = divisor;
_vertexBuffersDirty = true;
if (address != 0)
{
_vertexBuffersEnableMask |= 1u << index;
}
else
{
_vertexBuffersEnableMask &= ~(1u << index);
}
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Sets a transform feedback buffer on the graphics pipeline.
/// The output from the vertex transformation stages are written into the feedback buffer.
/// </summary>
/// <param name="index">Index of the transform feedback buffer</param>
/// <param name="gpuVa">Start GPU virtual address of the buffer</param>
/// <param name="size">Size in bytes of the transform feedback buffer</param>
public void SetTransformFeedbackBuffer(int index, ulong gpuVa, ulong size)
{
ulong address = TranslateAndCreateBuffer(gpuVa, size);
_transformFeedbackBuffers[index] = new BufferBounds(address, size);
_transformFeedbackBuffersDirty = true;
}
/// <summary>
/// Sets a storage buffer on the compute pipeline.
/// Storage buffers can be read and written to on shaders.
/// </summary>
/// <param name="index">Index of the storage buffer</param>
/// <param name="gpuVa">Start GPU virtual address of the buffer</param>
/// <param name="size">Size in bytes of the storage buffer</param>
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
/// <param name="flags">Buffer usage flags</param>
public void SetComputeStorageBuffer(int index, ulong gpuVa, ulong size, BufferUsageFlags flags)
2019-10-13 06:02:07 +00:00
{
size += gpuVa & ((ulong)_context.Capabilities.StorageBufferOffsetAlignment - 1);
2019-10-13 06:02:07 +00:00
gpuVa = BitUtils.AlignDown(gpuVa, _context.Capabilities.StorageBufferOffsetAlignment);
2019-10-13 06:02:07 +00:00
ulong address = TranslateAndCreateBuffer(gpuVa, size);
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
_cpStorageBuffers.SetBounds(index, address, size, flags);
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Sets a storage buffer on the graphics pipeline.
/// Storage buffers can be read and written to on shaders.
/// </summary>
/// <param name="stage">Index of the shader stage</param>
/// <param name="index">Index of the storage buffer</param>
/// <param name="gpuVa">Start GPU virtual address of the buffer</param>
/// <param name="size">Size in bytes of the storage buffer</param>
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
/// <param name="flags">Buffer usage flags</param>
public void SetGraphicsStorageBuffer(int stage, int index, ulong gpuVa, ulong size, BufferUsageFlags flags)
2019-10-13 06:02:07 +00:00
{
size += gpuVa & ((ulong)_context.Capabilities.StorageBufferOffsetAlignment - 1);
2019-10-13 06:02:07 +00:00
gpuVa = BitUtils.AlignDown(gpuVa, _context.Capabilities.StorageBufferOffsetAlignment);
2019-10-13 06:02:07 +00:00
ulong address = TranslateAndCreateBuffer(gpuVa, size);
if (_gpStorageBuffers[stage].Buffers[index].Address != address ||
_gpStorageBuffers[stage].Buffers[index].Size != size)
{
_gpStorageBuffersDirty = true;
}
2019-10-13 06:02:07 +00:00
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
_gpStorageBuffers[stage].SetBounds(index, address, size, flags);
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Sets a uniform buffer on the compute pipeline.
/// Uniform buffers are read-only from shaders, and have a small capacity.
/// </summary>
/// <param name="index">Index of the uniform buffer</param>
/// <param name="gpuVa">Start GPU virtual address of the buffer</param>
/// <param name="size">Size in bytes of the storage buffer</param>
2019-10-13 06:02:07 +00:00
public void SetComputeUniformBuffer(int index, ulong gpuVa, ulong size)
{
ulong address = TranslateAndCreateBuffer(gpuVa, size);
_cpUniformBuffers.SetBounds(index, address, size);
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Sets a uniform buffer on the graphics pipeline.
/// Uniform buffers are read-only from shaders, and have a small capacity.
/// </summary>
/// <param name="stage">Index of the shader stage</param>
/// <param name="index">Index of the uniform buffer</param>
/// <param name="gpuVa">Start GPU virtual address of the buffer</param>
/// <param name="size">Size in bytes of the storage buffer</param>
2019-10-13 06:02:07 +00:00
public void SetGraphicsUniformBuffer(int stage, int index, ulong gpuVa, ulong size)
{
ulong address = TranslateAndCreateBuffer(gpuVa, size);
_gpUniformBuffers[stage].SetBounds(index, address, size);
2019-10-13 06:02:07 +00:00
_gpUniformBuffersDirty = true;
}
/// <summary>
/// Sets the binding points for the storage buffers bound on the compute pipeline.
