Jayden/sharpemu
1
0
Fork 0
mirror of https://github.com/par274/sharpemu.git synced 2026-07-21 18:36:13 +08:00
Code Issues Packages Projects Releases Wiki Activity

[AGC] Quake rendering progress: WAIT_REG_MEM, draw fixes, VideoOut, and HLE improvements (#68)

Browse Source 
* [agc] WAIT_REG_MEM suspend/resume, draw packet fixes, new HLE exports, debug cleanup

Rebased onto upstream 79a7437 (par274/sharpemu, rewritten history).

- GpuWaitRegistry: DCBs suspended on unsatisfied WAIT_REG_MEM are re-polled
  against guest memory on every submit; fixed 64-bit and standard packet parse
  offsets, apply the mask, treat PM4 compare function 0 as "always".
- TryReadSubmittedDrawCount: accept the 5-dword ItDrawIndex2 form emitted by
  DcbDrawIndex (count at +4); menu draws were silently discarded before.
- sceAgcDriverSubmitMultiDcbs: reversed ABI (rdi=address array, rsi=dword
  sizes, rdx=count).
- VideoOut: vblank events, sceVideoOutGetFlipStatus, buffers registered via
  sceVideoOutRegisterBuffers are valid flip targets.
- New HLE: libc stdio (fopen/fread/fseek/ftell/fclose/fgets), Dinkumware
  _Getpctype ctype table, NpTrophy2 stubs, AMPR PAK sequential-read tracker,
  MsgDialog lifecycle, NGS2 alt NIDs + dummy vtable for handle objects,
  guarded memset intrinsic, abort()/strcasecmp null-arg recovery.
- Removed investigation-only code (INT3 breakpoints, qfont/mcpp dumps,
  error-candidate printf traces, unconditional debug logs).

First rendered frame: Quake (PPSA01880) presents a 1920x1080 guest frame.

