JinxRyu/Ryujinx.HLE/HOS/Services/Nv/NvDrvServices/NvHostChannel/NvHostChannelDeviceFile.cs

570 lines
21 KiB
C#
Raw Normal View History

using Ryujinx.Common.Logging;
using Ryujinx.Graphics.Gpu;
2019-10-13 06:02:07 +00:00
using Ryujinx.Graphics.Gpu.Memory;
using Ryujinx.HLE.HOS.Services.Nv.NvDrvServices.NvHostChannel.Types;
using Ryujinx.HLE.HOS.Services.Nv.NvDrvServices.NvHostCtrl;
using Ryujinx.HLE.HOS.Services.Nv.NvDrvServices.NvMap;
using Ryujinx.HLE.HOS.Services.Nv.Types;
using Ryujinx.Memory;
using System;
using System.Collections.Concurrent;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
namespace Ryujinx.HLE.HOS.Services.Nv.NvDrvServices.NvHostChannel
{
class NvHostChannelDeviceFile : NvDeviceFile
{
private static readonly ConcurrentDictionary<long, Host1xContext> _host1xContextRegistry = new();
private const uint MaxModuleSyncpoint = 16;
private uint _timeout;
private uint _submitTimeout;
private uint _timeslice;
private readonly Switch _device;
private readonly IVirtualMemoryManager _memory;
private readonly Host1xContext _host1xContext;
public GpuChannel Channel { get; }
public enum ResourcePolicy
{
Device,
Channel
}
protected static uint[] DeviceSyncpoints = new uint[MaxModuleSyncpoint];
protected uint[] ChannelSyncpoints;
protected static ResourcePolicy ChannelResourcePolicy = ResourcePolicy.Device;
private NvFence _channelSyncpoint;
public NvHostChannelDeviceFile(ServiceCtx context, IVirtualMemoryManager memory, long owner) : base(context, owner)
{
_device = context.Device;
_memory = memory;
_timeout = 3000;
_submitTimeout = 0;
_timeslice = 0;
_host1xContext = GetHost1XContext(context.Device.Gpu, owner);
Channel = _device.Gpu.CreateChannel();
ChannelInitialization.InitializeState(Channel);
ChannelSyncpoints = new uint[MaxModuleSyncpoint];
_channelSyncpoint.Id = _device.System.HostSyncpoint.AllocateSyncpoint(false);
_channelSyncpoint.UpdateValue(_device.System.HostSyncpoint);
}
public override NvInternalResult Ioctl(NvIoctl command, Span<byte> arguments)
{
NvInternalResult result = NvInternalResult.NotImplemented;
if (command.Type == NvIoctl.NvHostCustomMagic)
{
switch (command.Number)
{
case 0x01:
result = Submit(arguments);
break;
case 0x02:
result = CallIoctlMethod<GetParameterArguments>(GetSyncpoint, arguments);
break;
case 0x03:
result = CallIoctlMethod<GetParameterArguments>(GetWaitBase, arguments);
break;
case 0x07:
result = CallIoctlMethod<uint>(SetSubmitTimeout, arguments);
break;
case 0x09:
result = MapCommandBuffer(arguments);
break;
case 0x0a:
result = UnmapCommandBuffer(arguments);
break;
}
}
else if (command.Type == NvIoctl.NvHostMagic)
{
switch (command.Number)
{
case 0x01:
result = CallIoctlMethod<int>(SetNvMapFd, arguments);
break;
case 0x03:
result = CallIoctlMethod<uint>(SetTimeout, arguments);
break;
case 0x08:
result = SubmitGpfifo(arguments);
break;
case 0x09:
result = CallIoctlMethod<AllocObjCtxArguments>(AllocObjCtx, arguments);
break;
case 0x0b:
result = CallIoctlMethod<ZcullBindArguments>(ZcullBind, arguments);
break;
case 0x0c:
result = CallIoctlMethod<SetErrorNotifierArguments>(SetErrorNotifier, arguments);
break;
case 0x0d:
result = CallIoctlMethod<NvChannelPriority>(SetPriority, arguments);
break;
case 0x18:
result = CallIoctlMethod<AllocGpfifoExArguments>(AllocGpfifoEx, arguments);
break;
case 0x1a:
result = CallIoctlMethod<AllocGpfifoExArguments>(AllocGpfifoEx2, arguments);
break;
case 0x1d:
result = CallIoctlMethod<uint>(SetTimeslice, arguments);
break;
}
}
else if (command.Type == NvIoctl.NvGpuMagic)
{
switch (command.Number)
{
case 0x14:
result = CallIoctlMethod<ulong>(SetUserData, arguments);
break;
}
}
return result;
}
private NvInternalResult Submit(Span<byte> arguments)
{
SubmitArguments submitHeader = GetSpanAndSkip<SubmitArguments>(ref arguments, 1)[0];
Span<CommandBuffer> commandBuffers = GetSpanAndSkip<CommandBuffer>(ref arguments, submitHeader.CmdBufsCount);
Span<Reloc> relocs = GetSpanAndSkip<Reloc>(ref arguments, submitHeader.RelocsCount);
Span<uint> relocShifts = GetSpanAndSkip<uint>(ref arguments, submitHeader.RelocsCount);
Span<SyncptIncr> syncptIncrs = GetSpanAndSkip<SyncptIncr>(ref arguments, submitHeader.SyncptIncrsCount);
Span<SyncptIncr> waitChecks = GetSpanAndSkip<SyncptIncr>(ref arguments, submitHeader.SyncptIncrsCount); // ?