/// </summary>
/// <param name="descriptors">Buffer descriptors with the binding point values</param>
public void SetComputeStorageBufferBindings(ReadOnlyCollection<BufferDescriptor> descriptors)
2019-10-13 06:02:07 +00:00
{
_cpStorageBuffers.SetBindings(descriptors);
_cpStorageBufferBindings = descriptors.Count != 0 ? descriptors.Max(x => x.Binding) + 1 : 0;
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Sets the binding points for the storage buffers bound on the graphics pipeline.
/// </summary>
/// <param name="stage">Index of the shader stage</param>
/// <param name="descriptors">Buffer descriptors with the binding point values</param>
public void SetGraphicsStorageBufferBindings(int stage, ReadOnlyCollection<BufferDescriptor> descriptors)
2019-10-13 06:02:07 +00:00
{
_gpStorageBuffers[stage].SetBindings(descriptors);
2019-10-13 06:02:07 +00:00
_gpStorageBuffersDirty = true;
}
/// <summary>
/// Sets the total number of storage buffer bindings used.
/// </summary>
/// <param name="count">Number of storage buffer bindings used</param>
public void SetGraphicsStorageBufferBindingsCount(int count)
{
_gpStorageBufferBindings = count;
}
/// <summary>
/// Sets the binding points for the uniform buffers bound on the compute pipeline.
/// </summary>
/// <param name="descriptors">Buffer descriptors with the binding point values</param>
public void SetComputeUniformBufferBindings(ReadOnlyCollection<BufferDescriptor> descriptors)
{
_cpUniformBuffers.SetBindings(descriptors);
_cpUniformBufferBindings = descriptors.Count != 0 ? descriptors.Max(x => x.Binding) + 1 : 0;
}
/// <summary>
/// Sets the enabled uniform buffers mask on the graphics pipeline.
/// Each bit set on the mask indicates that the respective buffer index is enabled.
/// </summary>
/// <param name="stage">Index of the shader stage</param>
/// <param name="descriptors">Buffer descriptors with the binding point values</param>
public void SetGraphicsUniformBufferBindings(int stage, ReadOnlyCollection<BufferDescriptor> descriptors)
2019-10-13 06:02:07 +00:00
{
_gpUniformBuffers[stage].SetBindings(descriptors);
_gpUniformBuffersDirty = true;
}
/// <summary>
/// Sets the total number of uniform buffer bindings used.
/// </summary>
/// <param name="count">Number of uniform buffer bindings used</param>
public void SetGraphicsUniformBufferBindingsCount(int count)
{
_gpUniformBufferBindings = count;
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Gets a bit mask indicating which compute uniform buffers are currently bound.
/// </summary>
/// <returns>Mask where each bit set indicates a bound constant buffer</returns>
public uint GetComputeUniformBufferUseMask()
{
uint mask = 0;
for (int i = 0; i < _cpUniformBuffers.Buffers.Length; i++)
{
if (_cpUniformBuffers.Buffers[i].Address != 0)
{
mask |= 1u << i;
}
}
return mask;
}
/// <summary>
/// Gets a bit mask indicating which graphics uniform buffers are currently bound.
/// </summary>
/// <param name="stage">Index of the shader stage</param>
/// <returns>Mask where each bit set indicates a bound constant buffer</returns>
public uint GetGraphicsUniformBufferUseMask(int stage)
{
uint mask = 0;
for (int i = 0; i < _gpUniformBuffers[stage].Buffers.Length; i++)
{
if (_gpUniformBuffers[stage].Buffers[i].Address != 0)
{
mask |= 1u << i;
}
}
return mask;
}
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
/// <summary>
/// Handles removal of buffers written to a memory region being unmapped.
/// </summary>
/// <param name="sender">Sender object</param>
/// <param name="e">Event arguments</param>
public void MemoryUnmappedHandler(object sender, UnmapEventArgs e)
{
Buffer[] overlaps = new Buffer[10];
int overlapCount;
ulong address = _context.MemoryManager.Translate(e.Address);
ulong size = e.Size;
lock (_buffers)
{
overlapCount = _buffers.FindOverlaps(address, size, ref overlaps);
}
for (int i = 0; i < overlapCount; i++)
{
overlaps[i].Unmapped(address, size);
}
}
/// <summary>
/// Performs address translation of the GPU virtual address, and creates a
/// new buffer, if needed, for the specified range.
