Fix/deadcells crash (#262)

* Boot compatibility fixes for UE titles, GUI toggles, and DeS render/boot work

Checkpoint of the Monster Truck Championship and Demon's Souls boot work.
Each piece is independently useful and verified against the titles.

- playgo: scePlayGoGetLocus now returns BAD_CHUNK_ID for chunk ids outside
  the known set, matching real firmware. Titles enumerate chunk ids until
  that error; answering OK for every id made the scan wrap the ushort range
  and spin forever. Missing-sidecar and no-app0 fallbacks report a
  fully-installed single chunk 0 so scePlayGoOpen keeps succeeding.
- kernel: restore the SHARPEMU_WRITABLE_APP0 opt-in. Unpackaged UE dumps
  write their Saved tree under /app0 during PS5 component init and treat
  the denial as a fatal boot error.
- pad: accept handle 0 as the primary pad across all pad calls. Real
  firmware hands out small non-negative handles and some titles read state
  with handle 0.
- bthid: env-gated experiment hooks for the Thrustmaster wheel middleware
  investigation (fail-only-RegisterCallback modes and a synthetic
  enumeration callback with a zeroed event struct). All default off.
- gui: add SHARPEMU_LOG_IO and SHARPEMU_WRITABLE_APP0 toggles to the
  Environment tab.
- videoout: per-swapchain-image render-finished semaphores (the shared
  semaphore raced the swapchain); whole-mip-chain layout init for offscreen
  guest images (sampled binds read mips stuck in Undefined); GPU-resident
  texture availability now canonicalizes through the texture format table
  and accepts compatibility-class aliases, cutting per-frame CPU texture
  re-reads (143 GB -> 55 GB per 300 s in Demon's Souls, 0.2 -> 0.5 fps).
- hle: add sceSystemServiceGetNoticeScreenSkipFlag,
  sceSystemServiceGetMainAppTitleId (title id published from the runtime),
  and sceNpWebApi2CreateUserContext (refuses so the online layer backs off).
- rtc: SHARPEMU_RTC_PROBE_RANGE diagnostic dumps the code around a busy-wait
  caller of sceRtcGetCurrentTick once; costs nothing when unset.

* [ShaderCompiler] Fix VReadlaneB32 scalar destination field

The scalar destination lives in the low vdst byte (bits 0-7); it was read
from bits 8-14, the VOP3B carry-out field readlane does not have, sending
every readlane result to s0. Verified against raw gfx10 encodings and
LLVM's assembler tests (v_readlane_b32 s5, v1, s2 -> low byte 0x05).

* [VideoOut] Survive device loss and flip-order asserts without dying

Two ways a frame could take down the whole presenter:

- Device loss between any two Vulkan calls in a frame unwound the window
  thread, and the Dispose-time fence check then threw again, masking the
  original error. Catch the loss at the frame boundary, retire
  presentations and guest submissions whose fences can never signal, and
  keep the window loop pumping so the game (audio, logic) carries on.
- The ordered-flip capture invariant is violated ~100 times per run by
  Demon's Souls (PPSA01342); on debug builds the Debug.Assert fail-fasts
  the process with nothing in the log. Downgrade it to a once-per-version
  warning until the capture/wait ordering is understood.