Span<Fence> fences = GetSpanAndSkip<Fence>(ref arguments, submitHeader.FencesCount);
lock (_device)
{
for (int i = 0; i < syncptIncrs.Length; i++)
{
SyncptIncr syncptIncr = syncptIncrs[i];
uint id = syncptIncr.Id;
fences[i].Id = id;
fences[i].Thresh = Context.Device.System.HostSyncpoint.IncrementSyncpointMax(id, syncptIncr.Incrs);
}
foreach (CommandBuffer commandBuffer in commandBuffers)
{
NvMapHandle map = NvMapDeviceFile.GetMapFromHandle(Owner, commandBuffer.Mem);
var data = _memory.GetSpan(map.Address + commandBuffer.Offset, commandBuffer.WordsCount * 4);
_host1xContext.Host1x.Submit(MemoryMarshal.Cast<byte, int>(data));
}
}
fences[0].Thresh = Context.Device.System.HostSyncpoint.IncrementSyncpointMax(fences[0].Id, 1);
Span<int> tmpCmdBuff = stackalloc int[1];
tmpCmdBuff[0] = (4 << 28) | (int)fences[0].Id;
_host1xContext.Host1x.Submit(tmpCmdBuff);
return NvInternalResult.Success;
}
private Span<T> GetSpanAndSkip<T>(ref Span<byte> arguments, int count) where T : unmanaged
{
Span<T> output = MemoryMarshal.Cast<byte, T>(arguments).Slice(0, count);
arguments = arguments.Slice(Unsafe.SizeOf<T>() * count);
return output;
}
private NvInternalResult GetSyncpoint(ref GetParameterArguments arguments)
{
if (arguments.Parameter >= MaxModuleSyncpoint)
{
return NvInternalResult.InvalidInput;
}
if (ChannelResourcePolicy == ResourcePolicy.Device)
{
arguments.Value = GetSyncpointDevice(_device.System.HostSyncpoint, arguments.Parameter, false);
}
else
{
arguments.Value = GetSyncpointChannel(arguments.Parameter, false);
}
if (arguments.Value == 0)
{
return NvInternalResult.TryAgain;
}
return NvInternalResult.Success;
}
private NvInternalResult GetWaitBase(ref GetParameterArguments arguments)
{
arguments.Value = 0;
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult SetSubmitTimeout(ref uint submitTimeout)
{
_submitTimeout = submitTimeout;
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult MapCommandBuffer(Span<byte> arguments)
{
int headerSize = Unsafe.SizeOf<MapCommandBufferArguments>();
MapCommandBufferArguments commandBufferHeader = MemoryMarshal.Cast<byte, MapCommandBufferArguments>(arguments)[0];
Span<CommandBufferHandle> commandBufferEntries = MemoryMarshal.Cast<byte, CommandBufferHandle>(arguments.Slice(headerSize)).Slice(0, commandBufferHeader.NumEntries);
foreach (ref CommandBufferHandle commandBufferEntry in commandBufferEntries)
{
NvMapHandle map = NvMapDeviceFile.GetMapFromHandle(Owner, commandBufferEntry.MapHandle);
if (map == null)
{
Logger.Warning?.Print(LogClass.ServiceNv, $"Invalid handle 0x{commandBufferEntry.MapHandle:x8}!");
return NvInternalResult.InvalidInput;
}
lock (map)
{
if (map.DmaMapAddress == 0)
{
ulong va = _host1xContext.MemoryAllocator.GetFreeAddress((ulong)map.Size, out ulong freeAddressStartPosition, 1, MemoryManager.PageSize);
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 (va != NvMemoryAllocator.PteUnmapped && va <= uint.MaxValue && (va + (uint)map.Size) <= uint.MaxValue)
{
_host1xContext.MemoryAllocator.AllocateRange(va, (uint)map.Size, freeAddressStartPosition);
_host1xContext.Smmu.Map(map.Address, va, (uint)map.Size);
map.DmaMapAddress = va;
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
}
else
{
map.DmaMapAddress = NvMemoryAllocator.PteUnmapped;
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
}
}
commandBufferEntry.MapAddress = (int)map.DmaMapAddress;
}
}
return NvInternalResult.Success;
}