* Implemented a guarded native intrinsic (rep movsb) in DirectExecutionBackend to bypass HLE dispatch overhead, while preserving memory safety checks.
This commit is contained in:
Foued Attar
2026-07-11 23:22:48 +02:00
committed by GitHub
parent de4fc1e1a8
commit e1cf5b13ef
21 changed files with 2204 additions and 191 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/SharpEmu.Libs/VideoOut/VideoOutExports.cs
+169 -5
View File
@@ -38,6 +38,7 @@ public static class VideoOutExports
private const ulong SceVideoOutPixelFormatA2R10G10B10 = 0x88060000;
private const ulong SceVideoOutPixelFormatA2R10G10B10Srgb = 0x88000000;
private const ulong SceVideoOutPixelFormatA2R10G10B10Bt2020Pq = 0x88740000;
private const ulong SceVideoOutInternalEventVblank = 0x5;
private const ulong SceVideoOutInternalEventFlip = 0x6;
private const short OrbisKernelEventFilterVideoOut = -13;
@@ -53,9 +54,15 @@ public static class VideoOutExports
Environment.GetEnvironmentVariable("SHARPEMU_LOG_VIDEOOUT_FPS"),
"1",
StringComparison.Ordinal);
private static readonly bool _logVideoOutSync = string.Equals(
Environment.GetEnvironmentVariable("SHARPEMU_LOG_VIDEOOUT_SYNC"),
"1",
StringComparison.Ordinal);
private static long _frameRateWindowStart = Stopwatch.GetTimestamp();
private static long _submittedFrameCount;
private static long _presentedFrameCount;
private static long _vblankSignalCount;
private static long _flipSubmitCount;
public static void ConfigureApplicationInfo(string? title, string? titleId, string? version)
{
@@ -99,6 +106,7 @@ public static class VideoOutExports
public float Gamma { get; set; } = 1.0f;
public VideoOutBufferGroup?[] Groups { get; } = new VideoOutBufferGroup?[MaxDisplayBufferGroups];
public VideoOutBufferSlot[] BufferSlots { get; } = CreateBufferSlots();
public List<FlipEventRegistration> VblankEvents { get; } = new();
public List<FlipEventRegistration> FlipEvents { get; } = new();
}
@@ -312,11 +320,48 @@ public static class VideoOutExports
}
Thread.Sleep(1);
lock (_stateGate)
SignalVblank(port);
return (int)OrbisGen2Result.ORBIS_GEN2_OK;
}
[SysAbiExport(
Nid = "Xru92wHJRmg",
ExportName = "sceVideoOutAddVblankEvent",
Target = Generation.Gen4 | Generation.Gen5,
LibraryName = "libSceVideoOut")]
public static int VideoOutAddVblankEvent(CpuContext ctx)
{
var equeue = ctx[CpuRegister.Rdi];
var handle = unchecked((int)ctx[CpuRegister.Rsi]);
if (!TryGetPort(handle, out var port))
{
port.VblankCount++;
return OrbisVideoOutErrorInvalidHandle;
}
if (!KernelEventQueueCompatExports.IsValidEqueue(equeue))
{
return OrbisVideoOutErrorInvalidEventQueue;
}
var userData = ctx[CpuRegister.Rdx];
lock (_stateGate)
{
var existingIndex = port.VblankEvents.FindIndex(registration => registration.Equeue == equeue);
if (existingIndex >= 0)
{
port.VblankEvents[existingIndex] = new FlipEventRegistration(equeue, userData);
}
else
{
port.VblankEvents.Add(new FlipEventRegistration(equeue, userData));
}
}
// Some engines wait on this queue before issuing their first flip. Provide a first
// edge now; later calls to WaitVblank advance the same notification sequence.
SignalVblank(port);
TraceVideoOut($"videoout.add_vblank_event eq=0x{equeue:X16} handle={handle} udata=0x{userData:X16}");
return (int)OrbisGen2Result.ORBIS_GEN2_OK;
}
@@ -371,6 +416,53 @@ public static class VideoOutExports
return SubmitFlip(ctx, handle, bufferIndex, flipMode, flipArg, submitGpuImage: true);
}
// Struct layout matches the classic SceVideoOutFlipStatus (40 bytes):
// count, processTime, tsc, flipArg, currentBuffer, flipPendingNum.
[SysAbiExport(
Nid = "SbU3dwp80lQ",
ExportName = "sceVideoOutGetFlipStatus",
Target = Generation.Gen4 | Generation.Gen5,
LibraryName = "libSceVideoOut")]
public static int VideoOutGetFlipStatus(CpuContext ctx)
{
var handle = unchecked((int)ctx[CpuRegister.Rdi]);
var statusAddress = ctx[CpuRegister.Rsi];
if (statusAddress == 0)
{
return OrbisVideoOutErrorInvalidAddress;
}
VideoOutPortState? port;
lock (_stateGate)
{
_ports.TryGetValue(handle, out port);
}
if (port is null)
{
return OrbisVideoOutErrorInvalidHandle;
}
ulong count;
long flipArg;
uint currentBuffer;
lock (_stateGate)
{
count = port.FlipCount;
flipArg = 0;
currentBuffer = unchecked((uint)port.CurrentBuffer);
}
KernelMemoryCompatExports.TryWriteUInt64Compat(ctx, statusAddress + 0x00, count);
KernelMemoryCompatExports.TryWriteUInt64Compat(ctx, statusAddress + 0x08, 0);
KernelMemoryCompatExports.TryWriteUInt64Compat(ctx, statusAddress + 0x10, 0);
KernelMemoryCompatExports.TryWriteUInt64Compat(ctx, statusAddress + 0x18, unchecked((ulong)flipArg));
KernelMemoryCompatExports.TryWriteUInt64Compat(ctx, statusAddress + 0x20, currentBuffer);
