drashna/ryujinx
1
0
Fork 0
Code Activity Actions

Use a new approach for shader BRX targets (#2532)

Browse source 
* Use a new approach for shader BRX targets

* Make shader cache actually work

* Improve the shader pattern matching a bit

* Extend LDC search to predecessor blocks, catches more cases

* Nit

* Only save the amount of constant buffer data actually used. Avoids crashes on partially mapped buffers

* Ignore Rd on predicate instructions, as they do not have a Rd register (catches more cases)
This commit is contained in:
gdkchan 2021-08-11 15:59:42 -03:00 • committed by GitHub
parent 70f79e689b
commit d9d18439f6
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
12 changed files with 472 additions and 149 deletions
Download patch file Download diff file Expand all files Collapse all files

53
Ryujinx.Graphics.Gpu/Shader/Cache/CacheHelper.cs
View file

@ -371,11 +371,13 @@ namespace Ryujinx.Graphics.Gpu.Shader.Cache
/// <summary>
/// Create guest shader cache entries from the runtime contexts.
/// </summary>
/// <param name="memoryManager">The GPU memory manager in use</param>
/// <param name="channel">The GPU channel in use</param>
/// <param name="shaderContexts">The runtime contexts</param>
/// <returns>Guest shader cahe entries from the runtime contexts</returns>
public static GuestShaderCacheEntry[] CreateShaderCacheEntries(MemoryManager memoryManager, ReadOnlySpan<TranslatorContext> shaderContexts)
public static GuestShaderCacheEntry[] CreateShaderCacheEntries(GpuChannel channel, ReadOnlySpan<TranslatorContext> shaderContexts)
{
MemoryManager memoryManager = channel.MemoryManager;
int startIndex = shaderContexts.Length > 1 ? 1 : 0;
GuestShaderCacheEntry[] entries = new GuestShaderCacheEntry[shaderContexts.Length - startIndex];
@ -389,31 +391,66 @@ namespace Ryujinx.Graphics.Gpu.Shader.Cache
continue;
}
GpuAccessor gpuAccessor = context.GpuAccessor as GpuAccessor;
ulong cb1DataAddress;
int cb1DataSize = gpuAccessor?.Cb1DataSize ?? 0;
if (context.Stage == ShaderStage.Compute)
{
cb1DataAddress = channel.BufferManager.GetComputeUniformBufferAddress(1);
}
else
{
int stageIndex = context.Stage switch
{
ShaderStage.TessellationControl => 1,
ShaderStage.TessellationEvaluation => 2,
ShaderStage.Geometry => 3,
ShaderStage.Fragment => 4,
_ => 0
};
cb1DataAddress = channel.BufferManager.GetGraphicsUniformBufferAddress(stageIndex, 1);
}
int size = context.Size;
TranslatorContext translatorContext2 = i == 1 ? shaderContexts[0] : null;
int sizeA = translatorContext2 != null ? translatorContext2.Size : 0;
byte[] code = new byte[context.Size + sizeA];
byte[] code = new byte[size + cb1DataSize + sizeA];
memoryManager.GetSpan(context.Address, context.Size).CopyTo(code);
memoryManager.GetSpan(context.Address, size).CopyTo(code);
if (cb1DataAddress != 0 && cb1DataSize != 0)
{
memoryManager.Physical.GetSpan(cb1DataAddress, cb1DataSize).CopyTo(code.AsSpan().Slice(size, cb1DataSize));
}
if (translatorContext2 != null)
{
memoryManager.GetSpan(translatorContext2.Address, sizeA).CopyTo(code.AsSpan().Slice(context.Size, sizeA));
memoryManager.GetSpan(translatorContext2.Address, sizeA).CopyTo(code.AsSpan().Slice(size + cb1DataSize, sizeA));
}
GuestGpuAccessorHeader gpuAccessorHeader = CreateGuestGpuAccessorCache(context.GpuAccessor);
if (context.GpuAccessor is GpuAccessor)
if (gpuAccessor != null)
{
gpuAccessorHeader.TextureDescriptorCount = context.TextureHandlesForCache.Count;
}
GuestShaderCacheEntryHeader header = new GuestShaderCacheEntryHeader(context.Stage, context.Size, sizeA, gpuAccessorHeader);
GuestShaderCacheEntryHeader header = new GuestShaderCacheEntryHeader(
context.Stage,
size + cb1DataSize,
sizeA,
cb1DataSize,
gpuAccessorHeader);
GuestShaderCacheEntry entry = new GuestShaderCacheEntry(header, code);
if (context.GpuAccessor is GpuAccessor gpuAccessor)
if (gpuAccessor != null)
{
foreach (int textureHandle in context.TextureHandlesForCache)
{