private NvInternalResult UnmapCommandBuffer(Span<byte> arguments)
{
int headerSize = Unsafe.SizeOf<MapCommandBufferArguments>();
MapCommandBufferArguments commandBufferHeader = MemoryMarshal.Cast<byte, MapCommandBufferArguments>(arguments)[0];
Span<CommandBufferHandle> commandBufferEntries = MemoryMarshal.Cast<byte, CommandBufferHandle>(arguments.Slice(headerSize)).Slice(0, commandBufferHeader.NumEntries);
foreach (ref CommandBufferHandle commandBufferEntry in commandBufferEntries)
{
NvMapHandle map = NvMapDeviceFile.GetMapFromHandle(Owner, commandBufferEntry.MapHandle);
if (map == null)
{
Logger.Warning?.Print(LogClass.ServiceNv, $"Invalid handle 0x{commandBufferEntry.MapHandle:x8}!");
return NvInternalResult.InvalidInput;
}
lock (map)
{
if (map.DmaMapAddress != 0)
{
// FIXME:
// To make unmapping work, we need separate address space per channel.
// Right now NVDEC and VIC share the GPU address space which is not correct at all.
// _host1xContext.MemoryAllocator.Free((ulong)map.DmaMapAddress, (uint)map.Size);
// map.DmaMapAddress = 0;
}
}
}
return NvInternalResult.Success;
}
private NvInternalResult SetNvMapFd(ref int nvMapFd)
{
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult SetTimeout(ref uint timeout)
{
_timeout = timeout;
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult SubmitGpfifo(Span<byte> arguments)
{
int headerSize = Unsafe.SizeOf<SubmitGpfifoArguments>();
SubmitGpfifoArguments gpfifoSubmissionHeader = MemoryMarshal.Cast<byte, SubmitGpfifoArguments>(arguments)[0];
2019-10-13 06:02:07 +00:00
Span<ulong> gpfifoEntries = MemoryMarshal.Cast<byte, ulong>(arguments.Slice(headerSize)).Slice(0, gpfifoSubmissionHeader.NumEntries);
return SubmitGpfifo(ref gpfifoSubmissionHeader, gpfifoEntries);
}
private NvInternalResult AllocObjCtx(ref AllocObjCtxArguments arguments)
{
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult ZcullBind(ref ZcullBindArguments arguments)
{
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult SetErrorNotifier(ref SetErrorNotifierArguments arguments)
{
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult SetPriority(ref NvChannelPriority priority)
{
switch (priority)
{
case NvChannelPriority.Low:
_timeslice = 1300; // Timeslice low priority in micro-seconds
break;
case NvChannelPriority.Medium:
_timeslice = 2600; // Timeslice medium priority in micro-seconds
break;
case NvChannelPriority.High:
_timeslice = 5200; // Timeslice high priority in micro-seconds
break;
default:
return NvInternalResult.InvalidInput;
}
Logger.Stub?.PrintStub(LogClass.ServiceNv);
// TODO: disable and preempt channel when GPU scheduler will be implemented.
return NvInternalResult.Success;
}
private NvInternalResult AllocGpfifoEx(ref AllocGpfifoExArguments arguments)
{
_channelSyncpoint.UpdateValue(_device.System.HostSyncpoint);
arguments.Fence = _channelSyncpoint;
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult AllocGpfifoEx2(ref AllocGpfifoExArguments arguments)
{
_channelSyncpoint.UpdateValue(_device.System.HostSyncpoint);
arguments.Fence = _channelSyncpoint;
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
private NvInternalResult SetTimeslice(ref uint timeslice)
{
if (timeslice < 1000 || timeslice > 50000)
{
return NvInternalResult.InvalidInput;
}
_timeslice = timeslice; // in micro-seconds
Logger.Stub?.PrintStub(LogClass.ServiceNv);
// TODO: disable and preempt channel when GPU scheduler will be implemented.