* Fix Dead Cells shader cache regression

---------

Co-authored-by: StealUrKill <35749471+StealUrKill@users.noreply.github.com>
This commit is contained in:
Spooks
2026-07-16 04:57:16 -06:00
committed by GitHub
parent 7b91c964fc
commit b4b95014f1
14 changed files with 567 additions and 47 deletions
@@ -4160,6 +4160,48 @@ internal static unsafe class VulkanVideoPresenter
_submitTimeline;
}
private void TransitionNewGuestImageToSampled(Image image, uint mipLevels)
{
var commandBuffer = AllocateGuestCommandBuffer();
var beginInfo = new CommandBufferBeginInfo
{
SType = StructureType.CommandBufferBeginInfo,
Flags = CommandBufferUsageFlags.OneTimeSubmitBit,
};
Check(
_vk.BeginCommandBuffer(commandBuffer, &beginInfo),
"vkBeginCommandBuffer(guest image init)");
var barrier = new ImageMemoryBarrier
{
SType = StructureType.ImageMemoryBarrier,
SrcAccessMask = 0,
DstAccessMask = AccessFlags.ShaderReadBit,
OldLayout = ImageLayout.Undefined,
NewLayout = ImageLayout.ShaderReadOnlyOptimal,
SrcQueueFamilyIndex = Vk.QueueFamilyIgnored,
DstQueueFamilyIndex = Vk.QueueFamilyIgnored,
Image = image,
SubresourceRange = ColorSubresourceRange(0, mipLevels),
};
_vk.CmdPipelineBarrier(
commandBuffer,
PipelineStageFlags.TopOfPipeBit,
PipelineStageFlags.AllCommandsBit,
0,
0,
null,
0,
null,
1,
&barrier);
Check(
_vk.EndCommandBuffer(commandBuffer),
"vkEndCommandBuffer(guest image init)");
// Same-queue submission order makes the transition visible to any
// later use of the image; no CPU-side wait is needed.
SubmitGuestCommandBuffer(commandBuffer, [], []);
}
private void EnsureGuestSubmissionCapacity()
{
CollectCompletedGuestSubmissions(waitForOldest: false);
@@ -4220,23 +4262,43 @@ internal static unsafe class VulkanVideoPresenter
return;
}
Check(result, $"vkWaitForFences(guest: {oldest.DebugName})");
if (result == Result.ErrorDeviceLost)
{
_deviceLost = true;
}
else
{
Check(result, $"vkWaitForFences(guest: {oldest.DebugName})");
}
}
while (_pendingGuestSubmissions.TryPeek(out var submission))
{
var status = _vk.GetFenceStatus(_device, submission.Fence);
if (status == Result.NotReady)
if (status == Result.NotReady && !_deviceLost)
{
break;
}
Check(status, $"vkGetFenceStatus(guest: {submission.DebugName})");
if (status == Result.ErrorDeviceLost)
{
// Pending fences never signal on a lost device; retire the
// submission anyway so teardown and back-pressure survive.
_deviceLost = true;
}
else if (status != Result.NotReady)
{
Check(status, $"vkGetFenceStatus(guest: {submission.DebugName})");
}
_pendingGuestSubmissions.Dequeue();
foreach (var image in submission.TraceImages)
if (!_deviceLost)
{
TraceGuestImageContents(image);
foreach (var image in submission.TraceImages)
{
TraceGuestImageContents(image);
}
}
foreach (var resources in submission.Resources)
@@ -4514,13 +4576,20 @@ internal static unsafe class VulkanVideoPresenter
$"queue={_activeGuestQueue.Name} submission={_activeGuestQueue.SubmissionId} " +
$"handle={work.VideoOutHandle} index={work.DisplayBufferIndex} " +
$"capture_complete={(captured ? 1 : 0)}");
#if DEBUG
System.Diagnostics.Debug.Assert(
work.Version == 0 || captured,
"An ordered wait-safe marker must execute after its flip capture.");
#endif
// Demon's Souls executes wait-safe markers before their flip capture;
// an assert here would fail-fast the process, so warn once instead.
// Dedup on a flag, not the (per-frame-unique) version, to bound growth.
if (work.Version != 0 && !captured && !_loggedFlipWaitOrderViolation)
{
_loggedFlipWaitOrderViolation = true;
Console.Error.WriteLine(
$"[LOADER][WARN] vk.flip_wait_order version={work.Version} " +
"executed before its flip capture; continuing.");
}
}
private bool _loggedFlipWaitOrderViolation;
private GuestImageResource CreateGuestFlipSnapshot(
GuestImageResource source,
long version)
@@ -8300,7 +8369,7 @@ internal static unsafe class VulkanVideoPresenter
return (properties.OptimalTilingFeatures & FormatFeatureFlags.ColorAttachmentBit) != 0;
}
private static Format GetTextureFormat(uint format, uint numberType) =>
internal static Format GetTextureFormat(uint format, uint numberType) =>
(format, numberType) switch
{
(9, _) => Format.A2B10G10R10UnormPack32,
@@ -10170,6 +10239,9 @@ internal static unsafe class VulkanVideoPresenter
_vk.AllocateMemory(_device, &allocationInfo, null, out var memory),
"vkAllocateMemory(offscreen)");
Check(_vk.BindImageMemory(_device, image, memory, 0), "vkBindImageMemory(offscreen)");
// Rendering and uploads only define the mips they touch; define the whole
// chain once so full-chain sampled binds never read Undefined layout.
TransitionNewGuestImageToSampled(image, mipLevels);
var viewInfo = new ImageViewCreateInfo
{
@@ -10937,7 +11009,7 @@ internal static unsafe class VulkanVideoPresenter
return view;
}
private static bool IsCompatibleViewFormat(Format imageFormat, Format viewFormat)
internal static bool IsCompatibleViewFormat(Format imageFormat, Format viewFormat)
{
if (imageFormat == viewFormat)
{
@@ -11003,6 +11075,23 @@ internal static unsafe class VulkanVideoPresenter
}
private void Render(double _)
{
try
{
RenderCore();
}
catch (Exception exception)
{
// Device loss can strike between any two Vulkan calls in the frame;
// keep the window loop pumping instead of tearing the presenter down.
if (!TryMarkDeviceLost(exception))
{
throw;
}
}
}
private void RenderCore()
{
if (Volatile.Read(ref _presenterCloseRequested))
{
@@ -11016,6 +11105,18 @@ internal static unsafe class VulkanVideoPresenter
return;
}
if (_deviceLost)
{
// Drain queued work so producers aren't back-pressured, then
// return without any Vulkan call (fences never signal post-loss).
while (TryTakeGuestWork(out var lostWork))
{
CompleteGuestWork(lostWork);
}
return;
}
// Reuse of a frame slot waits only on that slot's fence, keeping
// up to MaxFramesInFlight frames pipelined between CPU and GPU.
var frameSlot = _currentFrameSlot;
@@ -13836,20 +13937,38 @@ internal static unsafe class VulkanVideoPresenter
return;
}
if (result == Result.ErrorDeviceLost)
{
throw new VulkanDeviceLostException(operation);
}
throw new InvalidOperationException($"{operation} failed with {result}.");
}
private static void Check(Result result, string operation)
{
if (result == Result.ErrorDeviceLost)
{
throw new VulkanDeviceLostException(operation);
}
if (result != Result.Success)
{
throw new InvalidOperationException($"{operation} failed with {result}.");
}
}
// Typed so the frame-boundary catch can recognize device loss without
// depending on the exact wording of the exception message.
private sealed class VulkanDeviceLostException(string operation)
: InvalidOperationException($"{operation} failed with {Result.ErrorDeviceLost}.");
private bool TryMarkDeviceLost(Exception exception)
{
if (!exception.Message.Contains(nameof(Result.ErrorDeviceLost), StringComparison.Ordinal))
// Prefer the typed signal; fall back to the message for losses that
// surface through other layers (e.g. Silk.NET bindings).
if (exception is not VulkanDeviceLostException &&
!exception.Message.Contains(nameof(Result.ErrorDeviceLost), StringComparison.Ordinal))
{
return false;
}