/// </summary>
/// <param name="gpuVa">Start GPU virtual address of the buffer</param>
/// <param name="size">Size in bytes of the buffer</param>
/// <returns>CPU virtual address of the buffer, after address translation</returns>
2019-10-13 06:02:07 +00:00
private ulong TranslateAndCreateBuffer(ulong gpuVa, ulong size)
{
if (gpuVa == 0)
{
return 0;
}
ulong address = _context.MemoryManager.Translate(gpuVa);
GPU - Improve Memory Allocation (#1722) * Implement TreeMap from scratch. Begin implementation of MemoryBlockManager * Implement GetFreePosition using MemoryBlocks * Implementation of Memory Management using a Tree. Still some issues to work around, but promising thus far. * Resolved invalid mapping issue. Performance appears promising. * Add tick metrics * Use the logger instead * Use debug loggin instead of info. * Remove unnecessary code. Add descriptions of added functions. * Improve memory allocation even further. As well as improve speed of position fetching. * Add TreeDictionary to Ryujinx Commons Removed Unnecessary Usigns * Add a Performance Profiler + Improve ReserveFixed * Begin transition to allocation in nvdrv * Create singleton nvmemallocator * Moved Allocation into Nv Related Files As requested by gdkchan, any allocation of memory has been moved into the driver files. Mapping remains in the GPU MemoryManager. * Remove unnecessary usings * Add missing descriptions * Correct descriptions * Fix formatting. * Remove unnecessary whitespace * Formatting / Convention Updates * Changes / Fixes Made syntax and convention changes as requested by gdkchan. Fixed an issue where IsRegionUsed would return the wrong boolean. Fixed an issue where GetFreePosition was asked for an address instead of a size. * Undo commenting of Assert in shader cache * Update Ryujinx.Common/Collections/TreeDictionary.cs Co-authored-by: gdkchan <gab.dark.100@gmail.com> * Resolved many suggestions * Implement Improved TreeDictionary Based off of Pseudo code and custom implementations. * Rename _set to _dictionary * Remove unused code * Remove unused code. * Remove unnecessary MapLow function. * Resolve data-structure based issues * Make adjustments to memory management. Deactive de-allocation for now, it causes more harm than good. * Minor refactorings + Re-implement deallocation Also cleaned up unnecessary code. * Add Tests for TreeDictionary * Update data structure to properly balance the tree * Experimental Implementation: 1. Reduce Time to Next Node to O(1) Runtime 2. Reduce While Loop Ct To 2 (In Most Cases) * Address issues w/ Deallocating Memory * Final Build + Fully Implement Dictionary Interface for new Data Structure + Cover All Memory Allocation Edge Cases, particularly w/ Games that De-Allocate a lot. * Minor Corrections Give TreeDictionary its own count (do not depend on inner dictionary) Properly remove adjacent allocations * Add AsList * Fix bug where internal dictionary wasn't being updated w/ new node for overwritten key. * Address comments in review. * Fix issue where block wouldn't break out (Fixes UE4 issues) * Update descriptions * Update descriptions * Reduce Node visibility to protect TreeDictionary Integrity + Remove usage of struct. * Update tests to use new TreeDictionary implementation. * Remove usage of dictionary in TreeDictionary * Refactoring / Renaming * Remove unneeded memoryblock class. * Add space for while * Add space for if * Formatting / descriptions * Clarified some descriptions * Reduce visibility of memory allocator * Edit method names to make more sense as memory blocks are no longer in use. * Make names consistent. * Protect against npe when sucessorof is called against keys that don't exist. (Not in use by memory manager, this is for other prs that might use this data structure) * Possible edge-case resolve * Update Ryujinx.Common/Collections/TreeDictionary.cs Co-authored-by: gdkchan <gab.dark.100@gmail.com> * Update Ryujinx.HLE/HOS/Services/Nv/NvMemoryAllocator.cs Co-authored-by: gdkchan <gab.dark.100@gmail.com> * Reduce # of unnecessary duplicate variables / Reduce visibility of variables only internally used. * Rename count to _count * Update Description of Add method. * Fix copypasta * Address comments * Address comments * Remove whitespace * Address comments, condense variables. * Consolidate vars * Fix whitespace. * Nit * Fix exception msg * Fix arrayIndex check * Fix arrayIndex check + indexer * Remove whitespace from cast Co-authored-by: gdkchan <gab.dark.100@gmail.com>
2020-12-09 22:26:05 +00:00
if (address == MemoryManager.PteUnmapped)
2019-10-13 06:02:07 +00:00
{
return 0;
}
CreateBuffer(address, size);
return address;
}
/// <summary>
/// Creates a new buffer for the specified range, if it does not yet exist.
/// This can be used to ensure the existance of a buffer.
/// </summary>
/// <param name="address">Address of the buffer in memory</param>
/// <param name="size">Size of the buffer in bytes</param>
public void CreateBuffer(ulong address, ulong size)
{
2019-10-13 06:02:07 +00:00
ulong endAddress = address + size;
ulong alignedAddress = address & ~BufferAlignmentMask;
ulong alignedEndAddress = (endAddress + BufferAlignmentMask) & ~BufferAlignmentMask;
// The buffer must have the size of at least one page.
if (alignedEndAddress == alignedAddress)
{
alignedEndAddress += BufferAlignmentSize;
}
CreateBufferAligned(alignedAddress, alignedEndAddress - alignedAddress);
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Creates a new buffer for the specified range, if needed.