return NvInternalResult.Success;
}
private NvInternalResult SetUserData(ref ulong userData)
{
Logger.Stub?.PrintStub(LogClass.ServiceNv);
return NvInternalResult.Success;
}
2019-10-13 06:02:07 +00:00
protected NvInternalResult SubmitGpfifo(ref SubmitGpfifoArguments header, Span<ulong> entries)
{
if (header.Flags.HasFlag(SubmitGpfifoFlags.FenceWait) && header.Flags.HasFlag(SubmitGpfifoFlags.IncrementWithValue))
{
return NvInternalResult.InvalidInput;
}
if (header.Flags.HasFlag(SubmitGpfifoFlags.FenceWait) && !_device.System.HostSyncpoint.IsSyncpointExpired(header.Fence.Id, header.Fence.Value))
{
Channel.PushHostCommandBuffer(CreateWaitCommandBuffer(header.Fence));
}
Channel.PushEntries(entries);
header.Fence.Id = _channelSyncpoint.Id;
if (header.Flags.HasFlag(SubmitGpfifoFlags.FenceIncrement) || header.Flags.HasFlag(SubmitGpfifoFlags.IncrementWithValue))
{
uint incrementCount = header.Flags.HasFlag(SubmitGpfifoFlags.FenceIncrement) ? 2u : 0u;
if (header.Flags.HasFlag(SubmitGpfifoFlags.IncrementWithValue))
{
incrementCount += header.Fence.Value;
}
header.Fence.Value = _device.System.HostSyncpoint.IncrementSyncpointMaxExt(header.Fence.Id, (int)incrementCount);
}
else
{
header.Fence.Value = _device.System.HostSyncpoint.ReadSyncpointMaxValue(header.Fence.Id);
}
if (header.Flags.HasFlag(SubmitGpfifoFlags.FenceIncrement))
{
Channel.PushHostCommandBuffer(CreateIncrementCommandBuffer(ref header.Fence, header.Flags));
}
header.Flags = SubmitGpfifoFlags.None;
_device.Gpu.GPFifo.SignalNewEntries();
return NvInternalResult.Success;
}
public uint GetSyncpointChannel(uint index, bool isClientManaged)
{
if (ChannelSyncpoints[index] != 0)
{
return ChannelSyncpoints[index];
}
ChannelSyncpoints[index] = _device.System.HostSyncpoint.AllocateSyncpoint(isClientManaged);
return ChannelSyncpoints[index];
}
public static uint GetSyncpointDevice(NvHostSyncpt syncpointManager, uint index, bool isClientManaged)
{
if (DeviceSyncpoints[index] != 0)
{
return DeviceSyncpoints[index];
}
DeviceSyncpoints[index] = syncpointManager.AllocateSyncpoint(isClientManaged);
return DeviceSyncpoints[index];
}
private static int[] CreateWaitCommandBuffer(NvFence fence)
{
int[] commandBuffer = new int[4];
// SyncpointValue = fence.Value;
commandBuffer[0] = 0x2001001C;
commandBuffer[1] = (int)fence.Value;
// SyncpointAction(fence.id, increment: false, switch_en: true);
commandBuffer[2] = 0x2001001D;
commandBuffer[3] = (((int)fence.Id << 8) | (0 << 0) | (1 << 4));
return commandBuffer;
}
private int[] CreateIncrementCommandBuffer(ref NvFence fence, SubmitGpfifoFlags flags)
{
bool hasWfi = !flags.HasFlag(SubmitGpfifoFlags.SuppressWfi);
int[] commandBuffer;
int offset = 0;
if (hasWfi)
{
commandBuffer = new int[8];
// WaitForInterrupt(handle)
commandBuffer[offset++] = 0x2001001E;
commandBuffer[offset++] = 0x0;
}
else
{
commandBuffer = new int[6];
}
// SyncpointValue = 0x0;
commandBuffer[offset++] = 0x2001001C;
commandBuffer[offset++] = 0x0;
// Increment the syncpoint 2 times. (mitigate a hardware bug)
// SyncpointAction(fence.id, increment: true, switch_en: false);
commandBuffer[offset++] = 0x2001001D;
commandBuffer[offset++] = (((int)fence.Id << 8) | (1 << 0) | (0 << 4));
// SyncpointAction(fence.id, increment: true, switch_en: false);
commandBuffer[offset++] = 0x2001001D;
commandBuffer[offset++] = (((int)fence.Id << 8) | (1 << 0) | (0 << 4));
return commandBuffer;
}
public override void Close()
{
Channel.Dispose();
}
private static Host1xContext GetHost1XContext(GpuContext gpu, long pid)
{
return _host1xContextRegistry.GetOrAdd(pid, (long key) => new Host1xContext(gpu, key));
}
public static void Destroy()
{
foreach (Host1xContext host1xContext in _host1xContextRegistry.Values)
{
host1xContext.Dispose();
}
_host1xContextRegistry.Clear();
}
}
}