TraceVideoOut($"videoout.get_flip_status handle={handle} count={count} currentBuffer={currentBuffer}");
return (int)OrbisGen2Result.ORBIS_GEN2_OK;
}
[SysAbiExport(
Nid = "zgXifHT9ErY",
ExportName = "sceVideoOutIsFlipPending",
@@ -407,7 +499,8 @@ public static class VideoOutExports
return (int)OrbisGen2Result.ORBIS_GEN2_ERROR_MEMORY_FAULT;
}
if (filter != OrbisKernelEventFilterVideoOut || ident != SceVideoOutInternalEventFlip)
if (filter != OrbisKernelEventFilterVideoOut ||
ident is not (SceVideoOutInternalEventVblank or SceVideoOutInternalEventFlip))
{
return OrbisVideoOutErrorInvalidEvent;
}
@@ -436,7 +529,8 @@ public static class VideoOutExports
return (int)OrbisGen2Result.ORBIS_GEN2_ERROR_MEMORY_FAULT;
}
if (filter != OrbisKernelEventFilterVideoOut || ident != SceVideoOutInternalEventFlip)
if (filter != OrbisKernelEventFilterVideoOut ||
ident is not (SceVideoOutInternalEventVblank or SceVideoOutInternalEventFlip))
{
return OrbisVideoOutErrorInvalidEvent;
}
@@ -756,6 +850,38 @@ public static class VideoOutExports
return groupIndex < 0 ? groupIndex : setIndex;
}
private static void SignalVblank(VideoOutPortState port)
{
List<FlipEventRegistration> vblankEvents;
ulong eventHint;
lock (_stateGate)
{
port.VblankCount++;
eventHint = SceVideoOutInternalEventVblank |
((port.VblankCount & 0x0000_FFFF_FFFF_FFFFUL) << 16);
vblankEvents = new List<FlipEventRegistration>(port.VblankEvents);
}
var signalCount = Interlocked.Increment(ref _vblankSignalCount);
foreach (var vblankEvent in vblankEvents)
{
_ = KernelEventQueueCompatExports.TriggerDisplayEvent(
vblankEvent.Equeue,
SceVideoOutInternalEventVblank,
OrbisKernelEventFilterVideoOut,
eventHint,
vblankEvent.UserData);
}
if (_logVideoOutSync && (signalCount <= 8 || signalCount % 60 == 0))
{
Console.Error.WriteLine(
$"[LOADER][SYNC] vblank#{signalCount} handle={port.Handle} count={port.VblankCount} " +
$"queues={vblankEvents.Count} hint=0x{eventHint:X16}");
}
}
private static int SubmitFlip(
CpuContext ctx,
int handle,
@@ -790,11 +916,14 @@ public static class VideoOutExports
flipEvents = new List<FlipEventRegistration>(port.FlipEvents);
}
var guestImageSubmitted = false;
ulong guestImageAddress = 0;
if (submitGpuImage &&
bufferIndex >= 0 &&
TryGetDisplayBufferInfo(handle, bufferIndex, out var displayBuffer))
{
_ = VulkanVideoPresenter.TrySubmitGuestImage(
guestImageAddress = displayBuffer.Address;
guestImageSubmitted = VulkanVideoPresenter.TrySubmitGuestImage(
displayBuffer.Address,
displayBuffer.Width,
displayBuffer.Height,
@@ -819,6 +948,15 @@ public static class VideoOutExports
flipEvent.UserData);
}
var flipCount = Interlocked.Increment(ref _flipSubmitCount);
if (_logVideoOutSync && (flipCount <= 8 || flipCount % 60 == 0))
{
Console.Error.WriteLine(
$"[LOADER][SYNC] flip#{flipCount} handle={handle} buffer={bufferIndex} " +
$"addr=0x{guestImageAddress:X16} submitted={guestImageSubmitted} " +
$"flipQueues={flipEvents.Count}");
}
TraceVideoOut($"videoout.submit_flip handle={handle} index={bufferIndex} mode={flipMode} arg={flipArg} events={flipEvents.Count}");
ReportFrameRate(presented: false);
return (int)OrbisGen2Result.ORBIS_GEN2_OK;
@@ -902,6 +1040,16 @@ public static class VideoOutExports
TraceVideoOut(
$"videoout.register_buffers handle={port.Handle} group={groupIndex} start={startIndex} count={addresses.Length} fmt=0x{attribute.PixelFormat:X} tile={attribute.TilingMode} {attribute.Width}x{attribute.Height} pitch={attribute.PitchInPixel}");
VulkanVideoPresenter.EnsureStarted(attribute.Width, attribute.Height);
var guestFormat = MapPixelFormatToGuestTextureFormat(attribute.PixelFormat);
if (guestFormat != 0)
{
foreach (var address in addresses)
{
VulkanVideoPresenter.RegisterKnownDisplayBuffer(address, guestFormat);
}
}
return groupIndex;
}
}
@@ -1123,6 +1271,22 @@ public static class VideoOutExports
? 4u
: 0u;
// Maps the PS5 VideoOut pixel format space to the AGC "guest texture format" tags
// VulkanVideoPresenter._availableGuestImages keys on (see VulkanVideoPresenter.
// GetGuestTextureFormat: format=10 => 56 for 8-bit RGBA variants, format=9 => 9 for 10-bit).
private static uint MapPixelFormatToGuestTextureFormat(ulong pixelFormat) =>
NormalizePixelFormat(pixelFormat) switch
{
SceVideoOutPixelFormatA8R8G8B8Srgb or
SceVideoOutPixelFormatA8B8G8R8Srgb or
SceVideoOutPixelFormatB8G8R8A8Unorm or
SceVideoOutPixelFormatR8G8B8A8Unorm => 56u,
SceVideoOutPixelFormatA2R10G10B10 or
SceVideoOutPixelFormatA2R10G10B10Srgb or
SceVideoOutPixelFormatA2R10G10B10Bt2020Pq => 9u,
_ => 0u,
};
private static ulong NormalizePixelFormat(ulong pixelFormat)
{
if (GetBytesPerPixel(pixelFormat) != 0)