/// If a buffer where this range can be fully contained already exists,
/// then the creation of a new buffer is not necessary.
/// </summary>
/// <param name="address">Address of the buffer in guest memory</param>
/// <param name="size">Size in bytes of the buffer</param>
private void CreateBufferAligned(ulong address, ulong size)
2019-10-13 06:02:07 +00:00
{
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
int overlapsCount;
lock (_buffers)
{
overlapsCount = _buffers.FindOverlapsNonOverlapping(address, size, ref _bufferOverlaps);
}
2019-10-13 06:02:07 +00:00
if (overlapsCount != 0)
2019-10-13 06:02:07 +00:00
{
// The buffer already exists. We can just return the existing buffer
// if the buffer we need is fully contained inside the overlapping buffer.
// Otherwise, we must delete the overlapping buffers and create a bigger buffer
// that fits all the data we need. We also need to copy the contents from the
// old buffer(s) to the new buffer.
ulong endAddress = address + size;
if (_bufferOverlaps[0].Address > address || _bufferOverlaps[0].EndAddress < endAddress)
2019-10-13 06:02:07 +00:00
{
for (int index = 0; index < overlapsCount; index++)
2019-10-13 06:02:07 +00:00
{
Buffer buffer = _bufferOverlaps[index];
2019-10-13 06:02:07 +00:00
address = Math.Min(address, buffer.Address);
endAddress = Math.Max(endAddress, buffer.EndAddress);
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
lock (_buffers)
{
_buffers.Remove(buffer);
}
2019-10-13 06:02:07 +00:00
}
Buffer newBuffer = new Buffer(_context, address, endAddress - address);
Memory Read/Write Tracking using Region Handles (#1272) * WIP Range Tracking - Texture invalidation seems to have large problems - Buffer/Pool invalidation may have problems - Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution. - Native project is in the messiest possible location. - [HACK] JIT memory access always uses native "fast" path - [HACK] Trying some things with texture invalidation and views. It works :) Still a few hacks, messy things, slow things More work in progress stuff (also move to memory project) Quite a bit faster now. - Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former. - The Virtual range list is now non-overlapping like the physical one. - Fixed some bugs where regions could leak. - Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road) Move some stuff. I think we'll eventually just put the dll and so for this in a nuget package. Fix rebase. [WIP] MultiRegionHandle variable size ranges - Avoid reprotecting regions that change often (needs some tweaking) - There's still a bug in buffers, somehow. - Might want different api for minimum granularity Fix rebase issue Commit everything needed for software only tracking. Remove native components. Remove more native stuff. Cleanup Use a separate window for the background context, update opentk. (fixes linux) Some experimental changes Should get things working up to scratch - still need to try some things with flush/modification and res scale. Include address with the region action. Initial work to make range tracking work Still a ton of bugs Fix some issues with the new stuff. * Fix texture flush instability There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it) * Find the destination texture for Buffer->Texture full copy Greatly improves performance for nvdec videos (with range tracking) * Further improve texture tracking * Disable Memory Tracking for view parents This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice) The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future. * Introduce some tracking tests. WIP * Complete base tests. * Add more tests for multiregion, fix existing test. * Cleanup Part 1 * Remove unnecessary code from memory tracking * Fix some inconsistencies with 3D texture rule. * Add dispose tests. * Use a background thread for the background context. Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster. Also nerf the multithreading test a bit. * Copy to texture with matching alignment This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size. * Track reads for buffer copies. Synchronize new buffers before copying overlaps. * Remove old texture flushing mechanisms. Range tracking all the way, baby. * Wake the background thread when disposing. Avoids a deadlock when games are closed. * Address Feedback 1 * Separate TextureCopy instance for background thread Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread. * Add missing XML docs. * Address Feedback * Maybe I should start drinking coffee. * Some more feedback. * Remove flush warning, Refocus window after making background context
2020-10-16 20:18:35 +00:00
newBuffer.SynchronizeMemory(address, endAddress - address);
2019-10-13 06:02:07 +00:00
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
lock (_buffers)
{
_buffers.Add(newBuffer);
}
2019-10-13 06:02:07 +00:00
for (int index = 0; index < overlapsCount; index++)
2019-10-13 06:02:07 +00:00
{
Buffer buffer = _bufferOverlaps[index];
2019-10-13 06:02:07 +00:00
int dstOffset = (int)(buffer.Address - newBuffer.Address);
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
buffer.SynchronizeMemory(buffer.Address, buffer.Size);
2019-10-13 06:02:07 +00:00
buffer.CopyTo(newBuffer, dstOffset);
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
newBuffer.InheritModifiedRanges(buffer);
2019-10-13 06:02:07 +00:00
buffer.Dispose();
}
// Existing buffers were modified, we need to rebind everything.
2019-10-13 06:02:07 +00:00
_rebind = true;
}
}
else
{
// No overlap, just create a new buffer.
Buffer buffer = new Buffer(_context, address, size);
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
lock (_buffers)
{
_buffers.Add(buffer);
}
2019-10-13 06:02:07 +00:00
}
ShrinkOverlapsBufferIfNeeded();
}
/// <summary>
/// Resizes the temporary buffer used for range list intersection results, if it has grown too much.
/// </summary>
private void ShrinkOverlapsBufferIfNeeded()
{
if (_bufferOverlaps.Length > OverlapsBufferMaxCapacity)
{
Array.Resize(ref _bufferOverlaps, OverlapsBufferMaxCapacity);
}
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Gets the address of the compute uniform buffer currently bound at the given index.
/// </summary>
/// <param name="index">Index of the uniform buffer binding</param>
/// <returns>The uniform buffer address, or an undefined value if the buffer is not currently bound</returns>
2019-10-13 06:02:07 +00:00
public ulong GetComputeUniformBufferAddress(int index)
{
return _cpUniformBuffers.Buffers[index].Address;
}
/// <summary>
/// Gets the address of the graphics uniform buffer currently bound at the given index.
/// </summary>
/// <param name="stage">Index of the shader stage</param>
/// <param name="index">Index of the uniform buffer binding</param>
/// <returns>The uniform buffer address, or an undefined value if the buffer is not currently bound</returns>
2019-10-13 06:02:07 +00:00
public ulong GetGraphicsUniformBufferAddress(int stage, int index)
{
return _gpUniformBuffers[stage].Buffers[index].Address;
}
/// <summary>
/// Ensures that the compute engine bindings are visible to the host GPU.
2020-01-01 15:39:09 +00:00
/// Note: this actually performs the binding using the host graphics API.
/// </summary>
2019-10-13 06:02:07 +00:00
public void CommitComputeBindings()
{
int sCount = _cpStorageBufferBindings;
2019-10-13 06:02:07 +00:00
Span<BufferRange> sRanges = sCount < StackToHeapThreshold ? stackalloc BufferRange[sCount] : new BufferRange[sCount];
for (int index = 0; index < _cpStorageBuffers.Count; index++)
2019-10-13 06:02:07 +00:00
{
ref var bindingInfo = ref _cpStorageBuffers.Bindings[index];
2019-10-13 06:02:07 +00:00
BufferBounds bounds = _cpStorageBuffers.Buffers[bindingInfo.Slot];
2019-10-13 06:02:07 +00:00
if (bounds.Address != 0)
2019-10-13 06:02:07 +00:00
{
// The storage buffer size is not reliable (it might be lower than the actual size),
// so we bind the entire buffer to allow otherwise out of range accesses to work.
sRanges[bindingInfo.Binding] = GetBufferRangeTillEnd(
bounds.Address,
bounds.Size,
bounds.Flags.HasFlag(BufferUsageFlags.Write));
2019-10-13 06:02:07 +00:00
}
}
2019-10-13 06:02:07 +00:00
_context.Renderer.Pipeline.SetStorageBuffers(sRanges);
2019-10-13 06:02:07 +00:00
int uCount = _cpUniformBufferBindings;
2019-10-13 06:02:07 +00:00
Span<BufferRange> uRanges = uCount < StackToHeapThreshold ? stackalloc BufferRange[uCount] : new BufferRange[uCount];
2019-10-13 06:02:07 +00:00
for (int index = 0; index < _cpUniformBuffers.Count; index++)
2019-10-13 06:02:07 +00:00
{
ref var bindingInfo = ref _cpUniformBuffers.Bindings[index];
2019-10-13 06:02:07 +00:00
BufferBounds bounds = _cpUniformBuffers.Buffers[bindingInfo.Slot];
2019-10-13 06:02:07 +00:00
if (bounds.Address != 0)
2019-10-13 06:02:07 +00:00
{
uRanges[bindingInfo.Binding] = GetBufferRange(bounds.Address, bounds.Size);
2019-10-13 06:02:07 +00:00
}
}
_context.Renderer.Pipeline.SetUniformBuffers(uRanges);
// Force rebind after doing compute work.
_rebind = true;
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Ensures that the graphics engine bindings are visible to the host GPU.
2020-01-01 15:39:09 +00:00
/// Note: this actually performs the binding using the host graphics API.
/// </summary>
public void CommitGraphicsBindings()
2019-10-13 06:02:07 +00:00
{
if (_indexBufferDirty || _rebind)
{
_indexBufferDirty = false;
if (_indexBuffer.Address != 0)
{
BufferRange buffer = GetBufferRange(_indexBuffer.Address, _indexBuffer.Size);
2019-12-29 17:41:50 +00:00
_context.Renderer.Pipeline.SetIndexBuffer(buffer, _indexBuffer.Type);
2019-10-13 06:02:07 +00:00
}
}
else if (_indexBuffer.Address != 0)
{
SynchronizeBufferRange(_indexBuffer.Address, _indexBuffer.Size);
}
uint vbEnableMask = _vertexBuffersEnableMask;
2019-10-13 06:02:07 +00:00
if (_vertexBuffersDirty || _rebind)
{
_vertexBuffersDirty = false;
Span<VertexBufferDescriptor> vertexBuffers = stackalloc VertexBufferDescriptor[Constants.TotalVertexBuffers];
2019-10-13 06:02:07 +00:00
for (int index = 0; (vbEnableMask >> index) != 0; index++)
2019-10-13 06:02:07 +00:00
{
VertexBuffer vb = _vertexBuffers[index];
if (vb.Address == 0)
{
continue;
}
BufferRange buffer = GetBufferRange(vb.Address, vb.Size);
vertexBuffers[index] = new VertexBufferDescriptor(buffer, vb.Stride, vb.Divisor);
}
2019-12-29 17:41:50 +00:00
_context.Renderer.Pipeline.SetVertexBuffers(vertexBuffers);
2019-10-13 06:02:07 +00:00
}
else
{
for (int index = 0; (vbEnableMask >> index) != 0; index++)
2019-10-13 06:02:07 +00:00
{
VertexBuffer vb = _vertexBuffers[index];
if (vb.Address == 0)
{
continue;
}
SynchronizeBufferRange(vb.Address, vb.Size);
}
}
if (_transformFeedbackBuffersDirty || _rebind)
{
_transformFeedbackBuffersDirty = false;
Span<BufferRange> tfbs = stackalloc BufferRange[Constants.TotalTransformFeedbackBuffers];
for (int index = 0; index < Constants.TotalTransformFeedbackBuffers; index++)
{
BufferBounds tfb = _transformFeedbackBuffers[index];
if (tfb.Address == 0)
{
tfbs[index] = BufferRange.Empty;
continue;
}
tfbs[index] = GetBufferRange(tfb.Address, tfb.Size);
}
_context.Renderer.Pipeline.SetTransformFeedbackBuffers(tfbs);
}
else
{
for (int index = 0; index < Constants.TotalTransformFeedbackBuffers; index++)
{
BufferBounds tfb = _transformFeedbackBuffers[index];
if (tfb.Address == 0)
{
continue;
}
SynchronizeBufferRange(tfb.Address, tfb.Size);
}
}
2019-10-13 06:02:07 +00:00
if (_gpStorageBuffersDirty || _rebind)
{
_gpStorageBuffersDirty = false;
BindBuffers(_gpStorageBuffers, isStorage: true);
}
else
{
UpdateBuffers(_gpStorageBuffers);
}
if (_gpUniformBuffersDirty || _rebind)
{
_gpUniformBuffersDirty = false;
BindBuffers(_gpUniformBuffers, isStorage: false);
}
else
{
UpdateBuffers(_gpUniformBuffers);
}
_rebind = false;
}
/// <summary>
/// Bind respective buffer bindings on the host API.
/// </summary>
/// <param name="bindings">Bindings to bind</param>
/// <param name="isStorage">True to bind as storage buffer, false to bind as uniform buffers</param>
2019-10-13 06:02:07 +00:00
private void BindBuffers(BuffersPerStage[] bindings, bool isStorage)
{
int count = isStorage ? _gpStorageBufferBindings : _gpUniformBufferBindings;
Span<BufferRange> ranges = count < StackToHeapThreshold ? stackalloc BufferRange[count] : new BufferRange[count];
for (ShaderStage stage = ShaderStage.Vertex; stage <= ShaderStage.Fragment; stage++)
{
ref var buffers = ref bindings[(int)stage - 1];
for (int index = 0; index < buffers.Count; index++)
{
ref var bindingInfo = ref buffers.Bindings[index];
BufferBounds bounds = buffers.Buffers[bindingInfo.Slot];
if (bounds.Address != 0)
{
ranges[bindingInfo.Binding] = isStorage
? GetBufferRangeTillEnd(bounds.Address, bounds.Size, bounds.Flags.HasFlag(BufferUsageFlags.Write))
: GetBufferRange(bounds.Address, bounds.Size, bounds.Flags.HasFlag(BufferUsageFlags.Write));
}
}
}
if (isStorage)
{
_context.Renderer.Pipeline.SetStorageBuffers(ranges);
}
else
{
_context.Renderer.Pipeline.SetUniformBuffers(ranges);
}
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Updates data for the already bound buffer bindings.
/// </summary>
/// <param name="bindings">Bindings to update</param>
2019-10-13 06:02:07 +00:00
private void UpdateBuffers(BuffersPerStage[] bindings)
{
for (ShaderStage stage = ShaderStage.Vertex; stage <= ShaderStage.Fragment; stage++)
{
ref var buffers = ref bindings[(int)stage - 1];
2019-10-13 06:02:07 +00:00
for (int index = 0; index < buffers.Count; index++)
2019-10-13 06:02:07 +00:00
{
ref var binding = ref buffers.Bindings[index];
2019-10-13 06:02:07 +00:00
BufferBounds bounds = buffers.Buffers[binding.Slot];
2019-10-13 06:02:07 +00:00
if (bounds.Address == 0)
{
continue;
}
SynchronizeBufferRange(bounds.Address, bounds.Size);
2019-10-13 06:02:07 +00:00
}
}
}
/// <summary>
/// Sets the buffer storage of a buffer texture.
/// </summary>
/// <param name="texture">Buffer texture</param>
/// <param name="address">Address of the buffer in memory</param>
/// <param name="size">Size of the buffer in bytes</param>
/// <param name="compute">Indicates if the buffer texture belongs to the compute or graphics pipeline</param>
public void SetBufferTextureStorage(ITexture texture, ulong address, ulong size, bool compute)
{
CreateBuffer(address, size);
if (_rebind)
{
// We probably had to modify existing buffers to create the texture buffer,
// so rebind everything to ensure we're using the new buffers for all bound resources.
if (compute)
{
CommitComputeBindings();
}
else
{
CommitGraphicsBindings();
}
}
texture.SetStorage(GetBufferRange(address, size));
}
/// <summary>
/// Copy a buffer data from a given address to another.
/// </summary>
2020-01-01 15:39:09 +00:00
/// <remarks>
/// This does a GPU side copy.
/// </remarks>
/// <param name="srcVa">GPU virtual address of the copy source</param>
/// <param name="dstVa">GPU virtual address of the copy destination</param>
/// <param name="size">Size in bytes of the copy</param>
2019-10-13 06:02:07 +00:00
public void CopyBuffer(GpuVa srcVa, GpuVa dstVa, ulong size)
{
ulong srcAddress = TranslateAndCreateBuffer(srcVa.Pack(), size);
ulong dstAddress = TranslateAndCreateBuffer(dstVa.Pack(), size);
2019-10-27 02:41:01 +00:00
Buffer srcBuffer = GetBuffer(srcAddress, size);
Buffer dstBuffer = GetBuffer(dstAddress, size);
2019-10-13 06:02:07 +00:00
2019-10-27 02:41:01 +00:00
int srcOffset = (int)(srcAddress - srcBuffer.Address);
int dstOffset = (int)(dstAddress - dstBuffer.Address);
_context.Renderer.Pipeline.CopyBuffer(
srcBuffer.Handle,
dstBuffer.Handle,
2019-10-27 02:41:01 +00:00
srcOffset,
dstOffset,
2019-10-13 06:02:07 +00:00
(int)size);
2019-10-27 02:41:01 +00:00
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
if (srcBuffer.IsModified(srcAddress, size))
{
dstBuffer.SignalModified(dstAddress, size);
}
else
{
// Optimization: If the data being copied is already in memory, then copy it directly instead of flushing from GPU.
dstBuffer.ClearModified(dstAddress, size);
_context.PhysicalMemory.WriteUntracked(dstAddress, _context.PhysicalMemory.GetSpan(srcAddress, (int)size));
}
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Clears a buffer at a given address with the specified value.
/// </summary>
/// <remarks>
/// Both the address and size must be aligned to 4 bytes.
/// </remarks>
/// <param name="gpuVa">GPU virtual address of the region to clear</param>
/// <param name="size">Number of bytes to clear</param>
/// <param name="value">Value to be written into the buffer</param>
public void ClearBuffer(GpuVa gpuVa, ulong size, uint value)
{
ulong address = TranslateAndCreateBuffer(gpuVa.Pack(), size);
Buffer buffer = GetBuffer(address, size);
int offset = (int)(address - buffer.Address);
_context.Renderer.Pipeline.ClearBuffer(buffer.Handle, offset, (int)size, value);
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
buffer.SignalModified(address, size);
}
/// <summary>
/// Gets a buffer sub-range starting at a given memory address.
/// </summary>
/// <param name="address">Start address of the memory range</param>
/// <param name="size">Size in bytes of the memory range</param>
/// <param name="write">Whether the buffer will be written to by this use</param>
/// <returns>The buffer sub-range starting at the given memory address</returns>
private BufferRange GetBufferRangeTillEnd(ulong address, ulong size, bool write = false)
{
return GetBuffer(address, size, write).GetRange(address);
}
/// <summary>
/// Gets a buffer sub-range for a given memory range.
/// </summary>
/// <param name="address">Start address of the memory range</param>
/// <param name="size">Size in bytes of the memory range</param>
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
/// <param name="write">Whether the buffer will be written to by this use</param>
/// <returns>The buffer sub-range for the given range</returns>
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
private BufferRange GetBufferRange(ulong address, ulong size, bool write = false)
2019-10-27 02:41:01 +00:00
{
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
return GetBuffer(address, size, write).GetRange(address, size);
2019-10-27 02:41:01 +00:00
}
/// <summary>
/// Gets a buffer for a given memory range.
/// A buffer overlapping with the specified range is assumed to already exist on the cache.
/// </summary>
/// <param name="address">Start address of the memory range</param>
/// <param name="size">Size in bytes of the memory range</param>
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
/// <param name="write">Whether the buffer will be written to by this use</param>
/// <returns>The buffer where the range is fully contained</returns>
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
private Buffer GetBuffer(ulong address, ulong size, bool write = false)
2019-10-13 06:02:07 +00:00
{
Buffer buffer;
if (size != 0)
{
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
lock (_buffers)
{
buffer = _buffers.FindFirstOverlap(address, size);
}
2019-10-13 06:02:07 +00:00
buffer.SynchronizeMemory(address, size);
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
if (write)
{
buffer.SignalModified(address, size);
}
2019-10-13 06:02:07 +00:00
}
else
{
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
lock (_buffers)
{
buffer = _buffers.FindFirstOverlap(address, 1);
}
2019-10-13 06:02:07 +00:00
}
2019-10-27 02:41:01 +00:00
return buffer;
2019-10-13 06:02:07 +00:00
}
/// <summary>
/// Performs guest to host memory synchronization of a given memory range.
/// </summary>
/// <param name="address">Start address of the memory range</param>
/// <param name="size">Size in bytes of the memory range</param>
2019-10-13 06:02:07 +00:00
private void SynchronizeBufferRange(ulong address, ulong size)
{
if (size != 0)
{
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
Buffer buffer;
lock (_buffers)
{
buffer = _buffers.FindFirstOverlap(address, size);
}
2019-10-13 06:02:07 +00:00
buffer.SynchronizeMemory(address, size);
}
}
2019-12-31 22:09:49 +00:00
/// <summary>
/// Disposes all buffers in the cache.
/// It's an error to use the buffer manager after disposal.
/// </summary>
public void Dispose()
{
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
lock (_buffers)
2019-12-31 22:09:49 +00:00
{
Implement lazy flush-on-read for Buffers (SSBO/Copy) (#1790) * Initial implementation of buffer flush (VERY WIP) * Host shaders need to be rebuilt for the SSBO write flag. * New approach with reserved regions and gl sync * Fix a ton of buffer issues. * Remove unused buffer unmapped behaviour * Revert "Remove unused buffer unmapped behaviour" This reverts commit f1700e52fb8760180ac5e0987a07d409d1e70ece. * Delete modified ranges on unmap Fixes potential crashes in Super Smash Bros, where a previously modified range could lie on either side of an unmap. * Cache some more delegates. * Dispose Sync on Close * Also create host sync for GPFifo syncpoint increment. * Copy buffer optimization, add docs * Fix race condition with OpenGL Sync * Enable read tracking on CommandBuffer, insert syncpoint on WaitForIdle * Performance: Only flush individual pages of SSBO at a time This avoids flushing large amounts of data when only a small amount is actually used. * Signal Modified rather than flushing after clear * Fix some docs and code style. * Introduce a new test for tracking memory protection. Sucessfully demonstrates that the bug causing write protection to be cleared by a read action has been fixed. (these tests fail on master) * Address Comments * Add host sync for SetReference This ensures that any indirect draws will correctly flush any related buffer data written before them. Fixes some flashing and misplaced world geometry in MH rise. * Make PageAlign static * Re-enable read tracking, for reads.
2021-01-17 20:08:06 +00:00
foreach (Buffer buffer in _buffers)
{
buffer.Dispose();
}
2019-12-31 22:09:49 +00:00
}
}
2019-10-13 06:02:07 +00:00
}
}